A study of self-accelerating decomposition temperature (SADT) using reaction calorimetry

Abstract

Self-accelerating decomposition temperature (SADT) is determined generally by one of four testing methods recommended by the UN orange book, and can be analytically and numerically evaluated by using the calorimetric results of ARC, Setaram C80D and other instruments. The SADTs evaluated by ARC must be under the assumption of zero order reaction kinetics, and require extrapolation to lower temperatures. Using the C80D, however, the reaction can easily be detected in the vicinity of the SADT for many reactive materials due to its higher sensitivity. Therefore, the SADTs evaluated are more accurate, especially for those reactive materials whose reaction mechanism, phase and so on change near the SADT.In the present study, the Setaram C80D and the ARC were used to investigate the decomposition of an asphalt–salts mixture, which had caused a fire in a nuclear fuel facility. The decomposition mechanism of this mixture was complex but the high sensitivity of the C80D enabled it to be elucidated, and a reasonable estimate of SADT was obtained. The estimated SADT from the ARC was about 70K higher, due to the first two steps of the decomposition being undetected.In the estimation of SADT, the value of acquiring kinetic data close to the SADT can hardly be overstated.

Full text

Journal of Loss Prevention in the Process Industries 14 (2001) 331–336
www.elsevier.com/locate/jlp
A study of self-accelerating decomposition temperature (SADT)
using reaction calorimetry
Jinhua Sun
*
, Yongfu Li
1
, Kazutoshi Hasegawa
National Research Institute of Fire and Disaster, 14-1,3-chome Nakahara, Mitaka, Tokyo 181-8633, Japan
Abstract
Self-accelerating decomposition temperature (SADT) is determined generally by one of four testing methods recommended by
the UN orange book, and can be analytically and numerically evaluated by using the calorimetric results of ARC, Setaram C80D
and other instruments. The SADTs evaluated by ARC must be under the assumption of zero order reaction kinetics, and require
extrapolation to lower temperatures. Using the C80D, however, the reaction can easily be detected in the vicinity of the SADT for
many reactive materials due to its higher sensitivity. Therefore, the SADTs evaluated are more accurate, especially for those reactive
materials whose reaction mechanism, phase and so on change near the SADT.
In the present study, the Setaram C80D and the ARC were used to investigate the decomposition of an asphalt–salts mixture,
which had caused a fire in a nuclear fuel facility. The decomposition mechanism of this mixture was complex but the high sensitivity
of the C80D enabled it to be elucidated, and a reasonable estimate of SADT was obtained. The estimated SADT from the ARC
was about 70 K higher, due to the first two steps of the decomposition being undetected.
In the estimation of SADT, the value of acquiring kinetic data close to the SADT can hardly be overstated.  2001 Elsevier
Science Ltd. All rights reserved.
Keywords: Self-accelerating decomposition temperature; Heat flux calorimeter; Accelerating rate calorimeter; Organic peroxides; Asphalt–salts mix-
ture
1. Introduction
Self-accelerating decomposition temperature (SADT)
is the lowest ambient temperature at which the tempera-
ture increase of a chemical substance is at least 6 K in
a specified commercial package during a period of seven
days or less (United Nations, 1999). The SADT is a very
important parameter for assessing the safety manage-
ment of reactive substances in storage, transportation
and use.
SADT is generally determined by one of four testing
methods recommended by the United Nations (UN)
orange book, which are the United States SADT test, the
* Corresponding author. Permanent address: Department of Chemi-
cal Engineering, Huainan Institute of Technology, Huainan, Anhui
232001, People’s Republic of China. Tel.: +81-422-44-8331; fax: +81-
422-42-7719.
E-mail address: sun@fri.go.jp (J. Sun).
1
Permanent address: Nanjing University of Science and Tech-
nology, Chemical College, Nanjing 210094, People’s Republic of
China.
0950-4230/01/$ - see front matter  2001 Elsevier Science Ltd. All rights reserved.
