Anal Bioanal Chem (2002) 372 : 1

DOI 10.1007/s00216-001-1166-x

E D I TO R I A L

Franois Mathey

Analytical and Bioanalytical Chemistry

Published online: 14 December 2001

Springer-Verlag 2001

Analytical and Bioanalytical Chemistry is the sixth mem-
ber of the European family of chemical journals. Result-
ing from a collaboration between ten European chemical
societies and two major commercial publishers, these
journals provide a combination of excellent scientific qual-
ity, reasonable subscription rates and high commercial ef-
ficiency. They were founded seven years ago by several
chemical societies in continental Europe with the inten-
tion of publishing the best chemistry from Europe and the
rest of the world, while promoting and preserving the de-
sirable and necessary diversity of the scientific publica-
tions. We are proud of the performance of the European
system and are quite sure that Analytical and Bioanalyti-
cal Chemistry will become a major force in the near fu-
ture.
The appearance of Analytical and Bioanalytical Chem-
istry is quite timely. At present, the European Commis-
sion in Brussels is preparing a directive dealing with new
safety and handling rules for commercial chemical prod-
ucts, some of which have been on the market for many
years without any problems. The balance between safety
in its broadest sense and the needs of the market is a deli-
cate one, and the importance of these new rules for the
chemical industry and its partners in academic and indus-
trial research is quite obvious. Naturally, the regulations
are based upon results obtained from analytical testing.
For this reason, Analytical and Bioanalytical Chemistry
has a major role to play in bridging the divide between the
needs of the chemical industry and the environment. Only
a reasonable, scientifically sound approach can provide
the compromise necessary for innovation and economic
viability within the European chemical industry whilst
minimising risks to the environment and human health.
This exciting and never-ending challenge is one to which
the community of analytical chemists will surely contribute
handsomely.
Franois Mathey
President of the French Chemi-
cal Society
Professor of Chemistry at cole
Polytechnique
Director of Research of C.N.R.S.
Member of the French Academy
of Sciences, Leopoldina Acad-
emy and Academia Europaea.

F. Mathey ()
Socit Franaise de Chimie,
250, rue Saint-Jacques, 75005 Paris, France
e-mail: francois.mathey@polytechnique.fr
Anal Bioanal Chem (2002) 372 : 23
DOI 10.1007/s00216-001-1165-y
E D I TO R I A L
Heindirk tom Dieck
A synthesis for analytics creation of Analytical
and Bioanalytical Chemistry
Published online: 14 December 2001
Springer-Verlag 2001
Some years ago, in 1997, several European Chemical So-
cieties decided on a very important step: They agreed to
join forces, to sacrifice their national general chemistry
journals and to create the commonly owned and edited Eu-
ropean Journals. This was decades after the Rome Treaty
establishing the European Community (1958), and even
the more difficult step of bringing about the European
Union was achieved earlier in 1992 with the Maastricht
Treaty. So, politics was more courageous than science?
Yes and No. Yes, in the sense of supporting a necessary
development to overcome too weak and small-scale struc-
tures. No, if one compares the irreversible and complete
merger of the old, traditional journals with the small quanta
of sovereignty transferred by individual nations to the
central European authorities.
The concept of natural sciences lies in the search for
truth, objectivity and reproducibility. Thus science should
develop along lines which do not depend on national bor-
ders and similar barriers. In reality, progress in science has
unfolded in parallel to societies, sometimes in a national
direction and sometimes la mode.
Once born and realised, the idea of creating European
Journals set further steps in motion. The British Royal So-
ciety of Chemistry built bridges to the continent with their
offer to form the journal Physical Chemistry Chemical
Physics. From the very beginning it was evident that the
field of analytical chemistry needed a corresponding ini-
tiative, although this most important scientific domain is
already weakened by so many specialised journals. At the
worlds largest book fair in October 2000 in Frankfurt the
idea was presented to publishing colleagues at Springer-
Verlag. It took only a few weeks to develop the outline for
a publishing project in Analytical and Bioanalytical Chem-
istry. Thereafter, the merger of the French journal Analu-
sis and the German journal Fresenius Journal of Analyt-
ical Chemistry was agreed on by the French Chemical
H. tom Dieck ()
Gesellschaft Deutscher Chemiker,
Postfach 900440, 60444 Frankfurt, Germany
e-mail: htd@gdch.de
Society (SFC), the German Chemical Society (GDCh)
and Springer-Verlag. At the moment of writing I know
that our Spanish colleagues, also, will very soon become a
member of the owners group and will merge their journal
Qumica Analtica. And many others, with no own na-
tional journal to merge, will be invited to become partners
of Analytical and Bioanalytical Chemistry. European co-
operation and European ownership have nothing to do
with a geographic delimitation: the authorship, readership
and the visibility of science published in Analytical and
Bioanalytical Chemistry will be world-wide.
During the 20th century the art of synthesis occupied a
premier position in chemistry. Analytical chemists in Ger-
many frequently complained about ever lesser impact of
their field in the university curricula or about the low es-
teem of their work in chemical companies. This was the
situation yesterday; synthesis may still dominate as long
as mass products and commodities are profitable for clas-
sical chemical industries and craftsmens techniques pre-
vail in education. But tomorrow added value will largely
depend on analytical science. Product liability and quality
management, GLP and GMP, accreditation and certifica-
tion all these are terms that demonstrate the leading role
of the analytical sciences and their great impact on syn-
thesis and production. Health, safety, and environmental
care are fields that relate to qualities, quantities, properties
and effects, which can only be defined, determined and
judged with analytical tools. Modern chemical biology is
a striking example of the dominance of analytical science.
The syntheses were developed and optimised by nature.
Synthetic tools to prepare GMOs are nature-borne. And it
is a mere analytical task to decipher the letters, the words
and the entire script of life. Genomics, proteomics and
forthcoming even more complex omics are chemical
and bioanalytical challenges. The efforts of mankind to
influence the cellular game, for example, in drug develop-
ment, from combinatorial chemistry to clinical testing
how much of this is synthesis, how dominant is analysis!
Even the world of economics will soon comprehend that
innovative value creation depends more on the analytical
sciences than on chemical synthesis.

Thus, the prospects for our new journal Analytical and
Bioanalytical Chemistry, the result of a synthesis from a
manifold of components the merged journals, the co-op-
erating societies, the experienced publisher are excel-
lent.
3
Heindirk tom Dieck is the Man-
aging Director of the Gesellschaft
Deutscher Chemiker, Frankfurt,
Germany, since 1991. He re-
ceived his Ph.D. in chemistry
(1966) after studies at the Uni-
versities of Gttingen and Mn-
chen. After his Habilitation early
in 1971 he became an Associate
Professor of Organometallic and
Coordination Chemistry at the
Goethe University of Frankfurt/
Main and, in 1977, Full Profes-
sor and Director of the Institute
of Inorganic and Applied Chem-
istry, University of Hamburg.
Anal Bioanal Chem (2002) 372 : 45
DOI 10.1007/s00216-001-1167-9
E D I TO R I A L
Christina E. Dyllick Peter Enders
Dear Readers
Published online: 14 December 2001
Springer-Verlag 2001
Welcome to the inaugural issue of Analytical and Bioana-
lytical Chemistry. Three chemical societies, GDCh, SFC
and SEQA, and Springer-Verlag have joined forces to cre-
ate this new journal, which is the merger of three distin-
guished European journals (Fresenius Journal of Analyt-
ical Chemistry, Analusis and Qumica Analtica). The in-
creasing need to develop fast, reliable and environmen-
tally friendly analytical procedures or even complete ana-
lytical strategies within a very narrow time frame is a task
which requires more than ever before the extended in-
terdisciplinary cooperation between experts in the fields.
Therefore, the mission of Analytical and Bioanalytical
Chemistry is: to establish a truly international forum for
analytical and bioanalytical science with a pronounced in-
terdisciplinary, and also a practice-oriented character, and
to publish the information fast in print and electronic for-
mat.
Consequently, the scope of the merged journal will be
broad, encompassing the entire range of analytical and bio-
analytical research and its application, including topics at
the interfaces with the materials sciences, environmental
science, the life sciences, and others. It will embody and
encourage a multidisciplinary approach to solving analyt-
ical and bioanalytical problems. Analytical and Bioanalyt-
ical Chemistry will give greater coverage to the flourishing
and vibrant field of bioanalysis than its predecessors, both
in regular issues and in a number of special issues. In fact,
the next issue will be a special issue dedicated to Biosen-
sors and Biochips.
C.E. Dyllick () P. Enders
Analytical and Bioanalytical Chemistry, Springer-Verlag,
Tiergartenstrasse 17, 69121 Heidelberg, Germany
e-mail: abc@springer.de
Analytical and Bioanalytical Chemistry is published
twice monthly under the direction of an eminent team of
six internationally renowned Editors: S. Daunert, P. Gar-
rigues, G. Gauglitz, K.G. Heumann, K. Jinno and S.A.
Wise. You can read their informative biographical por-
traits on the following pages. Committed to the publication
of high-quality and high-impact analytical research, the
Editors will ensure that especially innovative research pa-
pers are published rapidly within 8 weeks of receipt, as
Papers in Forefront. Furthermore, it is upon their initia-
tive that color images necessary to the understanding of a
contribution will now be printed free-of-charge to the au-
thor. Moreover, the work of the Editors will be strength-
ened by a newly formed International Advisory Board. Its
members represent 20 nations as well as the most im-
portant fields of analytical and bioanalytical research.
In this first issue we are pleased to present a number of
special contributions. A feature article describing the chal-
lenges facing entrepreneurs in bioanalytical chemistry is
followed by Trends 2002, with 17 invited overviews by
predominantly younger scientists spotlighting current
trends in diverse areas of research. Four critical reviews
then provide in-depth coverage of several topics of high
analytical and bioanalytical interest, including the chal-
lenges of ion mobility spectrometry and the potential of ap-
tasensors. The main part of the issue is devoted to the pre-
sentation of high-quality, peer-reviewed original research
papers. One of them is accompanied by a video chro-
matogram which can be viewed as supplementary elec-
tronic material under the journals homepage in Springers
LINK service (http://link.springer.de/journals/abc). The
issue closes with Industry news, which describes how new
biochemical methods are revolutionizing work in the in-
dustrial laboratory.
We invite you to turn the pages of this inaugural issue
and take a look at the breadth and variety of papers pub-
lished in Analytical and Bioanalytical Chemistry.

Christina Dyllick is Managing
Editor of Analytical and Bioan-
alytical Chemistry. She has been
involved in the editing and pub-
lishing of scientific books and
journals since 1981. She received
her Ph.D. in chemistry from ETH,
Zrich (1976) and her under-
graduate education in chemistry
from the University of Bristol
(UK).
5
Peter Enders is Senior Execu-
tive Editor Chemistry and Food
Science with Springer-Verlag in
Heidelberg, Germany. He has
been active in publishing since
1975. He received his education
in chemistry, forensic science and
toxicology at the University of
Erlangen (Germany).
Anal Bioanal Chem (2002) 372 : 68
DOI 10.1007/s00216-001-1168-8

E D I TO R S

Meet the Editors

Published online: 14 December 2001

Springer-Verlag 2001

Analytical and Bioanalytical Chemistry is prepared under the leadership of six eminent analytical and bioanalytical sci-
entists. We invite you to meet the Editors of the new journal in the following biographical portraits.

Sylvia Daunert is an Associate Professor of Chemistry and Pharmaceutical Sciences

at the University of Kentucky, Lexington, USA, and an Associate Member of the
Center of Membrane Sciences. She received a Pharm. D. degree from the University
of Barcelona, a M.S. in Medicinal Chemistry from the University of Michigan in
1985, and a Ph. D. in Bioanalytical Chemistry from the University of Barcelona in
1991. She has been a Fulbright Scholar, and has received among others the Juan
Abell Pascual Award in Biochemistry from the Spanish Royal Academy of Doctors,
the Van Slyke Research Award from the American Association for Clinical Chem-
istry, a National Science Foundation-CAREER Award, a Cottrell-Scholars Award,
and a Lilly Analytical Faculty Award. She is a member of the Editorial Advisory
Boards of Talanta, and of the A-Page Panel of Analytical Chemistry. Dr. Daunert has
served in several National Committees of the ACS, and is currently the Chair of the
Education Committee of the Division of Analytical Chemistry of the ACS. Her re-
search interests lie in the area of bioanalytical chemistry, at the interface between an-
alytical chemistry and molecular biology. More specifically, her group employs re-
combinant DNA technology to design new assays and biosensors based on geneti-
cally engineered proteins and cells. An additional research focus of her group is in the
design of sensing arrays for the detection of molecules in small volumes and mi-
crofluidic platforms. Dr. Daunert is the author/co-author of 100 original research pub-
lications, book chapters, and of several patents in the area of bioanalytical chemistry.

Philippe Garrigues is currently a CNRS Research Director and the Head of the
Environmental and Toxicological Chemistry Laboratory (UMR 5472 CNRS) at the
University of Bordeaux I, France. He received his degree in Chemical Engineering at
the University Louis Pasteur (Strasbourg) in 1978 and his Ph.D. degree in 1985 in
Physical Chemistry at the University of Bordeaux I. He also received a degree in Ma-
rine Ecology and Biology in 1979.
Dr. Garrigues research interests are the analytical aspects (chromatographic and
spectroscopic) related to the detection of organic pollutants and the fate of these com-
pounds in the environment. In particular, he has demonstrated the capability of low
temperature spectrofluorimetry (Shpolskii Effect) for the analysis of aromatic com-
pounds. In cooperation with colleagues in toxicology, he has developed the use of bio-
chemical markers as early warning systems for the evaluation of the ecosystem health
of aquatic ecosystems through the coordination of large research projects (BIOMAR,
BEEP), mainly supported by DG Research (European Commission, Brussels).
Dr. Garrigues has authored about 150 publications. He has been the President of
ISPAC (International Society of Polycyclic Aromatic Compounds; 19971999) and
member of the Board of Directors of SETAC-Europe (Society of Environmental Tox-
icology and Chemistry ; 19961998). He chaired the International Symposia on Poly-
cyclic Aromatic Compounds (ISPAC) (Bordeaux ; 1991 and 1999) ) and was the Sci-
entific chair of the 8th SETAC-Europe Meeting (Bordeaux, 1998). He is also the
French representative in the Division Chemistry in the Environment of the Federa-
tion of the European Chemical Societies (FECS), and is a member of various scien-
tific committees at the national (French Ministries of Research, Environment, Indus-
try ; CNRS) and European level dealing with analytical, toxicological and environ-
mental aspects. He also presently serves as Editor-in-Chief for the journal Polycyclic
Aromatic Compounds and is a member of the editorial boards of various international
journals dealing with analytical and environmental chemistry.
 7
Gnter Gauglitz is Professor in the Institute of Physical and Theoretical Chemistry
at the Eberhard - Karls - University, Tuebingen, Germany, since 1987. He received his
Master of Science in 1966, at the State University of Iowa, and his Ph.D. degree in
1972, in Tuebingen, Germany. Since 1992, he is a titular member of the IUPAC com-
mission V.4 in the Analytical Chemistry Division, and has chaired this commission
since 2000. In 2001, he was elected a member of the IUPAC Division of Analytical
Chemistry. Since 1994, Dr. Gauglitz is a Member of the Graduiertenkolleg Analyti-
cal Chemistry at the University of Tuebingen, and acts as chairman for this institution
since 2000.
In 1997, he received the Wallac Award of the Society of Biomolecular Screening for
his research in fluorescent high-throughput screening. He is a member of the board of
the Deutscher Arbeitskreis fr Angewandte Spektroskopie (DASp), and is chairman
for the working group of Chemometrics and Lab Data Processing within the Analyt-
ical Chemistry Division of the Gesellschaft Deutscher Chemiker (GDCh). For the
past two years he has chaired the working group Chemical and Biochemical Sensors.
He is a member of the Editorial Board of the journals Trends in Analytical Chemistry
and Chemometrics and Intelligent laboratory Systems.
At present, there are over 20 Ph.D. and Master students in his research group, work-
ing in the field of UV absorption and fluorescence spectrometry, in the polymeriza-
tion of silicon oligomers, photochemical processes in photochromic systems and
liquid crystals, spectral refractometry in micro volume sizes and in the develop-
ment of various types of optical detection principles for chemical and biochemical
sensing, such as reflectometric interference spectroscopy, surface plasmon resonance,
integrated optical Mach-Zehnder devices, total internal reflexion fluorescence, and
fluorescence resonance energy transfer. He uses spectral ellipsometry to characterize
interphases of polymers and biomembrane surfaces applied to the detection and quan-
tification of molecular and biomolecular interaction, such as, the measurement of
volatile organic compounds in combination with various methods of multivariate data
analysis using chemometrics, and in the field of biosensing. Antibody and antigene
interactions are used in the case of pesticides and estrogens, and recently DNA and
protein interaction have been added, providing methods for high-throughput screen-
ing in the area of drug screening, as well as aiming toward the areas of genomics and
proteomics. The scientific results of his research group have been published in over
185 publications, many book chapters, and 3 monographs and have also been pre-
sented in many lectures at national and international conferences, resulting in many
proceedings and short communications. In addition, more than 10 patents have been
obtained in the fields of actinometry, integrated optics, interferometry, and high-
throughput screening devices.

Klaus Gustav Heumann is Professor of Analytical Chemistry at the Johannes-Gu-

tenberg-University Mainz, Germany. He received his Diploma degree in Chemistry
at the Technical University Darmstadt in 1966. His Ph.D. thesis work was carried out
at the Institute of Nuclear Chemistry at the same university, where he developed an
interest in mass spectrometry, an instrumentation he and his group are now special-
ized in. After his Habilitation in Analytical Chemistry, he was appointed Professor
for Inorganic and Analytical Chemistry at the University of Regensburg, where he
taught 1974 to 1996. During this time he declined two offers as Professor for Analyt-
ical Chemistry, one from the Technical University of Graz/Austria, and another from
the University of Kassel/Germany, as well as an offer as head of the analytical de-
partment of the Bundesanstalt fr Materialforschung und -prfung (BAM) in Berlin.
He accepted the offer as Professor for Analytical Chemistry at the University of
Mainz in 1996.
Dr. Heumanns research interests lie in the development and application of analytical
methods for the determination of trace elements and trace amounts of elemental
species, using ICP-MS, TIMS, different types of optical atomic spectrometry, and
electroanalysis as detection methods, and HPLC or capillary GC as separation meth-
ods. His research group has considerable experience in isotope dilution mass spec-
trometry (IDMS). They were the first to use IDMS in connection with the negative
thermal ionization technique and to introduce on-line IDMS into hyphenated tech-
niques for element speciation. Dr. Heumann has received several awards that include
the Oc-van-der-Grinten award for a significant development in environmental ana-
lytical chemistry.
Dr. Heumann is a member of several national and international scientific societies,
e.g. the German Chemical Society and the German as well as the American Mass
Spectrometric Societies. He served as a member of the Editorial Board of Fresenius
Journal of Analytical Chemistry for many years. He was a member of the IUPAC
Commission on Atomic Weights and Isotopic Abundances for twelve years and
chairman from 1991 to 1995. In addition, he has served as a member of the IUPAC
Inorganic Chemistry Division.
8
Kiyokatsu Jinno is currently Professor and Chair of the School of Materials Science,
Toyohashi University of Technology, Toyohashi, Japan. He received his B.S. and M.S.
degrees from Nagoya University and his doctoral degree from the Department of Ap-
plied Chemistry in Nagoya University in 1973 on radioanalytical chemistry. He then
joined Toshiba Research and Development Center for LSI fabrication processes,
where he worked, in particular, on plasma dry etching, electron beam lithography and
related areas for five years. In 1978 he was invited to join the faculty of the newly
constructed national university at Toyohashi, Toyohashi University of Technology.
As a member of the faculty of analytical chemistry, his research initially focused on
the separation sciences and spectroscopic areas.
Dr. Jinno has now been working in the area of separation chemistry for over 20 years
and has published about 300 articles on high resolution separation methods, with a
special emphasis on capillary liquid phase separations including micro-LC, capillary
zone electrophoresis, and capillary electrochromatography. His work on novel sta-
tionary phases with size and shape selectivity is particularly well known. He has also
focused on hyphenated techniques between capillary-based separation methods and
spectroscopic methods such as IR, MS and ICP. His major contribution in the devel-
opment of a new interface approach in the LC-IR field is especially well known. He
has tried to use computers in chromatography for retention prediction, optimization
of separation conditions and also identification of solutes separated by chromatogra-
phy. The international analytical chemistry community has also acknowledged his re-
cent interests in sample preparation and its miniaturization. In recognition of his in-
novative work he has received several awards worldwide, such as the Tswett medal
from the Russian Academy of Sciences, the Golay Award from the International
Symposium on Capillary Chromatography, the Award of the Society for Chromato-
graphic Sciences (Japan), and also the Helsinki University Medal. He is currently
working as an editorial member of over 15 journals and magazines in this area.

Stephen A. Wise is the Leader of the Organic Analytical Methods Group within the
Analytical Chemistry Division of the National Institute of Standards and Technology
(NIST), Gaithersburg, Maryland, USA. He received a B.A. in Chemistry from Weber
State University (Ogden, Utah) and a Ph.D. in Analytical Chemistry from Arizona
State University (Tempe, Arizona). Dr. Wise joined NIST (then the National Bureau
of Standards) in 1976 as a research chemist working primarily in the area of chro-
matographic separations. His research interests have focused on: (1) the study of re-
tention and selectivity in reversed-phase liquid chromatography; (2) the development
of chromatographic methods for the determination of organic pollutants (e.g., poly-
cyclic aromatic hydrocarbons, polychlorinated biphenyls, and chlorinated pesticides)
in environmental matrices such as sediment, tissue, and air particulate matter; (3) the
development of Standard Reference Materials for trace organic constituents in envi-
ronmental, clinical, and food matrices; and (4) environmental specimen banking pro-
cedures. He has authored or co-authored over 220 publications in these areas.
Dr. Wise has been associated with Fresenius Journal of Analytical Chemistry since
1991, first as a member of the Advisory Board, and since 1995 as Regional Editor for
North America. He is currently the Topical Editor for Analytical Separation Tech-
niques for Polycyclic Aromatic Compounds. Dr. Wise served as Chairman of the
Division of Analytical Chemistry of the American Chemical Society in 1996 and as
a member of the Editorial Advisory Board for Analytical Chemistry (19962000)
and the Journal of Microcolumn Separations (19912001). Since 1988 he has been an
adjunct faculty member in the Department of Chemistry at Georgetown University
(Washington, DC).
Anal Bioanal Chem (2002) 372 : 911
DOI 10.1007/s00216-001-1143-4
F E AT U R E A RT I C L E
Patricia E. Eisenhardt
Bioentrepreneurship 2002: at home in your lab
Published online: 11 December 2001
Springer-Verlag 2001
As our economy is increasingly dominated by globaliza-
tion, at a time when connectivity is rapidly erasing age-old
geographic and cultural boundaries, a remarkable thing is
happening in the business world. There beneath the shadow
of world-wide corporate conglomerates, cottage industry
is resurgent...literally.
Inc. 500, the 2000 ranking of the fastest-growing pri-
vate companies in America, reports that 61% of the com-
panies ranked were started in the founders home, as op-
posed to 39% of companies listed in the Inc. 500 of 1995
[1]. Many of these companies are high-tech and yet, over
half had initial start-up capital of less than $20,000 (USD).
The interesting point here is that some of the most vigor-
ous companies around today are those that start out small
and independent. Even in the life-sciences sector it is evi-
dent that there are forces at work that favor success of
small ventures. Now, this is not necessarily to suggest that
you, the bioanalytical chemist, go set up a laboratory in
your kitchen, but rather to emphasize that you are blessed
with many viable opportunities for bioentrepreneurship.
Looking out from the grassy knoll of expansive re-
search parks where pharma giants roam free, this idea of a
cottage industry might seem incredulous. One has only to
glance at Focus on Fundamentals: The Biotechnology Re-
port, Ernst & Youngs 15th Annual Review, to see abundant
alliances among pharma and biotech that include record-
breaking deals. In todays global marketplace, cross-bor-
der consolidation is generally accepted as a key to sur-
vival and growth, but inherent in this situation is a fasci-
nating twist. Authors Scott Morrison and Glen Giovan-
netti explain that ...this global network has allowed com-
panies from smaller regional markets to grow and com-
pete [2]. So, as quickly as more mature life science com-
panies are consolidating, new ones are emerging and seem
to promise to achieve even greater growth and vertical in-
tegration than their forerunners. In European Life Sciences
P.E. Eisenhardt ()
ChipRx, Inc., Columbus, OH, USA
e-mail: eisenhardt@columbus.rr.com
2001: Integration, Jan Leschly, Chairman and CEO of Care
Capital, LLC recommends that biotech companies think
about...going up the value chain themselves, before going
to big pharma [3].
Often, the same dynamics that affect the business cli-
mate for biotechnology also affect that of bioanalytical
chemistry, bringing similar opportunities for company
formation. In an interview with the Wall Street Corporate
Reporter, Peter Kissinger, Chairman and CEO of Bioana-
lytical Systems, Inc. explains: there is...an ecosystem with
regards to drug development where pharmaceutical com-
panies...have numerous partnerships with biotechnology,
drug discovery, development and specialty chemical com-
panies to whom they will outsource discovery, toxicology,
clinical trials (etc.). According to Kissinger, the bottom
line is that this outsource approach will be far more pro-
ductive than building the infrastructure to do everything
internally[4]. So it is that big pharma scouts for small spe-
cialized shops that might be strung together in a network
of resources that collectively can make a new drug a suc-
cess in the market.
This climate is one to which the classical bioanalytical
chemistry business model of a contract research organiza-
tion is well suited. Traditionally, this is a boutique busi-
ness, wherein small independent laboratories boast of highly
specialized and diverse capabilities. In recent years, some
have formed alliances and manage to run impressive full
service operations. Industry expert Alex MacDonald of
North Caldwell, NJ remarks that, irrespective of size, con-
tract research firms succeed by making investments in
people, technology, and the information systems to man-
age it all [5]. Return on these investments tends to take
the form of a fee for service or revenue from sales of prod-
ucts such as software, reagents, or instruments originally
made for in-house use. While one can derive a healthy in-
come in this way, the ebb and flow of demand drives many
to search for alternative, stabilizing sources of cash flow.
As one expands ones view of ways in which bioanalyt-
ical chemistry capability could generate revenue, new op-
portunities become apparent. For example James Scozzie,
President of Ricerca, LLC explains that his firm is giving
10
Fig. 1 Schematic diagram of the ChipRx drug-delivery device
a new definition to product revenue by forming partner-
ships with companies in product development, not on a
fee for service basis, but in exchange for a cut of the prod-
uct sales [6]. While riskier than a classic fee-for-service
model, the approach offers compensation proportional to
the success of the product, revenue which could be far
greater than compensation for the development services
alone.
Another option for the bioentrepreneur is to take bio-
analytical chemistry beyond the borders of drug develop-
ment, upstream into the space of lead discovery. LabBook,
Inc. and its partners are developing the means to visualize
diverse molecular biology data in new ways, accelerating
lead recognition through improvement in knowledge man-
agement. LabBook Chief Science Officer Jeff Spitzner re-
flects, Scientists engaged in discovery have unique knowl-
edge and experience to identify needs. Bioentrepreneurs
have the vision and motivation to fulfill them [7].
Business opportunities in bioanalytical chemistry also
abound downstream of drug development. For instance,
chemical biosensors may serve to guide therapeutic regi-
mens to improve drug safety and efficacy. A case in
point is the company I helped to start-up, ChipRx, Inc.,
which is developing an implantable drug delivery device,
Fig. 1. Loaded with drug and equipped with miniature
valves and biosensors that regulate drug delivery, the de-
vice will be designed to sense changes in the patients phys-
iology and release drug into the body exactly as needed
[8]. Some day the companys unique biochemical sensor
systems and device design could revolutionize human and
veterinary medicine by enabling individualized therapy in
clinical areas where it is desperately needed.
These alternative approaches to the business of analyt-
ical chemistry can be pursued through mature corporations
and new ventures alike. Where one chooses to (hang a
shingle) really depends on ones temperament and de-
sired lifestyle. In Nature Biotechnology, Rochelle Young
describes the successful bioentrepreneur as curious, cre-
ative, committed and willing to change perspectives [9].
More than this, though, the bioentrepreneur must be pre-
pared to be a leader under conditions of substantial uncer-
tainty and ambiguity. Perhaps the title of Tom Peters
amusing and insightful Fast Company article sums it up
best, Rule #3: Leadership is Confusing as Hell [10].
The ChipRx team chose to start-up in nothing more
than my home and a university lab. Although we expect to
outgrow this arrangement soon, it has served us well in
our first year of business. Exhausting and exhilarating at
the same time (although rarely in equal portion) the start-
up of this life science business has been a high-contrast
experience. In a small company, there is no buffer. All the
decisions I make have a direct impact on the company and
I receive credit for these decisions, whether I want it or
not. There is little choice of roles, everyone at ChipRx is
asked to contribute to a variety of activities including op-
erations, business development, and technology develop-
ment. My daily life at ChipRx reminds me of the days
when I lived in Manhattan everything is in your face all
the time, and its all urgent. Despite my high tolerance for
messy situations, it has been no small task to acclimatize
to such an environment. Continuing to adjust, I have
learned to follow these few guiding principles.
Business is personal
The behavior of the people involved in a fledgling busi-
ness can have a tremendous impact on its success. There
are times of difficulty and confusion, when all our team
has to go on is our best judgement and lots of hard work.
The only option at such times is to trust myself and my
business partners. That sense of trust only goes as far as
the commitment of the individuals involved and when it is
strong, a team can achieve great things. In a similar fash-
ion, well-managed, honest relationships with early clients
or strategic partners builds loyalty that can carry through
even the roughest of times.
Perception is critical
A companys public image merits consideration as a cen-
tral component of corporate development strategy. It is
never too soon to start building a strong presence. ChipRx
benefits greatly from the ability of its scientific founders
to network. They attract the attention of the press and pro-
ject a focused, articulate story, infused with grace and
charisma. Over time, the public is becoming familiar with
the company; that will facilitate efforts to expand the busi-
ness and project market acceptance. Customers, partners,
and investors will invest in people and technology with
which they are comfortable.
Cash rules
Cash flow management can make or break a young com-
pany, so the smoother it is, the better. The company does
well always to be on the lookout for the next source of
funding, and to keep a creative mindset, applying impor-
tant assets such as intellectual property or a strong team to
generate opportunities. A venerable old banker once told
me, Keep your eye on the dollar sign...
In the end, bioentrepreneurship is a leap of faith that
arises from love for a technology and love for the role one
plays in making that technology valuable and accessible
to the public. Founders must nurture an emergent venture
 11

and often make great personal and financial sacrifices for A founder and Vice President of
the sake of future prosperity. Business start-up is challeng- ChipRx, Inc., Patricia Eisenhardt
facilitates daily operations and
ing, so it is important to build as good a home as possible strategic corporate development.
for it, even if it is only a modest cottage at first. Now, if Her work in the life sciences has
youll excuse me, I must go tend my work, to keep the spanned from bench to hospital
home fires burning. bedside, and from the teachers
desk of a high school classroom
to the halls of a high-tech busi-
ness incubator. Central to Patri-
References cia Eisenhardts career is her de-
sire to enhance the value and ac-
1. Gendron G (2000) (ed) Inc. 22:6465 cessibility of scientific discover-
2. Ernst and Young (2001) Focus on fundamentals: the biotech- ies to society, and thus to im-
nology report, Ernst and Youngs 15th annual review. Ernst and prove quality of life
Young, LLP, New York, NY
3. Ernst and Young (2001) European life sciences 2001: integra-
tion. Ernst and Young, Cambridge, UK
4. OHanlon J (2001) Wall Street Corporate Reporter 6:14
5. Interview notes. Alex MacDonald may be reached at
BeeMac201@aol.com
6. Interview notes. More information is available at http://www.
ricerca.com
7. Interview notes. More information is available at http://www.
labbook.com
8. More information is available at http://www.chiprx.com
9. Young R (2001) Nature Biotechnol 19:BE9-BE10
10. Peters T (2001) Fast Company 44:124140
Anal Bioanal Chem (2002) 372 : 1213
DOI 10.1007/s00216-001-1163-0
TRENDS
Andrew J. de Mello
Miniaturization
Published online: 8 December 2001
Springer-Verlag 2001
In his now celebrated lecture on the 29th December 1959
Richard Feynman pondered the potential of miniaturiza-
tion in the physical sciences [1]. His vision, based on
known technology, examined the limits set by physical
principles and proposed a variety of new nano-tools in-
cluding the concept of atom by atom fabrication. In the
intervening decades, many of these predictions have be-
come reality; microelectronic systems have shrunk to
sizes close to the molecular level, scanning probe micro-
scopes (e.g. STM and AFM) enable us to image and ma-
nipulate individual atoms, and the molecular machinery
of living systems is now being more fully understood and
harnessed.
Surprisingly, it is only within the last decade that the
concepts of miniaturization have been seriously applied to
chemical and biological problems. Of particular focus and
interest has been the development and application of lab-
on-a-chip technology. These microscale analytical instru-
ments employ micromachined features (such as channels,
electrodes, reactors, and filters) and are able to manipulate
fluid samples with high precision and efficiency. Micro-
fluidic chip devices have been used in a wide variety of
applications including nucleic acid separations, protein
analysis, process control, small-molecule organic synthe-
sis, DNA amplification, immunoassays, DNA sequenc-
ing, and cell manipulations [2, 3]. In a fundamental sense,
chip-based analytical systems have been shown to have
many advantages over their conventional (larger) ana-
logues. These include improved efficiency with regard to
sample size, response times, cost, analytical performance,
process control, integration, throughput, and automation.
Much of the pioneering work in microfluidics has fo-
cused on the successful transfer of established analytical
A.J. de Mello ()
AstraZeneca/SmithKline Beecham Centre
for Analytical Sciences, Department of Chemistry,
Imperial College of Science, Technology and Medicine,
Exhibition Road, South Kensington, London SW7 2AY, UK
e-mail: a.demello@ic.ac.uk
technologies from conventional to microfluidic (chip-
based) formats. In particular, huge leaps in the efficiency
and application of separation techniques have been facili-
tated by miniaturizing column dimensions and creating
monolithic fluidic networks on planar substrates. Indeed,
a survey of the current literature demonstrates that almost
all separation methods (based on electrophoretic or chro-
matographic partitioning mechanisms) have been success-
fully transferred to micromachined formats [3].
Small volumes of sample and reagent (pLnL) are rep-
resentative of most miniaturized systems. This character-
istic has clear advantages associated with cost and analyt-
ical throughput, but does pose constraints on appropriate
detection methods. Consequently, much research has re-
cently focused on the development of detection tech-
niques that are highly sensitive, universal, miniaturized,
and cost-effective. Although the vast majority still utilize
fluorescent schemes (because of exceptional sensitivity)
other approaches to on-chip detection have been intro-
duced. These include electrochemical and electrochemilu-
minescent methods, absorption, indirect fluorescence, re-
fractive index variation, plasma, and thermal conductivity
methods [4]. Although many novel detection methods
have been reported for on-chip applications, future studies
must focus on improving concentration detection limits
and complete integration of the detector with the rest of
the analytical system (Fig. 1).
Of notable current interest is the use of microfabricated
structures for chemical and biological synthesis. Because
of the unique environments afforded within microfluidic
networks, a variety of synthetic processes can be per-
formed in continuous flow and batch formats. Increased
efficiencies of mixing and separation combined with high
rates of thermal and mass transfer make microreactors
ideal for processing valuable or hazardous reaction com-
ponents and improving reaction selectivities (thus yield-
ing higher quality products). DNA amplification, com-
binatorial (and high throughput) chemistry, immuno- and
bioassays, and tissue and cell culture are areas which have
all recently benefited from advances in microfluidic chip
technology [2, 5, 6].

Fig. 1 An integrated analysis system
in which DNA and reagent solutions
are placed on the device and elec-
tronic signals corresponding to ge-
netic information are the primary out-
put. Image taken from Ref. [8]
One of the primary advantages and goals of microfab-
ricating analytical instruments is the large-scale integra-
tion of functional components on a single monolithic de-
vice. Indeed, the ability to extract the required informa-
tion from a chemical or biological system almost always
involves performing a number of distinct analytical oper-
ations. Consequently, a complete device will ideally in-
clude all relevant steps in a complete analytical procedure
(including sample handling, sample pre-treatment, reac-
tion, analytical separation, analyte detection, and product
isolation). Although, this objective is widely recognized
few studies have yet embraced the concept of integration
fully. Of these the focus has primarily centered on the am-
plification and analysis of DNA fragments [2, 7, 8]. Fu-
ture work in this area will need to address integration in a
more general sense and solve problems associated with
materials compatibility and the handling of real-world
samples. It is also likely than many future microfluidic
devices will be constructed from polymeric materials
(rather than silicon or glass), because of the wide avail-
ability of base materials and highly cost-effective fabrica-
tion methods (such as injection molding and hot-emboss-
ing) [9].
It is fair to say that the envisaged applications of mi-
crofluidic analysis chips have undoubtedly expanded well
13
beyond those originally defined in the early nineteen-
nineties. The first commercial devices have been intro-
duced to market (see Table 1 in Ref. [3]), and it is almost
certain that the base technologies defined over the past
decade will have a high future profile in the fields of pro-
teomics, drug discovery, cellomics, clinical diagnostics,
genomics, and chemical production [10].
References
1. Feynman RP (1960) Eng Sci 23:22
2. Krishnan M, Namasivayam V, Lin R, Pal R, Burns MA (2001)
Curr Opin Biotech 12:92
3. Jakeway SC, de Mello AJ, Russell EL (2000) Fresenius J Anal
Chem 366:525
4. Schwarz MA, Hauser PC (2001) Lab-on-a-Chip 1:1
5. Mitchell MC, Spikmans V, Manz A, de Mello AJ (2001)
J Chem Soc Perkin Trans 1:514
6. Jensen KF (2001) Chem Eng Sci 56:293
7. Woolley AT, Hadley D, Landre P, de Mello AJ, Mathies RA,
Northrup MA (1996) Anal Chem 68:4081
8. Burns MA, Johnson BN, Brahmasandra SN, Hanique K, Web-
ster JR, Krishnan M, Sammarco TS, Man PM, Jones D, Held-
singer D, Mastrangelo CH, Burke DT (1998) Science 282:484
9. Becker H, Gartner C (2000) Electrophoresis 21:12
10. Sanders GWH, Manz A (2000) Trends Anal Chem 19:364
Anal Bioanal Chem (2002) 372 : 1415
DOI 10.1007/s00216-001-1156-z

TRENDS

Yoshinobu Baba

Genomics

Published online: 8 December 2001

Springer-Verlag 2001

Because sequencing of the human genome is almost com-

plete [1, 2], the human genome project will soon enter the
post-genome-sequencing era, including single nucleotide
polymorphism (SNP) analysis, functional genomics, mu-
tation analysis, transcriptome analysis, proteome analysis,
and metabolome analysis directed, in the future, towards
personalized medicine [3, 4, 5, 6, 7, 8, 9]. Capillary array
electrophoresis with 96384 capillaries has played a vital
role in the genome sequencing era, and an 108- to 109-fold
increase in DNA sequencing efficiency was achieved dur-
ing the last three decades (Fig. 1). In the post genome se-
quencing era, however, further development of analytical
technology for DNA, mRNA, protein, and metabolites is
urgently required. Even exa (1018)-order genome sequenc-
ing, corresponding to a 103-fold increase every ten years,
is required in 2030, when personalized medicine, which
requires sequencing of the personal genome, is expected
to be realized (Fig. 1). There are 3 tera (1012) base pairs
(bp) for a thousand people, 3 peta (1015) bp for a million
people, and 3 exa (1018) bp for a billion people, because a
person has 3 giga bp. Inexpensive genome sequencing
should also be accomplished during the next three
decades. Sequencing costs have dropped 100-fold over
the last 10 years, corresponding to an approximately two-
fold decrease every 18 months. This rate is similar to
Moores law governing improvements in semiconduc-
tor manufacture. Personal genome sequencing still costs
ca. 300500 million US$, however. The cost of personal
sequencing should decrease to 1000 US$ in 2030 for per-
sonalized medicine.
Such substantial improvement in the speed and cost of
sequencing requires aggressive technological innovation
fuelled by major investment. Improvements are needed to
Y. Baba ()
Department of Medicinal Chemistry,
Faculty of Pharmaceutical Sciences,
The University of Tokushima, CREST,
Japan Science and Technology Corporation (JST),
Shomachi, Tokushima 7708505, Japan
e-mail: ymttbaba@ph.tokushima-u.ac.jp
Fig. 1 DNA sequencing road map towards future Exa (1018)-order
genome sequencing
move current di-deoxy sequencing to smaller volumes
and more rapid sequencing, based upon advances such as
microchannel electrophoresis. More revolutionary meth-
ods, e.g. mass spectrometry (MS) and nanotechnology in-
cluding single-molecule sequencing and nanopore ap-
proaches, have not been fully developed, but hold great
promise and are worthy of strong encouragement.
Microchip-based technology will be fundamental to
the post-genome sequencing era [10], because even revo-
lutionary methods and microchannel electrophoresis will
be able to be developed on a chip and will be applicable to
the analysis of DNA, mRNA, protein, and metabolites.
Recent progress in microfabrication and nanofabrication
technologies based on computer-chip technology have in-
troduced a new research area for integrated microchip and
nanochip technology. Investigation of integrated micro-
chip technology is developing towards the integration of
all the steps required for personalized medicine, i.e. ex-
tracting the genomic DNA from the cell, amplification
and specific reactions of DNA, separation of DNA, hy-
bridization of DNA, and detection of DNA for genome
analysis and extraction of protein from a cell, two-dimen-
sional separation, enzyme digestion, detection of protein,
and interfacing to MS for proteome analysis. In the near
future so-called exa-bioinformatics, which integrates both
ultra-fast information technology and exa-sequencing

technology based on microchip and nanotechnology, will
be launched and enable us to obtain the huge amounts of
information required for personalized medicine.
References
1. International Human Genome Sequencing Consortium (2001)
Nature 409:860921
2. Venter, JC et al. (2001) Science 291:13041351
15
3. Baba Y, Figeys D (2001) Genomics and proteomics for analyt-
ical chemists. John Wiley and Sons, New York
4. Vulkmirovic OG, Tilghman SM (2000) Nature 405:820822
5. Service RF (2000) Science 287:19541956
6. Pandey A, Mann M (2000) Nature 405:837846
7. Roses AD (2000) Nature 405:85786
8. Evans WE, Relling MV (1999) Science 286:487491
9. Sander C (2000) Science 287:19771978
10. Heller MH, Guttman A (2001) (eds) Integrated microfabricated
device technologies: advances in genomic, drug discovery, and
clinical diagnostic applications. Marcel Dekker, New York
Anal Bioanal Chem (2002) 372 : 1617
DOI 10.1007/s00216-001-1149-y

TRENDS

Gunther Wittstock

Sensor arrays and array sensors

Published online: 8 December 2001

Springer-Verlag 2001

Electrochemical sensors are based either on the measure- microelectrode arrays, with a larger inter-electrode dis-
ment of a potentiometric, impedimetric, capacitive, ampero- tance, from a variety of electrode materials [2, 3]. The
metric, or voltammetric response of a specifically designed commercially available RAM electrodes (random assem-
electrode. This contribution considers transducers based bly of microelectrodes) also fall in this category [4]. An
on amperometric and voltammetric principles. In contrast important application is sensors for anodic stripping analy-
with other detection principles these signals are generated sis of heavy metals which make use of the effective mass
by ongoing electrochemical conversion of the species to be transport during the electrolytic deposition process. De-
detected. If such sensors are fabricated with one dimen- tection in flow systems would be another field in which
sion smaller than the diffusion length of the analyte, their such structures have potential advantages because their
performance might be dramatically improved because of: response to fluctuations in the flow rate is much less sen-
sitive than that of macroscopic electrodes. Parallel ultra-
1. increased mass transport towards the transducer (radial
microelectrode arrays with small inter-electrode distances
diffusion);
behave similarly to conventional macroscopic electrodes
2. reduced double-layer capacity (small electrode area re-
on the typical time-scales used for sensors, because the
sults in lower background current); and
diffusion layers of the individual electrodes overlap. Dra-
3. reduced Ohmic drop (small currents).
matically improved detection limits have, however, been
The term ultramicroelectrode is used to characterize elec- demonstrated with electrode arrays template-synthesized
trodes with such properties. The use of such miniaturized from track-etch membranes [5], because of their extremely
systems in modern electroanalytical chemistry follows three
basic strategies (Fig. 1a):
An assembly of such microelectrodes is operated in par-
allel (array sensors) to exploit the advantageous response
properties of such electrodes while increasing the cur-
rent to a level easily measurable even outside a labora-
tory environment (Fig. 1a).
Two comb-shaped arrangements are interdigitated to en-
able redox recycling between the two combs acting as
anode and cathode (Fig. 1b).
Independent addressable sensors are used to construct
multianalyte sensors (sensor arrays; Fig. 1c).
There is already a rather large body of literature dealing
with the fabrication and characterization [1] of array sen-
sors. Recent reports describe efforts to produce parallel

G. Wittstock ()
Carl von Ossietzky University Oldenburg,
Department of Chemistry (FB9) and Institute of Chemistry
and Biology of the Marine Environment (ICBM),
PF 2503, 26111 Oldenburg, Germany
e-mail: gunther.wittstock@uni-oldenburg.de
Fig. 1 Schematic diagrams of: (a) in-parallel-operated ultramicro-
electrodes (array sensor), (b) an interdigitated electrode array for
sensitivity enhancement by redox recycling, and (c) independently
addressable ultramicroelectrodes (sensor arrays)

low double-layer capacity. Such electrodes are, further-
more, also intensively investigated as microcompartmen-
talized electrodes [6]. Such surfaces unite segregated mi-
croscopic domains for different functions, for example sup-
port of biomolecular layers and bare regions for efficient
electrochemical detection [7]. Sensors based on such prin-
ciples can have very short response times.
Interdigitated electrode arrays (IDA) consist of two elec-
trodes, one serving as a generator, the other operating as a
collector at which a reversible or quasi-reversible redox
couple can undergo multiple oxidation reduction cycles.
The different redox forms are transported by diffusion be-
tween the adjacent band electrodes. Theory predicts a dra-
matic increase of sensitivity as the inter-electrode distance
is reduced and the number of band electrodes per given
area is increased [8]. The integration of biological recog-
nition elements adjacent to or between the electrode array
has also been advanced. Redox-active species generated
by an label enzyme (e.g. 4-aminophenol formed by alka-
line phosphatase) are then redox-recycled at the IDA. Ef-
fort is currently being devoted to the construction of ar-
rays of IDA for the parallelization of such measurements
[9].
Arrays of individually addressable electrodes can be
used to determine different analytes or to provide spatially
resolved measurements. Exposing the cross sections of
bonding wires of conventional electronic circuits provides
convenient access to individually addressable electrode
arrays in explorative research [10]. Although modification
of the individual electrodes with different biological
recognition elements would enable the construction of
miniaturized biosensor arrays, the directed immobiliza-
tion of functional proteins on individual microscopic re-
gions is still a challenge. Scanning electrochemical mi-
croscopy (SECM) has become an ideal tool both for pro-
totyping of such structures and for the analysis of differ-
ent aspects of their performance [11]. Optimizing the lay-
17
out for suppression of cross-talk is another fundamental
problem. Cross-talk in electrochemical enzymatic sensor
arrays occurs if a product of an enzymatic reaction, e.g.
H2O2, diffuses to the neighboring array element causing a
signal even though the corresponding analyte is not pre-
sent. Strategies used to avoid this situation include:
use of interference-suppressing enzyme layers (e.g., con-
taining catalase to decompose H2O2) [12];
electrolysis of the interference-causing compound at
band electrodes placed between the individual sensors
[13]; or
splitting the sample flow into independent parallel
flows before reaching the individual sensors [14].
References
1. Wittstock G, Grndig B, Strehlitz B, Zimmer K (1998) Elec-
troanalysis 10:526531
2. Feeney R, Kounaves SP (2000) Electroanalysis 12:677684
3. Suzuki H (2000) Electroanalysis 12:703715
4. Fletcher S, Horn MD (1999) Electrochem Commun 1:502512
5. Menon VP, Martin CR (1995) Anal Chem 67:19201928
6. Ratcliff BB, Klancke JW, Koppang MD, Engstrom RC (1996)
Anal Chem 68:20102014
7. Nowall W, Dontha N, Kuhr WG (1998) Biosensors Bioelec-
tronics 13:12371244
8. Aoki K, Morita M, Niwa O, Tabei H (1988) J Electroanal
Chem 256:269282
9. Hintsche R, Albers J, Bernt H, Eder A (2000) Electroanalysis
12:660665 (and references cited therein)
10. Matos RC, Augelli MA, Lago CL, Angnes L (2000) Anal Chim
Acta 404:151157
11. Wittstock G (2001) Fresenius J Anal Chem 370:303315
12. Urban GA, Jobst G (1997) In Scheller FW, Schubert F,
Fedrowitz J (eds) Frontiers in biosensorics, vol 2. Birkhuser,
Basel, pp 161171
13. Strike DJ, Hengstenberg A, Quinto M, Kurzawa C, Koudelka-
Hep M, Schuhmann W (1999) Mikrochim Acta 131:4755
14. Kurita R, Tabei H, Hayashi K, Horiuchi T, Torimitsu K, Niwa
O (2001) Anal Chim Acta 441:165174
Anal Bioanal Chem (2002) 372 : 1819
DOI 10.1007/s00216-001-1151-4
TRENDS
Joseph J. Dalluge
Matrix-assisted laser desorption ionization-mass spectrometry
(MALDI-MS)
Published online: 8 December 2001
Springer-Verlag 2001
Abbreviations MALDI-MS Matrix-assisted laser
desorption ionization mass spectrometry GPC gel
permeation chromatography SNP single nucleotide
polymorphism
Matrix-assisted laser desorption ionization mass spectrom-
etry (MALDI-MS) [1] is one of two modern mass spec-
trometric techniques (the other being electrospray ioniza-
tion mass spectrometry) that have, over the past decade,
become essential tools in the analytical and life sciences
laboratories. Because the early history of MALDI-MS,
discussions of instrumental and experimental design, and
proposed ion desorption mechanisms are detailed in
nearly every reference cited herein, these topics will not
be reviewed. Instead, this article will focus on selected
scientific endeavors that rely on MALDI-MS as a stan-
dard tool for identification and characterization of the
wide variety of analytes that are measurable using this
technique. While this overview by no means encompasses
every area of research in which MALDI-MS is involved,
it should illustrate the versatility of the technique, and the
leading role MALDI mass spectrometry plays in the mod-
ern analytical laboratory.
MALDI-MS is a proven technology in the emerging
field of proteomics (functional genomics). With the eluci-
dation of the human genome complete, efforts are now
shifting toward the large-scale analysis of the function of
genes that is regulation of protein expression. Genome se-
quence information has provided a comprehensive re-
source for protein identification using data produced by
MALDI-MS. New methods for direct identification of pro-
teins in mixtures using enzymatic proteolysis, MALDI-MS,
and computer algorithms designed to match peptide mass
J.J. Dalluge ()
Central Research, Cargill, Incorporated,
P.O. 5702, Minneapolis, MN 554405702, USA
e-mail: joseph_dalluge@cargill.com
spectra to database sequences allow monitoring of the ex-
pression of any protein from any organism for which the
corresponding genome has been defined. In addition,
MALDI-MS has the capability of identifying posttransla-
tional modifications that alter the mass of a protein. As
such, MALDI-MS is being employed to characterize pro-
teinprotein interactions, cellular compartments, and sig-
naling pathways. Recent review articles detailing these
endeavors are available to the interested reader [2, 3, 4].
Some of the most compelling applications of MALDI-MS
involve direct characterization of genes themselves, partic-
ularly single nucleotide polymorphism (SNP) analysis [5].
SNPs are the most abundant type of variation found in the
human genome. As such, they are becoming popular ge-
netic markers for genome mapping studies, medical diag-
nostics, and identity testing. MALDI-MS has been demon-
strated as a promising tool for high-throughput screening
of SNPs [5, 6], with the potential to become the method of
choice for direct genetic analysis. In addition, MALDI-MS
has been employed effectively for peptide nucleic acid
(PNA) characterization [7]. PNAs are nucleic acid mimics
consisting of nucleobases connected via a peptide-like
backbone. Because PNAs are resistant to nuclease and pro-
tease digestion, they are being investigated as gene thera-
peutic drugs, and are employed in a variety of important
hybridization technologies [8].
Identification of microbes in their natural environment
has great importance in the diagnosis of disease, and char-
acterization of biological and environmental hazards. The
use of MALDI-MS for identification of bacteria based on
the measurement of cellular proteins has shown great
promise [9], but the limitations of microorganism identifi-
cation using mass spectrometry are recognized [9, 10].
Future development of improved methodologies and data
analysis techniques should advance MALDI-MS as a vital
technology in this area of research.
Characterization of synthetic polymer materials has tra-
ditionally been accomplished using chromatographic tech-
niques such as gel permeation chromatography (GPC).
Molecular weights of polymers obtained using GPC are
calculated based on relative comparison with molecular

weight standards. By contrast to GPC, MALDI-MS has
been shown to provide direct measurement of absolute
molecular masses of oligomers independent of polymer
structure [11]. It is also a more rapid technique than tradi-
tional chromatographic approaches, requiring only a few
minutes per analysis. The recognized disadvantage of
MALDI-MS for polymer characterization is that it fails to
provide correct molecular mass values for polydisperse
polymers. This limitation has been addressed through a
combination of MALDI-MS with GPC [12].
Future advances in the areas of research described
herein will certainly solidify the role of MALDI-MS as an
essential tool for the analyst.
19
References
1. Karas M., Hillenkamp F (1988) Anal Chem 60:2301
2. Mann M, Hendrickson RC, Pandey A (2001) Annu Rev
Biochem 70:437
3. Yates JR III (1998) J Mass Spectrom 33:1
4. Fenselau C (1997) Anal Chem 69:661A
5. Griffin TJ, Smith LM (2000) Trends Biotechnol 18:77
6. Li J, Butler JM, Tan Y, Lin H, Royer S, Ohler L, Shaler TA,
Hunter JM, Poliart DJ, Monforte JA, Becker CH (1999) Elec-
trophoresis 20:1258
7. Butler JM, Jiang-Baucom P, Huang M, Belgrader P, Girard J
(1996) Anal Chem 68:3283
8. Griffin TJ, Tang W, Smith LM (1997) Nat Biotechnol 15:1368
9. Dalluge JJ (2000) Fresenius J Anal Chem 366:701
10. Pineda FJ, Lin JS, Fenselau C, Demirev A (2000) Anal Chem
72:3739
11. Wu KJ, Odom RW (1998) Anal Chem 70:456A
12. Hanton SD, Liu XM (2000) Anal Chem 72:4550
Anal Bioanal Chem (2002) 372 : 2021
DOI 10.1007/s00216-001-1147-0

TRENDS

Frank Vanhaecke

ICPMS
Alternative means for the elimination of interferences

Published online: 8 December 2001

Springer-Verlag 2001

The success of ICPmass spectrometry is the result of its
low limits of detection (LOD), wide linear dynamic range,
multi-element capabilities, surveyable spectra, high sam-
ple throughput and ease with which it can be used with al-
ternative sample-introduction systems and chromatographic
separation techniques. Very soon after the introduction of
ICPMS, it became clear that non-spectral (matrix-in-
duced signal suppression or enhancement) and spectral
(overlap of the signals from ions with a mass difference
<0.5 u) interferences were its most prominent disadvan-
tages. Whereas non-spectral interferences could be fairly
easily coped with e.g., by means of sample dilution, the
use of (an) internal standard(s) or application of standard
additions or isotope dilution instead of external calibration
spectral interferences proved more troublesome. Espe-
cially for complex matrices and in the low mass-range
(80 u), obtaining reliable results or sufficiently low LOD
is, therefore, not straightforward. Of course, efforts were
made to cope with this problem and, from the beginning,
blank subtraction, mathematical correction or, if really nec-
essary, matrix/trace separation were used to warrant accu-
racy. Later, also improved aerosol desolvation devices and
cool plasma conditions were used.
The only universal means of overcoming spectral over-
lap was, however, using a double-focusing sector field
mass spectrometer in ICPMS instrumentation, instead of
the more traditional quadrupole filter. Sector field mass
spectrometers offer a much higher mass resolution than
do quadrupole filters, and so ions differing in mass by a
fraction of a mass unit only can still be separated (Fig. 1)
and accurate quantification becomes straightforward [1].
As a result, sector field ICPMS (ICPSFMS) is better
suited for analyzing demanding matrices. Accurate deter-
mination of the first-row transition metals a mass range
especially prone to spectral overlap was, e.g., demon-
strated for certified reference materials of different origin
[2]. Up to 60 elements could be determined in whole blood
after microwave-assisted acid digestion [3]. Increasing the
number of elements for which reliable data can be ob-
tained obviously leads to enhanced diagnostic capabilities
and a more profound insight into the consequences of en-
vironmental or occupational exposure. For soil digests also,
a multitude of spectral interferences complicate analysis
with quadrupole-based ICPMS (ICPQMS), whereas ap-
plication of higher mass resolution enables elegant avoid-
ance of these problems [4]. The only limitations are that
for a small minority of analyses the maximum resolution
offered by this instrumentation (up to 12,000) is still in-
sufficient, and there is an inverse relationship between
mass resolution and ion transmission efficiency (signal in-
tensity).
The success of ICPSFMS has, apparently, also trig-
gered important developments in ICPQMS instrumenta-
tion, the dynamic reaction cell (DRC) [5] and, to a lesser
extent, the hexapole and octopole collision cells [6] being
the most appealing. These devices enable suppression of
the signal intensity of unwanted ions by orders of magni-
tude as a result of the (selective!) reaction of the latter

Fig. 1 Schematic representation of the effect on the mass spec-

F. Vanhaecke ()
Laboratory of Analytical Chemistry, Ghent University,
Institute for Nuclear Sciences,
Proeftuinstraat 86, 9000 Ghent, Belgium
e-mail: frank.vanhaecke@rug.ac.be
trum of increasing the mass resolution. The analyte signal (light-
coloured) becomes resolved from the signals of (molecular) ions
that overlap with the analyte signal at low mass resolution. The de-
crease in signal intensity observed on increasing the mass resolu-
tion has been neglected

with an appropriate gas. New unwanted ions can, unfortu-
nately, also be created in this way. To avoid this, the gases
used with a hexapole or octopole collision cell are limited
to H2 and noble gases (mainly He for thermalization), so
the number of spectral interferences that can be tackled is
relatively limited. With the quadrupole set-up of the DRC,
on the other hand, the cell can be operated simultaneously as
a band-pass mass filter, removing these newly formed ion
species from the beam. As a result, in a DRC a large(r) va-
riety of gases (e.g., also NH3, CH4 or O2) can be used and
the application range becomes virtually unlimited. Chem-
ical resolution in a DRC can thus be used not only to over-
come more traditional spectral overlap (e.g., 40Ca+/40Ar+,
56Fe+/40Ar16O+ or 75As+/40Ar35Cl+) which can also be
avoided by using a collision cell [6] but even to resolve
the signals of isobaric isotopes, which would require a mass
resolution substantially exceeding the maximum mass res-
olution offered by modern ICPSFMS instrumentation. In
this context, the possibility for Sr isotopic analysis of ge-
ological samples without previous Rb/Sr separation has
been demonstrated [7] (isobaric overlap of the signals from
87Rb+ and 87Sr+ was avoided by selectively converting the
Sr+ ions into the corresponding SrF+ species by use of
CH3F).
To exploit the advantages of DRCICPMS fully, how-
ever, a collection of gases and considerable operator expe-
rience are required. It should, on the other hand, be stressed
that the use of NH3 as a single reaction gas enables elim-
ination of the most abundant interferences and enables
ng L1 LOD for traditionally difficult elements such as K,
Ca, and Fe. Advantages of mass resolution over chemical
resolution are: 1. the straightforwardness of the former
approach; 2. the direct, visual detection of the presence of
(molecular) ions that would interfere with the determina-
tion at low resolution and the possibility to estimate their
importance (Fig. 1); and 3. the simultaneous solution for
spectral interferences by different types of interfering ions.
ICPSFMS instrumentation is, however, significantly
more expensive than ICPQMS instrumentation equipped
with a DRC; this might explain why the number of DRC-
units sold so far is starting to approach the number of
ICPSFMS instruments used world-wide.
Despite the undeniable advantage, only a minority
(10%) of the ICPMS instruments used worldwide offer
21
high mass or chemical resolution. It can, of course, be
assumed that for ICPSFMS the purchase price (approxi-
mately twice that of an ICPQMS instrument) is often a
major obstacle. Larger numbers of sales will probably be
achieved if the analytical community is convinced that
ICPSFMS is sufficiently robust and that instrument op-
eration is not extremely complicated. The success of
DRCICPMS which was only introduced at the end of
the nineties on the other hand, will mainly depend on
the availability of data on the selective ionmolecule re-
actions that have to be used to tackle specific problems
and the maintenance of a purchase price that is suffi-
ciently different from (i.e. lower than) that of a sector
field unit.
The first ICPSFMS instrument equipped with a hexa-
pole collision cell was introduced very recently. This im-
plies that the analyst has both high mass-resolution and
chemical resolution at his or her disposal. A combination
of a sector field mass spectrometer and a DRC within the
same instrument would, on the other hand, result in the
most effective and versatile ICPmass spectrometer ever
produced. For this to become reality, however, some tech-
nical problems must be overcome, and difficulties related
to patents owned by different (competing) manufacturers
can be expected. Even if the latter issue could be resolved,
a marketing study assessing the interest from the analyti-
cal community and the money available for this instrumen-
tation would probably be very useful.
References
1. Jakubowski N, Moens L, Vanhaecke F (1998) Spectrochim Acta
B 53:17391763
2. Townsend AT (2000) J Anal At Spectrom 15:307317
3. Rodushkin I, Axelsson MD (2000) Sci Total Environ 250:83
100
4. Latkoczy C, Prohaska T, Stingeder G, Wenzel WW (2000) Fre-
senius J Anal Chem 368:256262
5. Tanner SD, Baranov VI (1999) At Spectrosc 20:4552
6. Feldmann I, Jakubowski N, Thomas C, Stuewer D (1999) Frese-
nius J Anal Chem 365:422428
7. Moens LJ, Vanhaecke FF, Bandura DR, Baranov VI, Tanner SD
(2001) J Anal At Spectrom, in press
Anal Bioanal Chem (2002) 372 : 22
DOI 10.1007/s00216-001-1152-3
TRENDS
Karen W. Phinney
Chiral separations
Published online: 8 December 2001
Springer-Verlag 2001
The pharmaceutical industry has largely been the driving
force behind the development of chiral separation tech-
niques. This focus helps to explain the growth in the area
of preparative scale enantioselective separations in recent
years. A few years ago, preparative scale chromatographic
separations would not have been considered a cost effective
approach to enantiomer production. However, the transition
from traditional batch processes to simulated moving bed
(SMB) technology has dramatically changed the acceptance
of chromatography for large-scale production. Both batch
and SMB methods utilize chiral stationary phases (CSPs) to
achieve the desired separation. Although most systems in-
corporate liquid eluents, the use of gases and supercritical
fluids as eluents has also been reported [1, 2].
Evolution continues to occur in the area of CSPs, pri-
marily for liquid chromatography (LC). The desirable prop-
erties of CSPs can differ significantly depending upon their
intended use. On one hand, stationary phases possessing
enantioselectivity towards a broad spectrum of compounds
are advantageous for analytical scale separations. This prop-
erty reduces the number of different columns that must be
evaluated to achieve the desired separation. Extremely high
stereoselectivity merely increases analysis time to unde-
sirable levels. On the other hand, a stationary phase with a
very high level of enantioselectivity towards a single tar-
get is often preferable for preparative scale separations.
Combinatorial chemistry is increasingly being used to iden-
tify chiral selectors with high levels of enantioselectivity
for a target compound. Suitable chiral selectors are identi-
fied by screening mixture libraries [3] or individual library
compounds [4] against the target.
Capillary electrophoresis continues to grow in popular-
ity for enantioselective separations, although reproducibil-
ity and sensitivity remain issues of concern. Neutral and
K.W. Phinney ()
Analytical Chemistry Division,
Chemical Science and Technology Laboratory,
National Institute of Standards and Technology,
Gaithersburg, MD 20899, USA
e-mail: karen.phinney@nist.gov
charged cyclodextrins are the most common chiral selec-
tors in CE, but new additives are also being identified [5].
Capillary electrochromatography (CEC) has not played a
major role thus far; however, the lure of high peak effi-
ciency associated with CE coupled with chromatographic
separation mechanisms continues to fuel research in this
area. Continuous free flow electrophoresis (CFFE) is a re-
cent addition to the field of chiral separations. This tech-
nique seeks to exploit the resolving power of electrophore-
sis for a continuous feed process [6]. As in CE, a chiral
additive is incorporated into the background electrolyte.
Applications of supercritical fluid chromatography
(SFC) to enantioselective separations have also expanded
because SFC offers an attractive alternative to liquid chro-
matographic techniques. Replacement of traditional liquid
eluents with supercritical fluids commonly results in in-
creased efficiency, simplified method development, and
reduced waste generation [7, 8]. In general, the CSPs used
for enantioselective separations in SFC are the same as
those used in LC.
Additional types of separation procedures, such as
membrane-based approaches, are gaining attention for enan-
tioseparations [9]. Clearly, no single technique can address
all the current challenges, and the need for high through-
put as well as low cost will likely continue to steer future
developments in chiral separations.
References
1. Francotte ER (2001) J Chromatogr A 906:379397
2. Juza M, Mazzotti M, Morbidelli M (2000) Trends Biotechnol
18:108118
3. Bluhm LH, Wang Y, Li T (2000) Anal Chem 72:52015205
4. Welch CJ, Bhat G, Protopopova MN (1999) J Comb Chem 1:
364367
5. Chankvetadze B, Blaschke G (2001) J Chromatogr A 906:309
363
6. Stalcup AM, Sutton RMC, Painuly P, Rodrigo JV, Gratz SR,
Yanes EG (2000) Analyst 125:17191724
7. Phinney KW (2000) Anal Chem 72:204A211A
8. Terfloth GJ (2001) J Chromatogr A 906:301307
9. Maier NM, Franco P, Lindner W (2001) J Chromatogr A 906:
333
Anal Bioanal Chem (2002) 372 : 2324
DOI 10.1007/s00216-001-1146-1
TRENDS
Ulrich Panne
Spectrochemical analysis with lasers
Published online: 8 December 2001
Springer-Verlag 2001
Introduction
Since the development of the first laser in 1960, lasers
have generated distinct fascination, because of their in-
tense, coherent, monochromatic radiation, which promised
new and exciting possibilities in spectroscopy. Not surpris-
ingly, numerous laser-based techniques for the detection
of molecules and atoms in solids, liquids, gases, or plas-
mas have been described. Despite the reported impressive
gains in selectivity and detection limits, the considerable
cost and sophistication of most laser sources often limit
their use in analytical instruments. The recent advent of
turn-key solid-state lasers and diode lasers improved the
situation considerably, but the universal laser source for
spectrochemical analysis still lies in the future. For dedi-
cated applications, however, lasers offer unique solutions
in analytical chemistry. Thus although laser spectrochem-
ical analysis is not a substitute for conventional techniques,
it is becoming a valuable and long promised addition
to the analytical chemists toolbox [1, 2, 3].
Elemental spectrochemical analysis
Laser ablation (LA) either for introduction of sample into
elemental analyzers such as the ICPMS or in combina-
tion with atomic emission spectroscopy (denoted laser-in-
duced breakdown or plasma spectroscopy, LIBS or LIPS)
is currently the dominating application of lasers in ele-
mental analysis [4, 5]. Ablation can be conducted in bulk
analysis mode (spot size >200 m) or microanalysis mode
(spot size 120 m). The latter rivals other microanalyti-
cal techniques in speed and ease of use. Whereas current
applications of LIPS have been developed extensively for
process analysis, LAICPMS is becoming an established
U. Panne ()
Institute of Hydrochemistry, Technical University of Munich,
Marchioninistr. 17, 81377 Munich, Germany
e-mail: ulrich.panne@ch.tum.de
routine method for solid-sample analysis. Other approaches,
e.g. diode laser absorption spectroscopy or laser-enhanced
atomic fluorescence (LEAF), are still hampered by avail-
able laser technology, but are promising areas of research
and could be turned into real multi-element techniques
with the advent of diode lasers in the UVvisible range
[6].
Molecular spectrochemical analysis
Absorption spectroscopy
Whereas diode lasers in classical absorption measure-
ments are well established for the visible and near infrared
(VIS/NIR) range, the new generation of quantum cascade
lasers for the mid-IR will provide exciting possibilities for
improved molecular identification. Photothermal spectros-
copy, e.g. thermal-lens spectroscopy (TLS) or photoacoustic
spectroscopy (PAS), with laser sources are an ultrasensitive
alternative for measurement of the absorbance of small
volumes. With the advent of Nd:YAG laser-pumped opti-
cal parametric oscillators (OPO) wide-range (220 nm to
2.4 m) wavelength-tunable systems are now entering the
market and promise new life for such applications, be-
cause of their ease of use. The increasing demand for ab-
sorption spectra from small volumes arises not only from
analytical problems in bioanalysis and microcolumn sepa-
ration techniques, but also in the process analysis of opaque
materials and liquids [7].
Fluorescence spectroscopy
Laser-induced fluorescence (LIF) for detection of native
fluorophores, such as the extensively studied polycyclic
aromatic hydrocarbons (PAH) or fluorescent proteins, and
labeled analytes, is a real old timer. Analytical demands
in the life sciences, in particular, have led to a renaissance
and commercial sophistication of LIF apparatus and tech-
niques, as has a wealth of new fluorescence labels. LIF is,
24
therefore, becoming a routine tool in microcolumn sepa-
ration and analysis of single (bio)molecules [8]. A further
impact is observed in fluorescence microscopy, where multi-
photon LIF and fluorescence correlation spectroscopy (FCS)
with femtosecond-lasers have enabled the creation of a new
user-friendly generation of these lasers [9]. Advances in
near-field microscopy with LIF (and other laser-based
techniques such as Raman spectroscopy) will provide a
nano-scale spectrochemical analysis beyond the confocal
limit.
Raman spectroscopy
Although Raman spectroscopy has been synonymous with
laser Raman spectroscopy since the mid 1960s, it was not
until holographic notch filters and sensitive CCD detec-
tors simplified and improved the experimental set-up that
Raman spectroscopy became a routine analytical tool [10].
Raman microscopy, process analysis with NIR diode lasers,
and UV resonance Raman at wavelengths below 255 nm
are not only currently receiving serious academic atten-
tion, but can already be seen in several real-world appli-
cations.
References
1. Imasaka T (1999) Talanta 48:305
2. Gooijer C, Mank AJG (1999) Anal Chim Acta 400:281
3. Panne U, Niessner R (1997) In: Gnzler H, Bahadir AM, Bors-
dorf R, Danzer K, Fresenius W, Galensa R, Huber W, Lin-
scheid, Lderwald I, Schwedt G, Tlg G, Wisser H (eds) Ana-
lytiker Taschenbuch, Vol. 16, Springer, Berlin, p 155
4. Demtrder W (1998) Laser spectroscopy. Springer, Berlin
5. Gnther D, Jackson SE, Longerich HP (1999) Spectrochim
Acta B 54:381
6. Rusak DA, Castle BC, Smith BW, Winefordner JD (1998)
Trends Anal Chem 17:453
7. Winefordner JD, Gornushkin IB, Pappas D, Matveev OI, Smith
BW (2000) J Anal At Spectrom 15:1161
8. Bialkowski SE (1996) Photothermal spectroscopy methods for
chemical analysis. John Wiley and Sons, New York
9. Zander C (2000) Fresenius J Anal Chem 366:745
10. Rigler R, Elson E (2001) (eds) Fluorescence correlation spec-
troscopy. Springer, Berlin
11. McCreery RL (2000) Raman spectroscopy for chemical analy-
sis. John Wiley and Sons, New York
Anal Bioanal Chem (2002) 372 : 2526
DOI 10.1007/s00216-001-1158-x

TRENDS

Klaus Albert Manfred Krucker Tobias Glaser

Alexandre Schefer Annette Lienau Daniel Zeeb

Hyphenated techniques

Published online: 8 December 2001

Springer-Verlag 2001

To fulfil the need for efficient and rapid on-line identifi-

cation of peaks from a chromatographic separation the hy- +2+&
+
+2 2 +2+& 2 +
phenation of chromatographic separation techniques with +2 ++ + + +2
&+2+
spectroscopic identification techniques is of increasing + 2+ 2+ +
2
importance.
The separation technique usually applied is high-per- 16 

formance liquid chromatography; the most useful spectro-

scopic techniques for elucidation of the structures of un-
known compounds are mass spectrometry and NMR spec- 61 
troscopy. Hyphenation of these two spectroscopic tech-
niques with HPLC will, therefore, furnish extremely valu-
able information.
Within the field of mass spectrometry new powerful
ionisation techniques have been developed. Electrospray
ionisation (ESI) and atmospheric-pressure chemical ioni-
sation (APCI) can be easily coupled with HPLC and
proved to yield extremely valuable information [1].
Within the field of NMR spectroscopy the application
of continuous-flow probes with a detection volume of ei-
ther 60 or 120 L is now routine, with over 200 installa-
tions world-wide. The on-line coupling of HPLC and
NMR can be performed in either the stopped-flow or con-
tinuous-flow mode. Current sensitivity levels are in the
500 ng range for one-dimensional 1H NMR spectra and in
the microgram range for two-dimensional spectra.
Combination of effective on-line coupling of HPLC
with ESI MS, APCI MS, and NMR has been shown to be
extremely helpful for the solution of a myriad of separa-
tion and identification problems in pharmacy, biotechnol-
ogy, and nutrition [2, 3, 4].
Miniaturisation of on-line coupling techniques to meet
the needs of effective research in the new millennium is a
major challenge. Whereas the use of capillaries is becom-
ing more popular in HPLCMS, capillary HPLCNMR is
still under development [5, 6, 7, 8, 9].
K. Albert () M. Krucker T. Glaser A. Schefer A. Lienau
D. Zeeb
Universitt Tbingen, Institut fr Organische Chemie,
Auf der Morgenstelle 18, 72076 Tbingen, Germany
e-mail: klaus.albert@uni-tuebingen.de
SSP       
Fig. 1 1H NMR spectrum (600 MHz) of 200 mmol L1 sucrose in
D2O recorded in a 750-nL capillary detection cell (one transient)
Figure 1 shows a sensitivity test of a 2.0-mm capillary
HPLCNMR probe with a detection volume of 750 nL.
This newly designed probe with a double-saddle Helm-
holtz coil results in significant improvements in signal
line shape and easy magnetic-field homogenisation. The
signal-to-noise ratio of 50:1 obtained for the anomeric
proton of 0.2 mol L1 solution of sucrose in D2O is suffi-
cient for elucidation of the structures of naturally occur-
ring substances.
Besides HPLC, electrically driven separation tech-
niques such as CE and CEC are increasing in importance.
Both techniques have already been successfully coupled
to spectroscopic detection techniques and show potential
for many real-life applications [10]. Overall, the on-line
coupling of spectroscopy and chromatography is a prereq-
uisite for the successful treatment of the problems en-
countered in genomics, proteomics, and metabolomics.
References
1. Wilson ID, Morgan ED, Lafont R, Shockcor JP, Lindon JC,
Nicholson JK, Wright B (1999) Chromatographia 49:374378
26
2. Sidelmann UG, Braumann U, Hofmann M, Spraul M, Lindon
JC, Nicholson JK, Hansen SH (1997) Anal Chem 69:607612
3. Lindon JC, Nicholson JK, Sidelmann UG, Wilson ID (1997)
Drug Metab Rev 29:705746
4. Sandvoss M, Pham LH, Levsen K, Preiss A, Mgge C, Wnsch
G (2000) Eur J Org Chem 12531262
5. Kautz RA, Lacey ME, Wolters AM, Foret F, Webb AG, Karger
BL, Sweedler JV (2001) J Am Chem Soc 123:31593160
6. Olson L, Peck TK, Webb AG, Magin RL, Sweedler JV (1995)
Science 270:19671970
7. Behnke B, Schlotterbeck G, Tallarek U, Strohschein S, Tseng
LH, Keller T, Albert K, Bayer E (1996) Anal Chem 68:1110
1115
8. Schlotterbeck G, Tseng LH, Hndel H, Braumann U, Albert K
(1997) Anal Chem 69:14211425
9. Webb AG (1997) Prog Nucl Magn Reson Spectrosc 31:142
10. Pusecker K, Schewitz J, Gfrer P, Tseng LH, Albert K, Bayer
E (1998) Anal Chem 70:32803285
Anal Bioanal Chem (2002) 372 : 2728
DOI 10.1007/s00216-001-1160-3
TRENDS
Uwe Karst
Detection techniques for liquid chromatography
Published online: 11 December 2001
Springer-Verlag 2001
Few analytical techniques have reached such a high degree
of maturity and applicability for routine analysis as liquid
chromatography (LC). From this point of view, it may be
surprising that there has been such a dynamic develop-
ment with respect to modern detection techniques for liq-
uid chromatography. On the other hand, due to the tremen-
dous impact of LC separations on such diverse topics as
the purification of pharmaceuticals, clinical chemistry or
environmental speciation analysis, the need for sensitive
and selective detectors becomes increasingly apparent.
In the 1990s, several detectors for LC were invented
or greatly improved; these included the evaporative light-
scattering detector as a universal device [1]. The com-
bination of LC and nuclear magnetic resonance (NMR)
spectroscopy allows one to carry out the structural eluci-
dation of analytes [2]. Laser-induced fluorescence (LIF)
detection of native or labeled analytes has become an im-
portant tool for bioanalysis at very low concentration lev-
els [3]. Despite these and other significant achievements,
the progress in liquid chromatography/mass spectrometry
(LC/MS) is outstanding [4].
Undoubtedly, the development of electrospray ioniza-
tion (ESI) [5] and atmospheric pressure chemical ioniza-
tion (APCI) [6] has triggered a revolution in the field of
detection in liquid chromatography, because LC/MS is
now readily available as a routine tool for pharmaceutical,
biomedical and environmental analysis as well as for com-
binatorial chemistry. In just over a decade, LC/MS has
completely lost its former reputation of being an expen-
sive method of limited sensitivity which could only be ap-
plied by specialists. As protonation and deprotonation are
the most abundant ionization mechanisms when using ESI
and APCI in the positive and negative mode, respectively,
polar or even charged analytes can be readily analyzed.
For this reason, selected groups of biomolecules are among
U. Karst ()
University of Twente, Faculty of Chemical Technology,
Chair of Chemical Analysis,
P.O. Box 217, 7500 AE Enschede, The Netherlands
e-mail: u.karst@ct.utwente.nl
the analytes that can be determined with the best sensitiv-
ity in LC/MS. On the other hand, analytes of low polarity
are only susceptible to these methods with moderate per-
formance. One of the major challenges for the future is
therefore the development of LC/MS methods and instru-
mentation that are capable of analyzing substances with
limited polarity at extremely low concentrations.
Very recently, some approaches for expanding the range
of applications to less polar compounds have been pub-
lished. Bruins et al. [7] have introduced atmospheric pres-
sure photoionization (APPI) to liquid chromatography/mass
spectrometry. The technique allows one to carry out sen-
sitive determination of non-polar compounds, e.g., poly-
cyclic aromatic hydrocarbons that can hardly be detected
under ESI or APCI conditions. Blair et al. [8] have inves-
tigated pentafluorobenzyl derivatives of biomolecules and
drugs and have observed dissociative electron capture
ionization with excellent sensitivity. Therefore the trans-
fer of derivatization techniques based on multihalogenated
derivatization reagents, which are frequently applied in gas
chromatography with electron-capture mass spectrometry
[9] appears to be possible. For nitroaromatic compounds,
similar electron capture effects are observed and have been
used for the development of derivatizing agents [10].
Another method of ionizing substances of limited po-
larity is their electrochemical oxidation at the ESI inter-
face [11] or an additional electrochemical cell [12]. Dedi-
cated derivatizing agents have been developed to be used
with electrochemistry/ESI-MS [11]. Brajter-Toth and co-
workers [13] performed on-line LC/particle beam-MS stud-
ies of the reaction products formed in an electrochemical
cell. Very recently, a LC/electrochemistry/MS system based
on the quantitative post-column oxidation of non-polar
analytes to charged products was realized [14].
In the coming years, it can be expected that there will
be an intense development of methods for the LC/MS
analysis of less polar substances. All of the approaches
presented above have strong advantages for selected
groups of analytes, but only few can be used universally
as they may either require particular properties of the an-
alytes (e.g., electrochemical activity) or a derivatization
28

step. The future will show if these or completely different 5. Whitehouse CM, Dreyer RN, Yamashita M, Fenn JB (1985)
techniques will become important complements to ESI and Anal Chem 57:675
6. Covey TR, Lee ED, Henion JD (1986) Anal Chem 58:2453
APCI with special emphasis on the analysis of non-polar 7. Robb DB, Covey TR, Bruins AP (2000) Anal Chem 72:3653
compounds. 8. Singh G, Gutierrez A, Xu KY, Blair IA (2000) Anal Chem 72:
3007
9. Giese RW (2000) J Chromatogr A 892:329
10. Hayen H, Jachmann N, Vogel M, Karst U (2001) German
References Patent Application DE 101 23 460.0
11. Van Berkel GJ, Quirke JME, Tigani RA, Dilley AS, Covey TR
1. Hsu BH, Orton E, Tang S-Y, Carlton RA (1999) J Chromatogr
(1998) Anal Chem 70:1544
B 725:103
12. Xu XM, Lu WZ, Cole RB (1996) Anal Chem 68:4244
2. Albert K (1999) J Chromatogr A 856:199
13. Regino MCS, Brajter-Toth A (1997) Anal Chem 69:5067
3. Van de Nesse RJ, Velthorst NH, Brinkman UAT, Gooijer C
14. Diehl G, Liesener A, Karst U (2001) Analyst 126:288
(1995) J Chromatogr A 704:1
4. Niessen WMA (1999) J Chromatogr A 856:179
Anal Bioanal Chem (2002) 372 : 2930
DOI 10.1007/s00216-001-1153-2
TRENDS
Bernd Speiser
Molecular electrochemistry
Published online: 8 December 2001
Springer-Verlag 2001
Electroanalytical chemistry traditionally involves:
1. detailed investigation of processes, including electron
transfer, at electrodes (molecular electrochemistry), and
2. determination of concentrations of redox-active species
by electrochemical means.
Common to both sub-fields is the use of the same arsenal
of electroanalytical techniques. This overview will mainly
deal with sub-field 1. Fairly broad coverage of the entire
field can be found bi-annually elsewhere [1]. Selected
trends which have emerged recently and are expected to
develop into important lines of research and applications
will be highlighted here.
First, the use of ultramicroelectrodes (UME, i.e. elec-
trodes with a characteristic dimension <25 m [2], or those
characterized by dimensions comparable with or smaller
than the diffusion layer thickness) has influenced electro-
analytical experiments in a variety of ways. Their minute
size enables investigation of electrochemical processes in
small lateral dimensions. Thus, local redox properties can
be probed, and an important development in this respect is
the scanning electrochemical microscope (SECM), in
which the UME is scanned along a surface. Although the
resolution of the SECM is, by necessity, limited to the di-
mensions of the diffusion layer, and is thus less than that
of other scanning probe techniques, its main advantage is
related to the flow of a Faradaic current which results
from a defined redox process. Consequently, the redox ac-
tivity of the surface can be separated from its topography
and is imaged, e.g. over biological samples [3].
Another advantage of the UME was recognized early
because of the small iR drop and double-layer charging
effects, high scan rate voltammograms (v>1 MV s1) are
available without such artifacts. One of the most striking
applications of the thin diffusion layers and high temporal
resolution attainable with such short time-scale experi-
B. Speiser ()
Institut fr Organische Chemie,
Auf der Morgenstelle 18, 72076 Tbingen, Germany
e-mail: bernd.speiser@uni-tuebingen.de
ments was recently presented [4]. A dendrimeric polymer
with redox-active Ru(tpy)2 groups was adsorbed on an
electrode, and the diffusion process through the molecules
(diameter5 nm) was dissected in the manner of an elec-
trochemical microtome.
Much UME work has recently been devoted to electro-
analytical experiments in media of low conductivity [5].
This opens the way to electrochemistry in non-polar me-
dia or even in the gas phase [2].
The use of electroanalytical techniques in molecular
electrochemistry is often accompanied by simulation of
experiments. Thus, a physical model of the real processes
(transport, electron transfer, chemical reactions, adsorp-
tion) is formulated in the form of mathematical (often dif-
ferential) equations and solved [6]. The result predicts the
experimental outcome on the basis of the physical model,
and, by comparison with the real experiment, enables qual-
itative testing of the validity of the model and quantitative
evaluation of mechanistic information.
General application of simulation in the context of cyclic
voltammetry was boosted a decade ago by DigiSim [7],
the first commercial software based on graphical user in-
teraction, which enables the formulation of the physical
model in the convenient form of chemical equations. The
sophistication of numerical solution methods continues to
increase [8, 9] and the development of computing facili-
ties and, in particular, the internet has initiated new possi-
bilities. Artificial intelligence approaches using a knowl-
edge-based system to elucidate reaction mechanisms have
been described [10]. Compton et al. provide a web-based
service for the evaluation of kinetic data [11]. Virtual in-
strumentation for electroanalytical experimentation en-
hances data generation and analysis [12]. Such applica-
tions certainly qualify as facets of complex software which
has been introduced as a problem solving environment
in electrochemistry [13]. It is expected that software of
this type will increasingly be developed in the future. In
the opinion of this author, this development could benefit
even more from open and co-operative development of
modular software integrating the approaches of different
authors.
30
References
1. Anderson JL, Coury LA Jr, Leddy J (2000) Anal Chem 72:
44974520
2. Heinze J (1993) Angew Chem 105:13271349; Angew Chem
Int Ed Engl 32:12681288
3. Hengstenberg A, Blchl A, Dietzel ID, Schuhmann W (2001)
Angew Chem 113:942946; Angew Chem Int Ed Engl 40:
905908
4. Amatore C, Bouret Y, Maisonhaute E, Goldsmith JI, Abrua
HD (2001) Chem Eur J 7:22062226
5. Myland JC, Oldham KB (2000) Anal Chem 72:39723980
6. Speiser B in Bard AJ, Rubinstein I (1996) (eds) Electroanalyti-
cal chemistry, vol 19. Marcel Dekker, New York, pp 1108
7. Rudolph M, Reddy DP, Feldberg SW (1994) Anal Chem 66:
589A600A
8. Bieniasz LK, Britz D (2001) J Electroanal Chem 503:141152
9. Harriman K, Gavaghan DJ, Houston P, Sli E (2000) Elec-
trochem Commun 2:150156
10. Pays MJ, Bos M, van der Linden WE (1993) Anal Chim Acta
283:811829
11. Alden JA, Compton RG (1998) Electroanalysis 10:207209
12. Economou AS, Volikakis GJ, Efstathiou CE (1999) J Autom
Method Manag Chem 21:3338
13. Bieniasz LK (1997) Comput Chem 21:112
Anal Bioanal Chem (2002) 372 : 3132
DOI 10.1007/s00216-001-1148-z
TRENDS
Detlef Gnther
Laser-ablation inductively-coupled plasma mass spectrometry
Published online: 12 December 2001
Springer-Verlag 2001
The first application of laser-ablation inductively-coupled
plasma mass spectrometry (LAICPMS) by Gray (1985)
marked the creation of one of the most versatile direct
solid-sampling techniques for both qualitative and quanti-
tative analysis in the last 16 years [1]. This method prof-
ited from early work performed in laser-induced break-
down spectroscopy (LIBS) and studies in laser-ablation
inductively-coupled plasma optical-emission spectrome-
try (LAICPOES) [2, 3].
From the beginning several keywords, e.g. rapid,
cheap, easy to use, bulk to micro-scale, trace ele- throughs in laser ablation which enabled simultaneous de-
ment, quasi-non-destructive, no sample preparation, tection of major and trace elements with a single mea-
and quantitative analysis, were applied to LAICPMS.
The general principle and the advantages of this technique
have not changed over the last two decades. A laser beam
is directed on to the surface of a sample in an air-tight ab-
lation cell. Interaction of the laser beam with the sample
leads to particles and/or vapour, which are transported by
a carrier gas flow into the ICP. The ions produced in the ICP
are extracted through the vacuum interface and analysed
by use of a variety of mass analysers and different detec-
tor systems.
A few years after Grays initial work the most severe
drawback of this technique, the non-sample-related varia-
tion of the relative signal response during analysis (so-called
elemental fractionation) was observed and the lack of ref-
erence materials for calibration was seen as a limitation of
this technique. This led to loss of some of the keywords
mentioned above, e.g. quantitative analysis. A milestone
report by Jackson et al. in 1992 [4] contained quantitative
results for a variety of elements in minerals, obtained by
use of an in-house-built laser-ablation system (1064 nm
Nd:YAG laser), with non-matrix matched calibration.
In the early nineties the first quadrupled 266 nm
Nd:YAG laser was used to show that lasersample inter-
D. Gnther ()
Eidgenssische Technische Hochschule Zrich,
Laboratorium fr Anorganische Chemie,
ETH Hnggerberg, HCI G113, 8093 Zrich, Switzerland
e-mail: guenther@inorg.chem.ethz.ch
action is less matrix-dependent if UV lasers are used (im-
proved absorption). This initiated studies using all avail-
able UV laser wavelengths (quadrupled 266 nm, quintu-
pled 213 nm Nd:YAG and excimer wavelengths such as
308 nm, 248 nm, 222 nm, 193 nm, and 157 nm) for
LAICPMS. This increased flexibility led to new appli-
cations. In addition to progress in laser technology, instru-
mental developments in ICPMS, especially higher sensi-
tivity, increased linear dynamic range, and the greatly im-
proved the speed of data acquisition, are further break-
surement. The amount of sample preparation is, however,
related to the spatial resolution required. For many appli-
cations determination of minor, trace, and ultra-trace con-
centrations at limits of detection in the low ng g1 range
(depending on the spatial resolution) by use of matrix-
matched and non-matrix-matched reference materials, has
been reported [5, 6]. Unfortunately, until now quantitative
analysis has required knowledge of the concentration of an
internal standard determined by another technique before
the analysis. Beside the progress in LAICPMS, ele-
mental fractionation [7] became the dominant keyword
with regard to calibration and accuracy. Despite many
successful approaches and even quantification without
standards (possible when all matrix elements can be mea-
sured simultaneously during analysis [8]), discussions on
LAICPMS remain focused on elemental fractionation.
The source of this phenomenon, which could occur during
the ablation process, during transport, or during atomisa-
tion and excitation, is still unknown. The number of vari-
ables determining the performance of LAICPMS makes
it very difficult to evaluate the individual effect of each.
Recent work by Russo et al. [9] has shown that changing
the laser conditions directly affects the analytical charac-
teristics. The transport efficiency is related to the carrier
gas [10] and to the laser wavelength [11] but is less influ-
enced by the transport system [12].
The ion source used in LAICPMS was originally de-
signed for solution nebulisation and not for dry-aerosol
introduction. Modifications of a set-up used for solution
32
Fig. 1 Limits of detection
of a commercially available
ICPMS for a selection of ele-
ments. Improved expansion
characteristics lead to en-
hanced sensitivity and reduced
background in a modified in-
strument coupled to laser abla-
tion (laser conditions were kept
constant)

emental fractionation. Both ICPMS were optimised to

provide optimum signal-to-noise ratio for the mass region
between m/z=7 and m/z=238. Additional care was taken
to minimise ICP background counts, oxide formation, and
doubly charged ions. It is important to notice that the data
shown (Fig. 2) were measured on a silicate matrix and
might be completely different for other samples.
More than 200 successful applications cannot hide the
need for fundamental studies to enable further develop-
ment of LAICPMS on the long and winding road to a
broader acceptance.

