THEMATIC NETWORK ON

HarsNet HAZARD ASSESSMENT OF HIGHLY REACTIVE SYSTEMS

4. Thermal Screening

This HarsBook text has been prepared by a HarsNet working

group. The text has been prepared in good faith but the authors
accept no responsibility for the consequences of application of the
information contained herein. The HarsNet Thematic network is
European Community Project BET2-0572 funded in part by
contract number BRRT-CT98-5066.

Keywords: thermal, screening, stability, exotherm, onset, decomposition, kinetics,

enthalpy

4.1. Introduction

The general objective of thermal screening of chemicals and mixtures is to identify

whether the sample can undergo an exotherm process and the temperature range of its
occurrence, which provides a preliminary indication of potential chemical reaction
hazards. The focus lies here on secondary reactions. Depending on the calorimeter used
additional information on the amount and rate of released heat, kinetics (normal or
autocatalytic) and pressure build up can be obtained. Thermal screening is the first
experimental stage in a hazard assessment.

4.2. Principle

Calorimeters for thermal screening are available from various producers in different
specifications, including DSC and various forms of DTA (single and twin set-up). All
screening calorimeters are characterised by the small sample size they require (mg- to g-
scale) and by the speed at which measurements can be performed. They are particularly
useful for:

· Screening of a large number of samples.

· Screening of highly unstable substances.
· Screening of samples, which are only available in laboratory quantities, such as
reactants, isolated intermediates, products and reaction mixtures.
· Unintended mixtures, e.g. contamination with rust

In a typical screening apparatus a small amount of sample is placed in a pressure

resistant metal or glass container and is heated at a constant rate (0.1 to 10K/min) in the
temperature range of -20 to 500°C. Besides this scanning mode, isothermal experiments
are also used for certain applications. Sensors measure the temperature of the sample or
sample container relative to a reference. In most of the available apparatuses this
temperature difference can be calibrated in units of rate of heat generation. Samples
which absorb a known amount of heat at a specific temperature, for instance the melting

4-1
INDUSTRIAL AND MATERIALS TECHNOLOGIES PROGRAMME OF THE EUROPEAN COMMISSION. PROJECT: BET2-0572
COORDINATION: Prof. Dr. R. Nomen Tel. +34-93-267 20 00 Fax. +34-93-205 62 66 http://www.iqs.url.es/harsnet
 THEMATIC NETWORK ON
HarsNet HAZARD ASSESSMENT OF HIGHLY REACTIVE SYSTEMS

of a known quantity of indium, are used for calibration. There is either no mixing or
only very limited mixing in the test cells.

A tabulated comparison of some screening calorimeters can be found in section 5.

Near-adiabatic gram-scale methods, such as RSST or ARC may also be used for
screening, but are not considered here. These methods are described in chapter 5 on
“Adiabatic calorimetry.”

4.3. The sample and sample container

For screening experiments closed high-pressure containers are typically used to

suppress/limit vaporisation and keep volatile decomposition products contained, which
may contribute to further decomposition. To ensure all compounds are kept in the
sample container, the mass of the filled sample container should be weighed before and
after a screening experiment.

The decomposition of a sample can depend strongly on its impurity profile. It is

therefore important that the screened chemicals or mixtures stem from the investigated
process. A sample taken from a heterogeneous mixture most likely does not represent
the mixture. This must be considered in the evaluation. The preparation of a
heterogeneous mixture in the sample container can generate more meaningful
information.

A disadvantage of using small sample sizes is that the sample has an unfavourable ratio
of surface to volume, which makes interactions with the container wall, the surrounding
air, etc. more likely. Catalytic as well as stabilising effects are possible, so that the
measured effect may not represent properties of the sample itself. For instance, many
compounds containing chlorine give artificial peaks when using steel or gold containers.
This kind of interaction can usually be eliminated by running screening experiments
with two containers of different material, e.g. steel and glass (Grewer, 1994).
Interactions with the surrounding air etc. can be eliminated by preparing the sample in a
glove-box with an inert atmosphere.

4.4. Evaluation

The previous section makes clear, that inherent features of thermal screening
measurements are considerable uncertainties and that an ideal evaluation should not be
based on one measurement only.

The interpretation and evaluation of screening experiments are often difficult because
the peak limits can not be clearly recognised and only rarely does an exothermic peak
appear independent from other peaks. However, using a different heating rate can
sometimes separate previously unseparated peaks. An illustration and interpretation of
different combinations of several peaks in screening experiments using some typical
examples of substances and mixtures can be found in /1, page 68/.

