Methods 76 (2015) 3–10

Contents lists available at ScienceDirect

Methods
journal homepage: www.elsevier.com/locate/ymeth

What does calorimetry and thermodynamics of living cells tell us?

Thomas Maskow ⇑, Sven Paufler
UFZ, Helmholtz Centre for Environmental Research, Dept. Environmental Microbiology, Permoserstr. 15, D-04318 Leipzig, Germany

a r t i c l e i n f o a b s t r a c t

Article history: This article presents and compares several thermodynamic methods for the quantitative interpretation of
Received 28 August 2014 data from calorimetric measurements. Heat generation and absorption are universal features of microbial
Received in revised form 23 October 2014 growth and product formation as well as of cell cultures from animals, plants and insects. The heat pro-
Accepted 28 October 2014
duction rate reflects metabolic changes in real time and is measurable on-line. The detection limit of
Available online 13 November 2014
commercially available calorimetric instruments can be low enough to measure the heat of 100,000 aer-
obically growing bacteria or of 100 myocardial cells. Heat can be monitored in reaction vessels ranging
Keywords:
from a few nanoliters up to many cubic meters. Most important the heat flux measurement does not
Calorimetry
Biothermodynamic
interfere with the biological process under investigation. The practical advantages of calorimetry include
Bioprocess control the waiver of labeling and reactants. It is further possible to assemble the thermal transducer in a pro-
Enthalpy balances tected way that reduces aging and thereby signal drifts. Calorimetry works with optically opaque solu-
Oxycaloric equivalent tions.
Law of Hess All of these advantages make calorimetry an interesting method for many applications in medicine,
environmental sciences, ecology, biochemistry and biotechnology, just to mention a few. However, in
many cases the heat signal is merely used to monitor biological processes but only rarely to quantita-
tively interpret the data. Therefore, a significant proportion of the information potential of calorimetry
remains unutilized. To fill this information gap and to motivate the reader using the full information
potential of calorimetry, various methods for quantitative data interpretations are presented, evaluated
and compared with each other. Possible errors of interpretation and limitations of quantitative data anal-
ysis are also discussed.
Ó 2014 Elsevier Inc. All rights reserved.

1. Introduction elaborated optimization and design of primer, probes and hybrid-

Heat is an inevitable by-product of any microbial growth and
product formation processes. Heat is directly related to growth
stoichiometry and heat production rate to the metabolic fluxes
via Law of Hess. Calorimetry measures the metabolic activity of a
population of cells. By this, any metabolic change can be monitored
on-line and in real time. These days, heat production rate can be
measured with a specific limit of detection of 105 W L1 which
corresponds to a microbial oxygen depletion of air saturated med-
ium (approx. 6.7 mg L1 at 37 °C) within 44 days. In case of aerobic
glucose combustion it relates to the tiny consumption rate of
0.03 lM h1 or 5 lg L1 h1 [1]. These data show the potential of
calorimetry. A good overview about the dependency of the limit
of detection on the considered reaction volume and on the applied
instrument is given by Zogg et al. [2]. Calorimetry works in opaque
media and doesn’t need any labeling agents. Long-lasting and
⇑ Corresponding author. Fax: +49 341 235 1351.
E-mail addresses: Thomas.maskow@ufz.de (T. Maskow), Sven.paufler@ufz.de
(S. Paufler).
ization conditions as known for molecular biology-based methods
[3] are not required in calorimetry. Finally, thermal transducers
may be, different to many other biochemical sensors, mounted in
a protected location. This keeps the transducer stable for long time
and reduces the signal drift. All of these advantages increase the
scientific and technical demand for calorimetric measurement
methods and data interpretation. As a result commercially avail-
able highly sensitive, high throughput calorimeters, able to mea-
sure up to 48 channels in parallel have been developed. Fields of
application ranges from medical research [4], environmental
microbiology [5–7] to the food industry [8]. Due to the develop-
ment of highly sensitive calorimeters with a high throughput and
their validation in different application fields the quantitative data
interpretation becomes more and more important. First reports on
enthalpy balances around systems of different scales which are
also applicable to calorimeters, bioreactors or even ecosystems
were published more than two decades ago [9,10]. The authors
correlated metabolic fluxes and growth stoichiometry with heat
flows via Law of Hess. However, advanced chemical analytics are
indispensable for such a data analysis. Quite often comprehensive

http://dx.doi.org/10.1016/j.ymeth.2014.10.035
1046-2023/Ó 2014 Elsevier Inc. All rights reserved.
4 T. Maskow, S. Paufler / Methods 76 (2015) 3–10
metabolic information are not available for interpretation of calo-
rimetric experiments. For this, the applicability of more simple
growth models assuming a constant heat production rate per cell
or per converted electron were successfully tested just recently
[11]. Finally, the recent tendency to simplify and to standardize
calorimetric measurements leads to potential misinterpretations
of calorimetric experiments [12,13].
This review relates different types of calorimetric data interpre-
tation to the level of available information. It explains weaknesses
and advantages of the evaluation methods. Also, causes of various
misinterpretations of calorimetric measurements are going to be
discussed. Finally, the compromise between modern trends in
respect to miniaturization and high-throughput measurements
and the requirement for more analytical information for a correct
calorimetric data interpretation will be discussed.
2. Extracting quantitative information from the calorimetric
signal
2.1. Data interpretations using the first law of thermodynamics
Every open system, independent of its size and the ability to
monitor the heat production, can be considered as a type of calo-
rimeter. A ‘‘system’’ can be a part of an ecosystem, a bioreactor,
an animal or a plant but it could also be a simple medical or food
sample containing active cells. Clearly defining the systems bound-
aries is essential for designing the experiments and later for data
evaluation. Fig. 1 shows a simplified chart of such an open system
exchanging energy and matter with the environment.
The general balance of such a system has the structure of:
Accumulation in the system ¼ Input  Output
 Chemical; biological conversions
ð1Þ
Every conserved quantity (e.g., elements, electrons, individual
chemical and biological species as well as enthalpy) can be bal-
anced. The usage of enthalpy balances for evaluation of calorimet-
ric results makes most sense if information about the metabolic
fluxes ri (Eq. (2)) or about the growth stoichiometric coefficients
Yi/X (Eq. (3)) is available or aspired. The difference between heat
and enthalpy is discussed in a later section of the article. However,
the main barrier to application of fully balanced systems is the
requirement of an exhaustive chemical analysis of the calorimetri-
cally monitored process. The following example shows the sim-
plest case of aerobic biomass formation (CHX1OX2NX3) from any
organic carbon source (CHS1OS2). The cell dry mass of biological
Fig. 1. Calorimeter as an open system exchanging energy and matter with the
environment. The calorimeter can incorporate parts of ecosystems, bioreactors or
simple samples with living matter. The system can be in steady state or in a transient.