Fig. 2 Th/U correlation plots for a transient signal acquired by use
of a 193-nm excimer laser system with different ICPMS instru-
ments. The ablation conditions were maintained constant for both
experiments. The slopes of approx. 1 and 2 acquired under opti-
mum operating conditions are indicative of ICP plasma-based ele-
mental fractionation.
nebulisation can lead to dramatic improvements in the de-
tection capability, as illustrated in Fig. 1. The portion of
elemental fractionation occurring in the ionisation source
is, therefore, unknown. The signal response from a 193 nm
ArF Excimer laser using two different ICP, shown in
Fig. 2, is indicative of an influence of the ion source on el-
References
1. Gray AL (1985) Analyst 110:551
2. Moenke-Blankenburg L (1989) Laser microanalysis. John Wiley
and Sons, Toronto
3. Moenke L (1993) Spectrochim Acta Rev 15:1
4. Jackson SE, Longerich HP, Dunning GR, Freyer BJ (1992)
Can Min 30:1049
5. Becker JS, Tenzler D (2001) Fresenius J Anal Chem 370:637
6. Durrant SF (1999) J Anal At Spectrom 14:1385
7. Fryer BJ, Jackson SE, Longerich HP (1995) Can Min 33:303
8. Leach AM, Hieftje GM (2000) J Anal At Spectrom 15:1121
9. Russo RE, Mao XL, Borisov OV, Liu HC (2000) J Anal At
Spectrom 15:1115
10. Mank A, Mason P (1999) J Anal Atom Spectrom 14:1143
10. Eggins SM, Kinsley LPJ, Shelley JMG (1998) Appl Surf Sci-
ence 129:278
11. Gnther D, Heinrich CA (1999) J Anal At Spectrom 14:1369
12. Bleiner D, Gnther D (2001) J Anal At Spectrom 16:449
Anal Bioanal Chem (2002) 372 : 3334
DOI 10.1007/s00216-001-1159-9
TRENDS
J. Bettmer
Elemental speciation
Published online: 11 December 2001
Springer-Verlag 2001
Interest in elemental or metal speciation has increased sig-
nificantly in analytical chemistry since the importance of
the determination of individual species was recognised1.
The identity of the species present has a decisive effect on
the role and properties of a metal in living organisms and
in dead matter. To acquire more knowledge about the ef-
fects and pathways of essential and toxic elements highly
sophisticated, sensitive and selective analytical methods
have been developed. Major target species include organo-
metallic compounds (e.g. methylmercury, butyltin), redox
systems (e.g. Cr(III)/Cr(VI), As(III)/As(V)), and com-
plexes of metals and metalloids with biomolecules (e.g.
metallothioneins). This paper briefly emphasises two de-
velopments in analytical chemistry which have started to
make important contributions to the field of elemental
speciation.
ICPMS (inductively coupled plasmamass spectrom-
etry) is well known as a sensitive, element-selective, and
multi-element detector in trace and ultra-trace speciation
analysis [2]. The analytical possibilities of ICPMS are
unique compared with other (optical) detectors, partly be-
cause of the ease with which it can be coupled to chro-
matographic or electrophoretic separation, and partly be-
cause it is possible to monitor the isotopes of the element
of interest. This additional information enables essential
investigations on artefact formation and species transfor-
mations during sample pre-treatment to be performed us-
ing isotopically labeled species [3, 4, 5, 6]. One example
in which transformations could influence the quality of re-
sults in speciation analysis is the determination of methyl-
mercury in certified reference materials, a topic about
which there is some controversy [7, 8]. A promising tool
1 Definitions
of the terms species, speciation, and speciation analysis
have been recommended by the IUPAC [1].
J. Bettmer ()
Johannes Gutenberg-Universitt Mainz,
Duesbergweg 1014, 55099 Mainz, Germany
e-mail: bettmer@mail.uni-mainz.de
for monitoring irregularities during the analytical process
seems to be isotope-dilution mass spectrometry (IDMS),
well-known as a relatively accurate and precise definitive
method, as demonstrated recently in the determination of
methylmercury [3]. Although degradation of methylmer-
cury to elemental mercury could be observed under some
conditions, species-specific IDMS enabled accurate analy-
sis of methylmercury.
During recent years the speciation of biometals com-
plexed with proteins and peptides, etc., has become a very
important task for analytical chemists [9]. Because com-
pletely characterised standards which can be used in the
same way as well-defined alkyl- or aryl-metal compounds
are rare, this field demands a different approach and, as a
consequence, analysis of these metal-containing biomole-
cules is more complex. The application of multidimen-
sional liquid chromatography (LC) or a combination of LC
and capillary electrophoresis (CE) is essential to guaran-
tee the purity of the separated compounds entering the de-
tection system(s). In addition to the element-selective in-
formation about the metalloproteins given by ICPMS as
detector, evaluation of the molecular structure requires
additional mass spectrometric coupling. Electrospray MS
has proven to be an essential tool for further identification
of macromolecules bound to metals [10], and the applica-
tion of more than one detector will significantly improve
our knowledge about the nature of metal complexes with
biomolecules in the near future. These investigations should
be performed in an interdisciplinary environment to guar-
antee progressive and effective continuation of our under-
standing the role of metals in our life.
References
1. Templeton DM, Ariese F, Cornelis R, Danielssen LG, Muntau
H, van Leeuwen HP, Lobinski R (2000) Pure Appl Chem 72:
14531470
2. Szpunar J, McSheeny S, Pole K, Vacchina V, Mounicou S,
Rodriguez I, Lobinski R (2000) Spectrochim Acta B 55:779
793
3. Demuth N, Heumann KG (2001) Anal Chem 73:40204027
34
4. Hintelmann H, Falter R, Ilgen G, Evans RD (1997) Fresenius J
Anal Chem 358:363370
5. Ruiz Encinar J, Monterde Villar MI, Gotor Santamaria V, Gar-
ca Alonso JI, Sanz-Medel A (2001) Anal Chem 73:31743180
6. Hill SJ, Pitts LJ, Fisher AS (2000) Trends Anal Chem 19:
120126
7. Quevauviller P, Horvat M (1999) Anal Chem 71:155A-156A
8. Bloom NS, Evans RD, Hintelmann H, Wilken RD (1999) Anal
Chem 71:575A
9. Lobinski R, Chassaigne H, Szpunar J (1998) Talanta 46:271
289
10. Chassaigne H, Vacchina V, Lobinski R (2000) Trends Anal
Chem 19:300313
Anal Bioanal Chem (2002) 372 : 3536
DOI 10.1007/s00216-001-1154-1
TRENDS
Christopher P. Palmer Vincent T. Remcho
Microscale liquid phase separations
Published online: 13 December 2001
Springer-Verlag 2001
Abbreviations CEC Capillary electrochromatography
CE capillary electrophoresis micro-LC microscale
liquid chromatography MEKC micellar electrokinetic
chromatography
Miniaturization and microfabrication of analytical instru-
mentation for chemical and biological analyses has been
proceeding at a rapid pace for a few decades, and has
gained momentum in the past decade with the introduc-
tion and application of microfabricated instruments. Moti-
vating factors for performing microscale separations and
analyses are reduced reagent, solvent and sample consump-
tion, reduced cost, improved analysis speed, sensitivity and/
or separation efficiency, and the ability to perform multi-
ple, often parallel, analyses on a large number of samples.
Examples of areas where this is of importance are ge-
nomics, proteomics, screening of combinatorial libraries,
and remote sampling and analysis.
The recent trend toward miniaturization is perhaps most
evident in the area of liquid phase analytical separations.
Rapid development in the areas of capillary electrophore-
sis (CE) and microscale liquid chromatography (micro-
LC) has led to their acceptance and routine use in the an-
alytical laboratory over the past 15 years. Research and
development continue with the introduction of new sepa-
ration modes and materials, sensitive and selective detec-
tion methods, new applications, and further miniaturiza-
tion and application of these techniques in microfabricated
devices.
C.P. Palmer ()
Department of Chemistry, The University of Montana,
Missoula MT 59812, USA
e-mail: palmerc@selway.umt.edu
V.T. Remcho
Department of Chemistry, Oregon State University,
Corvallis OR 97331, USA
Microfabricated instruments
The demonstration of CE separations in etched micro-
channels in a glass substrate in 1992 [1] has led to a very
large number of studies concerning the development of
microfabricated instruments (instruments and instrument
components on microchips) employing electrokinetically-
driven liquid-based separations. Microchips allow chemi-
cal and biological assays to be performed very quickly with
reduced sample consumption and cost. A variety of sepa-
ration modes have been demonstrated with these instru-
ments, including free zone and gel electrophoresis, electro-
chromatography, and micellar electrokinetic chromatogra-
phy (MEKC). Recent reports have demonstrated complete
CE separations in milliseconds [2]. Two dimensional sep-
arations combining either MEKC [3] or electrochromatog-
raphy [4] with free zone electrophoresis have also been
reported.
There has been a great deal of interest of late in fabri-
cation of microchips from plastics or polymeric materials
[5]. These materials allow easier, more cost-effective, and
more varied fabrication techniques to be utilized.
Detection technologies include off-chip UV or mass
spectrometric (MS) detection and on-chip fluorescence and
electrochemical detection. MS detection is an extremely
important capability for microscale separation devices be-
cause of its sensitivity, selectivity and qualitative capabil-
ities.
Sorbents and surface modification
The open channels produced in microchip devices may be
used without further modification, or with surface modifi-
cation to control electroosmotic flow in CE and MEKC.
Surface modifications can also be made to facilitate sorp-
tion of analytes on the channel walls, such that open tubu-
lar capillary electrochromatography (CEC) separations can
be achieved. The use of packed beds [6] or monoliths [7,
8] in microchannels or capillaries increases sample capac-
36
ity. These sorbents can be produced by a variety of meth-
ods, and can support a variety of different chromatographic
modes of separation. The porous nature of both conven-
tional and monolithic sorbents has led to the development
of models with reasonable ability to describe and predict
electroosmotic transport of eluents through pores [9, 10].
Novel materials have also been introduced and investi-
gated for application as pseudo-stationary phases in MEKC.
Ionic amphophilic polymers have generated considerable
interest due to applicability in a variety of buffer matrices,
unique chemical selectivity, and applicability with MS de-
tection [11].
On-column preconcentration
Several on-column preconcentration or stacking methods
have been introduced in the past decade to increase the de-
tection sensitivity of CE 101000 fold [12]. Electrophoretic
stacking mechanisms can efficiently stack ionic or ioniz-
able analytes. Analytes can also be preconcentrated on a
short section of chromatographic packing material at the
head of the capillary column or by using membrane con-
centrators. Ionic and neutral analytes can be preconcen-
trated using a relatively new method in which the sample
is swept or focused through interaction with a pseudosta-
tionary phase as it migrates through the sample zone. Con-
centration enhancement up to 5000-fold has been observed
for non-ionic hydrophobic solutes [13], and enhancement
of nearly a million-fold has been achieved through combi-
nation of this method and field amplified sample stacking
[14]. Several hundred-fold concentration enhancement has
also been reported using high conductivity sample buffers
[15].
Enrichment methods have also been employed in CEC
and capillary LC. El Rassi has used sequential frontal and
elution steps in electrochromatography to provide for trace
analysis [16]. Solid phase micro extraction has also been
employed on-line with microscale separation methods.
The higher loading capacity of capillary LC and CEC has
in general allowed those techniques to avoid some of the
detection sensitivity pitfalls addressed by preconcentra-
tion methods in CE. That said, the challenge of dealing
with small sample volumes in a quantitatively reproducible
manner still remains.
Applications
Microscale liquid phase separations methods continue to
see moderate growth in applications. Not surprisingly,
these methods are applied predominantly in cases where
the improved separation efficiency, selectivity and speed
are of particular importance, or where they can provide
complementary information to more established techniques.
Developments in CEC, capillary LC, CE and MEKC pro-
vide a range of selectivities and separation mechanisms
that extend the reach of microscale separations to encom-
pass a large pool of prospective analytes and matrices.
However, there remains concern about whether techniques
that rely on electroosmotic transport can generally meet
the reproducibility requirements of regulatory agencies.
The complementary nature of these techniques along
with their similar requirements of scale, geometry, and de-
tection will likely result in their use in various combina-
tions on microchips and in more conventional formats to
solve increasingly complex separations problems.
References
1. Harrison DJ, Manz A (1992) Anal Chem 64:1926
2. Jacobsen SC, Culbertson T, Daler JD, Ramsey JM (1998) Anal
Chem 70:3476
3. Rocklin RD, Ramsey RS, Ramsey JM (2000) Anal Chem 72:
5244
4. Gottschlich N, Jacobsen S, Culbertson CT, Ramsey JM (2001)
Anal Chem 73:2669
5. Soper SA, Ford SM, Qi S, McCarley RL, Kelly K, Murphy MC
(2000) Anal Chem 72:643A
6. Shultz-Lockyear LL, Colyer CL, Fan ZH, Roy KI, Harrison DJ
(1999) Electrophoresis 20:529
7. Chirica GS, Remcho VT (2001) J Chromatogr A 924:223
8. Svec F, Frechet JM (1996) J Die Makromolekulare Chemie,
Macromolecular symposia 110:203
9. Stol R, Poppe H, Kok W T (2000) J Chromatogr A 887:199
10. Vallano PT, Remcho VT (2000) Anal Chem 72:4255
11. Shamsi SA, Palmer CP, Warner IM (2001) Anal Chem 73:
140A
12. Osbourn DM, Weiss DJ, Lunte CE (2000) Electrophoresis 21:
2768
13. Quirino JP, Terabe S (1998) Science 282:465
14. Quirino JP, Terabe S (2000) Anal Chem 72:1023
15. Palmer J, Munro NJ, Landers JP (1999) Anal Chem 71:1679
16. Tegeler T, El Rassi Z (2001) Anal Chem 73:3365
Anal Bioanal Chem (2002) 372 : 3738
DOI 10.1007/s00216-001-1155-0
TRENDS
Peter Spgel Leif Schweitz Staffan Nilsson
Molecularly imprinted polymers
Published online: 12 December 2001
Springer-Verlag 2001
Abbreviations MIT Molecular imprinting technology
MIP molecularly imprinted polymer CEC capillary
electrochromatography SPE solid phase extraction
Molecular imprinting technology (MIT) offers a means of
preparing materials with cavities that are able to recognise
a certain molecule in terms of shape, size and chemical
functionality. In order to obtain a highly selective recog-
nition of a certain molecule, this molecule is incorporated
during the synthesis of the material as a template. When
the synthesis is completed, the molecule is extracted, thus
leaving behind a three dimensional physical and chemical
imprint of itself. The first appearance of MIT dates back
to the early 1930s [1] and the material used was silica
gels. In the 1940s the work mainly constituted fundamen-
tal studies on the mechanism of antibody antigen interac-
tions using MIT to prove the hypothesis that the antibody
adapted its selectivity first upon presentation to the anti-
gen [2]. This hypothesis of antibody formation was later
proven wrong, and as limitations of the silica material for
MIT were identified, there was a decline in MIT research.
In 1972 covalent MIT was introduced by Wulff at al. [3]
and in the early 1980s a successful preparation of a mole-
cularly imprinted polymer (MIP) using non-covalent MIT
was presented by Mosbach et al. [4]. These techniques of-
fered a means of preparing molecular imprints in organic
matrices (Fig. 1). The limitations of the silica material, i.e.
stability and reproducibility, were avoided and the interest
in MIT accelerated to yield an almost exponential increase
in published papers on MIT during the past few years.
The first works with these molecularly imprinted poly-
mers concerned chromatographic separations. In separa-
P. Spgel L. Schweitz S. Nilsson ()
Technical Analytical Chemistry,
Center for Chemistry and Chemical Engineering,
Lund University, PO Box 124, SE-221 00 Lund, Sweden
e-mail: Staffan.Nilsson@teknlk.lth.se
tion technology the gain of using MIPs lies in the ability
to perform demanding separations, e.g. enantiomer sepa-
rations, by using one of the tricky analytes as imprint
molecule in the preparation of the MIP. Bulk polymerised,
ground, sieved and slurry packed MIP HPLC columns have
been used for a broad range of demanding separations [5].
Recently, also capillary electrochromatography (CEC), that
offers improved separation efficiency when compared to
HPLC, has been used successfully in combination with
MIPs [6]. The miniaturised separation channels used in
this technique offers a diminution in chemical consump-
tion both in the preparation of the MIP, and then espe-
cially the precious template, and in the analysis. Also, the
miniaturised format has forced new MIP formats to be de-
veloped [7, 8, 9]. However, so far the expectations of the
MIP-CEC system in terms of efficiency have not been
fully achieved. During the past few years much effort has
been put into the development of MIP materials for solid
phase extraction (SPE) [10]. MIP-SPE has been shown
superior to ordinary SPE and liquid-liquid extraction in its
ability to produce much cleaner extracts [11]. The selec-
tivity and the stability of the MIP also offer faster method
development and cost efficiency. Leakage of imprint mol-
ecule from the MIP is a problem that is arising from the
difficulties in removing all of the template used in the
synthesis from the resulting polymer. This problem, that
can be disastrous when MIP-SPE is used in trace analysis,
can be circumvented by the use of an analogue to the im-
print molecule in the synthesis [11]. Using the template
analogue approach the difficulties in finding a pure sam-
ple of the analyte to be used as template can be circum-
vented. Apart from the more commonly used MIP based
separation techniques mentioned above, MIPs have also
been used in thin layer chromatography [12], membrane
separations [13] and bubble fractionation [14]. The anti-
body analogue features of the MIP have introduced it as a
complement to the biological antibodies in different types
of binding assays [10]. Another application of the MIP
that is growing in interest is the use of the MIP as the
recognition element in sensors [15]. Using a general de-
tection method, e.g. a mass sensitive sensor, the high se-
38
Fig. 1 Schematics of MIP
preparation. The non-covalent
approach: (A) Functional
monomers, cross-linking
monomers, radical initiator and
imprint molecule are mixed in
a proper solvent to allow com-
plexes to form between the
functional monomers and the
imprint molecule. (B) The
functional monomers are
locked in position by the poly-
merisation reaction, and (C) af-
ter having extracted the imprint
molecule (D) the MIP is able
to recognise the imprint mole-
cule. The covalent approach:
(A) Imprint molecules with
substitutions of polymerisable
groups are mixed with cross-
linking monomers and radical
initiator in a proper solvent and
polymerisation is initiated.
(C) After completed polymeri-
sation the covalent bonds be-
tween the imprint molecule
and the MIP can be broken and
(D) the MIP is able to rebind
the imprint molecule utilising
covalent bonds

lectivity of the MIP can be potentially useful in sensor ap-

plications.
Today MIT has spread into almost all the different ar-
eas of chemistry and particularly in analytical chemistry.
From being a technique only explored by a few academic
groups, it has spread tremendously during the past years
and today academia and industry in a broad range of re-
search areas are interested in the research. The develop-
ment started with stationary phases for chromatography
and today the application of the MIP in SPE seems to be
one of the most promising areas for MIT. The ability of-
fered by MIT to build selective stationary phases with
chemical building blocks seems very promising for future
chip applications. However, MIT is still very young and
its future might lie in other areas where the selectivity, sta-
bility and cost efficiency of the MIP could offer a solution
to a demanding analytical problem.
References
1. Polyakov MV (1931) Zhur Fiz Khim 2:799
2. Anonymous (1949) Chem Eng News 27:913
3. Wulff G, Sarhan A (1972) Angew Chem 84:364
4. Arshady R, Mosbach K (1981) Macromol Chem Phys 182:687
5. Remcho VT, Tan ZJ (1999) Anal Chem 71:248A
6. Schweitz L, Nilsson S (2001) In: Sellergren, B (ed) Molecu-
larly imprinted polymers man-made mimics of antibodies
and their applications in analytical chemistry. Elsevier, Am-
sterdam, p 377
7. Schweitz L, Andersson LI, Nilsson S (1997) Anal Chem
69:1179
8. Tan ZJ, Remcho VT (1998) Electrophoresis 19:2055
9. Schweitz L, Spgel P, Nilsson S (2000) Analyst 125:1899
10. Andersson LI (2000) J Chromatogr B 739:163
11. Andersson LI (2000) Analyst 125:1515
12. Kriz D, Berggren C, Andersson LI, Mosbach K (1994) Anal
Chem 66:2636
13. Mathew-Krotz J, Shea KJ (1996) J. Am Chem Soc 118:8154
14. Armstrong DW, Schneiderheinze JM, Hwang Y-S, Sellergren
B (1998) Anal Chem 70:3717
15. Dickert FL, Hayden O (1999) Fresenius J Anal Chem 364:506
Anal Bioanal Chem (2002) 372 : 3940
DOI 10.1007/s00216-001-1157-y
TRENDS
Valrie Camel
Extraction techniques
Published online: 11 December 2001
Springer-Verlag 2001
Over the past few years, efforts of the analysts focused on
the development of extraction techniques that allow effi-
cient extraction along with reduced solvent volumes and
time, and a high level of automation. The recent extrac-
tion techniques available, along with the traditional ones,
are presented below (Fig. 1).
When extracting liquid samples, the traditional liquid-
liquid extraction faces several limitations, namely use of an
extractant non-miscible with the sample, difficulty in ex-
tracting polar and ionic compounds from water, large or-
ganic solvent volumes resulting in a diluted extract. To pre-
vent these major drawbacks, two major techniques have
merged: solid-phase extraction (SPE) and solid-phase mi-
croextraction (SPME). In SPE [1, 2], the sample is perco-
lated through a solid phase which retains the solutes of in-
terest. They are further eluted with a small solvent vol-
ume. This technique offers the unique advantages of high
concentration of the final extract, selectivity, as well as
wide choice of the solid phase enabling the extraction of
virtually all compounds from aqueous or organic matrices.
In addition several systems are now available (manual mul-
tiple cartridge systems, disks and multiwell plates) with
possible automation and coupling to chromatography. The
main drawback is the sometimes non-compatibility of the
final extract solvent with the analytical system, leading to
a solvent evaporation and further dissolution of the residue
in an acceptable solvent. This problem is overcome by the
use of SPME, which is a real non-solvent technique [1, 3].
A fiber, covered with a polymeric phase, is put into the
sample or its headspace, and the solutes partition between
the sample and the phase. Desorption into a gas chromato-
graph injector enables analysis of the solutes to be carried
out. Recently, SPME coupling with liquid chromatography
has appeared, which should increase the use of this tech-
nique. In addition, SPME may also be used for gaseous
V. Camel ()
Institut National Agronomique Paris-Grignon,
Laboratoire de Chimie Analytique,
16 rue Claude Bernard, 75231 Paris Cedex 05, France
e-mail: camel@inapg.inra.fr
samples [4]. Of interest for certain applications is also stir
bar sorptive extraction (SBSE) due to its low cost and
simplicity [5]. This technique can be considered as a vari-
ant of SPME, as a stir-bar covered with a polymeric phase
is put into the sample, and partitioning of the solutes en-
ables their extraction. Analysis is further performed after
thermal desorption. However, both techniques are limited
by the nature of the sorbent phase available, thus restrict-
ing their applications. In addition, extractions are rarely
complete due to the low volume of the sorbent phase.
For solid matrices, from which solutes are difficult to
extract due to the frequently strong interactions with the
matrix, the traditional techniques (Soxhlet and ultrasonic
extractions) now face severe competition due to the de-
velopment of supercritical fluid extraction (SFE), pressur-
ized fluid extraction (PFE) and microwave-assisted ex-
traction (MAE). As the extraction temperature is a crucial
parameter, these recent techniques afford the opportunity
of elevated temperatures. In that way, rapid and efficient
extractions can be performed due to enhanced solubilities,
higher diffusion coefficients, lower viscosities, as well as
better desorptions of the solutes from the matrix (when
solutes are stable under high temperatures). The interest in
SFE [1, 6], which uses supercritical fluids as the extrac-
tant (with properties intermediate between those of gas
and liquids), has dramatically decreased in the past five
years, in spite of its unique elevated selectivity. This is
due to the high dependence of the extraction conditions on
the sample, leading to fastidious optimization procedures
and difficulty in using this technique routinely. The lower
interest in SFE is also partly due to the development of
PFE [1, 7], which uses a liquid solvent as the extractant.
The liquid state is maintained under elevated temperatures
upon application of a moderate pressure. Optimization of
extraction conditions is facilitated as organic solvents rec-
ommended in traditional techniques may be used most of
the time and pressure has a low influence. In addition, de-
veloped systems offer a high level of automation. Yet,
only one extraction at a time can be conducted. On the other
hand, MAE [1, 8] affords the opportunity of performing
several simultaneous extractions with closed-vessel sys-
40
Fig. 1 Main extraction tech-
niques for gaseous, liquid and
solid samples
tems (pressurized MAE), provided that similar extractions
are performed to avoid over-pressure in any cell. The cell
containing the sample and the liquid solvent is subjected
to microwave radiation that enables instantaneous and ef-
ficient heating in case of the presence of microwave-ab-
sorbing compounds. However, this technique requires fur-
ther filtration to obtain the final extract. Alternatively
MAE may be conducted in open-vessel systems (atmo-
spheric MAE), but as only limited extractions can be per-
formed simultaneously, few systems are in use. Finally,
the major drawback of PFE and MAE, which slows down
their development, is the high investment cost of the com-
mercialized systems [9].
References
1. Dean JR (1998) Extraction methods for environmental analysis.
John Wiley and Sons, Chichester
2. Thurman EM, Mills MS (1998) Solid phase extraction: princi-
ples and practice. John Wiley, New York
3. Pawliszyn J (1997) Solid phase microextraction: Theory and
practice. John Wiley, New York
4. Camel V, Caude M (1995) J Chromatogr A 710:319
5. Baltussen E, Sandra P, David F, Cramers C (1999) J Microcol-
umn Sep 11:737747
6. Ramsey ED (1998) Analytical supercritical fluid extraction tech-
niques. Kluwer Academic Publishers, London
7. Richter BE, Jones BA, Ezzell JL, Porter NL, Avdalovic N, Pohl
C (1996) Anal Chem 68:10331039
8. Camel V (2000) Trends Anal Chem 19:229248
9. Camel V (2001) Analyst 126:11821193
Anal Bioanal Chem (2002) 372 : 4142
DOI 10.1007/s00216-001-1161-2

TRENDS

Alexander Jung

DNA chip technology

Published online: 13 December 2001

Springer-Verlag 2001

particular, will probably dominate the market in the fu-

Introduction
In the last few years genomic research has demanded highly
parallel analytical approaches. Undoubtedly the most pow-
erful development has been the realization of so called
DNA chips segmented, planar arrays of immobilized DNA
fragments used in a wide field of applications from ex-
pression analysis (high-density arrays) to diagnostic chips
(low-density arrays).
All the components involved in a microarray experi-
ment influence the analytical validity of the results. These
include the components and technologies used to modify
the chip surface and perform the surface chemistry; the
hybridization procedure and the (micro-)fluidic devices;
the readout systems (scanners); and last but not least the
evaluation software. Each of these components has its
own specific problems and optimizable conditions. Instru-
ments for each experimental step of a microarray experi-
ment are, furthermore, commercially available. State of
the art are custom-made chips that are easily handled by
the customer and can be read out by almost every scanner.
Other approaches try to combine the complete experimen-
tal procedure in one instrument. An instrument, such as a
gene-sensor, could then be easily applied to almost any
specific analytical problem, for instance diagnostic appli-
cations and food testing. Initial solutions have already
been published [1] and much work is being invested in this
idea.
Fields of application
Application of DNA arrays in gene-expression analysis
[2] will continue to be very important; two new fields, in
A. Jung ()
Department of Physical Chemistry,
Faculty of Chemistry, University of Tbingen,
Auf der Morgenstelle 8, 72076 Tbingen, Germany
e-mail: aj@ipc.uni-tuebingen.de
ture. One is personalized medicine, which will make it
possible to repeat the complete sequencing of the genome
and search for the tens of thousands of single nucleotide
polymorphisms (SNP) for each individual. The platforms
for this must have an immensely high throughput and
much higher reliability than those of current systems. The
other main application is in the field of food control and
environmental monitoring and includes the detection of
pathogenic organisms and DNA chip-based toxicity mod-
els [3].
Microarray production
Arrays of immobilized DNA sequences can be obtained
by localized synthesis of probe sequences in situ on a
chip, as, for example, performed by Affymetrix. Photolitho-
graphic techniques have been established for the produc-
tion of highly resolved DNA patterns in high-density ar-
rays with 50 m feature size [4, 5]. Incomplete reactions
lead to failure sequences which cannot be removed and
limit the maximum sequence length and the fidelity of the
process. Another technology used to generate oligonu-
cleotide sequences on a DNA chip, the so-called -wet
printing, which combines high coupling-yields with fea-
ture sizes in the micron range, could become a very im-
portant technology in this field [6]. An alternative means
of spatially resolving in-situ synthesis is based on the con-
trolled delivery of activated nucleotides into selected hy-
drophobic wells on a structured surface [7].
The most widely used technology is, furthermore, the
localized immobilization of complete pre-assembled DNA
strands. Although this might, on the one hand, be more re-
stricted with regard to the maximum number of sequences
that can be handled, it does, on the other hand, enable im-
mobilization of stock DNA fragments for very economi-
cal manufacture of small arrays with a limited number of
immobilized probes. A good overview of the versatile pos-
sibilities of immobilization procedures is given by Beier
and Hoheisel [8]. Some very special applications have re-
42
sulted from the combination of modern optical technology
with, for instance, direct immobilization of DNA features
within a standard DNA-synthesizer by use of micro-mir-
ror arrays [9].
Detection methods
The complex analytical questions that must be solved by
DNA chips lead to the demand for improvement of the in-
formation depth of a DNA chip experiment. On commer-
cially available systems array detection is achieved almost
solely by use of fluorescently labeled oligonucleotides.
Several instruments are available which can be used to de-
tect hybridization on label-free systems (biosensors), for
example SPR (Biacore), resonant mirror (IAsys), or re-
flectometric interference spectroscopy (RIfS) [10]. Because
these surface-based detection methods do not require the
use of labeled compounds, one can detect directly, in ex-
tracts, PCR products or other complex matrices. Possible
influence of the interaction through the label, influences
of the surface on fluorescence intensity, or any unspecific
quenching effects or photobleaching can also be avoided.
Label-free measurement enables one to keep hybrid-
ization times short, because not only endpoint detection
but also dynamic measurement of association and dissoci-
ation are performed online. Sub-nanomolar concentra-
tions could be measured in label-free, flow-through hy-
bridization experiments.
The availability of this technology for simultaneous,
time-resolved detection of hundreds of hybridization
events on a DNA chip is an achievable goal in the near fu-
ture.
References
1. Anderson RC, Su X, Bogdan J, Fenton J (2000) Nucleic Acids
Res 28:e60
2. Freeman WM, Robertson DJ, Vrana KE (2000) Biotechniques
29:10421055
3. Blohm DH, Guiseppi-Elie A (2001) Curr Opin Biotech 12:41
47
4. http://www.affymetrix.com
5. Beier M, Hoheisel JD (2000) Nucleic Acids Res 28:e11
6. http://www.clondiag.com
7. Blanchard AP, Kaiser RJ, Hood LE (1996) Biosens Bioelectron
11:687690
8. Beier M, Hoheisel JD (1999) Nucleic Acids Res 27:19701977
9. Singh-Gasson S, Green RD, Yue Y, Nelson C, Blattner F,
Sussman MR, Cerrina F (1999) Nat Biotechnol 17:974978
10. Sauer M, Brecht A, Chariss A, Stemmler I, Gauglitz G, Bayer
E (1999) Anal Chem 71:28502857
Anal Bioanal Chem (2002) 372 : 43
DOI 10.1007/s00216-001-1150-5
TRENDS
John G. Westerfeld
Detection trends in high throughput screening
Published online: 11 December 2001
Springer-Verlag 2001
In the early days of high throughput screening (HTS) for
drug discovery the detection of macromolecules and the
ligands that bond to them have generally been limited to a
small number of techniques, namely absorbance (colori-
metric), radiolabeling and fluorescence intensity measure-
ments. As the art and science of HTS is evolving, many
more methods for detecting molecules of interest are be-
ing explored. One such method is time resolved fluores-
cence (TRF) and its progeny, homogeneous time resolved
fluorescence (hTRF) and fluorescence resonance energy
transfer (FRET) [1]. These are fluorometric methods that
use lanthanide chelates to give a long-term emission sig-
nal (TRF) or closely spaced moieties in order to transfer
light between a donor and acceptor (FRET) or a combina-
tion of the two (hTRF).
Another fluorescent method gaining wide utilization is
fluorescent polarization (FP) [2], in which large and small
molecule interactions enhance or diminish polarized fluo-
rescence signal. Its chief advantage is that it is a homoge-
neous assay. One recent survey of HTS research directors
indicates that in five years the three most widely used de-
tection labels will all be fluorometric [3].
One more fluorescence technique that is in its incipient
phase toward HTS is fluorescence correlation spectroscopy
(FCS) [4]. The principle is to measure the fluctuations in
fluorescence intensity caused by labeled molecules enter-
ing and leaving a small volume defined by a sharply focused
laser beam. This evaluation provides information on diffu-
sion times of molecules and aggregates in the sample and
number of molecules corresponding to a certain diffusion
time, which reflects molecular weight and structure. FCS
can measure a wide range of parameters, such as, protein
mass, equilibrium constants, and stoichiometry of binding.
Also on the rise as an alternative to radiometric and
fluorometric based methods is the use of chemi/biolumi-
nescence detection [5]. Generally the light produced by
these reactions is as, or more, sensitive than the former
methods with a wider dynamic range. With luminescence,
J.G. Westerfeld ()
Senior Account Specialist, BioWhittaker, Inc.,
Walkersville, MD 21793, USA
e-mail: John.Westerfeld@biowhittaker.com
waste disposal and handling are safer than other methods.
Also in this area of detection is electrochemilumines-
cence, in which certain chemical compounds emit light
when electrochemically stimulated, using ruthenium com-
plexed ions as the progenitor.
Other detection methods, which have been around for
some time, are being re-tooled for applications in HTS.
Capillary electrophoresis (CE) is undergoing research for
adaptation to HTS. Mass spectrometry (MS) and variations
of MS, such as, matrix-assisted laser desorption ioniza-
tion-time-of-flight MS (MALDI-TOF-MS) [6] are being
investigated as potential tools in protein and peptide mea-
surements in HTS. Another technology that holds hope
for the future is surface plasmon resonance (SPR) [7], like
CE and MS, a label free detection method. The principle
of SPR is that incident light of a certain wavelength that
impinges upon a metal surface to which specific mole-
cules, like a DNA fragment, are attached will interact with
the electrons of the metal and cause a plasmon phenome-
non. Any molecules that bond with the attached mole-
cules will shift the wavelength of the light, which can then
be measured and quantified. This is expected to have great
promise in DNA screening.
The future of HTS will change as the detection tech-
nologies change and the detection technologies are sure to
change as the knowledge of chemistry, physics and biol-
ogy continues to grow and new detection methods are dis-
covered.
References
1. Hemmil I, Webb S (1997) Drug Discov Today 2:373381
2. Prystasy L, Gosselin M, Banks P (2001) J Biomol Screening 6:
141150
3. Fox S, Sopchak L, Khoury R, Wang H (2001) Drug Discov Dev
4(4)
4. Rigler R, Elson E (2001) (eds) Fluorescence correlation spec-
troscopy. Springer, Berlin
5. Roda A, Pazzali M, Kricka LJ, Stanley PE, (eds) Biolumines-
cence and chemiluminescence: perspective for the 21st Century:
proceedings of the 10th international symposium on biolumines-
cence and chemiluminescence (1998) Sept. 48, Bologna, Italy,
John Wiley and Sons
6. Sinclair B (1999) The Scientist 13:18
7. www.biacore.com
Anal Bioanal Chem (2002) 372 : 4448
DOI 10.1007/s00216-001-1189-3
REVIEW
Ciara K. OSullivan
Aptasensors the future of biosensing?
Received: 4 September 2001 / Revised: 19 September 2001 / Accepted: 2 October 2001 / Published online: 13 December 2001
Springer-Verlag 2001
Abstract Aptamers are artificial nucleic acid ligands that
can be generated against amino acids, drugs, proteins and
other molecules. They are isolated from combinatorial li-
braries of synthetic nucleic acid by an iterative process of
adsorption, recovery and reamplification. Aptamers, first
reported in 1990, are attracting interest in the areas of
therapeutics and diagnostics and offer themselves as ideal
candidates for use as biocomponents in biosensors (ap-
tasensors), possessing many advantages over state of the
art affinity sensors. The properties of aptamers, their ap-
plicability to biosensor technology, current research and
future prospects are addressed in this short review.
Keywords Aptamers ELONA Aptasensors SELEX
Abbreviations SELEX systematic evolution of ligands
by exponential enrichment Aptasensors biosensors
utilising aptamers as biocomponent ELISA enzyme
linked immunosorbent assay ELONA enzyme linked
oligonucleotide assay FITC fluoroscein isothiocyanate
Introduction
In 1990, within months of each other, the laboratories of
G.F. Joyce (La Jolla, USA.) [1], J.W. Szostak (Boston,
USA.) [2] and L. Gold (Bolder, USA.) [3] independently
reported on the development of an in vitro selection and
amplification technique, which has allowed the discovery
of specific nucleic acid sequences that bind a wide array
of non-nucleic acid targets with high affinity and speci-
ficity. The technique by which these oligonucleotide lig-
ands are obtained was coined as SELEX systematic evo-
lution of ligands by exponential enrichment, and the result-
ing oligonucleotides are referred to as aptamers, derived
C.K. OSullivan ()
Bioelectrochemistry and Bioengineering Group,
School of Chemical Engineering,
Universitat Rovira i Virgili, Tarragona 43006, Spain
e-mail: ckosulli@etse.urv.es
from the latin aptus, meaning, to fit [4]. The last decade
has seen the selection of an extensive range of aptamers
with abilities to bind to small molecules, peptides, pro-
teins, cells etc. with selectivity, specificity and affinity
equal and often superior to those of antibodies [5, 6, 7, 8,
9]. Aptamers possess the ability to discriminate targets on
the basis of subtle structural differences such as the pres-
ence or absence of a methyl or hydroxy group or the D-
vs. L-enantiomeric configuration of the target [10]. In par-
allel, the body of literature covering immunosensor devel-
opment has exploded over the last twenty years with nu-
merous generic transduction platforms reported. How-
ever, as outlined in Table 1, antibody generation, particu-
larly for use in biosensors, has several fallbacks that are
addressed by aptamers and it can be envisaged that aptasen-
sors, using already developed or novel transduction plat-
forms, will be increasingly exploited in the coming years.
Aptamer selection The SELEX process
The SELEX process, depicted in Fig. 1, is a technique for
screening very large combinatorial libraries of oligonu-
cleotides by an iterative process of in vitro selection and
amplification. Nucleic acid libraries are easily obtained via
combinatorial chemistry synthesis. Each sequence synthe-
sised represents a linear oligomer of unique sequence and
the molecular diversity is dependent on the number of
randomised nucleotide positions. Typically libraries of at
least 10131018 independent sequences are employed, with
a variable region of 30 bases flanked by primers of 2025
bases normally being employed.
In the process, a random sequence oligonucleotide li-
brary is incubated with a target of interest in a buffer of
choice at a given temperature. In the initial cycles of se-
lection, a tiny percentage (0.10.5%) of individual se-
quences interact with the target. These sequences are sep-
arated from the rest of the library by using techniques
such as affinity chromatography or filter binding [11, 12].
This isolated population of sequences is then amplified to
obtain an enriched library to be used for the proceeding se-

Table 1 Advantages of aptamers over antibodies for analysis using enzyme linked oligonucleotide assays and aptasensors
Antibodies
Limitations against target representing constituents of the body
and toxic substances
Kinetic parameters of Ab-Ag interactions cannot be changed
on demand
Antibodies have limited shelf life and are sensitive to temperature
and may undergo denaturation
Identification of antibodies that recognise targets under conditions
other than physiological is not feasible
Antibodies often suffer from batch to batch variation
Requires the use of animals
Labelling of antibodies can cause loss in affinity
Fig. 1 Schematic indicating steps in-
volved in aptamer selection using the
SELEX process
45
Aptamers
Toxins as well as molecules that do not elicit good immune
response can be used to generate high affinity aptamers
Kinetic parameters such as on/off rates can be changed on demand
Denatured aptamers can be regenerated within minutes, aptamers
are stable to long term storage and can be transported at ambient
temperature
Selection conditions can be manipulated to obtain aptamers with
properties desirable for in vitro assay e.g. non-physiological buffer/T
Aptamers are produced by chemical synthesis resulting in little or
no batch to batch variation
Aptamers are identified through an in vitro process not requiring
animals
Reporter molecules can be attached to aptamers at precise
locations not involved in binding
46
lection/amplification cycle. The efficiency of enrichment
of high-affinity binders is governed by the stringency of
selection of each round. To this end, negative selection (re-
moval of aptamers that bind ligand support) and counter
selection (removal of aptamers that bind to structures sim-
ilar to that of the target) are employed. The length of in-
cubation time of library with target used is manipulated to
yield aptamers of desired kinetics. Upon achieving affin-
ity saturation, typically after 815 cycles of selection/am-
plification, the enriched library is cloned and sequenced
and individual sequences investigated for their ability to
bind to the target (e.g. using surface plasmon resonance)
[13]. Aptamers can then be truncated as desired, eliminat-
ing the fixed primer sequences as well as those identified
not to be part of the consensus motif (i.e. sequence required
for binding). When the desired sequence has been identi-
fied the aptamer can be produced in sizeable quantities by
chemical synthesis [4].
Nuclease resistant aptamers
Unmodified oligonucleotides, especially RNA, have a very
short lifetime in biological fluids (typically <10 min) due
to the action of nucleases [14]. Reports have been pub-
lished detailing approaches to make oligonucleotide se-
quences resistant to nucleases by modifying the oligonu-
cleotide backbone. However, most of these modifications
result in structures not recognised by the enzymes used in
the SELEX process. Recently however, modification of
Fig. 2 Schematic of possible assay
formats for use in ELONAs and ap-
tasensor
the 2 positions of pyrimidine nucleotides with amino/flu-
oro groups has been demonstrated to dramatically increase
the half life of aptamers in biological fluids with the mod-
ified structures still being recognised as substrates for the
enzymes utilised in the SELEX process [15, 16].
Aptamers in analysis
a. ELONA Enzyme linked oligonucleotide assay
As antibodies are used in ELISA, aptamers may be utilised,
but with a much greater degree of flexibility in ELONAs.
As depicted in Fig. 2, various formats can be exploited us-
ing, for example,
aptamers as capture molecules and antibodies as reporter
molecules
antibodies as capture molecules and labeled aptamers
against the antibody-analyte immunocomplex as reporter
molecules
aptamers as capture molecules and labeled aptamers
against the aptamer-analyte complex as reporter molecules
The use of aptamers thus facilitates the detection of small
and large molecular weight analytes in a sandwich-type
format overcoming the limitations of competitive-type as-
says, allowing the development of robust, sensitive assays.
The use of ELONA was patented by NexStar [17] (now
Gilead Sciences, Foster City California) in 1997 and a
handful of publications have also appeared [18].

b. Aptasensors
The application of aptamers as biocomponents in biosen-
sors offers a multitude of advantages over the state of the
art in affinity sensing. These advantages include:
kon and koff rates can be tailor designed according to the
nature of transduction and the desired assay time
small molecules can be detected using sandwich-type
formats eliminating need for competitive type assays
and the associated stringent assay requirements
selection of aptamers can be performed in conditions
akin to that of real matrix, particularly useful for envi-
ronmental and food aptasensors
modification of aptamers during immobilisation or when
labelling with reporter molecules can be facilitated with-
out affecting affinity
aptamers can, if desired, be subjected to repeated cycles
of denaturation and regeneration
Additionally, dependent upon the time required for aptasen-
soric detection, the use of nuclease resistant aptamers may
be avoided, a notable advantage over ELONA.
At the time of preparing this review, reports of aptasen-
sors have been limited, and have been almost exclusively
based on the use of fluorescence transduction, the field of
aptasensing still being in its infancy.
Kleinjung et al. (1998) [19] immobilised a biotinylated
RNA aptamer selected to L-adenosine on an optical fibre
surface derivatised with streptavidin. Real time measure-
ment was obtained using total internal reflection fluores-
cence in a fibre optic format and the sensor utilised a com-
petitive format using FITC-labeled L-adenosine as a re-
porter molecule. Quantitation in the submicromolar range
with selective binding to L-adenosine and a chiral discrim-
ination of 1700-fold was observed.
Potyrailo et al. (1998) [20] reported on the first ap-
tasensor that offers one-step direct detection of analyte.
An anti-thrombin DNA aptamer was labelled at the 5 end
with FITC and the 3 end with an alkyl amine so as to im-
mobilise it to a glass surface and fluorescence anisotropy
was used to detect the bound labelled aptamer probe-ana-
lyte binding event. The assay was completed in 10 min
and as little as 5 nM protein could be detected in an ad-
dressed volume of 5 nL.
Lee and Walt (2000) [21] also reported on the use of a
fibre-optic biosensor using an aptamer receptor for the
measurement of thrombin but used a competitive mode of
assay. An anti-thrombin DNA aptamer was immobilised
on the surface of silica microspheres the beads distributed
in microwells on the distal tip of an imaging fibre The
imaging fibre was coupled to a modified epifluorescence
microscope system, and the distal end of the fibre was in-
cubated with a fluorescein-labeled thrombin (F-thrombin)
solution and thrombin was detected via competitive assay.
The aptamer beads selectively bound to the target and could
be reused without any sensitivity change. The fibre-optic
microarray system had a detection limit of 1 nM, and each
test could be performed in 15 min, including regeneration
of the sensor.
47
Hesselberth et al. (2000) [22a,b] detailed the efforts of
their group to select signalling aptamers for use in ap-
tasensors ab initio, postulating that selecting for aptamers
already labelled with reporter molecules would minimise
any effect of post-selection modification on aptamer affin-
ity, but have yet (at time of manuscript preparation), to
publish conclusive results.
Stojanovic et al. (2001) [23a] reported a sensor for de-
tection of cocaine by ingeniously engineering an instabil-
ity in one stem of a three-way junction that formed the co-
caine binding pocket with the short stems labelled with a
fluorophore and a quencher. In the absence of cocaine, the
stems are open but in its presence they close and the three-
way junction forms. This major structural change brings
fluorescence and quencher together, thus signalling the
presence and concentration of analyte. The sensor was se-
lective for cocaine over its metabolites and was operable
in serum. The group had previously reported on extending
this ability to rationally design aptamers and constructed
bipartite aptamers that self-assemble in the presence of
their ligands [23b].
c. The use of molecular beacons in aptasensing
Aptasensors (in contrast to immunosensors) lend them-
selves readily to the transduction of target binding using
molecular beacons. Molecular beacons essentially contain
two structural components, a loop and a stem, with the
loop serving as a probe and the annealing of two comple-
mentary arm sequences that are flanked by the probe
forms the stem (see Fig. 3). A fluorophore and fluorescent
quencher are linked covalently at each end of the arm.
The stem of the beacon brings the fluorophore and
quencher into close proximity, resulting in no fluorescent
signal. When the molecular beacon encounters a target
molecule it forms a probe-target hybrid that is stronger
and more stable than the stem in the hairpin, with the re-
sulting conformational change forcing the arms apart,
thus permitting the fluorophore to fluoresce [24]. The use
of molecular beacons can be exploited for aptasensor de-
velopment, with several modes of assay possible, the most
straightforward being that the aptamer is incorporated in
an immobilisable hairpin structure. Yamamoto and Kumar
(2000) [25] have reported on the development of an im-
mobilisable molecular beacon aptamer that fluoresces
Fig. 3 Schematic of action of fluorescent molecular beacon. F flu-
orophore, Q quencher
48
specifically in the presence of TAT-1 protein, derived from
HIV-1 or HIV-2, but not in the presence of RNA binding
proteins and are working towards its incorporation in an
aptasensor.
Conclusion and outlook
This review gives an overall introduction into the selec-
tion of aptamers and their potential uses in analysis via
ELONAs and aptasensors, outlining their advantages over
the state of the art in affinity based analysis and reviewing
publications in the area. Despite the wealth of literature
detailing the use of aptamers in therapeutics, the use of
aptamers in in vitro diagnostics is a field that is still in its
infancy with the main focus of literature currently available
being on the detection of the aptamer-target binding event
using fluorescence transduction. However, with the advent
of automated platforms for aptamer selection coupled with
the advances being made in increasing the stability of ap-
tamers to nucleases, it is anticipated that aptamers will be-
come increasingly accessible to researchers in the biosen-
sor field.
The plethora of reported transduction platforms for
affinity sensing can be readily adapted to the use of ap-
tameric biocomponents. In addition, catalytic aptamers
(aptazymes) have been reported, offering further modes of
binding event detection. Moreover, it can be anticipated
that the use of immobilisable molecular beacons will be
exploited in the biosensor field (to date a handful of pub-
lications exist), a mode of detection highly suited to ap-
tasensors facilitating reagentless one step analysis.
Overall, the potential of aptasensors is immense, and
this exciting area is on the brink of exponential growth.
The ability to develop affinity based detection systems of
tailor designed characteristics, which can be applied to
analysis of analytes, unlimited, for example, by size, tox-
icity and matrix effects, offers the field of biosensing the
opportunity to explore new and dynamic routes of sensor
development and the availability of commercial aptasensors
by the end of the decade can definitively be anticipated.
Acknowledgements The author wishes to thank Jean-Jacques
Toulm of INSERM Unit 386 at the Universit Victor Segalen II,
Bordeaux, France and Ioanis Katakis of Universitat Rovira i Virgili,
Tarragona, Spain for insightful discussions on the potential of ap-
tasensors. The author acknowledges the financial assistance of Marie
Curie Fellowship program for funding of MCFI-200001246.
References
1. Robertson and Joyce, GF (1990) Nature 344:467468
2. Ellington, AD, and Szostak, JW (1990) Nature 346:818822
3. Tuerk, C and Gold, L (1990) Science 249:505510
4. Jayasena, S (1999) Clin Chem 45(9):16281650
5. Famulok, M (1994) J Am Chem Soc 116:16981706
6. Haller, AA and Sarnow, P (1997) Proc Natl Acad Sci 94:8521
8526
7. Gebhardt, K, Shokraei, A, Babane, G, Lindquist, BH (2000)
Biochemistry 39(24):72557265
8. Wilson, DS, Keefe, AD, Szostak, JW (2001) Proc Natl Acad
Sci 98(7):37503755
9. Kawakami, J, Imanaka, H, Yokota, Y, Sugimoto, N (2001)
J Inorg Biochem 82(14):197206
10. Geiger, A, Burgstaller, P, von der Eltz, H, Roeder, A, Famulok,
M. (1996) Nucleic Acids Res 24(6):10291036
11. Burke, D and Gold, L (1997) Nucleic Acids Res 25(10):2020
2024
12. Holeman, LA, Robinson, SL, Szostak, JW and Wilson, C (1998)
Folding & Design 3:423431
13. Jeong, S, Eom, TY, Kim, SJ, Lee, SW, Yu, J (2001) Biochem
Biophys Res Comm 281(1):237243
14. Kusser, W (2000) Rev Modern Biotech 74:2738
15. Kujau, MJ and Wlfl, S (1997) Nucleic Acids Res 26:1851
1853
16. Kubik, MF, Bell, C, Fitzwater, T, Watson, SR, Tasset, DM
(1997) J.Immuno 159:259267
17. Drolet, DW, Jayasena, SD, Gold, L (1998) US Patent 5 789
163
18. Ito, Y, Fujita, S, Kawazoe, N, Imanishi, Y (1998) Anal Chem
70(16):35103512
19. Kleinjung, F, Klussmann, S, Erdmann, VA, Scheller, FW,
Frste, JP, Bier, FF (1998) Anal Chem 70:328331
20. Potyrailo, RA, Conrad, RC, Ellington, AD, Hieftje, GM (1998)
Anal Chem 70:34193425
21. Lee, M, Walt, DR (2000) Anal Biochem 282(1):142146
22. a. Hesselberth, J, Robertson, MP, Jhaveri, S, Ellington, AD
(2000) Rev Modern Biotech 74:1525 b. Jhaveri, S, Rajendran,
M, Ellington, AD (2000) Nat Biotechnol 18(12):12931297
23. a. Stojanovic, MN, de Prada, P, Landry, DW (2001) J Am
Chem Soc 123:49284931 b. Stojanovic, MN, de Prada, P,
Landry, DW (2000) J Am Chem Soc 122:1154711548
24. Tyagi, S and Kramer, FR (1996) Nat Biotechnol 14:3038
25. Yamamoto R and Kumar, PKR (2000) Genes to Cells 5:389
396
Anal Bioanal Chem (2002) 372 : 4965
DOI 10.1007/s00216-001-1191-9
REVIEW
Ying Huang Elizabeth L. Mather Janice L. Bell
Marc Madou
MEMS-based sample preparation for molecular diagnostics
Received: 24 September 2001 / Accepted: 23 October 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract Completion of the Human Genome Project is
driving the rapid development of molecular diagnostics in
the laboratory. To accelerate the penetration of genetic
tests and other nucleic acid-based tests into clinical mar-
kets, simple, compact, automatic sample-preparation sys-
tems for molecular diagnostics must be developed. Micro-
electromechanical systems (MEMS) is a promising ap-
proach for the development of automated sample prepara-
tion for the clinical laboratory or point-of-care setting.
This review discusses MEMS-based components that
could be applied to the different stages of the sample-
preparation process such as cell separation, nucleic acid
purification, and nucleic acid amplification. Examples of
functional component integration are given. Issues dis-
cussed include partitioning of functions between the in-
strument and disposable unit, methods of propulsion of
fluids and particles, vapor and liquid barriers, and sample
size. Although further evaluation and development are
needed to provide practical solutions to some of these is-
sues, we conclude that MEMS-based components might
contribute to some components in a sample-preparation
system consisting of modular instruments and disposable
units, but will not provide a generic or a totally integrated
solution.
Keywords Sample preparation MEMS Molecular
diagnostics
Introduction
Critical advances in micromachining and genomics over
the last decade have brought the two fields to the point
where synergy can lead to revolutionary changes in bio-
medical diagnostics. Completion of the Human Genome
Y. Huang E.L. Mather J.L. Bell M. Madou ()
Nanogen Inc., 10398 Pacific Center Court,
San Diego, California, USA 92121
e-mail: Mmadou@nanogen.com
Project is driving the rapid development of molecular di-
agnostics in the laboratory. Micromachined instrumenta-
tion can revolutionize the commercialization of these di-
agnostic tests by development of compact, potentially dis-
posable, automated diagnostic systems that could be used
to perform molecular diagnostics in a clinical laboratory
or a point-of-care setting. Microelectromechanical sys-
tems (MEMS) have already permeated the molecular biol-
ogy and chemistry arenas and are having a profound im-
pact on instrument and assay development. The principal
advantages associated with a BIOMEMS approach in-
clude:
1. small instrument footprint
2. multiplexing so that several analyses can be completed
simultaneously
3. microfluidics for fluid manipulation, lessening the need
for traditional macro pumps
4. potential for integrating various functional components
on to the device
5. reduced volume requirements for sample and reagents
6. unsupervised automation
In this review we discuss MEMS contributions to sample
preparation in molecular diagnostics. Specifically, we ex-
amine the feasibility of sample preparation in a small, in-
tegrated footprint instrument capable of performing all the
aspects of a common DNA assay. Like packaging in the
semiconductor industry, sample preparation is an area that
has been pursued with less vigor than development of the
assay itself; perhaps it is perceived as somewhat less
glamorous just like packaging in the IC industry. But,
like IC packaging, sample preparation is of overwhelming
importance to the end customer.
Elements of sample preparation
for molecular diagnosis
Sample types and cell separation
Preparation of biomedical samples for DNA or RNA
analysis prior to the detection step tends to be relatively
50
Fig. 1 Block diagram showing the general sample-preparation
procedure for molecular diagnostics
complex. Multiple specimen types and different types of
target cell must be processed. Currently, the types of diag-
nostic assay that use molecular techniques include genetic
analysis, cancer-cell characterization, and infectious-
agent detection and quantification. Examples of speci-
mens types processed are whole blood, serum, urine,
saliva, cells, tissues from biopsies, spinal fluid, and stool.
Not unexpectedly, each specimen type has different sam-
ple-preparation requirements. Although sample prepara-
tion of DNA or RNA targets can be moderately simple
when target molecules are abundant, in certain types of
specimen, when target copies are rare or when the speci-
men is quite complex, the preparation can involve inten-
sive manual operations utilizing several different instru-
ments. Steps might include centrifugation or other separa-
tion methods, heating and cooling, mixing with reagents,
filtration, concentration, purification, and amplification
and detection. A general scheme is shown in Fig. 1.
Whole blood is a complex mixture containing high con-
centrations of serum proteins (6080 mg mL1), red blood
cells (RBC; approximately 5109 mL1), thrombocytes
(34108 mL1), and white blood cells (WBC; 510 of normal cells, urine requires a concentration step of 80-
106 mL1). The abundant RBC have no nuclear DNA, but
contain high levels of hemoglobin that interfere with DNA
amplification and other enzymatic manipulations. Histori-
cally, RBC have been separated from WBC by density
gradient centrifugation or by hypotonic lysis of the RBC
followed by centrifugation of the white blood cells. Today
many of these procedures have been replaced by methods
that combine lysis and purification of the nucleic acids.
(see below).
Some applications require a cell-separation step. When
the sample is complex and when the target copy number is
low, enrichment and sometimes purification of the target
cells are necessary. For example to detect 100 copies of
HIV in 1 mL whole blood, 0.5 mL blood is needed. That
0.5 mL blood will also contain approximately 2.5109 RBC
and 2.55106 WBC. Detection of this small amount of
viral DNA using enzymatic amplification requires re-
moval of the blood cells and concentration of the viral
particles.
Rare cancer cells or rare fetal cells in a mothers blood
must be highly enriched prior to DNA analysis or purified
for quantitative RNA analysis. Fluorescence-activated
cell sorting (FACS) [1, 2] and magnetic-activated cell
sorting (MACS) [3] have been used for this purpose.
These procedures involve tagging the desired cells by in-
cubating the total cell pool with specific cell-surface anti-
bodies conjugated either to fluorophores or magnetic
beads. The tagged cells can then be separated from large
pools of untagged cells by FACS or MACS.
Biopsy material might also require cell separation, be-
cause the presence of normal cells can complicate the
analysis of nucleic acids from the cancerous cells. One
approach used to obtain specific cell populations from
connective tissue cells is laser capture microdissection
[4]. This method involves placing a thin, heat-sensitive
film over a section of tissue while observing the sample
microscopically, and then selectively adhering the cells of
interest to the film using a short, focused pulse from an in-
fra-red laser. Images of each sample taken can be
recorded and the film with the cells attached can then be
placed directly into a tube for subsequent analysis.
Other sample types, such as stool, require extensive
purification. Besides removing large debris, food diges-
tion products and bacterial contaminants that can inhibit
amplification steps must also be removed [5]. In addition,
the large amount of normal bacterial flora can make the
detection of disease-causing species difficult. Although
immunomagnetic separations have been used to improve
the detection of low numbers of organisms, sample prepa-
ration still requires multiple steps before amplification
[6, 7].
Other less complex specimen types might require a
concentration step only. Urine, saliva, serum, or plasma
generally fall into this category whether one is assaying
for the presence of microorganisms or for the low levels
of DNA or cells that have been reported to be present in
these fluids [8]. The concentration required varies with
specimen type and the target species. For genetic analysis
to 100-fold [9], whereas saliva requires approximately
2025-fold concentration [10]. The concentration factor
for serum can be less. For example, 200 L plasma or
serum from patients diagnosed with breast cancer was
sufficient for detection of tumor-related alterations in
DNA [11].
Among the easier specimens for genetic analysis are
cheek cells or buccal swabs. These cells are moderately
pure and can be lysed directly by fairly standard proce-
dures. From the patients and medical technologists
viewpoints, an additional advantage of cheek cells (and of
urine and saliva specimens) is that their collection does
not require an invasive procedure.

Cell lysis and nucleic acid purification
After isolating the desired cells from the raw sample, the
cells must be lysed or disrupted to release the nucleic
acids. For blood, disruption of the nucleated white cells
can be performed by heating or by the addition of lysing
reagents such as detergents. With tissues, one or more en-
zymatic steps may also be used. A variety of approaches
can then be used to purify nucleic acids after the lysis
step. Traditionally DNA and RNA were purified by diges-
tion of proteins with proteinase K in the presence of de-
tergents, extraction with phenolchloroform, and concen-
tration by precipitation with alcohol in the presence of salt
[12]. Today DNA and RNA are often purified using
chaotropic reagents such as guanidinium hydrochloride or
guanidinium thiocyanate. This purification procedure
takes advantage of the observation that in the presence of
the chaotropic reagent, DNA and RNA bind to silica and
in its absence will be released [13, 14]. Non-nucleic acid
components of the cell lysates are washed away in the
presence of the chaotropic reagent and then the purified
nucleic acid can be eluted in a small volume of water or
low-salt buffer.
The efficient lysis of many bacteria, yeast, and spores is
more difficult. Common methods for lysing gram-negative
bacteria include treating the cells with alkaline buffer, with
heat, or with guanidinium thiocyanate [15, 16]. Gram-pos-
itive bacteria are more difficult to lyse efficiently. A vari-
ety of enzymes that are specific for gram-positive cell-wall
components have often been used to assist in the lysis
process. The enzymatic step is often followed by addition
of detergent and then heating [17, 18]. Other approaches
include boiling in the presence of a Chelex-100 (Bio-Rad)
[19] or the use of the bead beater, produced by Biospec
Products, in the presence of 100 L beads and Chelex-100
[20]. Sonication has also been used to lyse bacteria and
fungi [21, 22]. It should be noted that efficient lysis of mi-
croorganisms is critical for the direct analysis of clinical
specimens, because for many of these specimens the copy
number of microorganisms is quite low.
Amplification
The last step of sample preparation is amplification of low
copy numbers of target nucleic acids to higher copy num-
bers that will fall within the limits of the detection tech-
nology used. Several amplification methods are in use to-
day. The most widely used is the enzymatic polymerase
chain reaction (PCR) [23]. PCR is usually considered ca-
pable of amplifying as few as 20 copies of a target in any
given, purified, biological sample to over hundreds of
millions of copies. It generally uses temperatures above
92 C to denature the complementary strands of DNA and
then lower temperatures to enable annealing of primers
and enzymatic extension of the primers. Generally 2535
of these cycles are run. Other amplification procedures,
such as the ligase chain reaction (LCR), requires a similar
thermal cycling procedure [24]. Some methods of target
51
amplification are isothermal, e.g. the strand displacement
amplification (SDA) developed by Becton Dickenson
[25] and the transcription-mediated assays, TMA [26] and
NASBA [27] used by Gen-Probe and Organo-Teknica, re-
spectively.
A variety of signal-amplification methods have been
employed in commercial assays. These methods do not
produce more copies of target DNA, but instead have a
mechanism for amplifying a reporter molecule when the
target is present. Two examples of this type of assay are
the isothermal enzymatic Invader assay developed by
Third Wave [28] and the sequencing by synthesis devel-
oped by Pyrosequencing AB [29].
As can be seen from the discussion above, the sample
preparation process generally requires human intervention
either during a step or between steps. In addition to the
cost and the effect on workflow, possibilities for contami-
nation and operator error increase. It would, therefore, be
desirable to develop a small-footprint instrument with
test-specific disposable cartridges that can perform all the
sample-preparation steps in an automated and integrated
fashion.
MEMS-based sample-preparation
functional components
By adapting microfabrication techniques in IC to biologi-
cal applications, a variety of miniaturized devices have
been developed as sample-preparation functional compo-
nents. They will be discussed in the following sections ac-
cording to their function in the sample-preparation process.
Cell separation
As discussed above, some applications of sample prepara-
tion will require cell separation to reduce the complexity
of the original sample. Currently, several techniques for
cell separation have been or could be used on a micro
scale.
One technique, microfiltration, utilizes the size differ-
ences among cell types. Using conventional computer-
controlled (CNC) machining, different designs and sizes
of filters such as arrays of posts, tortuous channels [30]
comb-shaped filters [31] and weir-type filters [32] have
been fabricated from glasssilicon mainly for the purpose
of efficient isolation of white blood cells from whole
blood. Recently Yuen et al. [32] demonstrated that a weir-
type filter with a 3.5 m gap could isolate white cells with
reasonable capture efficiency (7%) of white cells and
>99.9% elimination of RBC. From the white blood cells,
a 226-bp region in the human coagulation Factor V gene
was successfully amplified in a microchip-based PCR re-
action. An advantage of microfiltration is that it does not
require specific buffer conditions. It does, however, re-
quire a size difference between the desirable and undesir-
able cell types. For example, it would not effectively sep-
arate the different types of white blood cell.
52
Dielectrophoresis (DEP) [33, 34, 35], which utilizes
the interactions between the intrinsic dielectric properties
of the cells and the applied a.c. electric field, is an active
separation approach. The dielectric property of a cell is a
function of the applied a.c. electric field. Different types
of cell have different responses to the applied electric
field. Therefore, by selecting the appropriate frequency of
the a.c. electric field for a sample type one can separate
different types of cell. A DEP device can be made by fab-
ricating microelectrodes on glass or silicon substrates by
photolithography. A mixture of cells can be focused to
different regions of the microelectrodes depending on
their individual dielectric responses to the applied electric
field. Because of its simplicity and flexibility, DEP has
been widely investigated for chip-based cell separations.
Using DEP, a variety of biological cells has been sepa-
rated including bacteria [36, 37, 38, 39], cancer cells [40,
41, 42, 43], stem cells [44, 45], and leukocyte subpopula-
tions [46]. Unlike microfiltration, DEP-based separation
is not limited by the size differences between different cell
types. In principle, different types of cell can be separated
with the same electrode design by selecting the appropri-
ate frequency for each cell type. DEP also has the advan-
tages of flexibility, controllability, and ease of application
to automation, and it does not require immuno-labeling.
Compared with microfiltration, most DEP separations re-
quire low conductivity separation buffer, which means
cells cannot be directly separated from their original sam-
ple solution, though Mller et al. [43] have reported the
separation of Jurkat cells from red blood cells in phos-
phate-buffered saline. To make DEP separation more
adaptable to real samples, more effort needs to be focused
on improving the sensitivity, efficiency, and throughput.
The Evotec Cytocon [47] is an example of a commercial
microsystem for cell separation and cell sorting that is
based on dielectrophoresis.
Both hydrodynamic flow [48, 49, 50] and electroos-
motic [51] flow have been used to transport and sort cells
on microchips. Micromachined cytometers that incorpo-
rate electrokinetic focusing to confine cells spatially have
been pursued by several groups [49, 52]. Another MEMS
method for cell concentration is based on ultrasonic stand-
ing-wave technology. Using acoustic waves, cells and
particles have been concentrated from bulk solutions us-
ing either batch systems or flow systems. Coakley et al.
reported separation of cells from whole blood using ultra-
sonic cell levitation and transportation in standing acoustic
waves [53]. A particularly interesting acoustic MEMS ap-
proach to cell manipulation is the use of flexural plate
wave devices (FPW) reported by Ballantine et al. [54].
Compact planar micropumps based on FPW could be
used to transport, mix, concentrate, and even lyse cells
(also see below).
Fu et al. recently sorted Escherichia coli cells express-
ing green fluorescent protein from a background of non-
fluorescent E. coli cells by use of a disposable microfabri-
cated fluorescence-activated cell sorter (FACS) [55].
This cell sorter was made using a soft lithograph proce-
dure [56]. FACS offers advantages over conventional
FACS, because of its low cost and potentially easy inte-
gration with detectors and optical filters.
Cell disruption and DNA purification
Many mechanical and non-mechanical methods can be
used for cell disruption; we will briefly review both types
of approach and consider their relative merit for integra-
tion in microsystems.
Mechanical methods
In mechanical disruption the cell envelope is physically
broken, releasing all intracellular components into the
surrounding medium. Several types of equipment for me-
chanical cell disruption are commercially available. In a
high-pressure homogenizer, a pump forces a slurry of
sample through a restricted-orifice valve. High pressure
(up to 150 Mpa) is followed by instant expansion through
a special exiting nozzle. Cell disruption is accomplished
by three different mechanisms in this system: impinge-
ment on the valve, high liquid shear in the orifice, and
sudden pressure drop upon discharge which causes explo-
sion of the cell.
In a bead mill cells are agitated in suspension with
small abrasive particles. Cells break because of shear
forces, grinding between beads, and collisions with beads.
This approach does not seem amenable to miniaturization.
High-frequency ultrasound is another method widely
used for cell lysis. Ultrasonication devices generate in-
tense sonic pressure waves in liquid media. Under the
right conditions, the pressure waves cause formation of
microbubbles which grow and collapse violently. The im-
plosion generates a shock wave with enough energy to
break cell membranes. Modern ultrasonic processors use
piezoelectric generators made of lead zirconate titanate
crystals. The vibrations are transmitted down a titanium
metal horn or probe tuned to make the processor unit res-
onate at 1525 kHz. Disadvantages of this method are that
ultrasonic disintegrators generate considerable heat dur-
ing processing; they also generate free radicals that can
react with biomolecules. The advantages of this method
are that it is suitable for hard-to-lyse cells and spores. Ul-
trasonic lysis is an attractive approach for miniaturization
of the cell-disruption process [57].
Mechanical disruption methods have several draw-
backs. Because cells are broken completely, all intracellu-
lar materials are released. Therefore, the DNA must be
separated from a complex mixture of proteins, and mem-
brane fragments. The cell debris produced by mechanical
lysis often consists of very small fragments, making the
solution difficult to clarify. Moreover, complete product
release often requires more than one pass through the
membrane disruption device, exacerbating the problem by
further reducing the size of the fragments.
 53

Non-mechanical methods
Non-mechanical lysis methods include chemical, enzy-
matic, osmotic, thermal, and electric methods. Chemical
methods extract intracellular components from microor-
ganisms by making the outer cell wall permeable with or-
ganic solvents that create channels through the cell mem-
brane. Solvents used for this purpose include toluene, ether,
phenylethyl alcohol, DMSO, benzene, methanol, and chlo-
roform. As discussed previously, chaotropic agents, deter-
gents, and enzymes have been used for cell lysis, as have
antibiotics, thionins, and chelating agents. One aspect of
the use of these agents is that they must be removed from
the sample before subsequent enzyme-mediated amplifi-
cation. This removal step could be a drawback in an
MEMS device. Another problem is that the amount of
lysing solution required might preclude on-chip storage.
Despite these drawbacks, Tian et al. have built and char-
acterized a miniaturized purification device that makes
use of silica resins and chaotropic agents to purify ge-
nomic DNA for subsequent PCR amplification [58].
Although cells exposed to slowly varying extracellular
osmotic pressure can usually adapt to such changes, cells
exposed to rapid changes in external osmolarity can be
mechanically injured. This procedure is typically con-
ducted by first letting the cells equilibrate internal and ex-
ternal osmotic pressure in a high sucrose medium, and
then rapidly diluting away the sucrose. The immediate
overpressure of the cytosol is assumed to damage the cell
membrane.
Repeated freezing and thawing or elevated temperature
will also lyse mammalian cells. In some applications the
target DNA can be released by elevated temperatures si-
multaneously with the denaturing step performed in PCR
[59]. These thermal methods are not, however, suitable
for samples that contain low levels of bacteria or hard-to-
lyse spores.
Cells can also be lysed electrically on microelectrode
arrays after pulsed high-electric-field treatment [38, 60].
In particular, E. coli lysate obtained by electric lysis has
been reported to be detected by electric-field-driven hy-
bridization on microelectronic chips [38].
Of the different types of non mechanical lysis, the elec-
tric and thermal approaches are the most amenable to
miniaturization. The chemical approaches might be ranked
according to the amount of chemicals required (the lower
the amount the more suited the approach to miniaturiza-
tion) as well as how critical it is to remove them. A possi-
ble solution when large volumes of lysing solution are re-
quired is to use an external storage bottle and implement a
flow-through system. Table 1 summarizes some of the ly-
sis methods, purification, and recovery steps.
Amplification
Most reports on miniaturized DNA amplification have fo-
cused on developing a micro PCR device using IC and
MEMS technologies. The first miniaturized DNA ampli-
fier reported by Northrup et al. [61] in 1993 had polysili-
con heaters patterned on a bottom Si3N4 membrane and
was sealed from the top by a glass slide. Since then, many
studies have been performed to enhance the design of the
microreaction chambers [62, 63, 64, 65, 66, 67, 68, 69,
70]. Efforts have been made to improve temperature con-
trol, enhance temperature ramping rates, and reduce
power consumption by exploring different materials such
as silicon [71], glass [72], and polyimide [73] and differ-
ent methods of heating such as Peltier thermoelectric ele-
ments [63], IR heating with a tungsten lamp [72], and so-

Table 1 Cell lysis, DNA purification, and recovery

Technical Cost Efficiency Ease of miniaturization
difficulty

Lysis method
Mechanical High High High Of the mechanical lysis methods reviewed,
ultrasonic is most amenable.
Chaotropic Medium Medium Medium Depends strongly on the volume of reagents involved.
Can be incorporated into a membrane, facilitating removal.
Thermal Low Low Low Easy to miniaturize, but not suitable for all cell types
Electronic Low Low Low Easy to miniaturize
Purification method
Ion exchange Medium Medium High Easy to implement on a miniaturized platform
Beads Medium Medium Medium Somewhat more cumbersome
Molecular weight cut-off Low Low Low Easy to implement on a miniaturized platform
(MWCO) membranes
Recovery method
Silica elution Low Low High If a flow-through system is used miniaturization
of low-salt buffer is easy to accommodate
Beads elution Low Low Medium More cumbersome
Membrane elution Low Low Low Easiest to miniaturize
54

Table 2 Prototype integrated DNA analysis devices

Sample preparation Biochemical reactions Detection Fabrication methods Ref.
for device

RNA purification from Reverse transcription-PCR, GeneChip CNC machining Affymetrix

100 mL of serum spiked DNase fragmentation, de- hybridization [86]
with different amounts phosphorylation, terminal
of a partial HIV sequence transferase labeling
White blood cells isolated PCR using integrated heater- N/A CNC machining (standard [32]
from <3 mL of human cooler to provide thermal microfabrication tech-
whole blood cycling niques for microchips)
NaOH lysis of whole blood PCR using thermo- Detection is per- CNC machining of poly- Tecan
or E. coli suspension electric devices on a formed separately carbonate or PMMA disc Boston [91]
spinning PC board platen by electrophoresis and lithography in PDMS
for smaller fluidic features
DEP collection of E. coli Strand displacement Electric field driven Laminated flexible sub- Nanogen
in a 70 mL sample amplification, Electronic DNA assay strates with electrode arrays [92]
DNA hybridization using IC lithography
Spores in paper products PCR on material Gel electrophoresis Acrylic or polyester Cepheid
(0.5 g in 10 mL water), eluted from a Si chip. in a separate step plastics and Si chips [94]
DNA purified on Si chip
(200 mm high columns,
18 mm in diameter and
a pitch of 34 mm)

Expanded from Ref.[32]

lution resistive heating [74]. Rapid thermal cycling of ap-

proximately 17 s cycle1 in a silicon microstructure [75]
and 16 s cycle1 in a polyimide structure has been demon-
strated [72]. Reaction volumes of several microliters [76]
or even picoliters [77] have been used to amplify a variety
of genes in these micro PCR devices. Multiple-sample
PCR amplification of four different targets has also been
achieved on a microchip [78].
Another approach used to perform multiple PCR reac-
tions on a microchip is to immobilize the individual sets
of primers within a polyacrylamide gel matrix poured on
to a glass slide. By use of this approach, mutations re-
sponsible for drug-resistance in three genes of M. tuber-
culosis were simultaneously identified [79, 80, 81].
The strand-displacement assay (SDA) is an alternative
DNA amplification method that has the advantage of be-
ing an isothermal reaction. This method has recently been
modified to improve its ability to amplify multiple targets
in a single reaction vesicle. This modified approach, an-
chored SDA, involves electrophoretically localizing sets
of primers on different electrodes on a microchip, using
an electric field to facilitate hybridization of target DNA
to the localized primers, and then allowing the SDA reac-
tion to proceed. By use of this approach five different an-
timicrobial resistance and bacterial identification genes
have been amplified and detected on a microelectric chip
[82, 83]. More recently 10 human genes were amplified
simultaneously on a microelectronic chip [84].
Examples of MEMS-based integrated
sample preparation
The previous section highlights individual MEMS func-
tional components for sample preparation. Here we will
review some of the attempts at integrating two or more of
these components in an integrated system. Over the last
five years, several attempts have been made to miniatur-
ize molecular diagnostics, inclusive of sample prepara-
tion, in the areas of biological warfare agents and point-
of-care [72, 76, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94].
Some of these efforts involve microfluidic systems, in
which the entire procedure from sample collection
through PCR, and sometimes, all the way to detection, is
embedded in one integrated instrument. Integration of cell
separation and PCR has been demonstrated on a weir-type
microfilter [32]. E. coli bacteria have been separated and
electrically lysed on a single microelectronic chip [38].
PCR amplification and capillary electrophoretic (CE) analy-
sis have been integrated on a microchip [70, 75, 95, 96].
A miniature analytical thermal cycling instrument has
been developed to amplify and detect DNA via PCR in
real-time [71]. Using this system, bacteria could be de-
tected in 7 min [59].
In Table 2, which has been expanded from Ref. [32],
we review some of these prototypes of MEMS-based inte-
grated systems and single out the Tecan, Nanogen, and
Cepheid approaches for more detailed discussion.
In the Tecan approach, the LabCD is used as a sam-
ple-preparation stage. The prototype combines sample
preparation and PCR and detection is performed sepa-
rately on an ethidium-bromide stained gel electrophoresis
slab. As shown in Fig. 2, the disposable disk consists of
 55

Fig. 2 LabCD for PCR (Tecan). Left: Top view of one area of 3. DNA amplification by strand-displacement amplifica-
the CD. The center of the disc would be in the upper left and the tion (SDA);
outer edge at the bottom. The elements are: (a) sample, (b) NaOH
for lysing cells, (c) tris-HCl from neutralizing the lysate, (d) capil- 4. desalting and histidine-buffer exchange;
lary valves, (e) mixing channels, (f) lysis chamber, (g) tris-HCl 5. denaturation of amplified/desalted amplicon; and
holding chamber, (h) neutralized lysate holding chamber, (i) PCR 6. DNA hybridization assay.
reagents, (j) thermal cycling chamber, and (k) air gap. Fluids
loaded in (a), (b), and (c) are driven at a first RPM into reservoirs At the core of this instrument is a small two-level-stacked
(g) and (f), at which time (g) is heated to 95 C. The RPM is in- microlaboratory that is 7676 mm2 and 2.77 mm thick.
creased and the fluids are driven into (h). The RPM is increased The stacked structure is constructed from a set of lami-
and fluids in (h) and (i) flow into (j). Right: The cross section
shows the disc body (m), air gap (k), sealing layers (n), heat-sink nated flexible substrates with fluidic cutouts, pressure-
(l), thermoelectric (p), PC-board (q), and thermistor (o). Reprinted sensitive adhesive layers, electrode arrays, and two Si
with kind permission from Kluwer Academic Press [91] chips. The completed stacked structure has the DEP chip
in the upper level chamber for dielectrophoretic collection
of bioparticles and the 1600-site chip in the lower level
for performing automated electric-field-driven assays
chambers and channels. These chambers and channels
(Fig. 3).
form the sample and reagent storage reservoirs, fluidic
The stacked structure was integrated into a fluidic sys-
controlling valves, fluidic mixing chambers, and heating
tem to form a prototype instrument (Fig. 4). In addition to
chambers. By rotating the disk at different speeds, cells
the stacked microstructure, the instrument also features
(bacteria or blood) and reagents can be sequentially moved
separate modules for isothermal SDA, desalting, and DNA
to different compartments to accomplish most sample-
denaturation to support DNA hybridization assays in the
preparation steps from cell lysis to DNA purification and
stacked structure. The SDA module is 76298.5 mm2 and
finally to the PCR amplification chamber. A centrifuge
consists of a denaturation chamber (Teflon tubing snaked
platform is an attractive sample-preparation stage because
through a channel in an aluminium heating block) and an
of the wide dynamic range and relative insensitivity to
amplification chamber (a polystyrene micro well in a sec-
fluid composition. It is also believed that the throughput
ond aluminium heating block) (Fig. 4).
could be augmented by placing multiple structures of the
Because electric-field-facilitated DNA hybridization
type shown in Fig. 2 on one plastic disc.
assays are performed in a low-salt environment, a desalt-
One major issue associated with this platform is yet to
ing module was designed to remove salts used in DNA
be resolved. Unless one works with all-dry reagents and
amplification and, simultaneously, to mix the amplified
reconstitutes them upon first use of the disk, the liquids
DNA with histidine buffer used in the DNA assay. The
held in the sample reservoirs will evaporate and distribute
2-stage desalting module (255776 mm3) consists of
as time passes, because there is no vapor barrier to stop
two chambers; salts are removed with ion-exchange resin
them from expanding into the open volume of the disk.
in the first chamber (BioRad) and histidine is exchanged
The fluid gating in the disk only works for fluids, not for
in the second. The amplified DNA is pumped sequentially
vapors.
through a hollow fiber (30,000 molecular weight cutoff,
At Nanogen, a sample-preparation prototype instru-
A/G Technology) coiled in each chamber to enable maxi-
ment was developed to perform both DNA hybridization
mum surface area for diffusion. Finally, the desalted DNA
assays and protein immunoassays from mixed samples
is denatured and pumped back to the stacked structure for
[92]. Sample preparation starts on one Si chip where di-
the DNA hybridization assay and fluorescent detection.
electrophoresis separates and concentrates intact E. coli
One sample-to-answer instrument that is beyond the
and/or protein toxins in the sample. After the dielec-
prototype stage has been developed by Cepheid. The
trophoresis step protein antigens are pumped to a second
GeneXpert instrument was introduced for pathogen detec-
Si chip for immunoassays. For the DNA assay the inte-
tion in 2001 [97]. This instrument is reported to process a
grated protocol consists of six steps:
raw biological sample and provide PCR results for the sam-
1. collection of E. coli by dielectrophoresis; ple in 30 min. Sample preparation occurs in disposable,
2. lysis of the bacteria and denaturation; single-use fluidic cartridges that separate and concentrate
56
Fig. 3 Stacked microlaboratory: (a) Fabrication of the stacked
structure 7676 mm2 square and 2.77 mm thick. The first patterned
Kapton layer features a flip-chip bonded CMOS chip (78 mm2
with 4040 electrodes) for DNA hybridization assays or im-
munoassays. Layers 2, 4, and 6 are double-sided pressure-sensitive
acrylic adhesive (PSA) layers (0.152 mm thickness per PSA layer),
some with fluidic cutouts. The third layer is an optically transpar-
ent poly(methyl methacrylate) (PMMA) film (0.178 mm thick) put
over the fluidic cutout to make a first enclosed channel/chamber.
The next patterned Kapton layer (5) holds the flip-chip-mounted
dielectrophoretic (DEP) chip (11 cm2 with sixteen large and nine
small platinum electrodes). Layer 7 is a glass plate (1.27 mm)
which encloses a second fluidic chamber. (b) Completed stacked
structure showing top and bottom views. Electrode patterns at the
opposite edges of the stacked laboratory enable electrical contact
to an external control circuit; fluidic connections are located on the
bottom of the stacked structure [92]
cells by means of filters. Cells are lysed using ultrasonic
techniques, and DNA or RNA are enriched and purified.
The nucleic acids and reagent mixes for PCR reactions are
delivered into a thin-walled disposable reaction chambers
that have an optical window suitable for real-time detec-
tion of PCR products. This chamber fits into the ICORE
module (ICORE for intelligent, cooling/heating, optical
reaction; Fig. 5). This module uses silicon-based, high-ef-
ficiency reaction chambers with integrated heaters and
simple, inexpensive electronics to control the reaction
temperatures. Four-color detection within the module en-
ables real-time identification and quantification of four
different targets in each reaction chamber. Multiple ICORE
modules can be loaded and run at same time in the
GeneXpert instrument. In contrast with traditional thermal
cycling systems in which all samples are subjected to the
same time/temperature/optical procedure, each I-CORE
unit can be operated and controlled independently en-
abling the user to run different cycling procedures.
Issues associated with MEMS-based
sample preparation
Partitioning of functions between instrument
and disposable unit
Microfluidic systems comprising nozzles, pumps, chan-
nels, reservoirs, columns, mixers, valves, etc. can be used
for a variety of sample-preparation applications. The ad-
vantages of these devices include lower fabrication costs,
enhancement of analytical performance, lower power re-

Fig. 4 Nanogens integrated fluidic system for the stacked mi-
crostructure. For a complete automated assay of the SLTI gene,
70 L E. coli suspension (109 mL1) is introduced into the fluidic
system through the micro-splitter valve (1) and delivered to the
DEP chip on the stacked structure (2) held in a clamping fixture.
For capturing the E. coli, an ac voltage of 5 kHz, 10 V peak to peak
is employed (52 min). With the ac voltage on, the chip is washed
with 50 mmol L1 histidine from the storage bottles (3). The ac
voltage is then turned off to release collected bacteria and the bac-
teria are mixed with SDA reagents and delivered to the SDA mod-
ule for bacterial lysis, DNA denaturation (4) and amplification of
SLT1 DNA (5). The DNA amplification product is desalted, ex-
changed with buffer (6) and denatured (7) for the DNA hybridiza-
tion assay back in the stacked microstructure (8). The hybridized
SLT1 gene is detected with a fluorescence-labeled DNA reporter
and viewed from the microscope lens above the stacked structure
(9) [92]
quirement, and lower consumption of chemicals than tra-
ditional systems [98]. The generic technical challenges in
developing a sample-preparation instrument based on
those MEMS components include the partitioning of the
different instrument functions such as power supply,
mechanism for getting the sample into and out of the in-
strument, sample concentration, providing fresh viable
reagents, and detection. At one extreme one might envi-
sion a totally integrated option i.e. a micro total analysis
system (-TAS). In such a monolithic MEMS approach,
electronics and MEMS elements are co-fabricated within
one single, lengthy sequential silicon process. In a hybrid
MEMS approach, the electronic parts are kept separate
from the sensor part and both Si and non-Si machining
processes are used. The device components in such a hy-
brid remain more serviceable, and because the packaging
and interconnects determine the final size of the instru-
ment, hybrids are often not much larger than their coun-
terparts based on a more integrated Si solution. Hybrids
are usually simpler to design and can be produced more
economically in smaller production volume runs. For
chemical and biosensor based instruments, they often rep-
resent the most viable means of implementing microma-
chining technology.
Partitioning is highly application-specific. Depending
on the application, e.g. clinical diagnostics or high-
57
throughput drug screening, the electronics, the power for
the propulsion mechanism, heaters, etc., can be integrated
with the disposable unit or be part of the instrument. In
high-throughput screening fluid handling is more sophis-
ticated and is typically incorporated into a large, table-top,
costly research instrument. In this case, the disposable
units for screening are less challenging than those for di-
agnostic application, because there is no need to contain
sealed fluid or dry stored chemicals on the units. For di-
agnostics the disposable unit should be inexpensive and
should be as uncomplicated as possible perhaps only
reagents, plastic substrate, and fluidic conduits. The de-
tection device should be inexpensive, small, preferably
hand-held, and user-friendly. A major challenge for diag-
nostic applications is to determine how to store wet or dry
chemicals in sealed reservoirs within the disposable pack-
age.
Choice of fluid and particle-propulsion method
A variety of technologies can be used to move small
quantities of fluids or suspended particles in microfluidic
channels and networks. Several standard methods have
been used to accomplish fluid transport including syringe
and peristaltic pumping and more sophisticated methods
based on electrochemical bubble generation, acoustics,
magnetics, DC and AC electrokinetics, and centripetal
forces. In Table 3 we compare characteristics of a number
of MEMS fluid propulsion methods [99]. Only the sim-
plest and least expensive pumping schemes should be in-
tegrated within the disposable unit. Here we review vari-
ous propulsion techniques with regard to their applicabil-
ity to sample preparation.
Mechanical pumps
The pressure that mechanical pumps must generate to pro-
pel fluids through capillaries becomes higher the narrower
the fluidic channels. With piezoelectric (acoustic), elec-
58
Fig. 5 Sample preparation and optical
real-time PCR (Cepheid) in ICORE
cartridge. Adapted from Ref. [97]
tro-osmotic (DC electrokinetic), electro-wetting (DC elec-
trokinetic), and electro-hydrodynamic-based pumping (AC
electrokinetic), on the other hand, pumping efficiency im-
proves with smaller capillary diameter. In other words
most non-mechanical pumping mechanisms scale more
favorably in the micro domain. Despite this, almost all
modern biotechnology equipment is based on traditional
external syringe or peristaltic pumps. Micromachined me-
chanical pumps have not achieved commercial success.
Although integrated micromachined pumps based on two
one-way valves can achieve precise flow control of the or-
der of 1 L min1 with fast response, high sensitivity, and
negligible dead volume, these pumps require complicated
fabrication processes, generate only modest flow rates and
low pressures, and consume a large amount of chip area
and a substantial amount of power. As a consequence of
the low pumping pressure, they often become useless
when the device dimensions are reduced and the resis-
tance to fluid flow and required driving pressure increase.
An example of a Si micromachined mechanical pump is
that from Debiotech for insulin infusion. This belt-worn
drug-delivery pump features a piezoelectric design pio-
neered at the University of Twente [100]. The size of most
Si micromachined pumps is quite large and because cost
advantages in Si manufacture are realized with small foot-
print devices only, plastic micro molding technology is
being explored as a viable alternative [101, 102].
Another example of mechanical fluid propulsion is
embodied in the Johnson and Johnson blister pouch HIV
test shown in Fig. 6 [103]. An automated mechanical
roller, similar to that used in credit card image transfer, is
used to squeeze a plastic pouch and move the reagents
through the miniature channels in it. This approach meets
the goal for a low-cost diagnostic disposable unit contain-
 59

Table 3 Comparison of fluid- and particle-propulsion methods in microfluidics

Fluid propulsion Valving Maturity Propulsion force Power source Materials Sample Flow rate
mechanism solved influenced by preparation
suitability

Centripetal force Yes for R&D Density, wetting, Rotary motor Plastics Good From
liquids level contact angle, <1 nL s1 to
and viscosity >100 mL s 1
Pressure (e.g. Yes for Product Generic Mechanical Plastics Good 20500 mL s 1
blister pouch) liquids level roller (Fig.6)
and vapor
Acoustic No R&D Generic 5 to 40 V Piezoelectrics High 20 mL s 1
level (peak to peak)
Electrokinetic Yes for Product pH, ionic strength 10 kV Glass, plastics Low 0.0011 mL s 1
(osmosis) liquids level
Electro-wetting Yes for R&D pH, ionic strength 110 kV Glass Low 0.3 mL s 1
liquids level
Dielectrophoresis No R&D Dielectric property 110 V Glass, silicon Good <100 mL s 1

Expanded from Ref. [99]

fluidic network in the pouch. The Johnson and Johnson

approach is a good example of clever partitioning and a
more integrated Si approach will have a tough time com-
peting in cost with this design. A disadvantage of the
pouch approach is that it is rather difficult to miniaturize
such pressure-driven systems further. The disposable unit
cartridge in the i-STAT blood analyzer device also makes
use of a blister pouch, which in this case also contains ma-
terials for calibration [104].
Mechanical methods are well suited for incorporation
into a sample preparation platform where the size of the
fluidic channels is not on the microscale. With decreasing
channel diameter, however, particle clogging can become
a serious problem with this approach. In this respect,
acoustic pumping (see below and Refs [53, 54]) and di-
electrophoresis, in which liquids or particles can be moved
on a substrate rather than in narrow channels, may offer
an advantage.

Acoustic pumping

Acoustic pumping is a constant (DC) fluid motion in-

Fig. 6 Blister pouch fluidics in diagnostics (J&J [103])
ing all the reagents and fluidics. The wet reagents are
cleverly stored in separate compartments within the plas-
tic pouch; small nips in the plastic close off these cham-
bers preventing liquid or vapor from migrating into the
duced by an oscillating sound field at a solidliquid inter-
face. The transducers need not be integrated with the dis-
posable unit although they can be. In a modular approach
the disposable unit is simply laid on top of the acoustic
pump network which is part of the reader instrument.
Acoustic pumping is well suited for mixing reagents [105]
and for channel-less particle concentration and separation
[53]. This method is, moreover, insensitive to the chemi-
cal nature of the sample fluids. Although the method is
still at the research stage and is considerably more com-
plex to implement than electro-osmotic pumping (see be-
low), the characteristics listed above make it a potentially
viable approach for cell separation and concentration.
60
Electrophoresis and electroosmosis
Electrokinetic pumping in a capillary is a method that
does not involve moving parts. The fluid is driven and
guided along the capillary by electrokinetic pressure a
combination of electrophoresis and electroosmosis in
which DNA molecules are moved by their charges and
fluid is moved by interaction with capillary wall surfaces
under the influence of an electrical potential applied along
the channel length. Typical electro-osmotic flow veloci-
ties are of the order of 1 mm s1 in a 1200 V cm1 applied
electric field. For example, Jorgenson et al. reported elec-
tro-osmotic flow of 1.7 mm s1 in free-flow capillary elec-
trophoresis [106]; this is fast enough for most analytical
purposes. The disadvantages of electroosmosis are the
high voltage needed (130 kV power supply), the direct
electrical-to-fluid contact, and the flow rate sensitivity to
the capillary wall charge, to the ionic strength, and to the
pH of the solution. It is, consequently, more difficult to
make it into a generic propulsion method. For example,
liquids with high ionic strength suffer from excessive
Joule heating and it is therefore difficult or impossible to
pump biological samples such as blood and urine.
Centrifugal pumping
Using a rotating disk, centrifugal pumping can provide
flow rates ranging from less than 10 nL s1 to greater than
100 L s1 depending on disk geometry, rotational rate
(RPM), and fluid properties [91]. Pumping is relatively in-
sensitive to chemical properties such as pH, ionic strength,
or chemical composition. Aqueous solutions, solvents (e.g.
DMSO), surfactants, and biological fluids (blood, milk,
urine) have all been pumped successfully. Fluid gating is
accomplished by using capillary valves in which capil-
lary forces retain fluids at an enlargement in a channel un-
til rotationally induced pressure is sufficient to overcome
the capillary pressure at the so-called burst frequency.
Compared with electrokinetic pumping, the ability to
move a variety of liquid types and volumes makes the
centripetal platform amenable to sample preparation tasks.
Vacuum pressure reservoir
Fluid transport can also be accomplished by employing
pressure difference between the sample, reagent reser-
voirs, and other chambers. Valves are required to separate
the evacuated or pressurized chambers from the rest of the
fluid path until fluid motion is required. Opening a valve
initiates the flow of fluid toward an evacuated chamber or
away from a pressurized chamber. The drawbacks of this
mechanism are that it can only be actuated once, it can be
difficult to implement efficient valving, and reliable man-
ufacture of pressurized or evacuated chambers can be
challenging. An advantage over the previously described
methods is that it is a straightforward means of generating
a large fluid-motive force high flow rates and high pres-
sures can be obtained. Another advantage is that the fluid
propulsion force does not depend on the type of sample.
These are attractive features for sample preparation.
Vapor and liquid barriers for storage of liquids
One major issue to be resolved with MEMS-based dispos-
able units is the incorporation of a fluidic network with
on-board reservoirs containing wet chemicals. If no vapor
barrier stops them from expanding into the available open
spaces, liquids held in reservoirs on a disposable unit can
evaporate and distribute slowly throughout the fluidic net-
work. For the disposable unit illustrated in Fig. 6, reagents
of interest are sealed in blister pouches in a plastic lami-
nate and small nips in the plastic channels seal the liquid
and vapor. Upon use, a mechanical roller breaks the liquid
and vapor seals. The nips form one-time-use valves only;
when the plastic nips are ruptured they cannot be closed
again. In most disposable diagnostic applications such
one-time-use valves will suffice. Most of the MEMS-
based diagnostic platforms proposed in the literature do
not provide a vapor barrier between channels and reser-
voirs and will not work in a practical diagnostic setting
where shelf lives of six months or longer are desirable. To
circumvent the vapor-valving problem one might consider
dry-storage of reagents. These reagents can be wetted by
the sample itself upon use of the diagnostic kit. Depend-
ing on the response time required and the specific
chemistries involved, however, this method might not be
suitable for all biological applications.
Sample-size issues
Sample sizes
A major difference between traditional clinical assays and
molecular assays is the amount of target material in a
sample. Chemical analytes are often present at between
1014 and 1020 entities mL1. Thus typical clinical chem-
istry assays can be readily performed with very small sam-
ple volumes, in the picoliter and microliter range. Typical
immunological assays are used to detect antigens present
at between 107 and 1018 species mL1. In contrast DNA
and RNA assays are used to detect far fewer targets
from less than 1 to 107 species mL1 (Fig. 7). These sen-
sitivity requirements set fundamental physical limitations
on the minimum quantities of the specimen analyzed.
Table 4 shows the minimum volumes necessary for
statistically valid results for several targets assayed by im-
munological assays and for some targets that could be as-
sayed by nucleic acid techniques. Despite the tremendous
advantage gained by being able to amplify nucleic acids
by factors of >108, when rare cells are sought, there are
still major issues to be overcome in terms of reducing
sample volumes to workable levels in microdevices.
 61
Fig. 7 Scaling of sample size in mole-
cular diagnostics. Reprinted from [90]

Table 4 Minimum amount of sample need for statistically valid sample with minimum dead volumes is one of the greatest
results assuming near-perfect efficiency of capture and detection challenges in micromachining. Very few machining tech-
Analyte Copies mL1 Type of assay Amount niques can provide the flexibility to cover the dynamic
blood of blood size-range required. Because the sample introduction
needed method can place important constraints on the manufac-
turing process, a few sample-introduction options are con-
Myoglobin ~21012 Immunoassay 25 fL
Troponin I or T ~2109 Immunoassay 25 pL
sidered here. Traditional sample-injection systems such as
Specific antibodies ~1108 Immunoassay 500 pL
flow-injection analysis (FIA) rely on some sort of com-
Single copy gene ~1107 DNA 5 nL mutating valve, either rotary or sliding, actuated manually
Virus such as HIV 5100 RNA 0.510 mL or mechanically. Harrow et al. [107], compared different
Blood infections <110 DNA 6>60 mL commercially available sample-injection systems and
Shoji et al. [108] were among the first to micromachine a
sample injector consisting of two three-way micro valves
Sample-introduction choices and a channel. These valve systems are, however, too
complex and expensive to consider integrating them in a
The exact mechanism for introducing samples into a mi- disposable unit. Perhaps the simplest way to introduce the
croinstrument depends on the sample type. Pretreatment sample into a micro-instrument is by creating wells into
of the sample might be required and relatively bulky sam- which the sample is dropped. For example, most of the
ples, often collected manually, must be introduced into the substrates with electro-osmotic pumps have plastic tubes
microscale channels of the device. Introduction of the glued to them to make sample reservoirs. The Biotrack
62
Fig. 8 (a) SEM photograph of
a fluidic coupler with a 110-m
thick sleeve around the bore to
prevent blocking of capillaries
with adhesive. The silicon is
partially cleaved to show the
sleeve. (b) SEM photograph
of capillaries inserted into the
sleeve couplers. Courtesy Pro-
fessor G. Kovacs, Stanford.
Reprinted from [110]
blood-coagulation device requires that a drop of sample
be placed in a depression on the instrument surface [109].
The sample is then transferred into an internal chamber by
capillary action through a wick. Problems with sample
wells include the large volume that they require, or the
need to align the sample accurately with the well if the
sample well is small. Because the wells are open, one
risks sample contamination or inclusion of air bubbles in
the sample. The latter are a major problem in microfluidic
systems, because of the associated surface tension, and
the addition of a bulky de-bubbler is not a suitable solu-
tion in microfluidic systems. Samples can also be injected
directly into the micro instrument or disposable cassette.
An example of this is the Johnson and Johnson pouch
HIV test in which an injection septum is provided in the
disposable pouch (Fig. 6). This reduces the chance of sam-
ple contamination but again requires exact alignment of
the injection needle with the sample-injection port. Manu-
facture of the pouch is, furthermore, complicated by the
need for an injection septum. In an alternate approach, the
collection device could be disposable. For example, the
diagnostic device could be manufactured in the form of a
blood-collection tube (Vacutainer, BectonDickinson), a
syringe, a pipette, or a swab. This would, however, con-
siderably complicate manufacture of the disposable unit
and might limit the design to a specific type.
Interfacing fluidic platforms with the outside world is
starting to be addressed by micromachinists. For example,
Gray et al. [110] have demonstrated different types of
minimum dead-space fluidic couplers for standard fused
silica tubing to MEMS devices. An example is shown in
Fig. 8.
Choice of manufacturing techniques for disposable units
Biological samples for molecular diagnostics are often
sampled at volumes that are not always compatible with
the microfluidic environment of MEMS devices. Although
silicon-based MEMS detectors, fluid propulsion systems,
and actuators have been successfully utilized and demon-
strated in a number of applications, interfacing MEMS
detection devices with sample acquisition remains a major
challenge in designing practical devices for molecular di-
agnostics. The principle advantages of Si-based manufac-
turing methods for making micro-instruments are that they
offer fabrication of intricate structures with very fine fea-
tures; thus heaters, electroosmosis electrodes, and acoustic
streaming pumps can be readily realized. On the other
hand, non-traditional microfabrication from plastic mate-
rials is continually improving and is becoming competi-
tive compared with silicon micromachining. If larger
structures are built, those methods could result in lower
manufacturing cost. Because of a broad range of applica-
tions and design requirements for each particular assay
and tremendous advancements in microfabrication tech-
nologies, it is difficult to provide general recommenda-
tions about which technology should be used to fabricate
a disposable unit. It is likely that both microfabrication
approaches would be used in a device. The cost of manu-
facturing the disposable unit will dictate the choice of
manufacturing technology and the partitioning of instru-
ment function.
Conclusions
The wide range of sample types and sizes in molecular di-
agnostics makes it very unlikely that a universal sample-
preparation solution can be developed. In general, a sam-
ple-preparation system should consist of a modular instru-
ment and disposable units. When designing a sample-
preparation system several considerations should be taken
into account.
1. Partitioning of functions between the instrument and
the disposable units are very application specific. For
molecular diagnostics, a modular instrument approach
is appropriate in order to have the capability to handle

a small or large number of samples of different sizes
and types and to perform different tests on the same or
multiple samples. The design of the disposable unit
should be simple and the manufacturing cost should be
low. A good balance between traditional machined
parts and micromachining of selected components
seems to be the optimal approach.
2. The method for fluid propulsion. The most popular
method based on mechanical pumps might not be reli-
able for microchannels, because of clogging. Other mi-
crofluidic methods, e.g. electrokinetic pumping, that
depend on the sample composition, pH, and ionic
strength could be restricted to particular applications
where those variables can be maintained constant.
New propulsion mechanisms, e.g. those based on rotat-
ing disks offer attractive alternatives for more general
sample-preparation schemes.
3. Particle manipulation. To manipulate target particles
such as cells in a raw sample open-pool methods such
as dielectrophoresis or acoustic methods have obvious
advantages over methods that rely on filters or narrow
channels. The open-pool methods might, however,
lack the selectivity or the speed required in different
molecular diagnostic applications. A combination of
these methods with other standard separation methods
might be a feasible means of satisfying the design re-
quirements of a particular sample-preparation system.
4. Integrated temperature control. Heating and cooling of
samples is preferably incorporated into the instrument
rather than into the disposable unit. When dealing with
small amounts of samples, however, integration of
miniaturized heating and cooling components can be
of real benefit in instrument design in terms of in-
creased speed and reduced power consumption.
5. There is a compromise between miniaturization and
a sufficient number of targets. A difficult issue in de-
veloping miniaturized devices for sample preparation
is to obtain sufficient target species for detection. For
some applications a sample-concentration step must
be incorporated into the sample-preparation step to
compensate for the small volumes handled by the de-
vice.
6. Fluidics storage. Most MEMS engineers have devel-
oped fluidics geared towards high throughput screen-
ing of drugs, in which case on-board storage of fluids
and cost is less of an issue. As pointed out, a major re-
maining challenge for developers of diagnostic plat-
forms based on microfluidics is the implementation of
both fluid and vapor gating. The vapor and fluid valv-
ing for long-term storage has not received the attention
it deserves. Although there are some practical solu-
tions, such as sealing the reagents in small blister
pouches with a plastic laminate, new principles and/or
devices with efficient fluid/vapor gating are needed.
7. Sample introduction. Equally significant is the need
for a proper interfacing technique to introduce the
sample into a disposable unit and the fluidic interface
between the instrument housing and the disposable
unit. This issue is starting to be addressed in the
63
MEMS community and will provide major market ad-
vantages to those who provide efficient solutions to
these problems.
Acknowledgements. The authors would like to thank Drs Dalibor
Hodko and Jim Zoval for critical reading of the manuscript and
helpful discussions.
References
1. Shapiro HM (1995) Practical flow cytometry. WileyLiss,
New York
2. Melamed MR, Lindmo T, Mendelsohn ML (1990) Cytometry
and sorting. WileyLiss, New York
3. Miltenyi S, Mller W, Weichel W, Radbruch W (1990) Cyto-
metry 11:231238
4. Emmert-Buck MR, Bonner RF, Smith PD, Chuaqui RF,
Zhuang Z, Goldstein SR, Weiss RA, Liotta LA (1996) Sci-
ence 274:9981001
5. Atkin W, Martin JP (2001) J Nat Cancer Inst 93:11
6. Nilsson HO, Aleljung P, Nilsson I, Tyszkiewicz Y, Wadstrom
T (1996) J Microbiol Methods 27:7379
7. Nunez YO, Fernandez MA, Torres-Nunez D, et al. (2001) Am
J Trop Med Hyg. 64:293297
8. Anker P, Mulcahy H, Chen XQ, Stroun M (1999) Cancer
Metastasis Rev 18:6573
9. Haufroid VA, Knoops CB, Bernard A, Lison D (1999) Clin
Chem 44:2210
10. Van Schie RCAA, Wilson ME (1997) J Immunol Methods
208:9192
11. Chen XQ, Bonnefoi H, Diebold-Berger S, Lyautey J, Leder-
rey C, Faltin-Traub E, Stroun M, Anker P (1999) Clin Cancer
Res 5:22972303
12. Sambrook J, Fritsch EF, Maniatis T (1989) Molecular
cloning, 2nd edn.
13. Chirgwin JM, Przybyla AE, MacDonald RJ, Rutter WJ (1979)
Biochemistry 18:52945299
14. Boom R, Sol CJ, Salimans MM, Jansen CL, Wertheim-van
Dillen PM, Van der Noordaa J (1990) J Clin Microbiol Mar
28:495503
15. Pitcher DG, Saunders NA, Owen RJ (1980) Letters Appl Mi-
crobiol 8:151156
16. Dutka-Malen S, Leclerc R, Coutant J, Courvalin P (1990) An-
timicrob Agents Chemother 34:18751879
17. Jordan JA, Durso MB (2000) J Clin Microbiol 38:25742578
18. Martineau F, Picard FJ, Menard C, Roy PH, Ouellette M,
Bergeron MG (2000) J Clin Microbiol 38:32803284
19. Drancourt M, Bollet C, Carlioz A, Martelin R, Gayral JP,
Raoult D (2000) J Clin Microbiol 38:36233630
20. Jaffe RI, Lane JD, Albury SV, Niemeyer DM (2000) J Clin
Microbiol 38:34073412
21. Buck GE, OHara LC, Summersgill JT (1992) J Clin Micro-
biol 30:13311334
22. Stockman LK, Clark A, Hunt JM, Roberts GD (1993) J Clin
Microbiol 31:845850
23. Saiki RK, Gelfand DH, Stoffel S, Scharf SJ, Higuchi R, Horn
GT, Mullis KB, Erlich HA (1988) Science 239:487491
24. Hsuih TCH, Park YN, Zaretsky C, Wu F, Tyagi S, Kramer
FR, Sperling R, Zhang DY (1996) J Clin Microbiol 34:501
507
25. Walker GT, Fraiser MS, Schram JL, Little MC, Nadeau JG,
Malinowski DP (1992) Nucleic Acids Res 20:16911696
26. Kacian DL, Fultz TJ (1995) US Patent 5,399,491
27. Compton J (1991) Nature 350:9199
28. Lyamichev V, Mast AL, Hall JG, Prudent JR, Kaiser MW,
Takova T, Kwiatkowski RW, Sander TJ, de Arruda M, Arco
DA, Neri BP, Brow MA (1999) Nat Biotechnol 17:292296
29. Ronaghi M, Karamohamed S, Petterson B, Uhlen M, Nyren P
(1996) Anal Biochem 242:8489
30. Wilding P, Pfahler J et al. (1994) Clin Chem 40:4347
64
31. Cheng JP, Fortina S, Sorrey LJ, Wilding P (1996) Molec Diag
1:183200
32. Yuen PK, Kricka LJ, Fortina P, Panaro NJ, Sakazume T,
Wilding P (2001) Genome Res 405412
33. Pohl HA (1978) Dielectrophoresis. Cambridge University
Press, Cambridge, UK
34. Jones TB (1995) Electromechnics of particles. Cambridge
University Press, Cambridge, UK
35. Pethig R (1996) Crit Rev Biotechnol 16:331348
36. Wang XB, Huang Y, Hlzel R, Burt JPH, Pethig R (1993)
J Phys D: Appl Phys 26:12781285
37. Markx GH, Huang Y, Zhou XF, Pethig R (1994) Microbiol-
ogy 140:585591
38. Cheng J, Sheldon EL, Wu L, Heller MJ, OConnell JP (1998)
Anal Chem 70:23212326
39. Huang Y, Ewalt KL, Tirado, M, Haigis R, Forster A, Ackley
D, Heller MJ, OConnell JP, Krihak M (2001) Anal Chem
73:15491559
40. Becker FF, Wang XB, Huang Y, Pethig R, Vykoukal J, Gas-
coyne PRC (1995) Proc Natl Acad Sci USA 92:860864
41. Gascoyne PRC, Wang XB, Huang Y, Becker FF (1997) IEEE
Ind Appl Soc Trans. 33:670678
42. Yang J, Huang Y, Wang XB, Becker FF, Gascoyne PRC
(1999) Anal Chem 71:911918
43. Muller T, Schnelle T, Gradl T, Shirley SG, Fuhr GJ (2000)
J Liq Chromatogr Rel Technol 23:4759
44. Stephens MM, Talary S, Pethig R, Burnett K, Mills I (1996)
Bone Marrow Transplant 18:777782
45. Huang Y, Yang J, Wang XB, Becker FF, Gascoyne PRC
(1999) J Hematother Stem Cell Res 8:481190
46. Yang J, Huang Y, Wang XB, Becker FF, Gascoyne PRC
(2000) Biophys J 78:26802689
47. Gradl G, Muller T, Pfennig A, Shirley S, Schnelle T, Fuhr G
(2000) In: Van den Berg A, Olthius W, Bergveld P (eds) Mi-
cro total analysis systems 2000. Kluwer Academic Publishers,
Enschede, The Netherlands, pp 443445
48. Blankenstein G, Larsen UD (1998) Biosensors Bioelectron
13:427438
49. Blankenstein G, Scampavia L, Ruzicka J, Christian GD (1996)
Cytometry 13:200204
50. http://www.micronics.net
51. Li PCH, Harrison DJ (1997) Anal Chem 69:15641568
52. Schrum DP, Culbertson CT, Jacobson SC, Ramsey JM (1999)
Anal Chem 71:41734177
53. Coakley WT, Hawkes JJ, Sobanski MA, Cousins CM, Speng-
ler J (2000) Ultrasonics 38:638641
54. Ballantine DS, White RM, Martin SJ, Ricco AJ, Zellers ET,
Frye GC, Wohltjen H (1997) Acoustic wave sensors, theory,
design and physico-chemical applications. Academic Press,
London, UK, p 111
55. Fu AY, Spence C, Scherer A, Arnold FH, Quake SR (1999)
Nat Biotechnol 17:11091111
56. Xia Y, Whitesides GM (1998) Angew Chem Int Ed 37:550
575
57. Fry GC, Martin SJ, Cernosek RW, Pfeifer KB, Anderson JS
(1991) IEEE Ultrason Symp Proc 1991:311
58. Tian H, Huhmer AFR, Landers JP (2000) Anal Biochem 283:
175191
59. Belgrader P, Benette W, Hadley D, Richards J, Statton P,
Mariella R Jr, Milanovich F (1999) Science 284:44950
60. Lee SW, Tai YC (1999) Sensors Actuators A 73:7479
61. Northrup MA, Ching MT, White RM, Watson RT (1993)
Proc IEEE Int Conf Solid-State Sensors Actuators 924926
62. Wilding P, Shoffner MA, Kricka L (1994) J Clin Chem 40:
18151818
63. Northrup MA, Gonzalez C, Hadley D, Hills RF, Landre P,
Lehew S, Saiki R, Sninsky JJ, Watson R (1995) Proc IEEE Int
Conf Solid-State Sensors Actuators, pp 764767
64. Shoffner MA, Cheng J, Hvichia GE, Kricka LJ, Wilding P
(1996) Nucleic Acids Res 24:375379
65. Hadd AG, Raymond DE, Halliwell JW, Jacobson SC, Ramsey
JM (1997) Anal Chem 69:34073412
66. Poser S, Schulz T, Dillner U, Baier V, Kohler JM, Schimkat
D, Mayer G, Siebert A (1997) Sensors Actuators A 62:672
675
67. Taylor TB, Winn-Deen ES, Picozza E, Woundenberg TM, Al-
bin M (1997) Nucleic Acids Res 25:31643168
68. Daniel J, Iqbal HS, Millington RB, Moore DF, Lowe CR,
Leslie DL, Lee MA, Pearce MJ (1998) Sensors Actuators A
71:8188
69. Kopp MU, de Mello AJ, Manz A (1998) Science 280:1046
1048
70. Lee TM, Hsing HIM, Lao AIK, Carles MC (2000) Anal Chem
72:42424247
71. Northrup MA, Benett B, Hadley D, Landre P, Lehew S,
Richards J, Stratton P (1998) Anal Chem 70:918922
72. Khandurina J, McKnight TE, Jacobson SC, Waters LC, Foote
RS, Ramsey JM (2000) Anal Chem 72:29953000
73. Gioradano BC, Ferrance J, Swedberg S, Hhmer AFR, Lan-
ders JP (2001) Anal Biochem 291:124132
74. Heap DM, Herrmann MG, Wittwer CT (2000) BioTechniques
29:10061012
75. Kopp MU, de Mello AJ, A Manz (1998) Science 280:1046
1048
76. Lagally ET, Medintz I, Mathies RA (2001) Anal Chem 73:
56570
77. Nagai H, Murakami Y, Morita Y, Yokoyama K, Tamiya E
(2001) Anal Chem 73:10431047
78. Waters LC, Jacobson SC, Kroutchinina N, Khandurina J,
Foote RS, Ramsey JM (1998) Anal Chem 70:51725176
79. Strizhkov BN, Drobyshev AL, Mikhailovich VM, Mirza-
bekov AD (2000) BioTechniques 29:844857
80. Tillib SV, Strizhkov BN, Mirzabekov AD (2001) Anal
Biochem 292:155160
81. Mikhailovich V, Lapa S, Gryadunov D, Sobolev A, Strizhkov
B, Chernyh N, Skotnikova O, Irtuganova A, Moroz A, Litvi-
nov V, Vladimirskii M, Perelman M, Chernousova L, Erokhin
V, Zasedatelev A, Mirzabekov A (2001) J Clin Microbiol 39:
25312540
82. Westin L, Xu X, Miller C, Wang L, Edman CF, Nerenberg M
(2000) Nat Biotech 18:199204
83. Westin L, Miller C, Vollmer D, Canter D, Radtkey R, Neren-
berg M, OConnell JP (2001) J Clin Microbiol 39:10971104
84. Personal communication, Karen Menge, Nanogen
85. Wilding P, Kricka L, Cheng J, Hvichia G, Shofner MA,
Fortina P (1998) Anal Biochem 257:95100
86. Anderson RC, Su X, Bogdan GJ, Fenton J (2000) Nucleic
Acids Res 28:60e
87. Han F, Lilard SJ (2000) Anal Chem 72:407379
88. Pourahmadi F, Taylor M, et al. (2000) Clin Chem 46:1511
1513
89. Krishnan M, Namasivayam V, Lin R, Pal R, Burns RA (2001)
Curr Opin Biotechnol 12:9298
90. Petersen KE, McMillan WA, Kovacs G, Northrup A, Christel
L, Pourahmadi F (1998) IVD Technology July/August:43
91. Kellogg GJ, Arnold TE, Carvalho BL, Duffy DC, Sheppard
NF Jr (2000) In: Van den Berg A, Olthius W, Bergveld P
(eds) Micro total analysis systems 2000. Kluwer Academic
Publishers, Enschede, The Netherlands, 239242
92. Yang JM, Bell J, Huang Y, Tirado M, Thomas D, Forster AH,
Haigis RW, Swanson PD, Wallace RB, Martinsons R, Krihak
M (2001) Submitted for publication
93. Duffy DC, Gillis HL, Lin J, Sheppard NF Jr, Kellogg G
(1999) Anal Chem 71:46694678
94. McMillan WA, Petersen KE, et al. (1999) Proc 8th Int Symp
Microbial Ecology, Halifax, Canada, Atlantic Canada Society
for Microbial Ecology
95. Waters L, Jacobsen S, Kroutchinina N, Khandurina J, Foote
R, Ramsey JM (1998) Anal Chem 70:158162
96. Woodley A, Hadley D, Landre P, deMello A, Mathies R,
Northrup M (1996) Anal Chem 68:40814086
97. http://www.cepheid.com
98. Madou MJ (1997) Fundamentals of microfabrication. CRC
Press

99. Madou MJ, Kellogg GJ (1998) SPIE biomedical optics. San
Jose, CA
100. van Lintel HT, van der Pol GFCM, Bouwstra S (1988) Sen-
sors Actuators 15:15367
101. Bustgens B, Bacher W, Menz W, Schomburg WK (1994)
Proc IEEE Micro Electro Mechanical Systems. vol January.
Oiso, Japan, pp 1821
102. Schomburg WK, Vollmer J, Bustgens B, Fahrenberg J, Hein
H, Menz W (1994) J Micromech Microeng 4:18691
103. Findlay JB, Atwood SM, Bergmeyer L, Chemelli J, Christy
K, Cummins T, Donish W, Ekeze T, Falvo J, Patterson D
(1993) Clin Chem 39:19271933
65
104. http://i-stat.com/
105. Tsao TR, Moroney RM, Martin BA, White RM (1991) IEEE
Ultrasonics Symp Proc, Orlando, FL, pp 93740
106. Jorgenson JW, Guthrie EJ (1983) J Chromatogr 255:335348
107. Harrow JJ, Janata J (1983) Anal Chem 55:24612463
108. Shoji S, Nakagawa S, Esashi M (1990) Sensors Actuators A
A21:189192
109. Hillman RS, Cobb ME, Allen JD, Gibbons I, Ostoich VE,
Winfrey LJ (1990) US Patent 4,963,498
110. Gray BL, Jaeggi D, Mourlas NJ, van Drieenhuizen BP,
Williams KR, Maluf NI, Kovacs GTA (1999) Sensors Actua-
tors 77:5765
Anal Bioanal Chem (2002) 372 : 6673
DOI 10.1007/s00216-001-1195-5
REVIEW
D. C. Collins M. L. Lee
Developments in ion mobility spectrometrymass spectrometry
Received: 19 July 2001 / Revised: 11 October 2001 / Accepted: 16 October 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract Ion mobility spectrometry (IMS) has been used
for over 30 years as a sensitive detector of organic com-
pounds. The following is a brief review of IMS and its
principles with an emphasis on its usage when coupled to
mass spectrometry. Since its inception, IMS has been in-
terfaced with quadrupole, time-of-flight, and Fourier-trans-
form ion cyclotron resonance mass spectrometry. These
hybrid instruments have been employed for the analysis
of a variety of target analytes, including biomolecules, ex-
plosives, chemical warfare degradation products, and il-
licit drugs.
Keywords Ion mobility spectrometry Mass
spectrometry Gas-phase electrophoresis Plasma
chromatography
Introduction
Ion-mobility spectrometry (IMS) was first introduced in
1970 under the name plasma chromatography. Initially
considered as an ion-separation technique with ion drift
times analogous to chromatographic retention times [1,
2], it later became more commonly viewed as a technique
for the selective detection of organic compounds [3, 4]. In
IMS, radioactive 63Ni (typically) is used to ionize vapors
of organic compounds through a series of ionmolecule
reactions. The ions are then carried under an electric po-
tential through a drift region to an ion collector (i.e. Fara-
day plate or mass spectrometer). Ions are selectively de-
tected on the basis of their unique drift times traveling
through the drift region.
D.C. Collins
Weber State University, Department of Criminal Justice,
Ogden, Utah 844081206, USA
M.L. Lee ()
Brigham Young University,
Department of Chemistry and Biochemistry,
Provo, Utah 846025700, USA
e-mail: milton_lee@byu.edu
Interest in IMS blossomed in the early 1970s because
of its analytical versatility, excellent detection limits, suit-
ability for real-time monitoring, and low cost. Despite
these attractive features, however, interest declined after
1976 [3]. Loss of interest was associated with universal
disappointment by unmet expectations and possible mis-
understandings. IMS was mistakenly compared with mass
spectrometry (MS), despite its much lower peak resolu-
tion and lack of mass information. In addition, chemical
aspects of ion formation at atmospheric pressure were not
fully appreciated or understood. This caused many to con-
clude that IMS was an interesting but not practical or us-
able technology [3, 4, 5, 6].
Between 1980 and the early 1990s, an interest in IMS
was renewed as a variety of modifications were made to
instrument design, resulting in its suitability for use for a
wide range of applications. During this time, IMS was fur-
ther employed as a detector for gas chromatography [7, 8]
and supercritical-fluid chromatography [9, 10]. It was also
used for the detection of bacteria [11] and military moni-
toring in hostile environments [3]. Many attempts were
also made to incorporate unconventional ionization sour-
ces, such as laser ionization [12] and a form of electrospray
ionization (i.e., coronaspray) into its design [13, 14].
Principles of IMS
The ability to distinguish ions occurs within the drift re-
gion. In IMS, ions are separated on the basis of their dif-
ferent velocities attained when accelerated through a drift
tube, by a constant electric field, against a counter-flow of
neutral gas. The average velocity of an ion, vd (cm s1), is
determined by the number of collisions it experiences
within the drift tube with the neutral drift-gas molecules,
which is directly proportional to the applied electric field
strength, E (V cm1)
vd = KE (1)
where K is the ion mobility constant in units of cm2 V1 s1
and is a combined property of both the nature of the ion of

interest and the drift gas. Equation 1 is valid only at low
field strengths (e.g. <1000 V cm1). At increasing field
strengths, K is no longer directly proportional to E [15].
K can also be calculated at low electric field strengths by
use of the MasonSchamp equation:
 
K = 3/16 ze/No (2/kT)1/2 (1/D ) (2)
where z is the integer charge of the ion, e is the charge of
an electron in Coulombs (i.e., 1.6021019 C), No is the
number density of the drift gas (molecules cm1), is the
reduced mass of a colliding iondrift-gas pair [=mM/
(m+M)], m is the mass of the ion, M is the mass of a neu-
tral drift molecule, k is Boltzmanns constant, T is the op-
erating temperature in Kelvin, and D is the ionneutral
average cross-sectional area; D is derived through a se-
ries of integrations that average the ionneutral collisions
over all possible scattering angles and energies. For rigid-
sphere collisions, integrating analytically yields D=d2,
where d is the sum of the ion and drift-gas radii. As can be
seen, assuming pressure and temperature are constant,
ze/1/2D is what is actually measured in IMS experiments.
For small atomic ions relative to the drift gas molecules
(e.g. N2, He, or air) mobility is largely controlled by re-
duced mass whereas for heavy ions is essentially equal
to M. The distinguishing property upon which relatively
large ions (the majority of ions analyzed) are separated is,
therefore, ionic size or D. Thus, ions of identical mass,
but different shapes (i.e., isomers), can be separated [3, 4,
5, 6].
Because K, as shown in Eq. (2), is found to be depen-
dent on both temperature and pressure (No is a function of
pressure), the value calculated by use of Eq. (1) is often
converted to a reduced mobility value, Ko
Ko = K(P/760)(273/T) (3)
where P is the operating pressure in Torr and T is the op-
erating temperature in Kelvin. This normalizes mobility
values calculated at different temperatures and pressures
for the purpose of comparison.
During its early years IMS was periodically compared
with time-of-flight mass spectrometry (TOFMS), which
was unfortunate because of its low resolving power and
Fig. 1 Schematic diagram of the Al-
pha II PC/MS combined system (Re-
produced with permission from Ref.
[17])
67
only approximate correlation with ion mass. Its unique ad-
vantage over mass spectrometry is its ability to operate at
atmospheric pressure or near atmospheric pressure. In ad-
dition, data acquired can be used to calculate structural in-
formation (e.g. average cross-sectional area) [16, 17, 18].
By interfacing IMS with MS, a somewhat orthogonal com-
prehensive separation (i.e., structural information) and de-
tection (i.e., mass information) device can be constructed.
IMSMS developments and applications
Most commercial IMS instruments use Faraday plate de-
tection; that is, ions are detected as they strike a metal plate
and induce a current. Faraday plate detection is simple, in-
expensive, and can be used for both positive and negative
ions. Unfortunately, such detection does not afford addi-
tional qualitative information associated with the separa-
tion of ions. Because of this, some have chosen to use MS
as a means of ion detection and identification.
The advantages of interfacing IMS with MS were un-
derstood early in the development of IMS, and the cou-
pling of the two techniques is virtually as old as IMS it-
self. In early 1971, when IMS was referred to as plasma
chromatography (PC), the Franklin GNO Corporation de-
veloped the first commercial ion mobility spectrome-
termass spectrometer (IMSMS) and demonstrated its
use as a detector for the identification and analysis of trace
amounts of oxygenated compounds (i.e., 1-octanol and
1-nonanol) [19]. The instrument was called the Alpha II
PC/MS. It was a two-gated IMS instrument made from
stacked rings that had been joined to a modified Finnigan
quadrupole mass spectrometer. The IMS section operated
at atmospheric pressure and the quadrupole at approxi-
mately 105 Torr (1 Torr=133.322 Pa). The interface between
the two included an ion lens for focusing of ions exiting
the aperture of the IMS into the quadrupole mass spec-
trometer. The ions were then detected with an electron mul-
tiplier placed at the end of the quadrupole, as in Fig. 1 [20].
The instrument was soon used for comparison of IMS
MS data obtained by 63Ni ionization in nitrogen or pure
air, with mass spectral data obtained by chemical ioniza-
68

tion [21]. Also, work was performed to enable better un-

Fig. 2 Total and specific IMS spectra observed for p-nitrophenol
using the Alpha II PC/MS system (Reproduced with permission
from Ref. [17])
derstanding of the formation of ions in the reaction region
of the IMS [20, 22]. In addition, many used the Alpha II
PC/MS to help clarify the validity of IMS mass-mobility
relationships [23, 24]. Several papers, before the above-
mentioned studies, attempted to create mass-mobility cor-
relation curves by speculative assignment of masses to the
identified IMS peaks without the use of a mass spectrom-
eter [25, 26, 27]. The use of the Alpha II PC/MS enabled
this former work to be refined and corrected.
The Alpha II PC/MS was equipped with three distinct
modes of operation. Each mode was dependent on the op-
eration of the entrance and exit gates of the IMS drift
tube. The entrance gate enabled ions to enter the IMS in-
strument for analysis, and the exit gate enabled them to
leave the drift tube and enter the mass spectrometer. Mass
spectral data were acquired by holding both of the gates
open, thus enabling all ions formed within the reaction
region of the IMS instrument to continually drift down the
tube into the quadrupole mass spectrometer. The mass
spectrometer was then scanned to produce a total mass
spectrum of ions present in the sample. Total IMS spectra
could be obtained by operating the entrance gate of the in-
strument in its normal gating fashion (enabling periodic
pulses of ions to enter the drift tube) with the exit gate con-
tinuously open. The mass spectrometer was then operated
in the total-ion-monitoring mode, enabling all ions leav-
ing the drift tube to be detected, but not mass analyzed.
Finally, mass identification of individual IMS peaks could
be achieved by operating the instrument with both gates

Fig. 3 Positive IMS spectra and

mass-identified IMS spectral data
for codeine (traces ac) and acetyl-
codeine (traces df) at 220 C
(Reproduced with permission from
Ref. [26])
 69

opening and closing in such a fashion that only ions of a trometermass spectrometer. Several models were ulti-
desired drift time could pass into the mass spectrometer. mately produced; in all of these the basic operation of the
The mass spectrometer would then be tuned to a previ- new instrument remained the same as that of its predeces-
ously assumed mass-to-charge ratio (m/z) for the purpose sor. To acquire both IMS and MS data the instrument was
of identification. In this manner, direct correlation could gated for a certain drift time and the mass spectrometer
be made between a given IMS peak and its corresponding was tuned to a specific m/z value for the analyte of inter-
mass (or masses, for more than one compound of distinct est. The IMSMS instrument was primarily regarded (and
m/z defining a given IMS peak). Figure 2 shows data col- used) as a selective detector.
lected using both the total- and selected-ion-monitoring Most of the limited IMSMS work in the 1980s was,
modes of the mass spectrometer [20]. as in the 1970s, performed using the commercially avail-
In 1975 Franklin GNO Corporation halted its manufac- able instrument supplied through PCP. Work with this in-
turing operation. Shortly after, however, two former foun- strument in the 1980s included the identification and char-
ders, M. J. Cohen and R. F. Wernlund, formed a new com- acterization of illicit and prescription drugs. It was thought
pany, PCP (West Palm Beach, FL, USA) to continue the that an IMSMS system could be used to aid rapid identi-
technology under the patent license of Franklin GNO. Un-
fortunately, as mentioned previously, interest in ion mobil-
ity spectrometry began to wane in the late 1970s and
through the mid-1980s. The 1980s brought a better under-
standing of IMS, however, and this, in turn, sparked a re-
newal of interest. At this time, many felt IMS was more a
spectrometric than chromatographic process. This new
consensus fostered a name change of the process from
plasma chromatography to ion mobility spectrometry. PCP
began manufacturing an early version of the Alpha II
PC/MS as the Phemto-Chem MMS-160 ion mobility spec-

Fig. 4 IMS spectra for DMMP at 220 C in N2 at a concentration

of (a) 15 g m3, (b) 50 g m3, and (c) 100 g m3 for ions pro-
duced by the 63Ni source and (d) 15 g m3 and (e) 100 g m3 Fig. 5 (a) IMS spectra of cytochrome c at 30, 90, and 200 C.
for ions produced by laser radiation at 266 nm (Reproduced with (b) Corresponding time-of-flight mass spectra at 30, 90, and
permission from Ref. [28]) 200 C (Reproduced with permission from Ref. [30])
70
Fig. 6 Contour plot of the
nested data recorded for the
(D)Phe-Xxx-Xxx-CONH2 pep-
tide library over an m/z range
of 375505. The insets show
IMS slices (and peak intensi-
ties) taken at m/z=399.5 and
476.0. Reproducible features
are indicated with an asterisk.
Sequences that are expected at
these m/z ratios are indicated
(Reproduced with permission
from Ref. [37])
fication of drug residues on the hands of patients admitted
to the hospital with a drug overdose [28, 29] (Fig. 3). The
instrument was also used to identify structurally different
ions with the same m/z values [30]. A modified version of
the instrument was used by Lubman and coworkers, after
laser desorption of solid samples, including 63Ni ioniza-
tion in the normal fashion. They used this method to de-
tect explosives (i.e., trinitrotoluene, cyclotetramethylene-
tetranitramine, and cyclotrimethylene-trinitramine). In ad-
dition, throughout their studies, 63Ni ionization was com-
pared with laser ionization (Fig. 4) [12, 31].
In the 1990s, work continued in the field of IMSMS,
but with more of a departure from the use of the commer-
cially available design offered through PCP. Guevremont
and coworkers purchased an IMSMS instrument from
PCP and modified it to accommodate a time-of-flight mass
spectrometer to study ions formed during the initial stages
of electrospray ionization (Fig. 5). A LeCroy digital oscil-
loscope (Model 9350, Chestnut Ridge, NY, USA) was used
for data collection and processing [32]. TOFMS enabled
complete mass spectra of individual peaks separated in
the IMS instrument to be obtained. This eliminated the te-
dious task of having to tune the mass spectrometer to the
assumed m/z value of an IMS peak.
In 1997, Clemmer et al. designed an IMSMS instru-
ment with a quadrupole mass spectrometer for characteri-
zation of oligosaccharides and proteins [33, 34, 35, 36].
Later, they constructed an IMSMS instrument with a
time-of-flight mass spectrometer to study biomolecules,
incorporating a unique time-to-digital converter data sys-
tem linked to, and operated by, a Pentium computer [37,
38, 39, 40, 41, 42, 43, 44, 45]. Data were presented as
drift times associated with a nested flight time. Sample
ionization was achieved by electrospray ionization. With
the information acquired, a contour plot of flight time
against drift time could be generated (Fig. 6). Clemmer
also performed high-resolution IMS studies. An ion trap
and octapole collision cell were incorporated into the de-
sign to increase the duty cycle and obtain fragment infor-
mation, respectively.
Recently, Hill et al. designed and constructed an in-
house electrospray IMS instrument, and interfaced it with
a C50-Q (ABB Extrel, Pittsburgh, PA, USA) quadrupole
mass spectrometer. The instrument was used successfully
for the study of isomeric peptides, illicit drugs, chemical
warfare degradation products, explosives, and IMS selec-
tivity [46, 47, 48, 49, 50, 51] (Fig. 7). High IMS resolution
values were also reported and associated with the very
small diameter (i.e., 40 m) entrance orifice into the mass
spectrometer, which enabled the sampling of ions from
the most homogenous region of the electric field within
the drift tube [52]. Modes of operation were equivalent to
those offered by the commercial IMSMS instrument
available through PCP.
Russell et al. and Ionwerks (Houston, TX, USA) recent-
ly collaborated in the development of an IMSTOFMS in-
strument with a high-pressure matrix-assisted laser des-
orption ionization source for sample ionization. The in-
strument was used for the analysis of peptide mixtures
(Fig. 8) [53]. Data were collected and processed in an ion-
counting mode by means of an Ionwerks time-to-digital
converter (Model TDCX4). They also designed a Fourier-
transform ion cyclotron resonance mass spectrometerion
mobility spectrometer with TOFMS detection [54].