The following subsections describe the evaluation of a single peak.

4-2
INDUSTRIAL AND MATERIALS TECHNOLOGIES PROGRAMME OF THE EUROPEAN COMMISSION. PROJECT: BET2-0572
COORDINATION: Prof. Dr. R. Nomen Tel. +34-93-267 20 00 Fax. +34-93-205 62 66 http://www.iqs.url.es/harsnet
 THEMATIC NETWORK ON
HarsNet HAZARD ASSESSMENT OF HIGHLY REACTIVE SYSTEMS

4.4.1. Qualitative Evaluation

The most obvious result of a screening test is the direction of the resulting peak
indicating an exothermic or endothermic process. The other information is the onset
temperature1, Tonset, the peak temperature and the peak shape, where the onset
temperature characterises the thermal stability, and the peak shape (size and sharpness)
gives an indication on the hazard potential of the sample (figure 4.1).

TO = Onset temperature TP
TP = Peak temperature
Q = Heat of reaction
heat generation rate

exo

Area Q

TO

temperature / time

Figure 4.1. Idealised screening curve of an exothermic reaction

In general, the larger and sharper the peak, the more hazardous the decomposition
reaction. However, low activation energy processes, which cause broad peaks, should be
considered when extrapolating the data to adiabatic conditions.

The preliminary assessment of the thermal stability of chemicals or mixtures at

production scale is often based on a rule-of-thumb using Tonset of the secondary
reaction2. Such a rule-of-thumb states that if the maximum operating temperature of a
process is X Kelvin lower than Tonset, the operation will not experience this secondary
reaction, and it is not necessary to obtain more detailed information by other means. For
instance, a safety margin of 100K is often used for DSC measurements. This kind of
rule-of-thumb can easily be misused, since the value of Tonset depends on the sensitivity
of the apparatus, the sample container, the heating rate and is sometimes difficult to
determine - especially if the signal is noisy and/or curvy. The DSC safety margin of
100K is based on the assumption that the decomposition follows zeroth order kinetics
and has a defined temperature dependence. It has been shown that reactions with
activation energies below 80kJ/mol will violate the 100K rule (Hofelich, 1989). Hence,
the 100K rule was extended to ensure that low activation energy decompositions are
excluded. Beside the 100K safety margin, a shift of the peak temperature of less then
40K is demanded when changing the heating rate by a factor of ten (Steinbach, 1995).

1 The observed Tonset is the temperature at which the sample shows the first observable instrumental
response due to a reaction
2
See remark at the end of this subsection!
4-3
INDUSTRIAL AND MATERIALS TECHNOLOGIES PROGRAMME OF THE EUROPEAN COMMISSION. PROJECT: BET2-0572
COORDINATION: Prof. Dr. R. Nomen Tel. +34-93-267 20 00 Fax. +34-93-205 62 66 http://www.iqs.url.es/harsnet
 THEMATIC NETWORK ON
HarsNet HAZARD ASSESSMENT OF HIGHLY REACTIVE SYSTEMS

Another rule-of-thumb for DSC measurements has been proposed. It assumes a

comparatively low activation energy of 50kJ/mol, zeroth order kinetics and a detection
limit of 20W/kg for the DSC (Keller, 1997). The maximum allowable process
temperature Tp,max is regarded as safe if the adiabatic time to maximum rate of the
secondary reaction is greater than or equal to 24 hours, which leads to the following
correlation:

TP ,max = 0.65 × Tonset + 50K . [Eq. 4.1]

A more general correlation between Tp,max and Tonset of a DSC measurement assuming
zeroth order kinetics (Dransfeld, 2001) is:

2
R × Tonset æ tmr24 × E A × q onset ö
TP ,max = Tonset - × lnçç ÷
2 ÷
[Eq. 4.2]
EA c
è P × R × ( Tonset + 50 K ) ø

where R is the universal gas constant, EA the activation energy, tmr24 the adiabatic time
to maximum rate of 24 hours, qonset the detection limit of the DSC and cp the heat
capacity.

The kind of kinetics (normal or autocatalytic) of the heat generation can be qualitatively
determined using the peak profile of an isothermal measurement.