It has to be adapted to the specific experimental conditions. Convective heat flows
due to exchange of matter is not considered. The figure is derived from [9].
organisms is designated in the following as biomass. The elemental
composition of X1 = 1.70; X2 = 0.42 and X3 = 0.25 is typical for bac-
terial biomass [14].
rS CHS1 OS2 þ rN NH3 þ rO2 O2 ! r X CHX1 OX2 NX3 þ r CO2 CO2
þ r H2 O H 2 O ð2Þ
ri
with Y i=X ¼ rX
follows Y S=X CHS1 OS2 þ Y N=X NH3 þ Y O2 =X O2 ! CHX1 OX2 NX3
ð3Þ
þ Y CO2 =X CO2 þ Y H2 O=X H2 O
The five unknown stoichiometric/yield coefficients Yi/X are not
independent of each other. They have to fulfill four elemental bal-
ances for C, H, O and N. This means the yield coefficients can be cal-
culated if one more conditional relation is known. This additional
relation can be the enthalpy balance which is also called the Law
of Hess (Eq. (4)).
X
n X
n
DR H X ¼ ðY i=X Df Hi Þ ¼  ðY i=X DC Hi Þ ð4Þ
i¼1 i¼1
The energy balance/Law of Hess correlates the measured
growth reaction enthalpy DRHX with the energy contributions of
each chemical species i involved in the growth reaction. DRHX is
the enthalpy released (negative sign) or consumed (positive sign)
during the formation of one mole biomass. The energy contribution
of each species can be described as enthalpies of formation DfHi or
of combustion DCHi (see Eq. (4)). However, the correct standard
and reference state of the enthalpy of all species have to be
selected and to be corrected for temperature. Von Stockar and
co-workers discussed consequences of wrong usage of reference
states and neglecting temperature corrections [10]. Note, the usage
of combustion enthalpies simplify the enthalpy calculation (in this
specific case) because the combustion enthalpies of CO2 and H2O
are zero (see Eq. (5))
DR HX ¼ ðDC HX  Y S=X DC HS  Y N=X DC HN Þ ð5Þ
The calorimetrically determined heat production rate Q_ con-
tains additional kinetic information[15,16]. Eq. (6) shows it at the
example of the growth rate rX.
Q_ ¼ r X DR HX ð6Þ
R
The growth reaction heat (DR Q ¼ Q_ dt) is equal to the growth
reaction enthalpy DRH for open systems with a constant pressure
(dp = 0). In case of calorimetric measurements in the often used
R
closed ampoules, a contribution of work ( V dp) has to be added.
In case of a typical ampoule of 4 mL with 2 mL cellular suspension
and 2 mL air space a pressure increase of 1 atm (101,325 Pa) pro-
vides only 0.2 J, which can often be neglected in comparison to
the growth heat.
Calorimetry measure the average metabolic activity of a popu-
lation of cells. An in-depth data interpretation can only be done
if enough analytical information is available. The calorimetric sys-
tems must be large enough for extensive sampling or to include
online sensors. This is the case for experiments in a reaction calo-
rimeter, in bioreactors combined with a flow through calorimeter
[17–20], or by performing a calorimetric experiment in parallel
to an equivalent experiment in a bioreactor. The latter two are
error-prone due to metabolic processes in the flow through line
or wall growth [21,22]. Differences between the physical environ-
ment in the calorimeter and in the bioreactor may distort the calo-
rimetric result. An experiment must be carefully designed to
minimize distortions of the calorimetric signal due to (i) the addi-
tion of substances with deviating temperatures, (ii) the heat of
neutralization by keeping the pH constant, (iii) heat of evaporation
due to aeration, (iv) heat effects of gas adsorption and release, and
 T. Maskow, S. Paufler / Methods 76 (2015) 3–10
Fig. 2. Heat production rate (top) and relative energy distribution between the
substrate (glucose) and the products (ethanol, biomass, intermediates and heat)
during the time course of growth of Saccharomyces cerevisiae (bottom). The energy
input (i.e., in form of glucose) was set to 100%. The dotted lines indicate sudden
changes in the heat production rate due to the consumption of glucose or ethanol,
respectively. The data were taken from [17].
(v) heat input by stirring of media with a potentially changing vis-
cosity have also to be taken into account. This can be done as for
instance described by von Stockar et al. and others (see for exam-
ple [10,23–26]). All these are unwanted energetic flows resulting in
increased noise, baseline offset and signal drift. They have a nega-
tive impact on the methods overall limit of detection and limit of
quantification that are closely related to the systems signal to noise
ratio.
In fundamental research fully balanced calorimetric experi-
ments can be applied to detect shifts in metabolic pathways,
because a given pathway is always connected to a certain growth
stoichiometry. The growth stoichiometry in turn can be calculated
in advance by knowledge of the respective pathway using the YATP
5
concept [27]. The potential of reaction calorimetry in this context
was demonstrated at the example of Saccharomyces cerevisiae
shifting from the respiratory to the respire-fermentative metabo-
lism [28] and at the example of the cleavage of aromatic ring struc-
tures in ortho- or meta position [29].
Other applications are the usage to quantify the energetic costs
of metabolic adaptations. This was demonstrated at the examples
of adaptation to a shortage of nutrients [30] or to stress exerted
by high salt concentrations [31–34]. For instance it was found that
the adaptation to salt stress using ectoine as protection molecule
does not burden the metabolism (for details see [33]). Also the
more complex haloadaptation process of the non-conventional
yeast Debaryomyces hansenii was revealed using enthalpy balances
[35].
How valuable the calorimetry in combination with enthalpy
balances is, shows the following very early and simple example
(Fig. 2). The energy distribution, during the growth of yeast, is
quantified in this work. Important changes in the assimilation
pathway from the usage of glucose to usage of ethanol are imme-
diately indicated by two main maxima in the heat production rate.