Fig. 7 (a) IMS spectrum obtained from a drug mixture of amphet-
amine, methamphetamine, cocaine, LSD, and THC. (b) Mass spec-
trum of the drug mixture (Reproduced with permission from Ref.
[45])
Fig. 8 3D plot of m/z against IMS
drift time of a tryptic digest of ther-
mally denatured chicken egg white
lysozyme (Reproduced with permis-
sion from Ref. [50])
71
Also recently, Lee et al. [55] developed an IMSTOFMS
instrument with an Ionwerks data system (Model TDCX4)
and a Jaguar TOFMS (LECO, St. Joseph, MI, USA). The
instrument incorporated a recently developed ambient
pressure and temperature electrospray ionization interface
[56]. This instrument was used for the analysis of selected
benzodiazepines, triazine herbicides, and combinatorial
chemistry samples. Their work emphasized the use of IMS
as a rapid separation technique capable of performing high-
throughput analysis of potential drug candidates synthe-
sized through combinatorial chemistry techniques. The
ability of the IMS to separate compounds was character-
ized in terms of separation figures of merit (i.e., separa-
tion efficiency and peak capacity) [52, 56]. Efficiencies of
35,00050,000 theoretical plates were achieved in less
than 50 ms. Figure 9 shows reconstructed selected-ion
IMS separations of selected benzodiazepines and triazine
herbicides using TOFMS detection.
IMSMS has also found its way into industrial appli-
cations. Probably the most successful application has been
in the semiconductor industry by Carr et al., while work-
ing at IBM. They showed, through a series of papers pre-
sented almost exclusively at conferences, that IMSMS
technology could be used to detect and identify surface
contaminants on semi-conductors and in headspace va-
pors in sealed electronic packages [57, 58, 59, 60].
High-field asymmetric waveform ion mobility spec-
trometry, similar to ion mobility spectrometry, acquires
data as a function of ion mobility constant values. Devel-
oped in the late 1990s, it exploits the change in an ions
mobility constant at increasing electric field strengths, and
operates as an atmospheric pressure ion filter. Guevremont
et al. coupled this instrument to both a quadrupole and a
triple quadrupole mass spectrometer [61, 62, 63, 64, 65,
66]. The hybrid instruments, employing primarily electro-
spray ionization, were successfully used to analyze chlori-
nated and brominated byproducts of drinking water disin-
fection and a variety of biomolecules.
72
Fig. 9 Reconstructed selected-ion IMS separations of (a) five ben-
zodiazepines and (b) five triazine herbicides using TOFMS detec-
tion. Conditions: 27 cm drift tube, 15 kV drift voltage, 20 kV elec-
trospray, 25 C, ~650 Torr (1 Torr=133.322 Pa), 0.5 L min1 sam-
ple, 0.2 ms gate pulse width, 35 Hz (IMS), 5 kHz (TOFMS), 30 s
data collection, N2 1500 mL min1. Peak identifications (in order
as above): (a) oxazepam, diazepam, nitrazepam, clonazepam,
and prazepam; (b) simetryn, ametryn, prometon, terbutryn, and
prometryn
Conclusions
IMS is a separation technique that affords qualitative in-
formation associated with the average cross-sectional area
of ions. When interfaced with a mass spectrometer, data
can be collected which gives additional qualitative infor-
mation (i.e., mass information). IMS has been success-
fully coupled to quadrupole, time-of-flight, and Fourier-
transform ion cyclotron resonance mass spectrometry us-
ing radioactive 63Ni, electrospray, laser, and matrix-assist-
ed laser desorption ionization sources. Applications have
been reported for the analysis of illicit drugs, explosives,
chemical warfare degradation products, biomolecules,
and combinatorial chemistry samples. Instrumentation has
also been used to study ions formed in an electrospray
source and to generate high-resolution IMS separations.
References
1. Karasek FW (1970) Res Dev 21:3437
2. Karasek FW (1970) Res Dev 30:2527
3. Eiceman GW (1991) Critical Rev Anal Chem 22:471489
4. Spangler GE, Campbell DN, Vora KN, Carrico JP (1984) ISA
Trans 23:1728
5. St Louis RH, Hill HH Jr (1990) Anal Chem 21:321355
6. Hill HH Jr, Siems WF, St Louis RH, McMinn DG (1990) Anal
Chem 62:12011209A
7. Balm MA, Hill HH Jr (1982) Anal Chem 54:3843
8. Snyder AP, Harden CS (1993) Anal Chem 65:299306
9. Morrissey MA, Hill HH Jr (1989) J Chromatogr Sci 27:529
533
10. Huang MX, Markides KE, Lee ML (1991) Chromatographia
31:163167
11. Snyder AP, Shoff DB (1991) Anal Chem 63:526529
12. Lubman DM, Kolaitis L (1986) Anal Chem 58:19932001
13. Shumate CB, Hill HH Jr (1989) Anal Chem 61:601606
14. McMinn DG, Kinzer JA, Shumate CB, Siems WF, Hill HH Jr
(1990) J Microcol Sep 2:188192
15. Revercomb HE, Mason EA (1975) Anal Chem 47:970983
16. Karpas Z (1986) Anal Chem 74:153
17. Karpas Z, Stimac RM, Rappoport R (1988) Int J Mass Spec-
trom Ion Proc 83:163
18. Karpas Z (1989) Anal Chem 61:684
19. Karasek FW, Cohen MJ, Carroll DI (1971) J Chromatogr Sci 9:
390392
20. Karasek FW, Kim SH, Hill HH Jr (1976) Anal Chem 48:1133
1137
21. Karasek FW, Denney DW, DeDecker EH (1974) Anal Chem
46:970973
22. Kim SH, Betty KR, Karasek FW (1978) Anal Chem 50:1784
1788
23. Carr TW (1977) Anal Chem 49:828831
24. Griffin GW, Dzidic I, Carroll DI, Stillwell RN, Horning EC
(1973) Anal Chem 45:12041209
25. Karasek FW, Kilpatrick WD, Cohen MJ (1971) Anal Chem
43:14411447
26. Kilpatrick WD (1971) Abstracts from the 19th Annual Confer-
ence on Mass Spectrometry, Atlanta, GA, p 320
27. Karasek FW, Tatone OS (1972) Anal Chem 44:17581763
28. Lawrence AH, Nanji AA (1988) Biomed Environ Mass Spec-
trom 16:345347
29. Lawrence AH (1986) Anal Chem 58:12691272
30. Kim SH, Spangler GE (1985) Anal Chem 57:567569
31. Lubman DM, Huang SD, Kolaitis L (1987) Appl Spectrosc 41:
13711376
32. Guevremont KW, Siu KWM, Wang J, Ding L (1997) Anal
Chem 69:39593965
33. Yansheng L, Clemmer DE (1997) Anal Chem 69:25042509
34. Hoaglund CH, Valentine SJ, Clemmer DE (1997) Anal Chem
69:41564161
35. Shelimov KB, Clemmer DE, Hudgins RR, Jarrold MF (1997)
J Am Chem Soc 119:22402248
36. Dugourd Ph, Hudgins RR, Clemmer DE, Jarrold MF (1997)
Rev Sci Instrum 68:11221129
37. Valentine SJ, Counterman AE, Hoaglund CS, Reilly JP, Clem-
mer DE (1998) J Am Soc Mass Spectrom 9:12131216
38. Hoaglund CS, Valentine SJ, Sporleder CR, Reilly JP, Clemmer
DE (1998) Anal Chem 70:22362242
39. Henderson SC, Valentine SJ, Counterman AE, Clemmer DE
(1999) Anal Chem 71:291301
40. Srebalus CA, Li J, Marshall WS, Clemmer DE (1999) Anal
Chem 71:39183927
41. Valentine SJ, Counterman AE, Clemmer DE (1999) J Am Soc
Mass Spectrom 10:11881211
42. Hoaglund-Hyzer CS, Li J, Clemmer DE (2000) Anal Chem 72:
27372740
43. Srebalus CA, Li J, Marshall WS, Clemmer DE (2000) J Am
Soc Mass Spectrom 11:352335
44. Srebalus Barnes CA, Clemmer DE (2001) Anal Chem 73:424
433
45. Hoaglund-Hyzer CS, Clemmer DE (2001) Anal Chem 73:177
184
46. Wu C, Siems WF, Asbury GR, Hill HH Jr (1998) Anal Chem
70:49294938
47. Asbury GR, Hill HH Jr (2000) Anal Chem 72:580584
48. Wu C, Siems WF, Hill HH Jr (2000) Anal Chem 72:396403

49. Asbury GR, Wu C, Siems WF, Hill HH Jr (2000) Anal Chim
Acta 404:273283
50. Asbury GR, Klasmeier J, Hill HH Jr (2000) Talanta 50:1291
1298
51. Wu C, Siems WF, Klasmeier J, Hill HH Jr (2000) Anal Chem
72:391395
52. Asbury GR, Hill HH Jr (2000) J Microcol Sep 12:172178
53. Gillig JK, Ruotolo B, Stone EG, Russell DH, Fuhrer K, Gonin
M, Schultz AJ (2000) Anal Chem 72:39653971
54. Bluhm BK, Gillig KJ, Russell DH (2000) Rev Sci Instrum 71:
40784086
55. Collins DC, Lee ML (2001) Ultrahigh speed separations using
gas-phase electrophoresis, 24th Int Symp Capillary Chro-
matography and Electrophoresis, Las Vegas, NV
56. Collins DC, Lee ML (2001) Fresenius J Anal Chem 369:225
233
57. Carr TW (1977) Thin Solid Films 45:115
58. Carr TW, Needham CD (1978) Analysis of organic surface
contamination by plasma chromatographymass spectrometry,
4th Int Symp Contamination Control, Washington DC
73
59. Carr TW (1980) Analysis of organic surface contaminants by
plasma chromatography-mass spectroscopy and Raman micro-
probe techniques, 8th Ann Proc IEEE, Reliability physics, New
York, p 186
60. Carr TW, Corl EA, Liu CL, Majtenyi CG (1978) Quantitative
H2O determination in components using a plasma chromato-
graph-mass spectrometer, 16th Ann Proc IEEE, Reliability
physics, New York, p 59
61. Purves RW, Guevremont R, Day S, Pipich CW, Matyjaszczyk
(1998) Rev Sci Instrum 69:112
62. Ells B, Barnett DA, Froese K, Purves RW, Hrudey S, Guevre-
mont R (1999) Anal Chem 71:47474752
63. Guevremont R, Purves RW (1999) J Am Soc Mass Spectrom
10:492501
64. Barnett DA, Ells B, Guevremont R (1999) J Am Soc Mass
Spectrom 10:12791284
65. Purves RW, Guevremont R (1999) Anal Chem 71:23462357
66. Purves RW, Barnett DA, Ells B, Guevremont R (2000) J Am
Soc Mass Spectrom 11:738745
Anal Bioanal Chem (2002) 372 : 7490
DOI 10.1007/s00216-001-1139-0
REVIEW
I. Rodriguez Pereiro A. Carro Daz
Speciation of mercury, tin, and lead compounds
by gas chromatography with microwave-induced plasma
and atomic-emission detection (GCMIPAED)
Received: 7 June 2001 / Revised: 10 September 2001 / Accepted: 20 September 2001 / Published online: 13 December 2001
Springer-Verlag 2001
Abstract Because of their high toxicity and widespread
distribution, the reliable selective quantification of alkyl
and aryl species containing mercury, tin, or lead has been
one of the goals of speciation analysis in recent years.
Since becoming commercially available, GCMIPAED
has been one of the most-used tools in this work. In this
paper, the value and limitations of GCMIPAED for the
speciation of Hg, Sn, and Pb compounds in environmental
samples are reviewed and compared with the analytical
characteristics of other hyphenated GC-based techniques.
Because quantification of Hg, Sn, and Pb species by GC
techniques normally requires complex sample preparation
involving several steps, the effect of sample-preparation
methods on the accuracy and precision of the results is
discussed. Finally, we describe the current status of a
rapid, low-cost GCMIPPED system specifically de-
signed for routine quantification of Hg, Sn, and Pb species
in environmental control laboratories.
Keywords Speciation Mercury Tin Lead Gas
chromatographymicrowave-induced plasmaatomic
emission
Introduction
Speciation analysis is the separation and quantification of
identifiable chemical species containing a particular ele-
ment [1]. Its aim is to determine the distribution of the
element among the different chemical species in a sample.
Few of the instrumental methods available for elemental
analysis have sufficient selectivity for direct quantifica-
tion of individual metal-containing species. Spectroscopic
techniques are very sensitive but usually determine only
I.R. Pereiro () A.C. Daz
Dpto. Qumica Analtica, Nutricin y Bromatologa,
Facultad de Qumica, Universidad de Santiago de Compostela,
Santiago 15782, Spain
e-mail: qnisaac@usc.es
the total amount of metal in the sample, and they must be
combined with a separation technique before individual
species can be quantified [2, 3]. Speciation analysis, there-
fore, normally relies on hyphenated techniques.
In this review we concentrate on what, with the excep-
tion of GCMS, has probably been the most successful
commercially marketed GC-based hyphenated technique,
the combination of GC with detection of atomic emission
stimulated by use of a microwave-induced helium plasma
(GCMIPAED).
Because microwave-induced helium plasmas (MIP)
are formed at lower temperatures than inductively cou-
pled argon plasmas (ICP), liquid samples cannot be intro-
duced into them, because they are extinguished by even
small amounts of organic vapor. When MIPAES is com-
bined with GC it is, therefore, necessary to vent the GC
solvent front before it arrives at the discharge tube. When
this problem had been solved, GCMIPAED proved to
have several advantages over GCICPAED. The most
important are, perhaps, the higher electron temperature
and ionization energy of MIP, which enable quantification
not only of metals but also of semi-metals, and even of
other organic compounds containing elements with high
ionization potentials (fluorine, chlorine, etc.) [4, 5, 6].
Other advantages of MIP are the small dead volume of the
discharge tube, the compatibility of the plasma with the
low carrier gas flows used in capillary GC columns, and
the consumption of less gas than ICP. AES detection of el-
ements excited by an MIP is, furthermore, both very se-
lective and often (e.g. for selenium and for most elements
with atomic masses lower than 40) less subject to interfer-
ence than the more sensitive ICPMS system [3, 7]. There
is no doubt that these properties that were responsible for
the worldwide commercial success of the GCMIPAED
system marketed by HewlettPackard at the end of the
nineteen-eighties.
One of the most important environmental applications
of GCMIPAED is the quantification of alkylmetal species
of low molecular-weight. These species are highly toxic at
low concentrations and are easily bioaccumulated; some
are regarded as endocrine system disrupters. In environ-
 75

Table 1 Summary of GCAES methods for the determination of inorganic mercury and organomercury species in various sample
matrices

Species; sample Sample preparation LOD CRM used Determination Date Ref.
matrix technique

MeHg; fish Alkaline digestion 0.4 pg MeHg as CRM 464, GCMIPAED 2000 [38]
tissue (KOHMeOH)+derivatization Hg DORM-2
(HClCH 2Cl2NaBPh 4)
MeHg, EtHg; (1) Ethylation (NaBEt4) (1) 0.04 pg GCMIPAED 2000 [24]
standard aqueous MeHg as Hg
solutions (2) Phenylation (NaBPh4) (2) 0.02 pg GCAFS
MeHg as Hg
(2) 0.03 pg GCMS
EtHg as Hg
MeHg, Hg2+; (1) Derivatization (Grignard reagent) 0.1 ng g1 IAEA 356, GCMIPAED 2000 [31]
soil, sediment (2) CH 2Cl2 extraction+ethylation MeHg CRM 580,
(NaBEt4)+Tenax collection PACS-1

MeHg, EtHg, Derivatization (Grignard reagent) 0.2 pg MeHg DORM-2, GCrf-HCGDAED 2000 [23]
Hg2+; dogfish and EtH, 0.3 pg DOLT-2
tissue Hg2+
MeHg; tuna fish, MAE (HAc)+phenylation (NaBPh4) 0.04 pg MeHg CRM 463, GCMIPAED 2000 [43]
cockle, mussel, as Hg CRM 464,
dogfish DORM-2
MeHg, Me2Hg, (1) Ethylation (NaBEt4) <1 ng L1 PTIGCMIPAED 1999 [49]
Hg2+; water, soil, (2) Hydride generation
sediment (NaBH 4)+preconcentration on
Chromosorb and thermal desorption
Me2Hg, Et2Hg, (1) Direct SPME (1) 144 pg mL1 GCMIPAED 1999 [34]
Ph2Hg; water, Me2Hg as Hg,
soil, sediment 30 pg mL 1
Et2Hg as Hg
(2) HS-SPME (2) 30 pg mL 1
Me2Hg as Hg,
25 pg mL 1
Et2Hg as Hg
MeHg, EtHg, Acid hydrolysis 1.3 ng mL 1 DORM-2, GCrf-GDAED, 1998 [22]
Hg2+; fish tissue (HCl)+DDTC+derivatization MeHg and DOLT-2 GCdc-GDAED
(Grignard reagent) EtHg, 3.0 ng
mL 1 Hg2+
MeHg, Me2Hg, Derivatization (Grignard reagent) (1) 0.5 (2) (1) GCFAPES 1998 [21]
Hg2+; natural gas 2.1 pg Me 2Hg
condensate (1) 0.9 (2) (2) GCMIPAED
4.1 pg MeHg
(1) 1.7 (2)
4.1 pg Hg2+
MeHg; grain, Acid hydrolysis (HCl)+Celite 545 0.24 pg as Hg GCAED 1998 [54]
cereal products, column+CH 2Cl2 elution+Kuderna
fruit, vegetables Danish concentrator+stannic
chlorideMeOH
MeHg; fish Alkaline digestion 5 pg MeHg PTIGCMIPAED 1998 [39]
(KOHMeOH)+ethylation
(NaBEt4)+Carbotrap column
MeHg; fish MAE (TMAH)+(1) ethylation (1) 3 pg g1 CRM 464 PTIGCMIPAED 1997 [41]
(NaBEt4) MeHg
MAE (TMAH)+(2) hydride (2) 12.5 pg g 1
generation (NaBH 4) MeHg
MeHg, Hg2+; TMAH digestion+ethylation 7 pg MeHg, DOLT-2, PTIGCFAPES 1997 [44]
fish tissues (NaBEt4)+Tenax-TA column+thermal 1 pg Hg2+ TORT-2,
desorption DORM-2
76

Table 1 (continued)

Species; sample Sample preparation LOD CRM used Determination Date Ref.
matrix technique

MeHg, Hg2+; Ethylation (NaBEt4) 6 pg MeHg, PTIGCMIPAED 1996 [71]

water 20 pg Hg 2+
MeHg; sediment (1) SFE (2) distillation (KClH 2SO 4) 0.1 ng g1 PACS-1 GCMIPAED 1996 [29]
MeHg; marine Extraction (Cu powder+CuSO 4)+ DORM-2, GCMIPAED 1996 [32]
tissues KBr+toluene TORT-1,
DOLT-2,
IAEA 350
MeHg, Me2Hg, (1) On-line amalgamation (1) 0.24 mg L 1 GCMIPAED 1996 [33]
Hg2+; natural gas trap+derivatization (Grignard reagent) Me2Hg,
condensate 0.56 mg L 1
MeHg and Hg2+
(2) HS-SPME+derivatization (2) 2 mg L 1
(Grignard reagent) Me2Hg,
3 mg L 1 MeHg,
2.3 mg L 1 Hg2+
MeHg, EtHg; Extraction (Cu powder+CuSO 4)+ 0.8 pg MeHg TORT-1, GPC-GCMIPAED 1996 [30]
marine tissues, KBr+toluene as Hg DOLT-2,
sediment IAEA 350,
DORM-2....
MeHg, EtHg, SPE (DTC- resin)+acid thiourea 0.4 pg MeHg, GCMIPAED 1995 [47]
Hg2+; sea water elution+derivatization (Grignard EtHg and Hg 2+
reagent)
MeHg, Hg2+; SPE (DTC- resin)+acid thiourea 0.04 ng MeHg, GCMIPAED 1995 [46]
natural water elution+derivatization (Grignard 0.28 ng Hg 2+
rich in humic reagent)
substances
MeHg, EtHg, Sulfhydryl cotton microcolumn+HCl 2 ng MeHg GCMIPAED 1995 [53]
Hg2+; water elution+phenylation (NaBPh 4) and EtHg,
3.2 ng Hg2+
MeHg, EtHg, Direct aqueous phase phenylation GCMIPAED 1995 [36]
PhHg, Hg2+; (NaBPh4)
biological tissues
MeHg; fish and (1) Alkaline digestion (NaOH-NaCl) From low ng g 1 DOLT-2, GCMIPAED 1994 [40]
other biological (2) Acid leaching (HCl-NaClsat)+ to mg g 1 TORT-2,
tissues derivatization (Grignard reagent) CRM 464,
CRM 463,
IAEA-MA
B3/TM
MeHg; fish, Acid hydrolysis (HCl)+cysteine 1.2 pg MeHg DORM-1 GCMIPAED, 1994 [20]
biological acetate aqueous solution+toluene GCECD
tissues, sediment
MeHg, EtHg, DTC-resin+elution with thiourea 0.025 ng MeHg Closed flow injection- 1993 [45]
Hg2+; natural solution+derivatization (Grignard and EtHg, GCMIPAED
water reagent) 0.075 ng Hg2+
MeHg; Tenax column+thermal elution into 3 pg MeHg GCICPAED 1992 [19]
atmospheric benzene+concentration with a as Hg
samples nitrogen gas flow
MeHg, EtHg; Acid leaching 0.8 pg MeHg as IAEA-MA GCECD, GCMIP a 1991 [37]
fish (HClNaClsat)+derivatization Hg, 1.3 pg EtHg B3/TM
(Grignard reagent) as Hg
MeHg; fish (1) Acid hydrolysis (HCl)+cysteine 3 ng/8 mL GCECD, GC-DCP 1987 [18]
acetate aqueous solution+Kuderna MeHg AED
Danish concentrator
(2) Extraction with diethyl etherlight
petroleum+Kuderna Danish
concentration

Table 1 (continued)
Species; sample Sample preparation LOD
matrix
MeHg, Me2Hg; Acid hydrolysis (HBrHI)+benzene 2 pg Me2Hg,
water, fish 0.5 pg MeHg
Organic Hg; Acid hydrolysis 0.1 ng MeHgCl
salmon (HCl)+HgCl2+Benzene+clean-up
with thiosulphate
a
Non-commercial microwave-induced plasma
mental samples they are normally present in polar species,
making their conversion to compounds that are less polar
and thermally stable a pre-requisite for gas chromato-
graphic separation [7]. CGMIPAED is one of the best
and most frequently used techniques for quantification of
the resulting derivatives.
This paper reviews the application of CGMIPAED
to the speciation of alkyl species containing mercury, tin,
and lead in environmental samples mainly water, sedi-
ments, and biological tissues. The scope and limitations of
the technique are discussed, and its analytical perfor-
mance is compared with that of other hyphenated tech-
niques used in species analysis, including GCAED sys-
tems with other plasma sources. Particular attention is
given to sample preparation procedures and derivatization
methods, because of their great influence on the precision
and accuracy of the analytical results.
Speciation of mercury species
The biogeochemistry and the health effects of mercury
(Hg) have received considerable attention because of the
toxicity of methylmercury (MeHg) and dimethylmercury
(Me2Hg), the accumulation of Hg in biota, and its bio-
magnification in aquatic food chains [8, 9]. Knowledge of
the concentration, transport, and dynamics of MeHg and
other organomercury compounds in aquatic ecosystems
is, therefore, needed to enable prediction of the potential
impact on human and aquatic life [8, 10].
In the nineteen-eighties it became clear that GC with
electron-capture detection (ECD) had poor selectivity for
organomercury species and often required elaborate, time-
consuming clean-up procedures before sample injection;
a simpler, more specific technique was required [11, 12].
Although early GCMIPAED systems had many draw-
backs (see below, Comparison of plasma sources and de-
tection techniques), the commercial MIPAED instru-
ment has many advantages over these earlier designs, par-
ticularly with regard to venting of the solvent to protect
the plasma, plasma stability, background correction, and
convenience of data handling. The use of a high-perfor-
mance monochromator with a photodiode-array spectro-
photometer as detector also improves sensitivity [13] (at-
mospheric-pressure MIP are more sensitive and signifi-
cantly more selective than low-pressure MIP for mercury;
to obtain maximum sensitivity for mercury with a low-
77
CRM used Determination Date Ref.
technique
GCECD, GC-MIPa 1975 [28]
GC-MIP a 1971 [27]
pressure MIP, hydrogen should be used as scavenger gas
[14].) Table 1 lists studies of the application of GC
MIPAED, and of some other GC methods, to the deter-
mination of Hg species in a variety of sample matrices.
Note that in each example the table specifies the certified
reference materials (CRM) used; in species analysis, as in
other areas of analytical chemistry, the use of CRM is es-
sential for good quality control and traceability [15, 16].
Although packed columns are sometimes useful, capil-
lary columns are generally essential for complex environ-
mental samples [17]. Because of its flexibility, ease of
use, inertness, and ability to retain a variety of coatings,
fused silica is the preferred material for capillary columns,
its inertness being especially important in the determina-
tion of chemically active compounds such as organomer-
cury compounds [14].
Plasma sources and detection systems
Early GCMIPAED systems had serious drawbacks
intolerance of large volumes of injected sample, plasma
instability, poor reproducibility, poor precision, and in-
ability to run unattended for long periods of time [18]. Be-
cause of this, at that time Panaro et al. developed a simple,
inexpensive GCAED method using a direct current
plasma (DCP) [18]; an isothermal packed-column GC
system dedicated to the DCP generator enabled routine
qualitative and quantitative determination of organomer-
cury species in complex food samples (when used to de-
termine MeHg in fish, the sample was subjected to an ex-
traction procedure based on the method of Hight and
coworkers [11, 12] and included concentration by use of a
Kuderna-Danish apparatus). Being element-selective, DCP
AED eliminated interferences and thus enabled more ac-
curate determination of MeHg than ECD.
MIPAED is very sensitive for volatile species con-
taining metals but, as noted above, the introduction of sol-
vent into the plasma quenches the discharge, making it
necessary either to vent the solvent vapor before it enters
the cavity or to ignite the plasma only after the solvent has
passed through the cavity. An ICP can, on the other hand,
be maintained in the presence of an organic solvent. Kato
et al. combined GC with an axially viewed inductively
coupled plasma (AXV-ICP) and AED [19]; air samples
were collected in a Tenax column and eluted with ben-
zene. AXV performed better than conventional side view-
78
ing in particular, the LOD for alkylmercury compounds
in the atmosphere was 20 times lower with the axial
method.
After an acid-hydrolysis extraction procedure, CGMIP
AED enables accurate quantification of MeHg in unclean
marine samples and samples with a high fat content (e.g.
mussels), which pose serious resolution problems for GC
ECD systems [20].
Frech et al. compared MIPAED and furnace-atomiza-
tion plasma-excitation spectrometry (FAPES) as detection
systems for use with high-resolution GC [21]. The test an-
alytes were the mercury species in a natural gas conden-
sate, which had been derivatized by use of butylmagne-
sium chloride in tetrahydrofuran. Better LOD were achieved
with FAPES, and MIPAED measurement of Me2Hg was,
in fact, impossible, because the plasma could not with-
stand loading with hydrocarbon solvents. Although base-
line disturbances occurred with both sources during sam-
ple elution, in FAPES they resulted from changes in back-
ground emission, a problem that could be ameliorated by
means of a suitable background correction system,
whereas in MIPAED they were because the plasma lost
excitation capability, a problem little can be done to solve.
Recently, the analytical potential of capillary GC with
radiofrequency (rf) and direct current (dc) glow dis-
charges (GD) and AED has been investigated for quantifi-
cation of low levels of MeHg, ethylmercury (EtHg) and
inorganic mercury (Hg2+) in fish tissues [22]. Both
dc-GDAED and rf-GDAED performed at a level similar
to that of more common AED methods. GCGDAED
enabled accurate determination of the Hg species without
the need to resort to standard addition techniques. When a
hollow-cathode was used to increase emission intensities,
optimization of discharge conditions (pressure, He flow
rate, rf power) afforded LOD that were 510 times better
than those obtained by use of flat-cathode GDAED and
were also lower than those obtained with MIPAED [23].
The hollow-cathode GD-AED system had a low construc-
tion cost and consumed 1020 times less He than MIP
AED.
The performance indices of GC in combination with
MIPAES, mass spectrometry (MS), or atomic fluorescence
spectrometry (AFS) have been evaluated for quantifica-
tion of MeHg and EtHg after aqueous derivatization with
sodium tetraethylborate and sodium tetraphenylborate
[24]. Both GCAFS and GCMIPAED had broad linear
ranges, although the AFS sensitivity setting had to be ad-
justed for each particular sample on the basis of its Hg
concentration. Phenylation seemed to be the better deriva-
tization method, because it distinguished better between
EtHg and Hg2+ and was less expensive than ethylation. It
was concluded that although GCMS is, perhaps, the
most important technique for identification and confirma-
tion, it is unlikely to provide the sensitivity required for
determination of Hg in most environmental samples.
Sample pretreatment for mercury speciation
with GCMIPAED
Extraction
The first extraction method developed specifically for
quantification of mercury species involved the leaching of
Hg compounds from the sample using hydrochloric acid
[25, 26]. Since the nineteen-seventies this method has
been used to prepare fish samples for the determination of
several organic mercury salts by GCMIPAED [27, 28]
(Talmis paper [28] describes a procedure that makes the
clean-up step of the West method unnecessary.) The to-
tal analysis time for MeHg in fish is less than 15 min, and
more than 30 benzene extracts can be analyzed per hour.
As mentioned above, a modified acid hydrolysis proce-
dure has also been applied to marine samples before quan-
tification of MeHg by CGMIPAED [20].
Supercritical-fluid extraction (SFE) has been used to
isolate MeHg from sediments before quantification by
CGMIPAED [29]. SFE is fast and reliable, and is more
sparing of solvents, labor and chemicals than other cur-
rently used methods such as steam distillation (the refer-
ence method in the study given in Ref. [29]).
Acid leaching procedures have been applied to both
marine tissues and sediments [30, 31]. Interfering sulfur-
containing species can be removed during sample extrac-
tion with copper powder, the extracts subsequently being
themselves extracted with toluene after acidification with
potassium bromide solution [30, 32]. High molecular-
weight pigments and lipids can be removed from the ex-
tracts by preparative gel-permeation chromatography
(GPC) before GC analysis [30].
Few studies have used solid-phase microextraction
(SPME) to prepare samples for quantification of Hg species
by GCMIPAED [33, 34]. SPME extracts volatile or
semivolatile organic compounds directly from an aqueous
or gaseous sample passed through a capillary or over a
fused-silica fiber coated with an appropriate stationary
phase. In combination with headspace sampling (HS),
SPME has proved suitable for extraction of Me2Hg and
diethylmercury (Et2Hg) in the analysis of aqueous and
soil samples. It is also possible to use direct SPME for
less volatile organomercury compounds in aqueous sam-
ples [34].
Derivatization
The derivatization of analytes before their determination
by GC with low-pressure or atmospheric-pressure MIP
AED avoids capillary column passivation problems even
if no prior extraction stage is used [35]. Direct ethylation
or phenylation in an aqueous medium, and butylation with
a Grignard reagent, have been compared with regard to
their improvement of the multi-element determination of
several organomethylates in biological tissue [36]. The
addition of traces of a mixture of oxygen and hydrogen to
the plasma improves the shape of the chromatogram peaks

by removing carbon deposits and preventing the forma-
tion of refractory oxides. For samples for which there is
no information about the species present, derivatization
by at least two procedures is recommended.
Extraction and derivatization
Fish and marine samples. Combining derivatization with
extraction improves the chromatographic characteristics
of the species to be separated. For MeHg and EtHg in fish,
samples can be extracted with saturated sodium chloride
solution and HCl; after the extracts have been shaken vig-
orously with sodium diethyldithiocarbamate (DDTC) and
toluene, the toluene phase can be derivatized with butyl-
magnesium chloride in tetrahydrofuran (THF) [37]. Col-
umn efficiency, detection limits, and resolution of MeHg
and EtHg are better in capillary GCMIPAED than
packed column GCECD.
Alkaline digestion with KOHmethanol is an obvious
alternative to acid treatment before extraction of Hg com-
pounds with dichloromethane then phenylation [38] or
ethylation [39]. When used to isolate MeHg from fish
samples, alkaline digestion has been found to eliminate
interferences with the derivatization reaction before GC
MIPAED determination [38, 39]. International intercali-
bration studies of the simultaneous determination of
MeHg and Hg2+ from reference and candidate reference
materials by CGMIPAED after acid leaching or alka-
line digestion and subsequent derivatization revealed very
large differences between the extraction efficiencies of
different sample work-up procedures. Artifact formation,
detector selectivity, chromatographic performance, and
the stability of Hg compounds are discussed on the basis
of their results [40].
Microwave-assisted extraction (MAE) is a well ac-
cepted technique in elemental trace analysis and has also
been used for quantification of organometallic compounds.
Under the usual conditions (high applied energy, high
pressures, and high temperatures) most compounds are
destroyed and all speciation information is lost [41], but
when performed under milder conditions in a closed sys-
tem with sophisticated pressure and temperature control it
achieves successful extraction of organomercury com-
pounds [42]. Alternatively, an open system using focused
microwaves has also been successfully applied to the di-
gestion of fish samples before Hg species analysis by
GCMIPAED [41]. Digestion of fish tissues with tetra-
methylammonium hydroxide (TMAH) is quick and sim-
ple, and losses of volatile analytes as a result of heating by
the microwave field are reasonably low and reproducible.
Two different derivatization and injection procedures
have been examined (ethylation, extraction into hexane,
and injection with a cooled injection system; and hydride
generation with sodium tetrahydroborate then purge-and-
trap injection (PTI)).
For determination of MeHg with the closed-vessel sys-
tem the extraction and derivatization conditions have been
optimized by use of a factorial experimental design [43].
79
Combination of MAE with a suitable derivatization pro-
cedure avoids the need for previous column treatment
with inorganic salts before GCMIPAED, and also re-
duces solvent volume and analysis time.
PTI has also been used in other studies of the applica-
tion of GCMIPAED [41] or GCFAPES [44] to quan-
tification of MeHg and Hg2+ in biological tissues. In the
latter technique samples were solubilized with TMAH
and the ionic species were purged from aqueous solution
after ethylation; the species were then preconcentrated on
Tenax-TA and thermally desorbed on to an isothermal GC
column.
Water samples. In the quantification of Hg compounds in
water samples, low concentration is without doubt the
most serious problem; occasionally very large sample vol-
umes must be processed. The method developed by Emte-
borg et al. [45] for simultaneous determination of mercury
species in natural waters involves large sample volumes,
preconcentration on a dithiocarbamate (DTC) resin-loaded
microcolumn, extraction of the eluted species into toluene,
and derivatization with a Grignard reagent; these deriva-
tives are separated and quantified by GCMIPAED. One
disadvantage of the method is that the preconcentration
step fails in the presence of high concentrations of humic
substances. To solve this problem a batch method has
been developed [46] Hg species enrichment is per-
formed by adding purified DTC resin directly to the water
sample. The accuracy of this method has been assessed, in
part, by means of interlaboratory comparison with results
from GCAFS, and the ability of the columns to retain
alkyl-Hg and Hg2+ compounds has been studied [47]. In a
later variant, large volumes of the sample are injected in a
packed-column GC system connected to a capillary GC
system coupled to a MIPAED device [48]. In this way
the Hg species are focused before separation in the ana-
lytical column; this minimizes the risk of the plasma be-
ing extinguished by excess solvent. This variant has been
applied to the determination of Hg species in river water.
In another study [49] Hg species were either ethylated
or were converted to hydride with NaBH4, stripped from
solution with He, pre-concentrated on Chromosorb at
room temperature or 160 C, released from the adsorbent
by thermal desorption, and quantified by PTIGCMIP
AED. Both these procedures were designed for analysis
of soil and sediments and water.
Because of problems associated with the production of
natural materials and the stability of Hg compounds in
aqueous media, very few reliable reference materials are
available for the quantification of mercury compounds in
natural waters [50, 51]. Because of this the possibility of
storage on a solid support has been studied [52]. The sta-
bility of Hg species from natural waters immobilized on
sulfhydryl cotton fibers (SCF) was examined by elution
into HCl solution, derivatization with NaBPh4, and deter-
mination by GCMIPAED [53].
Gaseous samples. Hg species present in natural gas con-
densate have been extracted for determination by GC
80
MIPAED using an on-line amalgamation trap [33 ]. The
Hg compounds can be collected from the column eluate
within a specific time window and subsequently passed to
the plasma as Hg vapor in a flow of pure helium. Use of
the amalgamation trap improves the determination of
mercury species in two ways. First, by separating analyte
material from the matrix before detection it enables the in-
troduction of samples containing large amounts of carbon
compounds without any need to use split injectors or di-
luted or smaller sample masses to preserve chromato-
graphic resolution and detector efficiency. Second, use of
the trap enables the addition of standards directly after the
gas chromatograph by injecting samples of air saturated
with mercury vapor; this procedure enables independent
calibration for standards separated by the chromatograph.
Results obtained by use of the trap were compared with
data obtained by HS-SPME and liquid sample injection
followed by direct measurement of the column eluate.
SPME might be a simple and useful alternative method
for the determination of derivatized polar Hg species [33].
Soil samples. In addition to the method mentioned above
in the section Water samples [49], extraction with di-
chloromethane then in situ ethylation and collection of the
species on Tenax has been used for determination of
MeHg in forest soils by GCMIPAED [31].
Food samples. Historically, Hg in food products other
than seafood was not subjected to species analysis be-
cause of the lack of adequate methodology. The high
specificity and selectivity of GCMIPAED, and the min-
imum analyte isolation required by this method (irrelevant
matrix components being transparent to the detector)
have, however, enabled its use for the determination of
MeHg in thirty-two samples of grains, cereal products,
fruits, and vegetables [54]. After acid hydrolysis with HCl
the resulting chlorinated species were eluted from a Celite
545 sample homogenate column with dichloromethane,
the eluate was treated with stannic chloride, and the ana-
lyte was isolated from co-extracted materials by use of a
wide-bore capillary column for GCMIPAED.
Speciation of tin compounds
Organotin compounds found in environmental samples
(water, sediments, particulate matter and living organ-
isms) have both anthropogenic and natural origins. The
most important anthropogenic compounds are butyl- and
phenyltin species. Tributyltin (TBT), used mainly as an
antifouling compound in paints, is highly toxic to both
target and non-target marine organisms, and has recently
been included in the list of endocrine system disrupters
[55]. Dialkyltins with butyl and octyl groups are used as
polymer stabilizers in, for example, PVC pipes, from
which they can be released into water [56], or even into
air, in fires or during processing of plastic materials at
high temperature [57]. Dibutyltin (DBT) and mono-
butyltin (MBT) can also be generated in the environment
by degradation of TBT. Triphenyltin (TPhT), although
also used as an antifouling agent in paints, is mainly em-
ployed as a fungicide and biocide in agriculture [58]. Nat-
ural degradation of TPhT to diphenyltin (DPhT) and mono-
phenyltin (MPhT) is well documented. Nowadays, use of
all these compounds (especially TBT) is restricted, but
they are still present in marine environments (mainly in
sediments, from which they are slowly released into the
aquatic environment). Naturally occurring organotin com-
pounds are mainly biomethylated species generated from
inorganic tin (Sn4+) in sediments and estuarine environ-
ments. Natural methylation of butyltin compounds has
also been reported [59].
The toxicity of organotin compounds depends on the
number and kind of organic groups bound to the Sn atom.
Species analysis is therefore essential for understanding
the biological effects and environmental impact of these
compounds. Accurate identification and quantification of
tin species in environmental samples needs reliable ana-
lytical methods based on the combination of adequate
sample preparation procedures with selective, sensitive
analytical techniques. Most viable methods are based on a
separation step followed by quantification, by use of an
appropriate detector. Gas chromatographic separation is
more widely used than HPLC because of the higher re-
solving power of GC columns, the smaller requirement for
hazardous solvents, and the availability of detectors that
are more sensitive and more selective. Among these tech-
niques, GCMIPAED has been one of the most widely
used in the past 10 years.
Experimental conditions and analytical features
of GCMIPAED
Optimization of MIPAED conditions for the analysis of
organotin compounds was first performed by Lobinski et
al. [60], and later by several others (Table 2). Although
there are small differences among the measurement con-
ditions used in these studies, there is general agreement
that a helium make-up flow rate higher than 200 mL min1
is necessary to prevent the interaction of tin compounds
with the quartz discharge tube and to achieve maximum
sensitivity. The presence of oxygen in the helium plasma
was also found to be necessary to prevent deposition of
carbon on the walls of the discharge tube, even though
oxygen leads to the formation of refractory tin oxides and
a consequent decrease in the emission signal. The nega-
tive effect of oxygen can be compensated by addition of
hydrogen to the plasma gas; hydrogen probably produces
very volatile hydrides that are easily atomized and ex-
cited, thereby reducing the detection limit of the tech-
nique [61, 62].
Although some authors claim absolute detection limits
<0.2 pg of tin [60, 63, 64, 65], most authors have reported
values between 0.5 and 5 pg. Slight variations in the he-
lium make-up flow and reagent gas pressure are probably
largely responsible for these small differences. With re-
gard to the effect of the emission line selected, Tutschku
 81

Table 2 Optimum MIPAED

conditions, and absolute detec- Emission He make-up H2 pressure O2 pressure ADL Ref.
tion limits (signal/noise=3) for line (nm) (mL min1) (Kpa) (Kpa) (pg Sn)
organotin compounds 270.651 220 448 207 15 [69, 145, 146]
303.419 240 400 160 ca 25 [61]
303.419 270 345 152 ca 23 [99]
270.651 240 414 138 0.5 [100]
326.23 240 200 200 0.8 [62]
150 450 150 24 [67]
303.419 200 5 [147]
303.419 240 345 138 0.05 [60]
Not available 303.419 240 345 138 0.15 [64, 65]

et al. [62] obtained lower detection limits at 326.23 nm With regard to the linearity of response, several papers
than with the more usual 270.651 and 303.419 nm lines. have reported GCMIPAED calibration curves that are
Table 2 summarizes MIPAED conditions and detection linear over the range 10 to 2500 ng tin mL1, with corre-
limits for organotin compounds. lation coefficients >0.999 [68, 69]. The detection limits
Table 3 lists the detection limits achieved by GC-based noted above and the elemental selectivity of CGMIP
systems other than GCMIPAED for quantification of AED have made this technique the workhorse for quan-
organotin compounds. Those achieved by GCMIPAED tification of organotin compounds in environmental sam-
are similar to those of the best GC-FPD instruments and ples.
of GCMS systems with a quadrupole mass analyzer op- In the discussion below the possibilities and limitations
erating in SIM mode (ion-trap mass analyzers are a valu- of GCMIPAED for the speciation of organotin com-
able alternative to high-sensitivity scanning, and have de- pounds in a variety of different matrices are discussed.
tection limits similar to those of quadrupole instruments Some comments on sample-preparation and derivatization
in SIM mode) [66]. In GCMIPAED, however, all strategies are also made.
organotin compounds are quantified under identical con-
ditions, whereas with quadrupole SIM MS detection sev-
eral different masses must be monitored [67]. The combi- Water samples
nation of gas chromatography with AAS is less sensitive
for tin compounds than GCMIPAED, because of the The levels of organotin compounds in polluted waters
high atomization temperatures of tin and the possible for- (rivers, estuaries, harbors, shipyard areas, and tap water in
mation of refractory oxides [63]. Detection limits one or contact with certain polymeric materials) are below the
two orders of magnitude lower than those of CGMIP detection limits of GCMIPAED or any other gas chro-
AED have only been achieved with customer developed matography-based technique used in organotin species
combinations of gas chromatography with ICPMS detec- analysis. A concentration step is therefore necessary be-
tion. fore analytical determination. In general, concentration
strategies are very similar for all gas chromatographic
Table 3 Absolute detection limits (S/N=3) for organotin com- techniques used in the speciation of tin, Table 4.
pounds achieved by gas chromatography systems other than Classical approaches are based on the extraction of
GCMIPAED organotin compounds (mainly butyl and phenyl com-
pounds) using a complexing agent (tropolone or sodium
System Instrumental detection Ref.
limits (pg tin) diethyldithiocarbamate) in combination with a volatile or-
ganic solvent, followed by derivatization with a Grignard
GC-FPD 3 [148] reagent [60, 67]. A faster option is the use of NaBEt4, and
0.2 [149] more recently NaBPr4, as derivatization agent (the latter
24 [150] enables the investigation of ethyltin compounds in water
0.20.5 [151] samples) [61, 64, 70]; in both reactions alkyl derivatives
0.318 [152] are formed in the aqueous medium and extracted simulta-
GC-QF AAS 3371 [63] neously into an organic solvent. NaBEt4 has also been
10100 [153] used to derivatize polar organotin species previously re-
GCMS 110 [66, 67] tained on a solid extraction cartridge [65].
1.58 [154] Methyltin compounds can be ethylated in aqueous
samples using NaBEt4, and simultaneously extracted and
GCICPMS 0.050 [155]
concentrated with a purge-and-trap device; these injectors
0.0150.034 [156]
can use cryogenically cooled capillary traps [71] or solid
0.0520.17 [112]
adsorbents operating at room temperature [49]. Polar butyl-
0.050.08 [157]
tin species can be converted by treatment with NaBH4 into
82
Table 4 Summary of GCMIPAED procedures for the quantification of organotin compounds in water samples
Compounds Sample Extraction technique Derivatization Detection Ref.
volume reagent limits
(mL) (ng L1)
Butyl and phenyltin 50 LLE (hexane) NaBEt4 0.1b [64]
Butyl and phenyltin 50 SPE, C18a NaBEt4 0.1b [65]
Butyl, methyl and diphenyltin 100250 LLE (hexanetropolone) EtMgBr 1015 [147]
Butyl and methyltin 1500 LLE (pentaneNaDDTC) PeMgBr [60]
Butyl and phenyltin 100 LLE (hexane) NaBEt4 1733 [61]
Butyl and phenyltin 100 LLE (hexane) NaBPr4 312 [70]
Methyltin 10 P&T NaBEt4 0.15 [71]
Butyltin 4 Direct SPME (PDMS fibers) NaBEt4 ca. 20 [73]
aDerivatization of polar compounds on the stationary phase of the bCalculated for an injection volume of 25 L, using a PTV injector
adsorbent cartridge Not available

volatile compounds that can be purged from water sam-
ples, an option that has normally involved the use of a
packed column coupled to a QF-AAS instrument [72]; as
far as we are aware, analysis of butyltin compounds as hy-
drides by GCMIPAED has not been reported in the lit-
erature.
The possibility of direct derivatization in aqueous sam-
ples enables the use of solid-phase microextraction
(SPME) as a concentration technique [73, 74, 75, 76].
Non-polar fibers are normally used (mainly with polydi-
methylsiloxane as stationary phase). The sampling step
can be performed directly in the aqueous phase or in the
headspace over the sample. The latter procedure enables
faster equilibration between sample and fiber for volatile
and semivolatile compounds, and also more selective ex-
traction. Although methyl- and butyltin species are nor-
mally sampled at room temperature [77, 78], for less
volatile species such as phenyltin compounds better sensi-
tivity is achieved at high temperature [79].
Despite the high concentration factors achieved with
these techniques, and their low or at best medium selec-
tivity, GCMIPAED suffers from no interferences at tin
emission lines (271, 303 and 326 nm). Some problems
with blanks have, however, been reported. Szpunar et al.
[80], who used Grignard derivatization with pentylmagne-
sium bromide, observed organotin signals in blanks when
several microliters of the organic phase were introduced
into the chromatography column by use of a PTV injector.
Blanks with organotin signals have also been detected
when NaBEt4 was used as derivatization agent; possible
sources of this contamination are plastic holders, the
polypropylene body of SPE cartridges [65], and reagents
used in sample preparation, including the NaBEt4 [81].
Szpunar et al. also reported that the use of sodium di-
ethyldithiocarbamate (NaDDTC) as a complexing agent
and then Grignard derivatization of the organic extract led
to noisy baselines, probably because of reaction of the
Grignard reagent with decomposition products of NaDDTC
[80]. Stb et al. [67] using GCMS have identified MBT
and DBT, at levels of 512 ng L1, in blank chromato-
grams corresponding to the analysis of water. Finally,
Rosenberg et al. have reported variations of 3040% in
the analysis of butyltin compounds at low concentrations
when using HS SPME with GCMIPAED [82]. Apart
from problems described by the authors in relation to the
SPME fiber surface, it is also possible that blank signals
are also responsible for this lack of repeatability when
organotin compounds are analyzed at concentrations be-
low the g L1 level.
Solid matrices
Sample preparation procedures for determination of organo-
tin compounds in solid matrices by gas chromatography
comprise several steps, the precise number of which de-
pends on the derivatization technique (Grignard reaction,
direct alkylation with sodium tetraalkylborates in aqueous
solution, or hydride generation) [83] and on whether a
clean-up step is needed, which in turn depends on the de-
rivatization technique (alkylborates normally lead to
cleaner extracts than Grignard reactions, which require
the use of complexing reagents such as tropolone [84]),
on the kind of sample, and above all on the selectivity of
the GC detector. For GCMIPAED systems, concentra-
tion of the final extracts from solid matrices is not nor-
mally necessary, because instrument sensitivity matches
target detection limits for environmental samples (approx.
1 ng g1) [85].
Sediment samples
Because organotin compounds are not involved in miner-
alogical processes and bind only to the surface of sedi-
ments, complete dissolution of the matrix is not consid-
ered necessary. Acid leaching combined with mechanical
shaking [86] or sonication [77], and microwave-assisted
leaching [87, 88], are, therefore, the basic approaches used
to release organotin compounds from sediments. Leach-
ing can be achieved by use of dilute acids with complex-
ing agents (tropolone or sodium diethyldithiocarbamate)
dissolved in an organic solvent [84, 89], or using only
aqueous or methanolic solutions of acetic or hydrochloric

acids [88]; normally the former choice is followed by
Grignard derivatization and the latter by direct alkylation
with NaBEt4 or NaBPr4. Methods based on hydride gen-
eration have been shown to be poorly efficient for extracts
with high sulfur, hydrocarbon, or inorganic metal content,
and are therefore not advisable for sediment analysis [72].
The element selectivity of GCMIPAED makes it un-
necessary to perform the time-consuming desulfuration pro-
cedures described in the literature for the determination of
organotin species in sediments by GC-FPD [90, 91]. In gen-
eral, no clean-up procedures are necessary with GCMIP
AED, although Ceulemans et al. [89] have recommended
that if tropolone is used as complexing reagent the final or-
ganic extract should be passed over basic alumina, or ex-
cess tropolone can contaminate the head of the column.
The analytical characteristics (sensitivity and selectiv-
ity) of GCMIPAED are, in principle, good enough to
solve the detection step in the analysis of tin compounds
in sediments. Problems mainly arise in sample prepara-
tion [92] and in the limited number of reference materials
(currently only reference sediment material PACS-2, with
certified concentrations of MBT, DBT, and TBT, is com-
mercially available). As a consequence, validation of ana-
lytical procedures for phenyltin species is a difficult task.
Some of the problems related with the analysis of organo-
tin compounds (mainly MBT and phenyltin species) in
sediments are well known and clearly described in the lit-
erature. Few sample extraction procedures for MBT claim
quantitative yields, and recoveries are sometimes esti-
mated by use of spiked samples [84, 93]. In real polluted
sediments interaction of MBT with the matrix, which is
directly related to sample sulfur and organic carbon con-
tent [84], is probably even stronger than in the spiked
samples, and the extraction yield accordingly lower. Be-
cause of its structure and polarity, similar behavior is ex-
pected of MPhT. Another problem is encountered in the
extraction of phenyltin compounds these species are
partially degraded by some leaching procedures designed
for butyltin compounds. Degradation is of special signifi-
cance when acid conditions are combined with high tem-
peratures. Despite their lability during sample prepara-
tion, however, surveys of phenyltin compounds, espe-
cially TPhT, in sediments are of environmental interest,
because of their toxicity and their persistence under the
anaerobic conditions obtaining in sediments [67].
Both problems (non-quantitative extraction and degra-
dation) could be compensated by use of isotope-labeled
organotin compounds as internal standards. If the label is
introduced in the tin atom [94] the excellent sensitivity
and multielemental capabilities of ICPMS might justify
marketing an interface between GC and ICPMS as a
complementary technique to GCMIPAED for organo-
tin speciation (and for volatile organometallic speciation
in general) [3]. A similar effect could be achieved with
13C or deuterated organotin standards; with these, deter-
mination can be achieved by use of conventional quadru-
pole GCMS [95], a technique available in most analyti-
cal chemistry laboratories which is sufficiently sensitive
for the analysis of organotin compounds in solid matrices.
83
Some authors have found that GCMIPAED chro-
matograms for some sediment samples show not only
methyl-, butyl-, and phenyltin compounds but also peaks
corresponding to minor organotin species [96]. Identifica-
tion of these species is the first step towards understand-
ing their origin and environmental behavior. GCMS is
currently the only hyphenated technique capable of pro-
viding information about their structures. Because the de-
tection limits of quadrupole mass analyzers, working in
scan mode, are still far above those achieved with MIP
AED, the use of ion-trap MS instruments is advisable
[66].
Biological tissues
Speciation of tin compounds in fish and other seafood tis-
sue has been studied for the past 15 years, since their neg-
ative effects on the reproduction of marine organisms
such as oysters and other mollusks were discovered [97,
98]. Nowadays, two reference materials are available
NIES-11 (fish tissue from the National Institute for Envi-
ronment Studies, Japan) and BCR-477 (mussel tissue pre-
pared by the former BCR). The concentration of TBT in
the former is certified and an approximate level is also in-
dicated for TPhT; the latter contains MBT, DBT, TBT,
MPhT, and TPhT, at least, although only the concentra-
tions of MBT, DBT, and TBT are certified.
As for sediments, sample preparation is usually per-
formed in several steps, the first of which is sample di-
gestion. Because organotin compounds can be incorpo-
rated in tissues, complete sample decomposition is a pre-
requisite for quantitative extraction. Acid solutions (mainly
acetic acid and dilute hydrochloric acid), modified super-
critical CO2, and mixtures of lipase and protease have all
been used to release organotin compounds from biologi-
cal tissues, but highest yields seem to be achieved with
aqueous TMAH solutions [99, 100, 101, 102, 103]. The
second step is derivatization and extraction of the organo-
tin derivatives into an organic solvent (except for hydride-
generation techniques, when cryogenic purge-and-trap de-
vices are used). Fast, integrated sample-preparation pro-
cedures have been described in which tissue digestion,
derivatization (ethylation), and extraction of the ethylated
derivatives into an organic solvent occur simultaneously
in only 3 min under the action of a microwave field [83].
The problems encountered in the sample preparation
step are similar to those reported for sediment samples:
sub-quantitative recovery and the degradation of phenyltin
compounds [101, 104]. TPhT seems to be more stable at
basic pH [79, 99] than in acid media [101], however,
which is another reason for preferring solubilization with
TMAH to acid leaching.
Irrespective of the digestion and derivatization agents
used, the final extracts from biological samples are col-
ored solutions containing large amounts of lipids. Clean-
up of the organic extract containing the derivatized organo-
tin species has therefore been recommended, to prevent
column deterioration and to eliminate baseline drift with
84
Fig. 1 GCMIPAED chromato-
grams obtained from a mussel-tissue
extract, with clean-up over alumina
(solid line) and without clean-up
(dotted line). Extraction was per-
formed as described elsewhere [99].
1. TPT (IS), 2. MPhT, 3. TPhT.
A. 248 nm carbon emission line.
B. 271 nm tin emission line
increasing oven temperature when relatively poorly selec-
tive detectors such as scan-mode MS or FPD are used. Po-
lar adsorbents such as silica [100], alumina [80, 99] and
Florisil [105, 106] are used for this purpose. Szpunar et al.
have shown that with CGMIPAED a stable, flat base-
line is obtained even without clean-up, although they also
noted that the response to ethylated TPhT depended on
whether or not clean-up had been performed [80]. In our
laboratory we have had the same problem with the deter-
mination of TPhT in biological materials; the 248 nm car-
bon channel, monitored simultaneously with the 271 nm
tin line, shows a large, broad signal at the retention time
of TPhT (Fig. 1). The species responsible for this non-spe-
cific signal (probably lipids) might temporarily modify
the internal wall of the quartz discharge tube, causing loss
of sensitivity in tin emission signals.
Speciation of lead compounds
Organolead compounds are ubiquitous pollutants in air,
atmospheric aerosols, water, and sediments. Tetraalkyllead
species (TAL; mainly Et4Pb, Me4Pb, and MenEt4nPb),
still used in some countries as anti-knocking additives in
gasoline [107], can penetrate the skin and biological
membranes and are readily absorbed via the lungs. The
toxicity of organolead compounds depends on the organic
groups bound to the lead atom methylated species are
less toxic than the corresponding ethylated compounds,
but are more stable and volatile [108].
TAL enter the atmosphere via vehicle exhaust pipes
and as a result of accidental spillage. There they are de-
composed to inorganic lead (Pb2+) via tri- and dialkyl
compounds. Photolysis and radical reactions are responsi-
ble for the fast decomposition of Me4Pb and Et4Pb
(t1/2=41 and 8 h, respectively). Ionic organolead species
(Me3Pb+, Et3Pb+, Et2Pb2+, and Me2Pb2+) are less toxic
than tetraalkylated species but have longer half-lives, en-
abling them to be transported considerable distances from
their anthropogenic sources [109].
The combination of gas chromatography with spectro-
scopic techniques (AAS, MIPAED, MS or ICPMS) is
common in the speciation of lead compounds. As for tin,
the separation of lead compounds by gas chromatography
requires prior derivatization of ionic alkylated species to
thermally stable volatile compounds. This reaction can be
 85

performed with Grignard reagents after chelation of these Analytical features of GCMIPAED
species and extraction into an organic solvent such as
hexane or pentane. Another option is to use borate reagents. Detection limits obtained in the speciation of lead com-
Although sodium tetraethylborate is only useful for meth- pounds by use of a variety of different GC-based hyphen-
ylated lead compounds (both inorganic and ethylated lead ated techniques are compared in Table 5. The detection
species yield Et4Pb [110]), Pawliszyn et al. have recent- limits of GCMIPAED are similar to those reported for
ly overcome this problem by using deuterium-labeled GCICPMS, and between two and three orders of mag-
NaBEt4 [111]; because the isotope-labeled ethyl group is not nitude better than those obtained using GCMS systems
present in environmental samples, it can be used to distin- or custom-made combinations of GC with AAS.
guish between ethylated inorganic lead and native ethyl- The GCMIPAED system enables measurement of
lead species. Lower-cost alternative derivatization reagents lead emission at 261.418 nm and 405.783 nm. The latter
are sodium tetrapropylborate [10, 112] and tetrabutylam- line is preferred because of its higher sensitivity, which is
monium tetrabutylborate [113, 114], both of which enable maximum with a helium make-up flow of approximately
simultaneous determination of most of the ionic alkyllead 300330 mL min1 [114, 116, 117, 118]. Oxygen and hy-
(IAL) species commonly found in environmental samples drogen are also necessary as reagent gases; oxygen pre-
(Me3Pb+, Et3Pb+, and MenEtmPb(4nm)+) [115]. vents carbon from being deposited on the walls of the dis-
charge tube (oxygen pressure is normally adjusted to 138
kPa (20 psig)) and, as for tin, the presence of hydrogen
improves sensitivity (pressures of 550650 kPa are re-
ported as optimum [116, 119, 120]).
With the 405.783 nm emission line, good linearity has
Table 5 Absolute detection limits (S/N=3) for organolead com- been obtained for injected amounts of organolead species
pounds achieved by GC-based hyphenated techniques
between 0.1100 pg, as Pb. If very high levels of organo-
Hyphenated Absolute detection limits Ref. lead species reach the plasma, lead can condense on the
technique (pg as lead) wall of the discharge tube, causing peak tailing and re-
GCMIPAED 0.031 [107, 116] duced sensitivity [116]. Because more than 99.9% of the
0.01 [123] lead in most samples is Pb2+ [121], and part of this inor-
0.040.08 [114] ganic lead can be derivatized and extracted with the native
organolead species, it is advisable to vent the chromato-
GCICPMS 0.010.016 [122]
graphic peak corresponding to native Pb2+, to prevent con-
0.7 [158]
tamination of the discharge tube [120].
0.040.09 [112]
GCAAS 4095 (Flame) [125]
3045 (Quartz furnace) [47] Water and other aqueous samples
12 (Quartz furnace) [159]
Because of their short half-lives and low solubility in wa-
GCMS 2 [110] ter, the presence of TAL compounds in aqueous samples
4 [130]
is not common [107]. The levels of IAL species in rain,
78 [131]
surface, and tap water (mainly dimethyl-, trimethyl-, di-

Table 6 Speciation of organolead compounds in water and other aqueous samples

Sample volume Extraction Derivatization Detection technique Detection limits Ref.
(mL) procedure reagent (ng L1)
100150 LLE hexane PrMgCl PTV GCMIPAED 0.1 [120]
75120 g (snow) LLE hexane PrMgCl PTV GCMIPAED 10 fg g1 [123]
6080 (wine) LLE hexane PrMgCl GCMIPAED 13 [119]
20 LLE hexane Bu4N+ Bu4B PTV GCMIPAED 0.040.08 [114]
100 LLE hexane PrMgCl GCMIPAED 0.50.9 [117]
25 SPME NaBEt4 GCICPMS 0.1 [74]
10 (urine) SPME NaBEt4 GCMSMS 7 [126]
100 LLE hexane NaBEt4 GCMS 2.5 [110]
20 SPME NaBEt4 GCMS 83130 [111]
500 LLE pentane BuMgCl GCQF AAS 12.2 [128]
1000 LLE pentane Bu4N+ Bu4B GCQF AAS 1030 [113]
25 SPE PrMgCl GCMS 14 [115]
50 Purge and trap NaBEt4 GCQF AAS [121]
50 LLE hexane PrMgCl GCICPMS 0.050.08 [122]
PTV: programmed temperature vaporization
Not available
86
ethyl-, and triethyllead) are at most a few ng L1 [47, 117,
118, 121, 122]. The concentration of organolead com-
pounds in polar snow is even lower, although it is indica-
tive of the atmospheric distribution of these pollutants and
of the world production of leaded petrol [120, 123, 124].
Because of these low concentrations, the first step in the
quantification of organolead compounds in aqueous sam-
ples is enrichment. Table 6 lists reported sample prepara-
tion procedures. The earliest concentration techniques
were based on liquidliquid extraction of IAL species
with hexane or pentane after complexation with sodium
diethyldithiocarbamate at pH 89; this was followed by
derivatization with Grignard reagents such as propyl- or
butylmagnesium chloride [117, 120, 122, 123, 124]. Si-
multaneous alkylation and extraction of IAL into organic
solvents can be performed by use of NaBEt4 [110, 111],
NaBPr4 [10, 112] or Bu4NBBu4 [113, 114], for all of
which the highest yield is achieved at pH 4. Irrespective
of whether a Grignard reagent or a borate is used, inor-
ganic lead present in the sample is co-extracted and de-
rivatized; to prevent large quantities of alkylated inor-
ganic lead from contaminating the MIP discharge tube (or
the ion source of GCMS systems), inorganic lead is nor-
mally masked with EDTA before extraction of IAL com-
pounds.
With initial sample volumes of 100200 mL and the
injection of 2550 L of the final organic extract into the
chromatographic column, GCMIPAED achieves detec-
tion limits of the order of 0.1 ng L1 (Table 6). At these
low concentrations, however, the accuracy and precision
of the analysis can easily be disturbed by blank signals
and the presence of artifacts. In fact, organolead com-
pounds are very often detected in blanks because of con-
tamination and transalkylation reactions. Steps can be
taken to prevent this:
For samples with concentrations below 0.1 pg g1, sam-
ple preparation should be performed in clean cabinets
to avoid contamination with airborne IAL [123].
Concentrated standards should be prepared and stored
in a different room from the sample [107].
Impurities in reagents (buffer solutions and complexing
and masking agents) can be removed by pre-extraction
with hexane [116, 117, 119, 123].
Derivatization reagents are another possible source of
problems. Purification of Grignard reagents is very dif-
ficult because of their low stability. PrMgCl is normally
preferred to BuMgCl and PeMgCl for quantification of
organolead compounds by GCMIPAED because it
causes less baseline disturbance [120].
If Grignard derivatization is used, Pb2+ must be masked
before extraction of organolead compounds because oth-
erwise it will react with the Grignard reagent, giving not
only the main product (e.g. Pr4Pb, if PrMgCl is used) but
also minor compounds such as MePr3Pb and EtPr3Pb,
which are formed by transalkylation reactions [80]. For
the alkylborate derivatization reagents NaBEt4, NaBPr4,
and Bu4NBBu4, which also react quantitatively with both
IAL and inorganic lead [111, 112, 121], we know of no re-
ports of transalkylation reactions; Pb2+ is, nevertheless,
usually masked with EDTA to reduce consumption of bo-
rate [110, 114]. The presence of EDTA does not affect the
derivatization of IAL because only inorganic lead forms
stable chelates [113].
Solid-phase procedures can be used as an alternative to
liquidliquid extraction for the quantification of lead
species in water samples [47, 115, 125]; adsorbents con-
taining dithizone or dithiocarbamate groups are usually
used. Purge-and-trap and SPME are also alternatives that
can be used after in-situ derivatization with borate
reagents [74, 111, 112, 121, 126, 127]; both techniques re-
quire smaller volumes of sample than LLE (Table 6) and
airborne contamination is also avoided, because samples
are processed in closed vessels; contamination problems
arising from derivatization reagents cannot, however, be
circumvented by use of these approaches.
Air samples
Determination of IAL and TAL compounds in the atmo-
sphere, in which they are found at levels below 1 ng m3,
requires the concentration of large volumes of air. Filters,
solid adsorbents such as Tenax, Porapak Q, and Amber-
lites, water-filled gas bubblers, and cryogenic traps have
all been used to extract organolead compounds from at-
mospheric aerosol and gas phases [108, 109, 128, 129].
Because of the possibility of decomposition of TAL com-
pounds in the presence of ozone, a layer of ferrous sul-
phate is normally placed before to the trap [129].
Solid samples
In addition to water and air, organolead compounds have
been found in soil, sediments, road dust, grass, and tree
leaves. Wet atmospheric deposition, direct contamination
(in the vicinity of emission sources) and absorption from
contaminated waters through plant roots are the mecha-
nisms responsible for the presence of alkyllead species in
these samples. The excellent sensitivity and selectivity of
GCMIPAED for lead compounds enables quantifica-
tion of levels in these samples without any need for high
concentration factors or extensive clean-up procedures.
Because organolead species are labile, however, care must
be taken to extract them without modifying their chemical
structure.
The extraction of lead compounds from sediments and
particulate matter can be performed under mild conditions
by use of organic solvents in the presence of complexing
agents [107, 130], supercritical CO2 modified with
methanol [131], buffer solutions of pH 4, or direct disso-
lution in water containing NaCl [121]. Most of these pro-
cedures give satisfactory yields for Me3Pb+ and Et3Pb+,
but not for dialkyllead species. Validation of the extrac-
tion procedure is, moreover, only possible for Me3Pb+, for
which there is a road dust reference material (BCR 605,
containing 7.91.2 g kg1 of Me3Pb+).

With regard to biological samples, TMAH has been
used as a tissue solubilizer for the speciation of lead com-
pounds in grass and tree leaves. The digestion step takes
several hours, and unless performed at low temperature
affords very low yields of dialkyllead compounds, pre-
sumably because of their transformation into inorganic
lead [132]. Enzymatic hydrolysis of biological tissues has
also been used samples are treated with a mixture of li-
pase and protease at 37 C for 2448 h [107]. SPME has
been described as a promising technique for the analysis
of Pb2+ and organolead compounds in urine and blood
after precipitation of erythrocytes (from blood), samples
are diluted with water and adjusted to pH 4, after which
lead compounds are derivatized in-situ and adsorbed on to
a 100 m polydimethylsiloxane-coated fused-silica SPME
fiber [126, 133].
Simultaneous analysis of organometallic compounds
by GCMIPAED
Although GCMIPAED is not really a multielemental
technique, it nevertheless enables simultaneous monitor-
ing of several emission lines, corresponding to one or sev-
eral elements, as long as they all fall within the spectral
region covered by the photodiode-array spectrophotome-
ter, which has a breadth of ca. 40 nm. Volatile lead, mer-
cury and tin compounds can therefore be analyzed in a
single chromatographic run by using the emission lines at
261.418, 253.652, and 270.651 nm, and because similar
sample preparation procedures are used for organometal-
lic compounds of these three elements (especially for wa-
ter samples), their simultaneous determination by GC
MIPAED is possible. It should, nevertheless, be empha-
sized that the multielemental detection mode is less sensi-
tive than the individual determination of each group of
compounds. The first reason for this is that the 261 nm
lead line is less sensitive than the 406 nm line. Secondly,
and more importantly, optimum make-up flows are differ-
ent for mercury, tin, and lead (ca. 50, 250, and 300 mL
min1, respectively) [36]. Tin and lead compounds interact
with the walls of the plasma discharge tube, making high
flows desirable to minimize this interaction; for mercury
species, for which no such interaction has been described,
high flows simply increase dilution and so reduce sensi-
tivity. There are similar differences among optimum hy-
drogen pressures, which are higher for tin and lead than
for mercury [36].
Despite these problems, applications of GCMIP
AED to simultaneous determination of Hg, Sn, and Pb
species have been described in the literature. Ceulemans
et al. [71] found that methyl species of mercury, lead, and
tin can be derivatized and purged from water samples un-
der similar conditions (ethylation with NaBEt4 at pH 5,
purge time 10 min), and proceeded to perform simultane-
ous quantification of tin, lead, and mercury methyl deriv-
atives in water samples by use of a cryogenic purge-and-
trap device. The MIPAED working conditions (helium
make-up flow and hydrogen pressure) were those previ-
87
ously found to be optimum for tin, and detection limits
lower than 1 ng L1 of metal were achieved for all com-
pounds with a sample intake of 10 mL. Schubert et al.
[70] have described the use of NaBEt4 and NaBPr4, then
liquidliquid extraction with hexane, for the simultaneous
determination of organotin (butyl and phenyl compounds)
and organolead (Et3Pb+ and Me3Pb+) in water samples;
the two groups of compounds were analyzed at 270.651
and 261.418 nm, and the 247.857 nm carbon line was also
monitored to obtain information about possible interfer-
ences from organic compounds. Finally, Minganti et al.
[36] have shown that lead, tin, and mercury speciation in
water and biological tissues is possible after extraction
and simultaneous derivatization with n-pentylmagnesium
bromide.
A low-cost GCMIPPED system
for rapid species analysis
The first commercial GCMIPAED instrument, mar-
keted by HewlettPackard in 1989, has enjoyed great suc-
cess in the speciation of volatile compounds. For some
years, an improved version of the same system has also
been available. This hyphenated technique is, however,
still beyond the financial means of most environmental
control laboratories, which need cheap, fast, reliable in-
strumentation for daily monitoring of volatile organo-
metallic species included in international lists of priority
pollutants. Accordingly, with the financial support of the
European Union, efforts have been made to develop a
low-cost instrument designed specifically for species
analysis of volatile compounds. The proposed system, the
automated speciation analyser (ASA), is based on a com-
bination of a multicapillary (MC) column (made of a bun-
dle of approximately nine hundred 1 m40 m i.d. capil-
laries housed in an isothermal oven) with a helium MIP
and a photoemission detector (PED) [134]. The analytical
performances of each component of this system, as de-
scribed in earlier papers, are described briefly below.
With regard to the separation step, the potential of MC
columns for speciation analysis has been demonstrated by
Lobinski et al. [135]. Theoretically, the efficiency of such
columns is a function of the internal diameter of each sin-
gle capillary, and sample capacity is determined by the
number of capillaries. For the MC column described
above maximum efficiency was achieved at flow rates be-
tween 50 and 150 mL min1, so the use MC columns with
a length of only one meter (rendering a large number of
theoretical plates per unit length), in combination with
high carrier gas flows, produced fast chromatographic
separations. The use of an isothermally operated MC col-
umn with the HewlettPackard MIP AES detector en-
abled quantification of tin, lead, and mercury species in
less than 1 min with detection limits equal to or better
than those achieved with a conventional 30 m0.32 mm
i.d. capillary column [136, 137]. Volatile organometallic
species are introduced at the head of the MC column as a
narrow band using a split injector or a custom-developed
88
cryogenic purge-and-trap injector [138, 139]. These pro-
cedures have been validated by analysis of reference ma-
terials (sediments, biological tissues, urine, and petrol)
with certified levels of tin, lead, and mercury compounds.
Other papers have reported the possibility of using iso-
thermal columns as short as 522 cm combined with
ICPMS for analysis of mercury and tin species [140,
141].
When the chromatographic separation stage had been
simplified and accelerated, attention was focused on the
selection of an inexpensive but reliable detector. The
choice made was an MIPPED. The helium MIP device
consists basically of a surfactron cavity, a rectangular
wave-guide in which waves can propagate only in the
transverse electronic mode (whereas the HewlettPackard
MIPAED system uses a Beenakker TM010 cavity), and a
ceramic discharge tube needing no cooling [142]. Dis-
crimination between plasma radiation and emission by an
analyte element is achieved by use of an oscillating nar-
row-band interference filter followed by a photomultiplier
tube [143]. This MIPPED assemblage is much cheaper
than the polychromator and photodiode-array spectrome-
ter used in the commercial MIPAED system. In combi-
nation with conventional GC columns the MIPPED sys-
tem had already been used for quantification of mercury
compounds [144]. In combination with MC columns, it
has achieved absolute detection limits of 34 pg Hg in the
determination of MeHg+ and Hg2+ in reference material
TORT-2 [134]. Further applications and the commercial
success of this low cost, dedicated, speciation system de-
pends on many factors, among them the development of a
range of interference filters for such elements as tin, lead,
and selenium which offer adequate flexibility, in particu-
lar with regard to maximum transmission wavelength and
bandwidth [134].
Conclusions
For quantification of volatile Hg, Sn, and Pb species in
environmental samples the performance of GCMIP
AED is currently unmatched, and only for water samples
is pre-concentration of the analytes necessary before
quantification. Nowadays, only GCICPMS gives simi-
lar or better detection limits than GCMIPAED in the
speciation of these three elements.
The ASA, a low-cost instrument designed for daily
monitoring of organometallic Hg, Sn, and Pb species in
control laboratories, and based on combination of a multi-
capillary column with a photoemission detector, has
achieved performance similar to that of the commercial
GCMIPAED system for Hg, but has yet to be validated
for Sn and Pb by use of certified reference materials cor-
responding to environmental samples.
The precision and accuracy of species-selective analy-
sis, especially for phenyltin and organolead species, is
limited by degradation problems and the formation of ar-
tifacts during sample preparation. The easiest way to
overcome these problems is to use isotope-labeled stan-
dards and MS detection. We regard this as the main ad-
vantage of GCICPMS over GCMIPAED and other
atomic emission detection systems for quantification of
mercury, tin, and lead species.
Acknowledgements Financial support from the Spanish DGICT
(project REN 20000984HIP) is acknowledged.
References
1. Guevremont R, Sturgeon RE (2000) J Anal At Spectrom 15:
3742
2. Adams F, Slaets S (2000) Trends Anal Chem 69:24112417
3. Lobinski R (1998) Spectrochim Acta Part B 53:177185
4. Gurka DF, Pyle S, Titus R (1997) Anal Chem 69:24112417
5. Sullivan JJ (1991) Trends Anal Chem 10:2326
6. Sullivan JJ, Quimby D (1992) In: Uden PC (ed) ACS Sympo-
sium Series, Element-specific chromatographic detection by
atomic emission spectroscopy. American Chemical Society,
Washington, DC, p 62
7. Lobinski R, Adams FC (1997) Spectrochim Acta Part B 52:
18651903
8. Horvat M, Wilken RD (2000) Fresenius J Anal Chem 366:
415416
9. EPA-452/R-97007 Mercury study report to congress. Vol V:
Health effects of mercury and mercury compounds. Office of
Air Quality Planning and Standards and Office of Research
and Development. US EPA. December 1997
10. Hempel M, Kuballa J, Jantzen E (2000) Fresenius J Anal
Chem 366:470475
11. Hight SC, Capar SG (1983) J Assoc Off Anal Chem 66:1121
1128
12. Hight SC, Corcoran MT (1987) J Assoc Off Anal Chem 70:
2430
13. Ceulemans M, Adams FC (1994) LCGC Int 7:694697
14. Olsen KB, Sklarew DS, Evans JC (1985) Spectrochim Acta
Part B 40:357365
15. Quevauviller P (1996) J Chromatogr A 750:2533
16. Horvat M (1999) Chemosphere 39:11671179
17. Rub E, Lorenzo RA, Casais C, Carro AM, Cela R (1992)
J Chromatogr 605:6980
18. Panaro KW, Erickson D, Krull IS (1987) Analyst
112:10971105
19. Kato T, Uheiro T, Yasuhara A, Morita M (1992) J Anal At
Spectrom 7:1518
20. Carro-Daz AM, Lorenzo-Ferreira RA, Cela-Torrijos R
(1994) J Chromatogr A 683:245252
21. Frech W, Snell JP, Sturgeon RE (1998) J Anal At Spectrom
13:13471353
22. Orellana-Velado NG, Pereiro R, Sanz-Medel A (1998) J Anal
At Spectrom 13:905909
23. Orellana-Velado NG, Pereiro R, Sanz-Medel A (2000) J Anal
At Spectrom 15:4953
24. Cai Y, Monsalud S, Jaffe R, Jones RD (2000) J Chromatogr A
876:147155
25. West G (1966) Acta Chem Scand 20:21312137
26. West G (1967) Acta Chem Scand 21:17901800
27. Bache CA, Lisk DJ (1971) Anal Chem 43:950952
28. Talmi Y (1975) Anal Chim Acta 74:107117
29. Emteborg H, Bjrklund E, dman F, Karlsson L, Mathiasson
L, Frech W, Baxter DC (1996) Analyst 121:1929
30. Donais MK, Uden PC, Schantz MM, Wise SA (1996) Anal
Chem 68:38593866
31. Qian J, Skyllberg U, Tu Q, Bleam WF, Frech W (2000) Fre-
senius J Anal Chem 367:467473
32. Behlke MK, Uden PC, Schantz MM (1996) Anal Commun
33:9192
33. Snell JP, Frech W, Thomassen Y (1996) Analyst 121:1055
1060

34. Mothes S, Wennrich R (1999) J High Resol Chromatogr 22:
181182
35. OReilly JE (1982) J Chromatogr 238:433444
36. Minganti V, Capelli R, De Pellegrini R (1995) Fresenius J
Anal Chem 351:471477
37. Bulska E, Baxter DC, Frech W (1991) Anal Chim Acta 249:
545554
38. Palmieri HEL, Leonel LV (2000) Fresenius J Anal Chem 366:
466469
39. Swan HB (1998) Bull Environ Contam Toxicol 60:511518
40. Emteborg H, Hadgu N, Baxter DC (1994) J Anal At Spectrom
9:297302
41. Gerbersmann C, Heisterkamp M, Adams FC, Broekaert JAC
(1997) Anal Chim Acta 350:273285
42. Vzquez MJ, Carro AM, Lorenzo RA, Cela R (1997) Anal
Chem 69:221225
43. Abun M, Carro AM, Lorenzo RA (2000) J Chromatogr A
889:185193
44. Jimnez MS, Sturgeon RE (1997) J Anal At Spectrom 12:597
601
45. Emteborg H, Baxter DC, Frech W (1993) Analyst 118:1007
1013
46. Emteborg H, Baxter DC, Sharp M, Frech W (1995) Analyst
120:6977
47. Johansson M, Emteborg H, Glad B, Reinholdsson F, Baxter
DC (1995) Fresenius J Anal Chem 351:461466
48. Hnstrm S, Briche C, Emteborg H, Baxter DC (1996) Ana-
lyst 121:16571663
49. Reuther R, Jaeger L, Allard B (1999) Anal Chim Acta 394:
259269
50. Carro AM, Mejuto MC (2000) J Chromatogr A 882:283307
51. Snell J, Qian J, Johansson M, Smit K, Frech W (1998) Ana-
lyst 123:905909
52. Mena ML, McLeod CW (1996) Mikrochim Acta 123:103
108
53. Mena ML, McLeod CW, Jones P, Withers A, Minganti V,
Capelli R, Quevauviller P (1995) Fresenius J Anal Chem 351:
456460
54. Madson MR, Thompson RD (1998) J Assoc Off Anal Chem
Int 81:808816
55. Takeuchi M, Mizuishi K, Hobo T (2000) Anal Sci 16:349
359
56. Becker G, Janak K, Colmsj A, stman C (1997) J Chro-
matogr A 775:295306
57. Vainiotalo S, Hyri L (1990) J Chromatogr 523:273280
58. Attar KM (1996) Appl Organomet Chem 10:317327
59. Rapsomanikis S, Harrison RM (1988) Appl Organomet Chem
2:151157
60. Lobinski R, Dirkx W, Ceulemans M, Adams FC (1992) Anal
Chem 64:159165
61. Girousi S, Rosenberg E, Voulgaropoulos A, Grasserbauer M
(1997) Fresenius J Anal Chem 358:828832
62. Tuschku S, Mothes S, Dittrich K (1994) J Chromatogr A
683:269276
63. Dirkx W, Lobinski R, Adams FC (1993) Anal Sci 9:273278
64. Ceulemans M, Lobinski R, Dirkx W, Adams FC (1993) Fre-
senius J Anal Chem 347:256262
65. Szpunar J, Ceulemans M, Lobinski R, Adams FC (1993) Anal
Chim Acta 278:99113
66. Stab JA, Brinckman UAT, Cofino WP (1994) Appl Organo-
met Chem 8:577585
67. Stb JA, Cofino WP, van Hattum B, Brinkman UAT (1993)
Fresenius J Anal Chem 347:247255
68. David F, Sandra P (1992) Analusis 20:5354
69. Liu Y, Lopez-Avila V, Alcaraz M, Beckert WF (1993) J High
Resol Chromatogr 16:106112
70. Schubert P, Rosenberg E, Grasserbauer M (2000) Fresenius J
Anal Chem 366:356360
71. Ceulemans M, Adams FC (1996) J Anal At Spectrom 11:201
206
72. Leroy MJF, Quevauviller P, Donard OFX, Astruc M (1998)
Pure Appl Chem 70:20512064
89
73. Tutschku S, Mothes S, Wennrich R (1996) Fresenius J Anal
Chem 354:587591
74. Moens L, De Smaele T, Dams R, Van Den Broeck P, Sandra
P (1997) Anal Chem 69:16041611
75. Lespes G, Desauziers V, Montigny C, Potin-Gautier M (1998)
J Chromatogr A 826:6776
76. Aguerre S, Montigny C, Lespes G, Potin-Gautier M (2000)
Analyst 125:263268
77. Millan E, Pawliszyn J (2000) J Chromatogr A 873:6371
78. Morcillo Y, Cai Y, Bayona JM (1995) J High Resol Chro-
matogr 18:767770
79. Vercauteren J, De Meester A, De Smaele T, Vanhaecke F,
Moens L, Dams R, Sandra P (2000) J Anal At Spectrom 15:
651656
80. Szpunar J, Ceulemans M, Dirkx W, Witte C, Lobinski R,
Adams FC (1994) Mikrochim Acta 113:287298
81. Tao H, Rajendram RB, Quetel CR, Nakazato T, Tominaga M,
Miyazaki A (1999) Anal Chem 71:42084215
82. Haberhauer-Troyer C, Crnoja M, Rosenberg E, Grasserbauer
M (2000) Fresenius J Anal Chem 366:329331
83. Rodriguez Pereiro I, Schmitt VO, Szpunar J, Donard OFX,
Lobinski R (1996) Anal Chem 68:41354140
84. Ceulemans M, Adams FC (1995) Anal Chim Acta 317:161
170
85. Dirkx W, Lobinski R, Adams FC (1994) Anal Chim Acta
286:309318
86. Dauchy X, Cottier R, Batel A, Borsier M, Astruc A, Astruc M
(1994) Environ Technol 15:569576
87. Rodrguez Pereiro I, Wasik A, Lobinski R (1999) Fresenius J
Anal Chem 363:460465
88. Szpunar J, Schmitt VO, Donard OFX, Lobinski R (1996)
Trends Anal Chem 15:181187
89. Ceulemans M, Slaets S, Adams F (1998) Talanta 46:395405
90. Schubert P, Fernndez-Escobar I, Rosenberg E, Bayona JM
(1998) J Chromatogr A 810:245251
91. Fernndez-Escobar I, Gibert M, Messeguer A, Bayona JM
(1998) Anal Chem 70:37033707
92. Abalos M, Bayona JM, Compa R, Granados M, Leal C,
Prat MD (1997) J Chromatogr A 788:149
93. Chau YK, Yang F, Maguire RJ (1996) Anal Chim Acta 320:
165169
94. Ruiz-Encinar J, Garca-Alonso JI, Snz-Medel A (2000) J Anal
At Spectrom 15:12331239
95. Arnold CG, Berg M, Mller SR, Domman U, Schwarzenbach
RP (1998) Anal Chem 70:30943101
96. Suzuki T, Kondo K, Uchiyama M, Murayama M (1999)
J Agric Food Chem 47:38863894
97. Alzieu C, Sanjuan J, Deltreil JP, Borel M (1986) Mar Pollut
Bull 17:494498
98. Morcillo Y, Porte C (1998) Trends Anal Chem 17:109116
99. Rodrguez I, Santamarina M, Bollan MH, Mejuto MC, Cela
R (1997) J Chromatogr A 774:379387
100. Chau YK, Yang F, Brown M (1997) Anal Chim Acta 338:
5155
101. Ceulemans M, Witte C, Lobinski R, Adams FC (1994) Appl
Organomet Chem 8:451461
102. Quevauviller P, Astruc M, Morabito R, Ariese F, Ebdon L
(2000) Trends Anal Chem 19:180188
103. Quevauviller P, Morabito R (2000) Trends Anal Chem 19:
8696
104. Szpunar J, Schmitt VO, Lobinski R, Monod JL (1996) J Anal
At Spectrom 11:193196
105. Gomez-Ariza JL, Morales E, Beltran R, Girldez I, Benitez
MR (1995) Analyst 120:11711174
106. Pellegrino C, Massanisso P, Morabito R (2000) Trends Anal
Chem 19:97106
107. Lobinski R, Dirkx WMR, Szpunar- Lobinska J, Adams FC
(1994) Anal Chim Acta 286:381390
108. Pyrzynska K (1996) Mikrochim Acta 122:279293
109. Metcalfe PJ (1989) Anal Proc 26:134136
110. Zufiaurre R, Pons B, Nern C (1997) J Chromatogr A 779:
299306
90
111. Yu X, Pawliszyn J (2000) Anal Chem 72:17881792
112. De Smaele T, Moens L, Dams R, Sandra P (1996) Fresenius J
Anal Chem 355:778782
113. Bergmann K, Neidhart B (1996) Fresenius J Anal Chem 356:
5761
114. Heisterkamp M, Adams FC (1998) Fresenius J Anal Chem
362:489493
115. Baena JR, Crdenas S, Gallego M, Valcrcel M (2000) Anal
Chem 72:15101517
116. Lobinski R, Adams FC (1992) Anal Chim Acta 262:285297
117. Paneli M, Rosenberg E, Grasserbauer M, Ceulemans M,
Adams FC (1997) Fresenius J Anal Chem 357:756762
118. Sadiki A, Williams DT (1996) Chemosphere 32:19831992
119. Lobinski R, Szpunar-Lobisnka J, Adams FC, Teissedre PL,
Cabanis JC (1993) J Assoc Off Anal Chem Int 76:12621267
120. Lobinski R, Adams FC (1992) J Anal At Spectrom 7:987992
121. Craig PJ, Dewick RJ, Elteren JT (1995) Fresenius J Anal
Chem 351:467470
122. Heisterkamp M, De Smaele T, Candelone JP, Moens L, Dams
R, Adamas FC (1997) J Anal At Spectrom 12:10771081
123. Lobinski R, Boutron CF, Candelone JP, Hong S, Szpunar-
Lobinska J, Adams FC (1993) Anal Chem 65:25102515
124. Lobinski R, Szpunar-Lobinska J, Adams FC (1994) Analusis
22:M54-M57
125. Salih B (2000) Spectrochim Acta Part B 55:11171127
126. Dunemann L, Hajimiragha H, Begerow J (1999) Fresenius J
Anal Chem 363:466468
127. Gorecki T, Pawliszyn J (1996) Anal Chem 68:30083014
128. Dirkx WM, Van Cleuvenbergen R, Adams FC (1992) Mikro-
chim Acta 109:133135
129. Allen AG, Radojevic M, Harrison RM (1988) Environ Sci
Technol 22:517522
130. Nern C, Pons B, Zufiaurre R (1995) Analyst 120:751754
131. Johansson M, Berglf T, Baxter DC, Frech W (1995) Analyst
120:755759
132. Van Cleuvenbergen R, Chakraborti D, Adams F (1990) Anal
Chim Acta 228:7784
133. Yu X, Yuan H, Grecki T, Pawliszyn J (1999) Anal Chem 71:
29983002
134. Rosenkranz B, Quevauviller P, Bettmer J (1999) Am Lab 31:
1724
135. Rodriguez-Pereiro I, Schmitt VO, Lobinski R (1997) Anal
Chem 69:47994807
136. Rodriguez-Pereiro I, Wasik, Lobinski R (1998) J Chromatogr
A 795:359370
137. Rodriguez-Pereiro I, Lobinski R (1997) J Anal At Spectrom
12:13811385
138. Wasik A, Rodriguez-Pereiro I, Lobinski R (1998) Spectro-
chim Acta Part B 53:867879
139. Wasik A, Rodriguez-Pereiro I, Dietz C, Szpunar J, Lobinski R
(1998) Anal Commun 35:331335
140. Slaets S, Adams F, Rodriguez-Pereiro I, Lobinski R (1999)
J Anal At Spectrom 14:851857
141. Rodriguez I, Mounicou S, Lobinski R, Sidelnikov, Patrushev
Y, Yamanaka M (1999) Anal Chem 71:45344543
142. Rosenkranz B, Bettmer J (2000) Trends Anal Chem 19:138
156
143. Mller H, Cammann K (1988) J Anal At Spectrom 3:907913
144. Bettmer J, Bradter M, Buscher W, Erber D, Rieping D, Cam-
mann K (1995) Appl Organomet Chem 8:541545
145. Lopez-Avila V, Liu Y, Beckert WF (1997) J Chromatogr A
785:279288
146. Liu Y, Lopez-Avila V, Alcaraz M, Beckert WF (1994) Anal
Chem 66:37883796
147. Gremm TJ, Frimmel FH (1992) Water Res 26:11631169
148. Mller M (1987) Anal Chem 59:617623
149. Michel P, Averty B (1991) Appl Organomet Chem 5:393397
150. M. Nagase, M. Toba, H. Kondo, K. Hasebe (1998) Analyst
123:10911094
151. Gmez-Ariza JL, Morales E, Girldez I, Beltrn R (1997) In-
tern J Environ Anal Chem 66:113
152. Jiang GB, Xu FZ (1996) Appl Organomet Chem 10:7782
153. Wilken RD, Kuballa J, Jantzen E (1994) Fresenius J Anal
Chem 350:7784
154. Quevauviller P, Maier EA, Griepink B (1995) Quality assur-
ance for environmental analysis. Elsevier, Amsterdam
155. Prange A, Jantzen E (1995) J Anal At Spectrom 10:105109
156. De Smaele T, Moens L, Dams R, Sandra P, Van der Eycken J,
Vandyck (1998) J Chromatogr A 793:99106
157. Aguerre S, Lespes G, Desauziers V, Potin-Gautier M (2001)
J Anal At Spectrom 16:263269
158. Kim AW, Foulkes ME, Ebdon L, Hill SJ, Patience RL, Bar-
wise AG, Rowland SJ (1992) J Anal At Spectrom 7:1147
1149
159. Forsyth DS, Hayward S (1995) Fresenius J Anal Chem 351:
403409
Anal Bioanal Chem (2002) 372 : 91100
DOI 10.1007/s00216-001-1197-3
T E C H N I C A L R E P O RT
C. Hnel G. Gauglitz
Comparison of reflectometric interference spectroscopy
with other instruments for label-free optical detection
Received: 22 August 2001 / Revised: 5 October 2001 / Accepted: 25 October 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract On the basis of kinetic measurements of bio-
molecular interactions, a reflectometric interference spec-
troscopy (RIfS) setup is compared with two commercial
instruments. These instruments are based on evanescent
wave techniques, surface plasmon resonance (SPR) (rep-
resented by BIAcore 2000) and resonant mirror (RM) tech-
nique (using IAsys plus). All methods allow a label-free
and time-resolved optical detection of biomolecular inter-
action.
These methods are mainly used in biomolecular inter-
action analysis (BIA). They provide practical techniques
for quantifying equilibrium constants and rate constants
over several orders of magnitude.
The general parameters of the three detectors, namely
baseline noise and drift as well as overall sensitivity and
limits of detection were compared. The fluid handling and
the related implications on the measurements have also
been considered.
The interaction between thrombine and thrombine in-
hibitor (TI) was investigated as a test system with the
three different methods and the kinetic rate constants were
determined and compared. For this TI was immobilized
on the surface and binding of thrombine was monitored
time-resolved. Determination of the kinetic rate constants
could prove that the RIfS set-up is comparable with SPR
using BIAcore 2000 and RM technique represented by
IAsys plus.
Keywords Label-free detection Reflectometric
interference spectroscopy Biomolecular interaction
Thrombine Thrombine inhibitor
C. Hnel () G. Gauglitz
Institute of Physical and Theoretical Chemistry,
University of Tbingen, Auf der Morgenstelle 8,
72076 Tbingen, Germany
e-mail: abc.gg@ipc.uni-tuebingen.de
Introduction
Biomolecular interaction analysis (BIA) is a term which
was established a few years ago by the Swedish Company
Pharmacia Biosensors of Sweden (now Biacore). It com-
prises the investigation of bio-molecular binding events in
real time without labelling by an optical detection system.
The advantage of the technology is that it provides a
direct and rapid readout of both the affinity and the kinet-
ics of interaction [1, 2]. The absence of labels means little
or no time-consuming pre-treatment of the components.
Optical sensors work continuously, reversibly and effi-
ciently. Principles in optical detection and applications are
well described in the literature [3, 4, 5].
Reflectometric interference spectroscopy (RIfS) as a
direct optical detection method is compared with surface
plasmon resonance (SPR) and resonant mirror (RM) tech-
niques, represented by two commercially available instru-
ments, BIAcore 2000 and IAsys plus.
Each of the three methods has been successfully used
for time-resolved detection of biomolecular interactions
as many references show, SPR using BIAcore [6, 7], RM
using IAsys [8, 9] and also RIfS [10, 11, 12], even in de-
tecting binding affinities of low molecular weight ana-
lytes [13].
The main objective in this study was to show that the
RIfS set-up fits well into the context of BIA and that it is
suitable for evaluation of biomolecular interactions and
determining kinetic constants.
The main components, a sample handling unit, a trans-
ducer with a sensing layer and the detection device re-
ceived attention throughout this paper. Fluid handling au-
thoritatively determines the reproducibility of the mea-
surements and the detection device influences the noise
and consequently the limit of detection (LOD). The sens-
ing layers should be suitable for the particular application
and also highly sensitive and specific. Altogether each
component can affect the merits of the instrument in a cer-
tain way and should be considered in this study.
A short description of the commercial instruments
should show their advantages and limitations and the RIfS
92
detection method used will be described in a little bit more
detail, because it is not as well known as the other princi-
ples.
The detection principle of the BIAcore instrument be-
longs, as do the RM techniques, to the refractometric tech-
niques. Beyond the critical angle, light will be totally in-
ternally reflected at interfaces with different refractive in-
dices. When this occurs, an electromagnetic field an
evanescent wave passes from the interface to the lower
refractive index medium, where it decays exponentially.
The resonance angle at which the evanescent field is
highly enhanced is extremely sensitive to the refractive
index at the sensor surface.
Surface plasmons are excited at the surface of a thin
metal layer (Au in the case of BIAcore) while the evanes-
cent wave interacts with free oscillating electrons. Energy
from the polarized incident light is lost to the metal film at
a well-defined angle of incidence depending on the refrac-
tive index of the medium beyond the chip surface [14].
Considering the fluid handling, BIAcore 2000 has a so-
phisticated, automated microfluidic system which is an
integral part of the biosensor. It has a very low dead vol-
ume and rapidly takes the sample to the sensing surface,
a constant and highly reproducible flow driven by highly
precise syringe pumps. It allows a consumption of sample
between 30 and 80 L/cycle by a microfluidic cartridge
system. Up to four channels can be used, so one of the
channels is mostly utilized as an in-line reference. Sam-
ples containing macromolecular aggregates or entire cells
could cause problems by plugging up the microchannels.
BIAcore is provided with a variety of surfaces for dif-
ferent applications. The most recommended and used sur-
face for the commercial systems is the carboxymethylated
dextran layer. It is a hydrophilic, three-dimensional, high
surface area matrix, bearing charged derivatizable carboxy-
late groups. It is used for most of the kinetic measurements
and evaluation of rate constants described in the literature
[15, 16]. These binding sites within the three-dimensional
layers provide different accessibility for ligands and there-
fore no homogeneously distributed binding kinetics.
RM, represented by IAsys plus, uses a dielectric cou-
pling layer of high refractive index on an integrated prism
as a waveguide. It is separated from the prism by a low in-
dex layer, which allows a coupling into the resonant layer
Fig. 1 a Principle of the detection of
affinity interactions by reflectometric
interference spectroscopy (RIfS)
(n and d are the refractive index and
the physical thickness of the layer,
I1 and I2 intensities of the light beams,
reflected at the interfaces of the layer).
b Spectral reflectance pattern due to
constructive and destructive interfer-
ence of the reflected radiation
for certain incident angles. The angles of excitation are sen-
sitive to changes in refractive index [17].
IAsys plus operates with a special cuvette system which
contains sample container and sensing layer. To avoid dif-
fusion limitation and ensure rapid sample mixing it includes
a microstirrer. Its sample consumption is about 150 L/
measurement. Also, dirty samples can be applied without
plugging the chamber. The cuvette includes two chambers,
so one could be used as a reference cell.
A disadvantage could be the user-dependent sample
handling in case of the lack of an autosampler. The sam-
ple has to be injected manually into the cuvette.
Considering the available surfaces, the same sensing
layers as the BIAcore instrument supplies are provided
and recommended.
Both evanescent field methods are strongly affected by
slight changes in temperature, so a highly sophisticated
instrumentation for referencing and temperature stabilisa-
tion is required.
RIfS, as a very simple and successful approach, sup-
plies a very specific method giving cheap, robust and reli-
able sensor elements. This reflectometric technique deter-
mines the apparent optical thickness (nd) of a thin layer
by measurements of interference of white light caused by
partial reflection at interfaces with a distance of approx-
imately 1 m. Any change in optical thickness causes a
shift in the resulting interference pattern [18].
In detail, a thin transparent film passed by a light beam
will cause at each of the interfaces a multitude of reflected
beams. A light beam passing a weakly reflecting thin film
will be reflected in part at each of the interfaces. As the
two reflected beams travel different optical paths, a phase
difference is introduced. If the film thickness is small (a
few micrometers) the difference in optical paths is minute
and even for short, coherent (white) light sources interfer-
ence effects can be observed. These lead to a modulation
of reflected light intensity due to constructive and de-
structive interference, as shown in Fig. 1.
The use of a simultaneous spectrometer allows fast
acquisition of the reflectance spectra. Any change in opti-
cal thickness will lead to a change in the interference
spectrum. A resolution of optical thickness (nd) of about
0.5 pm has been achieved [18]. Drift measured with the
set-up normally ranges smaller than 10 pm/h.