PII: S0950-4230(01)00024-9
adiabatic storage test, the isothermal storage test and the
heat accumulation storage test (United Nations, 1999).
There are two significant disadvantages of these four
methods. Firstly, the sample mass is so large, varying
from 400 g to 200 kg, that the test cannot be performed
without a special enclosure to control the hazards arising
from ignition, explosion and /or toxic products. Sec-
ondly, the time scale for the tests is long. These disad-
vantages encouraged the development of more con-
venient estimation methods (Fisher & Goetz, 1991,
1993; Wilberforce, 1981; Yu & Hasegawa, 1996; Whit-
more & Wilberforce, 1993; Li & Hasegawa, 1998;
Kotoyori, 1988, 1993). In these methods, SADTs are
analytically and numerically estimated from the results
of reaction calorimetry, such as an accelerating rate
calorimeter (ARC) and heat flux calorimeter (Setaram
C80D). These instruments have small sample size and/or
have their own protective casing, which enables safe
operation in an open laboratory.
By the use of ARC, Wilberforce (1981) estimated ana-
lytically the SADTs of organic peroxides by extrapolat-
ing the reaction rate data into the (lower) temperature

332 J. Sun et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 331–336
Nomenclature
A frequency factor (s
⫺1
)
Cp specific heat (J/g)
dH/dt over all heat flow (J/s)
E activation energy (J/mol)
⌬H heat of reaction (J/g)
M mass of reactant (g)
M
0
initial mass of reactant (g)
n reaction order
S contact area between drum surface and surroundings (m
2
)
R gas constant (J/K/mol)
T temperature of system (K)
T
NR
no return temperature (K)
T
0
surrounding temperature (K)
U overall heat transfer coefficient (J/m
2
/K/s)
⌽
thermal dilution
x conversion rate
range of the SADT under the assumption of a zero order
reaction and the models of Semenov. In many cases, the
estimated SADTs agreed well with the United States
SADT test. However, for some chemicals whose reac-
tions do not follow the assumed reaction mechanism
and/or whose phases change in the vicinity of their
SADT, the evaluated SADTs are not consistent with the
observed ones.
Whitmore and Wilberforce (1993) proposed an
improved method using the ARC in combination with
the thermal activity monitor (TAM). Because the TAM
has a high sensitivity, this method allows interpolative
rather than extrapolative estimation. Although this
method gives more reliable results than the method using
the ARC alone, it still has some problems when the reac-
tion mechanism is complex and/or a phase-change
occurs near the SADT. Hasegawa (1998) pointed out
that the ARC/TAM method is probably sufficient for
practical purposes of evaluating the SADT around 50°C,
but when the estimated SADT is far from the measuring
temperature range of the TAM, there is the danger that
the SADT includes a large error. Li and Hasegawa
(1998) solved numerically the heat balance equations for
a simple reaction, an auto-catalytic reaction and a
pseudo-autocatalytic reaction under the conditions of
both US-SADT test and the Dewar vessel test. They
showed that the calculated SADTs coincided with the
experimental ones within a few degrees.
Making the chemical and physical mechanism of
decomposition clear and obtaining the necessary thermal
data (specific heat, latent heat, heat of decomposition and
so on) in the temperature range of the SADT is essential
for a complete understanding of SADT. However,
obtaining all this data is not easy. Therefore, an instru-
ment, such as the C80D, is required to have both high
sensitivity, 10 µW at least, and many functions.
In this paper, the chemical kinetic parameters of a
complicated mixture were determined using heat flux
data measured in the C80D. SADT was estimated using
these kinetic parameters, based on the Semenov model.
The SADTs obtained by this method are compared to
those obtained using Wilberforce’s ARC method.