In order to solve the describing equation system for a calorim-
eter at least as many measured entities have to be available as the
number of degrees of freedom of the system. However, sometimes
even more analytical information are available than the degrees of
freedom of the system. In such cases, it is said that the system is
over-determined or contains redundant information. The redun-
dancy can be used for detection of gross and systematic errors
and data reconciliation. The bases for these applications were
founded more as two decades ago [36–39]. Taking the simple aer-
obic growth (see Eq. (3)) as an example, the five independent
growth yields are accompanied by four conservation equations.
Therefore, the degree of freedom is one. With the measurement
of the reaction heat is the system completely described and with
the additional measurement of CO2 evolution, or oxygen consump-
tion over-determined. Even first application of pure calorimetric
monitoring of bioreactors was reported very early [40] and
14 years later demonstrated at industrial scale [41]. The applica-
tion of redundancy analysis to calorimetrically monitored bioreac-
tors is discussed by van der Heijden et al. [37].
In contrast to systems with redundant information is the case of
the biodegradation of trace pollutants (in the lg/L range). Here it is
often difficult or even impossible to follow the degradation kinetics
of pollutants or the formation of biomass or intermediates by
chemical analysis due to the extreme small changes in concentra-
tion. The following example measured by isothermal titration cal-
orimetry will illustrate how calorimetry can, in combination with
enthalpy balances, yield new insights. The first step during the bio-
degradation of the herbicide 2,4-Dichlorphenoxybutyrate (2,4-DB)
ð7Þ
is an enzymatically catalyzed cleavage of the ether bond (Eq. (7)).
The question arises what happens with the cleavage products.
For clarifying that question a tiny amount of 2,4-DB was
added to a bacterial suspension and the degradation monitored
calorimetrically resulting in a reaction heat of 1149 ± 16 kJ/mol
6 T. Maskow, S. Paufler / Methods 76 (2015) 3–10
2,4-DB. This measured reaction heat was compared with calculated
reaction heats from enthalpy balances assuming different fates of
2,4-DB (Fig. 3). The comparison suggested that the metabolic end
products are dichlorophenol (DCP) and biomass [42]. The accumu-
lation of DCP during biodegradation of 2,4-DB has already been
reported [43], thus supporting this interpretation. Taking this as
an example the potential fate of biodegradation of other pollutants
can also be revealed.
2.2. Data interpretations using a constant energy per converted
electron
In the most common form of calorimetry (i.e., microcalorime-
try) but also with the newly emerging miniaturized or chip-calo-
rimeter it is difficult to fully characterize the biological process.
Furthermore, for many applications of calorimetry in medicine,
pharmacy, food industry, and environmental sciences is the real
time detection of metabolic activity sufficient. If the calorimetric
signal has to be interpreted without additional information, it is
a good idea to assume a constant heat production per converted
electron. If this assumption is correct, then the heat signal reflects
mainly the catabolic part of the metabolism because most elec-
trons are converted via electron transport phosphorylation. The
concept of a constant heat per converted electron goes back to
Thornton [44] who first described a linear relation of the combus-
tion enthalpy of multiple organic compounds DCHi versus degree of
reductance ci (Eq. (8)). Such a relation (in the following called
Thornton-Rule) was confirmed by many other authors [9,45–48].
DC Hi ¼ ci DC He þ di with ci ¼ 4 þ S1  S2 ð8Þ
DCHe, di describe the heat related with the conversion of one
mole electrons and the deviation between the linear relation and
the actual combustion enthalpy of the chemical compound i
respectively. ci in Eq. (8) is related to one carbon mole (C-mol) of
the compound i having the composition of CHS1OS2NS3. For
instance one mole of ethanol (C2H6O) corresponds to two carbon
mole of ethanol (CH3O0.5). From this linear relation and depending
on the applied data base a heat per mole converted electrons DCHe
between (107  120) kJ e-mol1 can be derived [14,44].
Therefore, another exciting question can be answered. The heat
Q released per mol oxygen respired r O2 is ((107  120) kJ/e-
mol  4 e-mol/mol-O2 = (428  480) kJ/mol-O2). That value is
Fig. 3. Comparison of the measured reaction enthalpy of the biodegradation of the
herbicide 2,4-DB by Rhodococcus erythropolis K2–3 with enthalpy balance calcula-
tions assuming different fate of 2,4-DB. Complete oxidation (A); Incomplete
oxidation (B: DCP is remaining and succinate is combusted; C: DCP is remaining
and succinate is used to produce biomass; D: DCP and succinate is remaining). The
data were taken from [42].
called oxycaloric equivalent and usually attains values between
430 and 480 kJ/mol-O2 with an average value of
(455 ± 15) kJ/mol-O2 [49].
Since the calorimetry in this simple picture reflects the cata-
bolic side of metabolism, it is interesting to compare the calorim-
etry with methods, which show the other sides of the
metabolism. For instance, the comparison with microscopy (mea-
sure of biovolume), the counting of colony forming units (measure
of fertility), fluorescence microscopy (live/dead distinction), ATP
content (energy charge of cells) has the potential to deliver mech-
anistic information about the metabolic target of biocides or anti-
biotics [7,50,51]. This can be applied for instance to characterize
toxic properties of chemicals, the action of antibiotics or of biolog-
ical agents.
Even more interesting for calorimetric data interpretation are
deviations in Q_ =r O2 from the oxycaloric equivalent. Three different
cases can be distinguished. First, if peroxides or other reactive oxy-
gen species (ROS) are formed, the oxycaloric equivalent is violated.
For example, Oroszi measured the photo-synthetically absorbed
heat of the diatom Phaeodactylum tricornutum and compared it
with oxygen production at different irradiances. He observed
strong deviations from oxycaloric equivalent and assigned this to
photosynthetic protection mechanisms that may be associated
with the synthesis of ROS or peroxides [52].
Second, and of more practical importance are deviations that
are caused by either partially anaerobic metabolism or anaerobic
zones in the considered bioreactor or ecosystem. Fig. 4 illustrates
the effect of anaerobiosis taking glucose as substrate, ignoring ana-
bolic reactions and assuming three different types of anaerobic cat-
abolic reactions (A – alcoholic fermentation, B – lactic acid
fermentation, C – homoacetic acid fermentation).
Fig. 4 shows clearly that a small percentage of anaerobiosis
(caused by anaerobic zones in a bioreactor or ecosystem or by
mixed fermentative metabolism) has only a little influence on
the Q_ =rO2 . This influence increases with increasing ratio of anaero-
biosis and approaches theoretically to infinity for 100% anaerobio-
sis. Indeed such deviations are reported by a few publications (e.g.,
[17]) reaching values of 11,000 kJ/mol-O2, which have been
attributed to lactic acid formation under anoxic conditions by the
authors [53].