In comparison to the described refractometric methods,
the reflectometric detection method selectively identifies
physically attached material on a surface [19] with no need
of referencing or temperature control unit.
The liquid handling for the RIfS system, providing a
single flow through cell, suffices because there is no ne-
cessity of signal referencing. The continuous flow is gained
by a robust fluidic system with either syringe pumps or
peristaltic pumps, which means a sample consumption of
200 L and 500 L respectively. Agglomerated samples
are not supposed to cause any problems by plugging up
the cell. A disadvantage could be the dead volume of 20 L
causing dispersion of the sample plug.
Also, there are suitable sensing layers for various ap-
plications. For kinetic measurements a plain, two dimen-
sional sensing layer is recommended, which shows homo-
geneous accessibility of binding sites and for that reason a
homogeneous kinetic behavior over the whole detection
area [20].
The data obtained with BIA can be used to derive un-
ambiguous answers to many qualitative questions; the
quantitative kinetic analysis is often difficult [21]. In some
cases there were multiphasic association and dissociation
sections found. The deviations were attributed to hetero-
geneity of the binding sites [22, 23]. So the sensing layer
of the transducer plays an important role.
In this study the interaction of thrombine and a throm-
bine inhibitor (TI) is observed. TI was immobilized to the
surface and the binding of thrombine was monitored re-
solved in time. Comparing general parameters such as
noise and drift and the evaluation of kinetic rate constants
should prove that determination of kinetic rate constants
obtained from the RIfS set-up is comparable with those
received from SPR and RM instruments.
Material and methods
Devices
For the SPR-based measurements a BIAcore 2000 device with au-
tosampler of the company Biacore (Uppsala, Sweden) was used;
Sensorchips CM5 (carboxylate dextran surface) were also deliv-
ered from there.
The RM technique was performed by IAsys plus of the com-
pany Labsystems (Affinity Sensors division). Carboxylate dextran
(CMD)-modified cuvettes have been supplied from there.
For the RIfS set-up components from several companies were
used:
1. Flow injection analysis device ASIA from Ismatec, Wertheim.
2. Diode array spectrophotometer SPEKOL, Analytik Jena, Ger-
many, modified according to [10] with a halogen reflector lamp
from Welch Allyn, New York, USA, and a polymer fibre of
PMMA 1 mm diameter from Ratioplast, Optoelectronics, Lhne,
Germany.
3. Optical transducers consisting of float glass coated with an in-
terference layer system (10 nm Ta2O5/330 nm SiO2), obtained
from Schott, Mainz, Germany. The term transducer provides
the additional information, that a physical property of the used
surface, here the interference-layer, is used to obtain the physi-
cal measurement parameter. Therefore the word chip is not
used in this context; nevertheless it could be used as a synonym.
4. The flow cell has the following dimensions 410.05 mm3,
< 200 nL.
93
5. Microcal Origin, version 5.0, Northampton, UK, was used to
evaluate kinetic data.
Materials
Chemicals and biochemicals were either from Sigma (Deisenho-
fen, Germany) or Fluka (Neu-Ulm, Germany). The TI (MW 393 g/
mol) was kindly put at our disposal by Dr. Friedrich, BASF, Lud-
wigshafen, Germany. Polyethyleneglycol (MW 2000 Da) was from
Rapp Polymere, Tbingen.
GA (Glutaric anhydride) was used to synthesize diamino-poly-
ethyleneglycol (DA-PEG)-GA surfaces by carboxylation of DA-
PEG-modified surfaces. Phosphate buffered saline (PBS) consist-
ed of 150 mM sodium chloride and 10 mM potassium phosphate at
pH 7.4. Ultrapure water (18.2 MOhmcm, SG, Barsbttel, Ger-
many) was used throughout this study.
Regeneration buffer used with BIAcore was TMM (2%Triton,
MgCl2 : Maleate:Tris, 1:1:1, 1 M NaCl), pH=6.8. Sodium acetate
buffer (NaAc) consisted of 10 mM sodium acetate in ultrapure wa-
ter at pH 4.6. For the activating solution 100 mM N-hydroxysuc-
cinimide (NHS) (MW 115.1 g/mol) was diluted in ultrapure water
and 400 mM N-ethyl-N-dimethylaminopropylcarbodiimide hydro-
chloride (EDC) (MW 191.7 g/mol) was diluted in ultrapure water
also. For activating surfaces both solutions were freshly mixed,
1:1. Blocking solution was ethanolamine in water 1 M, pH 8.0.
Surface chemistry for RIfS transducers
Pre-treatment of surfaces. All transducer chips (1.21.2 cm2) were
cleaned and modified with a reactive silane derivative by the same
procedure. Chips were first cleaned in NaOH in an ultrasonic bath
for 2 min, then washed with tap water. The surface was mechani-
cally cleaned and dried with Kim Wipes. Freshly prepared Piranha
solution (concentrated H2SO4 and H2O2 (30%), mixed 7+3) in an
ultrasonic bath for 15 min was used as the final cleaning step and
to ensure a surface rich in reactive silanol groups (security advice:
freshly prepared Piranha solution is extremely hot and aggressive.
Use a fume cupboard and appropriate security means). Afterwards,
transducers were washed thoroughly with ultrapure water and dried
in a nitrogen stream. To generate a surface provided with amino-
groups for coupling, the transducers were covered first with 10 L
3-glycidyloxypropyl-trimethoxysilane (GOPTS) for 1 h, followed
by cleaning with dry acetone and drying in a nitrogen stream, and
immediately after this pre-treatment 30 L of a solution of DA-
PEG in dimethylsulfoxide (4 mg/1 mL) was placed on each trans-
ducer and melted at 70 C overnight. Afterwards the transducers
were rinsed thoroughly with ultrapure water and dried. For con-
verting amino functions into carboxylate groups, 10 L of a GA/
N,N-dimethylformamide (DMF) solution (2 mg/1 L) were put
onto each transducer and than a second transducer was used to
cover (sandwich technique). The reaction needed 6 h within a
DMF saturated chamber. Then the transducers were washed with
DMF and ultrapure water and were dried.
Immobilization of TI on the carboxylic surfaces on the RIfS trans-
ducer. On the carboxymethylated surfaces, 10 L of a TI/DMF/di-
isopropylcarbodiimide solution (1 mg/10 L/1.5 L) were placed
and sandwiched with a second transducer. After 6 h in a DMF sat-
urated chamber the transducers were rinsed with DMF and ultra-
pure water and dried.
Characterization of immobilized TI with RIfS. Transducers were
characterized for maximum specific binding of thrombine by a
sample concentration of 50 g/ml and reached a maximum signal
of 500 pm, which is very low. A low capacity of binding sites is an
advantage to kinetic measurements, because measuring near the
limit of capacity reduces the probability of rebinding. Determina-
tion of dissociation rate constants is more precise.
For non-specific binding of another protein (below 10 pm),
ovalbumin from chicken egg was injected. For proof of the stabil-
ity of the immobilized TI layer,l several regenerations were carried
94
out. RIfS as a label-free optical detection technique allows the
monitoring of real time changes in the optical thickness (physical
thickness drefractive index n) of a thin transparent interference
layer coated to a glass carrier. The interference layer was exposed
to the sample, while it was illuminated with white light through the
glass carrier from the back side. Interference effects lead to a char-
acteristic spectral modulation of reflected intensity, which was
recorded by a diode array spectrometer. Changes in the reflectance
pattern allow on-line detection of binding events [24, 25, 26], with
a resolution down to some picograms per millimetres squared [10,
27]. The set-up used for RIfS-measurements consisted of a 20 W
tungsten halogen reflector lamp (Welch Allyn, New York), a bi-
furcated HCP silica fibre coupler (600 m fibre diameter) and a
diode array spectrometer MMS Spekol (Analytik Jena, Jena, Ger-
many). The transducer chips were mounted to a flow cell (41
0.05 mm3, <200 nL) and matched to the fibre with glycerol. A FIA
system (ASIA, Ismatec, Zrich, Switzerland) was used for sample
handling.
The maximum specific binding capacity of transducers was
characterized with solutions of 50 g/mL of thrombine in PBS:
1 mL of sample solution was loaded into the sample loop (0.8 mm
inner diameter, 800 L) of the liquid handling system and injected
with an initial flow rate of 240 L/min (10 s) followed by a flow
rate of 100 L/min (200 s). Subsequently the flow cell was rinsed
with PBS at 240 L/min (125 s). For the regeneration of the trans-
ducer surface it was rinsed with a pulse of low pH (0.01 M HCl,
250 s, 100 L/min). Non-specific interactions were checked using
the same injection protocol with 100 mg/mL of ovalbumin.
During all steps of the incubation the reflectance spectra were
recorded continuously and averaged over time intervals of 5 s.
From the resulting interference spectra the apparent change in op-
tical thickness was calculated on-line. For binding experiments the
signal shift from the pre-run baseline to the signal after binding
and rinsing was determined. For proteins a change in apparent op-
tical thickness of 1 pm corresponds to a surface loading of about
1.6 pg/mm2, which is calibrated by measurements of the specific
interaction between immobilized biotin and 125I-labelled strepta-
vidin. A publication regarding this is still in preparation.
Immobilization of TI on BIAcore chips. Immobilization of TI on
BIAcore chip CM5 was carried out on-line. The chip has a carbox-
ylate-modified surface as described above. TI (2.1 mg) was dis-
solved in 100 L NaAc buffer. After activation of the surface by a
200 L pulse of EDC/NHS (1:1), the injection block was washed
with running buffer. TI solution (80 L) was injected at a flow rate
of 10 L/min. Subsequently the surface was blocked by a 70 L
pulse of ethanolamine solution. After this treatment the chip was
regenerated with 20 L of 1 M NaCl solution.
The chip was given the ability to immobilize targets on four
different spots. Every spot can be accessed individually during
immobilization. During measurements any desired spot can be
switched in a row and used in parallel. So the first spot was chosen
for the reference measurement and the second one used for immo-
bilisation of the considered compound. So every sample hit the ref-
erence spot first and afterwards the measuring point in order not to
reduce the concentration of the sample before measuring.
Immobilization of TI in IAsys cuvettes. Every cuvette consists of
two measuring chambers. One was used as reference cell and the
other as measuring cell. For referencing the signal in both cells the
same liquid was injected but only in the measuring cell TI was im-
mobilized.
For measurements of thrombine kinetics with RM a CMD cu-
vette was used. In one of the two cuvette cells the immobilization
was accomplished. The other was used as reference cell. In order
to change the liquid in the cell completely every injection was car-
ried out three times. The liquid (always 50 L) was handled with a
pipette and the cuvette was sucked dry automatically. Stirrer speed
used was always 100 times /min and the immobilization was car-
ried out on-line.
After activating with EDC/NHS within 7 min the surface was
washed with PBS and afterwards changed to acetate buffer. Ten
microlitres of TI solution (1 mg/ml in acetate buffer) was added for
5 min and washed with PBS buffer. Subsequently the surface was
blocked by injection of ethanolamine for about 3 min. Then the cu-
vette was rinsed twice with NaCl (1 M in utrapure water) and in
between with PBS.
Measurement protocols
RIfS measurement protocol. For binding measurements of throm-
bine to immobilized TI several sample concentrations were in-
jected into the flow cell. During all these measurements an au-
tosampler was used and the measurements were carried out by pro-
gram. After a 120 s baseline measurement with PBS buffer the
sample was injected for 180 s at a flow rate of 100 L/min. After
that association period, buffer was injected for a dissociation pe-
riod of 150 s. Afterwards the surface was regenerated by a 250 s
pulse of low pH (0.01 M HCl, 30 L/min) and washed with buffer
for 120 s.
Measurement protocol for SPR. During measurements with BIAcore,
the autosampler was used and a flow rate of 10 L was steadily
run. After 120 s of buffer pre-run the sample was injected for 180 s.
Subsequently the surface was washed with running buffer for 270 s
and after that dissociation period the thrombine was removed off
the chip by a 180 s pulse of TMM solution. Afterwards the surface
was rinsed for 100 s with buffer. Spot 1 was used as reference and
spot 2 as measuring spot.
Measurement protocol using RM technique. Measurements with
IAsys plus were carried out without the use of an autosampler. All
injections were done by hand using an Eppendorf pipette. The in-
strument gave the ability to suck dry the cuvette. Every exchange
of liquid was done three times in order to ensure a complete re-
moval of the previous liquid.
After 2 min pre-run with buffer the sample was injected and
stayed in the cuvette for 5 min. Afterwards it was washed with
PBS buffer for 3 min and thrombine removed by low pH (0.01 M
HCl) for 2 min. After that the surface was washed with buffer for
2 min. Every step was carried out in both cells in parallel.
Results and discussion
Biomolecular interaction between immobilized TI and
thrombine using different techniques will be compared.
Kinetic data achieved from measurements with RIfS set-
up, SPR (using BIAcore 2000) and RM (using IAsys plus)
will be evaluated and compared.
Short description of the methods used
RIfS is one of the reflectometric techniques that have
been applied to direct immunosensing. The basic effect is
white light interference at thin transparent films. The prin-
ciple is extensively discussed elsewhere [11, 24].
SPR is also a direct optical method that is used in de-
tection of solid phase interactions. The basis of this tech-
nique is reflection of incident radiation at an interface be-
tween media with different refractive indices. Beyond a
critical angle the radiation will not pass through the inter-
face but will be totally reflected, the light is guided. An
evanescent field exists close to the interface in the me-
dium of lower refractive index and/or the transmittance.
The electric field of the reflected radiation depends on the
 95

refractive index, so any changes in these properties influ-
ence the evanescent field.
Surface plasmon waves are concerted oscillations of
electrons at a surface of a guiding material. The impact of
incident photons (only the parallel polarized part of the ra-
diation, TM mode) can excite such surface plasmons to
oscillate in resonance with the frequency of the radiation.
In general, a prism is used to excite plasmons in a sec-
ond interface (a metal like Au, Ag) which is coated on the
prism. The metal is used as a guiding material, incident
radiation penetrates this metal layer and excites the plas-
mons at the interface to the analyte. These plasmons ex-
hibit resonance in the visible range. Under conditions of
resonance, the intensity of reflected light is reduced like
an absorption peak at the scale of the angle of incidence
or wavelength of the exciting radiation.
So this technique gives a highly sensitive probe for de-
tecting changes in refractive index while binding to the
surface. But it needs a highly precise temperature control-
ling system in order to detect only changes in refractive
index caused from binding, not from media.
The third method used in this study is a RM technique,
which resembles the SPR sensor in its construction [17].
Light is totally reflected from the sensing surface by
means of a prism. At the sensing surface, the metal layer
is replaced by a dielectric resonant layer of high refractive
index. It is separated from the prism by a low index layer
thin enough that light may couple into the resonant layer
via evanescent field. Efficient coupling occurs only for
certain incident angles where phase matching between the
incident beam and the resonant modes of the high index
layer is achieved. At the resonance point, light couples
into the high index layer and propagates some distance
along the sensing layer interface before coupling back into
the prism. Resonance happens for TE and TM polariza-
tion and is observable as fine structure in reflected light.
The angle of excitation is sensitive to changes at the sens-
ing layer caused by changes of refractive index during in-
teractions at the surface. Light is reflected at any incident
angle, but a shift in the phase of reflected light during res-
onance can be detected.
Liquid handling
In flow-through systems, time resolution of the detection
is limited by sample handling, for example by the time
which is necessary for exchange of solution in the flow
cavity. In this study there were only two flow-through sys-
tems used for sample handling.
In order to evaluate data for determination of kinetic
rate constants it is necessary to know the region of maxi-

Fig. 2ac Determination of evaluable

region using the RIfS system limited
by distribution of the FIA peak within
the carrier liquid. (The noise of the
cy5-labelled antibody solution is
caused by the very low concentration
of the compound within the solution).
a Concentration profiles of the RIfS
fluidic system for determination of
the sample concentration by injecting
dye solutions (overlay of three differ-
ent molecular weight dyes). b En-
larged rise region of the injection,
limitation is caused by the construc-
tion of the fluidic system such as dead
volume. c Enlarged fall region of the
injection, distribution depending on
the size of the molecule
96
mum sample concentration. Concentration profiles were
measured with the RIfS system to register the peak of the
sample during the flow injection and specify the distribu-
tion. To this purpose, the reflected intensity was moni-
tored while dye solutions were injected under the same
conditions as the thrombine samples. Due to the different
diffusion dyes with different molecular weights used,
namely indigocarmin (MW 466 g/mol), dextranblue (MW
1 million g/mol) and Cy5-labelled antibody (150,000 g/
mol). Thrombine has a molecular weight of 39,000 g/mol.
Concentration profiles are given in Fig. 2. Maximum con-
centration was set to 100% in this plot.
Using this plot, the interesting evaluable region can be
defined. Maximum concentration is reached at nearly the
same time for each dye. Within 10 s up to 90% and within
20 s maximum concentration is reached using peristaltic
pumps. Removing the sample from the flow chamber
strongly depends on the molecular weight of the dissolved
dye. When the sample plug is directly flanked by carrier
liquid, diffusion takes place. Smaller molecules diffuse
faster than bigger molecules. Fast radial diffusion inhibits
dispersion of the sample plug caused by axial convection,
which means samples of small molecules do not mix with
the carrier stream as fast as larger molecules [28].
BIAcore 2000 has a very low dead volume and ensures
a highly precise exchange of sample in the flow cavity by
using a micro-flow cartridge. Because of successively fill-
ing the chamber, the different spots are reached one after
the other. The marginal resulting offset can be nearly
eliminated by software.
IAsys plus uses stirred liquid in a cuvette with two
cells. Limitations in this case could be not having a total
change of liquid while pipetting into the cuvettes, not fill-
ing both chambers simultaneously and evaporation during
long lasting equilibrium measurements because there is
no lid to cover the chambers.
By changing liquid three times, total exchange of liquid
is ensured. Nevertheless, a slight offset between reference
and measuring chamber caused by a time shift is unavoid-
able; a correction by software is possible. Temperature con-
trol and short measurement times keep evaporation low.
The comparison of the fluid handling systems shows
that RIfS and IAsys have to manage without a sophisti-
cated fluidic system. This could be a disadvantage in the
reproducibility of data and should be considered through-
out evaluation.
Comparison of performance criteria
The performance of a method first of all depends on gen-
eral parameters such as noise or drift. So first of all, noise
or drift obtained during the measurements should be com-
pared. For determination of LOD, signal calibration has to
be known. Measurements with radioactive labelled pro-
teins give the signal level for each method [29, 30] as
shown in Table 1.
The noise of the signal was derived from linear regres-
sion of recorded baselines as standard deviations of the
Table 1 Comparison of general parameters such as noise and drift
derived from baseline measurements with reflectometric interfer-
ence spectroscopy (RIfS), BIAcore 2000 and IAsys plus. Data were
taken during investigations of interaction between thrombine and
immobilized thrombine inhibitor (TI). For a better comparison the
results of the signal calibration by using radioactive labelled com-
pounds is shown. Signal calibration: 1 ng protein/mm2=1600 pm
optical thickness (RIfS), 163 arcsec (IAsys), 1000 RU (BIAcore)
Instru- rms-noise Limit of detec- Drift Drift
ment tion (3rms)/ (in units of
sensitivity concentration)
RIfS 0.80 pm 1.50 pg/mm2 10 pm/h 6 pg/h mm2
BIAcore 0.26 RU 0.78 pg/mm2 6 RU/h 6 pg/h mm2
IAsys 0.40 arcsec 7.36 pg/mm2 4 arcsec/h 24.5 pg/h mm2
signal after subtraction of linear drift. It was taken from
5 min pre-run baseline measurements with every instru-
ment using the surfaces with the immobilized TI.
Another limitation is the drift, which should not exceed
the signal. Detection of slow interactions is limited. The
drift is obtained from linear regression of this baseline
runs. Results are shown in Table 1.
Comparable data resulting from signal noise are the
LODs received from each instrument (3rms/sensitivity).
The lowest values were obtained from BIAcore 2000 in
this study. LOD of RIfS measurements is twice as high as
BIAcore data, and noise data achieved with IAsys plus in
this measurements gives an LOD more than 9 times
higher than the BIAcore data. All achieved values are
within one order of magnitude. Binding of proteins where
masses are in a range of several kilo Daltons can be easily
resolved and monitored.
Evaluation of rate constants
For the evaluation of rate constants several concentrations
of thrombine binding to the TI-modified surface were
monitored. Each sample concentration was measured three
times. For SPR and RM, referencing of the signal was nec-
essary. Signals obtained from the reference cell were sub-
tracted from that received from the measuring cell. RIfS
measurements were carried out without referencing. Bind-
ing curves for RIfS, referenced signals obtained with
BIAcore and also referenced binding curves for IAsys are
presented in Fig. 3.
Reproducibility of the binding curves
Considering the binding curves, the reproducibility of the
binding of thrombine with a certain concentration can be
estimated. Triple experiments have been carried out and
the standard deviation of the kobs values, which has been
percentally referred to the kobs values has been evaluated.
The average of the percental standard deviation is shown
in Table 2.
Data show that RIfS provides the lowest standard devi-
ations throughout the measurement. Despite the use of

Fig. 3 Binding curves of thrombine to immobilized TI received
with the RIfS set-up, with BIAcore 2000 and IAsys plus. Each con-
centration measured with an instrument is given within one plot to
show the reproducibility between repeated measurements
reference cells the reproducibility of the commercial in-
struments is not as good as the data obtained with RIfS.
The standard deviations are twice as high as received
from the RIfS-set-up.
97
Table 2 Comparison of the average of the percental standard de-
viationof kobs measured with RIfS, BIAcore 2000 and IAsys plus
Instrument Mean value of the standard deviation
of kobs in percent referred to kobs [%]
RIfS 8
BIAcore 2000 15
IAsys plus 15
Association rate constants
The kinetic of association of an analyte A binding to the
immobilized partner B to form the complex AB on the
surface is a bimolecular binding process. The rate of the
complex formation depends on the free concentration of
A and B components and on the stability of the formed
complex.
d[ A B]/dt = ka [ A][B] kd [ A B] (a)
ka is the association rate constant (M1s1) and kd the dis-
sociation rate constant (s1). [A] is in this case concentra-
tion of thrombine and [B] concentration of immobilized
TI. Second order kinetics and, if one component is con-
stantly replenished by laminar flow or stirred solution, re-
sulting in pseudo first order kinetic for binding reactions
to a surface are well described in literature [2, 31] and
[23]. According to these considerations, the change of sig-
nal R during time can be expressed as:
d R/dt = const. (ka A 0 + kd )R (1)
Solving the differential Eq. 1, the signal measured at a
time can be described as:
Rt = Req [1 ekobs t ] (2)
where
kobs = ka C + kd (3)
whereas Rt denotes the response at time t and Req the re-
sponse at equilibrium. A0 gives the initial molar concen-
tration of the component binding to the surface.
Each monitored binding curve derived from several
concentrations of analyte were exponential fitted using
Eq. 2 in order to get kobs which is shown as an example for
one of all monitored binding curves in Fig. 4. For evalua-
tion of each dataset Origin 5.0 was used.
For each binding curve a value for kobs were obtained.
According to Eq. 3 a linear fit was accomplished from
plot kobs versus initial concentration A0 for each dataset.
The slope results gives directly the association rate con-
stant ka as is shown in Fig. 5 for datasets monitored by
each method. Theoretically the dissociation rate constant
is also achieved from the intercept of the y axis. However
this intercept cannot be determined exactly when kd is low.
Limitations of the pseudo-first-order analysis are given
in [23]. In order to make sure of detecting the binding rate
instead of mass transport, derivative of the signal dR/dt is
plotted versus signal R. According to Eq. 1 the plotted
curve should be linear where mono-exponential behaviour
98

Fig. 4 a Binding curve with evaluable region limited by mass

transport. Pseudo first order kinetic is true only within the marked
section of the association, which is shown with a binding curve of
the RIfS measurement as an example. b Schematic drawing of the
sample concentration within the flow cell during determination of
dissociation. In order to reduce back binding effects kd is evaluated
from a binding curve received from a high sample concentration
because the lack of free binding sites

of the curve is true [31]. Aware of other limitations, all

data were evaluated in the same way to permit a compar-
ison among each other.
Another method to determine association rate con-
stants from binding curves is described in [31] using Eq.
1. Signal obtained from association dR/dt is plotted versus Fig. 5 Determination of association rate constants by linear re-
R as shown in Fig. 6. The slope value kobs can be intro- gression using plots from kobs vs. molar concentration for data
achieved with RIfS, BIAcore and IAsys
duced into a new plot versus analyte concentration with
Eq. 3 and the association constant is readily obtained from
the slope of the plot. Data obtained in this way are also higher. For evaluating association rate constants the first
given in Table 3. method is preferable.
The results obtained with this method are one order of Results achieved from association curves are within
magnitude below the results achieved with the method one order of magnitude for each instrument. Standard de-
described above. Linearization of an exponential relation viation of the fit for the dataset received with BIAcore
holds a loss of information. However, data compared gives the best results, 1/10 of the value of ka. For both
among each other demonstrate that association rate con- RIfS and IAsys, values for SD were 3 times higher.
stants are within one order of magnitude. The lowest stan-
dard deviation among repeated measurements is achieved
with the RIfS system; SD values obtained with IAsys plus
were about 4 times higher and for BIAcore about 11 times

Fig. 6 Determination of evaluable section of the association curve
and eliminating of the mass transport influenced regions by plot-
ting derivative of the signal dR/dt vs. signal R. The plot can also be
used to evaluate the association rate constant by the slope of the
linear fit, which provides kobs. The example has been taken from
kinetic measurements with the RIfS set-up
Table 3 Comparison of the determined association and dissocia-
tion rate constants of the interaction between thrombine and im-
mobilized TI measured with RIfS, BIAcore 2000 and IAsys plus
Instrument ka [104 M1s1] ka [103 M1s1] kd [102 s1]
derived from derived by
kobs-plot linearisation
RIfS 2.10.9 5.20.2 3.000.04
BIAcore 2000 3.20.3 7.32.3 0.700.01
IAsys plus 9.01.0 9.20.2 2.000.10
Dissociation rate constants
More favourable for achieving dissociation rate constants
then obtaining it from the y axis intercept is measuring
dissociation of bound analyte in buffer flow.
The rate equation is:
d Rt /dt = kd Rt (4a)
and the dissociation rate constant can be obtained by an
exponential fit of desorption data in terms of the integrat-
ed rate equation:
Rt = Req ekd t (4b)
as described also in the literature mentioned above. A typ-
ical fit taken from a RIfS-binding curve is shown in Fig. 4.
It has to be taken into account that that the sample con-
centration should be maximal. And as explained above, in
order to make sure of detecting the binding rate instead of
mass transport, the plotted curve of derivative of the sig-
nal dR/dt plotted versus signal R should be linear where
mono-exponential behaviour of the curve is true.
In order to avoid the influence of rebinding, data for
evaluation of dissociation rate constants were taken from
dissociation from highly loaded surface, i.e. after the bind-
ing of a highly concentrated sample. A comparison of dis-
sociation rates of thrombine from immobilized TI ob-
tained with RIfS, BIAcore 2000 and IAsys plus is given in
Table 3.
99
The values are also matchable. The lowest standard de-
viation is achieved with BIAcore, RIfS is 4 times higher
and IAsys 10 times higher.
Altogether the association and dissociation rate con-
stants obtained with RIfS fit very well into the data received
with the commercial instruments. The systems have been
compared as a whole, including all advantages and disad-
vantages, to prove the feasibility of kinetic measurements
with the RIfS-set-up.
BIAcore, with its highly sophisticated fluid handling,
and IAsys own the ability to reference the signal and to
reduce unspecific binding and drift. A temperature stabi-
lization reduces signal increase caused by the change of
refractive index with temperature. It also reduces the noise
of the detector. The RIFS-set-up, as a cheap and robust
device, has proved to obtain comparable data without a
sophisticated flow system and an extensive temperature
control.
To conclude, RIfS was compared with two other time-
resolved and label-free optical detection methods, SPR
and RM. Both were represented by commercially avail-
able instruments which are widely used. The RIfS set-up
used in our laboratory was compared regarding its suit-
ability for biomolecular interaction analysis with BIAcore
2000 and IAsys plus. Thus interaction between an immo-
bilized TI and thrombine was measured under the same
conditions with each instrument. Determination of affinity
rate constants shows that RIfS is a comparable technique
to the commercially available ones.
General parameters such as noise and drift are in good
agreement with those obtained by BIAcore 2000 and IAsys
plus. Regarding the LOD, the performance of RIfS is al-
most reaching BIAcore results and betters results achieved
with IAsys instrument throughout these measurements.
Although referencing the signal is not required for the
RIfS detection method, it could reduce drift caused by
light source and upgrade the system in that way. Refer-
encing has been tested in principle [32]. For reducing
noise a cooled detector could be used.
Sample consumption with the RIfS set-up at this time
is higher than consumption of the other devices. Improve-
ments in sample handling such as using a micro fluidic
system would decrease sample consumption. Using air
gaps dividing the sample plug from the carrier stream
would also reduce limitations given by sample handling.
Surface modifications suitable for various requirements
as mentioned above is offered by the companies involved.
Modifications of the RIfS transducer offer the same vari-
ety of surfaces and in addition a modification with poly-
ethyleneglycol suitable for kinetic binding measurements.
The very low price of the RIfS set-up compared to the
others is due to absence of temperature stabilization for
detection and low costs of components. Altogether, RIfS
is well comparable with other commercial available in-
struments.
Desirable for further development is a temperature con-
trolled flow chamber in order to evaluate temperature-
controlled binding, e.g. melting of DNA. Although the de-
tection method RIfS needs no temperature stabilization
100
for steady and reproducible measurements, it would be
favourable to evaluate temperature dependency of reac-
tions at the surface.
Further, a miniaturization of the device is possible. In-
stead of using white light providing the whole spectra for
the interference measurement, only four LEDs can be
used, detecting intensity at only four different wavelengths
with photodiodes instead of an diode array spectrometer,
which has already been tested in our workgroup [33].
Calibration of the signal by specific binding of protein
to the surface and solid phase detection of DNA-DNA in-
teraction using radioactive labelled compounds has also
been carried out.
Acknowledgements The help of Thomas Kast from the work-
group of Prof. Dr. Joachim-Volker Hltje, Department II (Bio-
chemistry) Max Planck Institute of Developmental Biology Tbin-
gen for committing the BIAcore 2000 is gratefully acknowledged.
References
1. Piehler J, Brecht A, Giersch T, Hock B, Gauglitz G (1997) J Im-
munol Methods 201:189206
2. OShannessy DJ et al (1993) Anal Biochem 249:457468
3. Brecht A, Gauglitz G (1997) Sensors Actuators B 3839:17
4. Baltes H, Gpel W, Hesse J (1996) Sensors Update, Sensor
Technology-Applications-Markets Vol. 1. VCH (Verlagsgesell-
schaft), Weinheim
5. Wolfbeis OS, Boisde G, Gauglitz G (1991) Optochemical sen-
sors. In: Gpel W, Hesse J, Zemel JN (eds) Sensors: a compre-
hensive survey vol. 2/I. VCH, Weinheim, chap 12
6. Karlsson R, Stahlberg R, (1995) Anal Biochem 228:27480
7. Malmqvist M, Karlsson R (1997) Curr Opinion Chem Biol
1:378383
8. Lowe PA, Clark T, Alwyn JH, Davies RJ, Edwards PR, Kin-
ning T, Yeung D (1998) J Mol Recognit 11:194199
9. Kiesau P (1999) Bioforum 22:623625
10. Schmitt H-M, Brecht A, Piehler J, Gauglitz G (1997) Biosens
Bioelectron 12:219 233
11. Brecht A, Gauglitz G, Polster J (1993) Biosens Bioelectron
8:387392
12. Piehler J, Schreiber G (2001) Anal Biochem 289:173186
13. Piehler J, Brecht A, Gauglitz G (1996) Anal Chem 68, 139143
14. Brigham-Burke M, OShannessy DJ, Smith (1993) Chromato-
graphia 35:459
15. Abraham R, Buxbaum S, Link J, Smith R, Venti C, Darsley M
(1995) J Immunol Methods 183:119125
16. George AJT, French RR, Glennie MJ (1995) J Immunol Meth-
ods 183:5163
17. Cush R, Cronin JM, Stewart WJ (1993) Biosens Bioelectron
8:347353
18. Brecht A, Gauglitz G (1994) Fresenius Z Anal Chem, 349:360
366
19. Brecht A. Gauglitz G (1995) Biosens Bioelectron 10:923936
20. Piehler J, Brecht B, Valiokas R, Liedberg B, Gauglitz G (2000)
Biosens Bioelectron 15:473481
21. Schuck P, Minton A (1996) Anal Biochem 240:262272
22. Khilko SN, Jelonek MT, Corr M, Boyd LF, Bothwell ALM,
Margulies DH (1995) J Immunol Methods 183:7794
23. OShannessy DJ, Winzor DJ (1996) Anal Biochem 236:275
283
24. Gauglitz G, Brecht A, Kraus G, Nahm W (1993) Sensors Actu-
ators B 11:2127
25. Piehler J, Brecht A, Gauglitz G, Zerlin M, Maul C, Thiericke R,
Grabley S (1997) Anal Biochem 249:94102
26. Sauer M, Brecht A, Charisse K, Stemmler I, Gauglitz G, Bayer E
(1999) Anal Chem 71:28502857
27. Piehler J, Brecht A, Gauglitz G (1996) Anal Chem 68:139143
28. Cussler EL (1984) Diffusion: mass transfer in fluid systems.
Cambridge University Press, UK
29. Internet (2001) http://www.biacore.com/biomol/pdfs/506.pdf
30. Affinity Sensors (1996) IAsys Methods Guide 2nd edn. Affin-
ity Sensors, Cambridge, UK
31. Karlsson R, Michaelsson A, Mattsson L (1991) J Immunol
Methods 145:229240
32. Haake H-M (2000) Monitoring von Festphasensynthesen und
biospezifischen Detektion in der HPLC mit Reflektometrischer
Interferenzspektroskopie. Dissertation, Tbingen
33. Reichl D, Krage R, Krummel C, Gauglitz G (2000) Appl Spec-
trosc 54:583586
Anal Bioanal Chem (2002) 372 : 101108
DOI 10.1007/s00216-001-1198-2
O R I G I N A L PA P E R
K. W. Phinney L. C. Sander
Preliminary evaluation of a standard reference material
for chiral stationary phases used in liquid
and supercritical fluid chromatography
Received: 20 August 2001 / Revised: 8 October 2001 / Accepted: 16 October 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract The applicability of a new Standard Reference
Material (SRM) for the evaluation of chiral stationary
phase (CSP) performance was demonstrated by utilizing
the SRM to characterize the chromatographic behavior of
eight commercially available CSPs in liquid and super-
critical fluid chromatography. The SRM consists of five
ethanolic solutions, each containing one chiral compound.
These test mixtures can be used to assess changes in col-
umn performance over time and to evaluate lot-to-lot vari-
ability in column manufacturing. The SRM was also used
to probe the effect of various parameters on column per-
formance.
Keywords Chiral stationary phase Enantiomers
Standard reference material
Introduction
Chiral stationary phases (CSPs) for liquid chromatogra-
phy (LC) have had a tremendous impact on enantioselec-
tive separations. Within the past two decades, the com-
mercialization of CSPs for LC has been accompanied by a
dramatic increase in the number of successful enantiose-
lective separations, and many chiral compounds can now
be resolved on more than one CSP [1]. The predominant
application of CSPs has been to the separation of drug
enantiomers, but the importance of stereochemical analy-
sis has been established in additional areas such as geo-
chemistry and environmental science [2, 3].
Contribution of the National Institute of Standards
and Technology. Not subject to copyright
K.W. Phinney () L.C. Sander
Analytical Chemistry Division,
Chemical Science and Technology Laboratory,
National Institute of Standards and Technology,
Gaithersburg, MD 20899, USA
e-mail: karen.phinney@nist.gov
Chromatographic separations of enantiomers on CSPs
depend upon the formation of transient diastereomeric
complexes between the enantiomers and a chiral selector
incorporated into the column packing material. Separation
of the enantiomers occurs when one enantiomer forms a
more stable complex and is retained longer than the other
enantiomer [4]. CSPs are often classified by the nature of
the chiral selector or by the type of interactions responsi-
ble for chiral recognition. Common chiral selectors in-
clude macrocyclic compounds, proteins, modified amino
acids and derivatized polysaccharides [5].
Evolution of CSPs has continued to occur as new CSPs
are prepared and commercialized. Chiral recognition stud-
ies have guided the development of modified versions of
early CSPs and the second generation versions of some
CSPs also offer improved stability [6, 7]. Many of the
CSPs initially developed for LC have now been employed
in supercritical fluid chromatography (SFC), where the
traditional liquid mobile phase is commonly replaced by a
carbon dioxide-based eluent [8].
The validity of chiral chromatographic results is con-
tingent upon the performance of the CSP. Once a suitable
CSP has been identified for the desired separation, enan-
tioselectivity is typically optimized through changes in
mobile phase composition, temperature, and flow rate [9,
10]. Inappropriate mobile phases, adsorption of sample
components, and leaching of bonded selectors can have
deleterious effects on column behavior. Changes in col-
umn performance often occur gradually and may not be
noticeable until the column fails to produce the desired
separation. Lot-to-lot variability in columns from the
same manufacturer may also have some bearing on the
suitability of a CSP for a particular analysis [11].
We have developed a Standard Reference Material
(SRM) for the evaluation of CSP performance in LC and
SFC. The SRM consists of five ethanol solutions, each
containing one racemic or enantiomerically enriched (non-
racemic) compound. The five test compounds are indap-
amide, ketoprofen, N-carbobenzyloxy-phenylalanine (N-CBZ-
phenylalanine), propranolol, and warfarin. The structures
of the test compounds are shown in Fig. 1, and the com-
102

Experimental
Certain commercial equipment, instruments, or materials are iden-
tified in this report to specify adequately the experimental proce-
dure. Such identification does not imply recommendation or en-
dorsement by the National Institute of Standards and Technology,
nor does it imply that the materials or equipment identified are
necessarily the best available for the purpose.

Reagents

Standard Reference Material 877, Chiral Selectivity Test Mix-

tures for Liquid Chromatography, was obtained from the Stan-
dard Reference Materials Program (NIST, Gaithersburg, MD,
USA). Carbon dioxide (SFC grade) was obtained from Scott Spe-
cialty Gases (Plumsteadville, PA, USA). All other reagents were
HPLC grade or better and were used as received.

Chiral stationary phases

The CSPs used in this study are listed in Table 2. These columns
were selected as a broad sample of dissimilar CSPs. Columns were
obtained from the following sources: Chiralcel OD and Chiralpak
AD (Chiral Technologies Inc., Exton, PA, USA), Chirex 3005 and
Chirex 3022 (Phenomenex, Torrance, CA, USA), Chirobiotic T,
Chirobiotic V, Cyclobond I 2000, and Cyclobond I 2000 RSP (Ad-
vanced Separation Technologies, Inc., Whippany, NJ, USA). All
columns were 4.6 mm25.0 cm. The same columns were used for
both LC and SFC. Columns were flushed with ethanol prior to
switching from one chromatographic technique (LC or SFC) to an-
other.

Chromatographic conditions
Fig. 1 Structures of components of SRM 877. The chiral center of
each compound is indicated by an asterisk Liquid chromatographic separations were performed at a tempera-
ture of 25 C and a flow rate of 1.0 mL/min. The injection volume
was 20 L. A solvent exchange to dissolve the test compound in
Table 1 Components and enantiomeric compositions of SRM 877
the mobile phase was necessary to avoid peak distortion with cer-
solutions
tain mobile phases. Supercritical fluid chromatography was per-
Compound Enantiomeric Major formed at a temperature of 30 C, a flow rate of 2.0 mL/min, and a
ratioa,b enantiomer pressure of 15 MPa. The injection volume was 5 L. Methanol
with 0.5% (v/v) isopropylamine was used as the modifier for all
Indapamide 1.0 N/Ac SFC separations. Detection in LC and SFC was performed by mea-
Ketoprofen 2.3 R surement of UV absorbance at 254 nm.
N-CBZ-phenylalanine 4.0 L
Propranolol hydrochloride 2.3 S
Warfarin 1.0 N/A Results and discussion
aData are provided for information purposes only and are not to be
A number of factors influence the performance of CSPs.
used for quantification
bEnantiomeric ratio (er)=mass E /mass E , where E is the major
1 2 1
Although the nature of the chiral selector has the greatest
enantiomer influence on the enantioselectivity of the CSP, other fac-
cNot applicable. The racemate was used
tors such as mobile phase composition, temperature, and
column age also have a bearing on column behavior. Suc-
cessful enantioselective separations are often achieved
position of the solutions is shown in Table 1. Potential ap- within a narrow range of mobile phase compositions, and
plications of this SRM include use as a control material CSPs are typically more sensitive to slight changes in mo-
for monitoring column performance, comparisons of bile phase composition than their nonchiral counterparts
columns having similar chiral selectors, and use in quality [11]. Temperature is an important parameter for chiral sep-
control for column manufacturing. We assessed the capa- arations in both LC and SFC because changes in tempera-
bilities of this SRM for testing column performance by ture can result in reversal of elution order [12]. In general,
using the test mixtures to evaluate the enantioselectivity enantioselectivity decreases with increasing temperature.
of eight commercial CSPs in LC and SFC. The performance of all CSPs degrades over time, but the
speed of this process varies. These changes may not be ap-
 103

Table 2 Chiral stationary phases evaluated with SRM 877

Column Particle Chiral selector Manufacturer
size
(mm)

Chiralcel OD 10 cellulose tris (3,5-dimethylphenylcarbamate) Daicel Chemical

Industries, Ltd.
Chiralpak AD 10 amylose tris (3,5-dimethylphenylcarbamate) Daicel Chemical
Industries, Ltd.
Chirex 3005 5 (R)-naphthylglycine and N-3,5-dinitrobenzoic acid Phenomenex
Chirex 3022 5 (S)-indoline-2-carboxylic acid and (R)-1-(a-naphthyl)- Phenomenex
ethylamine
Chirobiotic T 5 teicoplanin Advanced Separation
Technologies Inc. (Astec)
Chirobiotic V 5 vancomycin Advanced Separation
Technologies Inc. (Astec)
Cyclobond I 2000 5 b-cyclodextrin Advanced Separation
Technologies, Inc. (Astec)
Cyclobond I 2000 RSP 5 (R,S)-2-hydroxypropyl-b-cyclodextrin Advanced Separation
Technologies, Inc. (Astec)

Table 3 Suggested mobile phase compositions for liquid chro- Table 4 Suggested mobile phase compositions for supercritical
matographic separations of SRM 877 fluid chromatographic separations of SRM 877

Mobile Composition Mobile phase Composition

phase
A hexane:ethanol (50:50)
B hexane:2-propanol (95:5) with 0.1% diethylamine
C hexane:2-propanol (80:20) with 0.1% diethylamine
D hexane:2-propanol (95:5) with 1% trifluoroacetic acid
E hexane:2-propanol (80:20) with 1% trifluoroacetic acid
F hexane:1,2-dichloroethane:ethanol-trifluoroacetic acid
(85:10:5)a
G hexane:1,2-dichloroethane:ethanol-trifluoroacetic acid
(60:35:5)a
H 1% triethylammonium acetate, pH 4.1:acetonitrile
(90:10)b
J 1% triethylammonium acetate, pH 4.1:acetonitrile
(80:20)
K 1% triethylammonium acetate, pH 4.1:methanol (80:20)
L 1% triethylammonium acetate, pH 6.9:acetonitrile
(90:10)
M 1% triethylammonium acetate, pH 7.0:acetonitrile
(90:10)
N 20 mM sodium citrate, pH 6.3:tetrahydrofuran (90:10)
P 20 mM ammonium acetate in methanol
Q acetonitrile with 0.3% acetic acid and 0.2% triethyl-
amine
R methanol with 0.01% trifluoroacetic acid and 0.01%
ammonium hydroxide
S acetonitrile:methanol (95:5) with 0.3% acetic acid and
0.2% triethylamine
T acetonitrile:methanol (55:45) with 0.3% acetic acid and
0.2% triethylamine
aethanol and trifluoroacetic acid were premixed 20:1 (v/v)
b1% triethylammonium acetate was prepared by adding 1% (v/v)
triethylamine to water and adjusting pH to the desired value with
glacial acetic acid
parent without routine monitoring of parameters such as
retention factor, separation factor, and resolution.
Several test mixtures have been proposed for assess-
ment of enantioselectivity of CSPs for gas chromatogra-
AA carbon dioxide:methanol (95:5) with 0.5%
isopropylamine
BB carbon dioxide:methanol (90:10) with 0.5%
isopropylamine
CC carbon dioxide:methanol (85:15) with 0.5%
isopropylamine
DD carbon dioxide:methanol (80:20) with 0.5%
isopropylamine
EE carbon dioxide:methanol (70:30) with 0.5%
isopropylamine
phy (GC). In general, each test mixture has only been ap-
plicable to a particular type of CSP, such as modified cy-
clodextrin-based CSPs [13, 14]. Although some system
suitability tests for particular CSPs for LC have been rec-
ommended [15], no general approach to chiral column
evaluation currently exists. Many column vendors provide
a test chromatogram with new columns to demonstrate the
expected column performance, but the compound chosen
for this test may be one that is still well resolved even
when column performance has deteriorated. Hence, these
tests may not be an accurate prediction of column perfor-
mance for difficult separations. We elected to include five
different compounds in the SRM to allow the user a
choice of compounds for assessing column capabilities. In
addition, our goal was to develop test mixtures with ap-
plicability to as many different types of CSPs for LC and
SFC as possible.
Selection of the five test compounds for SRM 877 was
based upon a survey of chromatographic literature and hand-
books provided by column vendors. Factors that were
considered during the selection of the five test compounds
included commercial availability of the racemate and
availability of one enantiomeric form, stability of the
compound in solution, and the likelihood that the com-
104
Table 5 Liquid chromatographic data for selected commercial chiral stationary phases
Compound Chiralcel Chiralpak Chirex Chirex Chirobiotic Chirobiotic Cyclobond Cyclobond
OD AD 3005 3022 T V I 2000 I 2000 RSP
Indapamide
mobile phase A A G G K M H H
K(1)a 1.15 0.64 8.66 9.17 3.38 2.58 0.76 5.43
1.88 1.15 1.00 1.10 1.00 1.05 1.00 1.00
Rs 3.66 0.44 0.00 1.15 0.00 0.21 0.00 0.00
Ketoprofen
mobile phase D E P F K N H J
K(1) 3.23 1.22 3.39 1.67 3.73 0.94 3.12 4.97
1.00 1.20 1.10 1.00 1.07 1.11 1.00 1.00
Rs 0.00 1.56 0.99 0.00 0.41 0.33 0.00 0.00
Phenylalanine
mobile phase E E P F K H S J
K(1) 1.39 1.78 3.15 2.08 3.11 2.97 2.31 2.65
1.16 1.13 1.00 1.00 1.09 1.08 1.00 1.00
Rs 1.26 1.41 0.00 0.00 N.D.b N.D. 0.00 0.00
Propranolol
mobile phase C B G G T R S S
K(1) 1.84 2.01 3.77 7.00 3.20 8.99 4.72 4.19
1.88 1.00 1.00 1.09 1.13 1.11 1.00 1.00
Rs 5.16 0.00 0.00 0.85 2.53 1.77 0.00 0.00
Warfarin
mobile phase A A F F K H Q L
K(1) 0.38 0.52 6.59 5.71 5.94 4.46 0.40 4.87
2.47 1.46 1.00 1.09 1.10 1.48 1.14 1.00
Rs 3.63 2.11 0.00 0.78 0.68 3.24 0.59 0.00
a retention factor of first eluting enantiomer
b not determined

Table 6 Supercritical fluid chromatographic data for selected commercial chiral stationary phases
Compound Chiralcel Chiralpak Chirex Chirex Chirobiotic Chirobiotic Cyclobond Cyclobond
OD AD 3005 3022 T V I 2000 I 2000 RSP
Indapamide
mobile phase EE CC CC DD CC CC CC CC
K(1)a 3.16 12.14 14.15 11.56 19.16 22.59 9.31 18.08
1.35 1.00 1.00 1.05 1.03 1.04 1.00 1.00
Rs 5.55 0.00 0.00 1.40 1.08 1.12 0.00 0.00
Ketoprofen
mobile phase BB AA CC CC CC CC CC CC
K(1) 4.28 11.83 9.89 7.67 8.50 10.95 7.73 6.10
1.00 1.11 1.08 1.00 1.00 1.00 1.00 1.00
Rs 0.00 3.17 1.78 0.00 0.00 0.00 0.00 0.00
Phenylalanine
mobile phase BB BB DD DD CC CC CC CC
K(1) 8.38 5.50 15.91 12.62 18.69 19.20 19.00 16.26
1.10 1.11 1.00 1.00 1.04 1.00 1.00 1.00
Rs 2.18 2.66 0.00 0.00 1.33 0.00 0.00 0.00
Propranolol
mobile phase DD BB CC CC CC CC CC CC
K(1) 4.36 4.07 11.94 8.87 25.50 23.80 7.66 9.45
1.70 1.30 1.00 1.00 1.09 1.07 1.00 1.00
Rs 12.41 7.19 0.00 0.00 3.51 2.45 0.00 0.00
Warfarin
mobile phase DD DD CC CC CC CC CC CC
K(1) 2.80 2.12 9.95 12.14 9.93 10.64 7.54 8.31
1.72 1.94 1.07 1.00 1.00 1.02 1.00 1.00
Rs 5.64 14.72 1.33 0.00 0.00 0.73 0.00 0.00
aretention factor of the first eluting enantiomer

Fig. 2 Separations of indapamide enantiomers on commercial
chiral stationary phases by liquid chromatography
pounds enantiomers could be separated on CSPs used un-
der normal phase conditions as well as on CSPs designed
for reversed phase conditions. We chose to focus our ef-
forts on drug stereoisomers because they represent a ma-
jor application of chiral chromatographic techniques.
Three of the test mixtures were enriched in one enan-
tiomer to simplify identification of elution order. The en-
riched solutions were prepared by mixing one enan-
tiomeric form of the compound with the racemate to
achieve the desired enantiomeric composition. SRM 877
is designed to complement both quality assurance tests
performed by column manufacturers and system suitabil-
ity tests for chromatographic analytical procedures [16].
SRM 877 also builds on existing efforts at NIST to pro-
vide test mixtures to evaluate column performance [17].
Separation of the enantiomers of each of the five chiral
compounds was attempted on each of eight different
CSPs. Although the CSPs used in this study represent
only a small fraction of the total number of CSPs avail-
able, we anticipated that this study would provide prelim-
inary evidence of the value of the SRM for assessing col-
umn performance. The chromatographic conditions cho-
sen for each enantioselective separation were based upon
recommendations from column vendors. Exhaustive opti-
mization of each separation was not performed; therefore,
the results presented in this work may not reflect the high-
105
Fig. 3 Separation of ketoprofen enantiomers on a Chiralpak AD
CSP by LC (A) and SFC (B)
est separation factor () that could be achieved for each
test compound on a particular CSP. The mobile phases
used are summarized in Table 3 for LC and in Table 4 for
SFC. Chromatographic data for each of the five test com-
pounds on the eight CSPs is tabulated in Table 5 for LC
and in Table 6 for SFC. The data in Tables 5 and 6 are pro-
vided as a guide to selection of test compounds for a given
CSP and should not be interpreted as standard values. In a
few cases, poor peak shape precluded reliable determina-
tion of resolution (Rs).
The eight CSPs studied can be divided into groups by
the type of chiral selector. The following sections outline
the applicability of the test mixtures for evaluating each
type of CSP. It should be noted, however, that the use of
SRM 877 is not limited to the CSPs identified in this re-
port.
Polysaccharide-based CSPs
As shown in Table 5, four of the five test compounds were
enantioresolved in LC on the two polysaccharide-based
CSPs, the Chiralcel OD and Chiralpak AD columns. In
cases where more than one test compound is applicable to
monitoring the performance of a particular CSP, the selec-
tion of a test compound with closely eluting enantiomers
(Rs<1.5), such as N-CBZ-phenylalanine, may provide a
better indication of column changes than a compound that
is better than baseline resolved, such as propranolol. The
106
same number of test compounds was resolved on the two
CSPs in SFC, although the compounds resolved on the
Chiralpak AD CSP were not the same in LC and SFC.
Separations of indapamide enantiomers by LC on the
polysaccharide-based CSPs, as well as two other CSPs,
are shown in Fig. 2. In some cases, a reversal of elution
order was observed in the transition from LC to SFC.
Figure 3 illustrates the reversal of elution order that oc-
curred for ketoprofen enantiomers when a liquid eluent
in LC was replaced with a carbon dioxide-based eluent
in SFC.
Brush-type CSPs
The two brush-type (Pirkle-type) CSPs, Chirex 3005 and
Chirex 3022, incorporate modified amino acids as the chi-
ral selector [18]. Partial separation (Rs=1.0) of ketoprofen
enantiomers was observed on the Chirex 3005 CSP in LC.
Enantioselective separations for three of the test com-
pounds were obtained on the Chirex 3022 CSP in LC, but
resolution was less than 1.5 in each case. In contrast, both
ketoprofen and warfarin were resolved on the Chirex
3005 CSP in SFC. Only indapamide was enantioresolved
on the Chirex 3022 CSP in SFC.
SRM 877 was used to evaluate the differences between
two brush-type columns with the same chiral selector and
obtained from the same vendor. One column was pur-
chased specifically for this study. The second column had
been used for six months with various mobile phases. Fig-
ure 4 shows the separation of indapamide enantiomers on
two Chirex 3022 CSPs. As shown in the figure, slightly
better separation factor and resolution were observed on
the column that had been used for six months. The same
Fig. 4 Separation of indapamide enantiomers by LC on a new col-
umn (A), and a column that had been used for six months (B)
mobile phase was used for both columns, and therefore,
the variation shown likely did not arise from changes in
mobile phase composition.
Macrocyclic antibiotic-based CSPs
The Chirobiotic T and Chirobiotic V CSPs utilize slightly
different macrocycles as the chiral selector. Enantioselec-
tive separations of four of the five test mixtures were ob-
served on the Chirobiotic T CSP, but the low resolution
value for ketoprofen (Rs=0.41) makes this a poor choice
for column evaluations. In addition, Rs could not be reli-
ably determined for N-CBZ-phenylalanine. All five of the
test compounds were separated to some extent on the Chi-
robiotic V stationary phase, but the minimal degree of res-
olution obtained for some of the compounds makes them
poor candidates for column monitoring. Only propranolol
and warfarin were well resolved in LC. Separations of
propranolol enantiomers on the Chirobiotic T and V
CSPs, as well as two other CSPs, are illustrated in Fig. 5.
In SFC, three of the test compounds were enantioresolved
on each CSP. Although the enantiomers of indapamide
were not separated on the Chirobiotic T CSP in LC, a par-
tial separation was observed in SFC, as shown in Fig. 6.
Fig. 5 Separations of propranolol enantiomers on commercial chi-
ral stationary phases by liquid chromatography
 107

Fig. 6 Separations of indapamide enantiomers on commercial chi-

ral stationary phases by supercritical fluid chromatography

Table 7 Effect of temperature on the separation of propranolol Fig. 7 Separations of warfarin enantiomers on commercial chiral
enantiomers stationary phases by liquid chromatography

Temperature K(1) K(2) Rs

(C) farin was the only compound separated on the Cyclobond
I CSP in LC. Figure 7 illustrates the separation of war-
15 3.60 4.18 1.16 3.00
farin enantiomers on the Cyclobond I 2000 CSP and sev-
20 3.57 4.11 1.15 2.83
25 3.45 3.94 1.14 2.73
eral other stationary phases in LC. No enantioselective
30 3.40 3.85 1.13 2.59
separations of the test mixtures were observed on the Cy-
35 3.36 3.77 1.12 2.48 clobond I 2000 RSP CSP. For this CSP, SRM 877 pro-
vided evidence of retention behavior but could not be
used to evaluate enantioselectivity. It was not possible to
use SRM 877 to assess enantioselectivity of either cy-
The SRM was also used to study the effect of temper- clodextrin-based CSP in SFC because no enantioselective
ature on the chromatographic behavior of the Chirobiotic separations of the test mixtures were observed.
T CSP. Table 7 summarizes the decrease in retention,
separation factor, and resolution that occurred as the col-
umn temperature increased. It is important to note that Conclusions
the magnitude of temperature-induced changes in selec-
tivity will not be the same for all CSPs. In addition, tem- SRM 877 proved to be a convenient method of assessing
perature effects for a given CSP may be compound-spe- the performance of the majority of the CSPs studied. This
cific [19]. SRM represents an initial effort to design SRMs for chiral
stationary phases used in LC, and we expect that some
modifications may be made as we obtain feedback from
Cyclodextrin-based CSPs
SRM users. We also anticipate the development of addi-
tional SRMs to facilitate CSP characterization.
Two cyclodextrin-based CSPs, the Cyclobond I 2000 and
Cyclobond I 2000 RSP CSPs, were also examined. War-
108
References
1. Vandenbosch C, Massart DL, Lindner W (1992) J Pharm Bio-
med Anal 10:895908
2. Berthod A, Wang X, Gahm K-H, Armstrong DW (1998) Geo-
chim Cosmochim Acta 62:16191630
3. Armstrong DW, Reid GL, III, Hilton ML, Chang C-D (1993)
Environ Pollut 79:5158
4. Armstrong DW (1987) Anal Chem 59:84A-91A
5. Armstrong DW (1997) LC-GC S20-S28
6. Welch CJ (1994) J Chromatogr 666:326
7. Krstulovic AM, Vende JL (1989) Chirality 1:243245
8. Williams KL, Sander LC (1997) J Chromatogr A 785:149158
9. Kirkland KM (1995) J Chromatogr A 718:926
10. Pescher P, Caude M, Rosset R, Tambut A (1986) J Chroma-
togr 371:159175
11. Wozniak TJ, Bopp RJ, Jensen EC (1991) J Pharm Biomed Anal
9:363382
12. Stringham RW, Blackwell JA (1996) Anal Chem 68:2179
2185
13. Mayer S, Schmalzing D, Jung M, Schleimer M (1992) LC-GC
10:782785
14. Aichholz R, Blz U, Fischer P (1990) J High Resol Chro-
matogr 13:234238
15. De Lorenzi E, Fell AF, Caccialanza G, Massolini G, Kitsos M
(1992) J Pharm Biomed Anal 10:909915
16. Ng LL (2000) Am Pharm Rev 3:3134
17. SRM 869a, Column selectivity test mixture for liquid chro-
matography (Polycyclic aromatic hydrocarbons), and SRM
870, Column performance test mixture for liquid chromatog-
raphy, are designed to evaluate the performance characteris-
tics of reversed-phase LC columns
18. Cleveland T (1995) J Liq Chromatogr 18:649671
19. Smith RJ, Taylor DR, Wilkins SM (1995) J Chromatogr A 697:
591596
Anal Bioanal Chem (2002) 372 : 109114
DOI 10.1007/s00216-001-1129-2
O R I G I N A L PA P E R
A. Jurado-Lpez M. D. Luque de Castro
An atypical interlaboratory assay:
Looking for an updated hallmark (-jewelry) method
Received: 27 July 2001 / Accepted: 20 September 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract A comparison between a laser-induced break-
down spectrometry-partial least squares (LIBS-PLS) meth-
od and methods based on some well-known techniques,
such as induced-coupled plasma-atomic emission spec-
trometry (ICP-AES), flame atomic absorption spectrome-
try (FAAS), and atomic scanning microscopy (ASM) is pre-
sented in order to both validate the content of gold and sil-
ver in alloys to be used as solid standards and develop an
alternative to the established methods for the hallmark of
gold and silver in jewelry pieces. 17 alloys with gold con-
centrations ranging from 100 to 50% and 8 alloys with sil-
ver concentrations between 100 and 80% and variable con-
centrations of other metals usually present in jewels were
used as solid standards in LIBS in order to develop a meth-
od as general as possible. The results obtained in the anal-
ysis of some alloys (9 for gold and 7 for silver) show that
the proposed method is comparable with the official one.
Keywords LIBS Jewelry Hallmark Gold Silver
Introduction
A number of interlaboratory studies have been developed
in Europe, particularly in the last decade. Most of them
have been fostered in the light of the necessities creat-
ed by EU Directives. The different kinds of interlabora-
tory experiments depend on the aim for which they are
planned. Thus, when the performance of a single method
has to be tested the study is called collaborative trial or
method performance study [1]; the comparison of differ-
ent laboratories that perform comparable analyses with
their own individual methods is often called proficiency
testing or laboratory performance study [2], though some-
A. Jurado-Lpez M.D. Luque de Castro ()
Analytical Chemistry Division, Annex C-3,
Campus of Rabanales, University of Cordoba,
14071 Crdoba, Spain
e-mail: qa1lucam@uco.es
times the term round robin study is used [3]. Finally, an
interlaboratory study is used for the determination of the
reference value of a reference material or a certified refer-
ence material in which several independent methods known
for their accuracy are selected [4]. An issue of the Frese-
nius J. Anal. Chem. devoted to reference materials in both
biological and environmental fields has recently been
published [5].
Most of the analytical methods used in the systematic
analysis of solid samples are applied after wet digestion or
leaching and thus standard solutions commercially avail-
able can be used for calibration. These wet treatments
consist of more or less complex steps which are generally
critical parts of the overall analytical process because they
are responsible for the most significant errors. This is par-
ticularly true for solid samples, which have to be gener-
ally brought into either complete dissolution or leaching
in order to fulfill the requirements for sample introduction
of the majority of the spectroscopic instruments of general
use in routine laboratories [6]. Methods for direct solid
sample analysis are the most desirable for saving both
reagents and time.
Laser-induced breakdown spectrometry (LIBS) inte-
grates ablation of solid samples and excitation of the re-
sulting aerosol through plasma formation. Thus, the meth-
ods based on this technique are drastically simplified and
shortened as regards to methods involving sample disso-
lution. Examples of satisfactory applications of LIBS are
the determination of copper in aluminium samples [7], the
simultaneous quantification of Al, Cu, Fe, Pb and Sn in
zinc-based alloys [8] or the determination of aluminium in
Zn alloys [9]. The most remarkable aspects of LIBS are
rapidity (analysis time in the order of seconds), no appar-
ent sample destruction (only a small crater of micron-di-
ameter), in addition to the selectivity, sensitivity and pre-
cision characteristic of laser-based atomic techniques.
Nevertheless, LIBS shares with other techniques for direct
measurements in solids the lack of standards to run cali-
bration properly.
The official methods for the hallmark of gold and sil-
ver pieces are absolute methods (copelation and potentio-
110
metric titration, respectively). These methods involve se-
rious drawbacks, such as:
a) the samples must be completely destroyed,
b) the analytical procedures are tedious and time-con-
suming, involving multiple steps of high safety risk for
the operators, and
c) the energy costs are high.
Therefore, the development of reliable analytical methods
which, taking advantage of new analytical techniques, en-
ables the drawbacks of the conventional methods to be
circumvented should be promoted [10].
One of the present research guidelines of the authors
team is focused on the development of methods for help-
ing jewelry trade, one of the most significant industrial ar-
eas in Crdoba. One of the main authors aims is to de-
velop alternative methods which can be promoted for sub-
stituting those officially established for the hallmark of
gold and silver jewelry pieces. With this aim, we planned
to apply LIBS for establishing not only the content in gold
or silver in the target piece but also the content of other
components of the alloys used for making the jewel. One
of the main shortcomings to be circumvented was the lack
of the appropriate standards for running the calibration
curve. A series of nine Au and seven Ag alloy potential
standards were prepared with contents in gold and silver
ranging between 50100% and 79l00%, respectively. A
sui generis interlaboratory validation was planned with
the dual objective of assessing the metal concentrations in
the standards and comparing the results obtained using
well-established methods such as the official ones and
others based on ICP-AES, FAAS, etc. with those provided
by LIBS. For making this, the alloys were analysed by
five laboratories using different techniques.
Fig. 1 Experimental setup for
LIBS
The special characteristics of this validation study
(namely, 9 different materials for gold and 7 different ma-
terials for silver with different contents) made also man-
datory special statistical treatments of the data, different
from those used for conventional interlaboratory studies.
Materials and methods
Apparatus and instruments
The experimental setup for carrying out the laser-induced break-
down spectroscopic method is depicted in Fig. 1. A pulsed Nd:YAG
laser (Continuum, Model Minilite II) was used as excitation source,
operating at the second harmonic, wavelength 532 nm, with a laser
energy of 1.73 mJ with a 1 Hz repetition frequency. The pulse en-
ergy was measured with a pyroelectric joulimeter (Gentec, Model
ED-200L, nominal sensitivity of 9.86 VJ1) coupled to a digital
real-time oscilloscope (Tektronix, Model TDS 380). Owing to the
laser module does not permit the change of the beam energy, it was
modified by dispersing the beam (ca. 3 mm size at the output) us-
ing a 33 mm focal length plano-convex lens (Oriel, 36 mm outside
diameter, 236 mm aperture range). The beam was guided towards
the sample with a flat aluminised mirror (Oriel, 5050 mm, coated
with aluminium and overcoated with MgFa) and focused at normal
incidence on the target by a 135 mm focal length plano-convex lens
(Melles-Griot, 25 mm in diameter). The radiation emitted by the
plasma was collected and transmitted by a fiber optic to a Czemy-
Turner l/8 m MS125 spectrograph (Oriel, Model 77400), equipped
with an entrance slit of 25 m (Oriel, Model 77220) and a grating
of 2400 grooves mm1 (Oriel, Model 77420). The detector was a
charge-coupled device (Andor-Oriel, Instaspec IV Model 78430-V),
which consists of 1024128 elements and a total photoactive area
of 26 m2, connected via a 16-bit ISA card to a Pentium II micro-
processor computer. A multiple I/O box (Oriel, 10140) allowed
the connection between the Instaspec card and the laser power sup-
ply in order to synchronize the Q-switch with data acquisition, The
sample was placed in a manual X-Y-Z translation stage (Oriel,
Model 16921), which made possible sample manipulation as re-
quired. A 13.5 mm longitudinal movement could be achieved for

the three axes with a mechanical resolution of 10 m. An optical
distance laser sensor (M5L/50 Mikroelektronik, Germany) was
used for reproducible positioning of the sample.
The Andor CCD software (Andor Technology, version 2.0)
was used for management of the detection system. Data treatment
was carried out by the Pirouette software (Multivariate data analy-
sis for Windows 95/98 and NT, Infometrix, Inc., version 2.6).
The spectral region from 255 to 332 nm was considered for
plasma emission measurements. Five laser shots were delivered at
each sampling position and the independent spectra from the 2nd to
5th were accumulated to obtain the spectrum of the sample. The
emission signal was corrected by subtraction of the dark current at
the detector. Three sampling positions were made in all experi-
ments with three replicates each.
High-purity (>99.9%) gold and silver foils were used as stan-
dards. Au-Ag-Cu and Ag-Cu alloys with distinct compositions
were obtained from SAVECO (Crdoba, Spain), a Spanish Assay
Office. These alloys had a certificate value either in Au or Ag con-
tent and they were used for calibration purposes.
Reference procedure
The determination of gold in an alloy containing copper and silver
is carried out by melting it with Pb on a copela (a very porous cru-
cible made from bone ashes or Morganite). Pb and Cu are oxidised,
melted and absorbed by the porous copela, while gold and silver
content remains in it and is weighed. Then, nitric acid is added to
dissolve the silver and the residue is also weighed.
The amount of Pb added is crucial in order to obtain accurate
results, as an excess of Pb yields losses of noble metal (the losses
increase at high temperatures). On the other hand, the copela al-
ways absorbs amounts of noble metal as a function of the Pb de-
fect. The necessary amount of Pb will be as high as the alloy is rich
in gold. Moreover, the dissolution of the silver is only quantitative
when the resulting Ag-Au alloy ratio is 1:3 or higher. If the amount
of Ag is less than three times that of Au, the dissolution is incom-
plete. In this case, the addition of pure silver until the correct ratio
is mandatory. All these reasons make necessary a preliminary as-
say in order to know approximately the alloy composition before
doing the definitive analysis [11].
Results and discussion
The steps involved in this research were as follows: first,
the method for direct multielemental analysis of noble
metals was developed. This involved optimisation of the
parameters from which both the plasma formation and the
spectrum obtained depend on, and characterisation of the
method. Then, the comparison of the results with those
obtained by the official method in order to know the pre-
cision with which the content of Au and Ag in the stan-
dards can be established by LIBS. Finally, a comparison
of the results obtained by both the LIBS method and well-
known instrumental methods for metal determination with
the official method for the determination of Au and Ag in
jewelry was done.
Optimisation and characterisation of the LIBS method
Optimisation studies
Sample ablation (removal of particles from the sample
surface) occurs by the incidence of high power radiation
111
which leads to high temperatures in the sample surface.
Plasma formation takes place by several mechanisms in
which both vapour and solid phases are involved. In this
work three parameters influencing plasma formation were
optimised, namely: laser wavelength, laser energy and
working distance.
Optimisation of laser wavelength. Four wavelengths (1064,
532, 355, and 266 nm) are available from the Nd:YAG
laser. A visible radiation at 532 nm was used in the pre-
sent study since the optical elements required by the sys-
tem are of lower cost than those required in the UV region
and the beam trajectory is easily observed, thus decreas-
ing safety problems.
Optimisation of laser energy. The absence of an energy
selector in the laser device made mandatory the use of a
lens-diaphragm aperture binomial, optimised in a previ-
ous work [12] considering two aspects:
a) minimal relative standard deviation (RSD) with maxi-
mal signal-to-noise ratio (SNR) and
b) negligible deterioration of the sample.
The energy value chosen was 1.73 mJ, which provided a
satisfactory power density for sample ablation without ap-
preciable sample deterioration.
Optimisation of the working distance. The working dis-
tance (WD) can be defined as the distance between the fo-
calisation lens and the sample surface. This parameter was
optimised in order to obtain the lowest deterioration of the
sample. Thus, craters with a deep and narrow shape were
desirable. Hence, the WD was optimised moving up and
down the sample-laser impact zone, taking as reference
the nominal focal length of the lens. The optimal dimen-
sions of a crater were obtained at a working distance co-
incident with the nominal focal length of the lens [12]. In
consequence, this distance was selected for subsequent
experiments. The use of a distance sensor allowed faster
and more accurate measurements of the working distance
to be achieved taking into account the sample size.
Characterisation of the LIBS method
Multivariate calibration by PLS. Severe matrix effects
have been observed in the determination of metals in mul-
tielement samples [13, 14, 15] owing to the dependence of
the amount of ablated material on the physicochemical
properties of the sample conferred by the major compo-
nents. Therefore, the application of calibration strategies
such as internal standarisation is recommended in most
cases. According to a previous study developed by Ama-
dor-Hernndez et al. [10], partial least squares regression
can be considered as a satisfactory tool for the analysis of
the spectral information obtained by LIBS. In addition,
the application of this chemometric technique avoids the
necessity for using time-resolved spectra, thus reducing
the complexity and cost of the equipment.
112
Table 1 Content of gold and silver in jewelry pieces estimated by common metals Ag and Cu, and some others that can also
LIBS-PLS be found in jewels, such as Zn and Ni. LIBS spectra were
Pieces Nominal Predicted Relative obtained under the optimal experimental conditions: a laser
value (%) value (%) error (%) energy of 1.73 mJ and the delivering of five laser shots for
the accumulation of the last four. Three sampling positions
Gold 1 99.97 99.96 0.015
were considered, with three replicates each. Variables in
2 89.93 90.68 0.837
the whole spectral region selected (255332 nm) were
3 80.05 80.00 0.061
used as data (1024 experimental points). Interference of
4 76.0 75.39 0.800
the continuum radiation emitted at the earliest stage of
5 73.89 72.71 1.597
6 69.89 69.44 0.642
laser induced plasma was reduced by baseline correction.
7 60.07 59.95 0.198
Two common preprocessing algorithms (namely, mean
8 58.08 57.77 0.523 centering and autoscaling) were applied prior to construc-
9 50.09 52.57 4.951 tion of the statistical models by PLS-1, in order to define
the best calibration conditions.
Silver 1 99.68 99.57 0.107 After application of the PLS-1 algorithm to the plasma
2 97.70 97.61 0.062 emission data in every training set of samples, the cross-
3 95.67 95.76 0.085
validation leaving out one sample each time was devel-
4 92.71 90.30 2.600
oped and the PRESS (Prediction Error Sum of Squares)
5 89.63 89.50 0.152
versus the principal components plot was done. Then, the
6 84.66 84.77 0.128
F-statistic comparison of PRESS values [16, 17] was used
7 79.67 80.19 0.653
for selecting the optimum number of principal compo-
nents required for the construction of the calibration mod-
els. The prediction capability of the resulting calibration
In this work two sets of samples were used for the con- models was directly tested by their application to the an-
struction of the calibration matrix for each metal: 17 alloys alysis of nine gold alloys and seven silver alloys whose
with gold concentrations ranging from 100 to 50% and composition had been determined by the corresponding
8 alloys with silver concentrations ranging from 100 to official methods (copelation and potentiometric titration,
80%. In order to develop a method as general as possible, respectively). The results thus obtained are shown in
the alloys used for gold calibration included the most Table 1.

Table 2 Percent content of

gold in jewelry pieces obtained Pieces Nominal ICP-AES (1) ASM (2) ICP-AES (3) FAAS
by ICP-AES, ASM and FAAS value
(values in brackets correspond 1 99.91 91.09 (2.881) 99.25 (0.720) 96.99 (2.981) 93.33 (6.642)
to the relative error also in per-
cent) 2 89.93 88.93 (1.112) 90.87 (1.045) 94.99 (5.327) 86.87 (3.403)
3 80.05 80.19 (0.175) 82.42 (2.961) 82.13 (2.598) 78.97 (1.349)
4 76.00 76.26 (0.342) 78.48 (3.263) 77.69 (2.224) 74.08 (2.526)
Number in brackets after the 5 73.89 74.99 (1.489) 76.79 (3.925) 75.69 (2.436) 70.45 (4.656)
technique acronym indicates 6 69.89 69.69 (0.286) 73.03 (4.493) 70.86 (1.388) 66.56 (4.765)
the laboratory where the analy- 7 60.07 59.48 (0.982) 63.61 (5.893) 61.82 (2.913) 59.47 (0.999)
sis were developed. The ab- 8 58.08 59.02 (1.618) 62.24 (7.163) 57.97 (1.532) 56.19 (3.254)
sence of number indicates the 9 50.09 49.06 (2.056) 54.03 (7.866) 53.68 (7.167) 48.46 (3.254)
authors laboratory

Table 3 Content of silver of the target alloys obtained by ICP-AES, ASM and FAAS (values in brackets correspond to the relative er-
ror in percentage)
Pieces Nominal ICP-AES (1) ASM (2) FAAS (2) ICP-AES (2) ICP-AES(3) FAAS
value
1 99.68 99.98 (0.300) 99.92 (0.241) 100.70 (1.023) 102.23 (2.558) 95.99 (3.702) 100.86 (1.214)
2 97.70 98.44 (0.757) 97.95 (0.256) 99.73 (2.078) 99.94 (2.293) 98.21 (0.583) 96.81 (0.911)
3 95.61 96.55 (0.920) 96.31 (0.669) 98.79 (3.261) 99.16 (3.648) 98.69 (3.157) 93 35 (2.425)
4 92.11 93.40 (0.744) 92.70 (0.011) 93.28 (0.615) 95.17 (2.653) 93.59 (0.949) 91.16 (1.672)
5 89.63 90.26 (0.703) 89.46 (0.190) 91.61 (2.209) 94.57 (5.512) 95.55 (6.605) 88.72 (1.015)
6 84.66 85.40 (0.874) 83.49 (1.382) 84.79 (0.154) 83.43 (1.453) 84.61 (0.059) 82.03 (3.107)
7 79.67 19.77 (0.125) 79.27 (0.502) 80.59 (1.155) 79.45 (0.276) 78.77 (1.130) 77.08 (3.251)
Number in brackets after the technique acronym indicates the laboratory where the analysis were developed. The absence of number in-
dicates the authors laboratory.
 113

Comparison of methods
In addition to the analyses by the reference and authors
laboratories, where the target alloys were analysed both by
LIBS and FAAS, they were delivered to three other labo-
ratories in order to analyse them by other well-established
techniques for metals determination for their subsequent
comparison with the proposed LIBS method, taking as
true values those provided by the official method devel-
oped by the Spanish assay office. These laboratories were:
Istituto Superiore di Sanit, Rome (Italy), where the tech-
nique used for the analysis of both metals was ICP-AES
(labelled as 1); Analytical Chemistry Division, University
of Zaragoza (Spain), where the gold samples were anal-
ysed by ASM (Analytical Scanning Microscopy), while
the silver ones were analysed by ASM and also by FAAS
and ICP-AES (labelled as 2); Analytical Chemistry Divi-
sion, University of Malaga (Spain) where ICP-AES was Fig. 2 Plot of the slope and its standard deviation calculated from
used (labelled as 3). The results obtained for gold and sil- the application of the minimum squares regression method to the
comparison between the results obtained for gold with each tech-
ver are shown in Tables 2 and 3, respectively. Each result nique and copelation. Number in brackets indicates the laboratory
is the mean of four replicates, except in the case of the of- where the analysis was made; the absence of number indicates the
ficial method, which made ten replicates. authors laboratory (for details, see text)
The comparison of the results provided by the instru-
mental methods and those provided by the reference one
was done using a two-tailed t test to compare the means of The statistical parameters are shown in Table 4. The cal-
related (paired) values obtained from each method and the culated t-values were compared with the theoretical one at
official one in order to evaluate if the methods yield simi- a=0.05 and 8 degrees of freedom for gold (t=2.306) and
lar results at the 95% confidence level. The expressions 6 degrees of freedom for silver (t=2.447). As can be seen,
used were: ICP-AES and LIBS are the methods that yield values
    1/2 comparable to those obtained from the official method for
 d 2
d
2
/n
di i i gold, while for silver the ASM method can also be in-
d = sd =
cluded. Then, the results provided by every method were
n n1
  plotted versus those obtained by the official one and the
d minimum squares regression method was applied for test-
t= ing linearity between them. Taking into account the corre-
sd / (n)1/2
lation coefficient (because the absolute character of the
official method does not provide calibration curve, so no
slope is available for comparison), all the instrumental
Table 4 Statistical parameters obtained from the application of methods as well as LIBS had an acceptable linearity with
the two-tailed t-test to the comparison of the values provided by the official one. Then, each slope and its standard devia-
the official method and the instrumental ones tion were plotted to compare the instrumental methods
Instrumental d sd t among themselves. The results for gold are shown in Fig. 2.
method As can be seen, ICP-AES (1) and LIBS are the methods
that provide the more accurate results, while ASM has the
Gold ICP-AES (1) 0.362 1.212 0.897
minor standard deviation, but the slope has the higher dif-
ASM (2) 2.528 1.553 4.883
ference from 1. ICP-AES (3) has nearly the same slope as
ICP-AES (3) 1.539 2.250 2.052
ICP-AES (1), but the former has a higher standard devia-
FAAS 2.621 1.802 4.363
tion. The same was made for silver, as can be seen in Fig. 3,
LIBS 0.056 1.047 0.159
where the results obtained are similar to those provided
Silver ICP-AES (1) 0.583 0.278 5.543 from gold: the most accurate methods are those based on
ASM (2) 0.089 0.582 0.403 ICP-AES and LIBS, while the smaller standard deviations
FAAS (2) 1.396 1.030 3.580 correspond to ICP-AES (1) and ASM. According to Figs.
ICP-AES (2) 2.033 2.113 2.545 2 and 3, the proposed LIBS method provides values com-
ICP-AES (3) 0.821 3.027 0.718 parable to those obtained by the conventional techniques,
FAAS 1.387 1.351 2.716 being one of the more accurate and precise for both gold
LIBS 0.289 0.962 0.794
and silver.
Number in brackets after the technique acronym indicates the lab-
oratory where the analysis were developed. The absence of num-
ber indicates the authors laboratory.
114
Fig. 3 Plot of the slope and its standard deviation calculated from
the application of the minimum squares regression method to the
comparison between the results obtained for silver with each tech-
nique and potentiometric tritration. In this case, the correlation be-
tween ICP-AES (3) and the official method does not provide an ac-
ceptable value. Number in brackets indicates the laboratory where
the analysis was made; the absence of number indicates the au-
thors laboratory (for details, see text)
Potential use of LIBS for jewelry-hallmark
After this comparative study, the following aspects may
be counterbalanced for the potential use of LIBS for jew-
elry-hallmark. The LIBS method is not as accurate as the
official one but its use could endow jewelry-hallmark with
a series of advantages, such as:
Rapidity, due to the direct analysis of the sample which
also implies reagents saving.
No sample destruction, as the craters are of micron-di-
ameter, so the method can be considered non-destruc-
tive. This aspect is of unquestionable interest in valu-
able samples such as jewels.
Capability for identifying the alloy composition. Apart
from gold, silver and copper, there are some other met-
als usually found in jewelry alloys that can be identi-
fied, thus making possible the detection of commercial
frauds and presence of metals that can cause allergy
problems.
Spatial resolution, which allow us to know the distribu-
tion of metals, but can lead to errors on the overall com-
position when the alloy is not homogeneous.
Conclusions
The main factors influencing either instrumental or qual-
ity performance of LIBS have been taking into account in
the present study in order to achieve the objective for
which the method has been proposed: the establishment of
a rapid, non-destructive and accurate method for the de-
termination of noble metals in jewelry alloys. In spite of
the interferences present in the samples from all those
metals that can be found in this kind of alloys, (i.e., Cu,
Zn, Ni, etc.), the results confirm a good predictive ability
of the PLS models based on LIBS data for the analysis of
gold and silver in jewelry pieces in comparison with other
well-known instrumental methods for metal determina-
tion based on ICP-AES, ASM or FAAS. Moreover, the
proposed LIBS method can be used in the exploratory
analysis for the identification of the metals in an alloy.
Acknowledgements The Direcctin General de Investigacin
Cientfica y Tcnica (DGICyT) of Spain and the EU are thanked
for financial support (projects No. 95-0270-OP and lFD97- 0653).
The authors are grateful to the teams of Prof Dr S. Caroli, Prof Dr
J. R. Castillo and Prof Dr J. M. Cano-Pavn for the analysis.
SAVECO, Crdoba (Spain) is thanked for providing the certified
gold and silver alloys.
References
1. International Standard, Accuracy (Trueness and precision) of
measurement methods and results - Part 2: Basic method for
the determination of repeatability and reproducibility of a stan-
dard measurement method, ISO 5725-2-1994, Geneva, 1994
2. ISO/IEC Guide 43-1: 1997 (e), Proficiency testing by interlab-
oratory comparisons: Part 1: Development and operation of
proficiency testing schemes
3. Misra RK, Uthe JF, Musial CJ (1992) Analyst 117:10851091
4. Commission of the European Communities, Community Bu-
reau of Reference: New reference materials, improvements of
methods and measurements, EUR 9921 EN, 1985, Brussels
5. Biological and Environmental Reference materials, Eighth In-
ternational Symposium, W Fresenius (Ed.) (2001) Fresenius
Journal of Analytical Chemistry 370 (23)
6. Hoenig M, de Kersabiec AM (1996) Spectrochim Acta Part B
51:12971307
7. Kuzuya M, Mikami O (1992) J Anal At Spectr 7:493497
8. St-Onge L, Sabsabi M, Cielo P (1997) J Anal Atom Spectrom
12:9971004
9. Kim DE, Yoo KJ, Park HK, Oh KJ, Kim DW (1997) Appl
Spectrosc 51:2229
10. Amador-Hernndez J, Garca-Ayuso LE, Fernndez-Romero
JM, Luque de Castro MD (2000) J Anal At Spectrom 15:587
593
11. Treadwell FP, Tratado de Qumica Analtica, Manuel Marn
(editor) (1945) Barcelona, Spain
12. Amador-Hernndez J, Garca-Ayuso LE, Fernndez-Romero
JM, Luque de Castro MD, Anal Chim Acta (sent for publica-
tion)
13. Barbini R, Colao F, Fantoni R, Palucci A, Ribezzo S, van der
Seen HJL, Angelone M (1997) Appl Phys B 65:101
14. Nmet B, Kozma L (1995) J Anal At Spectrom 10:631636
15. Chalard C, Mauchien P, Andre N, Uebbing J, Lacour JL,
Geertsen C(1997) J Anal At Spectrom 12:183188
16. Thomas EV, Haaland DM (1990) Anal Chem 62:10911099
17. Haaland DM, Thomas EV (1988) Anal Chem 60:11931202
Anal Bioanal Chem (2002) 372 : 115121
DOI 10.1007/s00216-001-1119-4
O R I G I N A L PA P E R
F. Ehrentreich
Wavelet transform applications in analytical chemistry
Received: 31 May 2001 / Revised: 7 September 2001 / Accepted: 11 September 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract The wavelet transform has been established
with the Fourier transform as a data-processing method in
analytical chemistry. The main fields of application in an-
alytical chemistry are related to denoising, compression,
variable reduction, and signal suppression. Analytical ap-
plications were selected showing prospects and limita-
tions of the wavelet transform. An important aspect con-
sists in showing the advantage of wavelet transform over
Fourier transform with respect to dual localization of a
signal in both the original and the transformed domain en-
abling principal new application fields in comparison with
Fourier transform.
Keywords Wavelet transform Wavelet packet
transform Raman spectroscopy Spike recognition
Infra-red spectroscopy Spectra comparison
Introduction
Concerning wavelet transform methods diverse applica-
tion areas have been established in the information sci-
ences [1, 2, 3, 4, 5, 6, 7, 8], i.e. denoising, data compress-
ing, detection of discontinuities, edge detection, suppress-
ing of signals, locating and suppressing outlying values,
detection of missing data, detection of pure frequencies,
long-term evolution or self-similarity. Some special appli-
cations, such as fast multiplication of large matrices also
exist. The applications are not restricted to the one-di-
mensional case, but wavelet transform methods are suc-
cessfully applied for two-dimensional image analysis. In-
troductions and overviews from a chemometric point of
view have also been published [9, 10, 11, 12].
The main part of this work was performed at the TU Dresden,
Institute of Analytical Chemistry, working on the project
Eh 121/21 funded by the Deutsche Forschungsgemeinschaft.
F. Ehrentreich ()
Universitt zu Kln, Institut fr Biochemie,
Zlpicher Str. 47, 50674 Kln, Germany
e-mail: f.ehrentreich@uni-koeln.de
The wavelet transform (WT) is a mathematical trans-
formation for hierarchically decomposing functions. It led
to a description of a function in terms of a coarse overall
shape and details of a graded sequence.
In continuous wavelet analysis the wavelet coefficients
c are related with the signal according to:
  
1 x q
c ( p, q) = y (t)  dt (1)
p p
R
where x and y describe the original signal, p characterizes
the scale and resolution and q characterizes the transla-
tion.
Because we deal with discrete data vectors or matrices
we make use of the discrete wavelet transform. In discrete
wavelet transform (DWT), p and q take discrete values
according to: p = 2 j , q = k2 j , j N , k Z , leading to:

C (a, b) = C ( j, k) = y (t)  j,k (t) (2)
tZ
A real breakthrough regarding applications of the wavelet
transform in the applied sciences has only been achieved
after the work of Daubechies [1, 13], Mallat [14, 15] and
others. The first one includes suitable filter selection for
decomposition and reconstruction to solve the problems
of aliasing. The progress achieved by Mallat is due to the
development of efficient algorithms for performing the
discrete wavelet transform (DWT) as pyramidal algo-
rithms.
By wavelet analysis only the approximation coeffi-
cients, will be decomposed at each level j1 yielding new
approximation coefficients aj, that represent the low fre-
quencies or coarse shape of aj1, and detail coefficients dj
that correspond to the high-frequency constituent of the
preceding approximation. The original data might be con-
sidered as the approximation at level 0 by the a0 coeffi-
cients. Inversely, during wavelet synthesis, approximation
and detail coefficients from level j+1 will reconstruct the
approximation coefficients at the next finer level j. The
wavelet packet transform (WPT) is a generalization of the
WT [16, 17] because it leads to a complete decomposition
tree of the original data. It can be best explained as in Fig. 1.
116
Fig. 1 Schematic representation of the decomposition tree of the
wavelet packet transform (WPT) up to level 4. The solid thick
lines correspond to the sequence of approximation coefficients and
the dotted thick lines to the appropriate detail coefficients of the
wavelet transform
A schematic wavelet decomposition tree is drawn from
level 1 to level 4. The thick solid lines correspond to the
decomposed approximation coefficients of the WT and
the thick dotted lines to the details coefficients of the WT.
The other decomposition nodes are specific for the wavelet
packet transform.
Applications in analytical chemistry
Most of the applications in analytical chemistry belong to
a common processing scheme, i.e. they obey the sequence:
decomposition processing reconstruction. The wavelet
transform enriches the palette of procedures comprising
methods as principal component analysis and Fourier trans-
formation.
Despite the diverse analytical disciplines, two main
fields of wavelet applications in analytical chemistry may
be extracted that can be categorized to the above-men-
tioned general processing scheme. The first covers appli-
cations for improvement of signal-to-noise ratios [18, 19,
20, 21, 22, 23, 24, 25] and the second concerns data com-
pression [26, 27, 28, 29, 30]. Often, appropriate routines
are embedded in larger procedures as feature extraction
[26], peak recognition [21], calibration [27, 28] or spec-
tral library search [30].
In Fig. 2 a principal scheme is drawn that is applicable
for wavelet denoising as well as for compression. The
branches leading to soft thresholding (I) and hard thresh-
olding (II) are well investigated by mathematicians. If the
absolute values of detail coefficients |dj,k| are smaller than
a chosen threshold thrn those values are set to zero in both
cases, soft and hard thresholding. Soft and hard threshold-
ing differ with respect to processing detail coefficients
whose absolute values equal or exceed the threshold.
Hard thresholding does not change them, but soft thresh-
olding shrinks them towards zero avoiding a gap between
zero and the chosen threshold. Avoiding the discontinuity
is desirable for further data processing but does not make
much sense for pure compression purposes. The appropri-
ate formulas are given in Fig. 2. Pioneering work was
done by Donoho [31, 32, 33] making standard wavelet
transform procedures for denoising data available to the
scientific community.
Soft thresholding (path I in Fig. 2) of wavelet detail co-
efficients will generally be applied for data denoising. Due
to the fact that denoising is a procedure preceding further
data processing methods, the wavelet coefficients are of-
ten not in the focus by itself but the reconstructed signal
after removal of the noisy components.
Compression (path II in Fig. 2) is performed very sim-
ilar to denoising. Wavelet analysis does not lead to storage
compression by itself. During processing, the coefficients
below the threshold are set to zero. Underlying a common
data format for all coefficients, the zero coefficients need
as much storage space as the other ones. However, wavelet
decomposition with hard thresholding is an excellent pre-
processing step for compression methods such as run length
encoding or Huffman coding (a concise introduction is
given elsewhere [34]) because the procedure produces
large contiguous ranges of zeros. These processed data are
subsequently compressed efficiently. Temporarily the
compressed intermediate is of interest for storage. Be-
cause storage is not performed on his own the reconstruc-
tion of the data is the final point.
With the software used, both methods have in common
that they do not lead to compression of data in the computer
memory after thresholding. Admittedly, it could be possi-
ble to gain computer memory by using appropriate data
structures, e.g. by pointing only to non-zero coefficients.
An even more radical data reduction could be achieved
by crude wavelet analysis, removing all the detail coeffi-
cients and keeping only the approximation coefficients
(path III in Fig. 2). The dotted line in Fig. 2 should indi-
cate that also in that case a reconstruction is possible.
However, in general this strategy should be performed
with care because substantial data distortions may result.
Further, in contrast with the theory developed by Donoho

Fig. 2 Scheme of the sequence wave-
let decomposition processing re-
construction of signals y or approxi-
mation coefficients aj1. Three types
of processing wavelet detail coeffi-
cients are important: soft threshold-
ing, hard thresholding and crude pro-
cessing
et al. for thresholding wavelet coefficients [31, 32, 33]
crude processing is not founded on statistical theory.
Hence, it should only be performed with deep knowledge
of the subject, and analytical expertise should guide the
crude wavelet coefficient processing. Moreover, the ques-
tion arises, how the distortion of signals is comparable
with that of simple resolution reduction.
In Figs 3 and 4 resolution-dependent infrared spectra
of 4,8-dimethyl-3,7-nonadienoic acid methyl ester and
N-phenyl-2,3-naphthalenedicarboximide are drawn; the
experimental details are given elsewhere [35]. The upper
spectra are shown with the initial resolution of 2 cm1. In
the left column an ordinary resolution reduction was per-
formed (4, 8, 16, 32 cm1). At the right side successive
wavelet transformations from level 1 to level 4 based on
the Symlet basis with 8 vanishing moments (Sym8) are
shown corresponding in resolution to the ordinary resolu-
tion reduction. Concerning the preservation of important
spectral features, the spectra are comparable up to level 3
in both examples. At level 4 considerable signal distor-
tions are observed. In Fig. 3 the ordinary resolution reduc-
tion performs better and in Fig. 4 the Sym8 decomposition
lead to a better preservation of the spectral contours.
Because the behavior is not known in advance, it
seems advisable to perform even the preliminary data pro-
cessing at least with a resolution of 16 cm1. However, at
this level the preservation of significant signal features
obtained by crude wavelet processing is not substantially
117
better than those obtained by ordinary resolution decreas-
ing. At least, the advice should be given, to compare the
results with those of ordinary resolution decreasing.
Besides these main streams, other applications such as
recognition of signal discontinuities [36] or contrast en-
hancement of edges in image processing [37] have been
reported.
That application class shows a principal advantage of
the WT over Fourier transform (FT) and should be ex-
plained by Fig. 5. In Figs 5ae, a sine function is drawn
overlaid by a distortion (small spike). In Figs 5fq, results
of a Sym8 wavelet transform are represented. In Figs
5fh, the approximation coefficients are represented and
in Figs 5ik, the detail coefficients are represented. In Figs
5ln, the reconstruction of the approximations is shown
and in Figs 5oq, the reconstruction of the details is
shown. Fig. 5r represents the imaginary part of the Fourier
coefficients. For comparison in Fig. 5s, the undisturbed
sine is drawn and Fig. 5t represents the imaginary part of
the Fourier coefficients belonging to the pure sine. The
favorable property for signal and image processing con-
sists in the fact that a temporary instant is localized over
the different wavelet scales. The distortion of the sine is
clearly recognized not only in the original domain but also
in the frequency analogous scale representation of the
wavelet transform. That is a clear advantage with respect
to Fourier transform where the disruption is localized only
in the original domain.
118
Fig. 3 Infrared spectrum of 4,8-di-
methyl-3,7-nonadienoic acid methyl
ester in the range 4002446 cm1. Left
column conventional resolution re-
duction, right column wavelet ap-
proximation coefficients relying on
the Symlet basis with eight vanishing
moments

Fig. 4 Infrared spectrum of N-phenyl-

2,3-naphthalenedicarboximide in the
range 4002446 cm1. Left column
conventional resolution reduction,
right column wavelet approximation
coefficients relying on the Symlet
basis with eight vanishing moments

Fig. 5 Wavelet transformation
(Symlet with 8 vanishing moments)
and Fourier transformation of a sine
function overlaid by a spike. The sub-
figures are explained in the text
Fig. 6 Localization preservation in both original and wavelet do-
mains shown for the Raman spectrum of a graphite sample. The
wavelet transform was performed up to level 4 for the Symlet ba-
sis with 8 vanishing moments. From left to right: wavelet approx-
imation coefficients, wavelet details coefficients, reconstructed ap-
proximations, and reconstructed details
Figs 6 and 7 demonstrate the application of the local-
ization principle to Raman spectra of a graphite sample;
the experimental details are given elsewhere [38].
In the upper row of Fig. 6, the initial Raman spectra of
a graphite sample are represented for comparison. Below
119
are represented the wavelet transformations from level 1
to level 4. The columns from left to right represent wavelet
approximation coefficients, wavelet detail coefficients,
and reconstruction of the approximations and reconstruc-
tion of the details. It could be shown [38] that the level 1
details are suitable for discrimination between signal com-
ponents and disturbing spikes. The property of projection
between different wavelet representations enables the lo-
calization of the spiky components by the wavelet trans-
form and performing the spike removal afterwards in the
original domain by conventional methods as interpolation
or regression.
120

Fig. 7 Sequence of despiking and denoising Raman spectra: (a)

initial Raman spectrum of a graphite sample, (b) preliminary de-
noised spectrum, (c) despiked spectrum, (d) finally despiked and
(e) denoised Raman spectrum

Fig. 7 shows the usage of level 1 details as indicator

function for the spikes superimposed on the graphite spec-
trum. At the top (Fig. 7a), the initial Raman spectrum is
shown and next (Fig. 7b) the spectrum is presented in its
preliminary denoised form. In Fig. 7c the absolute values
of level 1 details (Sym8 transform) the spike indicator
function are shown. As spike indicator absolute values
of the level 1 details above a threshold of 0.01 has been
chosen. These positions were projected into the original
Raman spectrum. At the marked positions, the spikes have
been overwritten by interpolated values in the original do-
main. The last two subfigures show the spike removed
Raman spectrum (Fig. 7d) and at the bottom (Fig. 7e), the
finally despiked and denoised spectrum is drawn.
In some instances, the level 1 details could not surely
distinguish between tiny spikes and real signals. False Fig. 8 Application of coefficients of the wavelet packet transform
positive recognition of spikes is the result. Hence, it has as indicator function for spike recognition: (a) Raman spectrum;
been investigated whether wavelet packet transform could (b) wavelet decomposition by Coiflet with 5 vanishing moments
(level 1 details correspond to wavelet packet node (1,1) in Fig. 1);
be successful applied for an improvement of signal / spike (c) coefficients relying on Coiflet 5 WPT, node (3, 7); (d) coeffi-
discrimination. cients relying on Symlet 8 WPT, node (3, 7); (e) coefficients rely-
In Fig. 8, in extension to the wavelet transform [38], ing on Daubechies 12 WPT, node (3, 7); (f) coefficients relying on
different coefficients of a WPT have been investigated Symlet 8 WPT, node (4, 15); B Raman band, S spikes
with respect to its usability as indicator function for spike
recognition. Fig. 8a contains the Raman spectrum of a sil- WPT according to the decomposition structure (3,7) and
icon wafer. Figs 8bf show indicator functions, i.e. ab- (4,15) represents the true assignment.
solute values of WPT-coefficients, corresponding to dif- More general, the decomposition along the rightmost
ferent nodes given in Fig. 1. In Fig. 8b, a Coif5 WPT was sequence of the decomposition structure Fig. 1 are best
performed (Coiflet basis with five vanishing moments). suited as indicator functions for disruptions as spikes, i.e.
The coefficients correspond to node (1,1) of Fig. 1. In that the sequence (0,0)(1,1)(2,4)(3,7)(4,15). However,
special case, the coefficients are equivalent to level 1 de- further investigations are necessary for more important
tails of the Coif5 wavelet transform. In Fig. 8c, the coeffi- applications because the removal of tiny spikes could also
cients were chosen according to node (3, 7) from the de- be done by conventional procedures due to rules concern-
composition structure. Figs 8de differ from Fig. 8c due to ing the small difference from the baseline.
the chosen bases, i.e. Symlet 8 and Daubechies 12 WPTs
were performed. Fig. 8f relies also on Symlet 8 WPT;
however, the node (4, 15) was used. Conclusion
While the indicator function basing on the level 1 de-
tails of the WT show smaller values for spikes (S) as for Wavelet transform methods have been successfully intro-
Raman peak (R) the coefficients from the more complex duced into analytical chemistry applications and have en-

riched the arsenal of chemometrics. The most important
applications rely on wavelet detail coefficient suppression
in the wavelet domain. However, care should be taken
when leaving the path founded by theoretical considera-
tions as in the case of crude processing of wavelet detail
coefficients. In that case, strong analytical motivation
should guide that process and limitations of the method
should be considered.
Another category of applications makes use of the lo-
calization property of the wavelet transform. That proce-
dure does not follow the classical three-step scheme: de-
composition processing reconstruction, but according
to decomposition analysis reconstruction processing.
The selective recognition of instants in the transformed
domain clearly goes beyond the processing abilities
served by the Fourier transform.
Acknowledgements The support of the company Chemical Con-
cepts, Weinheim, FRG, for making the SpecInfo infrared spectral
database available to the author is grateful acknowledged.
References
1. Daubechies I (1992) Ten lectures on wavelets. SIAM, Philadel-
phia, PA
2. Kaiser G (1994) A friendly guide to wavelets. Birkhuser,
Boston, MA
3. Chui CK (1992) An introduction to wavelets. Academic Press,
New York
4. Strang GT, Nguyen T (1996) Wavelets and filter banks.
WellesleyCambridge Press, Wellesley, MA
5. Wickerhauser MV (1994) Adapted wavelet analysis from theory
to software algorithms. A. K. Peters, Wellesley, MA
6. Stollnitz EJ, DeRose TD, Salesin DH (1995) IEEE computer
graphics and applications 7684
7. Stollnitz EJ, DeRose TD, Salesin DH (1995) IEEE computer
graphics and applications 7585
8. Misiti M, Misiti Y, Oppenheim G, Poggi J-M (1997) Wavelet
toolbox users guide. Mathworks, Natick, MA
9. Walczak B (2000) (ed) Wavelets in chemistry, data handling in
science and technology, vol 22. Elsevier Science, Amsterdam
10. Walczak B, Massart DL (1997) Trends Anal Chem 16:451463
11. Alsberg BK, Woodward AM, Kell DB (1997) Chemom Intell
Lab Syst 37:215239
12. Leung AK-M, Chau F-T, Gao J-B (1998) Chemom Intell Lab
Syst 43:165184
121
13. Daubechies I (1988) Comm Pure Appl Math 41:909996
14. Mallat SG (1989) IEEE Trans Acoustics Speech Signal Proc
37:20912110
15. Mallat SG (1989) IEEE Trans Acoustics Speech Signal Proc
37:20912110
16. Wickerhauser MV (1991) INRIA lectures on wavelet packet al-
gorithms Proceedings ondelettes et paquets dondes. Rocquen-
court, France, pp 3199
17. Walczak B, van den Bogaert B, Massart DL (1996) Anal Chem
68:17421747
18. Barclay VJ, Bonner RF, Hamilton I P (1997) Anal Chem
69:7890
19. Woodward AM, Alsberg BK, Kell DB (1998) Chemom Intell
Lab Syst 40:101107
20. Mittermayr CR, Rosenberg E, Grasserbauer M (1997) Anal
Commun 34:7375
21. Ehrentreich F, Nikolov S, Wolkenstein M, Hutter H (1998)
Mikrochim Acta 128:241250
22. Nikolov SG, Hutter H, Grasserbauer M (1996) Chem Int Lab
Syst 34:263273
23. Cai C, de Harrington P (1998) J Chem Inf Comput Sci 6:
11611170
24. Walczak B, Massart DL (1997) Chemom Intell Lab Syst 36:
8194
25. Depczynski U, Jetter K, Molt K, Niemoller A (1999) Chemom
Intell Lab Syst 49:151161
26. Bos M, Vrielink JAM (1994) Chemom Intell Lab Syst 23:115
122
27. Jouan-Rimbaud D, Walczak B, Poppi RJ, de-Noord OE, Mas-
sart DL (1997) Anal Chem 69:43174323
28. Alsberg BK, Woodward AM, Winson MK, Rowland JJ, Kell
DB (1998) Anal Chim Acta 368:2944
29. Chau F-T, Gao J-B, Shih T-M, Wang J (1997) Appl Spectrosc
51:649659
30. Leung AK-M, Chau F-T, Gao J-B, Shih T-M (1998) Chemom
Intell Lab Syst 43:6988
31. Donoho DL (1995) IEEE Trans Inf Theory 41:613627
32. Donoho DL, Johnstone IM, Kerkyacharian G, Picard D (1995)
J Roy Stat Soc, series B 57:301337
33. Donoho DL, Johnstone IM (1994) Biometrika 81:425455
34. Press WH, Teukolsky SA, Vetterling WT, Flannery BP (1992)
Numerical recipes in C, 2nd edn. Cambridge University Press,
New York, pp 903909
35. Ehrentreich F (2001) Vib Spectrosc, submitted
36. Stork CL, Veltkamp DJ, Kowalski BR (1998) Appl Spectrosc
52:13481352
37. Wolkenstein M, Kolber T, Nikolov S, Hutter HJ (2000) Trace
Microprobe Tech 18:114
38. Ehrentreich F, Smmchen L (2001) Anal Chem 73:43644373
Anal Bioanal Chem (2002) 372 : 122127
DOI 10.1007/s00216-001-1132-7
O R I G I N A L PA P E R
Rudolf Tuckermann Sigurd Bauerecker
Bernd Neidhart
Evaporation rates of alkanes and alkanols
from acoustically levitated drops
Received: 24 June 2001 / Revised: 30 August 2001 / Accepted: 8 October 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract Evaporation constants of acoustically levitated
drops from the homologue series of n-alkanes and 1-alka-
nols in ambient air have been evaluated by size and tem-
perature measurements. The size of the pure liquid drops,
within a diameter range of 0.1 to 2.5 mm, was monitored
using a CCD camera, while temperature measurements
were carried out by IR thermography. During drop evapo-
ration, water from a humid environment with a relative
humidity between 5 and 80% was condensed on the drop
surface and in the case of n-pentane, the condensed water
froze as a result of the evaporative cooling.
Keywords Drop evaporation Acoustic levitation
Infrared (IR) thermography Drop freezing
Introduction
There is a general trend towards miniaturisation in mod-
ern analytical procedures in order to save analysis time
and cost as well as to reduce sample and waste volumes.
Analytical methods such as capillary gas chromatography,
capillary electrophoresis, atomic absorption spectroscopy
and mass spectrometry only require sample volumes in
the micro or submicroliter range. Handling of such small
sample volumes is difficult in view of analyte contamina-
tion and sorption processes on walls of the containers that
are usually used to hold the samples during analytical pro-
cedures. As shown in earlier work [1, 2, 3], the acoustic
levitation technique [4, 5] is a useful and easy tool for
contact-less sample pre-treatment (such as enrichment,
extraction and derivatisation) and for combination with an-
alytical techniques (such as optical spectrometry). Acous-
tically levitated liquid and solid samples are in a diameter
R. Tuckermann () S. Bauerecker B. Neidhart
Institut fr Kstenforschung/
Physikalische und Chemische Analytik,
GKSS-Forschungszentrum Geesthacht GmbH,
Max-Planck-Strasse 1, 21502 Geesthacht, Germany
e-mail: rudolf.tuckermann@gkss.de
range of 0.12.5 mm and they are suspended in a gaseous
environment by a stationary ultrasonic field of approxi-
mately 155 dB. Acoustic levitation avoids sample conta-
mination and sorption processes by container walls, but
suffers from evaporation and loss of solvents. To balance
evaporation and condensation on levitated drops during
the experiments, techniques of contact-less droplet size
monitoring and solvent and reagent supply have been de-
veloped [2]. For a quantitative analysis of evaporation pro-
cesses a general understanding of the properties of the drop
and gaseous environments and the technique of acoustic
levitation is needed. Therefore, we have performed evap-
oration experiments under different environmental condi-
tions and with different kinds of acoustically levitated
drops, i.e. distilled water and the homologues of alkanes
(from n-pentane to n-decane) and alkanols (from meth-
anol to 1-hexanol).
Experimental
Equipment
Acoustic levitator: APOS BA 10 (Dantec Invent, Erlangen, Ger-
many); telecentric IR illumination and program-controlled CCD
camera with software: TZB 30-IR, camera VC 11 (M1004), soft-
ware VC Win (Vision & Control GmbH, Suhl, Germany); 50 mm
objective: B5018A-3 (KA) (Pentax Handelsgesellschaft mbH, Ham-
burg, Germany); long-pass filter: Schott Farbglas RG 780 (ITOS
Gesellschaft fr Technische Optik mbH, Mainz, Germany); Video
monitor: Panasonic WV-BM 1410 (Dekom, Hamburg, Germany);
IR thermography system with macro lens and software: VarioScan
3011, software IribisPlus (InfraTec GmbH, Dresden, Germany);
0.25 mm Ni-CrNi mantle thermocouple (THERMOCOAX GmbH,
Hamburg, Germany).
Chemicals
The following chemicals were purchased from Merck KGaA,
Darmstadt, Germany: methanol p.a., ethanol p.a., 1-butanol p.a.,
n-pentane p.a. and n-hexane p.a., and from Fluka Feinchemikalien
GmbH, Buchs, Switzerland: 1-propanol (purum 99%), 1-pentanol
(purum 98%), 1-hexanol (purum 98%), n-heptane (purum 99%),
n-octane (purum 99%), n-nonane (purum 99%) and n-decane
(purum 98%).