2. Comparison of SADT
Hasegawa’s group (Yu & Hasegawa, 1996; Li &
Hasegawa, 1998) has developed a method to determine
the chemical kinetic parameters and to evaluate SADTs
for the self-reactive materials, whose reaction schemes
are simple, auto-catalytic and pseudo-autocatalytic. The
evaluated SADTs for organic peroxides and blowing
agents (Li & Hasegawa, 1998; Yu & Hasegawa, 1996)
are compared in Table 1 with those from the ARC
method and the US SADT test (Fisher & Goetz, 1991,
1993; Wilberforce, 1981).
It is clear from the Table 1 that the SADTs calculated
from C80D data are the same or a little lower than US
test SADTs, that is to say, there is no practical difference
between them. On the other hand, the SADTs calculated
from ARC data deviate from the US test ones more
widely.
Although the mean difference of ARC-US is approxi-
mately the same as the mean difference of C80D-US,
the standard deviation of the mean difference of ARC-
US is two times larger than that of C80D-US. Obviously
the C80D method gives generally more reliable SADTs
than the ARC method. This is because the higher sensi-

333J. Sun et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 331–336
Table 1
Comparison of SADT values (C80D versus US and ARC versus US)
Materials SADT for 25 kg Comparison
package (°C) (°C)
C80D
a
ARC
b
US
b
C80D- ARC-
US US
t-Butyl peroxybenzoate 57 53 57 0 ⫺4
Cumene hydroperoxide (83%) 79 89 79 0 10
t-Butyl peroxy acetate 65 70
Di-t-butyl peroxide 80 78 80 0 ⫺2
Benzoyl peroxide (75%) 80 86 ⬎49
Acetylacetoneperoxide 55 54 64 ⫺9 ⫺10
Azobisisobutyronitrile 49 60 50 ⫺110
p-Mentane hydroperoxide(53%) 67
Di-isopropylbenezene 65
hydroperoxide
α,α⬘-Bis(t-butylperoxy-m-80
isopropyl)benzene
Acetylacetonperoxide (37%) 51 65 ⫺14
t-Butyl peroxypivilate (75%) 22 27 ⫺5
Bis (2-chlorobenzoyl) peroxide 51.3 45 6.3
(49%)
Lauroyl peroxide 46.8 45 49 ⫺2.2 ⫺5
2,2⬘-Azobis 48 50 ⫺2
2-5-Dimethyl-2,5-bis (benzoyl 70 69 ⫺1
peroxy) hexane (75%) in H
2
O
Mean difference ⫺1.01 ⫺2.33
Standard deviation 3.99 7.94
Number of comparisons 8 9
a
Data from references (Li & Hasegawa, 1998; Yu & Hasegawa,
1996).
b
Data from references (Fisher & Goetz, 1991, 1993; Wilber-
force, 1981).
tivity and multi-functionality of the C80D removes the
uncertainty of the extrapolation required by the ARC
method.
3. Assessing the SADT of complicated mixture
3.1. Reactivity of asphalt–salts mixture
In this study an asphalt–salts mixture is studied. A fire
and explosion accident happened at the Bituminization
Demonstration Facility in the Tokai Works of the Power
Reactor and Nuclear Fuel Development Corporation in
Japan on 11 March, 1997. A 220 l steel drum was filled
with the asphalt–salts mixture at about 180°C and about
20 h later a fire broke out from the drum during a cooling
period. The asphalt–salts mixture was produced by mix-
ing low-level radioactive liquid waste, containing
NaNO
3
, NaNO
2
,Na
2
CO
3
,NaH
2
PO
4
and other salts, with
high temperature asphalt in an extruder. It was thought
that oxidation–reduction reactions between the asphalt
and a salt in the mixture, having produced under the
three specialized conditions, were initiated at the filling
temperature, and their heat release accumulated in the
drum, leading to a run-away reaction (Hasegawa & Li,
2000).
In order to elucidate this process of the fire-cause, an
asphalt–salts mixture was prepared in the laboratory and
its reactivity and heat generation were tested in both the
C80D and the ARC.