The third reason for potential deviations from the oxycaloric
equivalent of the total system could be deviations in
Fig. 4. Influence of anaerobiosis on deviation from the oxycaloric equivalent.
460 kJ mol1 was assumed as value for the oxycaloric equivalent. The graph
shows the percentage deviations from this value. The thermodynamic base data
were taken from [10].
 T. Maskow, S. Paufler / Methods 76 (2015) 3–10 7

Thornton-Rule for educts consumed or products formed during the Table 1

bioconversion. As an example, the ratio Q_ =r O2 provides theoreti- Limits of detection for the cell concentration as a function of the cell specific heat
production rate Q_ C . A typical Q_ D ¼ 100 nW (limit of detection of the applied
cally a real time measure for the biomass related yield coefficient calorimeter) and V = 2 mL (sample size) were assumed. Note, these values provide
YX/S (Eq. (9)). just a raw picture as they obviously depend on the respective growth conditions. Data
were taken from [1,87].
4½dS  Y X=S ðdX  X 3 dN Þ
Q_ =r O2 ¼ 4DC He þ ð9Þ Cell type Condition Q_ C Limit of detection
cS  Y X=S ðcX  3X 3 Þ
(pW cell1) (103 cells mL1)
dX, dS and dN describe the deviation of the biomass, substrate Bacteria
and ammonia as nitrogen source from the Thornton-Rule. The Escherichia coli Endogenous 0.05 1000
calculated typical dX and dS are less than 10% based on published Escherichia coli Anaerobic 0.2 250
values [10]. dN is approximately 2% taking the data of [54]. The Escherichia coli Aerobic 0.8 62
Staphylococcus Aerobic 2.5 20
total effect of these deviations on Q_ =rO2 is less than 10%. Indeed, aureus
the observed Q_ =r O2 ratio for complete aerobic growth without Mycoplasma hominis Aerobic 11 4.5
product formation vary between 385 and 495 kJ/mol-O2 [55]. Yeast
This means that for exploiting the relation in Eq. (9), the Q_ as well Schizosaccharomyces Aerobic 63 0.79
as rO2 have to be measured very accurately. However, Regestein pombe
and co-worker overcame that problem and showed in 2013 that Protozoa
deviations from the oxycaloric equivalent can be applied to moni- Tetrahymena Aerobic 3,300 0.015
tor biotechnological lysine formation in real time in a bioreactor pyriformis
[56]. Mammalian cell lines
Hepatozytena Aerobic 250 ± 6 0.2
Myocardial cells Aerobic 2,000 0.025
2.3. Data interpretations using a constant cell specific heat production
a
rate Own measurements.

of biomass concentration X with the cell specific growth rate l (Eq.

In medicine, pharmacy, food industry and water management,
the number of microorganisms is often more important than the
exact knowledge of their metabolic activities. This is also the case
for many toxicological studies in environmental sciences. The sim-
plest and most applied assumption in these cases is a constant cell
specific heat production rate Q_ C . We will later discuss how to get
and how to apply a more realistic estimation of Q_ C . But, related
to the number of cells, how sensitive can calorimetric measure-
ments be? A modern commercially available isothermal calorime-
ter achieves a detection limit of Q_ D ¼ 100 nW (see for instance:
www.tainstruments.com or [4]). With Q_ C a relation between the
measured heat signal Q_ , the cell concentration X and the sample
volume V can be written (Eq. (10)).
Q_ ¼ X Q_ C V
Putting Q_ D in Eq. (10) allows the calculation of the detection
limit for the cell concentration for a typical sample volume in
dependency on Q_ C (Table 1).
The cell concentration X in Eq. (10) is not constant. Its increase
can be described using the SKIP (sum kinetics with interaction
parameter) model [57–59] (Eq. (11)).
dX X
M
lmax;i Si
¼ lX with l¼ XM
dt i¼1 K S;i þ Si þ I S
j¼1 j;i j
Here are M the number of all nutrients considered as essential, Si
the concentration of these nutrients, and KS,i, lmax,i, Ij,i growth
parameters. At the beginning of the calorimetric experiment all
P
nutrients are sufficiently available and Si  K s;i þ Mj¼1 I j;i Sj . This
means, that l can be considered constant for a certain time period
(Eq. (12)).
Q_ ¼ X 0 Q_ C V expðltÞ
But during this period of time calorimetry can be applied to
determine the maximum specific heat production rate of a given
cell type in a defined medium simply by plotting the logarithm
of Q_ versus time t.
Of course a constant cell specific heat production rate can be
called into question. Based on Eq. (6), this controversy can be over-
come by considering the definition of the growth rate rX as product
(13)).
Q_ ¼ rX DR HX ¼ lX DR HX ð13Þ
The comparison of Eq. (13) with Eq. (12) shows a linear relation
between the cell specific heat production rate Q_ C and the specific
growth rate l (Eq. (14)).
Q_ C ¼ lQ C with Q C ¼ C C DR HX ð14Þ
The correlation factor QC can be understood as growth reaction
heat (in J) per cell. For estimation of QC the knowledge of the
enthalpy of growth reaction DRHX and the mean carbon content
(in mole) of a single cell CC is required. A simple expression for
the enthalpy change of growth reaction DRHX as function of degree
ð10Þ of reductance of the substrate cS and the biomass cX (Eq. (14))
results from applying Eq. (14) to growth stoichiometry.
DR HX ¼ 115kJ=C-molðcX  Y S=X cS Þ ð15Þ
Many authors (for instance, [60–64]) recognized that the yield
coefficient YS/X depends on the energy content of the substrate.
The energy content of the substrate depends on the relative degree
of reductance cS as known from the Rule of Thornton. The simplest
relation between the relative degree of reductance cS of the sub-
ð11Þ strate and the yield coefficient YS/X (Eq. (16)) is described by [65].
7:69
Y S=X ¼ for cS  4:67
cS ð16Þ
Y S=X ¼ 1:67 for cS > 4:67
The enthalpy change of the growth reaction DRHX can be calcu-
lated based there upon (Eq. (17)).