Fig. 1 Experimental setup for simultaneous size and temperature
measurements of acoustically levitated droplets via CCD camera
and IR thermography system, respectively
Experimental setup
The experimental setup used to evaporate pure liquid droplets in
ambient air is shown in Fig. 1. The evaporation process of acousti-
cally levitated drops was recorded simultaneously by a CCD cam-
era and an IR thermography system, both controlled by a personal
computer. The experiments were performed in an open acoustic
levitator, in ambient air, at ambient temperature and pressure. Tem-
perature, pressure and relative humidity of the environment were
monitored during the evaporation processes.
Acoustic levitator. The working frequency of the one-axial acoustic
levitator used for this investigation was 58 kHz corresponding to a
wavelength of = 5.7 mm in ambient air. The stationary ultrasonic
field was generated between an ultrasonic radiator, driven by a
piezoelectric transducer, and a concave reflector with a diameter of
18 mm. The reflector was positioned concentrically at a distance
d = n/2 (with n = 2, 3, ...)to the ultrasonic radiator in order to meet
resonance conditions. Acoustic levitation was possible at each pres-
sure node of the stationary ultrasonic field. In the present investi-
gation, a distance between radiator and reflector of n = 5 was ad-
justed, and the central pressure node was used for levitation. The
sound pressure level of the ultrasonic field was 150170 dB.
Size measurements of levitated drops. Horizontal and vertical di-
ameters of the suspended drops were measured by an intelligent
CCD camera from back-lit images of the drop about three times a
second [2]. Telecentric IR illumination of the levitated drop was
done at 880 nm and in combination with a 780 nm long pass filter
in front of the CCD camera. Size measurements of the drop ran
continuously using a program generated by the imaging software
VC and were calibrated with levitated polypropylene spheres with
a known diameter of 1.850.03 mm (Spherotech, Fulda, FRG).
The drop surface was calculated using the vertical and horizontal
diameters and assuming a rotationally ellipsoidal drop.
IR thermography system. IR thermography was chosen as a con-
tact-less method of temperature measurement with the advantage
of no additional heat transfer and high space resolution. The IR
123
thermography system is composed of an infrared camera operating
in the wavelength range of 812 m, equipped with an MCT-
detector cooled by liquid nitrogen and with a macro-lens. The
latter allowed a space resolution of 25 m per pixel. The image
frequency of the system was 1.2 Hz and in the temperature range
of 301200 C a temperature resolution of 0.1 C was feasible.
The whole system was controlled by special software. The series
of IR images were analysed by use of the emission coefficient of
the sample. For comparison, the temperature of suspended drops
could also be measured by a 0.25 mm mantle thermocouple.
Results and discussion
Evaporation processes of liquid droplets can be described
by a linear decrease of the droplet surface S with time t [6,
7]:
S = S0 K t (1)
where S0 is the initial surface at time t = 0. The propor-
tionality factor K, the so-called evaporation constant, is a
function of the liquid mass density l, the molecular mass
M, the binary diffusion coefficient D of the vapour in the
ambient gas, the partial vapour pressure p and tempera-
ture T at the drop surface (subscript s) and in the environ-
ment (subscript ):
 
8 D M ps p Sh
K = (2)
l R Ts T 2
with the gas constant R. The Sherwood number Sh de-
scribes the ratio of mass transfer with and without air
convection. Within a stationary ultrasonic field, acoustic
streaming leads to a higher mass transfer [8]. The acoustic
streaming is scaled by the intensity of the ultrasonic field.
As a result of that, the Sherwood number increases lin-
early with the intensity of the ultrasonic field in the case
of acoustically levitated drops [7]. This also leads to a lin-
ear increase of K with the intensity of the stationary ultra-
sonic field (see Fig. 2).
Fig. 2 Derived values of the evaporation constant K from size
measurements of evaporating n-hexane drops at different intensi-
ties of the stationary ultrasonic field scaled in arbitrary units. The
temperature and the relative humidity of the environment were
21 C and 49%, respectively
124
Table 1 Calculated values for the evaporation constant K of
n-alkanes and 1-alkanoles. Ksize is directly determined from size
measurements using Eq. (1), while Ktemp. is derived from the sur-
face temperature using Eq. (2)
Ksize (mm2s1) Ktemp. (mm2s1)
n-Pentane 2.0101 2.0101
n-Hexane 10.6102 9.3102
n-Heptane 4.9102 4.7102
n-Octane 2.0102 2.1102
n-Nonane 8.6103 7.7103
n-Decane 3.8103 2.6103
Methanol 4.5102 4.7102
Ethanol 3.1102 3.6102
1-Propanol 1.2102 1.4102
1-Butanol 7.0103 6.5103
1-Pentanol 4.0103 3.3103
1-Hexanol 1.7103 1.2103
Fig. 3 Temperature measurements of an evaporating water drop at
20 C and 26% relative humidity. The temperature was measured
simultaneously with a 0.25 mm thermocouple (symbols) sticking
In the following discussion, we present experimental in the levitated drop and with the IR thermography system (dashed
and solid lines). From the images of the IR thermography, the sur-
results of the evaporation of levitated pure liquid drops (wa- face temperature of the drop was calculated using different emis-
ter, alkanes and alkanols) in an environment of ambient sivity values. Optimum agreement was obtained with = 0.96, cor-
air, under ambient pressure and temperature conditions. responding well with the emissivity value of pure water as given in
The drop diameter at the beginning was in the range of the literature [12,13]
12 mm. From the experimental results we calculated K:
firstly from size measurements taken during the evapora- Evaporation of n-alkane drops
tion process using equation (1) and secondly from the
physical properties of the liquid drops and the ambient air, The emission coefficient of hydrocarbons is poorly de-
especially from the surface temperature Ts of the evapo- scribed in the literature, but is assumed to be close to the
rating drop using equation (2). D was calculated from [9] value for water, i.e. in the range of 0.91.0 [12]. By vary-
and the partial pressures were derived from the tempera- ing the emissivity during the analysis of the IR images
ture [10]. We chose a Sherwood number of 3.5 correspond- and by comparing the results with simultaneous measure-
ing to a sound pressure level of approximately 155 dB [7]. ments taken with a fine thermocouple we obtained the
The results are presented in Table 1 and mostly show best agreement with an emissivity value of 0.95 in the case
agreement between the evaluated values of K and refer- of n-heptane and 1-butanol drops. This value is confirmed
ence values from the literature [11]. There was an increas- in the literature for n-decane [12] and 1-propanol [14].
ing disagreement observed for higher alkanes and alka- Therefore, we used this value in the following analyses of
nols only. IR images of all hydrocarbons.

Evaporation of water drops

The measurements were validated with drops of distilled
water, whose emission coefficient = 0.96 is known from
the literature [12,13]. In Fig. 3, temperature measurements
of an evaporating water drop by a 0.25 mm-diameter ther-
mocouple and by the IR thermography system using dif-
ferent emissivity values are compared. The measurements
show good agreement for = 0.96 which corresponds to
the literature value.
To prove the validity of the method we calculated the
evaporation constant K from size measurements and the
surface temperature of evaporating water drops by vary-
ing the relative humidity of the environment. As shown in
Fig. 4, K depends linearly on the relative humidity. Fur-
thermore, we obtained good agreement between K calcu-
Fig. 4 Calculated values of evaporation constant K of water drops
lated from the surface temperature using Eq. (2) and K at different relative humidities of the environment and at a temper-
values determined directly from the size measurements ature of 21 C. The constant K was derived from size measure-
during the evaporation process. ments () and surface temperatures ()
 125

Fig. 5 Drop size and surface temperature of n-alkane and 1-alkanol drops evaporated in ambient air at 21 C and 38% relative humidity,
plotted as a function of time
126
A series of size and temperature measurements for
alkanes from n-pentane to n-nonane is shown in Figs. 5a
to e. The temperature and relative humidity of the envi-
ronment during these experiments was 21 C and 38%, re-
spectively. The surface of the suspended drops mostly de-
creased linearly with the evaporation time except at the
end of the measurements when the solvent was evapo-
rated. Here, a residual droplet consisting of impurities or
condensed water sometimes remains. The water is con-
densed from the humid environment onto the cold surface
of the evaporating drop. In case of n-pentane (see Fig. 5a),
where the surface temperature was close to 20 C, one
could clearly observe this effect. Furthermore, as a result
of the very low temperature of evaporating n-pentane
drops, ice particle formation from the condensed water
could be observed inside the suspended drop. In the ex-
ample of Fig. 5a, pure n-pentane evaporated into the hu-
mid environment. After approximately 33 s, all of the
pentane had evaporated and an ice particle formed for a
few seconds; the ice particle then melted and a water drop
remained. A step (between 34 and 38 s in the given ex-
ample) occurred in the time evolutions of drop surface
and temperature, corresponding to the phase change of wa-
ter. For the temperature calculation of the ice particle from
IR thermography images we assumed an emission coeffi-
cient of 0.85 [13]. Condensation of water during evapora-
tion processes is also remarkable in the case of n-hexane
(see Fig. 5b), where a small drop of condensed water re-
mained at the end of the n-hexane evaporation. For the
higher homologues n-heptane, n-octane and n-nonane, con-
densation of water was neglected because of the higher
surface temperature of the evaporating drops and the in-
solubility of water in alkanes. The condensed water with-
in a suspended alkane drop in the case of n-pentane and
n-hexane formed a separate droplet-shaped phase which
was completely surrounded by the alkane phase. Phase
separation of non-miscible solvents in acoustically levi-
tated drops could be visualised by video images and has
also been reported in the literature [15] in the case of
hexane and water. In that way, the enclosed water did not
essentially change drop surface and drop temperature evo-
lution during the evaporation process.
The surface temperature of the alkane drops during
their evaporation was nearly constant within a range of
1 C. The suspended drop was cooling down from ambient
temperature to its typical evaporation temperature at the
beginning of the measurements only. At the end, when all
alkane had evaporated, a jump in the temperature evolu-
tion occurred.
Evaporation of 1-alkanol drops
The evaporation of suspended alkanol drops was per-
formed under the same conditions (ambient air at 21 C
and with a relative humidity of 38%). The experimental
results are shown in Figs. 5f to j. Size and temperature
evolution of 1-butanol and 1-pentanol drops during their
evaporation were similar to those of the alkanes. Drop
surface decreased linearly as reported in the case of 1-bu-
tanol in the literature [16], and the surface temperature was
nearly constant during the evaporation process. 1-Butanol
and 1-pentanol were only partly miscible, while 1-propa-
nol, ethanol and methanol were miscible with water. Be-
cause of the condensed water, the surface and temperature
evolution changed in the case of miscible compared to non-
miscible solutions. The surface decrease changed slowly
from that of a pure alkanol drop, to that of a pure water
drop, and the temperature ran continuously from the typi-
cal evaporation temperature of a pure alkanol to that of a
pure water drop. Moreover, in comparison with alkane
drops, generally higher amounts of condensed water gath-
ered in alkanol drops under the same conditions (size and
surface temperature of the drop and temperature and rela-
tive humidity of the environment). An example is illus-
trated by n-heptane and ethanol (see Figs. 5c and g): only
in the case of ethanol drops could water condensation be
observed while surface temperature of the evaporating
drop and temperature and relative humidity of the envi-
ronment were similar.
Conclusions
Evaporation of acoustically levitated drops of pure water,
n-alkanes and 1-alkanols in an environment of ambient
air, temperature and pressure have been studied. The gen-
eral and quantitative understanding of this process is nec-
essary for the application of acoustic levitation techniques
for sample pre-treatment and for the combination with an-
alytical techniques. It has been shown that the evaporation
constant K can be calculated from both contact-less size
or temperature measurements. The method has been vali-
dated by using water drops and the evaluated values of K
for alkane and alkanol drops are in agreement. Therefore,
the method can also be used to determine binary diffusion
coefficients from the evaluated evaporation constant K.
Future work will focus on the evaporation processes of bi-
nary liquid drops. As seen in the case of condensed water
in alkane and alkanol drops, different effects of surface
and mixing processes have to be taken into account.
Evaporation rates can be changed by covering the drop
surfaces with monomolecular films [17].
Acknowledgment The authors thank Prof. Dr. H. K. Cammenga
for helpful discussions.
References
1. Welter E, Neidhart B (1997) Fresenius J Anal Chem 357:345
350
2. Eberhardt R, Neidhart B (1999) Fresenius J Anal Chem 365:
475479
3. Rohling O, Neidhart B (2000) Fresenius J Anal Chem 368:
125129
4. Lierke EG (1996) Acoustica 82:220237
5. Tuckermann R, Bauerecker S, Neidhart B (2001) Phys Unserer
Zeit 32:6975
6. Fuchs N (1934) Phys Z Sowjetunion 6:224 ff

7. Yarin L, Brenn G, Kastner O, Rensink D, Tropea C (1999)
J Fluid Mech 399:151204
8. Lee CP, Wang TG (1990) J Acoust Soc Am 88:23672375
9. Fuller EN, Schettler PD, Griddings JC (1966) Ind Eng Chem
58:18 ff
10. Reid RC, Prausnitz JM, Sherwood TK (1987) The Properties of
Gases and Liquids. McGraw-Hill, New York
11. Daidzic N (1995) Nichtlineare Tropfenschwingungen und -hy-
phenverdampfungen in einem Ultraschall-Positionierer, Dis-
sertation, Universitt Erlangen-Nrnberg
12. Daif A, Bouaziz M, Bresson J, Grisenti M (1999) J Thermo-
physics 13:553555
127
13. Schuster N, Kolobrodov VG (2000) Infrarot-Thermographie.
Wiley-VCH Verlag, Berlin
14. Konishi T, Ito A, Saito K (2000) Appl Optics 39:42784283
15. Eberhardt R (1999) Entwicklung einer Ultraschall-Levitations-
Apperatur fr die Probenvorbereitung in der Mikro- und Spuren-
analyse. Dissertation, Universitt Marburg
16. Erbil HY, Dogan M (2000) Langmuir 16:92679273
17. Stosch R, Cammenga HK (2000) J Colloid Interface Sci 230:
291297
Anal Bioanal Chem (2002) 372 : 128135
DOI 10.1007/s00216-001-1112-y
O R I G I N A L PA P E R
Jos R. Chirinos Kaveh Kahen Su-Ann E. OBrien
Akbar Montaser
Mixed-gas inductively coupled plasma atomic emission spectrometry
using a direct injection high efficiency nebulizer
Received: 14 June 2001 / Revised: 13 August 2001 / Accepted: 24 August 2001 / Published online: 14 December 2001
Springer-Verlag 2001
Abstract Experimental studies and computer simulations
were conducted to identify plasma operating conditions
and to explore and contrast the excitation conditions of
Ar, Ar-O2, and Ar-He inductively coupled plasmas (ICPs)
for the introduction of microliter volumes of sample solu-
tions with a direct injection high efficiency nebulizer
(DIHEN). The best MgII 280.270 nm/MgI 285.213 nm ratio
(6.6) measured with Ar ICP atomic emission spectrome-
try for the DIHEN (RF power = 1500 W; nebulizer gas
flow rate = 0.12 L min1) was less than the ratio (8.2) ac-
quired on the same instrument for conventional nebuliza-
tion (1500 W and 0.6 L min1). Addition of small amounts
of O2 or He (5%) to the outer gas flow improved excita-
tion conditions in the ICP, that is, a more robust condition
(a MgII/MgI ratio of up to 8.9) could be obtained by using
the DIHEN with Ar-O2 and Ar-He mixed-gas plasmas,
thereby minimizing some potential spectroscopic and ma-
trix interferences, in comparison to Ar ICPAES.
Keywords Direct injection high efficiency nebulizer
DIHEN Mixed-gas inductively coupled plasmas
Atomic emission spectrometry Matrix interferences
Computer simulation
Introduction
Argon inductively coupled plasmas (Ar ICPs) are com-
monly used as desolvation, vaporization, atomization, ex-
citation, ionization sources in analytical atomic emission
and mass spectrometry [1, 2]. Argon-supported ICP dis-
charges provide low detection limits, wide concentration
dynamic ranges, and minimal matrix effects. Mixed-gas
J.R. Chirinos K. Kahen S.-A.E. OBrien A. Montaser ()
Department of Chemistry, The George Washington University,
Washington, DC 20052, USA
e-mail: montaser@gwu.edu
J.R. Chirinos
On leave from Centro de Qumica Analtica,
Universidad Central de Venezuela, Caracas 1041a, Venezuela
plasmas are formed when N2, O2, air, H2, or He are added
to one or more gas flows of the ICP torch. Almost all
commercial ICP instruments are sustained in Ar, rather
than less expensive molecular gases such as air, N2, and
O2, mainly because the Ar ICP is formed more easily at
reasonable power levels (12 kW) and generally provides
better detection limits for elements possessing spectral
lines below 350 nm [3, 4]. However, mixed-gas plasmas
offer a superior ability to decompose refractory particles,
increased tolerance to higher solvent and analyte loads,
and less structured background spectra above 350 nm,
leading to improved analytical performance in certain ap-
plications [3, 5, 6, 7]. These characteristics allow for the
analysis of a wider range of samples than are possible us-
ing Ar alone. For example, introduction of O2 or He to the
outer gas flow of the Ar ICP is effective in suppressing
the background caused by organic solvents [8, 9] and in
improving the evaporation efficiency of slurries [10, 11]
when conventional nebulizer-spray chamber arrangements
are used. However, except for two recent conference pre-
sentations [12, 13], no report is available on studies of
mixed-gas plasmas for the direct injection of the sample.
The principal aim of this research is to explore the ex-
citation conditions of the mixed-gas ICPs for the direct in-
troduction of test solutions using the DIHEN. The DIHEN
is a low consumption micronebulizer (1100 L min1)
for ICP mass spectrometry (ICPMS) [14, 15, 16, 17, 18,
19, 20, 21, 22, 23, 24, 25] that operates at very low nebu-
lizer gas flow rates (ca 0.2 L min1) compared to conven-
tional pneumatic nebulizer-spray chamber arrangements
(ca 1.0 L min1). Initial studies describing the use of the
DIHEN for ICP atomic emission spectrometry (ICPAES)
were presented by Montaser and coworkers [12, 13] and
Todol and Mermet [26, 27]. The latter work, however,
was limited to Ar ICPAES measurements at a maximum
RF power of 1300 W. In contrast, our current and previ-
ous accounts [12, 13] consider the efficacy of the DIHEN
for atomic emission studies of the Ar ICP and the mixed-
gas plasmas operated at 1500 W. Such a study is signifi-
cant because plasma-related matrix effects are usually
more serious for direct liquid introduction [27] since the
 129

mass of sample (and solvent) presented to the plasma is
greater and usually has larger droplets than the aerosol in-
jected in conventional nebulization. It has been reported
that only 45% of the aerosol produced by the DIHEN are
composed of droplets less than 10 m [19]. Because the
DIHEN is delicate and relatively expensive, our study
was initially guided by computer simulation of Ar, Ar-O2,
and Ar-He plasmas to locate a thermally safe region to
place the DIHEN within the torch. To explore the excita-
tion condition or robustness of the mixed-gas ICP versus
the Ar ICP, the ratio of MgII/MgI was also measured for
the DIHEN with the plasma under the end-on viewing
condition and compared to results using conventional
nebulization. Finally, the effects of acid concentration and
easily ionized elements on the MgII/MgI ratios were ex-
amined for the DIHEN and contrasted to previous work
[26, 27]and our findings for conventional nebulization us-
ing the Ar ICP and the Ar-O2 plasma [12, 13].
dynamic equilibrium (LTE), (2) a steady state and laminar flow
prevails in the ICP torch, (3) the plasma is axially symmetric and
optically thin, (4) viscous dissipation and displacement currents
are negligible, and (5) flow, electromagnetic, temperature, and
concentration fields are two-dimensional.
Instrumentation and operating conditions
The instrumentation and operating conditions are shown in Table 1.
A free-running generator (Elan 5000 ICPMS generator), modified
in our laboratory [30], was used for all ICPAES studies. The outer
gas flow consisted of either Ar or Ar-O2 and Ar-He mixtures. Ar-
gon was used in the intermediate and nebulizer gas flows.
To estimate the robustness of the ICP [31], the line intensity
ratio MgII 280.270 nm/MgI 285.213 nm was measured using a
5 g mL1 solution of Mg prepared by diluting a 1000 g mL1
stock solution (Spex Certiprep Inc., Metuchen, NJ) in distilled,
deionized water (18 M cm). The effect of 2 M HNO3 (Optima
grade, Fisher Scientific, Pittsburgh, PA) and 0.1% and 1% NaCl
(ACS grade, Fisher Scientific) on plasma robustness was also in-
vestigated.

Experimental Results and discussion

Computer code for the simulation of plasma temperature
in the vicinity of the DIHEN nozzle The plasma operating conditions for DIHEN-ICPAES

Version 2.0 of the High Frequency Induction (HiFI) plasma code
[28] was used to calculate properties of Ar and mixed-gas plasmas
as a function of the operating conditions listed in Table 1. The
code predicts plasma temperature, gas velocity, power input, gas
enthalpy, and magnetic flux density [5, 28, 29]. Only the result for
plasma temperature was used in this work as an initial guide for lo-
cating a thermally safe region within the ICP torch to place the
DIHEN, that is, to prevent excessive heating of the DIHEN nozzle
due to the changes in the plasma operating conditions. The results
of this code must be treated as estimates because the model as-
sumes the following: (1) the plasma is in a state of local thermo-
In general, ignition of the Ar plasma for the DIHEN-
ICPAES measurement is straightforward and similar to
conditions used for conventional nebulization, that is, the
plasma is ignited at outer, intermediate, and nebulizer gas
flows of 15, 1.2, and 0 L min1, respectively. The Ar
ICPAES operating conditions under the end-on viewing
mode are generally comparable to those used for Ar
ICPMS [14, 15, 16, 17, 18, 19, 20, 21], that is, the RF
power, nebulizer gas flow rate, and solution uptake rate

Table 1 Operating conditions for DIHEN-ICPAES instrument

Operating parameter Description

RF generator PE-Sciex Elan 5000 (Perkin-Elmer Corporation, Norwalk, CT)

RF power (W) 1500
Nominal frequency (MHz) 32 a
RF generator type Free-running (Elan 5000 generator modified in our laboratory) [30]
Outer gas flow rate (L min1) 15 Ar, and mixtures (2.5, 5 and 10% v/v) of Ar-O 2 or Ar-He. Matheson gas proportioner
(Model MFMR-0800-AA, with tubes E200/E300) was used to mix gases.
Nebulizer gas flow (L min 1) 0.15
Intermediate gas flow rate (L min 1) 0.5, 1.2 (for plasma ignition and operation)
Solution flow mode Continuous
Nebulizer systems DIHEN (Model 170-AA, J E Meinhard Associates, Inc., Santa Ana, CA) and crossflow
nebulizer-Scott spray chamber.
Spectrometer Czerny-Turner monochromator (Model 2061, McPherson Instrument, Acton, MA) having
1.00 m of focal length and 1200 lines m1grating. The entrance and exit slits were 50 mm each.
Observation position The plasma was viewed end-on. It was imaged (1:5) onto the entrance slit of the spectrom-
eter using a 5.00 cm plano-convex lens having a focal distance of 25 cm. This arrangement
was not optically optimal, but was necessary to protect the available lens-monochromator
system from excessive heat generated by the ICP. Furthermore, no shear gas or a water-
cooled cone interface was used to remove the cool fringe of the plasma prior to observation.
Detection System Photomultiplier tube (Type FACT50, EMI Electronic Ltd., Middlesex, England) cooled to
28C and operated at 800 V.
a
Measured in this work. The manufacturer designed the generator as a 391 MHz power supply
130

are typically 1500 W, 0.15 L min1, and 80 L min1, re-

spectively, for the DIHEN positioned 2 mm below the top
of the intermediate tube of the ICP torch. Under such con-
ditions for the Ar ICP, no erosion of the DIHEN tip is ob-
served. As discussed below under computer simulation,
the DIHEN tip is well protected from excessive heating if
an intermediate gas flow of 1.2 L min1 is maintained for
operation with an Ar ICP. For studies of some mixed-gas
plasmas, the DIHEN tip had to be positioned 3 mm below
the top of the intermediate tube in order to shield it from
the hot zone of the ICP. This position was used throughout
this work for a consistent comparison of Ar ICP and
mixed gas plasmas. Visual observation of the well-known
yttrium bullet revealed that the aerosol was confined to
the axial channel of the plasma.

Robustness of the plasma generated in Ar ICP

using DIHEN

In ICPAES, the efficiency of the energy transfer in the

plasma is expressed by plasma robustness [31], as measured
by the intensity ratio MgII 280.270 nm/MgI 285.213 nm.
Higher ratios, indicative of greater plasma robustness [32,
33, 34, 35, 36, 37, 38, 39, 40], are achieved by applying
high RF power and by increasing the residence time of the
sample aerosol in the plasma. The use of axial viewing
generally leads to lower values of MgII/MgI ratios be-
cause the whole axial channel is probed, and the observa-
tion cannot be optimized for ionic lines, as performed by
radial viewing [41]. Under robust conditions, no signifi-
cant variation in the ratios is observed due to changes in
the matrix or reagent composition.
The previous DIHEN-Ar ICPAES studies were con-
ducted at 1300 W, a non-robust plasma condition, and re-
sulted in unsatisfactory MgII/MgI ratios (<6) [26, 27].
Furthermore, matrix effects were only significantly mini- Fig. 1 Experimental ratio of Mg II/Mg I as a function of (A) neb-
ulizer gas flow rate, (B) RF power and (C) solution uptake rate for
mized at low solution uptake rates (3040 L min1) for the Ar ICP using the DIHEN
lines possessing low sum values (Esum) of ionization and
excitation energies [26, 27]. In the present work, MgII/
MgI ratios are measured at 1500 W. Plots of MgII/MgI ra- the DIHEN is 6.6 (at input RF power and nebulizer gas
tios for Ar ICP are shown in Fig. 1 as a function of nebu- flow rate of 1500 W and 0.12 L min1, respectively), less
lizer gas flow rate, RF power, and solution uptake rate us- than 8.2 acquired by our ICPAES instrument for conven-
ing the DIHEN. The ratios are the mean of three measure- tional nebulization (at input RF power and nebulizer gas
ments on different days and the precision is within 5%. flow rate of 1500 W and 0.6 L min1, respectively). Also,
Fig. 1A and B illustrate that the highest MgII/MgI ratios the best ratio for the DIHEN is notably less than the val-
(ca 6.5) for the DIHEN are obtained at very low nebulizer ues (7.311.9) reported for ICPAES instruments such as
gas flow rates (0.15 L min1) and high power (1500 W). the Optima 3000XL [33, 37] (Perkin-Elmer Corporation)
However, at nebulizer gas flow rates less than 0.15 L min1, and Liberty Series II [38] (Varian Inc. Palo Alto, CA) op-
the plasma became unstable. The ratio is not strongly af- erated with conventional nebulizer-spray chamber arrange-
fected by the solution uptake rate (Fig. 1C), although ments. Accordingly, the Ar ICP is less robust in the pres-
slightly higher ratios were obtained at uptake rates below ence of the DIHEN compared to conventional nebuliza-
100 L min1. tion.
Table 2 presents a compilation of MgII/MgI data for a The data in column six of Table 2 present the percent-
few spectrometers using aqueous solution introduction by age decrease in the MgII/MgI ratio for changes in the
the DIHEN and conventional nebulizer-spray chamber plasma operating conditions towards the less robust state.
arrangements under the best operating conditions reported For the DIHEN, the MgII/MgI ratio decreases by 24%
for each nebulizer. The best MgII/MgI ratio obtained for when the RF power is reduced from 1500 to 1300 W. This
 131
Table 2 Ratios of MgII 280.270 nm/MgI 285.213 nm for various axially viewed Ar ICPs using DIHEN and conventional nebulizer-
spray chamber arrangements for aqueous solution introduction
Instrument Nebulization Input RF Nebulizer gas MgII/MgI Decrease of
power flow rate MgII/MgI
(W) (L min1) (%)b
This work DIHEN 1500 0.12 6.6 24
1300 0.12 5.0
This work Crossflow nebulizer-Scott type spray chamber 1500 0.60 8.2 11
1300 0.60 7.3
Optima [37] Crossflow nebulizer-Scott type spray chamber 1500 0.60 9.7 6
1350 0.60 9.1
Liberty [38] Concentric nebulizer-cyclonic spray chamber 1400 180a 11.9 4
1200 180a 11.4
Optima [33] Modified conespray-Scott type spray chamber 1450 0.60 7.3 19
950 0.95 5.9
aKpa; bDecrease of MgII/MgI=[(MgII/MgI at higher powerMgII/MgII at lower power)/(MgII/MgI at higher power)100]

is in contrast to the 419% decline measured for conven-

tional nebulization for various instrumental set-ups cited
in Table 2. The 24% decrease in the MgII/MgI ratio for the
DIHEN is still worse than the 19% decline reported for the
Optima instrument [33] when the RF power (1450 W
950 W) and nebulizer gas flow rate (0.600.95 L min1)
are changed significantly for conventional nebulization.
These results collectively suggest that the Ar ICP is less
robust for the direct introduction of aerosol, compared to
conventional nebulization, even at 1500 W the power
level available for most ICPAES instruments. This finding
is in line with the previous DIHEN-Ar ICPAES studies
conducted at 1300 W [26, 27].
Since little improvement in the MgII/MgI ratio could
be realized with a regular Ar ICP, that is, by increasing the
power and decreasing the DIHEN gas flow rate, one may
consider mixed-gas plasmas as a possible approach to en-
hance the robustness of the ICP. Based on prior experi-
mental studies with conventional nebulization [3, 42, 43,
44, 45, 46] introduction of molecular gases to the outer
gas flow reduces the diameter of the axial channel, thus
Fig. 2 Simulated plasma temperature (K) as a function of the dis-
enhancing the sample-plasma interaction and analytical tance from the end of the torch intermediate tube for Ar-O2 ICPs.
measurements. The magnitude of these effects is related RF Power, nebulizer gas flow and intermediate gas flows are
to the plasma gas composition and RF power. Further- 1500 W, 0.15 L min1 and 1.2 L min1, respectively
more, the position of the mixed-gas plasma shifts down-
ward in the ICP torch [3, 42, 43], a process which may
damage the DIHEN nozzle. To counteract this downward the outer gas flow. In all cases, the predicted axial channel
shift, the intermediate gas flow may be increased. In the temperatures rise steeply within the load coil region
following section, computer simulation was applied to (317 mm), reach a peak near the top turn of the induction
consider the cited effects collectively, and to find safe op- coil, and then decline gradually at distances of 2060 mm.
erating conditions for the DIHEN nozzle based on pre- At distances over 20 mm (Fig. 2A), the temperature de-
dicted plasma temperatures. clines as the O2 composition in the outer gas is increased,
resulting in a more dense plasma that is smaller than the
Ar ICP. Close to the intermediate tube of the torch, how-
Simulation of the axial temperature profiles, ever, the axial temperature rises as the O2 content is in-
as a function of gas composition, creased, especially when O2 completely replaces Ar. For
using Ar-O2 and Ar-He mixtures in the plasma gas flow example, at 2.5, 4.2, and 6.5 mm above the intermediate
tube, temperatures of 350, 1129, and 5931 K, respectively,
Fig. 2 shows simulated axial temperature distributions of are predicted for 100% O2 in the outer gas (Fig. 2B). For
Ar and Ar-O2 ICPs above the intermediate tube of the 5% O2 in the outer gas, the estimated temperatures are
torch for the introduction of 0, 5, 10, 20, and 100% O2 in 350, 438, and 2060 K (Fig. 2B) at 2.5, 4.2, and 6.5 mm
132
Fig. 3 Simulated plasma temperature (K) as a function of the dis-
tance from the end of the torch intermediate tube for Ar-He ICPs.
RF power, nebulizer gas flow and intermediate gas flows are
1500 W, 0.15 L min1 and 1.2 L min1, respectively
above the intermediate tube, respectively. The wide tem-
perature variations at the base of the plasma collectively
suggested that the DIHEN nozzle should be positioned at
larger distances below the intermediate tube when the per-
centage of O2 in the outer gas is extensively altered; oth-
erwise, the DIHEN nozzle would melt. Accordingly, the
DIHEN nozzle was placed 3 mm below the intermediate
tube for Ar-O2 ICP as compared to 2 mm for Ar ICP. One
must note, however, that the temperature rise predicted at
the base of the mixed-gas plasma is useful analytically be-
cause it enhances droplet desolvation and diminishes sol-
vent effects. Indeed, the cited issues are difficult to attack
when the sample aerosol is directly injected into the
plasma, especially for the large droplets.
Axial temperature profiles are shown in Fig. 3 for Ar
and Ar-He plasmas. At large distances from the nebulizer
(4060 mm), the plasma temperature is reduced with the
introduction of He. In contrast to the results obtained with
Ar-O2 plasmas, the predicted axial channel temperature
close to the intermediate tube (Fig. 3B) does not signifi-
cantly increase as more He replaces Ar in the outer gas,
suggesting that the DIHEN nozzle could be safely posi-
tioned 2 mm below the intermediate tube. Based on the
predicted temperatures, the spatial structures of the Ar-He
ICPs and the Ar ICP are nearly identical, hence the MgII/
MgI values of Ar-He ICPs should be less (see below) than
those measured in Ar-O2 plasmas.
Fig. 4 Simulated plasma temperature (K) as a function of the dis-
tance from the end of the torch intermediate tube for Ar-O2 ICP
with 10% O2 in the outer gas flow: (A) intermediate gas flow=
1.2 L min1 at different nebulizer gas flows and, (B) nebulizer gas
flow=0.15 L min1 at different intermediate gas flows
Simulation of the axial temperature profiles
at different nebulizer and intermediate gas flow rates
The predicted axial temperature profiles of the plasma at
different nebulizer gas flow rates are shown in Fig. 4A for
10% O2 in the outer tube. The plasma temperature in-
creases significantly close to the intermediate tube when
the nebulizer gas flow is reduced or stopped. These results
are significant for DIHEN operation, considering that the
DIHEN is typically operated at comparatively low nebu-
lizer gas flows (0.10.2 L min1). A nebulizer gas flow
rate of 0.15 L min1 was chosen for experimental work. At
lower nebulizer gas flows, either the DIHEN nozzle
would gradually clog due to excessive heating or the
aerosol would not penetrate the axial channel, resulting in
plasma instabilities.
The effect of the intermediate gas flow rate on axial
temperature distributions of the plasma is shown in Fig. 4B
for 10% O2 in the outer tube. The simulated profiles are
obtained at intermediate gas flow rates of 0.5, 1.2, and
2.0 L min1. The plasma temperature increases signifi-
cantly close to the intermediate tube when the intermedi-
ate gas is reduced from 2 to 0.5 L min1. For example, at
an intermediate gas flow of 2.0 L min1, the predicted
temperatures are 350, 414, and 1695 K at 2.5, 4.2, and
6.5 mm above the intermediate tube, respectively. At
0.5 L min1, the estimated temperatures are 350, 672, and
3875 K at 2.5, 4.2, and 6.5 mm above the intermediate
tube, respectively. These results collectively illustrate that
the intermediate gas flow plays, compared to conven-
 133

tional nebulization, a more critical role in operating a sta-

ble mixed-gas plasma with the DIHEN because the nebu-
lizer nozzle may easily clog from excessive heating. Sim-
ilar results are obtained with 10% He in the outer tube
(not presented here). Based on these results, an intermedi-
ate gas flow of 1.2 L min1 was used for subsequent ex-
periments.
The above simulated data were used to establish initial
operating conditions for DIHEN-ICPAES measurements
using mixed-gas plasmas. The predicted results were then
confirmed experimentally by using a dummy nozzle be-
fore installing the DIHEN. The dummy nozzle resembles
the DIHEN shell, except that it does not include a solution
capillary. In general, the dummy nozzle is a useful tool for
establishing safe operating conditions for the DIHEN
when the plasma is operated under conditions different
from those commonly used for the Ar ICP. One must note
that the plasma operated with a dummy nozzle has differ-
ent characteristics compared to the DIHEN because ef-
fects of aerosol and nebulizer gas flow are not considered.
However, the combined effort involving computer simula-
tion and the application of the dummy nozzle is most ef-
fective in establishing initial operating conditions.
Based on the studies described above, the DIHEN was
positioned 3 mm below the torch intermediate tube and Fig. 5 Experimental ratio of MgII/MgI as a function of nebulizer
was operated at the nebulizer and intermediate gas flow of gas flow rate; (A) Ar-O2 ICPs and (B) ArHe ICPs. RF power,
nebulizer gas flow, and intermediate gas flows were 1500 W,
0.15 L min1 or greater and 1.2 L min1, respectively, 0.15 L min1 and 1.2 L min1, respectively
when mixed-gas plasmas were used. Under such condi-
tions, no erosion in the DIHEN tip was noted. A recent re-
port refers to the deterioration of the DIHEN by the zle, increasing the risk of damage to the nebulizer. A mixed-
plasma at very low nebulizer gas flow rates (<0.3 L min1) gas ICP is a viable alternative to improve the excitation
and 1300 W [26, 27] however, in our view, this observa- conditions of the plasma in the presence of the DIHEN,
tion is attributed to the use of a low intermediate gas flow especially in this study, when the instrument is operated at
rate (0.5 L min1). maximum possible RF power and minimum possible neb-
ulizer gas flow rate.
Robustness of the plasma generated in Ar-O2
and Ar-He mixture using DIHEN Matrix effects in DIHEN-ICPAES

Plots of MgII/MgII ratios for Ar, ArO2, and Ar-He ICPs
are shown in Fig. 5 as a function of nebulizer gas flow rate
for the DIHEN. The ratios are the mean of three measure-
ments on different days and the precision is within 5%.
The Ar-O2 ICPs exhibit higher ratios compared to Ar ICP
(Fig. 5A). For example, the MgII/MgI ratio is approxi-
mately 8.5 and 6 for Ar-O2 ICP and Ar ICP, respectively,
at a nebulizer gas flow of 0.15 L min1. No significant dif-
ference was observed for mixed-gas plasmas having 2.5,
5, and 10% v/v of O2 in the outer gas flow. Fig. 5B shows
similar results for Ar-He plasmas, that is, Ar-He plasmas
exhibit higher MgII/MgI ratios compared to the Ar-ICP,
but the ratios are slightly lower than those obtained with
Ar-O2 ICP.
The above results jointly suggest that mixed-gas plas-
mas improve the robustness of the ICP when a relatively
small amount of the foreign gas is introduced (ca 2.5%).
For higher percentages of foreign gas, no significant im-
provement in the MgII/MgI ratios is realized, but the hot
zone of the plasma approaches the tip of the DIHEN noz-
Introduction of an inorganic acid, such as HNO3, or NaCl
is known to suppress or enhance analytical signal in
ICPAES, but the effects can be minimized for conven-
tional nebulization by using robust plasma conditions,
that is, reduced nebulizer gas flow and high RF power
[35, 37, 38, 39, 47]. Table 3 includes MgII/MgI ratios for
the introduction of HNO3 solutions by the DIHEN and
nebulizer-spray chamber arrangement operating under op-
timal operating conditions. Again, the ratios are the mean
of five measurements on different days and the precision
is within 5%. For the Ar ICP, the MgII/MgI ratios (Table 3)
measured with the DIHEN decrease only slightly (6.66.2
or 6%) as the HNO3 concentration is increased from 0 to
2 M. These data suggest that DIHEN-ICPAES is rela-
tively robust when Ar ICP is used at 1500 W. Todol and
Mermet also reported no matrix effect by nitric acid for a
1300 W Ar ICP, but a lower solution uptake rate (30
40 L min1) had to be used [26, 27]. Improved MgII/MgI
ratios are also noted for Ar-O2 ICP (Table 3), however the
level of suppression (6%) remained unchanged when the
134
Table 3 Ratio of MgII
280.270 nm/MgI 285.213 nm Nebulization system
for aqueous nitric acid and
sodium chloride solutions us-
ing DIHEN and conventional Plasma
nebulization Solution
Aqueous
Nitric acid (2 M)
NaCl (0.1%)
NaCl (1%)
Nebulizer gas flow (L min1)
Solution uptake (L min1)
RF power (W)
DIHEN was used. In contrast, with conventional nebu-
lization, acid effects are more pronounced with Ar-O2 ICP
compared to the Ar ICP, i.e., the mixed-gas plasma is less
robust than the Ar ICP. The main reason for this change in
plasma robustness is unclear to us. Obviously, the pres-
ence of the spray chamber makes interpretation of the re-
sult more difficult.
Table 3 also includes MgII/MgI ratios in the presence
of 0.1% NaCl introduced by the DIHEN and the conven-
tional nebulizer-spray chamber for Ar ICP and Ar-O2 ICP.
Solutions containing 1% Na were not used with the
DIHEN because the nebulizer is prone to clogging. The
data in Table 3 suggest that Na suppresses MgII/MgI
more severely than HNO3, both for the conventional neb-
ulization system and the DIHEN. Importantly, the use of
the Ar-O2 ICP was not very effective in reducing Na ma-
trix effects at 1500 W. In contrast, the Na matrix effect
was negligible at 1300 W and at a solution flow rate of
nearly 80 L min1 in Ar ICPAES for micronebulization
with a high efficiency nebulizer-cyclonic spray chamber
[26, 27]. Accordingly, to eliminate the Na matrix effect in
ICPAES, the DIHEN must present a smaller amount of
sample to the plasma at RF power level greater than 1500 W.
This conclusion is in line with the work of Todol and
Mermet [26, 27].
The above data on MgII/MgI ratios also brings some
uncertainty to the application of this ratio as an indicator
of plasma robustness [31]. At least for direct injection, it
is prudent to conduct further experiments to validate the
usage of MgII/MgI as an indicator of plasma robustness.
Conclusions
This paper describes ways to overcome the limitations of
the DIHEN. Computer simulation was applied to predict
the operating conditions of mixed-gas ICPs using the
DIHEN. The simulated temperature profiles show that
Ar-O2 and Ar-He plasmas provide better excitation condi-
tions than the Ar ICP but may create a hot zone close to
the DIHEN nozzle. For mixed-gas plasmas, the DIHEN
should be positioned farther below the torch intermediate
tube compared to its typical location in Ar ICPs. In gen-
eral, the DIHEN provides robust conditions (values of
DIHEN Crossflow nebulizer-Scott
type spray chamber
Ar Ar-O2 Ar Ar-O2
6.6 8.9 8.2 7.0
6.2 8.4 7.7 6.0
6.0 6.2 7.0 6.2
Nebulizer Nebulizer 5.9 5.7
clogs clogs
0.15 0.15 0.60 0.60
85 85 600 600
1500 1500 1500 1500
MgII/MgI greater than 6) for some Ar ICPAES measure-
ments at high RF power (ca 1500 W) and low nebulizer
gas flow (ca 0.15 L min1), however, the ratios were less
than those noted for a conventional nebulizer-spray cham-
ber arrangement. Slightly larger MgII/MgI ratios could be
obtained using Ar-O2 and Ar-He mixed-gas plasmas with
the DIHEN for the addition of the foreign gas to the outer
gas flow; nevertheless, the level of suppression (6%) for
the introduction of a solution of 2 M HNO3 remained un-
changed. With conventional nebulization, acid effects were
more pronounced with Ar-O2 ICP compared to the Ar ICP.
In general, results of these studies at 1500 W on mixed-
gas ICPAES with the DIHEN agree with the work of
Todol and Mermet who operated a 1300 W Ar ICP [26,
27]. Clearly, better MgII/MgI ratios were obtained be-
cause the ICP was operated at 1500 W. It was also demon-
strated that the plasma required RF power above 1500 W
in the presence of the DIHEN to run under robust condi-
tions, but only 1300 W was necessary for a pneumatic
nebulizer-spray chamber arrangement. Todol and Mer-
met also concluded that the acid effects are eliminated for
the DIHEN, but only at a solution uptake rate of 3040 L
min1 [27]. Operation at 1500 W, as demonstrated in this
work, allows introduction of 80 L min1 with the DIHEN
with nearly no HNO3 effect. At 1300 W, Todol and Mer-
met reported minimal Na matrix effect with the DIHEN,
but only for spectral lines possessing low Esum [27]. For
lines having high Esum values, as shown in this work, the
Na matrix effect could not be remedied at 1500 W or
through the use of mixed-gas plasma. To create a robust
plasma condition for direct injection, one must reduce the
amount of aerosol and decrease the droplet size intro-
duced into the ICP to the level prevalent in conventional
pneumatic nebulization (2040 mg min1), increase the
RF power beyond the level required for conventional neb-
ulization, and enhance plasmasample interaction by op-
erating a mixed-gas plasma.
It would be interesting to introduce a foreign gas to the
nebulizer gas flow as no report is currently available on
the efficiency of the DIHEN with a large proportion of
such a gas. In this case, the axial channel would be wider
[3], but an increase in thermal conductivity may result in
the central channel. Work in this area is in progress in our
laboratory.

Acknowledgements This research was sponsored by grants from
the US Department of Energy (DE-FG0293ER14320), the Na-
tional Science Foundation (CHE-9505726 and CHE-9512441),
and JE Meinhard Associates, Inc. The authors are grateful to John
A. McLean and Craig M. Benson for constructive discussions and
assistance and to Mr. Bill Rutkowski for excellent machine shop
service.
References
1. Montaser A, Golightly DW (1992) Inductively Coupled
Plasma in Analytical Atomic Spectrometry, 2nd edn. VCH-
Wiley, New York
2. Montaser A (1998) Inductively Coupled Plasma Mass Spectro-
metry. VCH-Wiley, New York
3. Montaser A, Van Hoven RL (1987) Crit Rev Anal Chem 18:
451031
4. Zander AT (1992) In: Montaser A, Golightly DW (eds) Induc-
tively Coupled Plasma in Analytical Atomic Spectrometry, 2nd
edn. VCH-Wiley, New York, pp 299339
5. Cai M, Haydar DA, Montaser A, Mostaghimi J (1997) Spectro-
chim Acta, Part B 52:369
6. Montaser A, Zhang, H (1998) In: Montaser A (ed) Inductively
Coupled Plasma Mass Spectrometry. VCH-Wiley, New York,
pp 809890
7. Sesi NN, MacKensei A, Shanks KE, Yang P, Hieftje GM (1994)
Spectrochim Acta, Part B 49:1259
8. Botto R (1993) J Anal At Spectrom 8:51
9. Botto R, Zhu JJ (1996) J Anal At Spectrom 11:675
10. Ebdon L, Foulkes M, Sutton K (1997) J Anal At Spectrom
12:213
11. Heitland P, Broekaert JAC (1997) J Anal At Spectrom 12:981
12. Chirinos JR, OBrien SE, Acon BW, Montaser A (2000) Mixed-
Gas Inductively Coupled Plasma Spectrometry Using a Direct
Injection High Efficiency Nebulizer, presented at the 27th An-
nual Conference of the Federation of Analytical Chemistry and
Spectroscopy Societies, Nashville, TN, September 2428
13. Montaser A (2001) Sample Introduction into Plasmas: The Sur-
vival of the Weakest, Invited Lecture at the European Winter
Conference on Plasma Spectrochemistry, Lillehammer, Nor-
way, February 48
14. McLean JA, Zhang H, Montaser A (1998) Anal Chem 70:1012
15. Montaser A, McLean JA, Kacsir JM (2000) US Patent 6 166
379
16. Singh J, McLean JA, Pritchard DE, Montaser A, Patierno SR
(1998) Toxicol Sci 46:220
17. Becker JS, Dietze HJ, McLean JA, Montaser A (1999) Anal
Chem 71:3077
18. Singh J, Pritchard DE, Carlisle DL, McLean JA, Montaser A,
Orenstein JM, Patierno SR (1999) Toxicol Appl Pharmacol
161:240
19. Acon BW, McLean JA, Montaser A (2000) Anal Chem 72:
1885
20. McLean JA, Acon BW, Montaser A, Singh J, Pritchard DE,
Patierno SR (2000) Appl Spectrosc 54:659
135
21. Minnich MG, Montaser A (2000) Appl Spectrosc 54:1261
22. McLean JA, Minnich MG, Su J, Lai W, Montaser A (2000)
Anal Chem 72:4796
23. Minnich MG, McLean JA, Montaser A (2001) Spectrochim
Acta, Part B 56:1113
24. Acon BW, McLean JA, Montaser A (2001) J Anal At Spectrom
16:852
25. Westphal CS, McLean JA, Acon BW, Allen LA, Montaser A
(2001) Anal Chem (submitted)
26. Todol J-L, Mermet J-M (2000) Advantages of Working with
Low Consumption Direct Injection Nebulizers in Plasma Spec-
trochemistry, presented at the 27th Annual Conference of the
Federation of Analytical Chemistry and Spectroscopy Soci-
eties, Nashville, TN, September 2428
27. Todol J-L, Mermet J-M (2001) J Anal At Spectrom 16:514
28. Boulos MI, Mostaghimi J, Proulx P (1989) HIFI Plasma Com-
puter Code, version 2.0, available from MI Boulos, Department
of Chemistry Engineering, University of Sherbrooke, Sheer-
broke, Que., J1 K 2R1, Canada
29. Mostaghimi J, Boulos MI (1992) In: Montaser A, Golightly
DW (eds) Inductively Coupled Plasma in Analytical Atomic
Spectrometry, 2nd edn. VCH-Wiley, New York, pp 949984
30. Iacone LA, Masamba WRL, Nam S-H, Zhang H, Minnich MG,
Okino A, Montaser A (2000) J Anal At Spectrom 15:491
31. Mermet J-M (1991) Anal Chim Acta 250:85
32. Poussel E, Mermet J-M, Samuel O (1993) Spectrochim Acta,
Part B 48:743
33. Ivaldi J, Tyson JF (1995) Spectrochim Acta, Part B 50:1207
34. Romero X, Poussel E, Mermet J-M (1997) Spectrochim Acta B
52: 487
35. Romero X, Poussel E, Mermet J-M (1997) Spectrochim Acta,
Part B 52:495
36. Brenner IB, Zander AT (2000) Spectrochim Acta, Part B 55:
1195
37. Brenner IB, Zischka M, Maichin B, Knapp G (1988) J Anal At
Spectrom 13:1257
38. Brenner IB, Zander A, Cole M, Wiseman A (1998) J Anal At
Spectrom 13:897
39. Todol J-L, Mermet J-M (1999) Spectrochim Acta, Part B 54:
895
40. Dennaud J, Howes A, Poussel E, Mermet J-M (2001) Spectro-
chim Acta, Part B 56:101
41. Dubuisson C, Poussel E, Mermet, J-M, Todol J-L (1998) J Anal
At Spectrom 13:63
42. Ishii I, Golightly DW, Montaser A (1988) J Anal At Spectrom
3:965
43. Ishii I, Montaser A (1990) J Anal At Spectrom 5:57
44. Montaser A, Fassel V, Zalewski J (1981) Appl Spectrosc 35:
292
45. Montaser A, Chan S, Vieira P, Huse G, Van Hoven R (1986)
Appl Spectrosc 40:473
46. Montaser A, Ishii I, Palmer BA, Layman LR (1990) Spec-
trochim Acta, Part B 45:603
47. Thompson M, Barnes RM (1992) In: Montaser A, Golightly
DW (eds) Inductively Coupled Plasma in Analytical Atomic
Spectrometry, 2nd edn. VCH-Wiley, New York, pp 249297
Anal Bioanal Chem (2002) 372 : 136140
DOI 10.1007/s00216-001-1136-3
O R I G I N A L PA P E R
Attila Gspr va Szles Harald Berndt
Analysis of submicroliter samples
using micro thermospray flame furnace atomic absorption spectrometry
Received: 2 August 2001 / Revised: 19 September 2001 / Accepted: 20 September 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract A recently described thermospray flame fur-
nace atomic absorption spectrometric (TS-FF-AAS) sys-
tem has been modified in order to extend the applicability
of the method for the determination of elemental traces in
very small sample volumes (microliter or submicroliter).
As an easily available, effective thermospray vaporizer,
a fused silica capillary was used and the liquid sample
was transported by 1 MPa (10 bar) gas pressure delivered
by a standard gas cylinder. A 0.3 L sample volume can
be analyzed with a higher power of detection than using
3 orders of magnitude larger sample volumes with con-
ventional flame atomic absorption spectrometry. The rela-
tive standard deviations (N=12) for 0.3 L volumes and
5 g/mL Pb samples amount to 3.1% and 3.8% in signal
height and signal area, respectively. The detection limit
was found to be 69 ng/mL. Initial experiments with other
elements (Cd, Hg, Tl, Zn) led to similar results.
Keywords Thermospray FFAAS Spectrometry,
atomic absorption, microthermospray
Abbreviations TS-FF-AAS Thermospray flame furnace
atomic absorption spectrometry FAAS flame atomic
absorption spectrometry ICP/MS inductively coupled
plasma/mass spectrometry GFAAS graphite furnace
atomic absorption spectrometry BIFF-AAS beam
injection flame furnace atomic absorption spectrometry
A. Gspr () . Szles
University of Debrecen,
Department of Inorganic and Analytical Chemistry,
POB. 21, 4010 Debrecen, Hungary
e-mail: gaspara@tigris.ktle.hu
H. Berndt
Institut fr Spektrochemie und Angewandte Spektroskopie,
Bunsen-Kirchhoff-Str. 11, 44139 Dortmund, Germany
Introduction
Although the importance of the highly sensitive atomic
spectrometric methods such as inductively coupled plasma/
mass spectrometry (ICP/MS) or graphite furnace atomic
absorption spectrometry (GFAAS) is significant, the much
cheaper and very simple flame atomic absorption spectro-
metric (FAAS) method remains a reliable element-selec-
tive method. In FAAS the analytical sensitivity can be im-
proved significantly by increasing the efficiency of aerosol
generation/transport and prolonging the residence time of
the free analyte atoms in the absorption volume [1]. In the
literature, many papers can be found concerning total
sample introduction or prolonged residence time (e.g.
Refs. [2, 3, 4, 5]). Recently, we published two techniques
that were based on both the total sample introduction and
a prolonged residence in the absorption volume of a flame-
heated tube furnace. These were beam injection flame fur-
nace atomic absorption spectrometry (BIFF-AAS) [6] and
thermospray flame furnace atomic absorption spectrome-
try (TS-FF-AAS) [7]. Due to both factors, a significant
improvement in the power of detection was achieved for
numerous elements. Using these flame furnace tech-
niques, 25 elements were determined, of which 17 ele-
ments exhibited a better power of detection than conven-
tional FAAS. In the case of BIFF-AAS, the sample was
introduced as a high speed liquid jet through a sample in-
troduction hole which then impacted on the inner surface
of a flame heated tube (flame furnace). The flame-heated
tube was placed in the air/acetylene flame on a standard
AAS burner head.
This flame-heated tube was also used in the thermo-
spray flame furnace AAS technique, where the liquid
sample to be analyzed is transported through a very hot
metal or ceramic capillary tip acting as a flame-heated
thermospray into the flame furnace. Compared to standard
flame AAS an improvement in the detection limits of be-
tween 14 and 67 for five investigated elements was found
[7]. The fundamental principle of this TS technique is
similar to that of the conventional thermospray techniques

[8, 9]. However, some remarkable differences exist. These
are the direct heating of the TS capillary with the flame
gases instead of a temperature-controlled electrical heat-
ing, the much higher temperature (up to 900 C), the short
length of the heated area (about 1 cm) with a strong tem-
perature gradient and a larger inner diameter of the capil-
lary. Another remarkable difference is that high pressure
is not needed for the transport of the liquids, which allows
the use of peristaltic pumps instead of HPLC pumps [7].
In this paper we show that it is possible to extend the ap-
plicability of the TS-FF-AAS method for determination of
elemental traces in submicroliter volumes of samples us-
ing a fused silica capillary as an easily available, effective
thermospray vaporizer and a standard gas cylinder as a
gas pressure pump for the transport of the sample.
Experimental
Spectrometer
A Philips model PU9200X atomic absorption spectrometer with
deuterium lamp background correction was used. Operating pa-
rameters (wavelength, slit, current for the hollow cathode lamp)
for the spectrometer were as recommended by the manufacturer. In
all experiments an acetylene/air flame was used. The original spec-
trometer had no suitable data/signal output to display transient sig-
nals, but with a separate personal computer and self-written soft-
ware, all signals could be visualized, recorded and stored. The
computer was interfaced to the analog output of the spectrometer
via an analog/digital converter unit. The software also permitted a
transfer of raw data to other computer programs, for example to a
chromatography software package (Eurochrom 2000, version 1.2,
Knauer, Berlin, Germany) for integration of transient signals and
to a spreadsheet program (Excel, Microsoft) for better displays of
the original signals.
TS-FF-AAS system
The schematic setup of the micro TS-FF-AAS system is shown in
Fig. 1. The arrangement of the system used in this work was simi-
lar to one already described [7]. There are two essential differ-
ences, the type of the pump and the TS capillary vaporizer. While
in the earlier work a peristaltic pump transported the sample, in the
present system a nitrogen or argon gas cylinder was used as a sim-
ple, pulsation-free device capable of forwarding small liquid sam-
ple plugs through a capillary with a small inner diameter into the
flame furnace. From the gas cylinder 0.12.5 MPa pressure gas
could be obtained via a standard pressure reduction valve. From
the point of view of our work the type of gas is not important. It
Fig. 1 Schematic setup of the micro TS-FF-AAS system for the
analysis of submicroliter samples
137
should be inert (no reaction with liquids or flame gases) and inex-
pensive (for instance nitrogen or technical argon). At the outlet of
the pressure reduction valve a special stainless steel connecting
piece (commercially available, Drger AG, 47807 Krefeld, Ger-
many) was used for a close connection between the reduction
valve and the PEEK capillary (o.d. 1/16, i.d. 0.5 mm) going to the
sample injection valve.
For sample introduction a six-port sample injection valve
(Knauer, Berlin, Germany) with various sample loops (0.350 L)
were applied. The sample loops were made using PEEK tubing
with various inner diameters. The smallest sample loop (0.3 L)
was prepared from the smallest inner diameter PEEK capillary
available on the market (64 m).
A flexible fused silica capillary (Polymicro Technologies,
Phoenix, USA) was tightened into the outlet of the sample intro-
duction valve. Since the outer diameter of the fused silica capillary
(365 m) was much smaller than that of the standard capillaries
(1/16) for these valves, it was screwed with the aid of a short
piece of standard PEEK capillary of 500 m inner diameter
(sheath piece) into the valve. The other end (outlet) of the fused
silica capillary was directed into the flame furnace through the in-
troduction hole. The length of the fused silica capillary was about
40 cm. In the experiments capillaries with inner diameters of 25,
50, 75, 100 and 200 m were used. The capillary was fixed with an
adjustable holder in front of the flame furnace.
The outlet end of the capillary can be considered as a thermo-
spray vaporizer. The last few millimeters of the capillary are
strongly heated due to the proximity of the glowing flame furnace.
A tight (hermetically sealed) connection between the end of the
capillary and the hole in the flame furnace is not needed, because
the vapor exits from the outlet of the capillary at high speed into
the tube furnace. The areas of both ends of the flame-heated tube
are much larger than the very small free remaining area of the sam-
ple introduction hole. Therefore, the vapor escaping from the fur-
nace will only exit at the ends of the tube and not through the in-
troduction hole. The end of the capillary can be easily put in and
out of the flame furnace through the introduction hole. The fused
silica capillary would melt and curve if it came into direct contact
with the flame gases, therefore, the flame-heated end of the capil-
lary was protected by means of a short piece (4 cm) of stainless
steel capillary (o.d. 1/8, i.d. 1.6 mm) welded into the introduction
hole of the furnace.
The arrangement of the flame furnace on the burner head of the
spectrometer, the geometry (10/7 mm o.d./i.d.) and the material
(super alloy, ASTM B 167) of the furnace was the same as in Refs.
[6, 7]. There was a hole of 3.2 mm centered in the flame furnace
for the introduction of the stainless steel protection tube and four
additional holes of 3 mm at the bottom side for ensuring the flame
gases enter the furnace. Small metal plates were screwed to both
ends of the burner head, in which two ceramic pins of 3 cm length
are fixed. The 10 cm flame-heated tube length was laid on these
pins and could be put into/out of the flame easily.
Thermospray aerosol generation
In all papers about TS sample introduction a electrically heated
capillary is used [8, 9], while in the TS-FF system the capillary is
heated only at one end by the acetylene/air flame over a length of
a few millimeters [7]. In our present work capillaries made from
synthetic fused silica with 25200 m inner diameter were used.
The flow in a capillary can be described by the Hagen-Poiseuille
law. In Fig. 2 the calculated effect of the inner diameter of the cap-
illary on the pressure to be applied and the flow rate of water at
20 C (=0.01 Poise) are shown. For these calculations a liquid
plug of 10 cm length in the capillaries was chosen. A liquid plug of
this length is equal to a volume of 0.19 L, 0.44 L, 0.78 L and
3.14 L in capillaries of 50 m, 75 m, 100 m and 200 m inner
diameter, respectively. According to this diagram, a pressure of
0.05 to 1 MPa and a capillary of 100 m inner diameter result in a
flow rate between 0.08 and 1.6 mL/min, which is a useful range for
TS-FF-AAS. Figure 3 shows the effect of various inner diameters
of the capillary on the signal height. The highest signals were ob-
138
Fig. 2 Influence of the flow rate and of the inner diameter of the
capillary on the pressure (calculated values according to Hagen-
Poiseuille law, the considered length of the liquid segment
amounts to 10 cm)
Fig. 3 Influence of various inner diameters of the capillary on the
signal height (1 MPa pressure was used for the introduction of the
1 L sample of 5 g/mL Pb into the flame furnace)
tained with capillaries of 75 and 100 m inner diameter. The sig-
nal obtained with a 200 m capillary was smaller because the TS
aerosol formation was not effective. Due to the very high flow rate
applying a pressure of 1 MPa, a 1 L sample (segment length:
3.2 cm) will be forced in an extremely short time (<1 ms) through
the hot end of the capillary. This time is probably too short for a
suitable heat transfer to the liquid resulting only in partial evapora-
tion. Reducing the inner diameter of the capillary to below 75 m
resulted in a drastic decreasing of the signals. The reason for this
could be that in this case the vapor generation is very effective
(longer throughput time of the sample through the heated part of
capillary, higher heat transfer) and, therefore, the vapor cannot
emerge immediately from the outlet of the capillary and creates
back-pressure against the flowing liquid, thus reducing the flow
rate of the liquid. In case of the capillary of only 25 m i.d., no sig-
nal was obtained probably due to the very small flow rate and the
high back-pressure. However, signals could be obtained with this
capillary if a higher pressure (2.02.5 MPa) was used to transport
the sample. In all the following experiments a capillary of 100 m
i.d. was used.
Results
Influence of pressure and sample volume
In Fig. 4 the signals of 0.3 L samples of 10 g/mL Pb us-
ing a 100 m i.d. capillary and different pressures are
Fig. 4 Signals of 0.3 L samples of 10 g/mL Pb obtained by
means of TS capillary with 100 m i.d. applying different pres-
sures
Fig. 5 Peak areas and peak heights obtained for different volumes
(0.3, 1, 10, 20 and 50 L) of samples (5 g/mL Pb, 100 m capil-
lary, 1 MPa)
shown. While the peak width changed with increasing
pressure (and thus the flow rate), the peak height remained
almost constant. This phenomenon was also observed in
our earlier work [6, 7]. The vaporization rate becomes
proportionally higher if the furnace is fed with a higher
flow rate. However, even if all the liquid is vaporized, the
average amount of sample inside the absorption volume
(flame furnace tube) cannot increase due to the constant
environment pressure (atmospheric pressure) around the
furnace of 0.1 MPa.
The peak areas and peak heights obtained for various
volumes (0.3, 1, 10, 20 and 50 L) of samples are shown
in Fig. 5. As it is typical in flow injection techniques, the
values of the signal area increase with the increasing sam-
ple volumes whereas the signal height reaches a plateau
(steady-state signal).
Comparison of TS-FF-AAS and conventional FAAS
After comparing the same small amount of sample solu-
tion (e.g. 10 L, 5 g/mL Pb) with the TS-FF-AAS and

Fig. 6 Comparison of the signal of a 0.3 L sample (20 g/ml Pb)
obtained with TS-FF-AAS and the signals of 100 L and 1.5 mL
(continuous sample introduction) sample volume determined with
conventional FAAS
conventional FAAS the superiority of TS-FF-AAS was
obvious. While a signal height of 1.05 absorbance units
could be obtained with TS-FF-AAS, the signal gained
with the conventional flame technique was hardly detect-
able. In order to have a more fair comparison, the mea-
surements with conventional flame AAS were done with
100 L and 1.5 mL sample volume, while for TS-FF-AAS
a volume of only 0.3 L was used. A 20 g/L Pb concen-
tration was chosen because it led to an easily measurable
signal using conventional flame AAS (Fig. 6). While the
maximal signal height in conventional flame AAS (1.5 mL
sample volume) amounted to 0.6 absorbance units,
TS-FF-AAS with only a sample volume of 0.3 L reached
a value of 1.7 (already in the curved part of the calibration
curve). Although the sample volumes analyzed were
23 orders of magnitude larger in case of standard FAAS,
the signals were still smaller than those obtained with the
TS-FF technique.
Analytical figures of merit
Table 1 contains the limits of detection (3 s values) ob-
tained from 25 measurements of a blank solution. The
LOD values for TS-FF-AAS were determined for three
small sample volumes (0.3, 1 and 10 L) while in con-
Table 1 Comparison of detec-
tion limits (N=25, 3 s values, Method Sample
peak heights) of Pb determined volume
by TS-FF-AAS (0.3 L, 1 L (L)
and 10 L sample volume; gas
pressure: 1 MPa; capillary:
100 m i.d.) and conventional
TS-FF-AAS 0.3
flame AAS (100 L samples,
discrete sampling technique) 1 52
a 10
LOD values obtained by
TS-FF-AAS related to that of FAAS 100
conventional FAAS
139
ventional flame AAS a larger volume was used (100 L,
discrete sampling technique [10]). The TS-FF determina-
tions were carried out using a capillary with 100 m
i.d., the applied pressure was 1 MPa. The concentration
values of LOD could be improved slightly by increasing
the sample volume. The improvement factors between
TS-FF-AAS and conventional FAAS amounted to 6.510.4,
however, the sample volume consumed was much larger
in case of FAAS. The improvement factors were remark-
ably higher if the absolute amounts were compared, be-
cause a 100 L sample volume was necessary for standard
FAAS measurements to obtain acceptable results. These
factors ranged from 105 to 2,250 using sample volumes
between 10 L and 0.3 L for TS-FF-AAS.
In Fig. 7 signals obtained for standard solutions of
0.510 g/mL Pb are shown. The applied sample volume
was 0.3 L. All standard solutions were injected three
times (calibration curve). The relative standard deviations
(N=12) for 0.3 L sample volume for 5 g/mL Pb amounted
to 3.1% and 3.8% in signal height and signal area, respec-
tively. Slightly better RSD% values were obtained for
larger sample volumes (2.9% and 3.2% for 1 and 10 L,
respectively). Figure 8 shows the signals of a number of
1 L samples containing 5 g/mL Pb as an example of the
reproducibility and the time consumption of the measure-
ments (approximately one sample/min). In Fig. 7 and
Fig. 8 small post-peaks can be seen following the analyti-
cal signal when the valve (in the inject position) was
changed to the load position to fill the sample loop
again with the next sample. These peaks disappeared after
a second introduction of a blank solution. A possible rea-
son for this memory effect could be very small amounts of
sample remaining in some parts of the HPLC sample in-
troduction valve. Comparing the area of the post peaks
with the area of the analytical signals (Fig. 7 and Fig. 8),
the volume responsible for these peaks lay in the region of
a few nanoliters. From analytical point of view these
peaks were negligible.
Discussion
The described micro TS-FF method is mainly advanta-
geous for the analysis of sample volumes smaller than
10 L. After the optimization of the operational param-
eters (capillary of 100 m i.d., 1 MPa pressure), samples
Detection limits
Concen- Improvement Absolute Improvement
tration of power amount of power
(ng/mL) of detectiona (pg) of detectiona
69 6.5 20 2,250
8.6 52 865
43 10.4 430 105
450 45,000
140
Fig. 7 Signals obtained for standard solutions of 0.510 g/mL Pb
(sample volume: 0.3 L, 100 m capillary, 1 MPa)
Fig. 8 Signals of repeated determinations of 1 L samples con-
taining 5 g/mL Pb (100 m capillary, 1 MPa)
of 0.3 L could be analyzed with a higher power of de-
tection than using conventional FAAS with sample vol-
umes 3 orders of magnitude larger. In case of submicro-
liter sample volumes, the velocity of the liquid segment is
very high (short length of liquid segment), therefore, the
liquid flow may be evaporated in the capillary only to a
very small extent. Due to the high speed collision of the
liquid jet onto the hot inner surface of the furnace, a com-
bined process of thermospray effects and jet impact va-
porization (JIV) [6] could be the real mechanism of the
aerosol generation.
The flow rate of the liquid sample plug in the TS cap-
illary is smaller than the values calculated using the Ha-
gen-Poiseuille law because of the back-pressure of the va-
por formed at the front end of the liquid reaching the end
of the hot capillary. The back-pressure becomes higher
when capillaries with smaller inner diameters are used.
The formation of the vapor generated over the last part of
the capillary and its back-pressure could be studied using
a standard Bunsen-burner with a flame of a propane-bu-
tane gas mixture. When the end of a capillary of 100 m
i.d. was pushed into the flame, fine vapor exiting at high
velocity could be observed (the flame was colored by the
use of sodium chloride solution). Using capillaries with
inner diameters smaller than 75 m and pressures lower
than about 0.5 MPa the flow rate of the liquid (qualita-
tively indicated by the intensity of the yellow color of the
flame) periodically changed, resulting in a poorer S/N ra-
tio. When a capillary with a 25 m i.d and 0.30.4 MPa
pressure was studied the direction of the flow changed,
bubbles moved backward toward the sample entrance of
the capillary, which could be observed through the trans-
parent wall of the fused silica capillary. This pulsation ef-
fect could be considerably reduced using a relatively high
pressure (1 MPa), capillaries of larger inner diameter
(75100 m) and only a very small sample volume (0.3
1 L). Working with a gas pressure of more than 2 MPa
and capillaries of 25 m i.d. satisfying results could also
be obtained, whereby the back-pressure of the vapor gen-
eration is fully compensated for by the higher pressure of
the transport gas.
As was already stated, FF-AAS provides better power
of detection compared to that of conventional FAAS for
17 elements [6,7]. In this work Pb was used as an exam-
ple for the determination of elemental traces by flame
AAS in very small sample volumes. This paper centers on
the fundamentals of this new technique. Therefore, the
main area of the investigations was the study on optimiza-
tion of parameters. In order to confirm the assumption,
that this micro TS-FF-AAS can be applied to the whole
group of TS-FF-AAS determinable elements, some provi-
sional experiments were also carried out with other ele-
ments (Cd, Hg, Tl, Zn). Very similar results and effects
were found, but more detailed studies are needed.
Acknowledgments Our work was supported by the Education
Ministry of Hungary (FKFP 23/2001), the Pro Regione Foundation
(Hungary) and the Bundesministerium fr Bildung, Wissenschaft,
Forschung und Technologie (Germany) and the Ministerium fr
Schule und Weiterbildung, Forschung und Technologie des Landes
Nordrhein-Westfalen (North Rhine Westphalia, Germany).
References
1. Welz B, Sperling M (1999) Atomic absorption spectrometry,
3rd edn. Wiley-VCH, Weinheim, p 20
2. Fuwa K, Vallee BL (1963) Anal Chem 35:942
3. Kahn HL, Peterson GE, Schallis JE (1968) At Absorpt Newsl
7:35
4. Delves HT (1970) Analyst 95:431
5. Matusiewicz H (1997) Spectrochim Acta Part B 52:1711
6. Gspr A, Berndt H (2000) Anal Chem 72:240
7. Gspr A, Berndt H (2000) Spectrochim Acta Part B 55:587
8. Koropchak JA, Veber M (1992) Crit Rev Anal Chem 23:113
9. Conver TS, Yang J, Koropchak JA (1997) Spectrochim Acta
Part B 52:1087
10. Berndt H, Jackwerth E (1975) Spectrochim Acta Part B 30:169
Anal Bioanal Chem (2002) 372 : 141147
DOI 10.1007/s00216-001-1196-4
O R I G I N A L PA P E R
Oliver Birkert Gnter Gauglitz
Development of an assay for label-free high-throughput screening
of thrombin inhibitors by use of reflectometric interference spectroscopy
Received: 28 August 2001 / Revised: 4 October 2001 / Accepted: 24 October 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract We describe the development of, and analytical
conditions used for, parallel affinity assay for thrombin
inhibitors adapted to the first label-free optical screening
HTS detection set-up fully integrable into a screening
platform. To achieve compatibility with pharmaceutical
libraries, an HTS-transducer was realized by gluing the
bottomless scaffolds of 96- and 384-well plastic micro-
plates on to transducer slides. The transducer are coated
with a dextran, to ensure biocompatibility and functional-
ity, and a known thrombin inhibitor was attached cova-
lently to it. By adapting reflectometric interference spec-
troscopy for simultaneous reading of the whole transducer
plate we were able to detect the binding of thrombin in all
the wells of the microplates on-line, in parallel, and time
resolved. By using an inhibition assay, the screening of
384 substances for thrombin activity can be performed
within an assay time of less than 15 min. We also show
that the data quality is high enough for parallel quantifica-
tion of the IC 50 values of the library substances.
Keywords High throughput screening Label-free
detection Thrombin Reflectometric interference
spectroscopy Assay development
Introduction
In recent years pharmaceutical research has been searched
for new ways to discover future drugs and agrochemical
products and has developed high-throughput screening
(HTS). The objective of this field of research is to screen
enormous numbers of compounds for activity against a
target of interest without the need for knowledge about
the structure of the target or the mechanism of the action
O. Birkert () G. Gauglitz
Institute of Physical Chemistry, University of Tbingen,
Auf der Morgenstelle 8, 72076 Tbingen, Germany
e-mail: oliver.birkert@ipc.uni-tuebingen.de
of these substances. The substances are pooled in sub-
stance-libraries which often contain 10 to 300 thousand,
mostly low-molecular-weight, compounds of high struc-
tural diversity [1].
One of the targets is the intensively investigated en-
zyme thrombin. Thrombin is a serine protease with a mol-
ecular weight of 34 kDa ( 40 kDa in cows) which consists
of two subunits. It catalyses the splitting of fibrinogen to
active fibrin and is the main component in the last step of
the blood-coagulation cascade. Exact regulation of the
blood coagulation is of special importance and determines
the balance between bleeding and thromboses. This
makes thrombin an important target in pharmaceutical
drug screening and much effort is being devoted to find-
ing effective thrombin inhibitors.
Whereas nowadays the number of interesting targets is
numbered in hundreds only, the human genome project
will lead consistently to the discovery of new targets
which, although not attributed to a discrete function, are
of great importance for the pharmaceutical industry. The
main targets actually tested are enzymes in the classes ki-
nases [2] or proteases [3] and membrane-bound or nucleo-
plasmatic receptors [4]. Screening of these targets is real-
ized either by functional assays or by binding assays.
Functional assays correlate the function of the target di-
rectly towards the concentration of the test substance.
This function can be as complex as the modulation of
gene expression or as simple as the activity of an enzyme.
Binding or affinity assays determine the biomolecular in-
teraction of two substances. These assays are less com-
plex and easier to perform but the target function cannot
be predicted. On the other side, binding assays supply ad-
ditional information if a competitive assay is used. For
binding assays the availability of very pure targets, e.g.
chromatographically purified, is necessary.
Among the physical methods used for HTS experi-
ments, imaging and scanning optical detection systems
are particularly widely used. The most important strate-
gies are fluorescence polarisation spectroscopy [5, 6],
fluorescence correlation spectroscopy [7], and fluores-
cence resonant energy transfer [8].
142

Although labelling of the binding partners is not neces-

sary and the binding data provide supplementary informa-
tion about affinity constants and kinetic data [9], highly
parallel label-free optical detection systems are still not
commercially available. Intensive work was undertaken to
adapt surface plasmon resonance [10] or reflectometric
interference spectroscopy [11] to the demands of HTS on
stability, automation, and parallelisation.
The RIfS screening arrangement presented is the first
label-free optical screening HTS detection system fully
integrable in a screening laboratory. It was used to de-
velop a binding assay to screen for compounds which re-
duce or inhibit thrombin activity. Parallel binding studies
were performed in special RIfS microplates with the stan-
dard 96- or 384-well format. These plates are coated with
a thin SiO2 layer which is required for signal detection
and on to which a dextran polymer layer was attached co-
valently. This amino functionalised dextran layer provides
biocompatibility [12]. In addition, the shield effect of this
polymer layer ensures that non-specific binding in the mi-
croplates was suppressed maximally. A thrombin inhibitor
similar to the inhibitor thrombstop [13] was covalently at-
tached to the dextran layer and enabled detection of inter-
action of thrombin at the surface for screening substance
libraries by means of a binding inhibition test for throm-
bin activity.

Experimental
Equipment and reagents

Thrombin and common chemicals and biochemicals were pur-

chased from Sigma (Deisenhofen, Germany) and Fluka (Neu-Ulm,
Germany). A thrombstop-analogous thrombin inhibitor with a free,
reactive amino group was supplied by Dr Peter Eckard, Knoll AG
(Ludwigshafen, Germany).
Partially amino-functionalized dextran with a average molecu-
lar mass of 260 kD was obtained from dextran by diol oxidation
and reductive amination as described elsewhere [14]. By this
method, approximately 10% of the anhydroglucose units were
functionalised.
Interference transducers were produced by Schott (Mainz, Ger-
many) by means of an ion-plating process.
The arrangement for optical detection and for gluing of the
RIfS microplates was performed by Carl Zeiss Jena GmbH, (Jena,
Germany). The parallel sample handling system for 96- and 384-
well plates was realized by CyBio (Jena, Germany).
Manual liquid handling for surface modification in the mi-
croplate wells was achieved by use of an eight-channel piston
stroke pipette (25200 L) from Eppendorf (Hamburg, Germany).
Fig. 1 Exploded schematic diagram of the detection unit. From
the white light source (a) six wavelengths are separated (b) se-
quentially and illuminate the RIfS microplate from underneath.
The light passes through the bulk carrier glass slide (c) and is re-
flected partially at the bottom and the top of the thin interference
layer (d, e) consisting of a 330 nm silicium dioxide layer and a
functionalized dextran layer. The reflected light (h) beam results in
a typical interference pattern, depending of the wavelengths of the
light. The intensity of the reflected beams is recorded by the CCD
camera (i) and enables the calculation of the optical thickness of
the interference layer. The plastic grid (f) has the dimensions of a
standard 96-well microplate. Liquid handling is realized by use of
a 96-pipette robot (g)
a nitrogen stream, they were glued to 96- or 384-well polystyrene
microplate grids.

Preparation of AMD RIfS microplates Immobilization of thrombin inhibitor

AMD was covalently attached to glass transducers by use of a
modification of the procedure of Piehler et al. [14]. The interfer-
ence transducers (73 mm108 mm) were cleaned in freshly pre-
pared, hot H2SO4H2O2, 6:4, rinsed with water and dried in a ni-
trogen stream. The activated transducers were silanised with glyci-
doxypropyltrimethoxysilane (400 L) for 1 h. AMD (2.5 g) was
dissolved in water (5 mL) and the solution was dropped on the pre-
treated transducers to furnish a uniform coating. Two such trans-
ducers were arranged in a sandwich. After a reaction time of 14 h
in a saturated-water atmosphere the sandwich was separated and
both transducers were rinsed with water. After drying the plates in
The wells of the microplates were filled with solid glutaric acid an-
hydride and the maximum amount of water was also added. The
plates are covered and shaken gently for 150 min. The wells were
rinsed thoroughly with water. A freshly prepared aqueous solution
of EDC (9.6 mg mL1, 0.05 mol L1) and NHS (5.9 mg mL1,
0.05 mol L1) was pipetted into the microplate wells. The plates
were covered and shaken gently for 20 min. The plates were
washed with water and immediately filled with 70 L (96-well
plate) or 30 L (384-well plate) of an aqueous inhibitor solution
(1.5 mg mL1; 3.8 mmol L1). The covered plates were shaken for
14 h, washed with water and dried in a nitrogen stream. The mod-

ified plates were stored at room temperature. The plates were sta-
ble for months.
Detection method
Parallel reflectometric interference spectroscopy (RIfS) enables
the optical on-line detection of specific biomolecular interaction in
special 96- and 384-well RIfS microplates without the need for la-
belling techniques. A schematic diagram of the detection and
pipetting unit is shown in Fig. 1. Basic information about single-
channel RIfS is available elsewhere [15, 16].
The parallel RIfS is based on the spectral distribution of re-
flectance from thin transparent layers on a glass substrate. The
beams reflected at both sides of an interference layer (330 nm
thickness) coated with a biopolymer adlayer lead to a interference
pattern from which the optical thickness, nd, of this thin layer may
be calculated. Changes in the optical thickness, nd, of the layer
caused either by changes in the reflective index or in the physical
thickness of the adlayer, e.g. because of specific binding to this
surface, leads to a shift of the interference pattern which is evalu-
ated in real-time.
The new parallel screening system was realized by a simulta-
neous illumination of the whole RIfS microplate glass bottom with
a halogen-light source. As a certain spectral resolution of the re-
flected light is necessary for signal evaluation, a monochromator
system was placed directly behind the light source. The mono-
chromator consists of a filter wheel which enables the subsequent
passage of monochromatic light of seven different wavelengths
(516.0, 543.5, 572.3, 606.8, 643.3, 685.9, and 707.0 nm). A total
turn of the filter wheel takes 18 s and determines the time resolu-
tion of the measurement. Diffusion and binding of protein to the
transducer surface normally occurs much more slowly. If the num-
ber of filters can be reduced to three; this leads to better time reso-
lution and increased signal noise.
The signal was recorded by means of a 12-bit CCD camera,
EEV type CCD0520, which traps the reflected light with the in-
terference information. The beam illuminates half of the CCD
chip. Simultaneously a reference beam was detected with the other
half of the chip. This reference beam enables correction for light
inhomogeneities caused by temporal intensity fluctuations of the
light source and by spatial differences in illumination intensity.
An automatic well mask generation program assigns the pixels
on the CCD chip to the wells of the microplate and accumulates
the signals corresponding to one well. Because the well mask gen-
eration tool is very flexible, masks can be created for standard 96-,
384-, and 1536-well microplates and for individual plate formats.
From these data the software calculates the average change in
optical thickness for each well and each measurement point in real
time. The data are saved in common file types. Signals from the
larger wells are calculated from a larger number of CCD camera
pixels; this results in an reduced noise. The root-mean-square
noise (RMS) is proportional to N0.5, where N is the number of il-
luminated CCD camera pixels.
Liquid handling was achieved by means of a modified pipet-
ting robot IGEL 250/96 from CyBio (Jena/Germany). Samples can
be transferred between four standard microplates and one RIfS mi-
croplate. The robot was modified in such a way that liquid han-
dling was possible without lifting the RIfS microplate from the de-
tection position. This enables continuous data recording totally in-
dependent of the pipetting steps.
Table 1 Data for drift and
RMS noise Plate type
96-Well aminodextran
96-Well without aminodextran
143
Changes in the refractive index
To examine the effect of changes in the refractive index, water and
glycerine solution (10% in water) were alternately measured in a
96-well microplate. Liquid handling was performed by means of
an automated pipettor control to ensure the plate was repeatedly
filled with both liquids for 600 s alternately. In this and all further
experiments data accumulated when the plates contained no liquid
were not used for evaluation.
Protein interaction
Protein adsorption was measured with the same pipetting program
for 96- and 384-well microplates, with the exception of the sample
volume. In the following, sample volumes are given for measure-
ments in 96-well plates, volumes for 384-well plates are given in
parentheses.
Before starting a measurement the plate was filled with PBS
for at least 10 min to prevent swelling effects within the data-
recording time. The plates were depleted and data recording and
the pipetting program were started simultaneously. First the RIfS
microplate was filled with 70 L (60 L) PBS and the baseline
was recorded for 300 s. The PBS was then removed and 70 L
(40 L) protein sample was pipetted from one of the storage mi-
croplates into the RIfS microplate. After a binding time of at least
900 s, the sample was replaced by 80 L (60 L) hydrochloric acid
to regenerate the plate. This enables re-use of the same plate for
more than 100 screening runs.
Screening experiments and evaluation of IC 50 value
For the binding inhibition test 96 samples of 100 mL PBS buffer
solution containing thrombin (6 units mL1) and chicken ovalbu-
min (500 g mL1) were pre-incubated with the inhibitors in a
standard polystyrene microplate. Ovalbumin was added in large
excess to prevent adsorption of thrombin within the pipettes and
microplates. These pre-incubated mixtures were measured with the
RIfS arrangement as described above.
As the binding event is controlled by diffusion of antibodies to
the surface, the calibrated signal is proportional to the concentra-
tion of free antibody in the solution at each time. After the binding
event the sensor was regenerated by pipetting hydrochloric acid
(pH 1.7) into the wells to prepare it for the next cycle.
For Screening experiments high, standardized inhibitor con-
centrations (60 ng mL1, 1.5107 mol L1) were used beside sam-
ples without added inhibitor.
For the determination of the IC 50 value, inhibitor concentra-
tion was varied between 4109 mol L1 and 1107 mol L1. Eval-
uation occurs by plotting the calibrated signal after a binding time
of 13 min against the inhibitor concentration.
Results and discussion
Signal drift and noise
As this is a new arrangement for parallel screening, fun-
damental data analysis was performed to specify the sig-
nal noise and the temporal signal drift. Both examinations
were conduct using plates with and without aminodextran
coating. Measurements are performed with dry plates and
with plates filled with water. Aminodextran is a hydrogel
RMS in air RMS in liquid Drift in air Drift in liquid
(pm) (pm) (nm h1) (nm h1)
1.620.003 19.35.1 0.020.02 0.400.26
1.410.003 23.60.9 0.040.02 0.290.52
144
Fig. 2 Signals arising from: (a) changes in refractive index, (b) non-
specific adsorption of calf serum, and (c) specific binding of
thrombin, with subsequent surface regeneration. One of 96 regis-
tered curves is given as an example
with the property of swelling in aqueous solution. To ob-
tain results unaffected by the swelling processes the plates
were equilibrated with water for at least 20 min before the
start of liquid measurements. For both types of plate re-
Fig. 3 Signal arising from changes in
refractive index (left) and from non-
specific adsorption of calf serum
(right)
sults are shown in Table 1. The quality of the data was not
affected by the biopolymer adlayer in air or in an aqueous
environment. Both drift and root-mean-square noise
(RMS) were at such a low level that they do not limit de-
tection of biomolecule adsorption.
Effect of changes in the refractive index
on sensor signals and signal calibration
The binding signal depends on the physical thickness, d,
and on the changes in the refractive index, n. Measure-
ments of liquids with different refractive indices were
used to characterize the sensor signals. During data regis-
tration the liquid in the RIfS microplate was changed be-
tween glycerine solution (10% in water; n=1.345) and wa-
ter (n=1.333). A signal curve for one well of a 96-well mi-
croplate is given in Fig. 2a The signal for the glycerine so-
lution is 6.3 nm higher than that for pure water. Data from
repeated measurements were highly reproducible. This
enables the direct comparison of sequentially recorded
plate screens in HTS experiments. Signal differences aris-
ing from the change in refractive index are given in Fig. 3
for the complete 96-well plate. Error bars are plotted as
above the columns. The errors result mainly from the
pipetting process. From this plot it is also apparent that
the responses differ from well to well. To standardise the
signals calibration was undertaken for each plate and all
further measurements were normalized relative to this cal-
ibration measurement. The calibrated signals for the mea-
surements in Fig. 3a are shown in Fig. 4a. The same cali-
bration was performed during the protein absorption mea-
surements.
Specific and non-specific binding of protein
Non-specific binding was examined with calve serum,
bovine albumin, and chicken ovalbumin. Because of the
shield effect of the aminodextran coating, non-specific
binding of bovine albumin and ovalbumin could be suc-
cessfully prevented. Even at high protein concentrations
of 1 mg mL1 non-specific binding signal vanishes in the
baseline noise. Only with calve serum at very high con-
 145

Fig. 4 Normalized signals for glycerine solution (a) and for the
specific binding of thrombin in a 96-well microplate (b) and in a
384-well plate (c)
centrations (1:10 diluted in water) non-specific binding
was detected. One binding curve for a 96-well plate is
shown in Fig. 2b. The high protein concentration results a
rapid coverage of the surface by protein absorption. Such
non-specific interaction often is characterized by irre-
versible binding. We found, in fact, that the baseline sig-
nal in Fig. 2b could not be reached again. This showed
that the non-specific interaction of calve serum could not
be regenerated under acidic conditions under which spe-
cific thrombin binding could be regenerated.
For the specific binding of thrombin, the plates were
covalently modified with a thrombin inhibitor. Thrombin
concentration was 6 units mL1. A single binding curve
for a 96-well plate is shown in Fig. 2c. The binding of
thrombin could be detected with sufficient time resolution
and the surface was completely regenerated with hydro-
chloric acid. The sensor surface was not affected by the
regeneration solution and stayed stable for more than
80 measurement cycles. Because the detected signals rep-
resent changes in the optical thickness, they are indepen-
dent of the size of the microplate wells and signals from
96-well plates can be compared with those from 384-well
plates without further conversion. Calibrated binding
curves for both types of plate are shown in Figs 4b and 4c.
Each well was filled with a thrombin solution of 6 units

Fig. 5 Screening signals on

96- and 384-well plates. Left,
non-calibrated binding curves
of screening runs for thrombin
inhibitors; right, calibrated
screening data; black, samples
containing inhibitor. For rea-
sons of clarity, only 96 binding
curves (systematically distrib-
uted on the plate) are plotted
for the 384-well microplate
146

Table 2 Screening results ob-

tained with the thrombin Re-found Re-found False False
thrombin inhibitor system with activity inactivity negative positive
different decision criteria 96-Well plate 60% criteria 9 of 9 85 of 87 0 2
96-Well plate 50% criteria 9 of 9 85 of 87 0 2
384-Well plate 60% criteria 87 of 88 288 of 296 1 10
384-Well plate 50% criteria 86 of 88 291 of 296 2 6

mL1. With few exceptions the signals are very homoge- binding signal or the calibrated binding signal, as is ap-
neous for both types of plate and enable well to well com- parent from Fig. 5. Depending on the criteria for hit iden-
parison. The system is, therefore, suitable for binding-in- tification 86 (50% criteria) or 87 (86% criteria) of the
hibition high-throughput screening assays. 88 inhibitor samples were identified on the 384-well
The better results from the 96 plates are a consequence plate. As mentioned above, false negative hits are most
of the lower signal noise, because of the greater well area. problematic in drug screening. The higher number of false
Signal derivation results mainly from the manual surface negative hits should, therefore, be accepted and the more
modification chemistry and might be prevented by an au- generous 60% criterion should be used so that as few hits
tomated preparation process; this is recommended when as possible are missed.
many plates are needed for screening of whole substance In addition to the experiments with the thrombstop-
libraries. analogous thrombin inhibitors, three other inhibitors of
unknown structure were examined; the results confirmed
the high assay stability. Screening was also performed
Screening experiments with 5% DMSO added to the samples. No influence of
DMSO was detected. Because library substances are nor-
To show the suitability of the parallel RIfS system throm- mally stored in DMSO, stability of the assay against
bin binding was examined for an unknown number of DMSO is essential even when a stabilisation of the assay
samples spiked with thrombin inhibitors. The results of compounds is not necessary.
this screening are shown in Fig. 5. For the 384-well plate
signals for every fourth microplate position are shown.
Samples without inhibitor give binding signals of more Evaluation of IC 50 value
than 0.3 nm whereas samples with inhibitor added re-
In addition to the determination of hit compounds, quanti-
sulted in signals <0.2 nm. For most of the positions, the
tative results such as IC50 value or binding constants of
inhibited samples could be re-found even without a cali-
drug candidates are of interest. The stability and reliabil-
bration step. The few curves which resulted in signals
ity of binding signals with the parallel RIfS-screening sys-
which could not be attributed with certainty could be as-
tem enable such examination. A binding inhibition test
signed after performing the standard calibration presented
was performed to determine the IC50 value of the thromb-
above. These results are shown in Fig. 5 as bar chart. The
stop-analogous thrombin inhibitor (Fig. 6). In this assay
samples with additional thrombin inhibitor are plotted in
black. For samples without inhibitors binding signals of
100% were expected whereas inhibited thrombin samples
should result in obviously distinguishable smaller signals.
For both plate types, nearly all of the pure thrombin sam-
ples gave binding signals of 80120% whereas the sam-
ples incubated with inhibitors mostly gave signals of
<40%. Re-finding rates are given in Table 2. To identify a
hit, two criteria of 40% and 50% signal inhibition were
stipulated. Except for two false-positive predictions, all
samples are classified correctly on the 96-well plate in
particular all the inhibitor samples were identified as hits,
which means that no false negative hits were found. This
is of primary importance, because each false negative hit
leads to the loss of a potential drug candidate. False posi-
tive hits in the 96-well plates occurred in wells in which
surface damage as a result of the manual surface coating
had previously been detected. In summary, the 96-well
plate gave absolutely stable and reliable results for the in-
hibitor screening. Fig. 6 Determination of the IC 50 value of a thrombin inhibitor by
The smaller well-size of the 384-well plates resulted in simultaneous detection of 96 different samples within less than
a higher signal noise but did not affect either the absolute 30 min. Inset, the structure of the thrombstop analogue inhibitor

the IC 50 value represents the concentration at which half
of the original binding signal was obtained. Thrombin
samples (20 g mL1) were incubated with inhibitor at
concentrations of 2109 mol L1 to 2107 mol L1. Each the University of Tbingen and by the Fonds der Chemischen In-
sample concentration was detected eightfold and simulta-
neously in one 96 RIfS microplate. After data calibration,
an IC 50 value of 50 nmol L1 could be determined from
the binding inhibition curve for this thrombin concentra-
tion (Fig. 6).
Conclusion
We have reported the development of a new binding assay
to screen for thrombin inhibitors with HTS methods. We
used a new, parallel, label-free detection method based on
the RIfS technique. With label-free optical methods it is
not necessary to wait for equilibrium binding to determine
the binding characteristic [17]. This enables shorter assay
times. Although we used binding times of 900 s, notice-
able and evaluable signals are available after <250 s.
The quality of screening data for 96- and 384-well
plates are excellent and promise successful further paral-
lelisation on denser plates, e.g. 1536-well plates. The
imaging system based on a CCD camera is not restricted
to a certain plate density and enables much denser plates
with marginal changes in the evaluation software. Be-
cause the signal does not depend on the well area and the
noise can be reduced by use of a high pixel-to-well ratio
in the CCD camera, even a chip-based arrangement can
be realised. The throughput of the system presented is
high enough to fulfill the specification of HTS and even
ultra-HTS which demands screening of 100,000 samples
per day [18].
Limitations of this new approach are the low sensitiv-
ity and therefore the restriction on sample molecules with
a high molecular mass, e.g. enzymes or receptors. Small
molecules, which are of main interest in drug discovery,
can be examined with an indirect binding-inhibition test.
147
Acknowledgements Part of this work was funded by BMBF pro-
ject Librarian II/0310838. Oliver Birkert was supported by the
DFG-Graduiertenkolleg Quantitative Analyse und Charakterisie-
rung pharmazeutischer und biochemisch relevanter Substanzen at
dustrie.
References
1. Parks CA, Crippen GM, Topliss JG (1998) J Comp Aided
Molec Design 12:441449
2. Hardcastle A, Boxall C, Newbatt Y, Rowlands MG, Turlais F,
Aherne GW (1999) British J Cancer 80:47
3. Zhang R, Beyer BM, Durkin J, Ingram R, Njoroge FG, Wind-
sor WT, Malcolm BA (1999) Anal Biochem 270:268275
4. Mellentin-Michelotti J, Evangelista LT, Swartzman EE, Mi-
raglia SJ, Werner WE, Yuan P-M (1999) Anal Biochem 272:
182190
5. Seethala R, Menzel R (1999) Anal Biochem 253:210218
6. Lynch BA, Loiacono KA, Tiong CL, Adams SE (1997) Anal
Biochem 247:7782
7. Brock R, Vamosi G, Vereb G, Jovin TM (1999) Proc Nat Acad
Sci USA 96:1012310128
8. Grahn S, Kurth T, Ullmann D, Jakubke H-D (1999) Biochim
Biophys Acta 1431:329337
9. Haake H, Schtz A, Gauglitz G (2000) Fresenius J Anal Chem
366:576585
10. Myszka DG, Rich LR (2000) Pharm Sci Technol Today 9:310
317
11. Rothmund M, Schutz A, Brecht A, Gauglitz G, Berthel G, Graefe
D (1997) Fresenius J Anal Chem 359:1522
12. Piehler J, Brecht A, Hehl K, Gauglitz G (1999) Colloids Sur-
faces B 13:325336
13. Klingler J, Friedrich T (1997) Biophys J 73:21952200
14. Piehler J, Brecht A, Gauglitz G (1996) Biosens Bioelectron
11:579590
15. Gauglitz G, Brecht A, Kraus G, Nahm W (1993) Sens Actua-
tors B 11:2127
16. Schmitt H-M, Brecht A, Piehler J, Gauglitz G (1997) Biosens.
Bioelectron. 12:809816
17. Edwards DA (1999) J Appl Math 63:89112
18. Koltermann A, Kettling U, Bieschke J, Winkler T (1998) Proc
Natl Acad Sci USA 95: 14211426
Anal Bioanal Chem (2002) 372 : 148154
DOI 10.1007/s00216-001-1190-x
O R I G I N A L PA P E R
R. M. Lzaro J. J. Garca A. Pi
Insulin receptor binding to erythrocytes during normal pregnancy:
an update of the method
Received: 3 September 2001 / Revised: 19 October 2001 / Accepted: 23 October 2001 / Published online: 13 December 2001
Springer-Verlag 2001
Abstract Normal values of the parameters of insulin re-
ceptors in erythrocytes were provided to make a control
group throughout gestation, 24 h postpartum and six weeks
after delivery with the aim of comparing them with other
parameters with insulin receptor-related pathologies. Thus,
one of the purposes of this study was the update of a
method to calculate the parameters of insulin receptors in
erythrocytes, carrying out several modifications that im-
proved the assay: during binding studies incubation was
in continuous rotation shaking, and increasing maximal
insulin binding. Other modifications were the increase in
the concentration of insulin 125I to 1 ng/mL, and the max-
imal concentration of unlabeled insulin, 100 ng/mL. Eryth-
rocyte age was considered by the intracellular creatine
content providing control values and allowing the normal-
ization of the parameters of insulin receptor during gesta-
tion. Data obtained in this study indicated that changes at
receptor binding level may be also considered to explain
insulin resistance: week 28 showed maximal insulin secre-
tion (16.701.44 U/mL) whereas plasma glucose concen-
trations remained almost constant (91.142.37 mg/100 mL)
with respect to the 1st and 2nd trimester of pregnancy
(89.731.38 and 91.712.10 mg/100 mL respectively);
insulin reached the point of maximal resistance, which is
explained by a decrease of maximum specific insulin bind-
ing, %Bo (6.320.51) at minimal values due to a decrease
of high-affinity receptor number per erythrocyte, N-AA
(162) at minimal values. Moreover, the negative correla-
tion between progesterone (31.20.2 ng/mL) and Ka-AA
(r=0.71) could possibly be related to this maximal resis-
tance.
Keywords Insulin binding to erythrocytes Insulin
receptor in pregnancy Insulin resistance in pregnancy
R.M. Lzaro () J.J. Garca A. Pi have been performed longitudinally throughout gestation,
Departament of Pharmacology and Physiology,
University of Zaragoza, Domingo Miral,
s/n 50002, Zaragoza, Spain
e-mail: rosamlg@yahoo.com
Abbreviations N-AA High-affinity receptor number
per erythrocyte Ka-AA affinity constant of high-affinity
receptors %Bo maximum specific insulin binding
Introduction
Human pregnancy is accompanied by profound changes
in carbohydrate and lipid metabolism [1, 2, 3, 4]. Although
plasma insulin levels rise during pregnancy [2], glucose
tolerance is reduced [5], which suggests a loss in the re-
sponsiveness of insulin-sensitive tissues [6]. The mecha-
nism that mediates this altered sensibility would be located
at the insulin binding level to its receptors, beginning with
receptor activation through ligand binding, which facili-
tates receptor dimerization and autophosphorylation of
specific tyrosine residues in the cytoplasmic portion. The
phosphotyrosine residues either enhance receptor catalytic
activity or provide docking sites for downstream signaling
proteins [6, 7, 8, 9, 10]. Although skeletal muscle, liver
and adipose tissue are the major target tissues for studies
on insulin receptor binding, ethical reasons limit the avail-
ability of cells from these tissues, particularly for longitu-
dinal studies in pregnant women. Therefore, most studies
were carried out on erythrocytes and monocytes. Both un-
changed [11, 12, 13, 14] and increased [15] insulin bind-
ing to erythrocytes from normal pregnancies have been
reported in agreement [11, 13] or in conflict [16] with re-
sults on monocytes. Moreover, monocytes need more blood
volume for binding measurements, representing only a very
small fraction of the whole blood cells. Finally, insulin re-
ceptors in erythrocytes exhibit many of the characteristics
of other tissues but the reported results on insulin binding
to erythrocytes in pregnancy were controversial, due to the
different time points during pregnancy at which data were
collected, rendering the comparison of the various results
difficult. Furthermore, only a limited number of studies
only one or two, and they did not allow reliable identifi-
cation of the changes in insulin binding in erythrocytes
throughout pregnancy. Therefore, the course of the serial

changes in parameters related to insulin binding to eryth-
rocytes has not yet been established.
This study was designed to obtain a control group data
of the parameters of insulin binding to its receptors in eryth-
rocytes during normal pregnancy.
Material and methods
Subjects
A group of 31 non-obese, healthy pregnant women, aged 20 to 40
was studied. Consent was obtained from all volunteers. They had
normal glucose tolerance according to OSullivans criterion [17].
Blood samples were obtained during basal fasting conditions after
a 12-hour fast by venous puncture at 12th, 20th, 28th, 32nd and
36th weeks of gestation, at 24 h postpartum and six weeks after de-
livery. Responses 60 min after a 50 g oral glucose overload were
evaluated at the 12th (first trimester) and 28th (third trimester) weeks
of pregnancy to study several parameters related to the diminished
sensitivity of insulin in the third trimester of pregnancy. This study
was completed with several hormones and metabolites involved in
glucose metabolism.
Chemicals and solutions
Monoiodinated A14Tyr-[125I] insulin came from New England Nu-
clear (approximate specific activity 100 Ci/g) and pig insulin
came from Novo Nordisk (Gentofte, Denmark). Ficoll-Paque solu-
tion was obtained from Amersham Pharmacia Biotech (Madrid,
Spain) (100 mL of solution contains 5.7 g ficoll 400 and 9 g sodium
diatrizoate). Ba(OH)28H2O was purchased from Fluka Chemika
(Madrid, Spain), and ZnSO47H2O from Merck (Madrid, Spain).
Dibutylphthalate, bovine albumin, diacetyl, -naphthol, p-chloro-
mercuribenzoic acid and creatine monohydrate came from Sigma
(Madrid, Spain). All other reagents and solvents were obtained at
the highest available purity.
Plasmatic insulin was measured by an Insik-5-Sorin kit; to quan-
tify directly 17--estradiol from serum or plasma we used the Im-
muchem Estradiol-17 (E2) kit from ICN Biomedicals, Inc Diag-
nostic Division. Progesterone was measured by Farmos Diagnos-
tica Progesterone 125I radioimmuno assay kit, testosterone by Izasa
(Distrib. Tecnicas S.A) testosterone RIA kit (five vials). Cortisol
by DPC Dipesa S.A cortisol RIA kit (five vials). Glucagon by A.
Menarini Diagnosticos S.A RIA kit. Plasma glucose was quanti-
fied by a SMA-II autoanalyzer, prolactine with AIA-600 (TOSOH)
autoanalyzer and free fatty acids by Hitachi 911 autoanalyzer from
Boehringer Ingelheim.
Radio receptor assay
Preparation of erythrocytes for binding studies. Erythrocytes were
isolated according to method of Gambhir et al. [18, 19] with sev-
eral modifications. Venous blood samples (10 mL) taken in the
morning (8 a.m.) after an overnight fast were drawn into an he-
parinized tube, then immediately centrifuged (400 g for 10 min at
20 C) and the plasma was discarded. The erythrocytes were iso-
lated by the Boyum method [20]. The cell pellet was mixed with
physiological saline (9 g/L), completing a 10 mL solution, and
then layered on 4 mL of Ficoll-Paque. After centrifuging at 400 g
for 20 min at 20 C, the saline and Ficoll-Paque phases, upper lay-
ers of the erythrocyte phase, containing mainly monocytes, granu-
locytes and lymphocytes were eliminated. The pellet containing
the erythrocytes was resuspended in physiological saline (1:1), and
the above procedure was once again repeated. The resulting washed
erythrocytes were resuspended in buffer G (2:1), pH = 8 (NaOH)
to equilibrate the cells. Buffer G was prepared according to the
method of Gambhir et al. [18], 50 mmol/L Tris, 50 mmol/L Hepes,
10 mmol/L MgCl2, 3 mmol/L Na2-EDTA, 10 mmol/L D-glucose,
149
10 mmol/L CaCl2, 50 mmol/L NaCl, 5 mmol/L KCl, 1 g/L bovine
albumin. After centrifuging the cell suspension (400 g, 10 min,
4 C) the buffer was aspirated (with leukocytes) and the cell pellet
resuspended in buffer G, adjusting the erythrocyte concentration to
4.4109 cells/mL (estimated in a STKS and DAX 72 of Technicon
and Cell Coulter System 8000 of Baker Instruments). Leukocyte
contamination: 525.6SD 205.5
Insulin binding studies. Isolated cells (400 L) were incubated with
50 L 125I-labeled insulin (final concentration 1 ng/mL) and 50 L
unlabeled insulin at different concentrations (0.5, 1, 2, 5, 10, 50,
100 ng/mL), at 4 C during 20 h and continuous rotation shaking
with a slope of 45. The incubation was finished by transferring
200 L duplicate aliquots to pre-chilled Eppendorf tubes contain-
ing 250 L buffer G and 250 L dibutylphthalate, then immedi-
ately centrifuged, 5000 g at 4 C in a Sigma 2MK centrifuge for
4 min. The supernatant was aspirated in order to remove the un-
bounded 125I-labeled insulin, and radioactive disintegration of the
cell pellet were counted in a Rackgamma Wallac 1270 (LKB).
Calculations. The average number of insulin binding sites per hu-
man erythrocyte was obtained by Scatchard analysis [21] of 125I-la-
beled insulin binding data. Using the amount of the insulin bound
(B), and total insulin concentration minus B as the free insulin con-
centration (F), a plot of B/F to B can be derived. The average num-
ber of estimated sites per cell was calculated using the following
proposed by Gambhir et al. [18]. The receptor number calculation
was performed accepting only one state for the receptor, with neg-
ative cooperativity (total receptors), accepting as high affinity the
receptor number that will bind to insulin at concentrations from 0.5
to 10 ng/mL.
Hormone measurements
Plasmatic levels of insulin, C-peptide, glucagon, cortisol, 17--estra-
diol, progesterone, prolactin, and testosterone, were determined by
radioimmuno-analysis; insulin according to Wilson and Miles
[22], C-peptide according to Heding [23], glucagon according to
Soybel et al. [24], cortisol according to Kehlet and Binder, Farmer
and Pierce [25, 26], 17--estradiol according to Abraham, Jurgen,
Hammond et al., Honjo et al. [27, 28, 29, 30], progesterone ac-
cording to Hammond et al., Abraham et al., Lebel and Grose, Rat-
cliffe et al. [29, 31, 32, 33] and testosterone according to Ismail et
al., Lox et al. [34, 35]. All of them were determined using com-
mercially available RIA kits following the protocols given by the
manufacturers. Prolactin was determined using enzyme immuno-
assay with an AIA-600 autoanalyzer.
Free fatty acids were determined according to Shimizu et al.
[36], based on a spectrophotometric method.
Creatine content in erythrocytes was performed according to
Griffiths and Fitzpatrick, Li et al. [37, 38, 39], based on the di-
acetyl--naphthol reaction and using a spectrophotometric method.
Statistical analysis
Values are shown as meanstandard error (SE). For the purpose of
inter-group comparison, three different statistical tests were used:
the non-parametric Wilcoxon test, the t-test and the Kruskal-Wallis
test. The level of significance was defined as p<0.05. The linear
correlation between two variables was calculated and tested by
Pearsons.
Results
Parameters of erythrocyte insulin binding receptor
Maximum specific insulin binding to erythrocyte recep-
tors (%Bo) (Fig. 1), increased from the 12th to 20th weeks
150
Fig. 1 Maximum specific insulin binding to erythrocytes, %Bo (A),
number of high-affinity receptors per erythrocyte, saturated at
10 ng/ml of unbound insulin N-AA (B) and affinity constant of
high-affinity, Ka-AA (C) from insulin receptors of erythrocytes
during pregnancy, 24 h postpartum and six weeks after delivery;
a denotes p<0.05 vs 12th week; c vs 28th week, e vs 36th week of
gestation and f vs postpartum
of gestation (6.390.44 to 6.940.66 respectively) because
of an increase in the number of high-affinity receptors
per erythrocyte saturated at 10 ng/mL unbound insulin
(N-AA), from 172 to 192. The %Bo decreased from the
20th to the 28th week (6.940.66 to 6.320.51 respec-
tively) because of a decrease of N-AA (192 to 162). %Bo
increased progressively throughout the third trimester, 28th,
32nd and 36th weeks (6.320.51, 6.800.48 and 7.38
0.60 respectively), due to an increase in N-AA from the
28th to the 32nd weeks (162 to 203) and an increase of
the affinity constant of high-affinity receptors (Ka-AA)
from the 32nd to the 36th week of pregnancy (7.500.66
to 9.491.37108 M1 respectively).
Carbohydrate metabolism during pregnancy
Fasting plasma insulin levels (Fig. 2) increased from the
12th (11.390.64 U/mL) to the 28th week of gestation
(16.701.44 U/mL, p<0.05), slightly decreased during
the 32nd week (15.301.08 U/mL) and remained almost
constant until the 36th week (15.711.36 U/mL). Fast-
ing plasma insulin levels postpartum (6.830.66 U/mL)
Fig. 2 Plasmatic concentrations of insulin (A) glucose (B) and free
fatty acids (C) throughout gestation, 24 h postpartum and six weeks
after delivery in 31 pregnant women. Blood samples were collected
in basal conditions; a denotes p<0.05 vs 12th week; b vs 20th week,
c vs 28th week, d vs 32nd week and e vs 36th week of gestation
were lower than that of the third trimester (p<0.05) and
increased in the puerperium (10.551.15 U/mL) at val-
ues lower than the first trimester. During the first and sec-
ond trimester of gestation plasma glucose concentrations
remained unchanged (Fig. 2) (89.731.38 and 91.71
2.10 mg/100 mL respectively). Then glucose decreased con-
tinuously as the third trimester progressed; 91.142.37 mg/
100 mL in the 28th; 87.391.70 mg/100 mL in the 32nd
and 85.271.39 mg/100 mL in the 36th week of preg-
nancy (p<0.05), until postpartum, with 83.582.33 mg/
100 mL (p<0.05) and increased puerperium, 90.331.91 mg/
100 mL with normal value. Fasting plasmatic free fatty
acids levels (Fig. 2) decreased from weeks 12th to 20th
(16.222.53 to 10.380.81 mmol/L, p<0.05) and increased
progressively during the third trimester (12.000.91,
15.272.53 and 15.772.12 mmol/L in the 28th, 32nd and
36th week, respectively). Values decreased at the postpar-
tum with statistical significance (p<0.05) and puerperium
(12.93.76, 11.580.93 mmol/L). Increased insulin secre-
tion during the first and second trimester of gestation was
associated with a decrease in plasma concentration of free
fatty acids (p<0.05). Estimates of amounts of insulin dur-
ing pregnancy have included quantification of the insulin
response to a fixed oral glucose overload in the first and