3.1.1. C80D test
As is well known, in a general heat analysis experi-
ment, when the programmed rate of temperature increase
is lower, the reaction can be detected at a lower tempera-
ture, at least up to a point. This is illustrated typically in
Fig. 1. Calculated heat flux curves of a general reaction
(assumed activation energy E=80,000 J/mol, frequency
factor A=10
5
s
⫺1
and ⌬H=⫺1000 J/g) at three different
heating rates are shown. It is seen that the maximum
heat release and the end of reaction are shifted to lower
temperatures as the heating rate is reduced. In other
words, a low heating rate experiment is appropriate to
examine reactive heat behavior at lower temperatures.
Therefore, in this study, the very slow temperature rise
rate (0.01°C/min) of which the C80D is capable was
used.
Fig. 2 shows the heat flux of the asphalt–salts mixture
(sample mass: 0.500 g, temperature rise rate:
0.01°C/min) measured in the C80D. The heat flux
increases very slowly and gradually with temperature
increase in the range 158–195°C, keeps almost constant
in the range 195–250°C, and increases sharply above
250°C. Therefore, the reaction of salts with asphalt can
be divided into three stages corresponding to the three
temperature ranges. In the first stage, liquid asphalt could
contact directly with oxidizing salts so the rate-determin-
ing step may be chemical reaction, as indicated by the
dependence of rate upon temperature. As this reaction
proceeds, it can be imagined that a layer of the reaction
product forms on the surface of the salt particles. Thus
in the second stage, when the layer of reaction product
Fig. 1. Calculated results of heat flow for a general reactive material
at different heating rates.

334 J. Sun et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 331–336
Fig. 2. Heat flux curve of asphalt–salts mixture, sample mass:
0.500 g; heating rate: 0.01°C/min.
becomes thicker, the rate becomes diffusion controlled
and almost independent of the temperature. In the third
stage, when the temperature exceeds the melting point
of oxidizing salts, the reaction mechanism changes from
a solid–liquid interface reaction into a liquid–liquid reac-
tion and the heat flux increases exponentially with tem-
perature.
3.1.2. ARC test
Fig. 3 shows the self-heating rate versus temperature
plot of the same sample tested in the ARC (sample mass
is 1.00 g,
⌽
=2.60). The heat release is only detected
when the temperature exceeds 276°C, just corresponding
to the third stage measured by C80D. The first two stages
do not appear in the record. This is because the ARC
has not only a low sensitivity of self-heating rate
(0.02°C/min), but also a high thermal inertia (phi value)
of 2.60.
The sensitivity of the ARC of 0.02°C/min is
Fig. 3. The relationship between self-heating rate and temperature
for asphalt–salts mixture, sample mass=1.00 g,
⌽
=2.60.
unchangeable, while the value of the phi depends on
sample mass. So, in order to measure the heat generation
at lower temperatures, it is favorable to increase the mass
of the asphalt–salts mixture. However, it should be
pointed out that if the mass of the sample is too large
there is a danger that the rapid reaction at higher tem-
peratures could exceed the insulating controlling ability
of the ARC and even lead to the bomb exploding.
We considered that the mass of the asphalt–salts mix-
ture could be safely increased to 1.20 g, giving
⌽
=2.33.
Fig. 4 shows the resulting self-heating rate versus tem-
perature plot. It is clear from Fig. 4 that self-heating can
be detected intermittently in the range of 215.5–275°C,
just corresponding to the second stage of the C80D
result, when the mass of the asphalt–salts mixture is
1.2 g. The measured self-heating rate is almost constant
in this temperature range. It is impossible to calculate the
SADT using such data, although such data do indicate a
low activation energy.
3.2. SADT for asphalt–salts mixture
Applying Wilberforce’s method (Wilberforce, 1981)
based on the Semenov model (Semenov, 1959) to the
ARC data of Fig. 3, the activation energy and SADT for
220-l drum filled with 250 kg asphalt–salts mixture can
be calculated as 203 kJ/mol and 245.5°C.