Remark:
The definition of an upper safe temperature limit using a safety margin or an adiabatic
time to maximum rate is based on zeroth order kinetics. This simplifying assumption is
conservative for normal (nth order) kinetics but not necessarily for autocatalytic kinetics.
This means, that the interpretation of screening traces to large scale manufacture
requires great care and experience in identifying reactions or decompositions that are not
proceeding via normal kinetics, i.e. screening tests are not like running a melting point
determination, where one just reads off the number from the trace!

4.4.2. Quantitative Evaluation

4.4.2.1. Heat Flow Profile

The measured heat flow contains the heat generation rate of the chemical reaction and
other thermal effects3 - the baseline. However, for evaluation only the heat generation
rate of the chemical reaction is of interest. Hence, the heat flow must be corrected prior
to an evaluation.

3E.g. many neat solvents reach their critical temperature (Acetone: 235°C; Isopropanol: 235°C; toluene:
319°C; water: 374°C) during screening measurements which causes an abrupt change in the baseline
profile. The same effect can be encountered when testing solutions.
4-4
INDUSTRIAL AND MATERIALS TECHNOLOGIES PROGRAMME OF THE EUROPEAN COMMISSION. PROJECT: BET2-0572
COORDINATION: Prof. Dr. R. Nomen Tel. +34-93-267 20 00 Fax. +34-93-205 62 66 http://www.iqs.url.es/harsnet
 THEMATIC NETWORK ON
HarsNet HAZARD ASSESSMENT OF HIGHLY REACTIVE SYSTEMS

a) Baseline Correction

The baseline represents the theoretical profile, which would have been measured if the
reaction/decomposition would have taken place without any heat release. The difference
between the measured heat flow and the baseline is the heat production rate of the
reaction. For a baseline correction the peak limits have to be fixed, which can be quite
difficult - especially with a noisy and/or curvy heat flow profile. In the reaction interval
the unknown baseline must be interpolated, which is often an error prone task.
Commercial evaluation software often offers a large variety of interpolation methods,
e.g. linear, integral, spline, horizontal etc. The baseline type determines the obtained
results and should be chosen carefully. For normal scanning experiments an integral
baseline, which shows a smooth profile, is recommended (figure 4.2). For normal
isothermal measurements a baseline horizontal from the peak end is recommended. If
there is any doubt about the peak limits or the baseline profile, a second experiment on
the sample may help with the determination of the peak limits and baseline profile.
heat generation rate

exo

tem perature / tim e

Figure 2: Idealised screening curve with integral baseline

Peak limits and interpolated baseline should be plausible and chosen in a way to give
rather conservative results, i.e. lowest onset temperature and largest heat of reaction or
decomposition.

b) Deconvolution

Ideally, the difference between measured heat flow and baseline, although representing
the heat production rate of the chemical reaction, should not directly be used for a
kinetic or TMR evaluation. This is caused by the convoluting effect of the sample
container acting as a heat buffer, which leads to a broader peak. However, it has to be
pointed out that a TMR evaluation without deconvolution usually leads to conservative
results. The convolution is usually characterised by a time lag constant, tsignal, which can
be obtained by the evaluation of the rising part of a negative melting peak. The “true”
heat production rate of the chemical reaction, q& , can be recalculated (Hemminger,
1989) from the baseline-corrected heat flow q& b using tsignal according to:

q& ( t ) = q& b ( t ) + t signal × q& b [Eq 4.3]

4-5
INDUSTRIAL AND MATERIALS TECHNOLOGIES PROGRAMME OF THE EUROPEAN COMMISSION. PROJECT: BET2-0572
COORDINATION: Prof. Dr. R. Nomen Tel. +34-93-267 20 00 Fax. +34-93-205 62 66 http://www.iqs.url.es/harsnet
 THEMATIC NETWORK ON
HarsNet HAZARD ASSESSMENT OF HIGHLY REACTIVE SYSTEMS

c) Heat of reaction

The integration of the baseline-corrected heat flow over time gives the heat of reaction,
which is a measure of the thermal hazard. Based on the amount of released heat
decisions are made how to proceed with further investigations.

The thermal hazard is usually related to the adiabatic temperature rise, which can be
calculated by dividing the heat of reaction with the heat capacity of the sample. An
adiabatic temperature rise of 50K, which equals a heat of reaction of about 100J/g for a
typical organic liquid, is often used as lower hazard limit. This means, if an exothermic
process generates less than 100J/g than this process is regarded as not hazardous
providing gaseous products or vapour are not generated in significant amounts. If the
sample undergoes several exothermic processes, the generated heat of each peak relative
to the temperature differences of their onset temperatures has to be considered.