DR HX ¼ 115kJ=C-molðcX  7:69Þ for cS  4:67
ð17Þ
ð12Þ DR HX ¼ 115kJ=C-molðcX  1:67cS Þ for cS > 4:67
This means that for many naturally occurring substrates (e.g.,
carbohydrates, approx. 70% of natural amino acids, proteins, DNA
and RNA) the reaction enthalpy per C-mol formed bacterial mass
is independent from the applied carbon-substrate. Describing
the elemental composition of biomass by CH1.70O0.42N0.25
(M = 23.99 g/C-mol) [14] approx. 440.5 kJ/C-mol results for
carbon sources with a degree of reduction 64.67.
8 T. Maskow, S. Paufler / Methods 76 (2015) 3–10
The second unknown information in correlation factor QC is the
carbon content of a single cell CC. This carbon content depends
obviously on the size of the considered cell. This value can be esti-
mated from the cell specific weight (in g) and the cell composition.
Taking Escherichia coli as a typical bacterial example a mean carbon
content per cell of 15  1015 mol was estimated by this way [66].
Thereof the mean cell specific growth enthalpy for bacteria of a
similar size as E. coli can be estimated to be 6.6 nJ/cell. With spe-
cific growth rates between 0.1 and 1 h1 cell specific heat produc-
tion rate for aerobe growing ‘‘typical’’ bacteria between 0.18 and
1.8 pW/cell can be estimated according to literature values (see
Table 1).
3. Reasons for possible misinformation
Sometimes the interpretability of the calorimetric measure-
ment reaches their limit due to peculiarities of the investigated
biological system or by characteristics of the design of the calori-
metric experiment. Two respective examples will be discussed in
the following.
3.1. Apparent discrepancies between enthalpy values measured with
high performance calorimetry and calculated from chemical analysis
Nowadays, high performance reaction calorimeter can measure
with a high accuracy and resolution. Depending on the experimen-
tal setup and being able to minimize undesired heat flows (see Sec-
tion 2.1) reaction calorimetry can achieve a limit of detection of
better than 5 mW/L and good long term baseline stability
(0.2 mW/(L h)). Such performance parameters allow to monitor
even low enthalpy yielding reactions (e.g., anaerobic fermenta-
tions) and to analyze them based on full enthalpy balances. How-
ever, new challenges of data interpretation may arise thereof.
There is no doubt that biological growth processes have to fulfill
the first law of thermodynamics. Any discrepancy between mea-
sured and calculated heat production rate based on enthalpy bal-
ances should indicate some not considered metabolic events. The
applied metabolic model might be over-simplified. However,
errors in chemical analysis of educts or metabolic products may
bias the result. In case of aerobic growth the substrate related
growth reaction enthalpy is often in the magnitude of the combus-
tion enthalpy of the substrate. For instance, the pure biological
combustion of glucose delivers 2813 kJ/mol whereas growth on
glucose (assuming a typical yield coefficient of 0.4 g/g) provides
1398.1 kJ/mol. The values where calculated applying the Law of
Hess (Eq. (5)), taking the already explained typical biomass compo-
sition and estimating the energy content of biomass using the
Thornton-Rule. Assuming an error in the growth yield coefficient
of 10% (maintaining the balances of redox and elements) caused
by chemical analysis an error in growth reaction enthalpy of
141 kJ/mol or 10.1% can be calculated. This corresponds with the
general experience that the heat production rate can be well pre-
dicted after knowing the relevant conversion rates. However the
situation changes for anaerobic systems. The simplest example is
the fermentation of glucose (Gluc) to ethanol (EtOH) and carbon
dioxide (Eq. (18)).
C6 H12 O6 ! Y EtOH=S C2 H6 O þ Y CO2 =S CO2ðgÞ with Y EtOH=S ¼ Y CO2 =S ¼ 2
ð18Þ
The reaction enthalpy is 100 kJ/mol of glucose. Already small
errors in chemical analysis e.g., in EtOH production (rate) measure-
ment, would translate into relatively large errors in calculated
reaction enthalpy (e.g., if the rate is 3.7% too big then DRHgluc = 0)
since combustion enthalpy values of the educts and products are
much larger than reaction enthalpy of the conversion. However it
could be argued that YEtOH/S and Y CO2 =S are not independent of each
other. They have to fulfill balances of the elements and redox
states. Indeed, for the alcoholic fermentation is YEtOH/S = Y CO2 =S = 2
fixed. What happens in a more complex reaction with analytical
errors which fulfill elemental and redox balances? A good example
might be the mixed acid fermentation of glucose to acetic (AC) and
butyric (BU) acid. This reaction can be conceptually separated into
two sub-reactions (Eqs. (19) and (20)).
C6 H12 O6 þ YH2 0=S H2 O ! YAC=S C 2 H3 O2 þ YCO2 =S CO2
þ YH2 =S H2 þ YH=S Hþ ð19Þ
C6 H12 O6 ! YBU=S C4 H2 O þ YCO2 =S CO2 þ YH2 =S H2 þ YH=S Hþ ð20Þ
The yield coefficients of each sub-reaction are fixed by the bal-
ances of the elements and redox states to YH2 O=S = YAC/
S = YCO2 =S = YH/S = 2; YH2 =S = 4 (Eq. (18)) and YBU/S = YH/S = 1;
YCO2 =S = YH2 =S = 2 (Eq. (19)). The respective reaction enthalpies
are 76.6 kJ/mol (Eq. (18)) and 57.6 kJ/mol (Eq. (19)). The advan-
tage of this approach is that analytical measurement error can be
understood as a shift of the ratio of the two reactions and that
thereby the elementary and redox balances are not violated.
Fig. 5 shows the influence of analytical errors (5% acetate rate)
on the theoretical error of the reaction enthalpy.
Depending on the ratio of the formation of acetate and buty-
rate, a large range of enthalpies swept. Even a change from exo-
thermic to an endothermic reaction happens at the ratio of 0.43.
This switching point is a discontinuity with the consequences
that a small error in chemical analysis results in infinite error
in reaction enthalpy. The upper part of Fig. 5 show the error
in reaction enthalpy for experimentally observed reaction ratios
[67]. Even in this range causes an error of chemical analysis of
5% an error in reaction enthalpy between 7% and 20%. The cho-
sen example demonstrates clearly that in case of anaerobic
growth, high performance calorimetry may provide more accu-
rate process information as the chemical reference analysis.
The chosen example certainly simplifies the realities. But even
if biomass formation is taken into account, for example, the
main message of the error estimation changes with little respect
(unpublished results).
