Arch. Environ. Contam. Toxicol.
29, 322-326 (1995)
Gas Phase Infrared Spectra of 209 Polychlorinated Biphenyl Congeners Using Gas
Chromatography with Fourier Transform Infrared Detection: Internal
StandardiTation with a 13C-Labelled Congener
B. Bush, E. L. Barnard
Wadsworth Center, New York State Department of Health, P.O. Box 509, Albany, New York 12201-0509, USA
Received: 17 January 1995/Revised: 20 March 1995
Abstract. The gas phase infrared spectra of biphenyl and 209
polychlorinated biphenyls (PCB) are observed; 30 are pre-
sented in graphical form so that their uniqueness can be appre-
ciated. The spectra were observed with a gas chromatograph
linked via a light pipe with a Fourier transform infrared spec-
trometer. Unequivocal library matches (with match factors >
990 in 1000) can be performed with 50 ng of a congener
injected into the chromatograph. Quantitation with the tech-
nique was briefly investigated with 2,2',5,5'-tetrachlorobiphe-
nyl (BZ no. 52). Selective wavelength chromatograms of the
most intense absorption (1457cm-1) were integrated. Septupli-
cate injections showed that 20 ng of the compound could be
determined with ---4% precision. The resolution of the spec-
trometer was sufficient to distinguish between the 13C-labelled
trichlorobiphenyl and the unlabelled material, indicating the
possibility of internal or surrogate standardization of PCB anal-
ysis.
Polychlorinated biphenyls (PCB) continue to be serious envi-
ronmental pollutants and pose a challenge to the analytical
chemist, in spite of the fact that their use was banned 15 years
ago. Since the inception of capillary gas chromatography (GC)
with electron capture detection (Ballschmitter and Zell 1982;
Bush e t a l . 1982; Schultz e t a l . 1989) or mass spectral detection
(Bonelli 1972), congener by congener analysis has been possi-
ble and much data has been produced on the environmental
occurrence and behavior of individual congeners and their indi-
vidual toxicities to various life forms, including humans (See-
gal and Schantz 1994).
In most circumstances, PCB congeners may be identified and
quantitated by the use of two GC columns using stationary
phases with disparate separating powers with an electron cap-
ture or mass spectral detection (Erickson 1986; Schultz et al.
Correspondence to: B. Bush
ARCHIVES OF
Environmental
Contamination
a n d Toxicology
1995 Springer-Verlag New York Inc.
1989). However, when a process such as microbial, thermal, or
photolytic degradation is involved, PCB congeners not nor-
mally found in the environment may be formed and will require
more stringent chemical structure determination. Mass spec-
trometry has been widely used for this purpose (Bonelli 1973),
but discrimination between isomers with the same number of
chlorine atoms and accurate quantitation can be of limited reli-
ability, especially with samples such as human milk, which
contain much hydrocarbon material. Unless a mass spectral
resolution > 10,000 is employed to resolve hydrocarbon inter-
ference from the target halogenated molecules with high mass
defect, reliable quantitation is not possible (Bush et al. 1983),
especially when single or multiple ion monitoring is employed
in attempts to improve detectability (Bush and Hong 1991).
Moreover, at both high and low resolution, interference occurs
between congeners differing by one chlorine substitutent, as
well as isomer groups that differ by two chlorine substitutions,
due to the contribution of D, 13C, and 170, and 37C1, respec-
tively.
By contrast, the infrared spectra of PCB congeners are sur-
prisingly characteristic, even between highly chlorinated iso-
mers with nearly identical chlorine substitution patterns
(Doumenq et al. 1992). Until recently, the technique has been
incompatible with capillary gas chromatography due to sensi-
tivity and chromatographic dead volume problems. Accessories
for adapting a relatively large horizontal optical bench for gas
chromatograph effluent monitoring are really only suitable to
column flow rates in the 40 ml/min range, using a make-up gas
to convert from the typical capillary flow rate of 2 ml/min,
producing such dilution as to make a fairly insensitive absopti-
ometric technique inadequate for environmental analysis (Coo-
per et al. 1986).
Two infrared techniques have recently become available
which overcome these problems: trapping the column effluent
on a liquid nitrogen cooled zinc/selenium disc, followed subse-
quently by Fourier transform attenuated reflectance spectros-
copy (Hembree et al. 1993), or the simpler approach of observ-
ing the column effluent in real time using a gold-plated light
pipe and an optical bench in the vertical plane (Gurka and Pyle
Infrared Spectra of PCBs
1988). The latter has been further enhanced recently by the
introduction of an array processor dedicated to the Fourier
transform which greatly increases the speed of the transforms.
This paper reports the gas phase infrared spectra of all 209
possible PCB congeners and evaluates the technique for quanti-
tative analysis of PCBs.
Materials and Methods
Polychlorinated Biphenyl Congeners
Standardized solutions (100 txg/ml) of individual PCB congeners dis-
solved in 2,2,4-trimethylpentane were purchased from Accustandard
Inc. (Newhaven, CT).