The investigated sample may be subject to transport regulations. The UN committee on

transport of dangerous goods has defined a value of 300J/g as the energy limit for tests
on self-reactivity. If this limit is exceeded, further tests are needed as outlined by the
United Nations recommendations.

If the heat of decomposition exceeds 500J/g, especially in combination with a sharp

peak, the sample may be an explosive4. If the decomposition energy is higher than
500J/g further tests on explosibility are recommended.

d) Kinetics

In principle, screening calorimeters offer the capability to determine kinetic parameters

from isothermal and scanning experiments. Calculation methods for establishing
reaction kinetics from several scanning experiments are proposed in the literature e.g.
/8-10/. It was shown that reliable kinetic parameters can be obtained by quantitative
screening methods, such as DSC (Wagner, 1996). However, the determination of
reliable kinetic parameters is limited to simple homogenous reactions. Kinetic
parameters obtained from complex reactions give only an indication and can not be used
for quantitative interpretations or predictions.

From one scanning measurement an estimation of an overall activation energy can be

obtained when plotting the natural logarithm of the corrected heat flow vs. the reciprocal
absolute temperature. The initial slope represents the negative activation energy divided
by the universal gas constant. From several scanning measurements at different heating
rates an estimated activation energy can be obtained using the Ozawa plot (Ozawa,
1970) - the natural logarithm of the heating rate vs. the reciprocal absolute peak
temperature. The slope of the resulting straight line represents the negative activation
energy divided by the universal gas constant.

4Thisenergy limit is based on the experience that an explosive with a decomposition energy of less than
500J/g is not known (Grewer. 1994).
4-6
INDUSTRIAL AND MATERIALS TECHNOLOGIES PROGRAMME OF THE EUROPEAN COMMISSION. PROJECT: BET2-0572
COORDINATION: Prof. Dr. R. Nomen Tel. +34-93-267 20 00 Fax. +34-93-205 62 66 http://www.iqs.url.es/harsnet
 THEMATIC NETWORK ON
HarsNet HAZARD ASSESSMENT OF HIGHLY REACTIVE SYSTEMS

In isothermal mode the activation energy can be obtained by plotting the natural
logarithm of the maximum heat production rate versus the reciprocal absolute
temperature. The slope of the resulting straight line represents the negative activation
energy divided by the universal gas constant.

e) Time to Maximum Rate (TMR)

TMR can be used as an alternative to the use of a safety margin in order to define an
upper safe temperature limit of a process. The calculation of TMR is based on several
isothermal measurements at different temperatures. Using the maximum heat production
rate and the isothermal temperature, the TMR for a given set-temperature Ts can be
approximated by extrapolation, assuming a reaction of second (Townsend, 1980) or
zeroth order (Grewer. 1994):

2
c × R × TS
TMR = P [Eq. 4.4]
q& (TS ) × E A

where cp is the heat capacity, R the universal gas constant, q& (Ts) the heat production
rate at Ts and EA the activation energy. Even though the application of TMR seems to be
straightforward, one has to consider that EA is typically determined from three
isothermal measurements, i.e. from three data points, and q& (Ts) is obtained by
extrapolation far outside the regression interval. Taking the temperature where TMR
equals 24 hours and a correlation coefficient for three points of 0.999, the time span can
have an uncertainty of three hours. For a correlation coefficient of 0.99 the uncertainty
may be as big as eight hours!

As a rule-of-thumb: A secondary reaction does not generally pose a threat to the process
if TMR is longer than 24 hours at the highest attainable process temperature.

4.4.2.2. Pressure Profile

Some calorimeters enable the measurement of the pressure on-line indicating the
amount and rate of gas generation, which can afford valuable information about
hazardous gas production. The plot of the logarithm of the pressure vs. the reciprocal
absolute temperature gives an indication whether the investigated sample generates
gaseous products. An approximately straight line in such an Antoine plot indicates that
the pressure, in the sample container, rises due to vaporisation rather than the generation
of non-condensable products.

Considerable deviations of the slope from 10.5 times the absolute boiling temperature of
the liquid5 indicate possible generation of non-condensable products.

5
Many organic (non-polar) liquids follow the Trouton rule reasonably well, i.e. the heat of evaporation
over the universal gas constants is approx. 10.5 times the absolute boiling temperature of the liquid.