Fig. 5. Influence of 5% analytical error in acetate formation rate on the error in
reaction enthalpy balance. The total reaction enthalpy as function of the ratio of
formation rates (acetate or butyrate) is also shown. For better resolution the part of
the graph where the most experiments are located is shown magnified in the top
right corner.
 T. Maskow, S. Paufler / Methods 76 (2015) 3–10
3.2. Distortions of calorimetrically derived growth parameter
Distortions of the calorimetric signal due to overlapping physi-
cal signals (e.g., stirring of the sample, evaporation, neutralization,
etc.) or instrumental errors (e.g., undefined heat flows, baseline
variations) can complicate the calorimetric data interpretation
but will be not discussed in the following. Here we will focus on
distortions of the metabolic activity caused by changing physical
environments. For simplicity, calorimetric measurements are often
performed (semi)automatically in sealed ampoules. This measur-
ing technique is in the following discussed. In closed ampoules
any concentration is expected to change over time as result of met-
abolic activity. Especially, relatively large cells (e.g., fungi, yeasts or
micro algae) have the tendency to sediment and to cause heteroge-
neity. The latter can be avoided by adjusting the density of the
measurement solution to the density of the biological entity or
by enhancing the viscosity of the measurement solution by adding
metabolically inactive additives [12]. More serious problems arise
from the oxygen consumption due to metabolic activities. In case
of heat produced by aerobic metabolic activity above certain
threshold (e.g., 100 J/L at 37 °C) oxygen diffusion and not the
metabolism alone starts to govern the heat trace. The threshold
is caused by oxygen solubility in the given medium and tempera-
ture and the oxycaloric equivalent [68]. Potential consequences of
ignoring the threshold are that important parameters like the spe-
cific growth rate or the growth reaction enthalpy might be deter-
mined incorrectly [13]. Although the heat trace suggests a
sequence of metabolic events (first aerobic later anaerobic) in real-
ity a mixed respiro-fermentative metabolism occurs. The specific
growth reaction rate and growth reaction enthalpy differ signifi-
cantly depending on the type of metabolism. More details are
given in [13].
4. Conclusions and outlook
Essentially, three types of calorimetric data interpretation
(enthalpy balances, constant heat per exchanged electron, and con-
stant heat per cell) have been established during the long history of
biocalorimetric research. What is applied at the end depends on
the scientific question but also on the available non-calorimetric
data. Beyond the formal modeling or hypothesis testing task of
calorimetric experiments, exploratory data analysis should be
developed to extract more information from the data. As a possible
example, calorimetric data of contaminated sites could be related
to soil microbiome data to better understand the relation between
degradation rate, degree of mineralization and the chemical prop-
erties of the pollutants, habitat characteristics and microbiome
composition. As statistical tools factor analysis (FA) principle com-
partment analysis (PCA), correspondence analysis (CA), and cluster
analysis are thinkable just to mention a few. First attempts to ana-
lyze antimicrobial properties [15,69–72] and interaction between
different microorganisms [73,74] using calorimetry, chemometry
and multivariate statistics have already been reported. We would
expect an essential contribution of calorimetry to the further
understanding of the performances of microbes in ecosystems,
human gut, and bioreactors if the calorimetric data are combined
with omic-technologies using models or even multivariate
statistics.
Despite their potential, further problems arise from the recent
trend of calorimeter miniaturization. The smaller the volume of
the calorimetric chamber, the more limited is the sampling. A solu-
tion is the so-called lab-on-a-chip (LOC) which integrates one or
several laboratory functions on a single chip of only millimeters
to a few square centimeters in size. LOCs and their further develop-
ment into Micro Total Analysis Systems (lTAS) aim to integrate the
9
total sequence of lab processes to perform biochemical and micro-
biological analysis. Indeed several thermal micro-sensors which
have the potential to be integrated into lTAS have already been
developed during the past two decades [75–77]. They achieve res-
olutions of a few nW and are able to manipulate samples of pL [78].
Even arrays of thermal LOCs are possible [79,80]. However, for bet-
ter data interpretation thermal LOC’s have to be combined with
other methods (e.g., micro-sensors for different substances or
impedance spectroscopy for biomass) to get real calorimetrically
based lTAS.
The other two promising ways to combine calorimetric signals
with elevated process analysis is performing the experiments in
optimized bioreactors where an energy balance is drawn around
the complete reactor or inserting a calorimetric immersion probe.
The first development was already initiated 25 years ago by von
Stockar [81] and led to the calorimetric monitoring of a bioreactor
in the m3 scale [82]. Recent developments achieve resolutions of
up to 20 mW/L [24] and are applicable to reactors of elevated pres-
sure [83]. However, this sensitivity is not good enough if, for exam-
ple, anaerobic bioprocesses have to be analyzed or even controlled.
Thus, methods have to be developed to increase sensitivity, accu-
racy and stability of calorimetrically monitored bioreactors. A first
step in this direction using an improved thermal shield was just
recently published [84].
A calorimetric immersion probe was suggested for the first time
in 1975 [85] and in 2012 developed on the basis of chip-calorime-
try [86]. The latter was sensitive enough for low cell concentrations
but failed at high cell densities due to technical limitations. The
development of a calorimetric immersion probes for microbial sus-
pensions of large activity range is still an open but also rewarding
task. Numerous applications in bioprocess engineering are
possible.
In general, problems also arise from the growing importance of
fed-batch, continuous and perfusion cultures in biotechnology.
Here, R & D work is required to reduce distortions of the calorimet-
ric signal due to convective heat flow and to mathematically model
the remaining effects (e.g., via multiphysics simulation of whole
calorimetric systems).
Beside the development of calorimetric instruments which pro-
vide more data the developments of mathematical models that
actually process those data is rewarding. For applications of calo-
rimetry in medicine or food industries which often trust on mea-
surement of simple closed ampoules an automated parameter fit
would be helpful. These parameters would provide the number
of bacterial cells at the beginning of the experiment and their
growth kinetics depending on biocide or antibiotic concentration,
physical parameters etc. First steps of such developments are
recently published [11] and deserve further development.
References
[1] T. Maskow, T. Schubert, A. Wolf, F. Buchholz, L. Regestein, J. Buechs, F. Mertens,
H. Harms, J. Lerchner, Appl. Microbiol. Biotechnol. 92 (2011) 55–66.
[2] A. Zogg, F. Stoessel, U. Fischer, K. Hungerbühler, Thermochim. Acta 419 (2004)
1–17.