Fig. 3 Plasmatic concentrations of insulin (A) and glucose (B)
throughout gestation, 24 h postpartum and six weeks after delivery
in 31 pregnant women. Blood samples were collected in basal con-
ditions and 1 h after a 50 g oral glucose overload, weeks 12 and 28;
a denotes p<0.05 vs overload in the 12th week
third trimesters (Fig. 3). Following oral glucose ingestion,
plasma insulin increased significantly (p<0.05) [40], in the
third trimester with respect to the first one, 68.157.73 U/
mL to 85.096.22 U/mL while glucose concentration did
not vary significantly, from 135.16.08 to 144.26.17 mg/
100mL (p=0.29). Fasting plasma glucagon levels (Fig. 4)
increased from the 12th to the 20th week of gestation
(151.221.8 to 196.127.5 pg/mL) with statistical signifi-
cance (p<0.05); it then decreased until week 28th (166.0
20.0 pg/mL) and increased progressively until it reached
the 36th week (172.624.8 and 20522.46 pg/mL to the
32nd and 36th week respectively, p<0.05). This parameter
decreased in the postpartum (160.617.1 pg/mL) and in-
creased in the puerperium, reaching the highest values of
the study (272.839.9 pg/mL). In the same figure, fasting
plasma cortisol levels increased progressively from the 12th
to the 28th week of gestation (24.331.02, 33.821.54
and 41.021.45 g/mL corresponding to the 12th, 20th and
28th week respectively, p<0.05) and decreased progres-
sively during the third trimester, postpartum and puerperium
(39.41.15, 35.411.52, 29.61.72 and 20.461.04 g/mL puerperium. The values obtained provided a control group
corresponding to the 32nd, 36th week, postpartum and
puerperium respectively).
151
Fig. 4 Plasmatic concentrations of glucagon (A) and serous con-
centrations of cortisol (B) throughout gestation, 24 h postpartum
and six weeks after delivery in 31 pregnant women. Blood samples
were collected in basal conditions; a denotes p<0.05 vs 12th week;
b vs 20th week, c vs 28th week, d vs 32nd week, e vs 36th week
of gestation and f vs postpartum
Gestational hormones
Gestational hormones, 17--estradiol, progesterone, pro-
lactin and testosterone were studied with the aim of com-
pleting the results of the control group of pregnant women,
correlating them with insulin receptor parameters and then
obtain some explanation about the resistance of insulin
action at receptor level. Fasting concentrations of 17--
estradiol, prolactin and testosterone increased progres-
sively (Fig. 5) from the 12th to the 36th week of gestation,
whereas progesterone levels kept elevated and almost con-
stant throughout gestation, reaching the maximal value at
week 20.
Creatine content in erythrocytes
Creatine concentration of erythrocytes increased progres-
sively (Fig. 6) from the 12th to the 36th week of gestation,
without changes in the postpartum, and decreased in the
to correct the values of the insulin receptor parameters de-
pending on erythrocytes age, using the formula proposed
by Muggeo [41].
152
Fig. 5 Serous concentrations of 17--estradiol (A), progesterone
(B), prolactin (C) and testosterone (D), throughout gestation, 24 h
postpartum and six weeks after delivery in 31 pregnant women.
Blood samples were collected in basal conditions; a denotes p<0.05
vs 12th week; b vs 20th week, c vs 28th week, d vs 32nd week, e vs
36th week of gestation and f vs postpartum
Discussion
In the bibliography, there are a lot of cross-sectional stud-
ies on insulin binding to monocytes and erythrocytes dur-
ing human pregnancy and the results were conflicting [42,
43, 44, 45, 46, 47, 48]. The different time points during
gestation at which data were collected in those investiga-
tions render the comparison of the various results diffi-
cult. Thus, it was decided to carry out a longitudinal study
on the changes of insulin binding to erythrocytes during
gestation, postpartum and puerperium in a group of healthy
pregnant women. Results showed that during early gesta-
tion, insulin increased; being a lipogenic hormone with
normal sensitivity, it produced a milieu that favors mater-
nal fat accumulation (Fig. 2), in preparation for the rise in
energy requirements associated with the rapid growth of
the fetal-placental unit during the second half of pregnancy.
In the third trimester, although plasmatic insulin levels are
increased, glucose tolerance was reduced, thus suggesting
a loss of responsiveness of insulin-sensitive tissues. Late
Fig. 6 Creatine in erythrocytes throughout gestation, 24 h postpar-
tum and six weeks after delivery in 31 pregnant women. Blood
samples were collected in basal conditions () and 1 h after a 50 g
oral glucose overload (), weeks 12 and 28; a denotes p<0.05 vs
12th week; b vs 20th week, c vs 28th week, d vs 32nd week and
e vs 36th week of gestation
gestation is characterized also by the rise in plasma con-
centrations of several diabetogenic hormones including
cortisol, progesterone and estrogens that also favored in-
sulin resistance [6, 49, 50]. There appears to be general
agreement that the second half of pregnancy is associated
with increasing insulin resistance in the periphery [4, 12,
51. 52, 53]. It was suggested [50, 53] that endocrine or
metabolic factors and/or post-receptor events, rather than
receptor defects, accounted for the insulin resistance in
late pregnancy. The data obtained in this study indicates
that changes in receptor binding might be also considered
to explain the differences, because week 28 of gestation
showed maximal plasma insulin concentration reaching
maximal pregnancy values whereas plasma glucose con-
centrations remained almost constant showing insulin re-
sistance which is explained by a decrease in %Bo reach-
ing the minimal values of the study due to a decrease of
N-AA with minimal values, and a decrease in Ka-AA.
Moreover and following oral glucose ingestion, peripheral
insulin sensitivity was reduced at the 28th week with re-
spect to the first one (Fig. 3). In addition, the negative cor-
relation between progesterone at this point and Ka-AA
(progesterone 31.170.23 ng/mL and Ka-AA 9.271.43
108 M1, r=0.71) can contribute to increase the insulin
resistance. During the third trimester, plasma insulin con-
centrations decreased, whereas binding to the receptor in-
creased continuously and plasma glucose concentration de-
creased progressively indicating an improvement of the
insulin resistance (Fig. 1 and 2). Calculations for glucose
and the parameters of insulin receptor during these weeks
of gestation have not been found in the reviewed bibliog-
raphy.
Although erythrocytes exhibit many of the characteris-
tics of other tissues, insulin receptor numbers in the eryth-

rocytes decrease with cell aging [41, 54, 55, 56], so in this
study, creatine concentration in erythrocytes was obtained
in all blood samples, with the aim of providing a control
group of this parameter during gestation, to correct the
values of the insulin receptor parameters depending on
erythrocytes age.
A critical point in this study was the method employed
to determine insulin receptor parameters because it was
originally of Gambhir et al. [18], although several varia-
tions have been made. Buffer G was made according to
Gambhir et al. [18], but the solution with albumin was eas-
ily contaminated, so it was of interest to prepare 500 mL
every 3 or 4 days without albumin, keeping it refrigerated,
extracting 100 mL and adding the albumin at the moment
of the study. Another point that improved the method was
the manipulation of blood samples during the preparation
of erythrocytes, maintaining the temperature at 4 C, work-
ing on a tray with ice. During binding studies, incubation
differs in part because Gambhir et al. incubated their ex-
periments at 15 C for 3.5 h without shaking, whereas the
incubation of this study was in continuous rotation shak-
ing, for 20 h with a slope of 45 and in a 4 C cold room;
this way, maximal insulin binding increased, obtaining a
statistic signification of 2 (p<0.01) [57] according to the
Wilcoxon test. Another modification was the increase in
the concentration of insulin 125I to 1 ng/mL [57], i.e. 12.5
times bigger than the one employed by Gambhir et al., ob-
taining more sensitivity in the assay. A further difference
was the diminished points of unlabeled insulin (0, 0.5, 1,
2, 5, 10, 50, 100 and 105 ng/mL) as regards Gambhir et
al.s method (0, 0.05, 0.5, 1, 2, 5, 8, 10, 50, 100, 500, 103,
104 and 105 ng/mL) thus, the study needed less blood for
binding studies of pregnant women. Moreover, maximal
concentration of unlabeled insulin was 100 ng/mL because,
at this concentration, it was observed that receptors of the
cellular surface were saturated. The Scatchard graphic
values, corresponding to the largest unbound insulin con-
centrations, were placed almost in parallel with the x-axis
[58]. Moreover, during insulin binding studies, Gambhir
et al. centrifuged pre-chilled Ependorf tubes containing the
incubate, buffer G, and dibutylphthalate during 2.5 min in
a normal centrifuge but kept it in a 4 C cold room, whereas
this study improved the method, centrifuging at 4 C in a
refrigerated centrifuge and 4 min, Gambhir et al. [18].
Conclusions
Parameter values of insulin receptors in erythrocytes were
determined, in a group of normal pregnant women, carry-
ing out numerous blood samples obtained throughout
pregnancy. Thus, it was possible to find that maximal re-
sistance to insulin binding described in the third trimester
of pregnancy coincides with the 28th week of gestation
when the receptor binding was the lowest of the study. If
we do not include this 28th week contained in the third
trimester (Fig. 1 and 2) it seems that this resistance to
the action of insulin is increased throughout the third
trimester, just as the reviewed bibliography describes.
153
Moreover, at this point of the study, there were high val-
ues of progesterone and the highest values of cortisol,
both well-known diabetogenic hormones.
These modifications described with regard to Gambhir
et al.s method have improved the development of the
technique and have facilitated the analysis of such a large
number of blood samples with minimum blood taking.
Acknowledgements This work was supported by a fellowship re-
ceived by Rosa Ma Lzaro from the Government of Aragn (Ref-
erence: BCM2590) and carried out at the Hormones Laboratory of
the Hospital Clnico Universitario Lozano Blesa of Zaragoza.
References
1. Kalkhoff RK, Kim HJ (1978) Ciba Found Symp 63:29
2. Desoye G, Schweditsch MO, Pfeiffer KP, Zechner R, Kostner
GM (1987) J Clin Endocrinol Metab 64:704
3. Boden G (1996) Obstet Gyn Clin N Am 23:1
4. Homko CJ, Sivan E, Reece EA, Boden G (1999) Semin Reprod
Endocr 17:119
5. Khl C (1975) Acta Endocrinologica 79:709
6. Yamashita H, Shao J, Friedman J (2000) Clin Obstet Gynecol
43:87
7. Liu F, Roth RA (1998) Mol Cell Biochem 182:73
8. Hubbard SR (1999) Prog Biophys Mol Biol 71:343
9. Clark SF, Molero JC, James DE (2000) J Biol Chem 275:3819
10. Ueno H, Kondo E, Yamamoto-Honda R, Tobe K, Nakamoto T,
Sasaki K, Mitani K, Furusaka A, Tanaka T, Tsujimoto Y, Kado-
waki T, Hirai H (2000) Mol Biol Cell 11:735
11. Tsibris JCM, Raynor LO, Buhi WC, Buggie J, Spellacy WN
(1980) J Clin Endocr Metab 51:711
12. Ryan EA, OSullivan MJ, Skyler JS (1985) Diabetes 34:380
13. Hjollund E, Pedersen O, Espersen T, Klebe J (1986) Diabetes
35:598
14. Toyoda N (1982) Am J Obstet Gynecol 144:679
15. Dwenger A, Mitzkat HJ, Holle W, Tost P, Trautschold J (1982)
J Clin Chem Biochem 20:273
16. Andersen O, Khl C (1986) Eur J Clin Invest 16:226
17. OSullivan JB, Mahan CM, Charles D, Dandrow RV (1973)
Am J Obstet Gynecol 116:895
18. Gambhir KK, Archer JA, Bradley CJ (1978) Diabetes 27:701
19. Gambhir KK, Archer JA, Carter L (1977) Clin Chem 23:1590
20. Boyum A (1968) Scand J Clin Invest 97:77
21. Scatchard G (1949) Ann NY Acad Sci 51:660
22. Wilson MA, Miles LEM (1977) In: Abraham GE (ed) Hand-
book of radioimmunoassay, Dekker, New York, p 257
23. Heding LG (1975) Diabetologia 11:541
24. Soybel D, Jaspan J, Polonsky K, Goldberg I, Rayfield E, Tager
H (1983) J Endocrinol 56:612
25. Kehlet H, Binder C (1973) J Clin Endocr Metab 36:330
26. Farmer RW, Pierce CE (1974) Clin Chem 20:411
27. Abraham GE (1975) In: Sturgis SH, Taymor ML (eds) Progress
in gynecology, vol 1. Grune and Stratton, New York, p 111
28. Jurgen H (1975) IC EN 40:19
29. Hammond G, Vinikka L, Vihkor (1977) Clin Chem 23:1250
30. Honjo H, Otsubo K, Yasuda J (1984) Acta End 106:374
31. Abraham GE, Swerdloff RS, Tulchinsky D, Odell WD (1971)
J Clin Endocr Metab 32:619
32. Lebel M, Grose J (1978) J Steroid Biochem 9:989
33. Ratcliffe WA, Corrie JET, Dalziel AH, MacPherson JS (1982)
Clin Chem 28:1314
34. Ismail AA, Niswender GD, Midgley AR (1972) J Clin Endocr
Metab 34:177
35. Lox CD, Christian CD, Heine MW (1974) Am J Obstet Gynecol
118:114
36. Shimizu S, Tani Y, Yamada H, Tabata M, Murachi T (1980)
Anal Biochem 107:193
37. Griffiths WJ, Fitzpatrick M (1967) Brit J Haematol 13:175
154
38. Griffiths WJ (1968) Brit J Haematol 15:389
39. Li PK, Lee JT, Li CS, Deshpande G (1982) Clin Chem 28:92
40. Spellacy WN, Goetz F, Greenberg BZ, Ells J (1965) Am J
Obstet Gynecol 92:11
41. Muggeo M, Dons RF, Ryan JL, Maghetti P, Valerio A (1985)
In: De Pirro R, Lauro R (eds) Handbook on receptor research.
Acta Medica, Rome, p 146
42. Dwenger A, Mitzkat HJ, Holle W, Tost P, Trautschold J (1982)
J Clin Chem Biochem 20:273
43. Lerario AC, Wajchenberg BL, El-Andere W, Ohnuma LY,
Monaci J, Sankowsky M, Toledo e Souza IT (1985) Germek
Diabetes 34:780
44. Andersen O; Khl C (1986) Eur J Clin Invest 16:226
45. Moore P, Kolterman O, Weyant J, Olefsky JM (1987) J Clin
Endocr Metab 52:937
46. Ramachandran A, Susheela L, Mohan V, Kuzhali D, Viswana-
than M (1989) Acta Diabetol 26:123
47. Andersen O (1990) Dan Med Bull 37:92
48. Schmon B, Desoye G, Friedl H, Hofmann H, Weiss PA, Hag-
mller K (1993) Clin Chim Acta 221:33
49. Ahmed SAM, Shalayel MHF (1999) East African Medical
Journal 76:465
50. Catalano PM, Tyzbir ED, Wolfe RR, Calles J, Roman NM,
Amini SB, Sims EAH (1993) Am J Physiol 264:60
51. Buchanan TA, Metzger BE, Freinkel N, Bergman RN (1990)
Am J Obstet Gynecol 162:1008
52. Sugden MC, Holness MJ (1998) Biochem Soc T 26:79
53. Puavilai G, Drobny EC, Domont LA, Baumann G (1982) J Clin
Endocr Metab 54:247
54. Eng J, Lee L, Yalow RS (1980) Diabetes 29:164
55. Kosmakos FC, Nagulesparan M, Bennett PH (1980) J Clin
Endocr Metab 51:46
56. Dons RF, Corash LM, Gorden P (1981) J Biol Chem 256:2982
57. Rosel P, Jorge S, Navarro MA, Bonnin MR (1983) Rev Diag
Biol 32:65
58. Pi J (1987) Doctoral thesis, Zaragoza
Anal Bioanal Chem (2002) 372 : 155163
DOI 10.1007/s00216-001-1164-z
O R I G I N A L PA P E R
Dirk Schaumlffel Andreas Prange Gizelher Marx
Klaus G. Heumann Peter Brtter
Characterization and quantification of metallothionein isoforms
by capillary electrophoresisinductively coupled
plasmaisotope-dilution mass spectrometry
Received: 10 July 2001 / Revised: 1 October 2001 / Accepted: 14 October 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract In a new approach to the characterization and
quantification of metallothionein isoforms an on-line iso-
tope-dilution method in combination with the coupling of
capillary electrophoresis (CE) to an inductively coupled
plasmasector field mass spectrometer (ICPSFMS) is re-
ported. Metallothionein (MT) isoforms are separated by
CE and the elements Cu, Zn, Cd, and S are detected si-
multaneously by use of ICPSFMS in the medium resolu-
tion mode. On-line isotope dilution is performed by con-
tinuous introduction of an isotopically enriched, species-
unspecific spike solution after the separation step. MT from
rabbit liver and a further purified MT-1 isoform were quan-
tified by determination of sulfur, and the stoichiometric
compositions of the metalloprotein complexes are charac-
terized by determination of their sulfur-to-metal ratios.
Keywords Inductively coupled plasma mass
spectrometry Capillary electrophoresis Isotope dilution
Metallothionein quantification
Introduction
In the field of bio-inorganic analytical chemistry the
analysis of metal compounds in living organisms to en-
able better understanding of their biochemical behavior
has become an area of growing interest. In biological sys-
tems metals are usually coordinated to bioligands with
high complexity, e.g. proteins or peptides [1].
D. Schaumlffel A. Prange ()
GKSS Research Center,
Max-Planck-Strae, 21502 Geesthacht, Germany
e-mail: andreas.prange@gkss.de
G. Marx K.G. Heumann
Institute of Inorganic Chemistry and Analytical Chemistry,
Johannes Gutenberg-University Mainz, 55099 Mainz, Germany
P. Brtter
Hahn-Meitner-Institute,
Glienicker Strae 100, 14109 Berlin, Germany
A frequently used analytical method for the detection
of metal complexes in biological matrices is the coupling
of a high-resolution separation technique with an induc-
tively coupled plasmamass spectrometer (ICPMS) as
element-specific detector. High-performance liquid chro-
matography (HPLC) is widely used as the separation tech-
nique [2] but the application of capillary electrophoresis
(CE) to biomolecules is also attracting more interest [3].
The separation efficiency of CE is high, the sample re-
quirement small, and the mechanism of separation is dif-
ferent from that of HPLC. CE therefore complements
HPLC as a separation technique for biomolecules. The
coupling of CE to ICPMS results in additional element-
specific information about the metal complexes. Majidi
[4] recently reviewed progress in the development of
CEICPMS interfaces. The interface introduced by
Schaumlffel and Prange in 1999 [5, 6] has been com-
mercialized. The design solves the problem of detrimental
effects of suction by the nebulizer on the CE-separation
and thus the migration times of the analytes are both sta-
ble and reproducible. Work with CEICPMS can, there-
fore, now focus on optimization of the CE separations and
on application to real samples [7]. Recent publications have
dealt with the application of CEICPMS to biomolecules,
especially to the characterization of metallothioneins [8, 9].
Metallothioneins (MT) are cysteine-rich proteins of
low molecular mass (67 kD) that occur in animal and hu-
man cells such as the liver, kidney or brain. They have a
variety of physiological functions, e.g. transport, storage,
and detoxification of metals or protection from metal tox-
icity. MT are involved in the metabolism of essential met-
als or act as free-radical scavengers. Four MT isoforms
are known and several subisoforms have been discussed.
The amino acid chain includes 21 covalently bonded sul-
fur atoms: 20 are bound to the cysteine and one to the
N-terminal methionine [10]. The twenty sulfhydryl
groups of the cysteines can coordinate seven bivalent ions
such as Zn2+ or Cd2+, or twelve monovalent ions, e.g. Cu+
[11].
There is still a lack of analytical methods for the quan-
tification of MT with high precision, accuracy, and iso-
156

form specificity [10]. Conventional methods in metallo-

thionein research are based on metal-binding assays [12, Experimental
13] that use the bonding of metals such as cadmium or Instrumentation
mercury to MT, or immunoassays such as radioim-
munoassays [14] or ELISA [15], in which antibodies are An Agilent 3D CE system containing fused-silica capillaries of
used. After isolation and purification, the MT can also be length 70 cm and inner diameter (i.d.) 75 m was used. The inter-
face (CEI-100; Cetac, Omaha, USA), which has been described in
quantified by freeze-drying [16] or by amino acid analysis detail in previous publications [5, 6], enables stable and robust
[17]. Capillary electrophoresis with UV detection, with coupling between CE and ICPMS without detrimental effects of
external calibration, has already been used for the quan- nebulizer suction on the CE capillary. An inductively coupled
tification of MT-1 and MT-2 [18, 19]. A recent publica- plasmasector-field mass spectrometer (ICPSFMS) (Element;
tion by Wolf et al. [20] reported quantification of the Finnigan MAT, Bremen, Germany) was used as element-specific
detector.
metal content of MT in human liver by use of HPLC cou-
pled to an ICPatomic-emission spectrometer.
Because ICPMS is not only a multi-element but also Instrument settings
a multi-isotope technique, it enables rapid isotope-ratio
The operating conditions of the instruments are summarized in
measurements, which thus enables the stable isotope dilu- Table 1.
tion technique to be employed. Isotope dilution (ID) is
possible for elements with at least two stable isotopes.
Accurate and precise determinations can be made by mea- CE
suring the isotope ratio of the isotope-diluted sample after The sample was injected hydrodynamically at 2 kPa s (1 kPa for
spiking with a solution enriched in an isotope of the ele- 2 s) which corresponds to a sample volume of 2.2 nL. The capil-
ment to be determined [21]. lary was rinsed with a buffer solution for 5 min at 1105 Pa before
As shown by Rottmann and Heumann in 1994 [22], an each run and additionally with 0.1 mol L1 KOH 1 min every 5th
HPLCICPMS coupling system in combination with an run, to condition the capillary. To prevent buffer depletion and to
reduce background levels of copper and zinc, the buffer was
on-line isotope dilution technique can be used for the ac- changed completely every run. The MT separations were per-
curate determination of elemental species. Even unknown formed under optimized conditions at 30 kV and 15 C.
species such as metal complexes of humic substances can
be quantified by adding a species-unspecific isotopically
ICPMS
enriched spike after the separation of the species. An im-
portant precondition in this case is that the elemental The ICPSFMS was operated in the medium-resolution mode
species and the species-unspecific spike must be com- (R=3000) to enable sulfur measurement simultaneously with the
pletely equilibrated [23]. The on-line IDMS technique has
also been applied to the determination of cadmium in bio-
logical and environmental materials [24, 25]. Current Table 1 Experimental conditions of the CEICPIDMS set-up
work uses the on-line isotope dilution technique with
HPLC coupled to an ICPsector field MS for the determi- Capillary electrophoresis
nation of sulfur in separated fractions of humic substances Capillary Fused silica, 70 cm length,
75 mm i.d.
[26].
Buffer solution 20 mmol L 1 Tris, pH 7.4;
In this paper a new approach to the quantification of
Temperature 15C
metallothionein isoforms is reported. The idea is to quan-
Voltage 30 kV
tify the MT via sulfur measurements. This is possible be-
cause the number of fixed sulfur atoms per MT molecule Injection Hydrodynamic, 2 kPa s
is known to be 21. Metallothionein isoforms are separated CEICPMS interface
by CE and the elements Cu, Zn, Cd, and S are detected si- Liquid flow rate 8.0 mL min 1
multaneously by use of ICPSFMS in the medium-resolu- Buffer solution 15 mmol L 1 ammonium nitrate,
tion mode. The medium mass-resolution of 3000 enables pH 7.4;
measurement of 32S and 34S without interference from Buffer solution for the 15 mmol L 1 ammonium nitrate,
16O + and 16O18O+, respectively. An on-line isotope-dilu- isotope dilution step pH 7.4+isotopically enriched spike
2
34
tion technique in combination with CEICPMS has been Spiked isotopes S, 65Cu, 68Zn, 116Cd
developed. On-line isotope dilution is performed by con- ICPSFMS
tinuous introduction of an isotopically enriched, species- Cooling gas 15 L min1
unspecific spiked solution after the separation step. This Auxiliary gas 0.9 L min1
enables quantification of unknown MT complexes. The Nebulizer gas 1.09 L min 1
stoichiometric compositions of the metal complexes can Power 1100 W
also be characterized by determination of the sulfur-to- Mass resolution 3000
metal ratios. The analytical performance regarding detec- Scan duration 880 ms
tion limits and precision is also discussed. 32
Isotopes monitored S, 34S, 63Cu, 65Cu, 64Zn, 68Zn,
114
Cd, 116Cd.

metal determination. The isotopes were scanned in the peak-hop-
ping mode. The acquisition method was optimized to a scan dura-
tion of 880 ms as a compromise between precise measurement,
which requires a longer acquisition time, and fast scanning, which
is necessary to acquire the sharp and narrow CE signals. In this
way approximately 15 scans per electrophoretic peak could be
achieved. The dead time of the detector was corrected to 25 ns in
accordance with Refs [25] and [27], by measuring the 114Cd/116Cd
isotope ratio at Cd concentrations between 5 and 2000 g L1,
which covered the expected concentration range in this work, and
detector dead times between 0 and 100 ns. Because of the low
transmission of the ICPSFMS in the medium-resolution mode,
however, the influence of the detector dead time on the isotope ra-
tio was low.
Chemicals and materials
A Tris buffer solution (20 mmol L1) was prepared by dissolving
1.21 g Tris(hydroxymethyl) aminomethane (p.A.; Merck, Darm-
stadt, Germany) in 500 mL water and adjusting the solution to pH
7.4 with nitric acid. The buffer solution was further cleaned from
metal contamination (especially Cu and Zn) by use of Chelex 100
ion-exchange resin (Fluka, Buchs, Switzerland). A solution of
potassium hydroxide (0.1 mol L1, Suprapur, Merck) was prepared
for conditioning the CE capillaries.
Fused silica capillaries, 75 m i.d.360 m o.d. were pur-
chased from Thermo Separation Products (TSP, Egelsbach, Ger-
many) and pre-conditioned by rinsing with 1 mol L1 KOH (Supra-
pur, Merck) for 5 min before use.
An ammonium nitrate buffer solution (15 mmol L1) was used
as make-up liquid. Ammonia solution (Suprapur, Merck; 25%,
0.78 mL) was diluted with 500 mL water and adjusted to pH 7.4
with nitric acid. Nitric acid (Suprapur, Merck) was cleaned by sub-
boiling in a quartz apparatus and then diluted with water. By using
ammonia and nitric acid of this high analytical grade as educts for
the make-up liquid, the background levels of Cu and Zn could be
kept lower than 4000 counts per second. This level cannot be
achieved by using cleaned Tris buffer as make-up liquid. The
make-up liquid prepared by this procedure was used for measure-
ment with natural isotope abundances.
Stock solutions of the enriched isotopes 34S, 65Cu, 68Zn, and
116Cd were provided courtesy of the group of Professor Heumann
(Johannes Gutenberg-University Mainz, Germany). The isotopic
composition of these spike solutions was 34S=85.9%, 65Cu=99.7%,
68Zn=91.0%, and 116Cd=83.4%. The species-unspecific spike solu-
Fig. 1 Calibration of the spike
mass flow MFspike. Injection of
54.3 nL of a standard solution
containing 3.0 g Cu mL1,
1.5 g Zn mL1, 3.0 g Cd
mL1, and 25 g S mL1 to cal-
culate the mass flow MFstd
from the base peak width ac-
cording to Eq. (1). MFspike is
calculated from MFstd accord-
ing to Ref. [21]
157
tion, used for the preparation of the corresponding make-up solu-
tion, was obtained by diluting these stock solutions to final con-
centrations of 300 g L1 for S, 90 g L1 for Cu, 60 g L1 for Zn,
and 280 g L1 for Cd. 2,6-diacetylpyridine-bis(N-methylenepyri-
diniohydrazone)dichloride was then added as chelating agent to a
final concentration of 0.2 mmol L1. After waiting 10 min to en-
able the formation of metal complexes, the solution was adjusted
to pH 7.4 with an ammonia solution (Suprapur, Merck, 25%). The
chelating agent serves as stabilizer for the metal ions at pH 7.4.
This spike solution was used as make-up liquid for CEICPMS
measurements with isotope dilution.
The chelating agent 2,6-diacetylpyridine-bis(N-methylenepyri-
diniohydrazone)dichloride was prepared according to a procedure
described elsewhere [28]. Thiocarbamide (p.A.) as sulfur com-
pound for calibration was obtained from Merck.
Two metallothionein samples were investigated a preparation
of rabbit-liver metallothionein (MT preparation, Lot 19H7812)
and further purified rabbit-liver metallothionein 1 (MT-1 prepara-
tion, Lot 127H7810), which were purchased from Sigma (St Louis,
MO, USA). Dilutions of metallothionein were stored under an ar-
gon atmosphere at 18 C.
All dilutions were prepared with filtered 18 M deionized wa-
ter from a Milli-Q system (Millipore, Milford, MA, USA)
CEICPMS system for on-line isotope dilution
The CEICPMS system was used to perform on-line isotope di-
lution. One important part of the coupling interface was a specially
designed nebulizer that worked in the-self aspiration mode at a sta-
ble flow rate of approximately 8.0 L min1. By self aspiration a
make-up liquid (NH4NO3 buffer, 15 mmol L1, pH 7.4) was sucked
continuously. This provided the electrical connection and adapted
the electro-osmotic flow rate inside the CE capillary to the flow
rate of the nebulizer. For use of this instrumental arrangement for
species-unspecific isotope dilution, the make-up liquid contained
the enriched isotopes 34S, 65Cu, 68Zn, and 116Cd. The eluent of the
CE capillary was continuously merged with the make-up liquid be-
fore the mixture was nebulized into the ICPMS for on-line iso-
tope ratio measurement of 32S/34S, 63Cu/65Cu, 64Zn/68Zn and
114Cd/116Cd. In this way, equilibration between spike and sample
was realized as an important precondition for IDMS analysis. The
concentrations of the enriched isotopes were optimized to obtain
isotope ratios in the range 0.12 in the isotope-diluted samples, to
keep the statistical error of the IDMS result low [21].
158
Isotope dilution calculations
On-line isotope dilution was performed in two steps. First, the
mass flow of the spike solution had to be calibrated by a reverse
isotope-dilution technique. A standard solution containing Cu, Zn,
Cd, and S with natural isotopic abundances was therefore injected
into the CE (50 kPa s; 54.3 nL) and electrophoresis was performed
(Fig. 1) under the conditions described above. The sulfur com-
pound thiourea was used, because it had a faster migration time
than SO42 under these conditions and no memory effect arising
from absorption on the capillary walls, as is observed for other or-
ganic sulfur compounds such as dithiothreitol, mercaptoethanol, or
cysteine. The metal ions had to be stabilized by complexing with a
strong ligand, where the resulting complex had a positive charge,
so that the metals could migrate towards the capillary outlet. The
compound 2,6-diacetylpyridine-bis(N-methylenepyridiniohydra-
zone), referred to above, fulfilled these requirements [29] and was
used to complex the metal ions. The mass flow MFstd of the stan-
dard was calculated from the base-peak width tpeak in Fig. 1, the in-
jection volume vinj and the concentration of the standard cstd using
Eq. (1).
C std Vi n j
M Fstd = (1)
t peak
MFstd, determined by this procedure, represented strictly just the
mass flow of the standard during the time of the measurement of
the corresponding peak in Fig. 1. Because of the constant electro-
osmotic flow and constant make-up flow it was assumed that MFstd
was also constant during IDMS measurement. Rcal is the average
isotope ratio of the particular peaks in Fig. 1.
The mass flow of the spike solution MFspike was calculated us-
ing the values of MFstd and Rcal determined, in accordance with
Ref. [22].
In a second step the mass flow of the sample MFsample as a func-
tion of time was calculated by use of Eq. (2), the derivation of
which is described elsewhere [22]. This calculation is demon-
strated using sulfur as an example:
 
spi ke h spi ke Rsample (t)
h 32 34
Msample
M Fsample (t) = M Fspi ke  
Mspi ke h 34 Rsample (t) h 32
sample sample
(2)
where Msample is the atomic mass of S in the sample, Mspike the
atomic mass of S in the spike solution, hisotopesample the isotope
abundance of the respective isotope in the sample, and hisotopespike
that of the respective isotope in the spike solution. The particu-
lar isotope abundances were determined by ICPMS before mea-
surement, whereby possible background values of the isotopes
Fig. 2 Recovery of a standard
solution containing Cu, Zn,
Cd, and S, measured by
CEICPIDMS
were taken into account. Rsample(t) is the on-line measured isotope
ratio during electrophoresis of the isotope-diluted sample. When
MFsample(t) was plotted against migration time a mass flow electro-
pherogram was obtained. Integration of the peaks resulted directly
in the mass of the respective element in the different species. Three
terms in Eq. (2), hisotopesample, hisotopespike, and Rsample(t), can be influ-
enced by mass discrimination effects. As Heumann et al. have
shown [27] the mass discrimination effects for IDMS analyses are
eliminated if the isotopic abundances are measured under the same
conditions. In this case, the three different isotopic terms in Eq. (2)
compensate each other. Accurate results can therefore be obtained
without any mass bias corrections.
Results and discussion
Verification of the method
The CEICPIDMS method developed was checked by
analyzing a standard solution. After calibrating the system
for the spike mass flow, a standard solution containing
3 g Cu mL1, 1.5 g Zn mL1, 3 g Cd mL1, and 25 g
S mL1 was injected as a sample and the recovery was de-
termined. The results are shown in Fig. 2. The recoveries
were between 101.7%, for Cu, and 106.6%, for Zn, with a
precision (n=4) of approximately 15%.
Quantification and characterization of MT-1
The electropherogram obtained for 32S from the rabbit
liver MT-1 preparation (Lot 127H7810) contained four
peaks (Fig. 3). The most intense, peak 3, corresponded to
the isoform MT-1. Peak 4 was a small residue of MT-2,
which indicated that during the clean-up procedure the
isoform MT-1 was not completely separated and cleaned
from MT-2 and other residues. This electropherogram
also showed that it was possible to measure 32S directly in
metallothioneins by ICPsector-field MS in the medium
resolution mode. The 34S curve represented the continu-
ous addition of the spike solution. Because the intensities
of both isotopes were of the same order of magnitude, the

Fig. 3 Sulfur-isotope electro-
pherograms (a) and corre-
sponding sulfur isotope ratio
electropherogram (b) measured
for MT-1 by ICPIDMS
optimum isotope ratio for 32S/34S in the range of 0.12
was achieved over the whole electropherogram (Fig. 3b).
Because of the 32S background values of the system, the
measured isotope ratio for 32S/34S at the base line in Fig. 3b
differed from the theoretical isotope ratio given by the
isotopic composition of the spike solution.
Simultaneously, the metal isotopes 63Cu, 65Cu, 64Zn,
68Zn, 114Cd and 116Cd were monitored during the elec-
trophoresis of MT-1. Figure 4 shows the isotope ratios for
the metals. At the MT-1 peak mainly Cd and Zn were de-
tected; peaks 5 and 6 in the Cu electropherogram (Fig. 4c)
were before and behind the MT-1 peak.
By applying Eq. (2) the isotope ratio electropherogram
of MT-1 for S, Cu, Zn and Cd was converted into a mass-
flow electropherogram (Fig. 5), where the Cu mass flow
was very low. Integration of the MT-1 peak in the sulfur
mass flow electropherogram resulted in the amount of sul-
fur. Three independent measurements were performed
159
with two calibrations of the mass flow of the spike solu-
tion. The average amount of sulfur was 0.460.05 ng. This
could be used to calculate the absolute quantity of rabbit
liver MT-1; this was 4.69 ng, assuming an average molec-
ular mass of 6870 D for MT (obtained from Ref. [8]). This
resulted in an MT-1 concentration of 2.13 mg mL1 for an
injection volume of 2.2 nL.
The amounts of the metals in MT-1 were quantified by
the same procedure, and found to be 0.430.07 ng Cd and
0.040.007 ng Zn. Thus the stoichiometric composition
of the MT-1 complex could be characterized by calculat-
ing the molar ratio of sulfur to metal, normalized to 21 sul-
fur atoms per MT molecule. This resulted in the rounded
ratios S:Cd:Zn=21:6:1 for MT-1. As a consequence it
could be stated that the rabbit liver MT-1 complex inves-
tigated in this preparation could be characterized as
Cd6Zn1-MT-1.
160
Fig. 4 Isotope-ratio electro-
pherograms of (a) Cd, (b) Zn,
and (c) Cu measured for MT-1
by ICPIDMS
Analytical characteristics of the MT-1 determination
From the measurements described above, the limits of de-
tection could be estimated as 6.8 g mL1 for sulfur and
72.6 g mL1 for MT-1, in accordance with the 3 crite-
rion. Note that 95% of the transmission is lost in
ICPSFMS when operating in the medium resolution
mode; this strongly influences detection limits.
The precision of the MT-1 quantification was 11.3%.
This value must be compared with the relative standard
deviation (RSD) of the peak area of MT-1 measured with-
out isotope dilution. In an independent experiment using
the same CEICPMS system the precision of the peak
area was determined as 9.9% [7]. The comparison indi-
cates that the main source of error in the reproducibility
lies in the electrophoresis, including sample injection and
not in the isotope-dilution technique, where more precise
results are expected [21]. To improve the precision of the
peak area an internal standard will be used in future in-
vestigations. The application of HPLC as an alternative
 161
Fig. 5 Mass-flow electrophe-
rograms of S, Cd, Zn, and Cu
for MT-1 (offset Cd: 0.015,
Zn: 0.0075, Cu: 0.001 for better
view)

Table 2 Comparison of total

element determination in the CEICPIDMS TXRF Amount (%, m/m) of Cu, Zn, and Cd in MT-1
MT-1 preparation (Lot (g mL1) (g mL1)
127H7810) measured by Measured by Measured by Specified by
CEICPIDMS and TXRF CEICPIDMS TXRF SigmaAldrich
Sulfur 772104 433118
Copper 41 6 18 5 0.5 0.4 n.d.
Zinc 52 4 54 15 0.7 1.2 1.1
Cadmium 691116 357 99 8.8 8.1 7.9

separation technique might also result in better precision,
but in comparison with HPLC capillary electrophoresis
has better separation efficiency for MT isoforms even
subisoforms can be separated.
Mass balance of the MT-1 determination
trast, the percentage of Cu, Zn, and Cd in MT-1 measured
by CEICPIDMS matched the TXRF results and also the
values specified by SigmaAldrich. Although the preci-
sion of the TXRF measurement was only approximately
30% for these determinations, the results suggested that
the total element contents were indeed represented by the
peaks in the mass flow electropherogram.

The MT-1 preparation (Lot 127H7810) investigated was

prepared to a concentration of 5 mg mL1, while the mea-
sured MT-1 quantity was 2.13 mg mL1. To check the
mass balance of the MT-1 determination, the total con-
tents of S, Cu, Zn, and Cd in the MT-1 preparation were
calculated by integration of the whole time window from
450 to 650 s in the mass flow electropherogram (Fig. 5).
These values were compared with total element contents
measured in the MT-1 preparation by total reflection
X-ray fluorescence (TXRF), which can determine element
concentrations in a 2 L sample. The results are presented
in Table 2. The values from TXRF were a factor of two
lower than those obtained by CEICPIDMS, probably
because of matrix effects from the Tris buffer, which
could have influenced the TXRF measurement. In con-
Quantification and characterization of MT isoforms
The rabbit liver MT preparation (Lot 19H7812) contains
the isoforms MT-1 and MT-2, some sub-isoforms and
degradation products, as indicated by at least seven peaks
in the electropherogram [7], depending on the separation
conditions. This MT preparation was investigated by
CEICPIDMS to characterize the composition of the
single peaks, in an attempt to give tentative formulas for
the metal complexes in the MT isoforms. The resulting
mass flow electropherogram, obtained by use of Eq. (2),
was therefore converted into an electropherogram in
moles (Fig. 6). Thus the molar ratios of sulfur to metals
can be obtained directly from the peak heights. Using the
162
Fig. 6 Electropherograms of S, Cd, Cu, and Zn of a rabbit liver
MT preparation; given in moles (offset Cd: 0.6, Cu: 0.11 for better
view)
peak heights instead of the peak areas has the advantage
that even peaks that are not fully resolved, such as peaks
4 and 5 in Fig. 6, can be evaluated.
The electropherogram in Fig. 6 shows a complex mix-
ture of different MT isoforms. Peak 4 and 5 were identi-
fied as MT-1 and peak 6 as MT-2 by comparison with the
electropherograms of purified MT-1 and MT-2 prepara-
tions [7]. Peak 5 in Fig. 6 (MT-1) corresponds to peak 3 in
Fig. 5 and peak 6 in Fig. 6 (MT-2) to peak 4 in Fig. 5. MT-1
and MT-2 in Fig. 6 have longer migration times than in
Fig. 5. The migration times are shifted in Fig. 6 because of
the larger amount of matrix present in the MT preparation
in comparison with the further purified MT-1 preparation.
The MT isoform belonging to peak 3 has not yet been
identified. Further investigations with molecule-specific
detectors such as electrospray MSMS will therefore be
necessary to identify this peak. Peaks 1 and 2 are probably
degradation products, because their intensity increases
with sample age. The quantity of sulfur of the species was
obtained from the mass-flow electropherogram; the re-
sults were 8.40.6 ng and 3.80.3 mg mL1 for MT-1 and
5.50.4 ng and 2.50.2 mg mL1 for MT-2. The sum of
Table 3 Molar ratios of sulfur to metals in MT, and suggested for-
mulas
MT (Fig. 6) Molar ratio S:Cd:Cu:Zn Suggested formula
Peak 3 21:5:4:0 Cd5Cu4-MT
Peak 4 21:5:1:1 Cd5Cu1Zn1-MT-1
Peak 5 21:7:5:1 (probably a mixture of
MT-1 subisoforms)
Peak 6 21:6:0:1 Cd6Zn1-MT-2
the measured MT quantity corresponding to peaks 16 in
Fig. 6 was 8.80.6 mg mL1, a good match with the sam-
ple investigated. This sample was prepared to a total MT
concentration of 10 mg mL1.
The molar ratios of sulfur to metals in the signals ob-
served in the electropherograms, and tentative suggestions
for their formulas, are summarized in Table 3. These re-
sults are based on three independent measurements and
two calibrations of the spike mass flow. Peaks 3 and 5
contain more than seven metal atoms per MT molecule,
because of the presence of copper. Peak 5 probably does
not represent a single compound. The unusual stoichio-
metric composition and the small peaks (including peak
4) close to peak 5 suggest the presence of more com-
pounds, probably a mixture of MT-1 subisoforms. For fur-
ther investigations the separation of these peaks will have
to be improved. In contrast, peak 6 can be characterized as
a Cd6Zn1-MT-2 isoform.
Conclusions
The coupling of capillary electrophoresis with ICPMS is
suitable for application of an on-line isotope dilution tech-
nique. This enables quantification of elements in unknown
species after CE separation, which can be used to investi-
gate metallothionein isoforms. Because of the known
number of 21 sulfur atoms per MT molecule, the sulfur
isotope-dilution technique is a promising approach to the
quantification of MT isoforms after CE separation. Addi-
tional measurements of the metals Cu, Zn, and Cd then
determination of the sulfur-to-metal ratio in MT, enables
characterization of the stoichiometric composition of the
metalloprotein complexes.
The quantification method presented is limited to pro-
teins such as MT, where the sulfur stoichiometry is known.
For unknown compounds additional structural informa-
tion, which can be obtained, e.g., from electrospray tan-
dem mass spectrometry (ESIMSMS), is necessary.
Further investigations will be performed to improve
the sensitivity of sulfur measurements by the use of an
octapole reaction system (ORS) in combination with an
ICPquadrupole MS and the use of more highly enriched
34S and 33S spikes. Details of the application of the
method to MT determination in animal and human tissues
will be published separately in the near future.

Acknowledgement The authors thank Dipl. Ing. Burkhard Erbslh
for performing the TXRF measurements.
References
1. Lobinski R (2001) Fresenius J Anal Chem 369:113114
2. Szpunar J (2000) Analyst 125:963988
3. Krylow S N, Dovichi N J (2000) Anal Chem 72:111R128R
4. Majidi V (2000) Microchem. J 66:316
5. Schaumlffel D, Prange A (1999) Fresenius J Anal Chem 364:
452456
6. Prange A, Schaumlffel D (1999) J Anal At Spectrom 14:1329
1332
7. Prange A, Schaumlffel D, Brtter P, Richarz A-N, Wolf C
(2001) Fresenius J Anal Chem 371:764774
8. Mounicou S, Polec K, Chassaigne H, Potin-Gautier M, Lobinski
R (2000) J Anal At Spectrom 15:635642
9. Taylor KA, Sharp BL, Lewis DJ, Crews HM (1998) J Anal At
Spectrom 13:10951100
10. Nordberg M (1998) Talanta 46:243254
11. Stillman MJ (1995) Coord Chem Rev 144:461511
12. Onosaka S, Cherian MG (1981) Toxicology 22:91
13. Eaton DL, Toal BF (1982) Toxicol Appl Pharmacol 66:134
142
14. Vandermallie RJ, Garvey JS (1979) J Biol Chem 254:8416
15. Garvey JS (1984) Environ Health Perspect 54:117
163
16. Kgi JHR, Himmelhoch SR, Whanger PD, Bethune JL, Vallee
BL (1974) J Biol Chem 249:3537
17. Nordberg GF, Nordberg M, Piscator M, Vesterberg O (1972)
Biochem J 126:491
18. Minami T, Tohno Y, Okazaki Y, Kubo K, Otaki N, Kimura M
(1998) Anal Chim Acta 372:241247
19. Beattie JH (1998) Talanta 46:255270
20. Wolf C, Rsick U, Brtter P (2000) Fresenius J Anal Chem
368:839843
21. Heumann KG (1988) In: Adams F, Gijbels R, van Grieken R
(eds) Inorganic Mass Spectrometry. Wiley, New York, pp 301
376
22. Rottmann L, Heumann KG (1994) Fresenius J Anal Chem 350:
221227
23. Heumann KG, Rottmann L, Vogl J (1994) J Anal At Spectrom
9:13511355
24. Valles Mota JP, Fernndez de la Campa MR, Garca Alonso JI,
Sanz-Medel A (1999) J Anal At Spectrom 14:113120
25. Valles Mota JP, Encinar JR, Fernndez de la Campa MR,
Garca Alonso JI, Sanz-Medel A (1999) J Anal At Spectrom
14:14671473
26. Marx G (2000) PhD thesis, Johannes Gutenberg-University
Mainz
27. Heumann KG, Gallus SM, Rdlinger G, Vogl J (1998) J Anal
At Spectrom 13: 10011008
28. Main MV, Fritz JS (1989) Anal Chem 61:12721275
29. Timerbaev AR, Semenova OP, Bonn GK, Fritz S (1994) Anal
Chim Acta 296:119128
Anal Bioanal Chem (2002) 372 : 164168
DOI 10.1007/s00216-001-1126-5
O R I G I N A L PA P E R
Yoshihiro Saito Motohiro Imaizumi Tsutomu Takeichi
Kiyokatsu Jinno
Miniaturized fiber-in-tube solid-phase extraction
as the sample preconcentration method
for microcolumn liquid-phase separations
Received: 23 July 2001 / Accepted: 20 September 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract Miniaturized fiber-in-tube solid-phase extrac-
tion (fiber-in-tube SPE) has been developed as a solvent-
less sample preconcentration technique for microcolumn
liquid-phase separation methods. Short capillaries packed
with polymer filaments were employed as the extraction
tube and the preconcentration power for phthalates in
aqueous solutions was studied. On the basis of the suc-
cessful on-line coupling of this preconcentration method
with liquid chromatography (LC), a more miniaturized
extraction cartridge, which is installed in the rotor of the
micro-injector, has been developed. With a modified com-
mercially available valve, on-line coupling of this sample
preconcentration method to capillary electrochromatogra-
phy (CEC) was also investigated.
Keywords Fiber-in-tube solid-phase extraction
Miniaturization Sample preparation Microcolumn
liquid chromatography Capillary
electrochromatography
Introduction
The development of miniaturized analytical systems has
been one of the most important projects in analytical
chemistry because it promises a solution to recent require-
ments and rapid analysis with low operation cost, but also
without environmental pollution. As a typical example of
the miniaturization of separation systems, microcolumn
liquid chromatography (micro-LC) and related techniques
have been widely investigated [1, 2]. Taking the advan-
tage of the low volumetric flow-rate of the mobile phase,
on-line coupling of micro-LC to various spectroscopic de-
tection systems, such as Fourier transform infrared spec-
Y. Saito M. Imaizumi T. Takeichi K. Jinno ()
School of Materials Science,
Toyohashi University of Technology,
Toyohashi 4418580, Japan
e-mail: jinno@chrom.tutms.tut.ac.jp
troscopy (FTIR), mass spectroscopy (MS) and inductively
coupled plasma spectroscopy (ICP), was developed and
specific detection with high sensitivities was accom-
plished by these hyphenated systems. The combination of
a specially designed stationary phase and a highly sensi-
tive/selective detector made it possible to analyze com-
plex sample mixtures.
In contrast with many successful applications of mi-
cro-scale liquid-phase separation techniques with high-
performance separation and detection, however, the re-
ports for this miniaturized sample preparation method,
which was specially designed for these separation systems
[3, 4, 5], have been somewhat limited because of the dif-
ficulties associated with on-line coupling. On the basis of
the previous results from the solid-phase microextraction
(SPME) technique of Pawliszyn et al. [6, 7, 8, 9, 10, 11,
12, 13, 14] and Jinno et al. [15, 16, 17, 18, 19, 20, 21, 22]
a novel miniaturized sample preparation method [23] was
recently developed. In the newly developed method fiber-
in-tube SPE, extraction has been carried out in short cap-
illaries packed with the filaments of synthetic polymer as
the extraction medium. The results have shown that the
effective contact of the sample solution with several hun-
dreds of the fine fibrous extraction media in the extraction
tube could provide miniaturization in this micro-scale
sample preconcentration device. Down-sizing of the ex-
traction device will also permit direct coupling of the ex-
traction process to microcolumn separation methods with-
out any disadvantages such as over-load injection and
poor resolution during separation.
As an extension of the previous work [23], miniatur-
ized fiber-in-tube SPE has been developed as a solvent-
less sample-preconcentration technique for microcolumn
liquid-phase separations, such as micro-LC and capillary
electrochromatography (CEC). For preparation of the ex-
traction tube a short capillary of polyetheretherketone
(PEEK) or polytetrafluoroethylene (PTFE) was packed
longitudinally with fine polymer filaments having both
solvent resistance and mechanical strength. The extraction
cartridge of fiber-in-tube SPE was incorporated in an in-
jection valve and used as the sample loop itself. Desorp-
 165

tion and injection are carried out simultaneously without

desorption solvent, which was required in the previous in-
vestigations [15, 16, 17, 18, 19, 20, 21, 22].

Experimental
Reagents and materials

Zylon fiber (HM type, Fig. 1) was obtained from Toyobo, Ohtsu,
Japan. All solvents were of analytical grade and obtained from
Kishida Chemical, Osaka, Japan. Dimethylphthalate (DMP), di-
ethylphthalate (DEP), di-n-propylphthalate (DPP), di-n-butylph-
thalate (DBP), and aromatic hydrocarbons were purchased from
Tokyo Chemical Industries, Tokyo, Japan. All PEEK and PTFE
tubing, and fused-silica capillaries were obtained from GL Sci-
ences, Tokyo, Japan. Water was purified by use of a Milli-Q Wa-
ter Purification System (Millipore, Tokyo, Japan). Fig. 2 Miniaturized fiber-in-tube SPEmicro-LC system. The ex-
traction was carried out in the extraction tube between the switch-
ing valve (model 7000, Rheodyne, Cotati, CA, USA) and the in-
Preparation of the extraction tube for fiber-in-tube SPE jection valve (model 7520, Rheodyne)

As shown in Table 1, three types of extraction tube were prepared

with longitudinal packing process of Zylon filaments. One was
prepared from a 24-mm length PEEK capillary packed with fila-
ments of the same length, and used as tubing between switching
valve and injection valves (Fig. 2). Another was incorporated in
the sample rotor of the micro-injection valve (Fig. 3), where the
fiber-packed capillary was sandwiched between two small pieces
of blank capillaries having a smaller i.d. to immobilize and to con-
tain the packed filaments. The other was designed for coupling to
electrokinetic separation methods, in which a commercially avail-
able 4-port 2-way valve was employed. The extraction tube was
placed in the rotor of the valve. The overview of the fiber-in-tube
SPECEC system and a close-up view of the extraction/injection
valve are shown in Fig. 4.

Fig. 3 In-rotor configuration of the extraction tube for micro-

LC applications. The extraction tube was incorporated in the sam-
ple rotor of the Rheodyne 7520 micro-injector

Fig. 1 Chemical structure of Zylon fiber

Micro-LC system
Table 1 Miniaturized fiber-in-tube SPE cartridges prepared in Micro-LC was performed with an Ultra-Plus II Capillary LC pump
this work (Micro-Tech Scientific, Sunnyvale, CA, USA) and a UVVis ab-
sorption detector (Model 875-UV, Jasco, Tokyo, Japan) with a
Tubing Dimensions Number Void Configu- home-made flow-cell of about 0.3-L volume. The detection
of packed volumea ration wavelength was set at 254 nm. With a modification in the rotor
filaments (sample loop), model 7520 micro-injector (Rheodyne) was used. A
laboratory-made packed capillary column (0.53 mm i.d.200 mm)
PEEK 0.25 mm i.d.24 mm 330 0.36 L Between was prepared with Develosil ODS-5 stationary phase (5 m, No-
valves mura Chemical, Seto, Japan) using a slurry packing method.
PEEK 0.50 mm i.d.5.0 mm 1490 0.21 Lb In rotor
of 7520c
PTFEd 0.25 mm i.d.5.0 mm 380 0.048 L In rotor CEC system
of HV4-1e
CEC measurements were performed with a laboratory-made sys-
aCalculated from the outer diameter of each filament, 11.5 m tem comprising a Model HCZE-30PN0.25 high-voltage power
bIncluding the void volume of blank PEEK tubing (0.25 mm
supply (Matsusada Precision, Kusatsu, Japan), a modified UVVis
i.d.1.0 mm) at the top and bottom of the extraction tube, the total absorption detector (Model 875-UV, Jasco) for on-column detec-
void volume will be about 0.31 L tion and an HV4-1 Plug Valve (Hamilton, Reno, NV, USA) for the
cThe cartridge was incorporated in the sample rotor of the Rheo-
preconcentration and injection of the sample. The typical applied
dyne 7520 valve voltage was 30 kV and the detection was made at 230 nm. As
dPTFE was selected for smoother rotation of the valve
shown in Fig. 4, the separation column of 0.15 mm i.d. packed with
eThe cartridge was incorporated in the rotor of the Hamilton HV4-1
Superiorex ODS (5 m, Shiseido, Yokohama, Japan) was con-
valve for CEC applications nected to the valve. The packed length was 50 mm (total length
166
Fig. 4 On-line coupling of miniaturized fiber-in-tube SPE and
CEC separation system
from the valve to the cathodic vial was 200 mm) and the window
for on-column detection was made just after the mid-frit. A pre-
column capillary of 0.25 mm i.d.150 mm length was connected to
the valve port on the other side, to supply the mobile phase from
anodic vial during the CEC separation. A PTFE tube from a sy-
ringe pump (Microfeeder MF-2, Azumadenki Kogyo, Tokyo,
Japan), to deliver the sample solution during the extraction
process, and a short fused-silica capillary (0.030 mm i.d.50 mm
length), for waste, were connected to the other two ports.
Data processing
For data acquisition and processing, Borwin chromatography data-
handling software (Jasco) running on a personal computer was
used. All measurements were carried out at least in triplicate and
the relative standard deviations (RSDs) were less than 4% for all
runs.
Results and discussion
First, the system shown in Fig. 2 was employed for the
analysis of phthalates. The extraction capillary was placed
between the switching valve and the injection valve, and
the extraction was carried out as follows. First, an aque-
ous sample solution was delivered to the extraction tube
by means of a syringe pump, and the analyte was ad-
sorbed onto the fibers. Next, a certain amount of the des-
orption solvent was also pumped after changing the
switching valve. At the same time the injector was
switched to fill the sample loop with the effluent from the
extraction tube. The injection was made just after sample
loading. All operational parameters, such as the flow-rate
Fig. 5 Typical chromatogram for the analysis of DBP with the
fiber-in-tube SPEmicro-LC system shown in Fig. 3. Conditions:
extraction tube, 0.50 mm i.d.5.0 mm PEEK packed with Zylon
(HM) filaments (1490 filaments); packing density, 80%; extrac-
tion, 32 L min1 for 15 min (480 L total volume); separation
column, 0.53 mm i.d.200 mm length packed with Develosil
ODS-5 stationary phase (5 m); mobile phase, methanolwater,
95:5 (4 L min1). Sample, DBP (1.0 g mL1) in water
and time for the extraction and desorption, were deter-
mined in preliminary experiments. For the fiber-packed
tube of 0.25 mm i.d.24 mm length (Table 1), the extrac-
tion flow-rate was 16 L min1 and desorption was per-
formed at a flow-rate of 2 L min1 for 4 min. The injec-
tion volume was 1 L, and the mobile phase was pumped
at 4 L min1 for the micro-LC analysis. The preconcen-
tration factor of this method was calculated by injecting
the standard solution in a conventional way, and was esti-
mated as about 101 for DBP with the extraction efficiency
of 21%. The volume of solvent needed for each analysis
was about 60 L.
No desorption solvent was needed for the extraction
device shown in Fig. 3. The extraction cartridge was in-
corporated in the rotor of the micro-injection valve. The
cartridge was prepared with a PEEK tube of 0.50 mm
i.d.5.0 mm. Two pieces of blank PEEK tubes, 0.25 mm
i.d.1.0 mm length, were employed to retain the fiber-
packed cartridge in the rotor. Furthermore, the modifica-
tion was made with a small piece of PEEK tubing to re-
duce the void volume of the remaining flow-pass in the
valve. During the extraction a certain volume of sample
solution was pumped by a syringe at the appropriate flow-
rate. The analytes of interest were adsorbed onto the sur-
face of the filaments packed in the extraction tube.
Changing the position of the injector the desorption was
carried out by the flow of the mobile phase. As shown in
Table 1, the total number of packed-filaments was about
1500, and the void volume of the fiber-packed injection
valve was about 0.31 L. Figure 5 shows the typical chro-
matogram obtained with an extraction cartridge of 5-mm
length. By this in-valve configuration, the further
miniaturization was accomplished as well as the solvent-
less sample preparation method for micro-LC separation.

Fig. 6 Typical chromatogram for the analysis of phthalates with
the fiber-in-tube SPECEC system. Conditions: extraction tube,
0.25 mm i.d.5.0 mm PTFE packed with Zylon (HM) filaments
(380 filaments); packing density, 80%; extraction, 4 L min1 for
5 min (20 L total volume); separation column, 0.15 mm i.d.
50-mm effective length (200 mm total length); stationary phase,
Superiorex ODS (5 m); precolumn capillary, 0.25 mm i.d.
100 mm; applied voltage, 30 kV; mobile phase, methanol
Tris-buffer (0.5 mmol L1), 90:10. Sample, phthalates mixture
(10 g mL1 each) in water
The total volume of the solvent required, including the
volume of the mobile phase, was reduced to about 40 L
for each analysis in this system.
On the basis of successful direct coupling of the fiber-
in-tube SPE to micro-LC separation, the application of
this extraction technique to sample preparation for CEC
analysis was also investigated with the system shown in
Fig. 4. As described in the Experimental section, the fiber-
packed cartridge was made with a 5-mm piece of PTFE
tubing (0.25 mm i.d.), and the tubing was incorporated
into the rotor of a small 2-way valve. Figure 6 depicts a
typical chromatogram for the analysis of phthalate mix-
tures with fiber-in-tube SPECEC system. Although neg-
ative baseline drift, probably due to the injection of a wa-
ter plug, was observed, a good separation was obtained.
No solvent was needed for the sample preparation, and
the total usage of organic solvent (as the mobile phase
component) for each analysis was about 2.5 L. The pre-
concentration factor was calculated based on comparison
of the peak areas with fiber-in-tube SPE preconcentration
and that without the preconcentration procedure. In the
latter case, the injection was made by pumping the stan-
dard sample prepared as the methanol solution, because in
the preliminary experiments no preconcentration effect
was observed with the methanol as the sample solvent.
Table 2 shows the calculated preconcentration factor and
the extraction efficiency. The results demonstrated that
good preconcentration was obtained only with a short
fiber-packed tubing, and a better preconcentration effect
was also observed for less-polar solutes. Analysis of the
167
Table 2 Preconcentration factor and extraction efficiency for di-
n-alkylphthalates [Ph(COOR)2] with fiber-in-tube SPECEC sys-
tem
Preconcentration Extraction
factor efficiency (%)
R=CH3 31.5 11.5
R=C2H5 38.8 14.2
R=n-C3H7 56.5 20.6
R=n-C4H9 80.2 29.4
Fig. 7 Plot of the preconcentration factor vs. log P value of phtha-
lates
relationships between the polarity of the solutes and pre-
concentration factor has been carried out. The plot in
Fig. 7 clearly indicates the linear relationship between log
P (where P is the 1-octanol/water partition coefficient)
[24] and the preconcentration factor, with a correlation
coefficient (r) of 0.990. The results are in good agreement
with our previous results from use of the fibrous packing
materials as the stationary phase in LC [25, 26] and CEC
[27, 28, 29], in which the elution order from the fiber-
packed column is the same as for conventional reversed-
phase packings.
Conclusions
A novel miniaturized sample-preparation method has
been developed with a short capillary and fine fibrous
packing materials as extraction media. Incorporating the
extraction cartridge into the injection valve, a solventless
sample preparation technique was realized for the analysis
of water samples by microcolumn liquid-phase separation
methods. The results also show the possibility of new fi-
brous extraction media having different chemical struc-
tures and/or coated with other materials, which are spe-
cially designed with the concept of molecular shape
recognition [30].
The applications of polymer-coated fibrous materials,
not only as the sample-preparation media but as the sta-
tionary phase materials for various chromatographic sepa-
ration techniques, such as GC or supercritical-fluid chro-
168
matography (SFC), are currently being investigated in our
laboratory along with the miniaturization of these tech-
niques.
Acknowledgements A part of this research was financially sup-
ported by Grant-in-Aids for Scientific Basic Research (C (2), No.
12640586) and for Young Scientists (A, No. 13740421) from
Japan Society for the Promotion of Sciences. The authors would
like to thank Toyobo Co. Ltd. for the donation of Zylon fiber. YS
also thanks Mr Yutaka Ohtsu (Basic Research Center, Shiseido,
Yokohama, Japan) and Dr Norikazu Nagae (Nomura Chemical,
Seto, Japan) for providing stationary phases.
References
1. Ishii D (1988) (ed) Introduction to microscale high-perfor-
mance liquid chromatography. VCH, New York, NY, USA
2. Yang FJ (1989) (ed) Microbore column chromatography,
Chromatographic Science Series, vol 45. Marcel Dekker, New
York, NY, USA
3. Yang Q, Tomlinson AD, Naylor S (1999) Anal Chem 71:
183A189A
4. Whang CW, Pawliszyn J (1998) Anal Commun 35:353356
5. Fritz JS, Masso JJ (2001) J Chromatogr A 909:7985
6. Chen J, Pawliszyn J (1995) Anal Chem 67:25302533
7. Boyd-Boland AA, Pawliszyn J (1996) Anal Chem 68:1521
1529
8. Eisert R, Pawliszyn J (1997) Anal Chem 69:31403147
9. Lord HL, Pawliszyn J (1998) LCGC 16:S41S46
10. Jia C, Luo Y, Pawliszyn J (1998) J Microcol Sep 10:167173
11. Mder M, Popp P, Pawliszyn J (1998) J Microcol Sep 10:
225234
12. Kataoka H, Lord HL, Pawliszyn J (1999) J Chromatogr B 731:
353359
13. Kataoka H, Narimatsu S, Lord HL, Pawliszyn J (1999) Anal
Chem 71:42374244
14. Kataoka H, Pawliszyn J (1999) Chromatographia 50:532538
15. Jinno K, Muramatsu T, Saito Y, Kiso Y, Magdic S, Pawliszyn
J (1996) J Chromatogr A 754:137144
16. Jinno K, Han Y, Sawada H, Taniguchi M (1997) Chro-
matographia 46:309314
17. Jinno K, Taniguchi M, Hayashida M (1998) J Pharm Biomed
Anal 17:10811091
18. Jinno K, Sawada H, Han Y (1998) Biomed Chromatogr
12:126127
19. Jinno K, Kawazoe M, Hayashida M (2000) Chromatographia
52:309313
20. Salleh SH, Saito Y, Jinno K (2000) Anal Chim Acta 418:6977
21. Salleh SH, Saito Y, Jinno K (2001) Anal Chim Acta
433:207215
22. Saito Y, Kawazoe M, Hayashida M, Jinno K (2000) Analyst
125:807809
23. Saito Y, Nakao Y, Imaizumi M, Takeichi T, Kiso Y, Jinno K
(2000) Fresenius J Anal Chem 368:641643
24. Hansch C, Leo A, Hoekman D (1995) (eds) Exploring QSAR
hydrophobic, electronic, and steric constants. ACS Profes-
sional Reference Book, American Chemical Society, Washing-
ton, DC, USA
25. Kiso Y, Jinno K, Nagoshi T (1986) J High Resol Chromatogr
Chromatogr Commun 9:763764
26. Kiso Y, Takayama K, Jinno K (1989) J High Resol Chromatogr
12:169173
27. Jinno K, Wu J, Sawada H, Kiso Y (1998) J High Resol Chro-
matogr 21:617619
28. Jinno K, Watanabe H, Kiso Y (2001) J Biochem Biophys
Methods 48:209218
29. Jinno K, Watanabe H, Saito Y, Takeichi T (2001) Electro-
phoresis 22:33713376
30. Jinno K (ed) (1996) Chromatographic separations based on
molecular recognition. WileyVCH, New York, NY, USA
Anal Bioanal Chem (2002) 372 : 169173
DOI 10.1007/s00216-001-1162-1
O R I G I N A L PA P E R
Michel Lamotte Philippe Fornier de Violet
Philippe Garrigues Michel Hardy
Evaluation of the possibility of detecting benzenic pollutants
by direct spectrophotometry on PDMS solid absorbent
Received: 20 August 2001 / Revised: 2 October 2001 / Accepted: 15 October 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract Solid-phase extraction has become one of the
most commonly used techniques for preconcentration of
analytes from environmental samples. In the standard use
of solid sorbent phases the extracted pollutants are subse-
quently eluted with a suitable organic solvent before chro-
matographic analysis. An alternative to this procedure is
analysis of the adsorbed and concentrated pollutants by
direct application of a spectroscopic method (fluorimetry
or absorptiometry) to the phase. Although this method can-
not be expected to give results as precise as those given by
chromatographic methods, it might have valuable applica-
tions, particularly for on site pollution monitoring. This
paper reports an evaluation of the capability of the method
for the spectrophotometric detection of BTX (benzene,
toluene, xylenes) in aqueous media and in contaminated
atmospheres, with polydimethylsiloxane (PDMS) as sor-
bent. The tests performed, with the estimated detection
limits, indicate that the method is relatively simple and
easy to operate and sensitive enough for application to the
monitoring of pollution both in water and in air in an in-
dustrial ambient atmosphere.
Keywords Solid sorbents BTX Pollutants
Absorptiometry Fluorimetry
Introduction
Because of the widespread environmental occurrence of
volatile organic compounds (VOC), and particularly ben-
zenic-type compounds such as benzene, toluene, and
xylenes (BTX), there is continuous interest in the devel-
M. Lamotte () P.F. de Violet P. Garrigues
Laboratoire de Physico-Toxico-Chimie, UMR CNRS 5472,
Universit de Bordeaux 1,
351 Cours de la Libration, 33405 Talence, France
e-mail: m.lamotte@lptc.u-bordeaux.fr
M. Hardy
Supelco SigmaAldrich, 38298 Saint Quentin Fallavier, France
opment of reliable, practical, and sensitive methods for
their monitoring and analysis in air and natural water.
The most sensitive methods yet developed for the de-
tection and determination of volatile organic pollutants
usually include an extraction and concentration step, usu-
ally on a solid phase, followed, after desorption or elution,
by chromatographic analysis. All these are laboratory meth-
ods and require sophisticated instruments and complex
operations and procedures if reliable results are to be ob-
tained, and it is usually difficult to adapt these methods
for on-site pollution control. Although liquidsolid ex-
traction has benefited from important progress made in
the development of new solid adsorbents [1, 2], the proce-
dure entails an elution step with an organic solvent, the
use of which is increasingly subject to legislative restric-
tion.
Besides these conventional methods, several systems
and instruments have been proposed for on site monitor-
ing of atmospheric VOC pollutants [3, 4, 5, 6] or for wa-
ter monitoring [7, 8]; most of these systems require com-
plex and heavy instrumentation, however, and although
they seem suitable for some specific applications, their sen-
sitivity or selectivity do not always meet the requirements
imposed by legislation.
As an alternative approach, during the last ten years
there have been attempts to evaluate methods based on the
detection and/or analysis of hydrophobic organic pollu-
tants from aqueous media by application of a spectro-
scopic method (absorptiometry or fluorimetry) directly on
a solid sorbent phase after extraction [9, 10, 11, 12, 13,
14, 15, 16, 17, 18] thus avoiding use of organic solvents
for elution and chromatography. Although this method
cannot be expected to give results as precise as those ob-
tained by use of chromatographic methods, it has some
advantages it is, for example, simple to operate and suit-
able for on site pollution monitoring.
Whatever the type of sorbent used, the analytical
process comprises two steps:
the adsorption and concentration step in which the hy-
drophobic pollutants are selectively adsorbed and con-
centrated on/or in the solid phase.
170
spectroscopic analysis which consists in recording the
fluorescence or absorption spectra of the solid phase
containing the adsorbed pollutants.
Among the sorbing phases which have been tested [1, 2],
silica C18 extraction disks with fiber glass matrix and flu-
orimetric detection and a polydimethylsiloxane (OV1-
type) block sorbent with absorptiometric or fluorimetric
detection seem to be the most suitable.
In this paper we report the evaluation of a method
tested for the detection of BTX in water and air; polydi-
methylsiloxane (PDMS) was used for extraction and con-
centration and UVvisible absorption spectroscopy for
measurement. Our goal was to assess the suitability of the
method for on-site analysis of polluted media, to monitor
compliance with tolerable concentration limits fixed by
legislation.
Experimental
Preparation of the PDMS phase
The PDMS phase (OV1 type from Supelco) was shaped into paral-
lelepiped blocks in a brass mold at reduced temperature (50 C).
Because of the size of the spectrophotometer analyzing beam and
because the analyte molecules are concentrated in a very thin depth
from the surface [16], the size of the sorbent block was set at
6 mm10 mm for the effective absorbing face and 2 mm for the
thickness of the optical path.
The same PDMS block could be used several times by remov-
ing the adsorbed volatile analytes by means of a dry nitrogen
stream at room temperature. The use of a hot nitrogen stream at
130 C, as applied by Wittkamp et al. [15], was avoided to mini-
mize any possible modification of the characteristics of the poly-
mer upon heating.
Procedure for BTX analysis in water
For analysis of BTX in water, test solutions (200 mL) were pre-
pared in a 250-mL stainless-steel container by adding 10 L of a
concentrated stock solution to water (Milli-Q Millipore). The stock
solutions were prepared in acetonitrile at suitable concentrations.
The PDMS attached to its holder was then immersed in the spiked
aqueous solutions for a fixed period of time. After adsorption, the
PDMS block holder was removed and immediately placed in a reg-
ular fused silica cell (1 cm1 cm) in the presence of pure water
(Milli-Q water from Millipore) for absorbance measurement. Water,
instead of the normal atmosphere, was used to surround the adsor-
bent in the measurement cell for two reasons:
1. it prevents desorption of the adsorbed volatile analytes; and
2. because the refractive index of the PDMS used (n1.37) is
closer to that of water than that of air, it greatly improves the
optical transparency of the solid phase.
Procedure for atmospheric BTX analysis
In the analysis of BTX in a contaminated atmosphere tests were
performed similarly to those described above for water analysis,
except that small volumes (<10 L) of suitably concentrated stock
solutions were introduced into a 1-L glass flask and left to evapo-
rate for 15 min. The PDMS attached on its holder was then hung in
the flask for a fixed period of time. After adsorption, the PDMS
holder was placed in a regular fused silica cell for absorption mea-
surement as described above.
Absorption spectra measurements
Absorption spectra were recorded with a Jasco (V-560) spec-
trophotometer in double-beam mode with air as reference and with
2 nm resolution. This apparatus was equipped with a double mono-
chromator which enabled precise absorbance measurement up to 3
absorbance units in the range 240300 nm and with good precision
( 0.0002 absorbance units with 2 nm band-pass width) because of
the very small amount of stray light (<0.0005% ) and if a single-
beam mode is adopted. The final spectra were obtained by differ-
ence between the spectra of the same PDMS block before and af-
ter adsorption. Because of the degradation of the optical properties
of the PDMS block during the adsorption process, the absorption
spectra of the analyte are always superimposed on a background
signal which increases gradually from the short wavelength side of
the spectrum to the long wavelength side. The absorbance was,
therefore, always measured on the first absorption band where the
most precise extrapolation of the underlying background could be
made.
All experimental work was performed in a thermostatted labo-
ratory at 22 C.
Results and discussion
BTX detection in water
The absorption kinetics of benzene at 100 mg m3 in wa-
ter, obtained by following the dependence of absorbance
by the PDMS block on contact time, is shown in Fig. 1.
The absorbance increases rapidly initially and then increases
continuously but linearly after approximately 60 min for
benzene and after shorter delays for the other derivatives.
As a compromise between measurement time and signal
intensity, a contact time of 60 min was chosen for all mea-
surements.
Calibration curves for benzene and toluene are shown
in Fig. 2. Similar good linearity for the dependence of ab-
sorbance on concentration was also obtained for p- and
o-xylenes; determination coefficients were >0.98. Figure 3
illustrates the absorbance amplification effect occurring as
a result of the extraction and concentration of the analyte
molecules from the dilute water solution on to the adsor-
bent block. A photometric enhancement factor (PEF) of ap-
Fig. 1 Kinetics of absorption of benzene, from a 100 mg m3 solu-
tion in water, by a PDMS (OV1) sorbent block, as measured from
the increase in absorbance of the solid phase at 262 nm

Fig. 2 Calibration curves showing the dependence of absorbance on
the PDMS sorbent block on the concentration in water, for benzene
(a) and toluene (b)
proximately 80 was found for benzene. Larger PEF val-
ues, approximately 200 and 300, respectively, were found
for toluene and p- and o-xylenes, in agreement with pre-
vious results obtained by Wittkamp et al. [15] with a sim-
ilar phase but different experimental conditions.
Limits of detection were determined from the slopes of
the calibration curves and the lowest absorbance signal
which could be detected. This signal is usually estimated
as three times the mean peak-to-peak noise signal. Because
of the degradation of the optical properties of the PDMS
block during the adsorption process, the LOD was, in
practice, limited by the increase in the background signal
towards the long wavelength side of the spectrum, which
makes it difficult to measure the contribution to the total
absorbance of a weak spectrum. This is illustrated in Fig. 4
where the absorption spectrum of benzene is reproduced
Fig. 3 Comparison of absorbance measured with a 100 mg m3 so-
lution of benzene in water (1 cm optical pathlength) and in the
PDMS block (0.2 cm optical pathlength) after immersion in the so-
lution for 60 min. A photometric enhancement factor, for equiva-
lent optical pathlengths, of 80 was estimated
171
at a concentration equal to approximately one third of the
estimated LOD.
Depending on the optical quality of the phase, in prac-
tice the weakest absorbance of a spectrum that could be
measured above background varied from 3 to 5104 ab-
sorbance units. The limits of detection deduced from these
values are indicated in Table 1.
The LOD estimated in this work are lower than those
reported previously. Our LOD for benzene is, neverthe-
less, slightly higher than the concentration in tap water
tolerated by law. It is, however, not very far from the con-
centration limits in industrial effluents in Europe and very
close to the benzene concentration found in Perrier water
in 1990 this was accidentally contaminated at levels rang-
ing from 12 to 20 mg m3 (1220 ppb).
For the other compounds, the estimated LOD are much
lower than the recommended concentration limits and the
method seems to be suitable for monitoring these pollu-
tants in most waters, including tap water.
BTX detection in air
In contrast with the kinetics of the absorption from water,
upon absorption of benzene from contaminated air the ab-
Fig. 4 Absorption spectrum of a PDMS block after immersion for
60 min in a 50 mg m3 solution of benzene in water
172

Table 1 Limits of detection (LOD), in mg m3 (ppb), estimated for monitoring of BTX in water by absorption on an OV1-type PDMS
block (thickness 2 mm) after contact for 60 min, and comparison with other data
BTX LOD LOD, other work Tolerated maximum Concentration limits Maximum values found
(mg m3) (mg m3) concentrations in tap water in cokeries (Europe) in tap water (USA,
(mg m3) (mg m3) Europe) (mg m3)
Benzene 1520 >50a 97b 5 (EPA), 1 (EEC), 10 (OMS) 332 0.3500
Toluene 46 >50a 10b 1000 (EPA), 500 (OMS) 375 14500
o-Xylene 24 >50a 7.8b 500 (OMS) 425440
p-Xylene 24 >50a 5.5b 500 (OMS) 425440
aUVspectroscopy by dynamic derivation of the spectra (from Ref. [19])
bUV spectroscopy on PDMS (from Ref. [15])
EPA: Environmental Protection Agency; OMS: Organisation Mondiale de la Sant; EEC: Economic European Community

Fig. 5 Kinetics of absorption of benzene, from 80 mg m3 conta-

minated air contained in a glass flask, by a PDMS (OV1) sorbent
block, as measured from the increase in absorbance of the solid
phase at 262 nm
Fig. 7 Absorbance signals measured on a PDMS block exposed for
7 min to exhaust gas emitted, respectively, from cold and warm
automobile engines equipped with a catalytic exhaust pipe system.
The absorbance spectra of benzene, toluene, and xylenes are dis-
played as reference spectra

tonitrile vapor but there is no experimental confirmation

Fig. 6 Calibration curves showing the dependence of absorbance
of benzene, toluene, and ortho- and para-xylene by the PDMS block
on the concentration in contaminated air, after 60 min contact time.
The determination coefficients were always >0.98
sorbance reaches a plateau after approximately 30 min con-
tact time for a concentration of approximately 80 mg m3
(Fig. 5). For the methyl derivatives at the same concentra-
tion, this plateau is reached after longer times approxi-
mately 50 min. This different behavior compared with
that for water might be because of the presence of ace-
of this. The calibration curves shown in Fig. 6 were ob-
tained for a contact time of 60 min. As observed for ab-
sorption from water, the extraction efficiency is the largest
for xylenes and the smallest for benzene in agreement
with the partition coefficients kow [20, 21] for these com-
pounds.
As depicted above, the limits of detection were esti-
mated from the calibration curves, and the weakest ab-
sorbance that we could be measured above the back-
ground. They were: benzene, 36 mg m3; toluene,
100200 g m3; o-xylene, 3060 g m3; p-xylene,
3060 g m3. Benzene concentrations in industrial ambi-
ent air in Europe and USA are of the order of 550 mg m3,
so it seems the method is sufficiently sensitive for moni-
toring benzene and its methyl derivatives in industrial areas.
As an application test, we have exposed a PDMS block
to the exhaust gas emitted by a car engine equipped with
a regular catalytic exhaust pipe system and operating with
lead-free gasoline. The signal obtained during the first
6 min after ignition is shown in Fig. 7. The system was at
ambient temperature and a signal in the absorption wave-
length range of BTX is clearly apparent. No such signal

was observed after 20 min of warming up when the cat-
alytic system has attained its working temperature.
Conclusions
The results presented above show that a direct spectro-
scopic detection on PDMS might constitute a method for
the detection of benzenic contaminants in water and pol-
luted atmospheres.
The evaluation tests and the estimated detection limits
obtained for benzene, toluene and xylenes indicate that
the method is relatively simple and easy to operate and sen-
sitive enough for application to pollution monitoring in
water and in industrial air.
In this work the applicability of the method has essen-
tially been evaluated on the basis of the limits of detection
which could be attained. The precision and repeatability
of the method for quantitative analysis have not been pre-
cisely determined. These characteristics are highly depen-
dent on the optical properties of the solid phase which in
turn are sensitive to chemical homogeneity, mechanical
stability with time, and optical transparency. The PDMS
material used in this work is specifically designed for chro-
matography, and not for spectrometric purposes. Accord-
ingly it is not totally suitable for the application described
in this work, particularly because of the instability of its
transparency. Despite this, calibration curves with deter-
mination coefficients >0.98 were always obtained, indi-
cating that the method can reasonably be expected to be
sufficiently precise for quantitative evaluation. A substan-
tial reduction of the limits of detection and an improve-
ment in the precision of the measurement would be ex-
173
pected from use of specifically prepared PDMS with suit-
able conditioning and shaping for optical application.
References
1. Hennion M.-C (1999) J Chromatogr A 856:354
2. Lika I (2000) J Chromatogr A 885:316
3. Mortgat B (1997) Environnement & Technique No 168, 2022
4. Di Benedetto D, Breuil P, Marchand JC, Rousson D (1999)
Actes du 9ime Congrs International de Mtrologie. pp 245
248
5. Mays P (1999) Environnement & Technique No 187, 4145
6. Aragon P, Atienza J, Climent MD (2000) Crit Rev Anal Chem
30:121151
7. Stuetz R (2001) Int Lab 31:1820
8. Thomas O, Theralaz F, Cerd V, Constant D, Quevauvillier Ph
(1997) Trends Anal Chem 16:419424
9. Carr JW, Harris JM (1988) Anal Chem 60:698702
10. Vilchez JL, del Olmo M, Avidad R, Capitn-Vallvey LF
(1994) Analyst 119:12111214
11. Pozomiek EJ (1991) Anal Lett 24:19131921
12. Eastwood D, Dominguez ME, Lidberg RL, Poziomek EJ (1994)
Analusis 22:305310
13. Arruda AF, Campiglia AD (1999) Anal Chim Acta 386:271
280
14. Wittkamp BL, Tilotta DC (1995) Anal Chem 67:600605
15. Wittkamp BL, Hawthorne SB (1997) Anal Chem 69:11971203
16. Fornier de Violet P, Garrigues P, Tyteca M, Hardy M (1998)
Forum Labo 98, mars 1998, CNIT PARIS
17. Algarra M, Lamotte M, Fornier de Violet Ph, Hernandez M,
Garrigues Ph (2000) Polycyclic Arom Compounds 19:241251
18. Algarra M, Radin C, Fornier de Violet Ph, Lamotte M, Garrigues
Ph, Hardy M, Gillard R (2000) J Fluoresc 10:355359
19. Vogt F, Tacke M, Bohmer W (1999) Spectrosc Eur 11:1218
20. Schwarzenbach RP, Geshwend PM, Imboden DM (1993) Envi-
ronmental organic chemistry. Wiley and Sons, New York, p 619
21. Chiou RP, Schmedding DW (1982) Environ Sci Technol 16:
410
Anal Bioanal Chem (2002) 372 : 174180
DOI 10.1007/s00216-001-1169-7
O R I G I N A L PA P E R
Lyndon L. E. Salins Elizabeth S. Goldsmith
C. Mark Ensor Sylvia Daunert
A fluorescence-based sensing system for the environmental monitoring
of nickel using the nickel binding protein from Escherichia coli
Received: 14 September 2001 / Revised: 22 October 2001 / Accepted: 24 October 2001 / Published online: 13 December 2001
Springer-Verlag 2001
Abstract A sensing system for nickel based on the nickel
binding protein (NBP) from Escherichia coli is shown to
be feasible. The versatility of NBP was demonstrated by its
use in three different assay formats. When the NBP binds
nickel, it undergoes a conformational change that can be
used as the basis for an optical sensing system for nickel.
The NBP gene was overexpressed in E. coli and the pro-
tein purified in a single step using perfusion anion-ex-
change chromatography. A unique cysteine residue at po-
sition 15 in the NBP was labeled with the fluorophore,
N-[2-(1-maleimidyl)ethyl]-7-(diethylamino)coumarin-3-
carboxamide (MDCC). In a spectrofluorimetric assay, there
was a maximum of 65% quenching of the fluorescence
signal produced by NBPMDCC in the presence of
nickel. A response curve for nickel using NBPMDCC
revealed a detection limit of 8108 mol L1. NBPMDCC
was also used to develop assays in microtiter plate and
fiber optic bundle formats. Detection limits for nickel us-
ing these formats were also in the submicromolar range.
Selectivity studies conducted with other divalent metals,
including copper, cobalt, iron, cadmium, and manganese,
showed that fluorescence quenching for cobalt was simi-
lar in magnitude but with a detection limit more than
10-fold higher than for nickel. The quenching responses
were lower for the other metals, with detection limits at
least 10 to 100 times higher than for nickel. These results
suggest that fluorescently labeled NBP is potentially use-
ful in the development of a sensing system for nickel.
Keywords Fluorescence-based sensor Environmental
monitoring Nickel-binding protein
Introduction
Nickel is important in many industrial applications, e.g. in
the making of stainless steel and other metal alloys. When
L.L.E. Salins E.S. Goldsmith C.M. Ensor S. Daunert ()
Departments of Chemistry and Pharmaceutical Sciences,
University of Kentucky, Lexington, KY 405060055, USA
e-mail: daunert@pop.uky.edu
this metal enters the environment it becomes a potential
health hazard to man and animals. Nickel is known to be
toxic to both prokaryotic and eukaryotic organisms [1, 2].
Nickel particles present in the air can settle in the ground
or be removed from the air by rain. As a result, a signifi-
cant amount of the metal can find its way into ground or
surface waters. Nickel toxicity can result in adverse health
effects ranging from allergic dermatitis to lung and nasal
sinus cancers. Because of the hazard of environmental
nickel, the Environmental Protection Agency (EPA) has es-
tablished that children should not ingest drinking water con-
taining nickel at concentrations of more than 0.04 mg L1
(6.8107 mol L1) [3].
Currently, there are several analytical methods for the
detection of nickel. The Department of Natural Resources
reports detection limits for nickel to be 1.7107 mol L1
with ICP and 3.4108 mol L1 using ICPMS [4]. These
traditional methods, including ICPAE spectrometry [5],
are sensitive enough to detect the metal at levels where it
is toxic to man; however, for field studies, where on-site
measurements are necessary, these methods are not prac-
tical. Determination of nickel via differential-pulse ad-
sorptive stripping voltammetry at a hanging mercury drop
electrode is a new technique that is sensitive to mol L1
concentrations of the metal [6], but for purposes of
screening for concentrations of nickel at which it is toxic
to man (7107 mol L1), this method might not be a fea-
sible option. In designing a sensing system for nickel,
high sensitivity, rapid results, and adaptability of the sys-
tem to miniaturization and use on-site are of critical im-
portance.
Nickel is an important nutrient for Escherichia coli
(E. coli) and other microorganisms. It is an essential com-
ponent of a number of microbial enzymes, which make
possible such reactions as ureolysis, methane biogenesis,
acetogenesis, and hydrogen metabolism [7]. Because of
this, the bacterium has developed a sensitive recognition
and transport system for nickel that is responsible for
monitoring and regulating the concentration of nickel
within the cell. The protein components of the nickel-
transport system in E. coli are encoded by the nik operon.

The amino acid sequences of the five overlapping nik
gene products are homologous to the proteins of peptide-
binding protein-transport systems in other Gram-negative
bacteria [8]. These five proteins together constitute a sys-
tem that not only enables the transport of nickel atoms be-
tween the periplasm and the cytoplasm but also makes
nickel chemotaxis possible.
The nikA gene codes for the periplasmic nickel-bind-
ing protein (NBP), the nickel-binding component of the
system. The NBP of E. coli consists of 524 amino acids.
The first 22 amino acids, which make up the signal pep-
tide necessary for transport into the periplasmic space, are
cleaved off when the protein is transported from the cyto-
plasm to the periplasm [9]. The molecular weight of the
mature protein is 56 kDa [8]. Quenching of intrinsic tryp-
tophan fluorescence and equilibrium dialysis reveal that
NBP binds a single nickel ion per molecule of protein
with a dissociation constant (Kd) <0.1 mol L1. This Kd
value reflects high specificity for nickel, because other di-
valent metal ions (Co2+, Cu2+, Fe2+) bind at least a factor
ten less tightly [10]. A preliminary X-ray diffraction study
of NBP achieved a maximum resolution of 3.0 [9]. Al-
though the detailed crystal structure of NBP is not cur-
rently available, the general structure of periplasmic bind-
ing proteins is known to consist of two similarly folded
globular regions, each made up of a -pleated sheet with
two or three -helices lining each side [9, 11, 12, 13].
Three short peptide segments in the protein forming a
cleft between the globular domains connect the two do-
mains. The binding site for the nickel lies within this cleft
[14, 15, 16]. The protein undergoes a conformational
change when it binds nickel, with the globular domains
approaching one another and reducing the size of the
groove around the ligand in the center, excluding mole-
cules of water or other solvent from the binding site [17].
This conformational change might be expected to cause
an observable change in the fluorescence intensity of a
fluorophore conjugated to the protein at the appropriate
site. This change in fluorescence could be directly corre-
lated with the concentration of nickel present in the sys-
tem, enabling the development of a sensitive assay for
nickel.
There is only one cysteine residue in NBP, located at
position fifteen. This seems to be a potentially good site
for conjugation with a sulfhydro-specific fluorophore such
as N-[2-(1-maleimidyl)ethyl]-7-(diethylamino)coumarin-
3-carboxamide (MDCC). In this study NBP was conju-
gated with MDCC, and steady-state fluorescence experi-
ments were conducted to verify there was a difference be-
tween the fluorescence signals emitted by the labeled pro-
tein in the absence and presence of nickel. Selectivity
studies were performed to evaluate the response charac-
teristics of NBP to other divalent metals that might be po-
tential ligands. The development of a fiber-optic sensor
for nickel was also investigated. A sensor of this kind is a
useful tool in environmental analysis, because it could fa-
cilitate remote sensing of the metal in areas not easily ac-
cessible for testing. The work presented discusses the de-
velopment of a novel fluorescence-based system that is
175
sensitive and can potentially be miniaturized and used for
on-site testing of water samples.
Experimental
Bacterial strains and plasmids
Plasmid p8602, containing the nikA gene, and E. coli strain B834
were provided by Dr Long-Fei Wu, at the Laboratoire de Chimie
Bacterienne, Marseilles, France.
Materials
Luria-Bertani (LB) medium was purchased from GibcoBRL
(Gaithersburg, MD, USA). Tris buffer ([tris(hydroxymethyl)ami-
nomethane]) was obtained from VWR Scientific (S. Plainfield, NJ,
USA). Ampicillin and bovine serum albumin (BSA) were bought
from Sigma (St Louis, MO, USA). Ethylenediaminetetraacetic
acid (EDTA), dithiothreitol (DTT), and all organic and inorganic
salts were obtained from either Fisher Scientific (Fair Lawn, NJ,
USA), VWR Scientific (S. Plainfield, NJ, USA), or Sigma. NBP
expression was induced with isopropyl--D-thiogalactoside (IPTG),
purchased from GibcoBRL (Gaithersburg, MD, USA). The micro
bicinchoninic acid (BCA) protein assay reagent kit from Pierce
(Rockford, IL, USA) was used to determine the concentration of
purified NBP. The fluorophore, N-[2-(1-maleimidyl)ethyl]-7-(di-
ethylamino)coumarin-3-carboxamide (MDCC), which was conju-
gated to NBP, was synthesized in our laboratory by the method of
Corrie [18]. Size-exclusion chromatography on a Sephadex G-25
column (Sigma) was used to separate conjugated protein from
excess fluorophore. Fluoronunc Maxisorp 96-well white microtiter
plates were purchased from Fisher Scientific (Norcross, GA,
USA).
Apparatus
Bacterial colonies were grown on agar plates at 37 C in a Fisher
Scientific Incubator. Cell cultures were grown in an Orbital Shaker
from Forma Scientific (Marietta, OH, USA) and pelleted by use of
a Beckman (Palo Alto, CA, USA) J2-MI Centrifuge. Unpurified
protein fractions were filtered with a 0.2 m filter from Nalgene
(Rochester, NY, USA).
The BioCAD Sprint Perfusion Chromatography System from
PerSeptive Biosystems (Cambridge, MA, USA) was used for pro-
tein purification. Liquid protein samples were lyophilized using
the VirTis Bench Top 3 Freeze Dryer (Gardiner, NY, USA). Pro-
teins were dialyzed against the proper buffer using a 1214,000
MWCO Spectra/POR molecular porous membrane from Spectrum
Medical Industries (Los Angeles, CA, USA). Protein purity was
verified by sodium dodecyl sulfate polyacrylamide gel electro-
phoresis (SDSPAGE) using the PhastSystem from Pharmacia
Biotech (Uppsala, Sweden). Protein absorbances in the BCA assay
were determined with a diode array spectrophotometer (model
8453) from HewlettPackard (Palo Alto, CA, USA). Fluorescence
studies were performed on a Spex Industries (Edison, NJ, USA)
Fluorolog-2 spectrofluorimeter, equipped with a 450-Watt xenon
arc lamp. A Cytofluor series 4000 fluorescence multi-plate reader
from Applied Biosystems (Foster City, CA, USA) was used for the
microtiter plate assay.
A bifurcated fiber-optic bundle and a dialysis membrane from
Spectra/Por3, MW cut-off 3500 from Spectrum Medical Industries
(Los Angeles, CA, USA) were used.
Optical probe, and type of dialysis membrane used
Expression of nickel-binding protein (NBP)
To determine the time course of NBP production in E. coli, a 5-mL
LB-ampicillin (100 g mL1) culture of E. coli B834 containing
176
p8602 was grown overnight in a shaker at 37 C and 250 rpm. The
next day LB-ampicillin (100 g mL1, 50 mL) were inoculated
with the overnight culture and incubated in a shaker at 37 C and
250 rpm. When the absorbance at 600 nm reached 0.6, IPTG (final
concentration of 0.2 mmol L1) was added to the culture. Samples
(2 mL) were taken 0, 1, 2, 3, 4, and 20.5 h after addition of IPTG.
Cells from 1 mL of each sample were pelleted in a microcentrifuge
tube. The supernatant was poured off and the pellet resuspended in
the drop of medium remaining in the tube. Chloroform (10 L)
was added to the resuspended cells, which were then vortex-mixed
briefly and left at room temperature for 15 min. TE buffer
(10 mmol L1 Tris-HCl, pH 7.0, 1 mmol L1 EDTA; 100 L) was
then added to the cells and the contents were mixed, and centrifuged
at high speed for 5 min. The upper periplasmic phase was transferred
to a clean microcentrifuge tube. A volume of 20 L of each frac-
tion was run on a 12.5% SDSPAGE, which was then stained with
coomassie brilliant blue R-250 to verify the presence of the NBP.
For purification of NBP a single isolated colony of E. coli
B834 containing p8602 was cultivated overnight in 3 mL LB-
ampicillin (100 g mL1) at 37 C and 250 rpm. This culture was
used to inoculate 500 mL LB-ampicillin (100 g mL1) which was
incubated at 37 C and 250 rpm until the absorbance at 600 nm was
0.6. At that point, IPTG was added to 0.2 mmol L1 final concen-
tration. The culture was then left to grow for a further 3 h. The
cells were pelleted by centrifugation at 3000 rpm and 4 C for
30 min. The isolated cells were then stored at 4 C.
Purification of NBP
Periplasmic NBP was released from cells by the chloroform method
of Ames et al [19]. The cells were resuspended in 5 mL chloroform
and kept at room temperature for 15 min. This solution was com-
bined with 50 mL of 40 mmol L1 Tris-HCl, 5 mmol L1 EDTA,
0.5 mmol L1 phenylmethylsulfonylfluoride (PMSF), pH 7.0. The
solution was centrifuged at 9000 rpm and 4 C for 15 min. The su-
pernatant was saved and lyophilized overnight. The powder was
dissolved in 1 mL deionized water and dialyzed against three
changes of 500 mL 10 mmol L1 Tris-HCl, pH 8.0 (Buffer A) over
12 h. The crude sample was centrifuged twice, each time for
15 min, at 9000 rpm and 4 C, and filtered through a 0.2 m mem-
brane filter. The protein was purified by HQ (quarternized poly-
ethyleneimine) perfusion chromatography. The column (2 mL bed
volume) was equilibrated with Buffer A. The flow rate was set at
8 mL min1. The periplasmic extract (1 mL) was injected onto the
column and the column washed with 10 column volumes of Buffer
A. The protein was eluted with a salt gradient from 0 to 100%
NaCl in Buffer A over 50 column volumes. Protein elution was
monitored by UV absorbance at 280 nm. Fractions containing the
purified protein were pooled and dialyzed against three changes of
500 mL 10 mmol L1 Tris-HCl, pH 8.0 over 16 h. The samples
were loaded on a 12.5% polyacrylamide gel that was then devel-
oped by silver staining. The SDSPAGE was used to determine
the presence (band appearing at a molecular weight of 56 kDa) and
purity of the NBP. The purified NBP was concentrated by
lyophilization. The product was dissolved in 3 mL distilled deion-
ized water and dialyzed against four changes of 10 mmol L1 Tris-
HCl, pH 8.0. The protein concentration was determined by the
BCA assay using bovine serum albumin as a standard and the en-
hanced test tube protocol with a working protein concentration
range of 5250 g mL1 according to the suppliers instructions.
A BCA assay revealed the concentration of NBP to be 13 mol L1.
Conjugation of NBP with MDCC
Purified NBP (1 mL) was dialyzed against 10 mmol L1 Tris-HCl,
pH 8.0, containing 10 mmol L1 DTT to eliminate disulfide link-
ages and free up the sulfhydryl group for reaction with the fluo-
rophore. The sample was then dialyzed against three changes of
500 mL 10 mmol L1 Tris-HCl, pH 8.0 over 6 h to remove any ex-
cess DTT (to prevent conjugation of that molecule with the fluo-
rophore, MDCC). In the dark at room temperature, 2.8 L of
25 mmol L1 MDCC were added, drop by drop, to the NBP solu-
tion to ensure that the final MDCC concentration would be
65 mol L1 (five times that of the NBP). The reaction was left to
proceed for 4 h at room temperature, after which time the sample
was loaded onto a Sephadex G-25 size-exclusion column for sepa-
ration of the conjugated protein from excess fluorophore. The con-
jugate was eluted with 10 mmol L1 Tris-HCl, pH 8.0. Fractions
containing the labeled protein were pooled, lyophilized, dissolved
in 3 mL deionized water, and dialyzed twice against 10 mmol L1
Tris-HCl, pH 8.0, in the dark at 4 C. The labeled protein was
stored at 4 C in the dark to keep the protein from degrading and
the fluorophore from photobleaching. The NBPMDCC concen-
tration was determined by BCA analysis to be 11.5 mol L1.
Fluorescence studies of NBPMDCC with nickel
Standard fluorimetric conditions
The fluorescence spectra of NBPMDCC were obtained by use of
a Fluorolog-2 spectrofluorimeter (Spex Industries, Edison NJ,
USA) equipped with a 450-W xenon arc lamp. The monochroma-
tor slit widths were set at 2 mm, and an excitation wavelength of
425 nm and an emission monitoring wavelength of 469 nm were
used unless otherwise noted. Different dilutions of NBPMDCC
were prepared and read in quartz cuvettes using a final reaction
volume of 1.5 mL. Spectrophotometric readings were taken at
room temperature (RT). From the dilution curve (data not shown)
an NBPMDCC concentration of 2.3108 mol L1 in 10 mmol L1
Tris-HCl, pH 8.0, was selected and used unless otherwise noted.
This concentration was selected because the signal obtained was
well above the background signal.
Steady-state study of NBPMDCC binding with nickel
in the presence and absence of bovine serum albumin (BSA)
A preliminary steady-state study of NBPMDCC was conducted to
obtain information about whether the presence of BSA would pre-
vent adsorption of the labeled protein by the walls of the cuvette.
The original 11.5 mol L1 stock of conjugated protein was diluted
100-fold in 10 mmol L1 Tris-HCl, pH 8.0, and divided into two
samples. One did not contain any BSA whereas 0.01% BSA was
included in the other to prevent adsorption of the protein by the
walls of the cuvette. The fluorescence spectra of both nickel-free
and nickel-bound (1.6104 mol L1 NiCl2, final concentration)
NBPMDCC were determined by use of the spectrofluorimeter af-
ter being left to incubate on a 400 rpm shaker for 10 min at 4 C.
Incubation-time study of the labeled protein with nickel
To determine the effect of incubation time on the change in fluo-
rescence response of NBPMDCC to nickel, the conjugated pro-
tein was incubated with nickel for different lengths of time before
determining fluorescence. The conjugate was incubated with
25 L 0.01 mol L1 NiCl2 in triplicate on a 400 rpm shaker at 4 C
for different periods of time. A volume of 25 L 10 mmol L1 Ni
was used in a total volume of 1.5 mL, which is equivalent to
1.6104 mol L1 Ni. This amount of Ni is sufficient to yield mea-
surable quenching in the signal (~35%) and hence was selected for
the study. Average fluorescence quenching values were plotted
against the incubation times in min.
Response curve of NBPMDCC to nickel
Samples (25 L) of different concentrations of NiCl2 were added
to NBPMDCC in triplicate samples and then incubated on a
400 rpm shaker at room temperature for 15 min, and analyzed in
the fluorimeter. Average fluorescence quenching was plotted in re-
lation to the log of the concentration of nickel in the sample.
 177
Selectivity studies of NBPMDCC with other divalent metals 0h 1h 2h 3h 4h 20.5 h Marker

Copper (II), cobalt, iron (II), cadmium, and manganese used were
in the chloride form, because chlorine is common in many organ-
isms and biosystems. Samples (25 L) containing different con- 60 kDa
centrations of each metal were added to NBPMDCC and incu- 50 kDa
bated on a 400 rpm shaker at RT for 15 min and analyzed in the 40 kDa
fluorimeter. Average fluorescence quenching was plotted in rela-
tion to the log of the metal concentration in the sample. 30 kDa

Steady-state fluorescence of MDCC with nickel

Samples (25 L) of 0.01 mol L1 and 0.1 mol L1 NiCl2 were added
to 1.5 mL samples of MDCC in 10 mmol L1 Tris-HCl, pH 8. Sam-
ples were analyzed in the fluorimeter as described above, except
that the emission wavelength was 477 nm.