Instrumentation
A Hewlett-Packard (H-P) Infrared Detector (5969B) linked to an H-P
5980 gas chromatograph was used (Gurka and Pyle 1988). A 9000
series 340 data system was employed with a 355 megabyte hard drive
and Pascal-based user software. Finished spectra were stored with an
H-P 9144 tape drive. A 25m Ultra-II 5% phenyl-methyloctadecylsilyl
bonded fused silica column (0.33 micron coating thickness 0.2mm
i.d.) (H-P, Avondale, PA) was employed using a 15C/min tempera-
ture ramp commencing 2 minutes after injection at 70C using an H-P
splitless injector (5ul injection volume). Flow rate was 2ml/min, and
light pipe temperature was 290C. The light pipe was connected (via a
make-up gas system to maintain atmospheric pressure) to a Mass
Selective Detector, but mass spectra are not reported in this work, as
they are widely available elsewhere (Erickson 1986). A wide-band
(550-4000cm -1) mercury cadmium telluride (MCT) infrared detector
cooled in boiling nitrogen liquid was used (H-P 5965B#100). The
PCB library developed here is available under copyright from Health
Research Incorporated, Albany, NY.
Results
30 spectra are presented in Figure 1 in the most visually compa-
rable format we could devise. BaUschmitter and Zell congener
numbering is used, and the chemical structures are given in a
format that allows the user to visualize the structure of each
compound easily. The remaining 179 spectra were equally
characteristic and will be provided by the authors on request.
The spectrum of 2,5,2',5'-tetrachlorobiphenyl (BZ# 52)
was observed at four different light pipe temperatures (260,
270, 280, and 300C) to determine the effect of temperature on
the gas phase spectrum. Each spectrum was matched with the
library spectrum obtained at 290C. No change in fit parameter
was observed indicating that the temperature effect on the spec-
trum was negligible. 260C was the lowest useable temperature
for this congener; below this temperature, the signal disap-
peared. 290C is used routinely to observe the highest chlorine-
containing congeners. The greatest infrared absorption at
1457cm-1 was integrated to determine the effect of temperature
on response. A regression response:
Area = -6.7(temperature C) + 3000
323
was obtained with a fit o f R 2 = 0.8. The temperature is there-
fore important in quantitating PCB with the IRD and must be
kept constant.
To determine the linearity of response, a series of injections
from 100 I~g/ml and 10 Ixg/ml solutions were made to yield 20,
50, 100, 250, and 500 ng per injection. The response is linear
over the range observed, and the fit is extremely close to a
straight line (R 2 = 0.994). The 20 ng injection was not visible
on the total response chromatogram, but was well above back-
ground on the selected wavelength chromatogram. Areas were
measured by demarking the start and stop integrate points man-
ually. The minimum detection level was determined by making
septuplicate injections of 50 ng using the EPA procedure for
determining MDL (EPA 1984). The standard deviation was +
2.3 ng, and hence the MDL (p = 0.01) was 7 ng. Unfortu-
nately, when 5 ng was injected into the instrument, the size of
the peak at 1457 cm -1 was just observable above background
but could not be integrated. The statistics-based MDL determi-
nation method recommended by the USEPA (EPA 1984) does
not take into account the characteristics of the integrator, which
in this analysis are the limiting factors. However, it was appar-
ent by visual observation that 20 ng was a robust measurement
and so could be taken at the realistic minimum detection level.
When the spectrum at 20 ng was searched against the library,
the match was achieved in only about 50% of the times it was
searched, whereas at 50 ng, a perfect match (with a match
factor >990 out of 1000) was always achieved. Hence for this
congener, 50 ng is the minimum detection level for identifica-
tion purposes, whereas the MDL is 20 ng for quantitation using
selected wavelength chromatography. This compares with 1 ng
of PCB by mass selective detector in the scanning mode (Hong
and Bush 1991) and '0.2 picograms with an electron capture
detector in our hands.
Two possible internal or surrogate standards were investi-
gated. The spectrum of benzophenone is shown in Figure 2.
Clearly the peak at 1683 c m - 1 is suitable for measurement as an
internal standard, being well clear of any PCB vibrations. ~3C-
labelled (99% enriched, Cambridge Isotopes Inc., Andover,
MA) 214,4'-trichlorobipheyl (BZ no. 28) spectrum is shown in
Figure 3 alone and as a 1:1 mix with the unlabelled compound.
The spectrometer is yielding sufficient resolution for this dis-
crimination, illustrating that heavy isotope dilution analysis is
another possibility with this infrared equipment.
Discussion
The most striking feature of the gas phase infrared spectra of the
PCB is their dissimilarity. When any spectrum is compared
with all the 209 spectra in the library using the Hewlett-Packard
search algorithm, an output shown in Figure 4 is produced. The
match factors are 996 for the first match and 888 for the second.
Depending on the quality of the spectra, match factors > 990
are achieved; visual comparison with the nearest-but-one match
makes the first choice the obvious only possible chemical struc-
ture, illustrating the power of visual interpretation in infrared
spectroscopy. This is then an exceptionally powerful tool for
discriminating between PCB isomers, particularly when com-
bined with high resolution chromatography. Quantitation can
be based on the integration of the selective wavelength chro-