[3] K. Nagarajan, K.C. Loh, Appl. Microbiol. Biotechnol. 98 (2014) 6907–6919.
[4] O. Braissant, D. Wirz, B. Göpfert, A. Daniels, Sensors 10 (2010) 9369–9383.
[5] X.-M. Rong, Q.-Y. Huang, D.-H. Jiang, P. Cai, W. Liang, Pedosphere 17 (2007)
137–145.
[6] O. Braissant, D. Wirz, B. Goepfert, A.U. Daniels, FEMS Microbiol. Lett. 303
(2010) 1–8.
[7] F. Buchholz, H. Harms, T. Maskow, Biotechnol. J. 5 (2010) 1339–1350.
[8] L. Wadsö, F. Gómez Galindo, Food Control 20 (2009) 956–961.
[9] S.I. Sandler, H. Orbey, Biotechnol. Bioeng. 38 (1991) 697–718.
[10] U. von Stockar, L. Gustafsson, C. Larsson, I. Marison, P. Tissot, E. Gnaiger,
Biochim. Biophys. Acta 1183 (1993) 221–240.
[11] O. Braissant, G. Bonkat, D. Wirz, A. Bachmann, Thermochim. Acta 555 (2013)
64–71.
[12] A.J. Fontana, L.D. Hansen, R.W. Breidenbach, R.S. Criddle, Thermochim. Acta
172 (1990) 105–113.
10 T. Maskow, S. Paufler / Methods 76 (2015) 3–10
[13] T. Maskow, F. Mariana Morais, L.F. Rosa, Y.G. Qian, F. Harnisch, RSC Advances 4
(2014) 32730–32737.
[14] J.L. Cordier, B.M. Butsch, B. Birou, U. Stockar, Appl. Microbiol. Biotechnol. 25
(1987) 305–312.
[15] M.A.A. O’Neill, A.E. Beezer, J. Tetteh, S. Gaisford, M. Dhuna, J. Phys. Chem. B 111
(2007) 8145–8149.
[16] L. Almeida e Sousa, A.E. Beezer, L.D. Hansen, D. Clapham, J.A. Connor, S.
Gaisford, J. Phys. Chem. B 116 (2012) 6356–6360.
[17] R. Brettel, I. Lamprecht, B. Schaarschmidt, Radiat. Environ. Biophys. 18 (1980)
301–309.
[18] C. Larsson, G. Lidén, C. Niklasson, L. Gustafsson, Biosyst. Eng. 7 (1991) 151–
155.
[19] C. Larsson, G. Liden, A. Blomberg, C. Niklasson, L. Gustafsson, Pure Appl. Chem.
65 (1993) 1933–1937.
[20] R.B. Kemp, P.M. Evans, Y. Guan, J. Therm. Anal. Calorim. 49 (1997) 755–770.
[21] B. Perry, A. Beezer, R. Miles, J. Appl. Bacteriol. 47 (1979) 527–537.
[22] T. Maskow, S. Müller, A. Lösche, H. Harms, R. Kemp, Biotechnol. Bioeng. 93
(2006) 541–552.
[23] I. Wadsö, Chem. Soc. Rev. 26 (1997) 79–86.
[24] T. Schubert, U. Breuer, H. Harms, T. Maskow, J. Biotechnol. 130 (2007) 24–31.
[25] M. Meier-Schneiders, U. Grosshans, C. Busch, G. Eigenberger, Appl. Microbiol.
Biotechnol. 43 (1995) 431–439.
[26] M. Meier-Schneiders, F. Schäfer, U. Grosshans, C. Busch, Thermochim. Acta 251
(1995) 85–97.
[27] W. Babel, H.W. van Verseveld, Microbial Growth on C1 Compounds, Springer,
1987. 210–219.
[28] P. Duboc, L.G. Cascao-Pereira, U. von Stockar, Biotechnol. Bioeng. 57 (1998)
610–619.
[29] T. Maskow, W. Babel, Thermochim. Acta 309 (1998) 97–103.
[30] T. Maskow, W. Babel, Appl. Microbiol. Biotechnol. 55 (2001) 234–238.
[31] A. Blomberg, C. Larsson, L. Gustafsson, J. Bacteriol. 170 (1988) 4562.
[32] T. Maskow, S. Kleinsteuber, Extremophiles 8 (2004) 133–141.
[33] T. Maskow, W. Babel, Biochim. Biophys. Acta 1527 (2001) 4–10.
[34] T. Maskow, W. Babel, Thermochim. Acta 382 (2002) 229–237.
[35] D. Schumer, U. Breuer, H. Harms, T. Maskow, Eng. Life Sci. 7 (2007) 322–
330.
[36] N.S. Wang, G. Stephanopoulos, Biotechnol. Bioeng. 25 (1983) 2177–2208.
[37] R. Van der Heijden, J. Heijnen, C. Hellinga, B. Romein, K. Luyben, Biotechnol.
Bioeng. 43 (1994) 3–10.
[38] R. Van der Heijden, B. Romein, J. Heijnen, C. Hellinga, K. Luyben, Biotechnol.
Bioeng. 43 (1994) 11–20.
[39] R. Van der Heijden, B. Romein, J. Heijnen, C. Hellinga, K.C.A. Luyben,
Biotechnol. Bioeng. 44 (1994) 781–791.
[40] H. Schuler, C.-U. Schmidt, Chem. Eng. Sci. 47 (1992) 899–913.
[41] A. Hocalar, M. Türker, S. Öztürk, AIChE J. 52 (2006) 3967–3980.
[42] F. Mariana, F. Buchholz, H. Harms, Z. Yong, J. Yao, T. Maskow, J. Microbiol.
Methods 82 (2010) 42–48.
[43] H. Mertingk, R.H. Müller, W. Babel, J. Basic Microbiol. 38 (1998) 257–267.
[44] W.M. Thornton, Philos. Mag. Ser. 6 (33) (1917) 196–203.
[45] A.C. Giese, Cell Physiology, Saunders, 1973.
[46] J. Roels, Energetics and Kinetics in Biotechnology, Elsevier Biomedical Press,
1983.
[47] K. Ho, W. Payne, Biotechnol. Bioeng. 21 (1979) 787–802.
[48] J. Luong, B. Volesky, Heat evolution during the microbial process—estimation,
measurement, and applications, in: A. Fiechter (Ed.), Microbial Activities,
Springer, Berlin, Heidelberg/New York/Tokyo, 1983, pp. 1–40.