Fluorescence assay on a multi-well plate reader Fig. 1 Time study of the induction of nikA with IPTG. The lanes
of SDSPAGE correspond to induction times of 0, 1, 2, 3, 4, and
A CytoFluor series 4000 fluorescence multi-well plate reader from 20.5 h, respectively. The final lane contains protein molecular-
PerSeptive Biosystems (Framingham, MA, USA) was used with a weight markers
96-well format and sample volumes of 200 L. A 420 nm excita-
tion filter with a band-pass of 50 nm and an emission filter of
460 nm with a band-pass of 40 nm were used to obtain fluores- 1 2 3
cence data. NBPMDCC was incubated with different concentra-
tions of nickel for 15 min at RT.

Fiber optic immobilization

70 kDa
An NBPMDCC concentration of 2.3108 mol L1 in 10 mmol L1
Tris-HCl, pH 8.0, was entrapped behind a dialysis membrane
60 kDa
at the tip of a glass bifurcated fiber optic bundle from an Oriel
(Stratford, CT, USA) modular spectrophotometer [20]. Details of 50 kDa
the fiber optic arrangement are given elsewhere [20]. The excita-
tion source was a tungsten-filament lamp. The fluorescence was
monitored by use of a grating monochromator set at an excitation 40 kDa
of 425 nm and using a 480 nm band-pass emission filter. A re-
sponse curve was constructed by dipping the probe first in buffer
alone to obtain a baseline signal. Nickel was then added in small 30 kDa
increments to the buffer while the latter was being stirred. The la-
beled protein was incubated with each addition of nickel for a pe-
riod of 15 min, until the fluorescence signal reached a maximum,
before additional nickel was added.

Results

Time course of NBP production in E. coli
There is considerable production of the 56 kDa NBP even
1 h after induction, with maximum production at approxi-
mately 3 h (Fig. 1). The 4 h time point showed no greater
quantity of NBP production than the 3 h time point. At
20.5 h, little NBP is present. Therefore a 3 h induction
was used for large-scale purification of NBP.
Purification of NBP
NBP was purified from a periplasmic extract in a single
step using perfusion chromatography with a quarternized
polyethyleneimine (HQ) anion-exchange column. The re-
sulting protein was >95% pure, as demonstrated by
SDSPAGE with silver staining (Fig. 2).
Fig. 2 SDSPAGE with silver staining verifying the purity of
NBP. Lane 1, protein marker; lane 2, crude periplasmic extract;
lane 3, purified protein
Steady-state study of NBPMDCC binding
with nickel in the presence and absence
of bovine serum albumin (BSA)
A steady-state study using non-conjugated MDCC indi-
cated no change in fluorescence signal with exposure to
nickel, demonstrating that nickel by itself does not quench
the fluorescence of MDCC. The results of the incubation
time study of NBPMDCC indicated significantly less
fluorescence quenching when BSA was included in the
assay. BSA contains a cysteine residue with a sulfhydryl
group which is negatively charged at pH 8 and which is
known to sequester cations. BSA can reduce quenching
178
Fig. 3 Effect of the incubation time with nickel on the NBP
MDCC fluorescence signal
Fig. 4 Response curve for nickel based on NBPMDCC using a
steady-state fluorimeter for detection
by binding nickel and preventing it from binding to the
NBPMDCC. We therefore conducted further experi-
ments without BSA.
Incubation time study of the labeled protein with nickel
The incubation time study of NBPMDCC with nickel
indicated significant (>30%) quenching even with only
2 min exposure to the ligand and maximum quenching of
65% after 30 min (Fig. 3). The chosen incubation time for
further fluorescence experimentation was 15 min, because
this value yielded significant quenching within a reason-
able length of time for a potential sensing mechanism.
Response curve of NBPMDCC to nickel
A response curve of NBPMDCC to nickel showed quench-
ing of the fluorescence signal that increased with increas-
ing concentrations of nickel (Fig. 4). The detection limit
of the system was 8108 mol L1 for nickel (calculated as
signal to noise ratio (S/N) = 3). The detection limit was
calculated using the background signal and three times its
associated standard deviation, which is not seen in the
Fig. 4 because it is small and the data points obstruct it.
Using these values a detection limit of 8108 mol L1 was
obtained.
Selectivity studies of NBPMDCC
with other divalent metals
Of the metals tested, only for cobalt were fluorescence
quenching results similar in magnitude to those for nickel.
Comparison of the response curves shows, however, that
the detection limit for cobalt (Table 1) is more than an or-
der of magnitude higher than for nickel, indicating less
specific response of the system to cobalt. Selectivity stud-
ies with other divalent metals, e.g. copper, iron, cadmium,
and manganese, showed the quenching response of NBP
MDCC to these ligands was significantly less than to
nickel. Although the detection limits obtained for cobalt
and copper are identical, the magnitude of quenching ob-
tained with copper was significantly less than with cobalt,
thus yielding a less sensitive curve. Table 1 also verifies
that the detection limits of the system for these ligands is
clearly higher than that for nickel.
Fluorescence assay on a multi-well plate reader
The next step in the development of the sensing system
for nickel is the miniaturization of the assay and hastening
the pace at which the data can be collected. For that, we
employed a 96-well fluorescence microtiter plate reader.
Reducing assay volumes to the microliter range reduced
the sensitivity of the system by approximately a factor of
three. Figure 5 indicates that the system still experiences
Table 1 Selectivity of the
NBPMDCC sensing system Interferent Detection limit
to divalent metals Copper 1.6106 mol L1
Cobalt 1.6106 mol L1
Iron 9.0106 mol L1
Cadmium 1.6106 mol L1
Data were obtained by use of a Manganese No response
spectrofluorimeter

Fig. 5 Response plot for nickel using the NBPMDCC sensing
system on a fluorescence multi-well plate reader
Fig. 6 Response curve for nickel obtained with the immobilized
NBPMDCC using an Oriel fiber-optic system
approximately 50% quenching of the emitted signal with
a detection limit for nickel of 2.4107 mol L1 (calculated
as S/N = 3), which is well within the levels needed to test
for levels of nickel that are toxic to man.
Fiber optic immobilization
To test the feasibility of using this system in a biosensor
apparatus, NBPMDCC in 10 mmol L1 Tris-HCl, pH 8.0
179
was entrapped behind a dialysis membrane at the tip of a
bifurcated fiber optic bundle. A response curve for nickel
was constructed as shown in Fig. 6. The maximum quench-
ing of the fluorescence signal was 35%, which was lower
than the 65% quenching observed for solution studies
with the spectrofluorimeter using milliliter sample vol-
umes. The detection limit for nickel employing this fiber
optic set-up was determined to be 1106 mol L1 (calcu-
lated as S/N = 3).
Stability study
A stability study was conducted over a period of a month
to observe change in the fluorescence signal of NBP
MDCC in relation to the time since its original conjuga-
tion. The deterioration of the signal was not sufficient to
affect experimental results. Two response studies of NBP
MDCC with nickel, conducted two weeks apart, showed
no significant change in quenching value (data not shown).
Discussion
Because a detailed crystal structure of NBP is not cur-
rently available, the best choice of a conjugation site for
labeling with MDCC was not clear. The NBP from E. coli
has a single cysteine residue located at position 15. This
unique cysteine provided an opportunity for site-specific
conjugation with a fluorophore to produce a labeled pro-
tein with fluorescent properties that would respond to dif-
ferent concentrations of nickel. The labeling of a single,
specific site can be an advantage over multiple-site label-
ing, which often results in a less sensitive assay. Multiple-
site labeling of a protein can result in increased back-
ground fluorescence, owing to multiple fluorescent labels
that do not respond to changes in conformation of the pro-
tein upon binding of ligand, because they are attached at
locations that see little change in local environment. Al-
though we did not have the detailed crystal structure for
NBP, we did have access to the structure of the dipeptide
binding protein (DPP) of E. coli, which shares 27% amino
acid sequence homology with NBP, and has only 30 fewer
amino acid residues [13, 21]. Pinpointing the residue at
position 15 in DPP indicated that its location was in a re-
gion that should experience much environmental change
as a result of a conformational change in the protein. The
lone cysteine-15 of NBP was, therefore, conjugated with
the sulfhydro-specific fluorophore, MDCC. MDCC is an
environment-sensitive fluorophore with unrestricted mo-
tion the two-carbon spacer arm between the maleimide
and the coumarin moiety gives flexibility to the fluoro-
phore, preventing any interference with the conforma-
tional change of the protein or its interaction with the lig-
and.
The addition of nickel to NBPMDCC resulted in up
to 65% quenching of the fluorescent signal. The system
was demonstrated to have a detection limit of 8108 mol L1
for nickel; as far as we are aware this is among the
180
most sensitive systems yet developed for nickel. The rea-
son for such significant quenching might be the theoreti-
cal location of the conjugated fluorophore. It is postulated
that when NBP is in the ligand-free form, MDCC might
be buried within a hydrophobic pocket of the protein
structure. When, however, the globular domains of the
protein engulf nickel, the fluorophore should be exposed
to solvent molecules, resulting in a weaker signal, owing
to quenching. Previous studies conducted by us with the
phosphate binding protein (PBP) revealed such a mecha-
nism of action [21]. Given the similar structures and func-
tion of NBP and PBP we hypothesize that the mechanism
of action should be similar. To elucidate the exact mecha-
nism by which the change in conformation of NBP affects
the fluorescence emission of the fluorophore, time-re-
solved anisotropy measurements should be obtained.
When NBPMDCC was used in a fiber optic biosensor
apparatus, maximum quenching of the fluorescence signal
was only 35%, compared with the 65% quenching ob-
served in solution studies with the spectrofluorimeter. The
detection limit for nickel employing this fiber optic set-up
was determined to be 1106 mol L1 (calculated as S/N = 3).
Although this is slightly above the level considered toxic
to humans, this sensor could still be used in remote loca-
tions to screen waters that have the potential of being con-
taminated with toxic levels of nickel. Further optimization
of the sensor is being performed, by varying different ex-
perimental conditions, to increase its sensitivity.
Of the metals tested, only for cobalt was fluorescence
quenching similar in magnitude to that for nickel, al-
though with a detection limit more than ten times higher
than that for nickel. Studies of intrinsic tryptophan quench-
ing in NBP upon binding to a ligand also demonstrated a
strong response to cobalt [22]. The remaining metals all
showed less maximum quenching than nickel and detec-
tion limits were more than 10 to 100 times higher than for
nickel. The 10-fold or tighter binding of nickel compared
with cobalt or copper means that the concentrations of
cobalt and copper in a test solution must be equal to or
less than that of nickel for the measurement to give a reli-
able measurement for nickel. Even though the specificity
of NBPMDCC for nickel is somewhat less than desired,
this molecule might be useful in sensitive sensing systems
used for prescreening for nickel. It might also find utility
in biosensors for areas where nickel contamination has
been previously found and it is desirable to monitor regu-
larly for the presence of this toxic metal.
Conclusions
In conclusion, a sensitive method for the detection of nickel
has been developed on the basis of bacterial nickel-binding
protein. A detection limit for nickel of 8108 mol L1 was
obtained using this method. The system was used to mea-
sure nickel in different ways in a conventional static
spectrofluorimeter, in a microtiter plate format, and in a
fiber-optic bundle format. The aim of this study was to
demonstrate that the system developed was versatile and
could be employed for in-situ measurements. The opti-
mization of the system for in-situ and remote sensing is a
part of future studies in our laboratory that involve minia-
turization of the set-up and the sensing system.
Acknowledgements We would like to thank the National Aero-
nautics and Space Administration for funding this research, and
the Kentucky Research Challenge Trust Fund for a fellowship pro-
vided to L.S. S.D. is a Cottrell Scholar and a Lilly Faculty
Awardee.
References
1. Babich H, Stotzky G (1983) Adv Appl Microbiol 29:195265
2. Brown SS, Sunderman FWJ (1988) Nickel toxicology. Acade-
mic Press, London, pp 14
3. US Department of Health and Human Services (1996) Toxico-
logical profile for nickel. US Department of Health and Human
Services, Public Health Service, Atlanta, GA
4. Environmental Protection Agency (1983) Methods for the
analysis of water and wastes. Environmental Protection Agency,
Washington DC
5. Molinero AL, Castillo JR (1998) Anal Lett 31:903911
6. Mrzljak RI, Bond AM, Cardwell TJ, Cattrall RW, Knight RW,
Newman OMG, Champion BR (1994) Analyst 119:1057
7. Hausinger RP (1987) Microbiol Rev 51:2242
8. Navarro C, Wu L-F, Mandrand-Berthelot M-A (1993) Mol
Microbiol 9:11811191
9. Charon M-H, Wu L-F, Piras C, de Pina K, Mandrand-Berthelot
M-A, Fontecilla-Camps JCJ (1994) Mol Biol 243:353355
10. de Pina K, Navarro C, McWalter L, Boxer DH, Price NC,
Kelly SM, Mandrand-Berthelot M-A, Wu L-F (1995) Eur J
Biochem 227:857865
11. Hirshberg M, Henrick K, Haire LL, Vasisht N, Brune M,
Corrie JET, Webb MR (1998) Biochemistry 37:1038110385
12. Brune M, Hunter JL, Howell SA, Martin SR, Hazlett TL,
Corrie JET, Webb MR (1998) Biochemistry 37:1037010380
13. Hsiao C-D, Sun Y-J, Rose J, Wang B-C (1996) J Mol Biol 262:
225242
14. Nickitenko AV, Trakhanov S, Quiocho FA (1995) Biochem-
istry 34:1658516595
15. Mowbray S L, Cole LB (1992) J Mol Biol 225:155157
16. Quiocho FA (1990) Phil Trans Roy Soc Ser B 326:341351
17. Spurlino JC, Lu GY, Quiocho FA (1991) J Biol Chem 266:
52025219
18. Adams MD, Oxender DL (1989) J Biol Chem 264:15739
15742
19. Corrie JET (1990) J Chem Soc Perkin Trans 1 21522152
20. Ames GF-L, Prody C, Kustu S (1984) J Bacteriol 160:1181
1183
21. Freeman MK, Bachas LG (1990) Anal Chim Acta 241:119
125
22. Lundgren JS, Salins LLE, Kaneva I, Daunert S (1999) Anal
Chem 71:589595
Anal Bioanal Chem (2002) 372 : 181186
DOI 10.1007/s00216-001-1199-1
O R I G I N A L PA P E R
Masanobu Mori Wenzhi Hu Paul R. Haddad
James S. Fritz Kazuhiko Tanaka Hirohito Tsue
Shunitz Tanaka
Capillary electrophoresis using high ionic strength background
electrolytes containing zwitterionic and non-ionic surfactants and its
application to direct determination of bromide and nitrate in seawater
Received: 12 September 2001 / Revised: 4 October 2001 / Accepted: 19 October 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract In capillary electrophoresis, it is commonly
considered that even a moderately high ionic concentra-
tion in the background electrolyte (BGE) leads to high
currents, resulting in Joule heating and serious peak dis-
tortion. As a new approach to overcome this problem,
zwitterionic (Zwittergent-314) and/or non-ionic (Tween
20) surfactants have been added to BGEs containing high
salt concentrations (e.g. 0.3 M NaCl) and have been
shown to result in acceptable separation currents (<200 A).
In turn, these BGEs could be applied to the separation
of samples containing high salt concentrations (such as
undiluted seawater) without the occurrence of any sig-
nificant peak broadening due to electrodispersion of the
sample. For example, a BGE comprising 10 mM Zwitter-
gent-314, 50 mM Tween 20, 0.3 M NaCl and 5 mM
phosphate (ph 7) could be used for the determination of
UV-absorbing anions in seawater, giving good peak
M. Mori () K. Tanaka
National Institute of Advanced Industrial Science
and Technology at Sero, Sero 489-0884, Japan
e-mail: masanobu-mori@aist.go.jp
H. Tsue S. Tanaka
Division of Material Science,
Graduate School of Environmental Earth Science,
Hokkaido University, Sapporo 060-0810, Japan
W. Hu
Department of Chemistry, Faculty of Science,
Hokkaido University, Sapporo 060-0810, Japan
P.R. Haddad
Australian Center for Research on Separation Science,
School of Chemistry, University of Tasmania,
GPO Box 252-75, Hobart 7001, Australia
J.S. Fritz
Department of Chemistry and Ames Laboratory,
Iowa State University, Ames, IA 50011, USA
Former address:
M. Mori
Division of Material Science,
Graduate School of Environmental Earth Science,
Hokkaido University, Sapporo 060-0810, Japan
shapes and detection limits of 0.8 M and 0.6 M for ni-
trate and bromide, respectively. The beneficial effects of
the non-ionic surfactant on the separation were attributed
largely to suppression of the electro-osmotic flow. On the
other hand, the zwitterionic surfactant was found to be ca-
pable of the incorporation of some anions in accordance
with the behaviour of these same surfactants in electrosta-
tic ion chromatography. This incorporation resulted in a
decreased conductivity of the BGE and also a change in
the separation selectivity of the system.
Keywords Bromide Nitrate Seawater Capillary
electrophoresis Surfactants
Introduction
Capillary electrophoresis (CE) offers several advantages
over high performance liquid chromatography (HPLC) in
that it is simple, gives high efficiency separations, uses a
small volume of electrolyte, and usually requires only a
short analysis time [1]. Most CE studies have been per-
formed on samples containing a relatively low total ion
concentration, with high salt samples, such as seawater,
receiving little attention. The main reason for this is that
these highly conducting samples cause electrodispersion
and broadening of the analyte peaks. Therefore, direct in-
jections of such samples are rarely used without some pre-
treatment such as dilution [2] or removal of salt in the
sample [3, 4]. High salt sample matrices also usually re-
sult in irreproducible injections and migration times [5].
One method to circumvent these problems has been to
raise the ionic strength of the background electrolyte
(BGE) such that electrostacking of the sample is restored
[6]. However, the addition of high salt concentration to
the BGE causes Joule heating and an increase in current,
leading to peak distortion and unstable baselines. Joule
heating can be minimised by lowering the applied volt-
age, increasing the capillary length, or reducing the inner
diameter of the capillary, but these approaches usually
lead to relatively long analysis times. It is therefore of in-
182

terest to develop methods, which permit the use of high

ionic strength BGEs but which do not also suffer from
problems of increased currents and Joule heating.
In this study, we have examined the use of several sur-
factants, especially zwitterionic surfactants, as additives
to the BGE in order to facilitate the separation of high salt
samples and to overcome some of the problems associated
with the use of the high salt BGEs. The zwitterionic sur-
factant was found to influence the resulting migration or-
der of the analyte anions by acting as a pseudo-stationary
phase and introducing a chromatographic component into
the separation. A particular advantage of zwitterionic
agents is that they can be used at high concentration with-
out affecting Joule heating.

Materials and methods

Apparatus and measurements. The CE was a HP3D capillary sys-
tem, which was obtained from Hewlett-Packard (Waldbronn, Ger-
many). A fused silica capillary (50 m i.d.) was obtained from GL
Science (Tokyo, Japan) and was used at total length 41 cm
(32.5 cm from inlet to detector). New capillaries were conditioned
by washing with methanol for 4 min, deionised and distilled water
for 3 min, 0.1 M NaOH for 10 min, water for 5 min and finally the
running buffer for 10 min. Prior to each run, the capillary was
rinsed with the running buffer for 5 min. Injection of the sample
solutions was performed by the pressure method from the negative
end of the capillary for 4 s for the model sample and for 20 s for
the seawater sample. A constant voltage of 15 kV was applied for Fig. 1 Electropherogram of five anions in a background elec-
the separation of sample anions, and separations were performed at trolyte (BGE) comprising 0.3 M NaCl and 5 mM phosphate (pH 7).
25 C. Conductivity measurements of the phosphate buffer con- 1 Br, 2 I, 3 NO2, 4 NO3, 5 MoO42. Capillary: fused silica,
taining salt and surfactants were measured using a TOA CM-8ET 41 cm total and 35.5 cm to detector (50 m i.d.); injection time,
Conductivity Meter (TOA Electronics, Tokyo, Japan) at 25 C. 4 s; voltage, 10 kV; current, 300 A; detector, UV at 215 nm
Reagents. The zwitterionic surfactant, N-tetradecyl-N,N-dimethyl-
3-ammonio-1-propanesulfonate (Zwittergent-314) was purchased
from Calbiochem (La Jolla, Calif., USA). Non-ionic surfactants concentration [6]. In this way, electrodispersion of the an-
[polyoxyethylene (23) lauryl ether (Brij 35) and polyoxyethylene alytes is minimised. However, as shown in Fig. 1, where
(20) sorbitan monolaurate (Tween 20)] and a cationic surfactant
[cetyltrimethylammonium chloride (CTAC)] were purchased from both the sample and the BGE contained 0.3 M NaCl,
Sigma (St. Louis, Mo., USA) and Aldrich (Milwaukee, Wis., USA), some analyte anions (e.g. nitrite and molybdate) give poor
respectively. The anionic surfactant sodium dodecylsulfate (SDS) peak shapes under the conditions applied. Even with a re-
was purchased from Wako Chemicals (Osaka, Japan). After addi- duced applied voltage, the current increased rapidly to
tion of the surfactants to 5 mM sodium phosphate buffer (pH 7),
the resulting solutions were filtered through 0.2 m cellulose ac-
300 A, leading to excessive Joule heating. Moreover, the
etate filter before use. Methanol was used as the electroosmotic analysis time was relatively long.
flow (EOF) marker.
Bromide (KBr), nitrate (KNO3), nitrite (KNO2), iodide (KI)
and molybdate (Na2MoO4) salts were purchased as analytical Effects of adding surfactants to high salt content BGEs
grade reagents from Wako Chemicals (Osaka, Japan), and the
model sample mixture comprised an aqueous solution containing
0.02 mM each of Br, NO3, NO2, I and 0.01 mM MoO42. Actual Figure 2 shows the influence on the separation current re-
seawater samples were taken from the surface of the Tokyo Bay sulting from the addition of various surfactants to a BGE
near the shore (Japan). A standard sample in an artificial seawater comprising a 5 mM phosphate buffer. In the cases of the
matrix [7] was obtained by dissolving KBr, KNO3, KNO2 and KI
to give concentrations of Br, NO3, NO2 and I of 2.0 M, re- anionic surfactant (SDS) and the cationic surfactant
spectively. (CTAC), the current increased with increasing concentra-
tion of surfactants. On the other hand, in a zwitterionic
surfactant (Zwittergent-314) or a non-ionic surfactant
Results and discussion (Brij 35 or Tween 20), the current remained constant or
decreased slightly over the concentration range examined.
Separation of samples of high ionic strength Figure 3 shows the separation currents resulting from the
addition of increasing concentrations of NaCl to 5 mM
The separation of inorganic anions under conditions where phosphate buffer containing 0.1 M of each surfactant.
the sample contains a high salt concentration is best ap- These currents were the average values observed over a
proached by using a BGE, which also contains a high salt measurement period of 10 min. In the absence of surfac-

Fig. 2 Changes in current with increasing concentration of several
surfactants added to 5 mM phosphate buffer at pH 7. These plots
were the average of currents measured over a 10 min period.
Cetyltrimethylammonium chloride (CTAC) (white square), sodium
dodecylsulfate (SDS) (black square), Zwittergent-314 (black cir-
cle), Brij 35 (black diamond) and Tween 20 (white circle). Capil-
lary length, 41 cm (32.5 cm to detector); voltage, 20 kV; tempera-
ture, 25 C
tant or in the presence of SDS and CTAC, the current in-
creased rapidly with increasing amounts of NaCl, reach-
ing 300 A at around 0.1 M NaCl. In contrast, in Brij 35,
Tween 20 and Zwittergent-314, this current was not
reached until about 0.5 M NaCl had been added to the
BGE. This effect of electrically neutral surfactants can re-
sult from changes of viscosity and permittivity of the
BGE, or the partition of salt into the neutral micelles in
the BGE.
Conductivity measurements for high salt content BGEs
with the various surfactants
Figure 4 indicates the variation of conductivity after addi-
tion of increasing concentrations of CTAC, Tween 20 and
Zwittergent-314 to a phosphate BGE containing 0.1 M
of different salts. The data are plotted as relative conduc-
tivity measurements. It can be seen that an increase in the
concentration of CTAC caused conductivity increase for
all salt types, whereas a decrease in conductivity was ob-
served for Tween 20 and Zwittergent-314. In the case of
Tween 20 (Fig. 4b) there was only a small difference in
the behaviour between BGEs containing different salts,
and the decreased conductivity can be attributed to in-
creased viscosity of the solution. Similar behaviour can be
expected for Brij 35 and any differences between Brij 35
and Tween 20, such as those shown in Figs. 2 and 3, seem
to be dependent on the molecular size of the surfactant [8,
183
Fig. 3 Changes in currents with increasing concentration of NaCl
added to BGEs containing 100 mM of various surfactants and
5 mM phosphate buffer (pH 7). No surfactant (white triangle),
CTAC (white square), SDS (black square), Zwittergent-314
(black circle), Brij 35 (black diamond) and Tween 20 (white cir-
cle). Other experimental conditions were as described in Fig. 2
9, 10]. On the other hand, both CTAC and Zwittergent-
314 showed different effects on conductivity in the pres-
ence of the various salts, especially when the counter-an-
ion was varied (Fig. 4a, c). However, precipitation was
observed above 50 mM CTAC (for 0.1 M NaNO3), 5 mM
CTAC (for 0.1 M NaI) and 2 mM CTAC (for 0.1 M
NaClO4), which imposed a restriction on the working
concentration ranges with BGEs containing these an-
ions. Zwitterionic surfactants did not show these effects
and therefore offered significant advantages. In the case
of Zwittergent 314, the decrease in conductivity
caused by addition of the surfactant depended on the
type of salt present in the BGE and followed the order
Na2SO4<NaCl<CaCl2<NaNO3<NaI<NaClO4. This order
corresponds to the order of retention of ion-pairs con-
taining these species in electrostatic ion chromatography
(EIC) using zwitterionic surfactants [11, 12].
Mechanism of uptake of ions
by the zwitterionic surfactant
The similarity between the above behaviour for the zwit-
terionic surfactant and that observed in EIC suggests that
the mechanism of uptake of ions by the surfactant might
be explained using retention models for EIC. Conven-
tional anion-exchange effects occurring at the quaternary
ammonium group on the zwitterion cannot explain the ob-
served results because the same behaviour would be ex-
pected with the cationic surfactant [13]. In EIC, cations
184

Fig. 4ac Plots of relative conductivities against increasing con-

centration of a CTAC, b Tween 20, and c Zwittergent-314 added
to 5 mM phosphate buffer containing 0.1 M of various salts. Tem-
perature, 25 C. Identities: NaCl (white square), CaCl2 (black cir-
cle), Na2SO4 (white circle), NaNO3 (white diamond), NaI (), and
NaClO4 (+)

and anions are retained by electrostatic interactions of

these species with the positively and negatively charged
groups of the zwitterionic surfactant. A complex retention
mechanism results, which involves repulsion effects (sim-
ilar to ion-exclusion chromatography) as well as electro-
static attraction effects [14, 15]. Analyte anions which are
polarisable (or chaotropic) show the strongest retention.
Thus, perchlorate and iodide can be expected to be incor-
porated into the zwitterion more strongly than other ions,
and divalent cations should assist this incorporation more
readily than monovalent cations, as observed in Fig. 4c for
Zwittergent-314.

The use of mixtures of Zwittergent-314

and Tween 20 for high salt samples

The above results suggest that single surfactants or mix-

tures of surfactants could be useful additives to the BGE
in a CE separation of anions in high salt matrices. Ex-
pected benefits would be the ability to use relatively high
concentrations of salt in the BGE (to aid electrostacking)
without concomitant high currents (and therefore Joule
heating) as well as some effects on the separation selec-
tivity. The mixed Tween 20 and Zwittergent-314 system
was applied to the determination of a standard anion con-
taining 0.3 M NaCl, with the separation of five anions Fig. 5ac Separation of five inorganic anions using a 60 mM
Tween 20, b 10 mM Zwittergent-314 and 50 mM Tween 20, and
(bromide, iodide, nitrite, nitrate and molybdate) being c 60 mM Zwittergent-314 added to 5 mM phosphate with 0.3 M
performed in a 5 mM phosphate buffer with 0.3 M NaCl NaCl. 1 Br, 2 I, 3 NO2, 4 NO3, 5 MoO42. Capillary length, 41 cm
at ph 7 with direct UV detection at 215 nm. There are sev- (32.5 cm to detector); voltage, 15 kV; current: 149 (a), 196 (b),
eral advantages to be gained by enhancing the effects of and 241 A (c); injection time, 4 s
the zwitterionic surfactant by addition of the non-ionic
surfactant. First, the velocity of EOF is reduced by adding
Tween 20. Second, the current in this system can be easily should not interfere with any anion partitioning into the
controlled (see Fig. 2 and Fig. 3). Third, even if a rela- zwitterionic surfactant since there is virtually no interac-
tively high concentration of Tween 20 is added to the tion between the anions and the non-ionic surfactant (see
BGE, precipitates do not form. Finally, the Tween 20 Fig. 4). Figure 5 shows electropherograms performed us-

Fig. 6a, b Electropherograms of a an actual seawater sample and
b the same sample spiked with 2 M of the analyte anions. BGE:
0.3 M NaCl, 10 mM Zwittergent-314, 50 mM Tween 20 and
5 mM phosphate (pH 7). 1 Br, 2 NO2, 3 NO3, 4 unknown, 5 I.
Voltage, 15 kV; current, 197 A; detector, UV at 215 nm; injec-
tion time, 20 s
ing Tween 20 and Zwittergent-314 alone, and also as a
mixture. In Fig. 5a, the separation was performed using
60 mM Tween 20, giving a current of 149 A and an analy-
sis time of less than 2 min as a result of increased effec-
tive velocities of the analyte anions due to the decreased
EOF. However, separation was poor and the baseline was
distorted. In Fig. 5b, 10 mM of Zwittergent-314 plus
50 mM Tween 20 was used, giving a current of 192 A
but also providing reproducible analysis time, resolution
and a stable baseline. The NaCl concentration added to
the sample could be increased to about 0.6 M without
detrimental effect. Finally, if Zwittergent-314 (60 mM)
alone was added to the BGE (Fig. 5c), the baseline resolu-
tion of all anions could be achieved but a long analysis
time resulted due to retention of anions on the zwitterionic
surfactant pseudo-stationary phase.
A further noteworthy behaviour evident from Fig. 5 is
the effect of the surfactants on the separation selectivity.
In Fig. 5a, the selectivity (as reflected by the migration or-
der) is the same as that for Fig. 1 and is the result of the in-
herent electrophoretic mobilities of the analyte ions.
185
There is no evidence of any interaction between the ana-
lyte anions and the non-ionic surfactant. However, in
Fig. 5b there is a change in migration order, especially for
iodide and nitrate, indicating that these analytes are inter-
acting with the zwitterionic surfactant. This interaction is
even stronger in Fig. 5c due to the increased concentration
of the zwitterionic surfactant, and is evident for all ana-
lytes except perhaps for molybdate.
Determination of anions in actual seawater
As a practical application, seawater samples were analysed
using the mixed surfactant system. Ion chromatography
[15, 16, 17] and electrochemical [18] methods have been
used to determine trace anions in seawater, but few stud-
ies have been done on direct seawater analysis by CE [6,
19, 20]. The difficulties in the CE method arise from elec-
trodispersion within the sample zone, and the fact that the
mobility of chloride is similar to those of anions of inter-
est, making separation difficult. Figure 6a shows the sep-
aration of undiluted seawater using a BGE containing
mixed surfactants and 0.3 M NaCl. A good bromide peak
was obtained at 1.2 min, followed by smaller nitrite and
nitrate peaks, but iodide and iodate could not be detected
with direct UV detection under the conditions used. Detec-
tion limits for bromide and nitrate were 0.8 and 0.6 M,
respectively, at 215 nm (S/N=3). Values of percentage rel-
ative standard deviation for three consecutive runs were in
the range 0.20.5% and 5.97.0% for migration time and
peak area, respectively. A standard addition method (Fig.
6b) was used to quantify the analytes and these measure-
ments indicated that the concentrations of bromide and ni-
trate were 56.5 and 1.63 M, respectively.
In conclusion, limiting the separation current is very
important in CE to avoid the generation of the Joule heat-
ing. It was shown that the addition of mixtures of zwitte-
rionic and non-ionic surfactants to the BGE permits the
use of high salt concentrations in the BGE. In turn, this
enables the direct analysis of samples with high ionic
strength, such as seawater. Results from conductivity
measurements indicated that the partitioning of anions
into the zwitterionic surfactant differed from that ob-
served for cationic or non-ionic surfactants and that this
partitioning behaviour was consistent with that observed
in electrostatic ion chromatography using hydrophobic
stationary phases coated with a zwitterionic surfactant.
Use of the zwitterionic surfactant Zwittergent 314 as a
mixture with the non-ionic surfactant Tween 20 allowed
the direct analysis of seawater with good peak shapes and
moderately low separation currents.
References
1. Fukushi K, Takeda S, Chayama K, Wakida S (1999) J Chro-
matogr A 834:349362
2. Tangen A, Lund W, Frederiksen RB (1997) J Chromatogr A
767:311317
186
3. Belgrader P, Devaney JM, Del Rio SA, Turner KA, Weaver
KR, Marino MA (1996) J Chromatogr B 683:109114
4. Clarke NJ, Tomlinson AJ, Schomburg G, Naylor S (1997) Anal
Chem 69:27862792
5. Ueda T, Maekawa T, Sadamitzu D, Oshita S, Ogino K, Naka-
mura K (1995) Electrophoresis 16:10021004
6. Ding W, Thornton MJ, Fritz JS (1998) Electrophoresis 19:
21332139
7. Lyman J, Fleming RH (1940) J Mar Res 3:134
8. Ajith S, Rakshit AK (1995) J Phys Chem 99:1477814783
9. Ghosh S, Moulik SP (1999) Indian J Chem Sect A 38:1016
10. Acharya KR, Bhattacharya SC, Moulik SP (1997) J Photochem
Photobiol A 109:2934
11. Hu W, Takeuchi T, Harguchi H (1993) Anal Chem 65:2204
2208
12. Hu W, Haddad PR, Hasebe K, Tanaka K (1999) Anal Commun
36:309312
13. Kaneta T, Tanaka S, Taga M, Yoshida H (1992) Anal Chem
64:798801
14. Hu W (1999) Langmiur 15:71687171
15. Hu W, Haddad PR, Hasebe K, Tanaka K, Tong P, Khoo C
(1999) Anal Chem 71:16171620
16. Ito K (1997) Anal Chem 69:36283632
17. Butler ECV, Gershey RM (1984) Anal Chim Acta 164:153
161
18. Nakayama E, Kimoto T, Okazaki S (1985) Anal Chem 57:
11571160
19. Fukushi K, Tada K, Takeda S, Wakida S, Yamane M, Higashi
K, Hiiro K (1999) J Chromatogr A 838:303311
20. Fukushi K, Watanabe K, Takeda S, Wakida S, Yamane M, Hi-
gashi K, Hiiro K (1998) J Chromatogr A 802:211217
Anal Bioanal Chem (2002) 372 : 187195
DOI 10.1007/s00216-001-1131-8
O R I G I N A L PA P E R
Beate Lilleengen Grethe Wibetoe
Graphite furnace atomic absorption spectrometry used for determination
of total, EDTA and acetic acid extractable chromium and cobalt in soils
Received: 28 May 2001 / Revised: 6 August 2001 / Accepted: 20 September 2001 / Published online: 13 December 2001
Springer-Verlag 2001
Abstract Methods for determination of both the total and
extractable content of Cr and Co in soil samples were in-
vestigated. For the total content of metal, ultrasonic slurry
sampling graphite furnace atomic absorption spectrome-
try was used and compared with conventional analyses af-
ter microwave digestion. The influence of grinding, leach-
ing and homogeneity for slurry sampling was also exam-
ined. The concentration of the elements in the analyzed
materials were in the range: 50 g g1 0.4% for Cr and
814 g g1 for Co. Relative standard deviations (slurry
sampling) were in the range 3%12% for Cr and 0.3%
6% for Co determinations. The detection limit and char-
acteristic mass (peak-area measurements) for Co were
0.14 g g1 and 12.6 pg, respectively. For Cr less sensitive
wavelengths were used. Excellent agreement with certi-
fied reference material was found for total Cr and Co us-
ing slurry sampling.
EDTA and acetic acid extractions were performed, us-
ing protocols given by the Measurement and Testing Pro-
gramme of the European Commission. The percentages
extracted for the different soil samples were 0.31.0 for
Cr and 2.524 for Co. To validate the accuracy of the ex-
tractable Cr, CRM 483 Sewage sludge amended soil was
analyzed. The values found were 37% and 32% higher than
the certified value for EDTA and acetic acid extractable
Cr, respectively. The precision for extractable concentra-
tion of Cr and Co was about 6% or less. External calibra-
tion with aqueous standards, matched to contain the same
reagents as the samples, was employed.
Keywords Chromium Cobalt EDTA/acedic acid
extraction Spectrometry, atomic absorption Graphite
furnace
B. Lilleengen G. Wibetoe ()
Department of Chemistry, University of Oslo,
P.O. Box 1033, 0315 Oslo, Norway
e-mail: grethe.wibetoe@kjemi.uio.no
Abbreviations GFAAS Graphite furnace atomic
absorption spectrometry ETAAS electrothermal atomic
absorption spectrometry CRM certified reference
material EDTA ethylenediaminetetraacetic acid
DTPA diethylenetriaminepentaacetic acid
Introduction
In the present work we focus both on methods for deter-
mination of total content of metal as well as on methods to
determine the fraction available for plant uptake, i.e. the
bioavailable part of the element. For the determination
of total Cr and Co in soil, samples were analyzed both as
slurries and after digestions using different digestion pro-
cedures. To obtain accurate results for decomposed soil
samples a total digestion is normally required, which usu-
ally includes use of hydrofluoric acid in addition to nitric
acid. For slurry sampling, other parameters are of impor-
tance for the accuracy and precision. The effect of some
of these parameters, i.e. grinding, homogeneity and per-
centage extracted into aqueous phase of the slurry was ex-
amined and discussed. Slurry sampling graphite furnace
atomic absorption spectrometry (GFAAS) has previously
been developed and used for determination of the total
content of various metals in soils and sediments, includ-
ing some work on Co [1, 2, 3] and Cr [1, 4, 5, 6, 7, 8].
Several extraction agents and a variety of procedures
have so far been used to determine the bioavailable frac-
tion of metal in soils, the extraction agents include acids,
chelating agents and buffered or unbuffered salt solutions.
Rauret [9] summarizes, in a recent paper, state-of-the-art
extraction procedures used for heavy metal determination
in contaminated soil and sediments. For Cr and Co, ethyl-
enediaminetetraacetic acid (EDTA) [10, 11, 12] or acetic
acid [10, 11, 13, 14] have most often been used as extrac-
tion agents.
To judge which extraction agent and procedure that most
likely will give the best estimate of the fraction available
for plant uptake is difficult and not a topic for the present
work. Moreover, at present there is no consensus about a
188

unique method that always gives the best estimate for the Table 1 Furnace programs used for the determination of Cr and
bioavailable fraction because the availability of metal to Co
plants depends on a large number of properties concern- Cr Co
ing the soil, plant and climate.
Fortunately, the problem with the different extraction Wavelengths (nm) 357.9, 429.0a, 425.4a 240.7
agents and procedures used has been dealt with by the Spectral bandpass (nm) 0.7 0.2
Measurement and Testing Programme of the European Injection volume (L) 10 20
Commission, which recently launched a project aiming at Step 1: Drying
harmonizing measurements for extractable trace metal Temperature (C) 90 90
content in soil. Protocols for determination of extractable Ramp time (s) 1 1
trace elements in soils have been described, and reference Hold time (s) 5 5
materials certified for their EDTA, diethylenetriaminepen- Step 2: Drying
taacetic acid (DTPA) and acetic acid extractable trace ele- Temperature (C) 110 110
ments have been prepared [15, 16, 17, 18]. Published re- Ramp time (s) 20 20
sults for the use of these procedures and CRMs are re- Hold time (s) 10 20
quested to see what to expect regarding the accuracy of Step 3: Pyrolysis
these operationally defined fractions. Temperature (C) 1200 1200 (1100b)
In the present work the protocols for EDTA and acetic Ramp time (s) 10 10
acid extraction were used to determine the concentration Hold time (s) 10 10
of metal to estimate the bioavailable part of Cr and Co in Step 4: Atomizationc
agricultural soil and sludge amended soil. The results for Temperature (C) 2300 2400 (1900b)
Cr were validated by analyzing CRM 483 Sewage sludge Ramp time (s) 0 0
amended soil that is certified for EDTA and acetic acid Hold time (s) 3 3 (2b)
extractable part for some trace elements including Cr (but Step 5: Clean out
not Co). Temperature (C) 2650 2650
Except for papers in connection with the development Ramp time (s) 1 1
of the CRM, no results for Cr and Co in CRM 483 Sewage Hold time (s) 4 3
sludge has, to the authors best knowledge, been published
When slurry sampling was used, the slurries were homogenized by
so far. automatic agitation for 30 s just before injection.
aLess sensitive lines used for total content of Cr.
bUsed for analysis of EDTA and acetic acid extracts.
cWall atomization; peak-area measurements; gas stop.
Experimental
Apparatus

Determinations were performed using a PE 5000 Zeeman atomic
absorption spectrometer with an HGA 400 graphite furnace, an
AS40 autosampler with 0.5 mm (i.d.) tubing and Intensitron hol-
low cathode lamps (all from Perkin-Elmer, Norwalk CT, USA).
An ultrasonic processor, VT 50 T Vibra cell (Sonics and Materials,
Danbury, CT, USA) with a 3 mm (o.d.) titanium probe was incor-
porated into the autosampler to provide automatic homogenizing
of the slurry. Zeeman background corrector was used, and the an-
alytical signal corrected for background and the background signal
were recorded with the Perkin-Elmer 3600 data system and
Anadex DP 9500 B printer. Pyrolytically coated graphite tubes
without platform (Applied Optics AB, Mlnlycke, Sweden) were
used. Electrothermal programs used are shown in Table 1. The vials
used in the autosampler were commercial 2 mL polystyrene vials
(Sarstedt, Ski, Norway) and in some cases in-house-made 15 mL
borosilicate vials. For grinding of samples a MM 200 Mixer Mill
(Retsch, Haan, Germany) was used. Grinding cups (10 mL) and
grinding beads (12 mm diameter, one in each cup) were made of
zirconia.
Particle size measurements were made using a Coulter Multi-
sizer II (Coulter Corporations, Miami, FL, USA) with Sampling
Stand IIA and an orifice tube with 70 m aperture diameter, cov-
ering the particle diameter range 1.442 m. For other details of
particle measurements see reference [19].
Microwave-assisted decompositions were performed in 100 mL
PFA Teflon vessels using ETHOS 1600 Advanced Microwave
Station from Milestone (Sorisole, Italy) with an HPR-1000/10 S
rotor. An in-house-made end-over-end shaker (360o rotation, four
samples capacity) and a Jouan B4i centrifuge (Nantes, France)
were used for mixing and separation of suspensions, respectively.
Reagents and samples
Water (>15 M), purified by Elix S water purification system
from Millipore (Bedford, USA), was used throughout. All reagents
were of analytical grade. The acids used were 65% nitric acid, 48%
hydrofluoric acid, 37% hydrochloric acid and 100% acetic acid, all
of pro analysis quality from Merck (Darmstadt, Germany) plus
30% hydrogen peroxide from Fluka (Buchs, Switzerland). Element
stock standard solutions of 1000 mg L1 concentration (Teknolab,
Drbak, Norway) were used to prepare calibration standards. Other
chemicals used were EDTA, free acid (Fluka, Buchs, Switzerland),
about 25% ammonia solution (AnalaR sp. gr.) from BDH Labo-
ratory Supplies (Poole, England) and Triton X-100 (Sigma, St.
Louis, USA).
The following reference materials were used: CRM 07411 Chi-
nese Soil (Standard Materials of Soils Component, Harbin, China)
and CRM 483 Sewage sludge amended soil (IRMM, Geel, Bel-
gium).
Two real soil samples were also analyzed; an agriculture soil
(soil-1) and a sludge amended soil (soil-2), both collected in Nor-
way.
Procedures
All samples and reference materials were dried to constant weight
(110 C) and real soil samples were sieved, using a 2 mm diameter
sieve.
For the digestion of samples, 0.25 g aliquots of the material
were accurately weighed into the decomposition vessels. Two dif-

ferent mixtures of concentrated acids were used: (1) 5.0 mL nitric
acid plus 1.0 mL hydrogen peroxide and (2) 4.0 mL nitric acid,
1.0 mL hydrogen peroxide plus 2.0 mL hydrofluoric acid. The sam-
ples were digested in closed vessels (max. pressure set to 30 bar,
i.e. 300 kPa) in the microwave oven (program: linear increase
from ambient temperature to 150 C for 20 min and then 23 min at
210 C). After digestion and cooling, all samples were diluted to
100 mL. For determination of Cr further dilution was required.
For preparation of slurries, the samples and reference material
CRM 483 were ground in the mixer mill for a selected time with
the amplitude set at maximum. For studying the effect of parti-
cle size, different grinding times were used. For studying the effect
of acid on the percentage extractable into the aqueous part of the
slurries, samples ground for 30 min were used. For the final analy-
ses, the samples were, with some exceptions, ground for 10 min:
CRM 07411 (particle diameter <42 m) was analyzed as received
and soil-1 was ground for 60 min. Slurries were prepared by
weighing 24 mg material (0.01 mg accuracy) into the vials of the
autosampler before diluent (0.22 M HNO3) was added. For CRM
483 the diluent also contained 0.005% (v/v) Triton X-100. The
slurry concentration and diluents used for final analysis are given
in Table 2.
For particle size measurements, the slurries were made in the
same way as for analyses (see Table 2). After ultrasonic agitation
for 20 s (30% full effect), 200 L of the suspension was transferred
to the beaker of the Coulter Multisizer, containing 200 mL Isoton
II electrolyte solution, for particle measurements. For more infor-
mation on the particle size measurements see Refs. [19, 20].
For the determination of the percentage of metals extracted into
the liquid phase of the slurry, the elements were first determined in
the slurry and then in the supernatant after centrifuging.
For determination of the EDTA and acetic acid extractable con-
tent of Cr and Co, the preparation of reagents and extraction was
performed (with some minor exception) as described in the proto-
col given by the European Commission for CRM 483 [17]. The ex-
traction procedures were also performed with 5 and 25 times
downscaling. The extraction procedures were as follows (param-
eters for 25 times downscaling are given in parentheses):
1. EDTA extraction: 5 g (0.2 g) of soil sample was transferred to
an 250 mL (10 mL) extraction bottle to which 50 mL (2 mL) of
0.05 mol L1 EDTA was added. The obtained mixture was
shaken on an end-over-end shaker operating at 3010 rpm for
1 h in a room at 202 C.
2. Acetic acid extraction: 5 g (0.2 g) of soil sample was transferred
to a 250 mL (10 mL) extraction bottle to which 200 mL (8 mL)
of 0.43 mol L1 acetic acid was added. The obtained mixture
was shaken on an end-over-end shaker operating at 3010 rpm
for 16 h in a room at 202 C.
Table 2 Slurry concentration used for analyses
Sample Slurry concentration
Cr Co
(mg mL1) (mg mL1)
CRM 07411 1.32.7 1.32.7
CRM 483 0.0150.03b 1.32.7
Soil-1 1.32.7 1.32.7
Soil-2 0.030.05 1.32.7
The volume of diluent (0.22 M HNO3a) used was 1.5 mL except
for the determination of Cr in CRM 483 and soil-2, where 10 mL
diluent was used.
aThe CRM 483 also contained 0.005% (v/v) Triton X-100.
bThe slurry was prepared in two steps: 1.53 mg of sample was
transferred to the vial and 10 mL diluent added. After homoge-
nization by ultrasonic agitation, 1.0 mL of this slurry was trans-
ferred to a second 15 mL vial and 9.0 mL added diluent.
189
The extracts were immediately filtered or centrifuged. A 0.45 m
filter was used for filtering. Centrifuging was performed for 5 min
at 4000 rpm. For more details of the procedure see reference [17].
In the final analyses, 0.2 g and 1.0 g sample were used for CRM
and real samples, respectively.
After extraction (in some cases) the supernatant was washed
with small portions of the extraction agent, dried at 110 C and an-
alyzed by slurry sampling to determine the non-bioavailable frac-
tion of Co.
External calibration with aqueous standards was used for
quantitative analyses. The calibration graphs covered the range
10150 g L1 for Cr and 1040 g L1 for Co. The aqueous stan-
dards and blanks used for slurry sampling, decomposed samples
and extracts (EDTA and acetic acid) contained the same concen-
tration of acids and reagents as the samples analyzed. The graphite
furnace programs used for all analyses are given in Table 1. The
optimized furnace programs are based on pyrolyses and atomiza-
tion curves for standard solutions and slurries of the different ma-
terials. Because of the relatively high content of Cr in the samples,
the less sensitive wavelengths 429.0 nm and 425.4 nm were used
for the determination of total Cr.
Results and discussions
Determination of total content by slurry sampling
Effect of acid concentration on leaching, accuracy and
precision. Nitric acid of various concentrations is nor-
mally chosen as a diluent for slurry sampling GFAAS. In
some cases, hydrofluoric acid has been used as diluent
with the intention of mobilizing a larger fraction of ana-
lyte into the liquid phase, and, thus improving the accu-
racy and precision (see, e.g. Ref. [3]). However, strong
acid is not always needed for a large fraction to be ex-
tracted; in some cases nearly quantitative extraction is
found using diluted nitric acid or even pure water (e.g.
Refs. [20, 21]). On the other hand, in the slurry analysis a
high degree of extraction is usually not a condition to ob-
tain reliable results; good accuracy and precision have
been obtained with most of the analyte retained in the
solid phase of the slurry [22]. Thus, more work is needed
to learn more about these relations for different samples
and analytes.
In the present work the effect of various concentrations
of nitric acid on the percentage extracted and the accuracy
and precision of the final result was examined for deter-
mination of Cr and Co in the various soils. As can be seen
in Fig. 1, the percentage extracted increases with nitric
acid concentration specially in the range from 01.1 M
acid. For Cr the maximum extracted was in all cases less
than 24% for CRM 07411 and soil-1, but was nearly
quantitative for CRM 483 when 2.2 M nitric acid was
used as diluent. The reason for the high percentage leak-
age for CRM 483 Sewage sludge amended soil, found in
the present work, could be that the added sludge contains
more soluble forms of Cr than normally found in soils.
Data from previous work for extraction of Cr from soil
and sediments, in connection with slurry sampling, are
found in the literature. Klemm and Bombach [5] found
26% leakage from diluted nitric acid (pH 2) for polluted
river sediments; Dobrowolski [6] found 14%15% using
0.05% nitric acid and 24%47% leakage using 5% nitric
190
Fig. 1 Percentage of metal extracted (one standard deviation,
n=3) into the liquid phase of the slurry as a function of nitric acid
concentration for: a Cr, and b Co. Median particle diameter was
4.5, 5.7 and 8.7 m for CRM 07411, CRM 483 and soil-1, respec-
tively. The nitric acid concentration used for quantitative measure-
ments was 0.22 M (vertical line)
acid from various types of soils; and Mierzwa et al. [7]
experienced 40%46% and 30%38% leakage for soil
and sediments, respectively, using 4% (w/w) nitric acid.
Except for CRM 483, the percentage leakage of Co was
in all cases higher than for Cr. The maximum percentage
extracted was 75%, 53% and 47% for CRM 07411,
CRM 483 and soil-1, respectively. No data from extrac-
tion of Co from soil and sediments slurries using nitric
acid was found in the literature, but results using 50% (v/v)
hydrofluoric acid gave 75%95% leakage of Co from sed-
iments [3].
As seen in Fig. 2, in spite of the difference in percent-
age extracted, analyses of the slurries of different acid con-
centration did not seem to effect the accuracy and preci-
sion significantly. In the present case, good results were ob-
tained even when pure water was used as diluent. In spite
of that, 0.22 M nitric acid was chosen as a diluent for the
final analysis as observations show that diluted nitric acid
gives a better wetting of the particles, and probably a more
rugged method. When using 0.22 M nitric acid as diluent,
the percentage leakage for Cr was less than 15% for
CRM 07411 and soil-1 and around 50% for CRM 483,
while for Co, the percentage extracted was in the range
from about 40% to 70%.
Fig. 2 Concentration of metal found (mean value one standard
deviation, n=3) analyzing slurries of various nitric acid concentra-
tion for: a Cr, and b Co
Homogeneity. In spite of the very small amount of sam-
ples used for slurry sampling, sample heterogeneity is sel-
dom the limiting factor for reliable results. However, in
certain cases one or several analytes have heterogeneous
distributions and the desired accuracy and precision may
not be obtained for these elements. The sample may con-
tain rare particles with very high contents of the analyte,
so-called nuggets, as described by Kurfrst [23]. In the
present analyses some outliers appeared during the Cr
determinations and the presence of nuggets in some of the
samples was suspected. A homogeneity study was, there-
fore, performed for the different materials. Ten sample
replicates, i.e. different slurry preparations (three to six
measuring replicates of each) were measured for Cr in
the CRMs and soil-1. CRM 07411 was analyzed as re-
ceived (median particle diameter () 7.9 m), CRM 483
(=8.6 m) and soil-1 was ground for 10 min (=10.7 m)
and the latter also for 60 min (=5.5 m) before analyses.
The results in Fig. 3 show that CRM 07411 and soil-1
(ground for 10 min) probably contain nuggets. For Cr in

Fig. 3 Concentration of Cr in ten sample replicates (three to six
measurements of each slurry) for: a CRM 07411, b CRM 483,
c soil-1 (ground for 10 min), and d soil-1 (ground for 60 min).
Each bar represents a single replicate. Median particle diameter
was 7.9 m for CRM 07411, 8.6 m for CRM 483, 10.7 and
5.5 m for soil-1, ground for 10 min and 60 min, respectively. For
CRM 07411, the certified value for Cr (mean value one standard
deviation) is shown (horizontal lines)
CRM 483, no outliers were experienced, in spite of the
fact that the number of particles injected for this sample
was about two orders of magnitude less than for the other
two; i.e. the particle size distributions were similar for the
three samples but the slurries injected were 100 times more
dilute for CRM 483 (see Table 2). Cr in CRM 483 Sewage
sludge amended soil, which contains much higher con-
centration of Cr than the other samples, is probably more
evenly distributed than in the other two samples. The quan-
titative results for the determination of Cr with and with-
out the outliers are shown in Table 3. The precision is
greatly improved when the outliers are omitted, especially
for soil-1, but because the outliers are few compared to
191
the number of replicates, the average content of Cr is not
significantly lowered. In routine analyses, when only few
replicates are taken for analyses outliers may not be expe-
rienced even when the materials contain nuggets, but when
they occasionally appear they may contribute significantly
to the result. Thus, when nuggets are expected, several
replicates should be taken. Kurfrst [24] has published a
paper dealing with the statistical treatment of data from
electrothermal atomic absorption spectrometry (ETAAS)
analyses of heterogeneous solid samples.
Another way to deal with the problem is to reduce
the effect of nuggets by increasing the amount of sample
or decreasing the particle size by further grinding. In the
present case, the latter was most practical and soil-1 was
ground for 60 min before analyses. As seen in Fig. 3d and
Table 3, the outliers disappeared, and better precision and
higher concentration were obtained for Cr in soil-1 when
the mean particle diameter was reduced from 10.7 m to
5.5 m.
No problem with heterogeneity was experienced for the
Co determination. No outliers were seen and good preci-
sion was obtained for all the samples. The homogeneity
192

Table 3 Results for Cr (mean

value one standard deviation) Sample Grinding Median Concentration na RSD Concentration na RSD
with and without omitting the time particle (all replicates) (%) (outliers omitted) (%)
outliers (min) diameter (g g1)
(m)
a Total number of replicates:
sample replicates (different CRM 07411 7.9 65 15 49 22.5 626 47 9.1
preparations of slurries) and Soil-1 10 10.7 46 25 40 54 403.5 38 8.6
measure replicates (different Soil-1 60 5.5 56.4 3.8 40 6.7 No outliers
injections of same slurry).

for Co in CRM 07411 (analyzed as received) is shown in less sensitive wavelength, 0.4% Cr was determined with
Fig. 4. good precision (5% RSD).

Results for total content. The results for total content of

Cr and Co analyzed by slurry sampling and after diges- Extractable part of metal
tion, with two different mixtures of acids, are compared in
Fig. 5 and Fig. 6. (Soil-2 was only analyzed by slurry sam- The proposed procedures for EDTA extraction and acetic
pling.) The use of nitric acid plus hydrogen peroxide did acid extraction from the European Commission [17] were
not digest the samples completely, but the results were used for determination of Cr and Co in soil. The EDTA
maximum 25% lower than when hydrofluoric acid or
slurry sampling was used. The additional use of hydroflu-
oric acid digested all samples completely, except CRM
07411 where a small, non-dissolved remainder was ob-
served. The results for digested samples using hydrofluo-
ric acid and slurry sampling were very similar. Determi-
nation of Co and Cr in CRM 07411 using slurry sampling
also confirmed the accuracy of the method. The values
found were 61.82.1 g g1 for Cr (mean value1 stan-
dard deviation, n =17) and 11.60.6 g g1 for Co (n=18),
thus they are in excellent agreement with the certified val-
ues 59.65 g g1 and 11.61.4 g g1.
The detection limit for Co was 0.14 g g1 based on
3.3 times standard deviation of the blank and the use of
4 mg mL1 slurry concentration. When soil slurry concen-
trations larger than about 4 mg mL1 were used the signal
started to be deformed and then precision decreased. The
characteristic mass (peak-area measurements) for Co was
Fig. 5 Results for total content of Cr after microwave digestion of
12.6 pg. For Cr the upper limit of determination is proba- samples, with various acid mixtures, and slurry sampling. (Be aware
bly of more interest than the detection limit. By using a of a logarithmic scale on the y-axis)

Fig. 4 Concentration of Co in ten sample replicates (four measure-

ments of each slurry) for CRM 07411. Each bar represents a single
replicate. Median particle diameter was 7.9 m. The certified value
for Co (mean value one standard deviation) is shown (horizontal Fig. 6 Results for total content of Co after microwave digestion of
lines) samples, with various acid mixtures, and slurry sampling
 193

Table 4 Extractable Cr and

Co (mean value one standard EDTA extractable Cr EDTA extractable Co Acetic acid extractable Co
deviation; three extractions
with three measurements of (g g1) (% RSD) (g g1) (% RSD) (g g1) (% RSD)
each) in CRM 483 using cen- Centrifuged 38.80.7 1.8 1.790.05 2.7 2.080.10 4.8
trifuging and filtering in the
procedure Filtered 38.40.7 1.8 1.790.04 2.4 2.060.07 3.6

extraction protocol was used for two CRMs and two real extractions were performed. It could be of interest to de-
samples, while the acetic acid extraction was used only termine also this fraction, which could represent the non-
for CRM 483 Sewage sludge amended soil. The latter bioavailable content of the metal. Another reason to de-
CRM is certified both for the EDTA and acetic acid ex- termine this remainder was part of the validation of the
tractable content of Cd, Cr, Cu, Ni, Pb and Zn. The ex- method, to see if the amount extracted plus non-extracted
tractable fractions are operationally defined and it is, added up to the total amount. As Fig. 8 shows, the results
therefore, of outermost importance to follow the given for extractable and non-extractable added up to the same
procedure exactly, because small changes may effect the concentration as the total content, determined by slurry
final results. It is also important that the protocols for op- sampling, both for EDTA and acetic acid extraction.
erationally defined fractions are practical, rugged and eco-
nomical. For this purpose, we looked at the possibility to
do a few changes in the protocol without effecting the re-
sults; the possibilities to use centrifuging instead of the
recommended filtering and to scale down the procedure
were examined.
Our experience is that it is more practical to centrifuge
the soil mixture instead of filtering. Centrifuging is quicker
and the risk of contamination is less, especially when the
samples are centrifuged in the same bottles as used for the
extraction (as used in the present case). The results are
given in Table 4. No significant difference was found, and
centrifuging could preferably be used instead of filtering.
According to the given protocol, 5 g of sample should
be used for extractions. As the CRMs are very expensive,
it is advantageous to reduce analyzed sample amounts.
The procedures were, therefore, scaled down 25 times. The
results for the extractable part of Cr and Co in CRM 483,
using 5 g and 0.2 g sample, shown in Fig. 7, are not sig-
nificantly different (t-test, 95% confidence level).
The quantitative results for extractable (concentration
and percentage of total) and total metals are given in
Table 5 and Table 6. The percentage extractable Cr was
small for all the materials, only 0.3%1.0% of the total.
The percentage extractable Co was higher, 2.5%6.9% for
three of the samples, while for CRM 483 Sewage sludge
amended soil, 21.5% and 24.4% were extracted using
EDTA and acetic acid, respectively. The result for the ex-
tractable concentration of Cr in CRM 483 was higher than
the certified values both for EDTA and acetic acid extrac-
tion, i.e. 37% and 32% higher, respectively. The same de-
gree of accuracy should not be expected for the opera-
tionally defined fractions of metal compared to the total
content, as only small differences in the performance of
the procedure, between laboratories, are inevitable.

Determination of the non-extracted remaining Co

using slurry sampling
Slurry sampling was used to determine the concentration
of Co retained in the samples after EDTA or acetic acid
Fig. 7 EDTA and acetic acid extractable metal (mean value one
standard deviation, based on dry mass) in CRM 483 (mean value
one standard deviation) using 5 g (standard procedure) and 0.2 g
of material (25 times downscaling) for: a Cr, and b Co
194

Table 5 Extractable (concen-

tration and percentage) and to- Extractable RSDe na Total RSDt Extractable RSDc
tal content of Cr in soils (mean concentration (%) (g g1) (%) percentageb (%)
value one standard deviation) (g g1) (%)
EDTA extract
CRM 07411 0.5600.009 1.5 3 61.82.1 3.3 0.910.03 3.7
The certified value with 95% CRM 483 39.31.8 4.7 19 3925195 5.0 1.000.07 6.8
confidence level is given in (28.62.6)a
bold types in brackets. All re-
sults are based on dry mass. Soil-1 0.330.02 5.7 3 56.43.8 11.6 0.580.10 8.9
aNumber of extractions (three Soil-2 0.360.02 4.3 3 111.010.1 9.1 0.330.03 10.0
measurements of each). Acetic acid extract
b(Extractable concentration/
CRM 483 24.71.5 6.3 10 3925195 5.0 0.660.06 8.7
total) 100%. (18.71.0)a
cRSD2=RSD 2+RSD 2.
e t

Table 6 Extractable (concen-

tration and percentage) and to- Extractable RSDe na Total RSDt Extractable RSDc
tal content of Co in soils (mean concentration (%) (g g1) (%) percentageb (%)
value one standard deviation) (g g1) (%)
EDTA extract
CRM 07411 0.590.02 3.3 3 11.60.6 5.0 5.10.3 6.0
All results are based on dry CRM 483 1.760.07 3.8 9 8.20.5 5.7 21.51.5 6.9
mass.
aNumber of extractions (three Soil-1 0.590.01 2.0 3 8.50.4 4.7 6.90.3 5.1
measurements of each). Soil-2 0.350.02 4.6 3 13.60.05 0.3 2.50.1 4.6
b(Extractable concentration/
Acetic acid extract
total)100%. CRM 483 2.0 0.1 5.7 7 8.20.5 5.7 24.42.0 8.0
cRSD2=RSD 2+RSD 2.
e t

using simple aqueous standards is possible. The concen-

Fig. 8 Results for concentration of Co in soil-1 for: a EDTA ex-
tracted plus non-extracted, b acetic acid-extracted plus non-ex-
tracted, and c total. Determinations of non-extracted and total Co
were performed using slurry sampling
Conclusions
Slurry sampling GFAAS is well suited for the determina-
tion of the total content of Co and Cr in soil samples. The
method is simple, rapid and rugged. External calibration
tration of nitric acid used as diluent is not critical, and ac-
curate results are obtained even when only 24 mg of sam-
ple is used for sample preparation and only a small frac-
tion of the metal is leached into the liquid phase of the
slurry. In the cases where nuggets were observed for Cr,
sufficient grinding of the soil samples could be used to
improve the precision. Digestion with nitric acid (plus hy-
drogen peroxide) resulted in a recovery minimum of 75%,
which may be accurate enough for some environmental
monitoring purposes. However, to obtain more accurate
results for decomposed samples, hydrofluoric acid has to
be used. But, even with the additional use of hydrofluoric
acid, a complete digestion of sample could not always be
ensured.
The precision of the EDTA and acetic acid extractable
metal was good, about 6% or better for Cr and Co. The
values found for the extractable Cr in CRM 483 Sewage
sludge amended soil were 37% and 32% higher than the
certified value for EDTA and acetic acid extractable Cr,
respectively. This accuracy could be considered satisfac-
tory, as the fractions are operationally defined. In addi-
tion, when information of the part available for plant is re-
quested, the operationally defined fraction will in any case
be a rough estimate of the bioavailable fraction. More
published results for the operationally defined fraction of
metal in the CRMs are needed to draw conclusions of the
optimal expected accuracy.

References
1. Hoenig M, Regnier P, Wollast R (1989) J Anal At Spectrom 4:
631
2. Mierzwa J, Sun YC, Yang MH (1997) Anal Chim Acta 355:
277
3. Lpez-Garca I, Snchez-Merlos M, Hernndez-Crdoba M
(1999) Mikrochim Acta 130:295
4. Carrondo MJT, Reboredo F, Ganho RMB, Oliveira JFS (1984)
Talanta 31: 561
5. Klemm W, Bombach G (1995) Fresenius J Anal Chem 353:12
6. Dobrowolski R (1996) Spectrochim Acta Part B 51:221
7. Mierzwa J, Sun YC, Yang MH (1998) Spectrochim Acta Part B
53:63
8. Akker AH van den, Heuvel H van den (1992) At Spectrosc
13:72
9. Rauret G (1998) Talanta 46:449
10. Sahuquillo A, Lpez-Snchez JF, Rubio R, Rauret G (1995)
Microchim Acta 119:251
11. Sahuquillo A, Bosch H, Rauret G, Muntau H (1998) Fresenius
J Anal Chem 360:304
12. Andreu V, Gimeno-Garca E (1996) Commun Soil Sci Plant
Anal 27:2633
195
13. Mitchell MC, Berrow ML, Shand CA (1987) J Anal At Spec-
trom 2:261
14. Houba VJG, Lexmond ThM, Novozamsky I, Lee JJ van der
(1996) Science Tot Environ 178: 21
15. Quevauviller Ph (1998) Trends Anal Chem 17:289
16. Quevauviller Ph (1998) Trends Anal Chem 17:632
17. Office for Official Publications of the European Communities,
Luxembourg (1997) The certification of the EDTA- and acetic
acid-extractable contents (mass fractions) of Cd, Cr, Cu, Ni, Pb
and Zn in Sewage sludge amended soils, CRMs 483 and 484.
Report EUR 17127 EN (ISSN 1018-5593), Office for Official
Publications of the European Communities, Luxembourg
18. Quevauviller Ph, Lachica M, Barahona E, Gomez A, Rauret G,
Ure A, Muntau H (1998) Fresenius J Anal Chem 360:505
19. Wibetoe G, Takuwa DT, Lund W, Sawula G (1999) Fresenius
J Anal Chem 363:46
20. Engelsen C, Wibetoe G (2000) Fresenius J Anal Chem 366:494
21. Miller-Ihli NJ (1990) Fresenius J Anal Chem 337:271
22. Miller-Ihli NJ (1997) J Anal At Spectrom 12:205
23. Kurfrst U (ed) (1998) Solid sample analysis. Springer Verlag,
Berlin
24. Kurfrst U (1991) Pure & Appl Chem 63:1205
Anal Bioanal Chem (2002) 372 : 196204
DOI 10.1007/s00216-001-1170-1
O R I G I N A L PA P E R
S. Thompson H. Budzinski K. LeMenach
M. Letellier P. Garrigues
Multi-residue analysis of polycyclic aromatic hydrocarbons,
polychlorobiphenyls, and organochlorine pesticides in marine sediments
Received: 20 August 2001 / Revised: 9 October 2001 / Accepted: 11 October 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract A multi-residue analysis procedure using micro-
wave-assisted extraction and pre-purification has been de-
veloped for the combined analysis of polycyclic aromatic
hydrocarbons (PAH), polychlorobiphenyls (PCB), and
organochlorine pesticides (OCP) in marine sediments. This
procedure has been validated with certified marine sedi-
ment. Several surrogate standards have been employed
and the use of octachloronaphthalene (OCN) as a surro-
gate standard for organochlorine determination in this ma-
trix is discussed. The recoveries of all compounds were
high (>70%) and the relative standard deviations are of
the same order as the certified values. Different analytical
problems are discussed, including DDT degradation in
gas chromatography and laboratory PCB background lev-
els. Quantification problems encountered for two pesticides
(cis-chlordane and trans-nonachlor) were attributed to PAH
interference in the GCECD chromatogram.
Keywords Aromatic compounds Organochlorine
compounds Multi-residue analysis
Introduction
Polycyclic aromatic hydrocarbons (PAH), polychlorinated
biphenyls (PCB), and organochlorine pesticides (OCP)
such as -hexachlorocyclohexane (HCH) and DDT, are
three classes of persistent and ubiquitous organic contam-
inants. These compounds have similar physicochemical
properties such as hydrophobicity, low degradability, and
high affinity for lipids. Because of these properties, these
contaminants are readily accumulated by living organisms.
They are often found together in environmental samples,
because they are mostly generated by anthropogenic ac-
S. Thompson H. Budzinski K. LeMenach M. Letellier
P. Garrigues ()
Laboratory of Environmental and Toxicological Chemistry,
UMR 5472 CNRS, University of Bordeaux,
33405 Talence Cedex, France
e-mail: p.garrigues@lptc.u-bordeaux.fr
tivity. Industrial activity and combustion of fossil fuels are
the main sources of PAH in the environment. PCB were
widely used in the electrical industry and as additives in a
variety of materials such as plastics, inks, and carbon
copy paper. OCP were used as insecticides, although most
have been withdrawn because of their environmental per-
sistence. Atmospheric or fluvial processes can transport
these contaminants into the marine environment, where they
can be adsorbed on to particulate matter and sediment.
To evaluate the impact of pollution on living organ-
isms, a variety of biomarkers [1, 2] has been developed for
rapid screening of exposure to environmental contaminants.
These techniques analyse the enzyme activity (ethylre-
sorufine-O-deethylase, EROD, acetylcholinesterase AChE)
induced by the mechanism of metabolism of xenobiotics.
Such activity is not, however, compound-specific and syn-
ergetic or antagonistic effects can be observed. For exam-
ple, planar compounds such as PAH and certain PCB [3]
both induce EROD activity. To validate the use of biomark-
ers in environmental evaluation detailed chemical analy-
sis should be performed and there is, therefore, a need for
a single analytical procedure which enables the determi-
nation of several classes of contaminants.
Several environmental monitoring programmes (Na-
tional Status and Trends Program [4], Reseau National
dObservation [5]) specify the routine analysis of marine
sediments and many environmental studies evaluate a large
number of samples. To satisfy the demand for increased
productivity and faster analysis of these contaminants an
analytical procedure has been developed in this work which
enables the determination of the three classes from the
same purified extract of sediment.
Quantitative analysis of these compounds remains a
challenging task, because they often occur at very low
concentrations (ng g1) compared with endogenous mate-
rial, and several purification steps may be necessary to
obtain a chromatogram free from interference. Gas chro-
matographic separation of all of the PCB is currently im-
possible. As Larsen [6] pointed out, no single column is
capable of separating even the seven priority PCB (28, 52,
101, 118, 138, 153, and 180). Indeed, several congeners can
 197

coelute in gas chromatography and they cannot be distin-

guished by electron capture detection (ECD). Compounds
of other classes can also interfere, e.g. OCP, which also
give an ECD signal.
Sample-preparation procedures result in losses of the
compounds to be analysed. Surrogate standards that have
physicochemical properties similar to those of the studied
compounds will be subject to the same losses and thus en-
able accurate determination of the original concentrations.
Octachloronaphthalene (OCN) is often used as a surrogate
standard for PCB quantification [7, 8, 9], although it does
not have the same chemical structure as PCB. Also some
authors have reported the presence of polychlorinated naph-
thalenes (PCN), also called chlorowaxes, in environmen-
tal samples [10, 11]. Although used to a lesser extent than
PCB, PCN industrial formulations have been used since
the beginning of the century, and are also present in PCB
industrial mixtures. PCB congeners which are absent from
technical mixtures because of their thermodynamically
unstable substitution patterns seem more suitable as surro-
gate standards, because they are also absent from environ-
mental samples and their physicochemical properties are
similar to the other PCB congeners.
Some authors have recently published results from mea-
surements of PCB in indoor air [12, 13, 14, 15, 16], which
can be high because of the extensive use of PCB in the
past. Background levels of PCB in indoor air can be high
enough to interfere with measurements and even contam-
inate samples [14], and as such must necessarily be deter-
mined in analytical procedures. Blank measurements re-
veal levels of PCB, which can bias results.
Similarly, the analysis of DDT by gas chromatographic
methods can be a challenging task [17, 18]. DDT is a rel-
atively thermolabile compound and the high injector tem-
peratures used in gas chromatography can cause degrada- Fig. 1 The sample-preparation procedure
tion during the analysis. Our laboratory has studied this
problem and the chromatographic conditions have been
optimised to limit DDT degradation. were purified with sulfuric acid under a microwave field. A second
purification was performed on silica gel, after which the extract was
Quality-control procedures are important to ensure cor- submitted to a liquid chromatographic step to yield two fractions.
rect quantification, and the use of certified reference ma-
terials enables determination of the accuracy of an analyt-
ical procedure. The precision of a method can be deter- Sample preparation
mined by replicate analyses of a homogeneous material. The surrogate standards PCB 30, 103, 155, 198 and OCN, 4,4-DDT
The developed procedure has been validated with the ref- d8 and the perdeuterated PAH were added to the sediment before
erence material SRM 1941.a from the National Institute extraction. The sample (2 g mass of dry sediment) was extracted
of Standards and Technology, Gaithersburg, MD, USA by microwave-assisted extraction with dichloromethane (30 mL)
(NIST), a marine sediment. The recoveries and repeatabil- by use of the Maxidigest 350 Prolabo apparatus (Prolabo, Paris,
France), which operates with open cells under atmospheric pres-
ity for PAH, PCB, and pesticide analysis are reported. The sure. The extraction conditions were: 30% (w/w) moisture; time
results obtained for the PCB are compared with those cal- 10 min; power 30 W. These microwave-assisted extraction condi-
culated with OCN as a surrogate and the suitability of this tions were optimised previously [19, 20].
compound for PCB quantification is discussed. PCB blank After extraction, the sample was filtered and purified with sul-
furic acid (9 mol L1, 10 mL) under a microwave field (operating
levels are also discussed with regard to possible sources conditions: time 10 min; power 30 W) [21]. The organic and acid
of indoor contamination. phases were separated and the organic extract was neutralised with
de-ionised water, and dried with anhydrous sodium sulfate. The
extract was concentrated under a gentle flow of nitrogen, and puri-
fied a second time on a column (10 cm1 cm i.d.) containing 9 cm
Experimental activated silica on top of 1 cm activated copper for elimination of
elemental sulfur. The PAH, PCB, and OCP were eluted from this
The experimental procedure is summarised in Fig. 1. The samples column with 10: 90 dichloromethanepentane (15 mL). The ex-
were treated by microwave-assisted extraction and after filtration tract was concentrated under nitrogen and transferred to 0.5 mL
198
isooctane. The solution was further re-concentrated to 0.05 mL in
a conical injection vial.
Liquid chromatography
The extract was further purified by HPLC with an aminosilane col-
umn (NH2 5 m, 250 mm; 4.6 mm i.d.; Stagroma, Switzerland)
with pentane as mobile phase at a flow rate of 1 mL min1 and UV
detection at 254 nm. The sample was introduced into the sample
loop (100 L) by means of a single injection of 0.05 mL. Two
fractions were collected. The first was collected from the solvent
peak at a retention time of 3.5 min until the retention time corre-
sponding to PCB 128. This PCB was the last of the studied PCB to
elute. This first fraction contains the PCB and the least polar pesti-
cides. The second fraction was collected from the cut off point of
the first fraction until the last eluting PAH dibenz[a,h]anthracene
(retention time 22 min). The second fraction contains the PAH
from phenanthrene to dibenz[a,h]anthracene and the polar pesti-
cides. The HPLC column was calibrated for retention times by use
of a standard solution of PCB 128 and dibenz[a,h]anthracene.
Gas chromatography
Fraction 1, which contained organochlorine compounds only, was
analysed by GCECD and Fraction 2 was analysed by GCECD
for the remaining organochlorine compounds and by GCMS for
PAH determination.
GCECD analyses were performed on an HewlettPackard (HP;
Avondale, MA, USA) 5890 series II gas chromatograph equipped
with a 63Ni electron-capture detector, an automatic HP 7673 injec-
tor, and a 60 m0.25 mm i.d.0.25 m film thickness HP5-Trace
(5% phenyl methylsiloxane) capillary column (HewlettPackard).
The GC conditions were: splitless injection (1 L); purge off at
injection and on after 1.5 min; injector temperature, 280 C for
Fraction 1, 250 C for Fraction 2; detector temperature, 290 C; ini-
tial oven temperature, 60 C; held for 2 min, heated to 120 C at
6 min1 and held for 5 min, then heated to 280 C at 2 min1 and
held for 20 min, total time 117 min. Helium was used as the carrier
gas at a column head pressure of 140 kPa for Fraction 1 and
190 kPa for Fraction 2, and the ECD make-up gas was nitrogen at
60 mL min1. The anode purge gas used was helium at a flow rate
of 12 mL min1. The relative response factors of the different com-
pounds were determined by injecting a standard solution (SRM)
spiked with the same solution of surrogate standards as that used
for spiking the sediments. The response factors were determined
after each four samples. Blank injections of isooctane were per-
formed between each injection of a sample to ensure the cleanli-
ness of the injector.
Electron ionisation (70 eV) GCMS analyses were performed
on an HP 5890 series II GC equipped with an automatic HP 7673
injector and a 60 m0.25 mm i.d.0.25 m film thickness HP5
(5% phenyl methylsiloxane) capillary column (Hewlett-Packard).
The injector was maintained at 270 C. The temperature program
was: 50 C (2 min) to 290 C (20 min) at 5 min1. The carrier gas
was helium at a constant flow rate of 1 mL min1. The GC was
coupled to an HP 5972 mass selective detector. Single-ion-moni-
toring mode was employed and the signal was acquired on the
molecular ions at 1.4 scan s1 with the electron multiplier at 1500 V.
The ions acquired were: m/z=178, 188, 202, 212, 228, 240,
252, 264, 276, 278, 288. The interface temperature was 240 C.
The PAH studied ranged from tri-aromatic (phenanthrene, an-
thracene) to hexa-aromatic (benzo[ghi]perylene) and included:
phenanthrene; anthracene; fluoranthene; pyrene; benzo[a]anthra-
cene; chrysene; triphenylene; benzo[b]fluoranthene; benzo[j]fluo-
ranthene; benzo[k]fluoranthene; benzo[e]pyrene; benzo[a]pyrene;
perylene; indeno[1,2,3-cd]pyrene; benzo[ghi]perylene; dibenz[a,h]-
anthracene; dibenz[a,c]anthracene. Response factors were deter-
mined relative to surrogate standards by injecting SRM 2260
preparations of PAH. Some of the PAH co-eluted; for these the re-
sults are given as the sum of the compounds.
Chemicals and reagents
The compounds used as surrogate standards for PAH determina-
tion were perdeuterated PAH. P-d10, BaP-d12 and Bper-d12 were pur-
chased from Cambridge Isotope Laboratories (CIL, Cambridge, MA,
USA), Fluo-d12, and Chyr-d12 from MSD Isotopes (Division of Merck
Frost Canada, Montreal, Canada) and Py-d12 from NIST (Gaithers-
burg, MD, USA). The Standard Reference Material SRM 2260, a
standard solution of 24 aromatic hydrocarbons was provided by
the NIST.
The PCB were analysed as individual congeners and the stan-
dards were purchased from Promochem, (Promochem, Molsheim,
France) in solution at 99%+ purity. The surrogate standards PCB 30,
103, 155, and 198 and OCN were also obtained from Promochem
(Molsheim, France) as crystals of 99%+ purity. A certified solu-
tion of PCB, SRM 2262 PCB in Hexane was obtained from
NIST.
The surrogate standard for pesticide analysis was 4,4-DDT d8
obtained from CIL, (Cambridge, MA, USA). The Standard Refer-
ence Material SRM 2261, a standard solution of chlorinated pesti-
cides was provided by the NIST (Gaithersburg, MD, USA).
All glassware was rigorously cleaned with detergent, then by
overnight heating at 450 C. The anhydrous sodium sulfate and the
concentrated sulfuric acid was from Fluka Chemie (Buchs, Switzer-
land), and the silica gel (particle size 0.0630.2 mm) from Merck
(Darmstadt, Germany). The sodium sulfate and silica gel were pre-
extracted with dichloromethane (Prolabo, Fontenay sous Bois,
France, pesticide grade) in an ultrasonic bath, dried and were rinsed
with dichloromethane just before use. The silica gel was activated
by overnight storage in an oven at 140 C. The copper (40 mesh,
95.5% purity, Aldrich, Strasbourg, France) was activated with hy-
drochloric acid (1 mol L1), neutralised with water, and rinsed with
acetone and dichloromethane. Isooctane of spectroscopy grade
quality, and pentane and acetone of HPLC grade were from Schar-
lau, ICS (St. Mdard en Jalles, France).
Procedural blanks were conducted in different laboratories and
buildings with exactly the same material and glassware under the
same conditions as the samples but with no matrix added to the
solvent for the extraction. The solvent tests were conducted by
evaporating 24 mL solvent under a gentle flow of nitrogen.
Results and discussion
Surrogate standards for PCB determination
The principle of surrogate standard quantification is well
known and described elsewhere [22, 23]. The standards
are added to the sample matrix and are taken through the
entire analytical procedure extraction, clean-up, chro-
matography. This method assumes that the recoveries of
the surrogate standard and the analyte are equivalent and
enables estimation of handling efficiency. The choice of
surrogate standard thus depends on the similarity of its
physicochemical properties with those of the analyte. The
ideal surrogate standards must also be absent from the en-
vironment, and free from interference in the chro-
matogram.
For this study, we have investigated the use of several
surrogate standards for the determination of PCB PCB 30,
(2,4,6-trichlorobiphenyl), PCB 103, (2,2-,4,5-,6-penta-
chlorobiphenyl), PCB 155, (2,2-,4,4-,6,6-hexachloro-
biphenyl), and PCB 198, (2,2-,3,3-,4,5,5-,6-octachloro-
biphenyl). All of these have a thermodynamically un-
favourable 2,4,6 chloro-substitution pattern and have never
been identified in technical mixtures [24]. These standards
 199
Table 1 The organic compounds analysed, the surrogate standard
used for each, and the HPLC Fraction
Compound Internal Standard Fraction
Phenanthrene Phenanthrene d10 2
Anthracene Phenanthrene d10 2
Fluoranthene Fluoranthene d10 2
Pyrene Pyrene d12 2
Benzo[a]anthracene Chrysene d12 2
Chrysene Chrysene d12 2
Triphenylene Chrysene d12 2
Benzo[b]fluoranthene Benzo[b]fluoranthene d12 2
Benzo[k]fluoranthene Benzo[b]fluoranthene d12 2
Benzo[j]fluoranthene Benzo[b]fluoranthene d12 2
Benzo[e]pyrene Benzo[e]pyrene d12 2
Benzo[a]pyrene Benzo[a]pyrene d12 2
Perylene Benzo[a]pyrene d12 2
Fig. 2 Percentage error in the quantification of some PCB with
Indeno[1,2,3-cd]pyrene Benzo[g,h,i]perylene d12 2 different surrogate standards
Dibenz[a,h]anthracene Benzo[g,h,i]perylene d12 2
Dibenz[a,c]anthracene Benzo[g,h,i]perylene d12 2
Benzo[g,h,i]pyrene Benzo[g,h,i]perylene d12 2 the more chlorinated congeners. The less chlorinated con-
PCB 28 PCB 30 1 geners are more volatile than the higher chlorinated con-
PCB 44 PCB 103 1 geners, as seen by comparing Henrys Law constant, H,
PCB 52 PCB 103 1 for the different congeners [25] and will be more subject
PCB 66 PCB 155 1 to loss during the analytical procedure. Thus a surrogate
PCB 87 PCB 155 1 standard with similar physicochemical properties, such as
PCB 101 PCB 155 1 number of chlorine atoms or volatility, should be lost to a
PCB 105 PCB 198 1 similar extent. Similarly the other surrogate standards are
PCB 118 PCB 198 1 highly suitable for the higher chlorinated congeners. PCB
PCB 128 PCB 198 2 198 gives good results for a large range of PCB, from hexa-
PCB 138 PCB 198 1 chlorobiphenyl to decachlorobiphenyl. OCN also gives
PCB 170 PCB 198 1 reasonable values for the range of PCB.
PCB 180 PCB 198 1 As a consequence of these results PCB 30 was used to
PCB 206 PCB 198 1 quantify, in elution order, PCB 8 to PCB 28, PCB 103 to
PCB 209 PCB 198 1 quantify PCB 52 to PCB 44, PCB 155 to quantify PCB 66
Hexachlorobenzene (HCB) PCB 198 1 to PCB 77 and PCB 198 to quantify PCB 118 to PCB 209
2,4-DDE PCB 198 1 (Table 1). The use of several surrogate standards in a nat-
4,4-DDE PCB 198 1 ural sample also enables determination of whether there is
4,4-DDD 4,4-DDT d3 2
an interference problem with one of the standards. For ex-
4,4-DDT 4,4-DDT d3 2
ample, if there is a large difference between the value of
cis-Chlordane 4,4-DDT d3 2
an analyte calculated by use of two different surrogate
trans-Nanochlor 4,4-DDT d3 2
standards there might be a co-elution problem with one of
the standards.

represent the full range of PCB both in degree of chlorine

substitution and in the range of elution times in the gas
chromatogram. Octachloronaphthalene was also investi-
gated for comparison purposes. This study considered in-
dividual PCB, and the congeners studied are shown in
Table 1.
The PCB surrogate standards were used to quantify a
standard solution of PCB, SRM 2262 PCB in Hexane,
which had undergone an evaporation step only. This ex-
periment mimics volatilisation/adsorption that might oc-
cur during re-concentration of a sample. The dependence
of the error (%) of the quantified value of a few selected
PCB, representative of the full range of PCB studied, on
the surrogate standard used for their determination is shown
in Fig. 2. PCB 30 results in the lowest error for the tri-
chloro- and tetrachloro-congeners and a larger error for
OCN as surrogate standard for PCB determination
in SRM 1941.a
Quantification using OCN as surrogate standard for the
marine sediment SRM 1941.a gave very poor results. Re-
covery was approximately 200% for all the compounds,
which suggests either that the area of the chromatographic
peak for OCN is too large or that the concentration as-
signed is too low.
A sample of SRM 1941.a was qualitatively extracted
without addition of surrogate standards. The extract was
then spiked with the standards one by one and analysed by
GCECD to reveal any possible matrix interference in the
chromatogram. All the PCB congeners were free from in-
terference but OCN co-eluted with an unknown impurity
200
Fig. 3 (a) GCECD Chromatogram obtained from a standard so-
lution, showing OCN. (b) GCECD chromatogram obtained from
a marine sediment extract before spiking with OCN
which resulted in a large peak (Fig. 3). OCN cannot be
used as an surrogate standard for this matrix because its
use will result in systematic overestimation of the concen-
trations of the analytes (area of the standard too high).
Optimisation of gas chromatographic injector conditions
for the analysis of DDT
DDT is a thermolabile compound that can be degraded
during gas chromatographic analysis [17, 18]. For example,
the analysis of 4,4-DDT-d8, the surrogate standard used
for the quantification of chlorinated pesticides under the
conditions used for the analysis of PCB leads to the de-
tection of three peaks (Fig. 4). These peaks have retention
times corresponding to 4,4-DDE-d8, 4,4-DDD-d8 and
4,4-DDT-d8. The DDT compound represents only 60% of
the total area of the three peaks. The analysis was per-
formed at different injector temperatures to determine the
optimum conditions for the GC analysis of 4,4-DDT-d8.
Reduction of the injector temperature to 250 C increases
the area of the 4,4-DDT-d8 peak to 67%. The gas flow
through the injector can be increased so that the com-
Fig. 4 GCECD Chromatograms of 4,4-DDT d8 under different
chromatographic conditions. (a) Injector temperature=280 C; col-
umn head pressure=140 kPa. (b) Injector temperature=250 C; col-
umn head pressure=140 kPa. (c) Injector temperature=250 C; col-
umn head pressure=190 kPa
pound spends less time in the hot glass insert of the injec-
tor and is rapidly deposited in the cooler (oven at 60 C)
column head. Under these conditions (injector tempera-
ture=250 C, and column head pressure=190 kPa) the
degradation is minimal, 4,4-DDT-d8 represents 96% of
the total area. For this sample-preparation procedure the
degradation of DDT can be determined, because the
deuterated standard will also be degraded and will lead to
the corresponding deuterated degradation product peaks,
which are not present in the environmental sample.
PAH interference in GCECD chromatograms
The ECD is a highly sensitive detector that responds to
electronegative compounds; the signal is proportional to the
electron affinity of the analyte. the signal from organo-

Fig. 5 GCECD Chromatogram of SRM 2260 PAH in Toluene
chlorine compounds such as PCB and the chlorinated pes-
ticides depends on the degree of chlorination. On a nor-
malised scale, hydrocarbons have a relative sensitivity of
1, monochlorinated compounds 100 and polychlorinated
compounds 106 [26]. The electron affinities of condensed
aromatic compounds are, however, much greater than
those of aliphatic hydrocarbons, and the ECD has been
proposed as a selective detector for the PAH [27, 28], be-
cause there are large differences between the electron affini-
ties of the different isomers. For example, benzo[a]pyrene
has a relative response factor more than twice that of
benzo[e]pyrene. Indeed, this electron-accepting property
is thought to be responsible for their toxic behaviour [29].
A GCECD chromatogram obtained from an SRM
2260 standard solution of PAH is presented in Fig. 5. The
average individual compound concentration in this solu-
tion is 50 ng L1. The peaks of the PAH are not very in-
tense, but the areas are significant compared with those of
the peaks resulting from low concentrations of organo-
chlorine compounds. Comparison of the retention times of
the standards of the three classes of contaminant reveals
that several co-elutions are possible. Because the HPLC
step in the sample preparation separates the PCB from the
PAH, however, only the pesticides that remain in Fraction
2 are affected. Quantification problems were encountered
201
for two pesticides, cis-chlordane and trans-nonachlor, in
GCECD analysis. These two chlorinated pesticides can
co-elute with pyrene and deuterated pyrene. Other authors
have reported the interference of pyrene with chlorinated
pesticides in gas chromatographic analysis [30]. Although
PAH have low response factors compared with organo-
chlorine compounds, because the PAH can be present at
concentrations 100 times greater, their signal can be suffi-
cient to create interference.
Procedural blanks
Procedural blanks are a standard procedure in analytical
chemistry and are a necessary step for exact quantifica-
tion. The procedural blanks performed during these vali-
dation experiments revealed a background level of PCB.
There were no significant peaks in the pesticide and PAH
fraction. The results obtained are shown in Table 2 in ng/
sample and the overall standard deviation is low (<5%).
The chromatographic profile of these analyses is also con-
stant. There are few tri-chlorinated congeners that are typ-
ical of an air source of PCB. The distribution is dominated
by tetra- to hexachlorinated PCB, which suggests an indoor
source such as sealant [15] or capacitors [13]. PCB were
used in building material (insulators, electrical transformer
fluids) until the 1970s and, as other authors have demon-
strated [14], it is probable that the air of the laboratory is
202
Table 2 Results (ng/sample) obtained for procedural blanks of Chemistry) the construction of which dates from approxi-
PCB mately 1990. No research has been conducted on PCB in
PCB Laboratory RSD Chemistry school this building. All the material (glassware, silica gel, sodium
sulfate) was prepared in our laboratory but was washed
8 0.27 29.6 0.36 with dichloromethane just before use in the Chemistry
18 0.29 3.4 0.2 School. The results are shown in Table 2, the level of PCB
28 0.53 64.2 0.0 is approximately a third of that in laboratory in the Depart-
44 1.31 29.0 0.39 ment of Chemistry (11 ng/sample compared with 28 ng/
52 2.62 18.7 0.0 sample). Because the glassware and material was prepared
66 4.27 3.3 1.47
in the contaminated laboratory, it is not surprising traces
87 1.66 13.3 0.66
of PCB were found, but the difference is significant. The
101 3.41 20.8 1.51
chromatographic profile is different, however. Although
105 1.36 15.4 0.58
the penta- and hexachlorobiphenyls still dominate, the tri-
118 3.42 6.1 1.85
and tetra-chlorinated congeners were absent. This could
128 0.47 27.6 0.24
138 2.37 26.2 1.54
be explained by the stronger adsorption of heavier PCB
153 1.98 25.3 1.72
congeners (penta- and hexachlorobiphenyls) by glassware
170 0.16 18.7 0.07 compared with the lighter PCB.
180 0.29 33.3 0.25 These results demonstrate that PCB persist in indoor
187 0.15 20.0 0.09 environments even after about 15 years since they were
banned. Indoor sources seem to be the cause of the high lev-
Total 24.56 6.8 11.09 els of PCB, as shown by the distribution of congeners. This
contamination particularly affects buildings that date from
before the PCB ban (19301980), i.e. most public buildings.
contaminated. The laboratory is situated in the Department
of Chemistry building that was constructed in the early
1960s, the period during which PCB were extensively em- Validation of the analytical procedure
ployed. Procedural blanks were performed in other laborato-
ries in the same building, where no PCB research had been The reference material, SRM 1941.a Organics in Marine
performed. The results showed that background levels were Sediment was analysed to determine the precision of the
the same, with the same chromatographic profile, thus the procedure. Three separate extractions were performed ac-
contamination was not caused by the manipulation of PCB. cording to the sample-preparation procedure (Fig. 1), en-
It was suggested by Wallace et al. [12] that contamina- abling calculation of the standard deviation of the mea-
tion of procedural blanks occurs during clean-up and treat- surements. The recoveries were corrected for the values
ment of the sample. This is demonstrated by comparing obtained in the procedural blanks.
the values obtained for solvent, 0.14 ng mL1 for the sum
of 18 PCB, compared with 24.6 ng/sample for a proce-
dural blank (purification steps included), in which a max- PAH
imum of 30 mL solvent were used.
To verify these results a procedural blank was performed Sixteen PAH were found in Fraction 2; their concentra-
in another building on the University campus (School of tions and recoveries are given in Table 3. Some of the

Table 3 Recoveries and rela-

tive standard deviations for PAH SRM 1941.a RSD (%) Experimental Recovery RSD (%)
polycyclic aromatic hydrocar- (ng g1) (ng g1) (%)
bons in SRM 1941.a PHE 489 23 4.7 432146 88.3 33.8
AN 184 14 7.5 108 23 58.4 21.3
FLUO 981 78 8.0 847149 86.3 17.6
PYR 811 24 3.0 665114 82.1 18.2
BaA 427 25 5.9 299 51 70.0 17.1
CHRY+TRIPH 577 35 6.1 567 43 98.4 7.6
BbF+BkF+BjF 1442150 10.4 1160 56 80.4 4.8
BeP 553 59 10.7 462 41 83.6 8.9
BaP 628+ 52 8.3 463 21 73.7 4.5
PER 452+ 58 12.8 366 23 81.0 6.3
IP 501 72 14.4 572 30 114.3 5.2
BP 525+ 67 12.8 509 25 97.0 4.9
DaA 117+ 13 11.1 77 6 65.8 7.8
Total 7687650 8.5 6527728 85.0 11.2
 203

Table 4 Recoveries and rela-

tive standard deviations for PCB SRM 1941.a RSD (%) Experimental Recovery RSD (%)
polychlorinated biphenyls in (ng g1) (ng g1) (%)
SRM 1941.a 28a 9.80 3.70 37.8 10.561.35 107.8 12.8
44 4.80 0.62 12.9 4.821.43 100.5 29.9
52 6.89 0.56 8.1 8.560.82 124.3 9.3
66 6.80 1.40 20.6 6.520.62 95.9 9.5
87 6.70 0.37 5.5 3.550.34 53.0 9.6
101 11.00 1.60 14.5 11.730.35 106.7 3.0
105 3.65 0.27 7.4 3.600.05 98.6 1.4
118 10.00 1.10 11.0 10.640.51 106.4 4.8
128 1.87 0.32 17.1 1.600.19 85.6 11.9
138+163+164 13.38 0.97 7.2 11.010.79 82.3 7.2
153 17.60 1.90 10.8 16.020.36 91.0 2.2
170+190 3.00 0.46 15.3 4.320.09 143.9 2.1
180 5.83 0.58 9.9 8.680.42 148.9 4.8
206 3.67 0.87 23.7 4.300.41 117.3 9.5
209 8.34 0.49 5.9 10.180.63 122.0 6.2

aIndicative
Total 113.5315.21 13.4 116.098.36 102.4 7.2
concentration only

compounds are given as sums of two or three individual
compounds because of co-elution from the gas chromato-
graphic column (chrysene + triphenylene; benzo[b]fluor-
anthene + benzo[j]fluoranthene + benzo[k]fluoranthene;
dibenz[a,h]anthracene + dibenz[a,c]anthracene). Recov-
eries range from 58% to 114%, with an average of 85%
for all the compounds (Table 3). Relative standard devia-
tions are quite high for the lighter compounds (phenan-
threne, anthracene, fluoranthene, and pyrene) ranging from
17 to 34%, for phenanthrene. The heavier compounds, from
chrysene to benzo[g,h,i]perylene have relative standard
deviations equivalent to or lower than those given with
the certified values. It is possible that the acid purification
step leads to slight degradation of the smaller tri-aromatic
PAH, leading to the low recovery for anthracene. The
lowest recoveries were found for the PAH anthracene and
dibenzanthracenes. These compounds are also the least
abundant in the matrix with concentrations of only 117
and 184 ng g1. The recoveries of the PAH are, however,
comparable with those in other studies [17, 18], and the
results are reasonable for environmental analysis.
PCB
All the PCB are found in Fraction 1 except for PCB 128,
which is in Fraction 2. For most of the PCB, and using the
surrogate standards described, the recoveries are high (usu-
ally >80%); the results are presented in Table 4. The av-
erage recovery of all the PCB is 102.4%. Only for PCB 87
is the recovery low 53%; the relative standard deviation
of 10% is, however, not very high compared with those of
the other PCB congeners. The recovery of PCB 52 is high
124%; this can be explained by the close elution of a
large interfering peak which is not removed by the HPLC
purification step. High recoveries, approximately 120% to
150%, were obtained for PCB 170 to PCB 209. The peaks
of these congeners are well-resolved in the chromatogram
and co-eluting impurities do not seem to be present. Their
quantification by use of other surrogate standards does not
significantly change the result. The standard deviations
calculated on three measures are comparable with those of
the certified concentrations, except for PCB 44. The re-
sults are an improvement on previously published results
[21] for some congeners, because of the extra HPLC pu-
rification step. This removes the polar pesticides which
can interfere with the PCB, e.g. 4,4-DDT which co-elutes
with PCB 138.
Pesticides
Six chlorinated pesticides are certified in the marine sedi-
ment SRM 1941.a, and an indicative concentration is given
for 4,4-DDT. The recoveries and relative standard devia-
tions are shown in Table 5. The recoveries of hexachloro-
benzene, 4,4-DDE, 4,4-DDD and 4,4-DDT were good,
and ranged from 102% to 124%.
The relative standard deviations are quite high for this
class of compound, except for hexachlorobenzene. This is
probably because of the low concentrations of some of
these pesticides in this matrix. The average recovery for
the sum of the pesticides is 118.5%. The recovery of 2,4-
DDE is too high at 278%. This compound is not totally
separated from interference, and its low concentration in
this matrix, 0.73 ng g1, leads to a chromatographic peak
which is small relative to those of other closely eluting
compounds such as PCB 101, the concentration of which
is 11.0 ng g1. To improve the recovery of this compound
a larger amount of certified material should be extracted.
Recovery of cis-chlordane and trans-nonachlor from
this matrix are very high, 420% and 250%, respectively.
When analysed without addition of PAH surrogates re-
coveries of these compounds are much improved 130%
and 70%, respectively. cis-Chlordane and trans-nonachlor
are not entirely separated in the chromatogram and are co-
eluted with deuterated pyrene and pyrene. The certified
values of these compounds are determined by mass spec-
204