On the other hand, in order to calculate the chemical
kinetic parameters of reaction materials by using C80D
data, a simple reaction mechanism is assumed to be
dependent on the Arrhenius law, so the rate expression
for the consumption of reaction can be defined as Eq. (1)
dx
dt
⫽A exp(⫺E/RT)(1⫺x)
n
(1)
where x=
M
0
−M
M
0
is conversion rate.
Fig. 4. The relationship between self-heating rate and temperature
for asphalt–salts mixture, sample mass=1.20 g,
⌽
=2.33.

335J. Sun et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 331–336
Substituting x into Eq. (1), the next equation can be
easily obtained
⫺
dM
M
0
dt
⫽A exp(⫺E/RT)
冉
M
M
0
冊
n
(2)
At the initial stage the reactant consumption should
be negligible. Therefore, M must be approximately equal
to M
0
. By multiplying Eq. (2) by the heat of reaction
⌬H, the heat flow of the reaction is obtained as
dH
M
0
dt
⫽⌬HA exp(⫺E/RT). (3)
By plotting the curve of ln
dH/dt
⌬HM
0
versus inverse tem-
perature (T
⫺1
) as Fig. 5. The activation energy (E) and
frequency factor (A) can be easily calculated according
to the reference (Li & Hasegawa, 1998).
According to the Semenov model (Semenov, 1959),
the uniform temperature rise rate in a reaction system is
established by the difference between the rate of heat
generation form the system and the rate of heat transfer
to the environment as the following equation:
C
p
M
0
dT
dt
⫽⌬HM
n
0
A exp(⫺E/RT)⫺US(T⫺T
0
) (4)
At the no return temperature (T
NR
) both the dT/dt=0
and d(dT/dt)/dT=0 must be held. The solution of sur-
rounding temperature (T
0
) must be equal to the SADT
SADT⫽T
0
⫽T
NR
⫺
冉
RT
2
NR
E
冊
(5)
The SADT of the 220-l drum (the drum; 56 cm in
diameter and 71 cm in filled height) filled with 250 kg
Fig. 5. Thermo-kinetic parameters estimated from relationship
between ln(dH/dt)/M
0
/⌬H and T
⫺1
.
Table 2
Kinetic parameters and SADTs of the drum filled with asphalt–salts
mixture (heat of reaction, 1352 J g
⫺1
, overall heat transfer coefficient
of a drum, 4.388 J m
⫺2
K
⫺1
s
⫺1
)
Method Activation Frequency SADT (°C)
energy (E) factor (A)s
⫺1
kJ/mol
C80D 125 1.26×10
7
173.0
ARC 203 – 245.5
asphalt–salts mixture was calculated under the assump-
tion of Semenov’s model. The kinetic parameters and
SADT so obtained are compared with those obtained by
applying Wilberforce’s model to the ARC data in
Table 2.
4. Discussion
The SADT of 173°C obtained from the C80D data is
consistent with the incident where self-heating occurred
from the filling temperature of 180°C. The ARC-derived
SADT is much higher and, irrespective of the cause of
the incident, is palpably in error. It is clear that in this
case, the low sensitivity of the ARC coupled with the
complexity of the decomposition process led to the very
large error, and that in such cases the use of ARC data
alone to estimate SADT is inappropriate. However, it
has to be said that in the absence of other data, such
complexity will not be apparent, so it seems difficult to
justify the use of ARC data alone in any case.
Generally speaking, however, it also has to be said
that the C80D may not always be sensitive enough to
estimate the SADT. If the thermal conductivity was very
low and/or the mass very large, the C80D could experi-
ence similar problems to the ARC.
In Fig. 6, three cases of extrapolation are shown. The
Fig. 6. Three cases of the errors in SADTs as a general rule.

336 J. Sun et al. / Journal of Loss Prevention in the Process Industries 14 (2001) 331–336
solid line A–B shows the relationship between the self-
heating rate and temperature as measured by ARC.