4-7
INDUSTRIAL AND MATERIALS TECHNOLOGIES PROGRAMME OF THE EUROPEAN COMMISSION. PROJECT: BET2-0572
COORDINATION: Prof. Dr. R. Nomen Tel. +34-93-267 20 00 Fax. +34-93-205 62 66 http://www.iqs.url.es/harsnet
 THEMATIC NETWORK ON
HarsNet HAZARD ASSESSMENT OF HIGHLY REACTIVE SYSTEMS

Since the pressure measurement is not convoluted, the time lag constant, tsignal, can be
checked or determined by measuring a simple gas producing reaction, e.g. an azo
decomposition. The profile of the heat production rate and pressure rate should be
identical if the correct tsignal was used for deconvolution.

4-8
INDUSTRIAL AND MATERIALS TECHNOLOGIES PROGRAMME OF THE EUROPEAN COMMISSION. PROJECT: BET2-0572
COORDINATION: Prof. Dr. R. Nomen Tel. +34-93-267 20 00 Fax. +34-93-205 62 66 http://www.iqs.url.es/harsnet
 THEMATIC NETWORK ON
HarsNet HAZARD ASSESSMENT OF HIGHLY REACTIVE SYSTEMS

4.5. Comparison of Thermal Screening Calorimeters

Temperature

Raw data as
Calorimeter

sample size

range in °C

function of

Sensitivity

in mW/kg
Tabulated
operation
Mode of

Price in
Stirring
Typical

kEuro
time
DSC
Heat Flow 2-10 mg -50 to Isothermal no T, q 20 35
700 T-ramp
Power 2-10 mg -50 to Isothermal no T, q 20 35
Comp. 700 T-ramp
Twin DTA
TAM 0.5-2 g 20 to 80 Isothermal yes Q 10-4 70
C80 0.5 –1.5 20 to Isothermal yes6 T, q, p 1 60
g 300 T-ramp
IET T-ramp T, DT
Single DTA
Carius tube 5-20 g 20 to T-ramp no T, DT, p
400
DPT T-ramp T, p
Sikarex 1
Radex 1-3 g 20-400
T-ramp no T, DT
Sedex 2-100 g 0-400
Isothermal yes T, DT, p 80
T-ramp
Adiabatic
u
TS 1–7g 20 to Isothermal & no T, DT, p 26
400 Several T –
ramps
6
Mixing performance is limited. On-line pressure measurement not possible when
mixing.

4-9
INDUSTRIAL AND MATERIALS TECHNOLOGIES PROGRAMME OF THE EUROPEAN COMMISSION. PROJECT: BET2-0572
COORDINATION: Prof. Dr. R. Nomen Tel. +34-93-267 20 00 Fax. +34-93-205 62 66 http://www.iqs.url.es/harsnet
 THEMATIC NETWORK ON
HarsNet HAZARD ASSESSMENT OF HIGHLY REACTIVE SYSTEMS

4.6. References

Grewer, T. Thermal Hazards of Chemical Reactions; Elsevier B.V., 1994

Hemminger, W.F.; Cammenga, H.K. Methoden der thermischen Analyse. Springer-

Verlag, 1989

Hofelich, T.C.; Thomas, R.C. Int. Symp. on Runaway Reactions; AIChE 1989; p.74

Hugo, P. et al; Z. Naturforsch. 50a (1995); p.549

Keller, A. et al; J. Loss Prev. Ind. 10 (1997); p. 31

Kissinger, H.; Anal. Chem. 29 (1957); p.1702

Ozawa, T.; J. Therm. Anal. 2 (1970); p. 301

Ozawa, T.; J. Therm. Anal. 9 (1976); p.369

Steinbach, J.; Chemische Sicherheitstechnik; VCH, 1995

Townsend, D.I..; Tou, J.C.; Thermochim. Acta 37 (1980); p. 1

United Nations; Recommendations on the transport of dangerous goods; Test and

criteria, 2nd revised Edition, 1995

Wagner, S.; Experimental and theoretical investigations of the use of Differential

Scanning Calorimetry as a thermokinetic measuring method (in German); PhD Thesis;
TU Berlin, 1996

4-10
INDUSTRIAL AND MATERIALS TECHNOLOGIES PROGRAMME OF THE EUROPEAN COMMISSION. PROJECT: BET2-0572
COORDINATION: Prof. Dr. R. Nomen Tel. +34-93-267 20 00 Fax. +34-93-205 62 66 http://www.iqs.url.es/harsnet