Table 5 Recoveries and rela-

tive standard deviations for Pesticide SRM 1941.a RSD (%) Experimental Recovery RSD (%)
pesticides in SRM 1941.a (ng g1) (ng g1) (%)
HCB 70 25 35.7 83.0 1.7 118.6 2.0
2,4-DDE 0.73 0.11 15.1 2.030.35 278.1 17.2
4,4-DDE 6.59 0.56 8.5 6.781.90 102.9 28.0
4,4-DDD 5.06 0.58 11.5 6.281.40 124.1 22.3
4,4-DDTa 1.25 0.10 8.0 1.280.40 102.4 31.3
cis-Chlordane 2.33 0.56 24.0 9.782.43 419.7 24.8
3.070.18b 131.8 5.9
aIndicativeconcentration only trans-Nanochlor 1.26 0.13 10.3 3.180.42 252.4 13.2
bConcentration determined 0.900.04b 71.4 4.4
without addition of PAH Total 87.2227.04 31.0 103.345.97b 118.5 5.8
deuterated standards

trometry, which can clearly distinguish between the com-
pounds as a result of their different molecular mass, m/z=
202, pyrene; 212, pyrene-d12; 373, cis-chlordane; 409, trans-
nonachlor. In this matrix pyrene is present at a concentra-
tion of 81124 ng g1, compared with 2.330.56 ng g1 for
cis-chlordane and 1.260.13 ng g1 for trans-nonachlor.
To avoid interference problems GCMS is the preferred
method for detection of these compounds, although the
higher detection limits of GCMS are not always practical
for environmental samples. The determination of these
pesticides with the developed sample preparation proce-
dure and analysis by GCECD remains a suitable tech-
nique which leads to good recoveries, with the condition
that deuterated pyrene is not used as an surrogate stan-
dard.
Conclusion
The analytical procedure combining microwave-assisted
extraction and pre-purification steps developed for the de-
termination of PAH, PCB, and chlorinated pesticides in
sediment matrices enables the precise determination of
most of these compounds. This procedure enables the de-
termination of the three classes of contaminant in the
same extract and is thus a timesaving method.
Some problems persist however for certain compounds
in these matrices, notably for the two pesticides cis-chlor-
dane and trans-nonachlor. The interference of PAH in the
GCECD chromatogram can lead to over-estimation of
organochlorine compounds. Although the response factors
of the PAH are low compared with those of chlorinated
compounds, the PAH are often present at concentrations
100 times greater. Elimination of deuterated pyrene as an
surrogate standard for PAH determination limits the inter-
ference and enables good recovery of these compounds.
The use of PCB congeners which are absent from en-
vironmental samples as surrogate standards improves the
precision of PCB quantification. Similarly the use of sev-
eral surrogate standards simultaneously enables the ana-
lyst to identify possible co-elutions with the standards, as
can be shown by the example of OCN for SRM 1941.a.
Care should be taken with the use of OCN as an surrogate
standard for PCB analysis the possible presence of in-
terference should always be investigated.
Acknowledgements Prolabo is acknowledged for the loan of the
microwave apparatus. NIST is thanked for the SRM samples.
References
1. Lafaurie M, Narbonne JF, Galgani F (1992) Analusis Magazine
20:6, 27
2. Garrigues Ph, Burth H, Wallcer CH, Narbonne JF (2001) Bio-
markers in Marine Organisms. Elsevier, Amsterdam, 500 p
3. Lang V (1992) J Chromatogr 595:1
4. Farrington JW, Goldberg ED, Risebrough RW, Martin JM,
Bowen VT (1983) Environ Sci Technol 17:490
5. Claisse D, Joanny M, Quintin JY (1992) Analusis 20:M19
6. Larsen BR (1995) J High Resol Chromatogr 18:141
7. Mullin MD, Pochini CM, McCrindle S, Romkes M, Safe SH,
Safe LM (1984) Environ Sci Technol 18:468
8. Bergen BJ, Nelson WG, Pruell RJ (1993) Environ Sci Technol
27:938
9. Fuoco R, Colombini MP, Samcova E (1993) Chromatographia
36:65
10. Jarnberg U, Asplund L, De Wit C, Grafsom A, Haglund P,
Jansson B, Lexen K, Strandell M, Olsson M, Jonsson B (1993)
Environ Sci Technol 27:1364
11. Haglund P, Jakobsson E, Asplund L, Athanasiadou M, Berg-
man A (1993) J Chromatogr 634:79
12. Wallace JC, Basu I, Hites RA (1996) Environ Sci Technol
30:2730
13. Vorhees DJ, Cullen AC, Altshul LM (1997) Environ Sci Tech-
nol 31:3612
14. Alcock RE, Halsall CJ, Harris CA, Johnston AE, Lead WA,
Sanders G, Jones KC (1994) Environ Sci Technol 28:1838
15. Balfanz E, Fuchs J, Kieper H (1993) Chemosphere 26:871
16. Oatman L, Roy R (1986) Bull Environ Contam Toxicol 37:461
17. Wylie PL, Klein KJ, Thompson MQ, Hermann BW (1992)
J High Resol Chromatogr 15:763
18. Grob K, Wagner C (1993) J High Resol Chromatogr 16:464
19. Letellier M, Budzinski H, Garrigues P, Wise S (1996/1997)
Spectroscopy 13:71
20. H Budzinski, M Letellier, K LeMenach, P Garrigues (1999)
J Chromatogr A 837:187200
21. Thompson S, Budzinski H (1999) Anal Chem 76:4960
22. Quilliam MA, Hardstaff WR, Anacleto JF, Leblanc MD, Ster-
giopoulos V, Dick KL, Bowser MT, Curtis JM, Embree DJ,
Sim PG, Boyd RK (1994) Fresenius J Anal Chem 350:109
23. Baumard P, Budzinski H (1997) Analusis 25:246
24. Frame GM (1997) Fresenius J Anal Chem 357:714
25. Shiu WY, Mackay DJ (1986) Phys Chem Ref Data 15:911
26. Hewlett-Packard Reference Manual 7 (1990) 12
27. Cantuti V, Cartoni GP, Liberti A, Torri AG (1965) J Chro-
matogr 17:60
28. Lovelock JE (1961) Nature 189:729
29. Lovelock JE (1962) Nature 190:540
30. Pyle SM, Marcus AB, Robertson GL (1998) Environ Sci Tech-
nol 32:3213
Anal Bioanal Chem (2002) 372 : 205215
DOI 10.1007/s00216-001-1125-6
O R I G I N A L PA P E R
Sean J. Hart Gregory J. Hall Jonathan E. Kenny
A laser-induced fluorescence dual-fiber optic array detector
applied to the rapid HPLC separation
of polycyclic aromatic hydrocarbons
Received: 19 June 2001 / Revised: 17 August 2001 / Accepted: 20 September 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract A multi-channel detection system utilizing fiber
optics has been developed for the laser-induced fluores-
cence (LIF) analysis of chromatographic eluents. It has
been applied to the detection of polycyclic aromatic hy-
drocarbons (PAH) in a chromatographically overlapped
standard mixture and to a complex soil sample extract ob-
tained during fieldwork. The instrument utilizes dual-fiber
optic arrays, one to deliver multiple excitation wave-
lengths (258342 nm) generated by a Raman shifter, and
the other to collect fluorescence generated by the sample
at each excitation wavelength; the collected fluorescence
is dispersed and detected with a spectrograph/CCD com-
bination. The resulting data were arranged into excitation
emission matrices (EEM) for visualization and data analy-
sis. Rapid characterization of PAH mixtures was achieved
under isocratic chromatographic conditions (1.5 mL min1
and 80% acetonitrile in water), with mid g L1 detection
limits, in less than 4 minutes. The ability of the instrument
to identify co-eluting compounds was demonstrated by
identifying and quantifying analytes in the rapid analysis
of a 17 component laboratory-prepared PAH mixture and
a soil extracted sample. Identification and quantification
were accomplished using rank annihilation factor analysis
(RAFA) using pure component standards and the EEMs
of mixtures measured during the rapid high-performance
Electronic supplementary material to this paper can be obtained
by using the Springer LINK server located at
http://dx.doi.org/10.1007/s00216-001-1125-6.
S.J. Hart () G.J. Hall J.E. Kenny
Tufts University, Chemistry Department,
Medford, MA, 02155, USA
e-mail: shart@ccf.nrl.navy.mil
Current address:
S.J. Hart
Naval Research Laboratory, Chemistry Division,
4555 Overlook Ave, Washington, DC 203755342, USA
Current address:
G.J. Hall
Department of Science, US Coast Guard Academy,
15 Mohegan Ave, New London, CT 06320, USA
liquid chromatography (HPLC) method as the unknowns.
The percentage errors of the retention times (RTs) deter-
mined using RAFA compared to the known RTs measured
with a standard absorbance detector were between 0 and
11%. For the standard PAH mixture, all 17 components
were identified correctly and for the soil extracted sample,
all 8 analytes present were correctly identified with only
one false positive. Overall, the system achieved excellent
qualitative performance with semi-quantitative results in
the concentration predictions of both the standard mixture
and the real-world sample. Electronic supplementary ma-
terial to this paper can be obtained by using the Springer
LINK server located at http://dx.doi.org/10.1007/s00216-
001-1125-6.
Keywords Hydrocarbons, polycyclic aromatic
Fluorescence, laser-induced Fiber optic array
Chromatography HPLC
Introduction
The analysis of complex mixtures by HPLC is a powerful
technique for compound identification and quantification.
Traditionally, compounds are chromatographically sepa-
rated and then identified by one of several means, includ-
ing refractive index, absorbance (single/dual wavelength
responses, or full absorption spectra with a photodiode ar-
ray (PDA) detector), and fluorescence (single excitation
and emission wavelength) measurements. Of these detec-
tion techniques, fluorescence is the most sensitive, while a
PDA detector provides the most information (a full ab-
sorption spectrum at each time point) for compound iden-
tification. Numerous other techniques have been devel-
oped, including fluorescence lifetime [1, 2], fluorescence
excitation-emission matrix (EEM) [3, 4, 5, 6, 7], low-an-
gle light scattering (LALLS) [8], chemiluminescence [9,
10], mass spectrometry [11], and electrochemical tech-
niques [12]. Advanced detection systems for HPLC can
offer several benefits, including accurate and rapid com-
pound identification and/or quantification, faster analysis
206
Table 1 Standard compounds, abbreviations, regression coefficients and LODs (3 of the baseline)
Compound Abbre- Slope y-intercept r2 ex max em max LOD
viation (nm) (nm) (g/L)
Naphthalene N 183.3 6.1 0.99998 266 322 131.6
p-Xylene X 66.5 3.3 0.99997 266 286 286.3
1-Methyl-naphthalene 1-MN 102.1 121.9 0.99786 266 354 186.6
2-Methyl-naphthalene 2-MN 197.1 10.9 0.99997 266 354 109.5
Acenaphthene ACE 857.4 111.9 0.99979 299 322 31.1
Fluorene FLE 1285.1 51.4 1.00000 266 304 22.7
Phenanthrene Ph 53.7 8.6 0.99836 289 347 236.5
Anthracene A 111.2 13.2 0.99987 342 380 205.6
Fluoranthene FLA 45.9 4.8 0.99992 289 445 415.2
Pyrene P 525.2 28.4 0.99998 315 372 38.7
Benz[a]anthracene BA 307.5 64.4 0.99978 289 385 61.9
Chrysene C 219.6 19.9 0.99995 266 362 92.5
Benzo[b]fluoranthene BBF 88.8 2.4 0.99995 299 446 286.1
Benzo[k]fluoranthene BKF 487.5 231.1 0.99856 299 409 41.7
Benzo[a]pyrene BAP 521.5 28.4 0.99998 299 403 43.8
Dibenz[a,h]anthracene DA 188.8 13.4 0.99850 289 394 114.4
Benzo[ghi]perylene BP 173.1 72.1 0.99830 299 404 154.1

times, lower detection limits, elucidation of peak overlap,
and the ability to measure more complex samples.
Several researchers have developed various systems
intended to exploit the information contained in EEM data
to enhance HPLC measurements. A lamp-based video flu-
orometer was developed by Christian et al. [13] and uti-
lized by Hershberger and Christian with a laminar flow
cell to perform EEM analysis of HPLC separations [3].
During the chromatographic separation, EEMs were mea-
sured (10 s total time required per EEM) and component
identification and quantification were achieved by least
squares fitting and rank annihilation analysis. An updated
version [4] of this system, using a dye-laser source capa-
ble of rapidly scanning excitation wavelengths for opti-
mum sample excitation with an intensified array detector,
has been developed. This was used to measure fluores-
cence emission spectra during chromatographic elution.
Data analysis was accomplished using least squares fitting
of emission intensity-time matrices for 4 PAHs with quan-
titative results. Jalkian and Denton developed a variable
excitation wavelength detection system for HPLC [5].
The system was capable of collecting an EEM in 30 s, in
a static mode, but could only vary the excitation wave-
length for optimum emission during an HPLC separation,
and did not attempt quantification of overlapped analytes.
The measurement of EEMs during a chromatographic
separation using a multiple-wavelength laser system and
fiber optics has been accomplished in our laboratory using
an aluminium flow cell to spread out the eluent as it
passes the excitation and emission fibers [14].
The analysis of EEM data has received much attention
in the literature beginning with Warner et al., with least
squares [15] and rank annihilation [16] methods of analy-
sis. This first attempt at rank annihilation was later im-
proved by Lorber with RAFA [17,18]. The use of the gen-
eralized rank annihilation method (GRAM) [19, 20] has
been demonstrated by Kowalski and coworkers. GRAM
was applied to bimodal chromatographic data from a stan-
dard PDA [21, 22]. This method involved the analysis of
heavily overlapped components on two different columns
under different conditions (HPLC mobile phase composi-
tion). The separation of components occurs differently in
each case (e.g., a sample having complete overlap on one
column will have only partial overlap on the other). Using
the two separations and single component standards mea-
sured on the same columns, the overlap can be deconvo-
luted using GRAM. Results for concentrations of three
PAHs (BBF, BKF and perylene; for acronyms employed,
see Table 1) in a laboratory-prepared sample were in
agreement with known values, with errors between 3%
and 15%.
Our group has been involved with the in situ LIF de-
tection of PAHs using fiber optics. LIF-EEM measure-
ments are made through sapphire windows in a probe that
is hydraulically pushed by a cone penetrometer testing
(CPT) vehicle into the subsurface [23, 24]. Our motiva-
tion for developing a dual-fiber optic detector for HPLC
was to enable us to use on-site multidimensional LIF-
HPLC measurements for the calibration and verification
of in situ results. Dual-fiber optic measurements made in
situ are subject to soil matrix effects that can alter the in-
tensities and spectral properties of the analytes of interest.
These effects include dependence of penetration depth on
particle size, medium-induced shifts in both absorption
and emission spectra, energy transfer effects, and contam-
ination by fluorescent species not of interest (e.g., humic
materials). The correlation of in situ measured EEMs with
those measured by LIF-EEM HPLC after soil extraction
could aid in the identification and correction of matrix ef-
fects. In addition, a comparable measurement on a sample
extracted from a known quantity of soil could allow cali-
bration of the in situ fluorescence measurements to pro-
vide concentrations (mg of analyte per kg of soil). During
the development of the LIF-EEM HPLC system it became

apparent that it had substantial power for the characteriza-
tion of complex mixtures with significant chromatographic
overlap. Therefore, measurements of standard HPLC sep-
arations, rapid HPLC separations with limited chromato-
graphic separation, and flow injection analysis (FIA) [25]
of mixtures were investigated. To the best of our knowl-
edge, this is the first HPLC detection system reported in
the literature to employ dual fibers or fiber arrays (one for
excitation and another for emission) probing the contents
of a capillary, although a similar strategy was recently em-
ployed in a DNA-sequencing scheme using capillary elec-
trophoresis [26]. Many other researchers have employed
fibers for excitation light delivery [27, 28] or the collec-
tion of fluorescence emission light [29, 30, 31, 32] in
chromatographic separations, but not both simultaneously.
In this paper, a standard HPLC analysis for 16 PAHs
and xylene (abbreviations given in Table 1) which typi-
cally requires 30 minute analysis times, expensive columns,
and gradient elution was accomplished in under 4 min-
utes, isocratically, with a short, inexpensive column. Such
a rapid separation is possible using EEM spectroscopy,
where multiple emission spectra are simultaneously
recorded at multiple excitation wavelengths. The addi-
tional information contained in the EEMs of the heavily
overlapped chromatographic peaks can be mathematically
deconvoluted using multi-dimensional analysis tech-
niques. Rank annihilation factor analysis is used to iden-
tify and estimate concentrations of species contained in
the EEMs measured during the rapid HPLC analysis [17,
18, 19, 20, 21, 22].
Fig. 1 LIF-EEM HPLC sys-
tem: laser source focused into
a Raman shifter, prisms used
separate the stimulated Raman
scattered beams, fiber optics
used to deliver laser beams of
differing wavelengths to the
capillary containing the chro-
matographically (HPLC) sepa-
rated mixtures. An additional
set of fiber optics is used to
collect fluorescence emission
for delivery to an imaging
spectrograph and CCD detec-
tion system
207
Experimental
The primary excitation source in our LIF system is a Q-switched
flash lamp-pumped Nd:YAG laser (Surelite I, Continuum, Santa
Clara, CA) operated at 266 nm, the 4th harmonic, at a frequency of
20 Hz. The 266 nm UV radiation serves as the pump beam for a
hydrogen/methane Raman shifter, generating many beams (>20)
of different wavelengths, through stimulated Raman scattering
[33], that can then be separated and used. In this work, seven
beams were chosen based upon their high excitation energies [34]
(the highest being chosen) and an appropriate spectral distribution
(total excitation wavelength range 84 nm) sufficient to differenti-
ate PAHs. The excitation wavelengths used were 257, 266, 278,
289, 299, 315, and 342 nm. The excitation source [23, 24, 34] and
its application to HPLC detection are shown in Fig. 1. The multiple
beams emerging from the prism system after the Raman shifter are
launched into optical fibers (held in chucks) that deliver the light
to a capillary conducting mobile phase and analytes separated by
the HPLC column.
The detection system is composed of an imaging spectrograph
(SpectraPro 150, Acton Research Corporation, Acton, MA) and a
CCD detector (TE/CCD-576EMUV, Princeton Instruments, Tren-
ton, NJ). At the entrance slit, the detection fibers are held in a col-
umn by an aluminium plug, which is interfaced to the spectro-
graph. The grating is set so that the CCD detects wavelengths in
the region of 250 nm to 500 nm. The image of the fibers is focused
onto the CCD chip held at the exit plane of the spectrograph.
A shutter between the CCD and spectrograph controls the CCD
chip exposure time, which was 3 s for these experiments. The ex-
posure time was selected to allow the collection of no less than
3 EEMs per peak at full width at half maximum (FWHM). A de-
crease in the exposure time would increase the chromatographic
resolution at the expense of detector sensitivity. For this work,
which demonstrates the detection of components within over-
lapped peaks, detector sensitivity was chosen over temporal reso-
lution.
The CCD detector has a UV coated chip with 576368 pixels,
which is thermoelectrically cooled to 40 C. The long axis of the
CCD chip is set horizontally and used as the wavelength axis. The
CCD is binned into seven channels to detect the fluorescence from
208
the seven detection fibers. Each channel is composed of 1820
rows of pixels, with five to ten unused rows between channels to
reduce the cross-talk due to charge blooming and spectrograph
aberrations. The pixels along the long axis were binned by four,
yielding 144 superpixels per channel. The fiber core diameters
were 200 m, defining the emission bandpass, which was 4 nm.
The software used to collect spectroscopic data was Winspec for
Windows 3.1 (Princeton Instruments, Trenton, NJ).
The detector used for PAH measurements relies on the fluores-
cence of analytes passing through a fused silica capillary tubing.
A linear array of fibers used to deliver different excitation wave-
lengths (excitation fiber array) was positioned next to the fused sil-
ica capillary so that the axis of the excitation beam would pass
through the centerline of the tubing. A second fiber array (emis-
sion collection array) was similarly aligned along the capillary, but
perpendicular to the excitation fiber array to avoid coupling exci-
tation light. Both fiber arrays have the same fiber spacing, and the
first excitation fiber is aligned with the first emission fiber and
therefore all fibers are so aligned. The capillary tubing was con-
nected to the standard stainless steel HPLC tubing using a standard
stainless steel HPLC 1/16 i.d. (1 inch =2.54 cm) union and a
0.4 mm i.d. Vespel-graphite ferrule (Alltech Associates, Deerfield,
IL) for mating the stainless steel hardware to the glass capillary
tubing. Analytes passed out of the column flow into the capillary
tubing and passed the excitation and emission fiber arrays posi-
tioned along the capillary. As an analyte band passes the fiber ar-
ray, the analyte molecules can be excited by the incident UV radi-
ation from a given excitation fiber. Any fluorescence from the an-
alyte will be coupled into the corresponding emission fiber.
The fibers used in this work have core, cladding, and jacket di-
ameters of 200 m, 220 m, and 240 m, respectively, and a nu-
merical aperture (NA) of 0.22 (Polymicro Technologies, Phoenix,
AZ). An important criterion in the design of the fiber arrays is the
total capillary volume probed by the fiber arrays. The length of the
array used in this work was 2.7 mm. This is made up of 7 fibers,
240 m o.d, and 6 blank or spacer fibers of 180 m o.d., with
ends painted black. The capillary used was a polyimide-coated
fused silica type, 320 m i.d., 420 m o.d. (TSP320420, Polymi-
cro Technologies, Phoenix, AZ). The excitation light cones are de-
fined by the relationship, NA = nsin, where n is the refractive in-
dex of the medium surrounding the fiber tip, and is the half an-
Fig. 2 Sketch of the fiber optic
arrays and the horizontally ori-
ented detection capillary show-
ing the excitation distal cones
and the emission collection
fibers oriented perpendicularly
to one another. Analyte bands
(size modified for illustrative
purposes) are shown traveling
in the capillary past the fiber
arrays
gle of light exiting the fiber. The total volume probed by the fiber
arrays is 0.23 L, defined by the length of the capillary that con-
tains excitation light: 2.70 mm + (2 0.05 mm). The additional
0.05 mm on either end of the array is due to the fiber optic cones
expanding over the i.d. of the capillary, calculated from the fiber
NA and the refractive index of water (n =1.33).
The optimum distance between the fibers was determined by
ray tracing, as follows. The 180 m separation gap between fibers
was chosen so that the excitation cones would overlap only in the
region beyond the internal diameter of the capillary, thus limiting
cross-talk between fiber channels. In Fig. 2, a drawing of the capil-
lary fiber array detector is shown. The essential elements are de-
picted, including the fused silica capillary, excitation fiber optic ar-
ray, collection fiber optic array, and the overlap of the fiber optic
light cones. The HPLC eluent is shown to travel through the hori-
zontally oriented capillary with analyte bands passing the detection
region. During the measurements, the detection region was iso-
lated using baffles and the room was darkened to prevent ambient
light from entering the collection fibers. The detection region
along the capillary was prepared by removing the capillary poly-
imide coating using a flame then cleaning with methanol.
The fiber arrays are fixed with epoxy resin (Tra-Con, F114-SC,
Bedford, MA) into array holders machined out of aluminium. The
holders are 50 mm 25 mm 7 mm with a 1.2 mm deep by 15 mm
wide slot milled into one surface spanning the length of the holder.
A hole is drilled through the body at the center near one edge so
that the holder can be mounted onto a micro x-y-z translation stage
(MT-XYZ, Newport, Irvine, CA). The fibers and spacers are
placed one by one onto double-sided tape placed in the slot of the
array holder. The tape holds them in place until the positions are
fixed. The ends of the fibers protrude slightly over the edge of the
slot and are covered with epoxy. Once dry, the epoxy bead cover-
ing the fiber array is filed flat using fine sandpaper, and the fiber
faces are polished with lapping paper. The excitation and emission
fibers each had a total length of 1 m, the emission fibers consisted
of two half-meter sections, joined by a standard fiber optic con-
nector (ST-type, Fiber Instrument Sales, Inc., Oriskany, NY), for
modularity.
The standard HPLC used was a Perkin-Elmer system consist-
ing of an autosampler (model ISS-200), binary pump (model 250),
column oven (model 101), and photodiode array detector (model

Fig. 3 EEMs of standard compounds used for RAFA. a) X, b) N,
c) 1-MN, d) 2-MN, e) ACE, f) FLE, g) Ph, h) A, i) FLA, j) P, k)
BA, l) C, m) BBF, n) BKF, o) BAP, p) DA, q) BP
LC-235C). The column used was a 3.3 cm diameter, 5 m particle
size, C18 column (Perkin-Elmer, Brownlee Columns, Norwalk,
CT). A C18 guard column (Alltech, Waukegan, MI) was used be-
fore the analytical column. The separation conditions were as fol-
lows: flow rate of 1.5 mL min1, isocratic elution (80% acetoni-
trile/20% water) for 5 min. All injection volumes were 25 L and
the column oven temperature was set at 30 C. This system and
method were used to measure the standard EEMs (single compo-
nents), the 17 component mixture, and the soil extract.
Another HPLC method was used for the purposes of determin-
ing the LIF-EEM systems sensitivity towards the various PAHs
and the conventional analysis of the unknown soil extract sample.
This required complete separation of the analyte peaks and was ac-
complished by using more rigorous separation conditions coupled
with a more selective column (Green PAH, Hypersil, UK). The
separation conditions were isocratic elution (50% acetonitrile/50%
water) for 5 minutes, and then a linear gradient elution (to 100%
acetonitrile) for 25 min. A column equilibration time of 5 min was
used between runs. All other parameters were the same as in the
other method. The detection limits obtained using the isocratic elu-
tion method are expected to be somewhat higher due to the greater
band broadening associated with isocratic separations. However,
the retention times with the rapid method are short (less than
4 min) and this tends to reduce the band broadening effect.
Detector characterization
An important consideration regarding detector performance in a
multi-channel instrument is channel separation, both for excitation
209
and emission. There can be cross-talk between channels at either
the CCD detector or the fiber optic array. In our system, excitation
light from each channel is detected in neighboring channels be-
cause of charge blooming on the CCD chip, which is the spilling
over of photon-generated electrons from pixels whose well capac-
ity has been exceeded by excess excitation light [35]. While filters
are often used to reject scattered light in traditional single wave-
length fluorescence experiments, in a multi-channel LIF system a
different long-pass filter would be required to efficiently remove
the scattered light from each channel while not absorbing any of
the fluorescence. The placement of multiple filters is not readily
possible in such a system while maintaining optimum sensitivity
[6]. However, we have found that subtraction of a blank EEM re-
moves the majority of the scattered light resulting in a 94% de-
crease in the height of a contaminating scattered light peak. Data
post-processing routines can remove any residual scattered light
artifacts. The cross-talk of fluorescence from channel to channel
would represent a greater challenge, because of the contamination
of the analytical information sought. Measurements were per-
formed to determine whether fluorescence cross-talk between
channels was significant. This was done by exciting the fluores-
cence of quinine sulfate (0.1 g L1 in ethanol), static in the capil-
lary, using one excitation channel at a time while collecting data
from all the emission fibers. The fluorescence cross-talk, defined
here as the peak fluorescence level in the emission channel being
tested divided by the peak fluorescence level in the emission chan-
nel corresponding to the excitation, was <0.9% for all cases.
Data processing
The LIF-EEM data were acquired in the autostore mode of the
CCD software. In this mode, the software collects data continu-
ously and saves each frame, consisting of seven emission spectra,
one at each of seven excitation wavelengths, i.e., an EEM. A pre-
210
Fig. 4 Relative intensity plots for rapid HPLC of 17 component
mixture. a) PDA UV chromatogram (255 nm) (dashed line), total
summed fluorescence (solid line), summed fluorescence at differ-
ent excitation wavelengths b) 258 nm, c) 266 nm, d) 278 nm, e)
289 nm, f) 299 nm, g) 315 nm, h) 342 nm
viously measured file containing the CCD dark current and any
other possible light reaching the detector, excluding the laser exci-
tation lines, was subtracted in real time by the software on a per
frame basis. In this application the integration time was 3 s and the
file processing/writing was accomplished during the subsequent
integration period. During the typical 5 minute HPLC separation,
100 files were collected at 37 kB each, totaling 3.7 MB of data.
The raw data were processed using programs written in C++.
Certain functions (standard deviation and smoothing) were called
upon from a commercial C++ function library package (Science,
Engineering, and Graphics Tools, Quinn-Curtis, Inc., Needham,
MA). The programs were designed to allow rapid background sub-
traction and scattered light removal for the large number of files
generated during each measurement [25, 36]. This was accom-
plished by first subtracting a blank EEM, i.e., that of pure mobile
phase measured when no analytes were present in the detection re-
gion. Small residual scattered light peaks are left behind after
blank subtraction, due to variations in the mobile phase composi-
tion. These are removed using an automated C++ program, which
scans the derivative spectrum to find transition points that define
scattered light peaks (<10 nm), which can then be removed and in-
terpolated data substituted [25, 36].
Each EEM was power normalized using the peak calibrant
(0.10 g L1 quinine sulfate in ethanol) fluorescence intensity (at
365 nm) in detector counts on a per channel basis. In order to iden-
tify temporal regions of interest, the data are reduced by summa-
tion in the emission wavelength dimension. A C++ program has
been written to read the processed files (background, scattered
light removed, and normalized) and sum the intensities at all emis-
sion wavelengths, outputting channel (excitation wavelength)
sums versus time or frame number. The total fluorescence chro-
matogram is simply the plot of the sum of the normalized excita-
tion channel sums as a function of time. The RAFA calculations
[17, 18] were programmed in MatLab [37] (MathWorks Inc., Sher-
born MA), and were run on a Pentium 166 MHz PC. RAFA was
selected as the analysis method of choice as the added complexity
of multi-analyte GRAM analysis was deemed unnecessary for this
application. The RAFA calculations on the PAH mixtures were
performed sequentially using EEMs of the pure components as
standards.
Reagents
The following PAHs were of 98% purity or higher, with the ex-
ception of 1-MN (95%), and were used without further purifica-
tion: N, ACE, FLE, Ph, A, P, BBF, 1-MN, 2-MN (Aldrich Chem-
ical Co., Milwaukee, WI), FLA, BKF, BP, BA, C, DA, and BAP
(Acros Chemical, NJ) (See Table 1 for compound abbreviations).
p-Xylene (98%) (J.T. Baker, Phillipsburg, NJ) was also used with-
out further purification. The mobile phase used was HPLC grade
acetonitrile (Fisher Scientific, Fairlawn, NJ) with distilled water
prepared in house. All PAHs were prepared in the mobile phase
used for analysis (80% acetonitrile/20% water).

Fig. 5 Selected EEMs from the rapid separation of the 17 compo-
nent mixture. The position of the EEM in the chromatogram is
given by the lettering of the normalized summed fluorescence
point that corresponds to that EEM (excitation wavelength is Ex ,
emission wavelength is Em )
An unknown mixture of significant complexity was chosen to
highlight the capability of the instrument to identify analytes with
significant spectral overlap and fluorescence contamination. The
sample chosen was an extract of a soil sample from Otis Air Na-
tional Guard Base (OANGB) in Cape Cod Massachusetts. The
sample was extracted according to a procedure using ultra-sonica-
tion and reversed-phase column clean-up described elsewhere
[36]. Conventional HPLC with PDA detection was performed on
the sample to determine the identities and concentrations of the
PAHs present.
Results and discussion
Continuous monitoring of the source intensities over half-
hour periods was performed to ascertain excitation inten-
sity stability. This was accomplished using three second
exposures of the CCD (each integrating the results of 60
laser shots) with a standard solution of quinine sulfate
present in the capillary, the fluorescence being propor-
211
tional to the excitation energy. In order to account for
variations in the source intensity during the approximately
4 minute HPLC analyses, three measurements of quinine
sulfate (0.10 g L1 in ethanol) fluorescence were made be-
fore and after each analysis. The average of all six mea-
surements served to normalize the excitation energy de-
pendence of the fluorescence sums and EEMs for a given
analysis. The % standard deviation for all six measure-
ments was roughly 7% for all cases. The variation in any
given channels excitation energy over several days was
30%. While the variations between measurements are
accounted for by the normalization procedure, short-term
variations occurring in between the calibration measure-
ments remain unaccounted for and may contribute to er-
rors in the subsequent analysis. Due to the spatial and
temporal variations of the Raman shifted beams [36] and
our fiber launch conditions (overfilled to prevent dam-
age), the variations may be greater than measured. These
unaccounted variations may be responsible for the degra-
dation of quantitative information produced by the sys-
tem. Further work is being performed to improve the sta-
bility of this novel excitation source.
212

Limits of detection etc. General, qualitative comparisons between EEMs

A measure of the systems performance and potential ap-
plications are the LODs for the different PAHs. Four mix-
tures, containing different concentrations of the 17 com-
pounds, were subjected to a standard PAH analysis, de-
scribed above, with complete separation of all compo-
nents. The calibration of the instruments response was
determined by measuring the EEMs at four concentrations
of each PAH and p-xylene. The intensity at the peak exci-
tation and emission wavelengths was chosen for each
molecule. The standard compounds, abbreviations, cali-
bration curve summaries (regression coefficients, and r2
value), wavelengths of peak excitation and emission, and
LODs are given in Table 1. As can be seen, the response
is linear (the concentrations were typically in the range
0.550 mg L1 for all compounds) with the square of the
correlation coefficient ranging from 0.998 to 1.000. The
LODs were all <420 g L1, with some of the more fluo-
rescent compounds having LODs in the 20100 g L1
range.
LIF-EEM analysis
For qualitative comparisons, the EEMs of the standards
are shown in Fig. 3 aq. These EEMs are used in the
RAFA of the rapidly analyzed 17 component mixture and
the soil extract, discussed below. At the concentrations
used in this work (low mg L1), no resonance energy
transfer was observed between co-eluting PAHs. The
spectral variations in the EEMs are clearly evident, with
short wavelength excitation and emission for the early
eluting compounds such as X, N, 2-MN, 1-MN, Ph, FLE,
and ACE and mid- to long- wavelength excitation and emis-
sion for the larger compounds such as A, FLA, P, BA C,
measured using our fiber optic system may be made with
EEMs collected using a commercial spectrometer pro-
vided that both are excitation power normalized as dis-
cussed in the experimental section.
A standard 17 component mixture consisting of 16 PAHs
and p-xylene was injected into the HPLC system for rapid
separation and LIF-EEM analysis. This resulted in a chro-
matogram that contained several heavily overlapped
peaks. The total normalized summed fluorescence and
UV absorbance (255 nm) chromatograms are compared in
Fig. 4 a. The summed fluorescence plots for individual ex-
citation wavelengths are shown in Fig. 4 bh. As can be
seen, both the UV and fluorescence chromatograms con-
tain between 8 and 10 peaks distributed over 17 com-
pounds, indicating significant overlap in several of the
peaks. The peak positions in the UV chromatogram com-
pare well with those in the fluorescence chromatogram,
but the relative intensities are different, as expected.
EEMs from selected times during the rapid separation
of the 17 component mixture are shown in Fig. 5 ar, with
the summed total fluorescence plot labeled with the corre-
sponding EEM letters. A visual comparison with the stan-
dard EEMs in Fig. 3 allows the identification of some of
the regions of chromatographic overlap. For example,
in Fig. 5 b and 5 c, there are two emission peaks, one at
280 nm and the other at 325 nm. The first is due to the X
peak and the second is due to N. Similarly, the overlap of
FLA and P can easily be seen in the EEMs in Fig. 5 ik.
The first EEM is FLA, followed by a lower concentration
mix of the two and finally a dominant P with a small
amount of residual FLA in EEM k. A final example is that
of BBF and BKF between EEMs n and o in Fig. 5. EEM n
is a mixture of the two analytes (BBF at higher concentra-
tion), seen by observing the large long wavelength fluo-
rescence with a doublet forming at 420 nm. The next

Table 2 Comparison of known

retention times and concentra- Analyte PDA RTa RAFA RTb Ratioc Known conc RAFA Ratiod
tion values with RAFA de- (min) (min) (mg/L) Calculated
termined values for the stan- conc (mg/L)
dard 17 component PAH mix-
ture X 0.69 0.70 1.02 12.9 19.1 1.48
N 0.64 0.65 1.02 5.5 8.2 1.49
1-MN 0.76 0.75 0.99 50.0 59.9 1.20
2-MN 0.78 0.70 0.89 4.8 7.3 1.52
ACE 0.84 0.85 1.01 5.1 3.0 0.59
FLE 0.82 0.80 0.97 4.6 2.4 0.52
Ph 0.90 0.90 1.00 4.5 5.9 1.31
A 0.95 0.90 0.95 6.0 9.7 1.62
FLA 1.07 1.10 1.03 6.4 9.2 1.44
Symbols: RT = retention time; P 1.23 1.20 0.97 4.2 4.1 0.98
aphotodiode array detector de-
BA 1.56 1.50 0.96 4.6 6.0 1.30
termined RTs by single analyte
injection, bpeak concentration C 1.57 1.55 0.99 5.6 13.1 2.34
time from rank annihilation BBF 2.09 2.10 1.00 4.5 6.2 1.38
factor analysis, cratio of RAFA BKF 2.27 2.20 0.97 4.7 3.3 0.70
determined RT to PDA deter- BAP 2.54 2.55 1.01 3.9 5.4 1.38
mined RT, dratio of RAFA de- DA 3.09 3.30 1.07 3.9 4.0 1.03
termined concentration to
known value BP 3.58 3.60 1.00 4.4 4.0 0.91

Fig. 6 Selected EEMs from the rapid separation of the ONAGB
soil sample extract. The position of the EEM in the chromatogram
is given by the lettering of the normalized summed fluorescence
point that corresponds to that EEM (excitation wavelength is Ex ,
emission wavelength is Em )
EEM is dominated by BKF with a smaller amount of BBF
present, obvious by observing the same doublet with
peaks at 420 nm and 435 nm.
There are two performance comparisons to be made:
between the measured analyte RTs and concentrations and
the known RTs and concentrations. Table 2 lists the
known RTs and analyte concentrations and compares
them with the RAFA determined values. The RAFA RTs
are the times at which the peak concentrations of analytes
were determined. All 17 analytes were correctly identified
and the RAFA-predicted RTs agreed with the known val-
213
ues to within 11% in all cases. It should be noted that the
RTs of the individual analytes were unknown (refer to Fig.
4 a) and had to be determined through 17 separate HPLC
analyses (described above) of pure standards injected
into the same system under identical conditions (requiring
85 min of analysis time).
Overall, the system achieved semi-quantitative perfor-
mance, with RAFA-calculated concentrations generally
(22 out of 25) being within a factor of 2 from the known
values and many (18 out of 25) being within a factor of
1.5. All but three concentrations determined by RAFA
were within 50% of the known concentrations, with the
RAFA results being generally higher. This is not unex-
pected, since there is no constraint in RAFA against at-
tributing the observed fluorescence sequentially to more
than one analyte. Three of the values (P, DA, and BP)
were determined to within 10% error (ratio, 0.91.0).
214

Table 3 Comparison of
known retention times and Analyte PDA RTa RAFA RTb Ratioc PDA (Known) RAFA Calculated Ratiod
concentration values with (min) (min) Conc (mg L1) Conc (mg L1)
RAFA determined values for X nd nd nd nd
the ONAGB soil sample ex-
tract N nd nd nd nd
1-MN nd nd nd nd
2-MN nd nd nd nd
ACE nd nd nd nd
FLE nd 0.75 nd 0.13
Ph nd nd nd nd
A 0.95 1.00 1.05 1.45 1.46 1.01
Symbols: nd = non-detect, FLA 1.07 1.00 0.93 1.05 1.95 1.86
RT = retention time; aphoto- P 1.23 1.20 0.98 1.99 2.33 1.17
diode array detector deter- BA 1.56 1.55 0.99 0.27 0.94 3.48
mined RTs by single analyte
injection, bpeak concentration C 1.57 1.55 0.99 0.79 2.11 2.67
time from rank annihilation BBF 2.09 2.10 1.00 1.47 0.94 0.64
factor analysis, cratio of RAFA BKF 2.27 2.20 0.97 0.62 1.07 1.73
determined RT to PDA deter- BAP 2.54 2.55 1.00 0.31 0.40 1.29
mined RT, dratio of RAFA DA nd nd nd nd
determined concentration to
known value BP nd nd nd nd

A linear regression of measured and RAFA calculated
concentrations versus known analyte concentration shows
a general linear trend, as indicated by a r2 value of 0.97.
The rapid separation technique with RAFA analysis of
EEMs may therefore be said to provide semi-quantitative
results for specific chemical substances.
A soil extract sample collected at OANGB was ana-
lyzed to determine the systems ability to characterize a
complex real world sample. The normalized summed flu-
orescence chromatogram and selected EEMs are shown in
Fig. 6. The summed fluorescence chromatogram reveals a
region of elevated baseline from 0.5 min to 2.5 min with
several peaks overlaid. The first EEM at 0.65 min (Fig. 6 a)
contains broad fluorescence emission (350490 nm) that
does not match any EEM in the library (Fig. 3). The sec-
ond EEM at 0.95 min (Fig. 6 b), contains the characteristic
anthracene signal similar to the standard EEM shown in
Fig. 3 h. There is additional, unaccounted, fluorescence
emission from 313 nm to 427 nm (324 nm peak) at exci-
tation wavelengths 266288 nm. The third EEM at
1.15 min (Fig. 6 c), contains the fluorescence profile of
fluoranthene with long wavelength emission from 356 nm
to 506 nm. There is additional fluorescence near the same
region as in the previous EEM: 310340 nm (peak at
323 nm). This may indicate a common type of interfer-
ence that was co-extracted from the sample with the PAHs
that is contributing to the elevated baseline. The peak of
the summed fluorescence plot is defined by EEM d, which
shows the fluorescence profile of pyrene, similar to the
standard in Fig. 3 j. The complete chromatographic over-
lap of BA and C can be seen in Fig. 6 e, where there is sig-
nificant spectral overlap (refer to standard EEMs, Fig. 3 k
and 3 l). The spectral signature of BBF can be seen in
EEM f in Fig. 6 with additional emission signal (possibly
due to a co-eluting unknown) in the 290330 nm region at
excitation wavelengths 258278 nm.
The RAFA results and comparison with conventional
HPLC for the OANGB soil extract sample are given in
Table 3. The PDA and RAFA determined RTs for the
OANGB soil extract agreed to within 9%. All eight ana-
lytes present (determined by HPLC) were correctly iden-
tified with one additional compound (FLE) found that was
not detected using the standard HPLC analysis. All RAFA
determined concentrations were within a factor of 2 of the
known values with the exception of BA and C, ratios of
3.5 and 2.7, respectively (the poorer result may be a result
of spectral overlap). The other analytes had ratios ranging
from 0.6 to 1.7. Four (A, P, BKF, and BAP) out of eight
analytes had ratios that were within 0.51.5. It is reason-
able to expect that the performance of the system should
be poorer for BA and C due to spectroscopic overlap [36].
Conclusion
There are relatively few chromatographic detection tech-
niques that can determine analyte identity and concentra-
tion in an unseparated or partially separated mixture; most
require complete chromatographic resolution of neighbor-
ing peaks. Our dual-fiber optic LIF-EEM detector has
proven effective in the rapid characterization of complex
mixtures. Optical fibers as channel separators and light
collection devices are beginning to be used in commercial
analytical devices. In a commercially available UV-VIS
PDA detector (Thermo Separation Products, UV6000LP,
San Jose, CA), fibers are used to transform the light
emerging from a reflective light pipe flow cell (circular
beam profile) into the rectangular shape required by the
diode array detector, thus minimizing losses associated
with a circular beam incident on a slit [38].
The dual-fiber optic capacity makes our detection sys-
tem flexible and potentially applicable in several areas.
These include FIA (flow injection analysis) [25], CE
(capillary electrophoresis), and manufacturing process
control [39] (e.g., fuel production, natively fluorescent or
tagged bio-polymeric products). The use of dual-fiber op-

tic arrays allows the instrumentation (laser, optics, spec-
trograph, and CCD) to be completely detached from the
detection region, so that hazardous or inaccessible sys-
tems in an industrial setting may be monitored without en-
dangering the instrumentation or operator.
The EEM-HPLC combination, using standard columns
for complete separation of analytes, has promise as an on-
site calibration tool for LIF-CPT site characterizations.
The rapid HPLC/EEM detection technique has several ad-
vantages including simplicity (dual-fiber optic), relative
speed (3 s or less per measurement), and the ability to de-
termine sample composition and estimate concentrations
without complete separation of species of interest. This
technique has the potential to be applied in many different
screening applications because of its inherent flexibility.
In a static or chemical flow process (HPLC, CE, flow
process monitoring, etc.) a dual-fiber optic system capa-
ble of being deployed in a dangerous or inaccessible loca-
tion would have significant value. In addition, its utility
for method development and peak purity analysis would
be substantial for qualitative identification purposes and
quantity estimation. Even with the quantitative limitations
of the current system, there is considerable value in the
current level of qualitative and semi-quantitative perfor-
mance.
The difficulties found with quantitative performance
are addressable and we believe could be remedied with
further development efforts. Efforts to reduce variations
and uncertainty in the measurements might enable the
quantitative determination of more compounds than was
achieved in this study. With some modifications, the vari-
ations in excitation energy could be more carefully moni-
tored or reduced. These adjustments include power nor-
malization on a frame by frame basis, use of tapered
fibers for complete excitation beam launching, and Ra-
man shifter gas re-circulation for improved beam quality.
Alternatively, the use of a high power lamp would avoid
the need to reduce excitation energy instability associated
with a novel excitation source, and would reduce the cost,
at the expense of detection limits. With improvements to
the current design, the system could achieve lower detec-
tion limits and better quantitative performance while
maintaining the current simplicity and speed, and provid-
ing the same high performance qualitative capabilities.
Acknowledgements Funding for this work was provided in part
by the National Risk Management Research Laboratory of the US
Environmental Protection Agency, and in part by the United States
Department of Defense under Grant No. DACA 39-93-1-001 to Rice
University for the Advanced Applied Technology Demonstration
Facility for Environmental Technology Program (AATDF). This
work does not necessarily reflect the position of these organiza-
tions and no official endorsement should be inferred. SJH would
like to thank William C. Green for his helpful suggestions and in-
sights.
215
References
1. Dvorak MA, Oswald GA, Van Benthem MH, Gillispie GD
(1997) Anal Chem 69:34583464
2. Smalley MB, McGown LB (1995) Anal Chem 67:13711376
3. Hershberger LW, Callis JB, Christian GD (1981) Anal Chem
53:971975
4. Skoropinski B, Callis JB, Danielson JDS, Christian GD (1986)
Anal Chem 58:28312839
5. Jalkian R, Denton B (1989) SPIE 1054:91102
6. Muroski AR, Booksh KS, Myrick ML (1996) Anal Chem 68:
35343538
7. Booksh KS, Muroski AR, Myrick ML (1996) Anal Chem 68:
35393544
8. Mhatre R, Stuting HH, Krull IS (1989) Trends in Anal Chem
8:260268
9. Miyawa JH, Schulman SG, Perrin JH (1997) Biomed Chro-
matogr, 11:224229
10. Zhang Z, Baeyens WRG, Zhang X, Zhao Y, Van Der Weken D
(1997) Anal Chim Acta 347:325332
11. Whitehouse CM, Dreyer M, Yamashita M, Fenn JB (1985)
Anal Chem 57:675679
12. Chao M, Huang H, Anal Chem (1997) 69:463470
13. Johnson DW, Gladden JA, Callis JB, Christian GD (1979) Rev
Sci Instrum 50:1
14. Taylor TA, Kenny JE (1996) US Patent 5 491 344
15. Christian GD, Davidson ER, Warner IM (1977) Anal Chem 19:
21522159
16. Ho CN, Christian GD, Davidson ER (1980) Anal Chem 52:
10711079
17. Lorber A, Anal Chim Acta (1984) 164:293297
18. Lorber A, Anal Chem (1985) 57:23952397
19. Sanchez E, Kowalski BR (1986) Anal Chem 58:498499
20. Sanchez E, Kowalski BR, Wilson BE (1989) J Chemom 3:
493498
21. Sanchez E, Ramos LS, Kowalski BR (1987) J Chromatogr 385:
151164
22. Ramos LS, Sanchez E, Kowalski BR (1987) J Chromatogr 385:
165180
23. Hart SJ, Chen YM, Lien BK, Kenny JE (1996) SPIE 2835:
7382
24. Hart SJ, Chen YM, Lien BK, Best TW, Kenny JE (1997) Field
Anal Chem Tech 6:2538
25. Hart SJ, Hall GJ, Kenny JE (1998) SPIE 3534:601
26. Kheterpal I, Mathies RA (1999) Anal Chem 71:31A-37A
27. Abbas AA, Shelly DC (1995) J Chromatogr A 691:3753
28. Edkins TJ, Shelly DC (1998) J Chromatogr 459:109118
29. Sepaniak MJ, Yeung ES (1980) J Chromatogr 190:377383
30. Van de Nesse R, Hoornweg G, Gooijer C, Brinkman UATh,
Velthorst NH (1993) Anal Chim Acta 281:373383
31. Sepaniak MJ, Yeung ES (1980) J Chromatogr 190:377383
32. Nishi S, Nomura A, Yamada J, Tsunoda K (1990) Anal Chim
Acta, 229:37
33. Jarvis GB, Mathew S, Kenny JE (1994) Appl Opt 33:4938
34. Lin J, Hart SJ, Taylor TA, Kenny JE (1994) SPIE 2367:7079
35. Sweedler JV (1993) Crit Rev Anal Chem 24:5998
36. Hart SJ, PhD Dissertation (1998) Tufts University
37. Pepper JW, Wright AO, Kenny JE (2001) Spectro Chim Acta
Part A, in press
38. DeLand P (1997) American Laboratory, November, 1924
39. Cooley RE, Stevenson CE (1992) Process Control and Quality,
2:4353
Anal Bioanal Chem (2002) 372 : 216220
DOI 10.1007/s00216-001-1123-8
O R I G I N A L PA P E R
Christopher J. Smith Wenlin Huang
Charisse J. Walcott Wayman Turner James Grainger
Donald G. Patterson Jr
Quantification of monohydroxy-PAH metabolites in urine
by solid-phase extraction with isotope dilution-GCMS
Received: 8 June 2001 / Revised: 28 July 2001 / Accepted: 25 August 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract For measurement of biomarkers from poly-
cyclic aromatic hydrocarbon (PAH) exposure, an analyti-
cal method is described quantifying hydroxylated PAH
(OH-PAH) in urine samples. This method determined
monohydroxy metabolites of naphthalene, fluorene, phen-
anthrene, fluoranthene, pyrene, chrysene, benzo[c]phen-
anthrene, and benz[a]anthracene. The sample preparation
consisted of enzymatic hydrolysis, solid-phase extraction
and derivatization with a silylating reagent. Five carbon-13
labeled standards were used for isotope dilution. Ana-
lytes were separated by gas chromatography (GC) and
quantified with high-resolution mass spectrometry (HRMS).
This method produced good recoveries (4170%), linear-
ity, and specificity. Data were corrected for blank levels
from the naphthalene, fluorene, and phenanthrene metab-
olites. Method detection limits ranged from 2 ng L1 for
1-hydroxypyrene to 43.5 ng L1 for 1-hydroxynaphtha-
lene. Using quality control charts from two urine pools,
the method can be readily applied to biomonitoring PAH
exposure.
Keywords Polyaromatic hydrocarbons Isotope dilution
High resolution mass spectrometry Human urine
Solid-phase extraction
Introduction
Many classes of chemicals are thought to be carcinogenic
or genotoxic. Among these, PAH are one of the most sig-
nificant priority pollutants based upon the amounts pro-
duced by combustion of biogenic and anthropogenic ma-
terials [1]. Human exposure to PAH is generally caused
by airborne particulates and by dietary consumption of
C.J. Smith () W. Huang C.J. Walcott W. Turner
J. Grainger D.G. Patterson Jr
Centers for Disease Control and Prevention,
4770 Buford Highway, MS F-17, Atlanta, GA 30341, USA
e-mail: cps6@cdc.gov
food prepared by grilling, smoking, or storage in charred
wooden casks. PAH exposure is also caused by cigarette
and cigar smoking [2]. Biomarkers of human exposure to
PAH are necessary for the accurate determination of ex-
posure and body burden from the above-mentioned
sources. By measuring metabolites of PAH (PAHm) in
urine, short term (1 to 3 days) exposure may be evaluated.
A widely adopted method was reported by Jongenee-
len and coworkers for the quantification of 1-hydroxy-
pyrene from urine [3]. Their protocol includes enzymatic
hydrolysis of the conjugated metabolites, solid-phase ex-
traction (SPE) with C-18 media, separation with high-per-
formance liquid chromatography (HPLC), and detection
by fluorescence emission. Recent exposure assessments
using this method include workers at a gasworks plant,
firefighters at a training facility, and children from an in-
dustrial region [4, 5, 6]. Another HPLC method that de-
termines hydroxypyrene, hydroxyphenanthrene, hydroxy-
benz[a]anthracene, and hydroxybenzo[a]pyrene has been
developed by Angerer and coworkers [7, 8]. They applied
their method to monitor occupational exposure of workers
in a fireproof stone-producing plant. Exposures to PAH
were found with the conclusion that the workers expo-
sure was too high when compared with other industrial
plants and the exposure limits for benzo[a]pyrene.
Several groups have worked on the analysis of PAHm
using GCMS techniques and incorporate varying sample
preparation methods. With solid-phase microextraction,
Gmeiner and coworkers developed a screening method
for PAHm with GCMS to detect 10 monohydroxy PAHm
in urine samples [9]. Korfmacher and coworkers have
used GCMS with chemical ionization to analyze metab-
olites of nitro-PAH from microsomes [10]. Incorporating
liquidliquid extraction and fractionation with a silica gel
column, Wilson and coworkers have used GCMS to
monitor PAH exposures in children of low-income fami-
lies [11]. A productive method has been applied by Grim-
mer and coworkers who have used liquid extraction and
GCMS to determine metabolites of phenanthrene, fluo-
ranthene, pyrene, chrysene, and benzo[a]pyrene in urine
[12, 13, 14, 15]. They have demonstrated that the relative
 217

ratios between metabolites from a PAH are invariant in an

individual; an important result for the basis of correlating
PAH exposure and health effects.
Using immunoaffinity chromatography and HPLC,
Strickland and coworkers have given several reports on
PAH exposure results and comparisons of three analytical
methods [16, 17, 18, 19]. Hurtubise and coworkers have
used solid-matrix luminescence as a method to detect
PAHm in purified HPLC fractions, and they examined ap-
plications of capillary electrophoresis [20, 21]. Additional
methods with applications to metabolites in urine, DNA
adducts, and protein adducts can be found in an article by
Angerer, Mannschreck and Gundel and a review by Jon-
geneelen [22, 23].
With the above HPLC and GC procedures, there is still
a need to develop a straightforward method that can mon-
itor exposure to many different PAH. Such a technique
could then examine PAHm profiles to determine correla-
tions between the PAH source and the resulting metabo-
lites, changes in the sources of exposure, and whether ex-
posure reduction practices were effective. The approach
taken herein focused on OH-PAH with significant urinary
excretion and the expectation of successful detection at
background (non-occupationally exposed) levels. These
technical and laboratory specific considerations led to the
selection of 2-, 3- and 4-ring OH-PAH as the target ana-
lytes. This paper describes an isotope dilution-GCHRMS
method for quantifying the monohydroxy metabolites of
PAH found in urine. For accurate calibration, carbon-13
labeled standards were used for isotope dilution. The pro-
cedure involves sample preparation by SPE followed by
trimethylsilyl derivatization. OH-PAH from 8 parent PAH
can be determined. The choice of HRMS produced excel-
lent sensitivity and target analyte specificity. The method-
ology was applied to a set of pooled urine samples from
non-occupationally exposed persons to generate quality-
control data. As a validated method, this procedure can be
readily used to monitor PAH exposure by quantitation of
the PAHm in urine.
Experimental
Chemicals
The list of analytes is presented in Table 1. Standard materials for
the OH-PAH were obtained from Aldrich Chemical (Milwaukee,
WI, USA), Midwest Research Institute (Kansas City, MO, USA),
AccuStandard (New Haven, CT, USA), and Promochem (Wesel,
Germany). Internal standards were 13C6 labeled with 99% chemi-
cal purity and greater than 98% isotopic purity. The carbon-13 la-
bels were obtained as follows: 3-fla and 1-pyr from TerraChem
(Merriam, KS, USA), 3-phen and 6-chr from Cambridge Isotope
Laboratories (Andover, MA, USA), and 1-nap was synthesized at
the Centers for Disease Control and Prevention. Stock solutions of
the individual OH-PAH were made up in acetonitrile, and working
solutions of the 18 OH-PAH mixture were made by combining the
aliquots of the stock solutions and dilution in acetonitrile. A work-
ing solution of 100 pg L1 per carbon-13 labeled standard was
prepared. The solutions were kept in silanized amber glass vials
and stored at 70 C. Argon was used to displace the air in the vial
headspace. Except as noted, all other chemicals were obtained
from Sigma Chemical (St Louis, MO, USA).
Sample preparation
Urine donations from an anonymous volunteer group were pooled.
To prepare samples for enzymatic hydrolysis, 3 mL of the urine
pool was added to 10 L of the working labeled solution, 5 L of
beta-glucuronidase/arylsulfatase (Roche Molecular Biochemicals,
Indianapolis, IN, USA) and 5 mL of 0.1 mol L1, pH 5 sodium ac-
etate buffer. The samples were incubated at 37 C for 3 h.
Using a vacuum manifold holding 12 cartridges, SPE car-
tridges (EnvirElut PAH, Varian, Harbor City, CA, USA) were con-
ditioned with methanol followed by deionized purified water. Hy-
drolyzed urine samples were loaded onto the cartridges, that were
then rinsed with deionized water and 50:50 (v/v) watermethanol.
After drying using vacuum for 10 min, the cartridges were washed
with 50:50 (v/v) acetonitrilemethanol. To remove residual water,
the acetonitrilemethanol eluent was passed through a cartridge
loaded with 2.5 g anhydrous sodium sulfate. Two milliliters of
acetonitrile was then used to rinse the sodium sulfate. The com-
bined eluent was evaporated under a stream of nitrogen in a water
bath at 55 C. The residue was reconstituted with 20 L toluene
and placed in a 700-L tapered autosampler vial.

Table 1 List of analytes, ab-

breviations, method detection No. Analyte Abbre- MDL QC level CV
limits (MDL), calculated levels viation (ng L1) (ng L1) (%)
and coefficients of variation 1 1-Hydroxynaphthalene 1-nap 43.5 889 5
(CV) from QC samples with a
target of 800 ng L1 of each 2 2-Hydroxynaphthalene 2-nap 19.3 870 6.5
OH-PAH 3 2-Hydroxyfluorene 2-fle 15.2 1135 10.9
4 3-Hydroxyfluorene 3-fle 9.5 1099 10.6
5 1-Hydroxyphenanthrene 1-phen 15.0 na na
6 2-Hydroxyphenanthrene 2-phen 11.2 1052 5.8
7 3-Hydroxyphenanthrene 3-phen 15.3 1051 2.6
8 3-Hydroxyfluoranthene 3-fla 3.5 664 1.7
9 1-Hydroxypyrene 1-pyr 2.0 824 3.2
10 1-Hydroxybenzo[c]phenanthrene 1-bcp 5.7 565 19.4
11 2-Hydroxybenzo[c]phenanthrene 2-bcp 6.8 488 9.9
12 3-Hydroxybenzo[c]phenanthrene 3-bcp 5.0 476 10.4
13 3-Hydroxychrysene 3-chr 9.9 470 20
14 6-Hydroxychrysene 6-chr 3.4 837 2.0
15 1-Hydroxybenz[a]anthracene 1-baa 4.6 475 9.3
16 3-Hydroxybenz[a]anthracene 3-baa 5.4 478 7.9
218
The derivatization was performed by adding 5 L of N-methyl-
N-(trimethylsilyl)-trifluoroacetamide (MSTFA) to the autosampler
vial, displacing the headspace air with argon, crimp sealing, and
incubating at 60 C for 1 h. For a set of 12 hydrolyzed urine sam-
ples, the processing time until GCMS analysis was approximately
2.75 h. Derivatized samples could be stored in a refrigerator for
one week.
GCMS conditions
The chemical analyses were performed on a Finnigan MAT 95
high-resolution mass spectrometer (Bremen, Germany) with an
Agilent Technologies 6890 gas chromatograph (Palo Alto, CA,
USA). The GC injector and transfer line were kept at 290 C. The
injector was operated in the splitless mode for 2 min, and a 2 mm
inner diameter glass liner was used. The chromatography column
(J&W Scientific, Folsom, CA, USA) was a 25 m DB-5MS with
0.25 mm inner diameter, and 0.25 m film. With a constant He
flow of 1 mL min1, the oven temperature was held at 100 C for
2 min, ramped to 160 C in 4 min, ramped from 160 C to 295 C
in 14 min, and then held at 295 C for 1.5 min.
The mass spectrometer was set to monitor the molecular ions
from the derivatized analytes using electron impact at 40 eV. This
electron energy produced more intense signals from the molecular
ions than the more commonly used 70 eV energy. The ion-source
temperature was 260 C. Similar ion intensities were generated
from the molecular ion and from loss of methyl. Four time-win-
dows were set for the multiple ion detection at a resolution of
10,000 (10% valley). The chromatograms were analyzed using the
software provided by the mass spectrometer manufacturer. Data
files of the integrated results were uploaded to a network for inclu-
sion into a database and further analysis with statistical software.
Results and discussion
Calibration and method detection limits
The GCMS analyses were performed with the deriva-
tized extract. The derivatization reduces the polarity of
the analytes and produces a number of benefits including
better detection sensitivity and no carry over between
runs. As shown in Fig. 1, each OH-PAH analyte was re-
solved for reliable quantification.
The carbon-13 labels were used to obtain response fac-
tors for each analyte. 1-nap(13C6) was used as the standard
for the hydroxynaphthols; 3-phen(13C6) referenced the hy-
droxyfluorenes and hydroxyphenanthrenes; 3-flra(13C6)
and 1-pyr(13C6) referenced their respective unlabeled com-
pounds, and 6-chr(13C6) was used for the calibration of the
hydroxychrysenes, hydroxybenz[a]anthracenes and hy-
droxybenzo[c]phenanthrenes. In this manner, true iso-
tope-dilution was obtained for five of the OH-PAH and
the remaining OH-PAH would be categorized as quanti-
fied by internal standard addition.
Calibration was performed by taking spiked water
samples through each step of the sample clean-up, includ-
ing enzymatic hydrolysis, and analyzing on the GCMS
instrument. Six calibration levels (10, 50, 150, 400, 700,
and 1200 ng L1) were used to generate calibration curves
having at least 8 points at each level. Blank levels were
observed for the naphthols, hydroxy-fluorenes, and hy-
droxy-phenanthrenes. By examining the reagents and any
potential carry-over from the GC instrument, the naphthol
Fig. 1 GCMS chromatogram of an extract from a spiked urine
sample. Peaks are identified by the numbers as listed in Table 1
background was believed to arise from environmental
sources inside and outside the laboratory (such as building
construction, road paving, etc.), and the blank levels of
the hydroxy-fluorenes and hydroxy-phenanthrenes were
thought to have been caused by ventilation of the high-
concentration standard solutions throughout the fume
hood system. A benefit of analyzing for the PAHm is that
they are much less prevalent in the environment than the
native PAH and most PAHm will have no blank levels in
the reagents.
The average blank level was subtracted from each cal-
ibration point to produce corrected calibration curves. By
fitting with a weighted linear regression algorithm, where
the weighting for each point was the inverse concentra-
tion, linear fits of the curves were good. The correlation
coefficients varied from 0.95 to 0.997, and the residual
value of the data points from the best-fit line indicated no
significant bias at either end of the calibration curve. The
curve for 3-chr produced the lowest correlation coeffi-
cient, and as discussed later, 3-chr has the highest varia-
tion in the quality control samples.
The method detection limits (MDL) for the OH-PAH
are presented in Table 1. Since the instrument sensitivity
was high enough to detect airborne and environmental
contamination for some of the OH-PAH, the MDL for an-
alytes with blank levels was obtained by adding the aver-
age blank level to 3 times the blank standard deviation.
MDL for the other OH-PAH were calculated according to
methods suggested by Taylor: extrapolating to obtain a
value for 3s0 (three times the standard deviation at zero
concentration) or by using three times the standard devia-
tion at a concentration level near the detection limit [24].
The detection frequency of 1-nap and 2-nap from pre-
vious reference range studies was near 85%, and for 1-pyr
the detection frequency was close to 7% for non-occupa-
tionally exposed samples [25, 26]. The method from Ref.
[23] had a detection limit near 70 ng L1 of 1-pyr, thus us-
ing the current isotope-dilution GCMS method with a

MDL of 2 ng L1, the detection frequency should increase
significantly. With a preliminary screening of urine sam-
ples using the current method, the OH-PAH from chry-
sene, benzo[c]phenanthrene, and benz[a]anthracene still
have low detection frequencies.
Recovery
Recoveries for the OH-PAH were calculated by compar-
ing peak areas for two sets of samples. The first set was
spiked with the OH-PAH and the carbon-13 labels, prior
to sample preparation. During sample preparation, the
dried extracts were reconstituted with 20 L toluene. The
second set of samples was spiked only with the OH-PAH
prior to sample preparation, and their dried extracts were
reconstituted with a 20 L toluene solution containing the
carbon-13 labels at the same amount as spiked into the
first set. The extracts from both sets were then derivatized
and analyzed on the GCMS system. Average recoveries
were calculated for the labels and then the recoveries for
the unlabeled OH-PAH could be determined. With seven
samples from each set, the average recovery went from 41
to 70% for 6-chr and 1-nap, respectively. Part of the loss
in the analyte recovery was most likely due to absorption
of the polar analytes to the SPE bed and components of
the urine matrix that may interfere with the extraction.
While the glassware was silanized, analyte adsorption could
occur to an extent affecting the recovery at ultra-trace lev-
els.
Quality-control samples
Urine donations from non-occupationally exposed indi-
viduals were collected and combined. From the combined
donations, pools of approximately 1 L were prepared at
target amounts for each individual OH-PAH. Due to back-
ground levels in the donated urine, several of the analytes
in the pools exceeded the target level. At the time of
preparation, we did not have a sufficient amount of
1-phen to add in each pool, thus data regarding this ana-
lyte in the QC pools were not available. After characteriz-
ing each pool with multiple analyses, control charts, con-
fidence limits, and coefficients of variation (CV) were
generated. Coefficients of variation and the calculated
amounts for the OH-PAH are listed in Table 1, and Fig. 2
contains a control chart for 1-pyr from a pool with a target
of 800 ng L1. The data reflect both inter-run and intra-run
variation over several weeks. The coefficient of variabil-
ity decreases as the analyst becomes familiar with the ex-
traction methodology.
The QC values for the 1-nap and 2-nap levels were sur-
prisingly low, since significant levels of these analytes
were expected in urine donations [25, 26]. By monitoring
the free OH-PAH in samples after varying hydrolysis
times, only the naphthols increased after three hours. The
QC values for the naphthols were slightly above the target
219
Fig. 2 Quality-control chart for 1-pyr from a QC sample with a
target value of 800 ng L1. The dashed lines correspond to 99%
confidence limits, the dotted lines correspond to 95% confidence
limits, and the dash-dotted line is the mean value
amounts; even after overnight (16 h) hydrolysis. It has
been noted earlier by Rhodes and Houston, that enzymatic
activity of beta-glucuronidase towards naphthol glu-
curonide is significantly different from the often used ac-
tivity towards phenolphthalein glucuronide [27]. In addi-
tion, we obtained QC samples from the non-persistent
pesticide group at the Centers for Disease Control and
Prevention, who analyze urine samples for 1-nap and
2-nap to determine exposure from carbaryl. Our analyses
of the pesticide QC material produced results for 1-nap
and 2-nap within the established 99% confidence range
for the material. Until the reason for the anomalously low
1-nap and 2-nap results from our QC samples has been
uncovered, levels of 1-nap and 2-nap from this SPE
GCMS protocol will not be reported.
From Table 1, the QC samples had mean levels of
2-fle, 3-fle, 2-phen, and 3-phen ranging from 1023 to
1092 ng L1. This range reflects contributions from both
the spiked amount of OH-PAH and the background levels
from the pooled urine. The mean QC values for 1-pyr,
3-fla, and 6-chr were near their target value, although
3-fla was lower than anticipated. The other OH-PAH that
used the carbon-13 labeled 6-chr as their internal standard
had levels near 59% of the targeted amount. The low val-
ues indicate that, despite apparent structural and chemical
similarity, the internal standard underestimated the recov-
ery. This underestimation most likely results from the
change in matrix of the calibration standards in water to
the urine matrix for the QC samples. Generating calibra-
tion data from a pre-screened urine pool for OH-PAH lev-
els may also provide better agreement between the tar-
geted and calculated QC values. With the pre-screened
urine samples, only those OH-PAH having levels near the
detection limit could be used for calibration in urine. The
need for isotopic labels has been shown to be critical in
providing accurate results, thus, several more carbon-13
labels are being synthesized for these analytes.
220
The CV for each analyte is listed in Table 1. For
metabolites that have their own labeled isotopic standard
(1-nap, 3-phen, 3-fla, 1-pyr and 6-chr), the CV values are
quite low and range from 1.7 to 5%. These values com-
pare well with those from other standard clinical methods.
The analytes that use another label as their internal stan-
dard have CVs from 5 to 20%, which is still quite accept-
able for an ultra-trace method from a urine matrix. The
highest CV was for 3-chr, which also had the lowest cor-
relation coefficient for its calibration curve. Despite being
a positional isomer of labeled 6-chr, the imprecision was
highest for 3-chr over the other structural isomers.
Summary and conclusions
An isotope-dilution GCMS ultra-trace method for the
quantification of the hydroxylated metabolites of PAH in
urine has been developed and validated. By targeting 2-,
3-, and 4-ring OH-PAH in urine, background detection of
most OH-PAH is possible. While the need for a method
that incorporates PAHm having greater than 4 fused rings
still exists, the reported protocol can be used as a clinical
assay for studies on the relationships between PAH expo-
sure and health outcomes. Portability of the method to
other laboratories may be an issue since access to a high
resolution mass spectrometer is not widespread, but use-
ful MDL should be obtained with triple quadrupole mass
spectrometers. The MDL with HRMS are in the low ng L1
level, and should ensure sufficient detection frequency for
exposure models. To establish a reference range for the
U.S. population, we are presently applying this method to
a set of NHANES (National Health and Nutrition Exami-
nation Survey) samples.
Acknowledgements Cheryl McClure of the Dioxin and Related
Compounds Group at the Centers for Disease Control and Preven-
tion deserves our earnest thanks for developing the data transfer
and database programming. The use of trade names is for identifi-
cation only and does not imply endorsement by the U.S. Depart-
ment of Health and Human Services or the Centers for Disease
Control and Prevention.
References
1. International Agency for Research on Cancer (1983) IARC
monographs on the evaluation of carcinogenic risk of chemi-
cals to man: polycyclic aromatic compounds, part 1, chemical
and environmental data, vol 32. IARC, Lyon
2. International Agency for Research on Cancer (1985) IARC
monographs on the evaluation of carcinogenic risk of chemi-
cals to man: tobacco smoking, vol 38. IARC, Lyon
3. Jongeneelen FJ, Anzion RBM, Henderson PTh (1987) J Chro-
matogr 413:227
4. Frdric D, Haguenoer J-M, Zmirou D, Empereur-Bissonnet P,
Jongeneelen FJ, Nedellec V, Person A, Ferguson C, Dab W
(2000) J Occup Environ Med 42:391
5. Moen BE, vreb S (1997) J Occup Environ Med 39:515
6. Siwinska E, Mielzynska D, Smolik E, Bubak A, Kwapulinski J
(1998) Sci Total Environ 217:175
7. Gndel J, Schaller KH, Angerer J (2000) Int Arch Occup Envi-
ron Health 73:270
8. Gndel J, Angerer J (2000) J Chromatogr B 728:47
9. Gmeiner G, Krassnig C, Schmid E, Tausch H (1998) J Chro-
matogr B 705:132
10. Korfmacher WA, Djuric Z, Fifer EK, Beland FA (1988) Spec-
trosc Int J 6:1
11. Chaung JC, Callahan PJ, Lyu CW, Wilson NK (1999) J Expos
Anal Environ Epidemiol 2:85
12. Jacob J, Grimmer G (1987) Rev Anal Chem 9:49
13. Grimmer G, Jacob J, Dettbarn G, Naujack K-W, Heinrich U
(1995) Exp Toxicol Pathol 47:421
14. Jacob J, Grimmer G (1996) Centr Eur J Publ Hlth 4 Suppl 33
15. Grimmer G, Jacob J, Dettbarn G, Naujack K-W (1997) Int
Arch Occup Environ Health 69:231
16. Strickland P, Kang D, Sithisarankul P (1996) Environ Health
Perspect Suppl 5 104:927
17. Strickland PT, Kang D, Bowman ED, Fitzwilliam A, Downing
TE, Rothman N, Groopman JD, Weston A (1994) Carcinogen-
esis 15:483
18. Strickland P, Kang D (1999) Toxicol Lett 108:191
19. Lee C-K, Cho S-H, Kang J-W, Lee S-J, Ju Y-S, Sung J, Strick-
land PT, Kang D (1999) Toxicol Lett 108:209
20. Ramasamy SM, Hurtubise RJ, Weston A (1997) Appl Spec-
trosc 51:1377
21. Xu X, Hurtubise RJ (1999) J Chromatogr A 829:289
22. Angerer J, Mannschreck C, Gndel J (1997) Int Arch Environ
Health 70:365
23. Jongeneelen FJ (1997) Sci Total Environ 199:141
24. Taylor JK, (1987) Quality assurance of chemical measure-
ments. CRC Press, Boca Raton, FL, chap 9
25. Hansen AM, Christensen JM, Sherson D (1995) Sci Total Envi-
ron 163:211
26. Hill RH, Head SL, Baker S, Gregg M, Shealy DB, Bailey SL,
Williams CC, Sampson EJ, Needham LL (1995) Environ Res
71:99
27. Rhodes JC, Houston JB (1981) Xenobiotica 11:63
Anal Bioanal Chem (2002) 372 : 221229
DOI 10.1007/s00216-001-1183-9
O R I G I N A L PA P E R
Piero Mirti Patrizia Davit Monica Gulmini
Colourants and opacifiers in seventh and eighth century glass
investigated by spectroscopic techniques
Received: 23 July 2001 / Revised: 24 September 2001 / Accepted: 13 October 2001 / Published online: 12 December 2001
Springer-Verlag 2001
Abstract Glass fragments dating from the seventh and
eighth century AD were excavated in the Crypta Balbi in
Rome. They were studied to detect agents involved in
colour development and opacification. Reflectance spec-
tra recorded on powdered samples revealed the contribu-
tion of Fe(II), Fe(III), Mn(III), Cu(II), and Co(II) ions in
determining colour hues. The effect of the Mn/Fe atomic
ratio on glass colour is discussed. It is apparent that me-
dieval glassmakers in Italy could obtain a wide range of
colours by exploiting the presence of iron and manganese
as contaminants of sand and flux and controlling the
amount of oxygen let into the furnace. X-ray diffraction
and scanning electron microscopy coupled with energy-
dispersive X-ray analysis were used to study opaque frag-
ments. The presence of calcium antimonate was detected
in white, blue, and blue-green fragments, and elemental
copper was detected in a red glass.
Keywords Ancient glass Colourants Opacifiers
Reflectance spectroscopy x-raydiffraction SEM-EDS
Archaeometry
Introduction
Coloured and opacified glass has been produced since an-
cient times. Glass colour can be determined by the pres-
ence of metal ions dissolved within the glass matrix,
which absorb visible light as a result of electronic transi-
tions [1]. The most common colouring agents present in
ancient glass are Fe(II), Fe(III), Mn(III), Cu(II), and
P. Mirti () P. Davit M. Gulmini
Dipartimento di Chimica Analitica,
via P. Giuria 5, 10125 Torino, Italy
e-mail: mirti@ch.unito.it
Present address:
P. Davit
Dipartimento di Chimica IFM, Universit di Torino,
via P. Giuria 7, 10125 Torino, Italy
Co(II) ions. In some instances glass colour can arise from
light scattering, because of the presence of very minute
particles dispersed in the glass matrix; as an example, trans-
parent ruby glass is coloured by gold or copper particles
[1]. Depending on their size, particles which segregate
from a transparent matrix can cause opacity of the glass
elemental copper or copper(I) oxide particles are found in
red opaque glass [2, 3]. Apart from these, opacifying
agents exploited since ancient times include calcium anti-
monate and tin dioxide (which generate a white opaque
glass), and lead antimonate and lead stannate (which gen-
erate a yellow opaque glass) [1, 4].
Fe(II) and Fe(III) ions are the most frequent colouring
agents in ancient glass, because iron is normally present
in a glass batch as a contaminant of sand; otherwise, addi-
tional iron can be added intentionally to enhance colour
saturation. Depending on the relative amount of Fe(II)
and Fe(III) ions, different colour hues can be obtained [1,
5, 6]. Glassmakers could, therefore, produce glass of dif-
ferent colours by modifying the redox conditions which
determine electron transfer between the two iron oxida-
tion states. This could be achieved by controlling the
amount of oxygen present in the furnace, or by exploiting
the presence of other redox systems in the batch. In this
respect the most important equilibria are those involving
manganese ions, which could be established as a conse-
quence of the introduction of this element either occasion-
ally, as a contaminant of raw ingredients, or intentionally.
Oxidation of Fe(II) to Fe(III) by oxygen, Mn(III) ions, or
other suitable agents (for instance Sb(V)), causes a colour
shift from bluegreen towards green, yellowgreen, and
yellow. Fe(II) renders a glass blue whereas Fe(III) colours
it pale yellow and the combined presence of the two ions
in different ratios determines the actual tint of the glass. In
addition, a colourless or nearly colourless glass can be ob-
tained when iron is prevalently present in its oxidised
form and its total concentration is low, because of the very
low absorption coefficient of Fe(III) ions; in such circum-
stances the presence of Mn(III) in moderate excess can
add to the colourless effect, because its purple hue coun-
terbalances the pale yellow hue of Fe(III). Finally, a pur-
222

ple glass is generated when Mn(III) is present in greater flectance spectroscopy is the considerable diminution of
excess. colour saturation, and its dependence on sample grain
Apart from iron and manganese, copper and cobalt size. Hue, however, is not affected by sample grinding.
were important colouring agents in the ancient glass in- Opacifying agents can be identified by X-ray diffrac-
dustry. Both Cu(II) and Co(II) ions generate a blue glass, tion, because their crystal structure generates peculiar dif-
but the former determines in a rather greenish hue and the fraction patterns. X-ray powder diffractograms can be
latter a rather purplish hue. Co(II) ions, moreover, have a recorded using of the same powder used for reflectance
definitively higher absorption coefficient than Cu(II) ions, spectroscopy, because both techniques are analytically
so cobalt still plays a major role in colour development non-destructive. Whenever powders are difficult to ob-
even when its concentration is significantly lower than tain, diffraction patterns can be recorded on small samples
that of copper, as is frequently found in blue ancient glass. by use of a microdiffractometer. Apart from minute items,
Occasionally the ubiquitous presence of iron can some- which do not provide enough sample for X-ray powder
how affect the actual hue of a blue glass, in particular diffraction (XRPD), microdiffractometry (MXRD) is par-
when copper is the main blue colourant. ticular useful for studying non-destructively glass frag-
Calcium antimonate, either Ca2Sb2O7 or CaSb2O6, was ments bearing particular decorative features, for example
the first opacifying agent used in glass production [1, 4]. vessels decorated with filaments of opaque glass.
It is formed by reaction of an antimony-containing ingre- Scanning electron microscopy (SEM) is another tech-
dient with calcium present in a soda-lime matrix and nique which is successfully used to study ancient glass.
causes the development of a white opaque glass. Since By taking images in the secondary electron (SE) or back-
Roman times antimony has gradually been replaced by scattered electron (BSE) mode one can observe particles
tin, which forms tin dioxide as a white opacifier, but its dispersed within the glass matrix. Coupling with an en-
use is still documented at the end of the first millennium ergy-dispersive detector (SEMEDS) also enables the
AD [4, 7]. Both antimony and tin could be used to pro- analysis of selected areas of the glass sample. Because of
duce a yellow opaque glass, when added to a lead glass, their small size (down to 1 m or even less), opacifying
because of the precipitation of yellow lead antimonate particles might not always be reliably analysed quantita-
(Pb2Sb2O7) and lead stannates (Pb2Sn2O4 or PbSnO3) re- tively, because the electron beam can also excite X-ray
spectively [4]. Other colours could be obtained in opaque emission in the surrounding matrix. Usually, however, the
glass as a result of the joint effects of opacifying and
colouring agents; for instance, blue and green opaque Table 1 Date, form, and colour of the studied fragments
glass can be obtained in the presence of Cu(II) with a
white or yellow opacifier, respectively. Finally, a red Sample Century Typology Colour
opaque glass is obtained by segregation of elemental cop- CB1 7th Foot of chalice Blue
per or Cu(I) oxide particles. CB4 7th Foot of chalice Blue
UVvisibleNIR spectroscopy, X-ray diffraction, and CB7 7th Foot of chalice Blue-green
electron microscopy are suitable techniques for studying CB10 7th Foot of chalice Green
colourants and opacifiers in glass. UVvisibleNIR spec- CB13 7th Foot of chalice Yellow
troscopy enables assignment of absorption bands to spe- CB15 7th Rim of lamp Blue-green
cific colouring agents [6, 8, 9, 10, 11]. Fe(II) ions gener- CB18 7th Base of lamp Yellow
ate a wide absorption band centred between 1050 and CB19 7th Base of lamp Yellow-green
1100 nm, with a tail in the visible region, and Fe(III) ions CB21 7th Base of lamp Green
are characterised by absorption at approximately 380 nm. CB24 7th Window pane Colourless
In addition, Fe(III) also gives bands at approximately 420 CB26 7th Window pane Yellow-green
and 435 nm, where bands from Mn(II) may also occur. In CB29 7th Window pane Green
contrast, Mn(III) is recognisable from characteristic ab- CB31 7th Decorative inlay Blue, translucent
sorption at approximately 490500 nm. Cu(II) ions are CB33 7th Decorative inlay Blue-green, opaque
recognised by a band centred at approximately 800 nm, CB34 7th Decorative inlay White, opaque
and Co(II) gives a characteristic triplet between 530 and CB35 7th Decorative inlay Pale blue, opaque
650 nm; the same ion also gives absorption bands in the CB36 7th Decorative inlay Red, opaque
NIR region, at approximately 1250, 1490 and 1750 nm. CB41 8th Foot of chalice Red-purple
Studies on ancient glass fragments generally deal with CB42 8th Foot of chalice Blue
samples with curved surfaces, of irregular thickness, CB43 8th Foot of chalice Blue-green
and/or with weathering products, which prevent the cor- CB45 8th Foot of chalice Yellow
rect collection of transmitted light for recording electronic CB46 8th Foot of chalice Blue-green
spectra. This problem is easily overcome by collecting re- CB47 8th Foot of chalice Green
flected instead of transmitted light, which also enables the CB49 8th Rim of chalice Green
recording of spectra for samples of opaque glass. Re- CB51fil 8th Chalice decoration White, opaque
flectance spectra may be recorded on abraded surfaces, or, CB52 8th Rim of lamp Green
when the form of the glass is complex, on a powdered CB53 8th Rim of lamp Blue
sample. The major drawback of grinding samples for re- CB56 8th Window pane Yellow-green

detection of key elements can suffice to disclose the na-
ture of the opacifier, which can be confirmed by X-ray
diffraction. In this respect, a positive aspect of the joint
use of SEMEDS and MXRD is the possibility of using
the same sample for both kinds of investigation.
This paper reports the use of reflectance spectroscopy,
X-ray diffraction, and SEMEDS to study colourants and
opacifiers in glass fragments dating from the seventh and
eighth century AD. The fragments were excavated in the
exedra of the Crypta Balbi in Rome [12, 13]. Transparent
and translucent glass included fragments of chalices,
lamps, window panes, and inlays; they were blue, blue
green, green, yellowgreen, yellow, redpurple or colour-
less. Opaque glass was represented by fragments of deco-
rative inlays, coloured white, pale blue, bluegreen or red.
Some of the eighth century chalices and lamps were,
moreover, decorated with filaments of white opaque glass;
a tiny fragment was sampled from one of these for study-
ing the opaque decoration.
Table 1 gives date, form, and colour of the studied frag-
ments.
Experimental
Reflectance spectra
Small pieces with a total weight of a few hundred milligrams were
taken from each archaeological fragment by means of a diamond
coated saw, after removal of any external layer by use of a dia-
mond-coated wheel. The pieces were washed twice in an ultra-
sonic bath with twice-distilled water, dried at 120 C and ground to
a fine powder in an agate mortar. Powders were inserted into cylin-
drical cells and pressed against an optical glass or quartz window
to expose a reproducible surface to the light beam of the spectrom-
eter. Reflectance spectra in the range 400700 nm were first
recorded by use of a Minolta CM-508i portable spectrometer to
obtain colour coordinates. This was equipped with a pulsed xenon
arc lamp and a barium sulfate integrating sphere to illuminate the
sample surface diffusely. The light reflected by the sample (specu-
lar component included) was detected by means of a silicon photo-
diode array, with wavelength pitches of 20 nm. Spectra were
processed by means of the Minolta SpectraMagic software, which
enables choice among several illuminants, use of either the CIE 2
or 10 standard observer, and expression of colour coordinates in
different systems. The illuminant D65 (average solar light), a 10
viewing angle and the CIEL*a*b* colour system were used in this
work. Results were the mean of three measurements in succession.
UVvisibleNIR reflectance spectra were recorded, for further
probing into colour mechanisms, by means of a Varian Cary
5 spectrophotometer equipped with a polytetrafluoroethylene inte-
grating sphere and providing a continuous scan between 200 and
2500 nm. The geometry used excluded the specular component of
the reflected light, detecting the diffusely-reflected light only.
SEMEDS examination of opaque fragments
A small chip was taken from opaque fragments. This was mounted
in epoxy resin and polished by use of a Buehler Minimet 100 in-
strument, with silicon carbide papers (500 and 1200 grit) and dia-
mond paste (6 and 1 m), in sequence. The polished samples were
coated with a graphite layer, by use of an Emitech K 950 appara-
tus. SEMEDS investigation was performed by means of a Cam-
bridge Instrument Stereoscan 360 scanning electron microscope
coupled with a Link Analytical QX-2000 energy-dispersive sys-
tem. The operating conditions were: acceleration voltage 15 kV,
223
filament current 2.43.0 A, working distance 25 mm. SEM images
were obtained by use of either secondary or back-scattered elec-
trons. EDS analyses were performed on opaque inclusions, by fo-
cusing the electron beam on 5-m diameter areas; as already men-
tioned, this usually enabled the identification of the opacifier by
detection of key elements. Analytical data were obtained by use of
the Oxford QX 2000 software and a ZAF-4 correction program,
using pure elements or oxides as standards. Cobalt was used for in-
strument calibration.
X-ray powder diffraction and X-ray microdiffraction
These techniques were used to detect crystalline phases in the
opaque fragments. XRPD was performed with a Siemens D5000
diffractometer, with graphite-monochromatised copper radiation
and collecting patterns in the 550 2 range. MXRD was per-
formed on the same samples prepared for the SEMEDS study, af-
ter removal of the graphite layer with silicon carbide paper. Dif-
fractograms were obtained by use of a Rigaku PSPC/MDG dif-
fractometer equipped with a rotating copper anode; X-rays were
collimated on to 100 m diameter areas, and patterns were ob-
tained in the 2080 2 interval.
Results and discussion
Composition of the fragments
The composition of the studied fragments (21 elements)
was obtained previously by inductively coupled plasma
optical emission spectroscopy (ICPOES) and is reported
elsewhere [12, 13]. All the fragments are soda-lime glass;
a particular feature of some of them is a fairly high cop-
per, antimony and/or lead content, irrespective of colour
and opacity. The composition of sample CB51fil was not
determined, because of the limited size of the opaque glass
decoration. As concerns samples CB33, CB35, and CB36,
analyses performed by SEMEDS suggested to verify the
data previously obtained for the phosphorus content.
These glasses were then analysed anew by ICP-OES ac-
cording to the procedure reported in [13] and the results
indicated that previous data suffered from copper interfer-
ence; actually, P2O5 levels below 0.2% were found in
these three samples.
Colourants
Figure 1 gives the colour of the samples considered (with
the exception of CB51fil), projected on the a*b* plane of
the CIEL*a*b* colour space. As usually found with an-
cient soda-lime glass, most of the samples are in the NW
quadrant of the diagram, which corresponds to hues rang-
ing from bluegreen to yellow, through green and yel-
lowgreen. Other colours are blue, red, and redpurple.
Iron and manganese contents, and their relative
amounts, are highly variable among the fragments consid-
ered. Relatively low concentrations of both elements
(<1%) are mainly found in blue and bluegreen samples;
for these both elements might have entered the batch as
contaminants of the main ingredients. Low iron and man-
ganese, and low Mn/Fe ratios, are, however, also found in
224
Fig. 1 Projection of the points representing glass fragments on the
a*b*plane of the CIEL*a*b* colour space. Symbols: circles, chal-
ices; triangles, lamps; squares, window panes; diamonds, inlays;
open symbols, seventh century fragments; closed symbols, eighth
century fragments
some green samples. High Mn/Fe ratios are usually found
in yellow glasses and in the redpurple sample, whereas
some yellowgreen glasses are characterised by high iron
and low Mn/Fe ratios.
It is apparent from Fig. 2 that increasing the Mn/Fe ra-
tio usually results in a shift from bluegreen towards
green and yellowgreen hues, and that, apart from the
redpurple glass, the higher ratios are observed for yellow,
colourless, and opaque-white fragments. This accords with
the progressive increase of the relative amount of Fe(III)
relative to Fe(II) because of oxidation of the latter by
Mn(III). For sample CB41, the large excess of manganese
means that enough Mn(III) remains in the glass to colour
it red-purple. In contrast to this general trend, some green
(CB10, CB47, CB21) and yellowgreen samples (CB19,
CB26, CB56) are characterised by relatively low Mn/Fe
ratios (<1); here proper control of the atmosphere of the
furnace might have enabled glassmakers to obtain all the
same glasses characterised by the presence of a significant
amount of Fe(III). The generally small amount of anti-
mony, and the low Sb/Fe ratio, would exclude that Sb(V)
might have significantly contributed to iron oxidation in
these glasses.
Reflectance spectra show that different mechanisms
can contribute to the development of similar hues in
glasses containing different amounts of iron, manganese,
and copper. For bluegreen fragments (Fig. 3 and Ref.
[12]) intense absorption due to Fe(II) ions is the most
characteristic feature of samples CB7 and CB15; the dif-
ferent contribution by Fe(II) and Fe(III) ions, as recognis-
able in the spectra, determines the more bluish hue of
fragment CB7 and the more greenish hue of CB15. Sam-
ples CB43 and CB46 are intermediate in hue between the
first two, but their spectra are different from each other
Fig. 2 Dependence of Mn/Fe atomic ratio on Fe2O3 concentration.
Colours of the fragments: squares, blue; circles, blue-green; trian-
gles, green; inverted triangles, yellow-green; diamonds, yellow;
dotted squares, red; dotted diamonds, red-purple; hexagons,
colourless or white; open symbols, seventh century fragments;
closed symbols, eighth century fragments
and from those of CB7 and CB15. The spectrum of CB46
is dominated by the absorption band of Fe(II), whereas
the contribution of Fe(III) is extremely small; peculiar to
this spectrum is the Fe(III) pattern between 380 and 440
nm the intensities of all three expected bands are similar
(the inner one barely appreciable). In contrast, the spec-
trum of CB43 does not reveal a similarly high contribu-
tion by Fe(II) ions, while Fe(III) determines a signifi-
cantly higher absorption. In this glass the CuO concentra-
tion is 0.87% and the absorption band of Cu(II) ions con-
tributes substantially to the bluegreen hue. A similar sit-
uation is observed for the opaque inlay CB33, for which
more intense bands are observed for both Fe(III) and
Cu(II) (here, CuO amounts to 1.56%). The same spectral
features are observed for another opaque inlay, CB35 [12],
but a higher CuO content (2.85%) and a higher Cu/Fe ra-
tio render the glass pale blue rather than bluegreen.
A further example of how copper ions can affect glass
colour is given by sample CB49 (Fig. 4), where a still dif-
ferent Cu/Fe ratio renders the glass green. Thus, progres-
sive diminution of the Cu/Fe atomic ratio (5.57, 2.20,
1.10, and 0.84 in CB35, CB33, CB43, and CB49, respec-
tively) determines a shift from blue to bluegreen and
green. For the other green samples (Fig. 4 and Ref. [12]),
the spectrum of CB21 also shows a contribution by Cu(II)
 225
Fig. 3 Reflectance spectra of
bluegreen samples

Fig. 4 Reflectance spectra of green

samples

Fig. 5 Reflectance spectra of yellow

and yellowgreen samples
226
Fig. 6 Reflectance spectra of blue
fragments
ions (present as 0.25% CuO), but further reveals the pat-
tern of Co(II) ions. This pattern is also appreciable in the
spectra of CB52, where the band of copper (present as
0.30% CuO) is masked by the Fe(II) band.
A typical spectrum of a glass coloured green by iron is
shown by CB29, with bands of similar intensity for Fe(II)
and Fe(III) ions. Still different, and more similar to the
spectrum of the blue-green fragment CB46, are those of
CB10 and CB47. Here again there is intense absorption
due to Fe(II), with the pattern between 380 and 440 nm of
significantly lower intensity. The greater contribution of
the Fe(III) pattern compared with CB46 would account
for the development of the green hue.
The spectra of the yellow-green fragments CB19, CB26,
and CB56 (Fig. 5 and Ref. [12]) are all dominated by the
bands of the Fe(II) and Fe(III) ions, and resemble that of
the green fragment CB29. The increased contribution by
Fe(III) ions relative to Fe(II), however, accounts for the
development of a yellowgreen rather than a green hue.
The spectra of the yellow glasses (Fig. 5) are finally
characterised by still more prevalent Fe(III) absorption
relative to Fe(II). The colour of fragment CB18 is more
shifted towards orange hues compared with CB13 and
CB45, and this might be ascribed to the absence of appre-
ciable absorption by Fe(II) ions. Similar to those of the
yellow fragments are the spectra (not reported) of CB24,
colourless, and CB34, opaque white; here a low total iron
content, and the prevalence of the oxidized form, account
for the colourless effect an opaque white, which turns into
in CB34 because of the presence of calcium antimonate
(see below).
As observed above, cobalt and/or copper ions can
modify the hue of a glass mainly coloured by Fe(II) and
Fe(III). In this context, it is worth noting that the typical
pattern of Co(II) is discernible in a large number of the
recorded spectra; this is because of the high absorption
coefficient of this chromophore, which enables recogni-
tion of its pattern even when the element could not be de-
tected by ICPOES analysis.
As for copper, relatively high levels in the bluegreen
and green glasses examined are frequently matched by
relatively high levels of both antimony and lead. It seems
plausible that copper was not added intentionally to affect
the final hue, but that it might be present as a consequence
of the addition of mosaic tesserae as recycled cullet to the
batch [13, 14, 15]. Indiscriminate use of white, blue, yel-
low, and green tesserae might have caused the contempo-
rary presence of copper, antimony and lead in the final
glass, and Cu(II) ions to affect the resulting colour. In this
context, one might further consider that Sb(V), present as
a white or yellow opacifier in the recycled tesserae, might
have added its contribution to those of manganese and/or
oxygen in oxidising Fe(II) to Fe(III).
Apart from sample CB35 discussed above, spectra of
blue glasses (Fig. 6 and Ref. [12]) are generally charac-
terised by the prevalent contribution of Co(II) relative to
Cu(II) ions, despite the lower concentration of the former;
only the spectrum of CB53 reveals a comparable contri-
bution of the two chromophores. The cobalt content of
blue fragments ranges from approximately 0.01% (CB1
and CB4) to 0.11% (CB31) CoO (apart from CB35, in
which it was not detected), whereas copper content
mainly varies between 0.10% (CB1) and 0.92% (CB53)
CuO (but reaches 2.85% in CB35). Its higher absorption
coefficient makes the Co(II) ion a more effective chro-
mophore than the Cu (II) ion; the increase of the copper to
cobalt ratio, however, usually causes a shift from more
purplish towards more greenish hues. Fe(III) can further
contribute to the more greenish hue of some samples.
Fig. 7 finally shows the spectra of the red and redpur-
ple samples. The latter is dominated by the intense ab-
sorption band of Mn(III) ions. The former is characterised
by the removal of the wavelengths below 600 nm as a re-
sult of light scattering by copper particles (see below);
spectral features between 800 and 1200 nm also suggest
the presence of Fe(II) and Cu(II) ions. Fe(II) ions would
have formed from metal iron added to create reducing
conditions in the system (see below), but the presence of
 227
Fig. 7 Reflectance spectra of red and
redpurple samples

some Cu(II) would indicate that copper reduction was not

complete.

Opacifiers

The studied opaque samples include two white, one blue-

green, one pale blue, and one red glass. One of the two
white samples (CB51fil) consisted of filaments of opaque
glass decorating a yellowgreen, almost colourless glass.
Back-scattered electron images of glass CB34 and of the
white filament of glass CB51 reveal the presence of inclu-
sions generally less than 10 m across (Fig. 8). EDS spec-
tra indicated the presence of calcium and antimony in
these particles, suggesting they are calcium antimonate.
The actual nature of the white opacifier was revealed by
X-ray diffraction, which unambiguously indicated that
calcium antimonate in the form CaSb2O6 was present in
both samples; in addition, the pattern of the form
Ca2Sb2O7 was also particularly evident in the diffrac-
togram of CB51fil. Both XRPD and MXRD were used on
CB34, but only MXRD was used on CB51fil, because of
the limited amount of opaque glass present on the frag-
ment. Air bubbles, with sizes up to more than 100 m
across, were also observed in SEM images of both
glasses.
As in the case of the white samples, inclusions gener- Fig. 8 Back-scattered electron images of samples CB34 (1000)
ally less than 10 m across were observed in SEM images (above) and CB51 (100) (below) showing inclusions of calcium
of CB33 and CB35 (Fig. 9); these were found to contain antimonate and air bubbles
Ca and Sb by EDS analysis; X-ray diffraction confirmed
the presence of CaSb2O6 as the main white opacifier. As
already said, these samples are both coloured by Cu(II)
ions together with Fe(III) in different relative amounts; in cated variable amounts of silicon, aluminum, calcium,
both cases the colour saturation expected on the basis of and sodium, suggesting they are devitrification products.
the high copper concentration is reduced by the white X-ray diffraction proved unable to detect these products,
opacifier. Air bubbles, up to several hundred micrometers probably because their concentration is below the detec-
across, were also present in these glasses. Inclusions of ir- tion limit of the technique. Both devitrification products
regular shape were also observed; these appear darker or and air bubbles contribute to the overall opacifying effect
brighter than the glass matrix in BSE images. EDS indi- [4].
228
Fig. 9 Back-scattered electron images of samples CB33 (500)
(above) and CB35 (200) (below), showing inclusions of calcium
antimonate, air bubbles and devitrification products. A particle of
tin dioxide is visible in the upper right part of the CB33 image
BSE images of CB33 and CB35 also revealed sporadic
bright particles, which EDS analysis suggested were in-
clusions of tin dioxide. Because of its low concentration,
however (tin content amounts to 0.09 and 0.14% SnO2 in
CB33 and CB35, respectively), this compound was not
detected by X-ray diffraction. This low tin content makes
it unlikely it was intentionally added for opacifying pur-
poses, even though its presence might add to the overall
opacifying effect. It seems more plausible that tin entered
the glass batch with copper, if bronze powder or scraps
were used to introduce the latter as a colouring agent. Tin
to copper weight ratios (approximately 0.06 and 0.05 in
CB33 and CB35, respectively) might support this hypoth-
esis.
Red opaque glass is known to owe both colour and
opacity to the presence of minute particles of elemental
copper or Cu(I) oxide within the glass matrix [2, 3]. SEM
images of fragment CB36 revealed the presence of very
small spherules less than 1 m across, dispersed in a ho-
mogeneous matrix (Fig. 10); these particles were found to
contain copper by EDS and their spherical form strongly
suggests they are elemental copper, because Cu(I) oxide is
most often found in the form of dendritic aggregates [2,
3]. This was confirmed by both XRPD and MXRD, which
gave a diffraction pattern matching that of elemental cop-
per. The size of the copper particles accounts for light
Fig. 10 Back-scattered electron image of sample CB36 (10,000)
showing inclusions of elemental copper
scattering, which generates the red colour while rendering
the glass opaque.
Whether copper had entered the glass as recycled
bronze, a Cu(II) mineral, or a blue frit coloured by Cu(II),
reducing conditions had to be created in the glass batch to
ensure the formation of elemental copper and avoid the
formation of a blue transparent glass. Tin, antimony, and
iron in proper oxidation states could be used as reducing
agents to achieve this purpose [2, 16]. In sample CB36
this role was almost certainly played by iron, because of
its high concentration (2.19% Fe2O3, with an iron-to-cop-
per atomic ratio of 1.05), and the low level of both anti-
mony and tin. The presence of Cu(II) in the glass matrix,
as suggested by the reflectance spectra (see above), with
tin undetected by ICPOES, would indicate the use of a
mineral or frit containing Cu(II), rather than recourse to
recycled bronze.
Conclusions
The identity of colouring and opacifying agents in seventh
and eighth century glass fragments has been revealed by
the joint use of UVvisibleNIR reflectance spectroscopy,
X-ray diffraction and SEMEDS investigation. It has been
found that early medieval glassmakers in Italy were skilful
enough to obtain a wide range of coloured glass by ex-
ploiting the presence of iron and manganese as contami-
nants of sand and flux, and controlling the amount of oxy-
gen let into the furnace. Occasionally, introduction of ad-
ditional manganese and/or iron could enable them to ob-
tain particular colour hues or to increase colour saturation.
Copper was found in some blue-green and green frag-
ments, in which it can affect the colour hue, but it is un-
likely that it was intentionally added for this purpose. In
contrast, because of the contemporary occurrence of lead
and/or antimony in these glasses, it is possible that opaque
glass, in particular mosaic tesserae, was recycled while
melting the final batch. Indiscriminate use of white, yel-
low, green, and blue tesserae might account for the pres-
ence of all three elements in some of the studied samples.

Calcium antimonate has been found as the main opaci-
fier in white, blue, and bluegreen glass. Its presence at a
rather late date during the first millennium AD suggests
continuity with Roman glass production; in fact, at this
date tin dioxide was already used as an alternative opaci-
fying agent, even though use of calcium antimonate has
been documented down to at least the end of the millen-
nium. Air bubbles and devitrification products add to the
overall opacifying effect, and this might account for the
concentration of antimony, which is not particularly high
compared with levels in other ancient opaque glasses. In-
deed, antimony could derive from more ancient opaque
cullet; in contrast with transparent glasses, however, lead
was found at a significant level in only one of the exam-
ined opaque fragments, and copper only in the blue and
bluegreen inlays, to which it was added as a colouring
agent. The occurrence of tin in the latter two fragments,
and the tin to copper ratios observed, might suggest that
glassmakers used bronze scraps to introduce copper into the
melting batch.
A red opaque glass has been found to owe both colour
and opacity to the presence of small particles of elemental
copper, less than 1 m across. This indicates that glass-
makers well knew how to create strongly reducing condi-
tions; in this instance these were almost certainly obtained
by adding iron to the final batch.
Acknowledgements Thanks are given to Dr Lucia Sagu (Dipar-
timento di Scienze Storiche, Archeologiche e Antropologiche
dellAntichit, Universit di Roma La Sapienza) for providing
the glass fragments and discussing their archaeological features.
The authors are also grateful to Mr Dario Vaudan (Soprintendenza
ai Beni Culturali, Aosta) for the MXRD analyses and to Professor
Giacomo Chiari (Dipartimento di Scienze Mineralogiche e Petro-
logiche, Universit di Torino) for XRPD analysis; to Dr Emanuele
Costa, Dr Raffaella Ruffini, and Mr Carmelo Sibio (Dipartimento
di Scienze Mineralogiche e Petrologiche, Universit di Torino) for
help in the use of the SEMEDS equipment. Financial support
from the Italian National Council of Researches (CNR) and from
229
the Italian Ministry for University and Scientific Research
(MURST) is acknowledged.
References
1. Newton R, Davison S (1996) Conservation of glass. Butter-
worth-Heinemann, Oxford
2. Freestone IC (1987) Composition and microstructure of early
opaque red glass. In: Bimson M, Freestone IC (eds) Early vit-
reous materials. British Museum, London, pp 173191
3. Brill RH, Cahill ND (1988) J Glass Stud 30:1627
4. Henderson J (1985) Oxford J Archaeol 4:267291
5. Newton RG (1978) Glass Technol 19:5960
6. Sellner C, Oel HJ, Camara B (1979) Glastechn Ber 52:255264
7. Freestone IC (1993) Compositions and origins of glasses from
Romanesque champlev enamels. In: Stratford N. (ed) Cata-
logue of medieval enamels in the British Museum. British Mu-
seum, London, vol II, pp 3745
8. Sanderson DCW, Hutchings JB (1987) Glass Technol 28:99
105
9. Mirti P, Ferrari RP, Laurenti E, Casoli A (1993) Spectrochim
Acta 49A:13611371
10. Orlando A, Olmi F, Vaggelli G, Bacci M (1996) Analyst
121:553558
11. Bartecki A, Burgess J (2000) The colour of metal compounds.
Gordon and Breach Science Publisher, Amsterdam, chap 8,
pp 195209
12. Mirti P, Lepora A, Sagu L (2000) Archaeometry 42:359374
13. Mirti P, Davit P, Gulmini M, Sagu L (2001) Archaeometry
43:491502
14. Freestone IC (1992) Theophilus and the composition of me-
dieval glass. In: Vandiver PB, Druzik JR, Wheeler GS, Free-
stone IC (eds) Materials issues in art and archaeology III. Ma-
terials Research Society, Pittsburgh, pp 739745
15. Henderson J (1995) Le verre de Dorestad. Continuit tech-
nologique ou innovation? In: Foy D (ed) Le verre de lantiquit
tardive et du haut Moyen Age. Typologie-chronologie-diffu-
sion. Muse Archologique Dpartemental du Val dOise,
Guiry-en-Vexin, pp 5155
16. Cable M, Smedley JW (1987) The replication of an opaque red
glass from Nimrud. In: Bimson M, Freestone IC (eds) Early
vitreous materials. British Museum, London, pp 151156
Anal Bioanal Chem (2002) 372 : 230231
DOI 10.1007/s00216-001-1171-0
I N D U S T RY N E W S
Ralph G. Weyandt
The role of modern biochemical methods in the determination
of hygienically relevant microorganisms in foodstuffs and (drinking) water
Published online: 11 December 2001
Springer-Verlag 2001
Abreviations BCYE buffered charcoal yeast extract
BSE bovine spongiform encephalopathy ELISA enzyme-
linked immunosorbent assay GMO genetically modified
organism GVPC BCYE-agar with glycine and three
antibiotics (vancomycin, polymyxin, cycloheximide)
LIA lumino-immuno-assay PCR polymerase chain
reaction
In the field of drinking water and the production and stor-
age of foodstuffs a variety of bacteria and fungi play a
considerable role with regard to reduction of quality and,
even more, with regard to health risks. Various species of
fungi produce so-called mycotoxins in cereals that have
been harvested and stored. These mycotoxins are suspected
of causing tumours in man and animals. Fresh egg prod-
ucts, fresh mayonnaise, and desserts containing egg can
suffer from a mass increase in Salmonella, which cause
fatal infections. Types of cheese contaminated with Liste-
ria have repeatedly led to deaths. Hygienic strains, viruses,
and pathogenic protozoa, which are responsible at least lo-
cally for severe epidemics, are frequently spread via drink-
ing water, the most important foodstuff. The list of exam-
ples of serious microbial contamination of foodstuffs and
drinking water is very long and, as media reports show, is
increasing almost daily.
As a method for identification of such often-occurring
pathogens and putrefiers, culture techniques with subse-
quent biochemical and/or serologic diagnostics partially
established over many decades have proven themselves.
Progress in this classical field of application of microbiol-
ogy was mainly achieved by optimization of the demands
on nutrients and cultures by means of selective enrichment
of the respective organisms and by (partial) automation of
the steps necessary for identification.
R.G. Weyandt ()
Institut Fresenius GmbH,
Im Maisel 14, 65232 Taunusstein, Germany
e-mail: weyandt@rud.fresenius.com
Until now these standardized and, because of the rela-
tively low technical requirements on the equipment, wide-
spread methods are among the most important routine pro-
cedures in private and government laboratories and author-
ities for investigation of foodstuff microbiology around the
world.
Times have, however, obviously changed abruptly
because of increased knowledge about molecular biology
and biochemistry in the last ten years, application-friendly
solutions for sensitive and mostly rapid determination of
germs have developed from research activities. These so-
lutions have revolutionized the laboratory landscape for
providers of analytical services and will continue to do so.
PCR-based rapid determination of hygienic strains is al-
most part of the standard repertoire, just as are the varied
immunochemical determination methods via ELISA or
LIA. Because of new types of biomarker, biochip, and mi-
cro-array, fully automated robotic systems, and multiplex
methods, these techniques are just the beginning.
A variety of conditions have led to this development:
widespread availability of highly pure biochemicals and
enzymes,
the decoding of genetic target sequences,
tailor-made production of gene probes, primers, markers,
and antibodies,
industrial production of robust laboratory equipment
which guarantees automated and miniaturised process-
ing of many samples simultaneously,
development of a new branch of industry operating
world-wide, which, via its R&D activities, its coopera-
tion with universities, and, finally, via a large number of
spin-off-companies, can produce new ideas and imple-
ment them rapidly, and
changes in the foodstuffs field have led to changes in
foodstuff analysis, e.g. the introduction of GMOs and
the effects of the on-going BSE-crisis.
The emerging revolution can be seen by way of the ex-
ample of the newest methods for Legionella analysis.
Whereas up until 3 years ago the only method of determi-
nation available to the routine laboratory was the classical

Legionella analysis according to ISO 11731 [1] (including
the steps:
1. membrane filtration,
2. treatment with an acid buffer, and
3. cultivation on, e.g., GVPC-agar and on cysteine-free
nutrient agar (e.g. BCYE-Cys); incubation period takes
7 to 10 days.
Today the provider of analytical services can choose from
a wide range of detection equipment available on the mar-
ket:
optimised classical determination,
determination by means of PCR after enrichment via
filtration [2],
determination microscopically (epifluorescence) after
treatment of the filter residue with gene probes [3], and
immunochemical determination in accordance with the
ELISA principle; even suitable as a fast on-the-spot de-
termination [4].
The same applies currently to Salmonella, Listeria, Campy-
lobacter, Escherichia coli O157, Clostridium spec. or Cryp-
tosporidium. The list of manufacturers is also very long,
and it is clear that the availability of additional methods
depends only on the needs of, and acceptance by, the mar-
ket.
It is encouraging that these ready-made methods of de-
termination usually only guarantee the manufacturer a
monopoly for a short time, if at all. As long as the analy-
sis is commercially relevant alternatives bringing further
refinement or time-saving are developed quickly.
In any case this development has led not only to differ-
entiated and problem-related application from the provider
of analysis services, or to corresponding niche occupation
231
on the part of the manufacturer, but has also contributed to
further propagation of the technology itself, to which, on
the other hand, new methods of determination were adapted.
Finally, and we are now in this exciting phase of develop-
ment not only hard competition among the new prod-
ucts but also integration of individual determinations in
automated system solutions led, after an initial phase, to
considerable price cuts, so that the new products now
hardly need to fear competition from classical methods.
An obstacle that unfortunately still exists, and that has
for a long time prevented more rapid introduction of the
new technologies, is the usually conservative legitimiza-
tion structures at national and international levels deter-
mination procedures that have once been proven and stan-
dardized become the gold standard. As long as the new
procedures are not expressly accepted into the official col-
lection of methods, acceptance by foodstuffs manufactur-
ing, by foodstuffs marketing, and by short-sighted providers
of analytical services remains comparatively low. It is,
therefore, desirable that, with such rapid development of
new analytical techniques (particularly in the field of hy-
giene microbiology) and in view of the urgent need for
quick, reliable methods, administrative and legislative ne-
cessities do not become a bottleneck hampering develop-
ment.
References
1. ISO 11731 Water quality-detection and enumeration of legionella,
1998
2. http://www.iqproducts.nl
3. http://www.vermicon.com
4. http://www.siemens.de/gebaeudemanagement