[49] L.D. Hansen, C. Macfarlane, N. McKinnon, B.N. Smith, R.S. Criddle, Thermochim.
Acta 422 (2004) 55–61.
[50] F. Buchholz, A. Wolf, J. Lerchner, F. Mertens, H. Harms, T. Maskow, Antimicrob.
Agents Chemother. 54 (2010) 312–319.
[51] F. Mariana, F. Buchholz, J. Lerchner, T.R. Neu, H. Harms, T. Maskow, Int. J. Med.
Microbiol. 303 (2013) 18–165.
[52] S. Oroszi, T. Jakob, C. Wilhelm, H. Harms, T. Maskow, J. Therm. Anal. Calorim.
(2011) 223–231.
[53] A. Schön, I. Wadsö, J. Therm. Anal. Calorim. 33 (1988) 47–54.
[54] C.E. Vanderzee, M. Mansson, I. Wadsö, S. Sunner, J. Chem. Thermodyn. 4 (1972)
541–550.
[55] B. Birou, I.W. Marison, U.V. Stockar, Biotechnol. Bioeng. 30 (1987) 650–660.
[56] L. Regestein, T. Maskow, A. Tack, I. Knabben, M. Wunderlich, J. Lerchner, J.
Büchs, Biotechnol. Bioeng. 110 (2013) 1386–1395.
[57] H. Yoon, G. Klinzing, H.W. Blanch, Biotechnol. Bioeng. 19 (1977) 1193–1210.
[58] J.B. Rogers, K.F. Reardon, Biotechnol. Bioeng. 70 (2000) 428–435.
[59] K.F. Reardon, D.C. Mosteller, J.D.B. Roger, Biotechnol. Bioeng. 69 (2000) 385–
400.
[60] J. Roels, Biotechnol. Bioeng. 22 (1980) 2457–2514.
[61] J. Heijnen, J. Van Dijken, Biotechnol. Bioeng. 39 (1992) 833–858.
[62] J. Heijnen, in: M.C. Flickinger, S.W. Drew (Eds.), Encyclopedia of Bioprocess
Technology: Fermentation, Biocatalysis, Bioseparation, John Wiley & Sons Inc.,
1999, pp. 267–291.
[63] P.L. McCarty, Biotechnol. Bioeng. 97 (2007) 377–388.
[64] U. von Stockar, V. Vojinovic, T. Maskow, J. Liu, Chem. Eng. Proc. 47 (2008) 980–
990.
[65] J.S. Liu, V. Vojinović, R. Patiño, T. Maskow, U. von Stockar, Thermochim. Acta
458 (2007) 38–46.
[66] T. Maskow, K. Wolf, W. Kunze, S. Enders, H. Harms, Thermochim. Acta 543
(2012) 273–280.
[67] D.T. Jones, D.R. Woods, Microbiol. Rev. 50 (1986) 484–524.
[68] E. Gnaiger, R.B. Kemp, Biochim. Biophys. Acta 1016 (1990) 328–332.
[69] W.J. Kong, J.B. Wang, C. Jin, Y.L. Zhao, C.M. Dai, X.H. Xiao, Z.L. Li, Appl.
Microbiol. Biotechnol. 83 (2009) 1183–1190.
[70] W. Kong, Z. Li, X. Xiao, Y. Zhao, P. Zhang, J. Therm. Anal. Calorim. 102 (2010)
331–336.
[71] J. Wang, D. Cheng, N. Zeng, H. Xia, Y. Fu, D. Yan, Y. Zhao, X. Xiao, J. Therm. Anal.
Calorim. 102 (2010) 137–142.
[72] S. Liu, Y. Zhao, N. Zeng, T. Liu, Y. Zhang, B. Han, J. Li, L. Wang, R. Wang, M. Gong,
J. Therm. Anal. Calorim. 116 (2014) 491–497.
[73] W.-J. Kong, X.-Y. Xing, X.-H. Xiao, Y.-L. Zhao, J.-H. Wei, J.-B. Wang, R.-C. Yang,
M.-H. Yang, Appl. Microbiol. Biotechnol. 96 (2012) 503–510.
[74] C. Vázquez, N. Lago, J.L. Legido, I. Arias, L.M. Casás, M.M. Mato, J. Therm. Anal.
Calorim. 113 (2013) 1415–1420.
[75] P. Bataillard, E. Steffgen, S. Haemmerli, A. Manz, H.M. Widmer, Biosens.
Bioelectron. 8 (1993) 89–98.
[76] J.M. Köhler, E. Kessler, G. Steinhage, B. Gründig, K. Cammann, Microchim. Acta
120 (1995) 309–319.
[77] J. Lerchner, J. Seidel, G. Wolf, E. Weber, Sens. Actuators, B 32 (1996) 71–75.
[78] W. Lee, W. Fon, B.W. Axelrod, M.L. Roukes, Proc. Natl. Acad. Sci. U.S.A. 106
(2009) 15225–15230.
[79] F.E. Torres, P. Kuhn, D. De Bruyker, A.G. Bell, M.V. Wolkin, E. Peeters, J.R.
Williamson, G.B. Anderson, G.P. Schmitz, M.I. Recht, S. Schweizer, L.G. Scott,
J.H. Ho, S.A. Elrod, P.G. Schultz, R.A. Lerner, R.H. Bruce, Proc. Natl. Acad. Sci.
U.S.A. 101 (2004) 9517–9522.
[80] F.E. Torres, M.I. Recht, J.E. Coyle, R.H. Bruce, G. Williams, Curr. Opin. Struct.
Biol. 20 (2010) 598–605.
[81] U. von Stockar, I.W. Marison, Adv. Biochem. Eng. Biotechnol. 40 (1989) 94–
140.
[82] M. Türker, Chem. Eng. Commun. 190 (2003) 573–598.
[83] L. Regestein, H. Giese, M. Zavrel, J. Büchs, Biotechnol. Bioeng. 110 (2013) 180–
190.
[84] S. Paufler, M.-T. Weichler, H. Harms, T. Maskow, Thermochim. Acta 569 (2013)
71–77.
[85] L. Wegstedt, US Patent 4054056, 1975.
[86] L. Regestein, A. Wolf, H.J. Schneider, T. Maskow, F. Mertens, J. Büchs, J.
Lerchner, Thermochim. Acta 544 (2012) 10–16.
[87] A.E. Beezer, K.A. Bettelheim, R.D. Newell, J. Stevens, Sci. Tools 21 (1974) 13–16.