When the SADT is estimated, the reaction is extrapo-
lated into the lower temperature range (straight line A–
A
0
). Three cases of real curves can be generally con-
sidered. The first case (A–A
1
) means that there is no
change in physical phenomenon and chemical reaction
mechanism below the temperature of point A. The calcu-
lated SADT must agree well with the measured SADT
in this case. In the second case (A–A
2
), the reaction is,
for example, pseudo-autocatalytic due to a phase tran-
sition below the point A temperature. The heat gener-
ation of the reaction is mostly counterbalanced by the
heat absorption in the phase transition such as melting.
Therefore, the extrapolated line (A–A
0
) is higher than
the real curve (A–A
2
). In this case, the calculated SADT
will be lower than the real SADT, which will be no prob-
lem in the practical use of SADT from the viewpoint of
safety. On the other hand, there is a big problem in the
third case where the real curve (A–A
3
) of self-heating
rate versus temperature takes position below the line A–
A
0
below the temperature of point A (such as the
asphalt–salts mixture). The calculated SADT will be
higher than the real SADT, which is a very dangerous
situation.
5. Conclusions
In this paper, two SADT-evaluating methods, one
using ARC data and one using C80D data, are compared.
For reactions which follow a simple reaction scheme
without phase change or chemical reaction mechanism
change near the SADT, the SADT obtained by ARC data
is in good agreement with the SADT determined by a
direct measurement test such as in the United States test
method. However, when a physical phenomenon such as
melting or evaporating occurs and/or a chemical reaction
mechanism changes occur between the onset tempera-
ture and the SADT, the reaction-extrapolating method of
the ARC method must lead the evaluated SADT into
error.
With the C80D, the reaction can be usually detected
in the vicinity of the SADT due to its high sensitivity.
Therefore, the C80D method gives generally more
accurate SADTs than ARC method. However, as a gen-
eral rule the sensitivity of C80D may not always be high
enough to evaluate the SADTs, especially for those
reactive materials whose thermal conductivity is very
low or whose mass is huge.
Acknowledgements
The authors thank Mr M.W. Whitmore for his com-
ments and amendments to this paper.
References
Fisher, H. G., & Goetz, D. D. (1991). Determination of self-accelerat-
ing decomposition temperatures using the accelerating rate calor-
imeter. Journal of Loss Prevention in the Process Industries, 4 (4),
305–316.
Fisher, H. G., & Goetz, D. D. (1993). Determination of self-accelerat-
ing decomposition temperatures for self-reactive substances. Jour-
nal of Loss Prevention in the Process Industries, 6 (3), 183–193.
Hasegawa, K. (1998). Evaluating methods for SADTs by the use of
reaction calorimeters. In International Workshop on Safety in the
Transport, Storage and Use of Hazardous Materials. Japan.
Hasegawa, K., & Li, Y. F. (2000). Explosion investigation of asphalt–
salt mixtures in a reprocessing plant. Journal of Hazardous
Materials, A79, 241–267.
Kotoyori, T. (1988). Critical ignition temperatures of chemical sub-
stances. Journal of Loss Prevention in the Process Industries, 2
(1), 16–21.
Kotoyori, T. (1993). Critical temperature for the thermal explosion of
liquids. Combustion and Flame, 95, 307–312.
Li, Y. F., & Hasegawa, K. (1998). On the thermal decomposition
mechanism of self-reaction materials and the evaluating method for
their SADTs. In Ninth International Symposium on Loss Prevention
and Safety Promotion in the Process Industries (pp. 555–569).
Spain.
Semenov, N. N. (1959). Some problems in chemical kinetics and reac-
tivity (Vol. 2, Chap. 8). Princeton University Press.
United Nations. (1999). Recommendations on the transport of danger-
ous goods, manual of tests and criteria (3rd revised ed.).
Wilberforce, J. K. (1981). The use of the accelerating rate calorimeter
to determine the SADT of organic peroxides. Columbia Scientific
Corp., Internal report.
Whitmore, M. W., & Wilberforce, J. K. (1993). Use of the accelerating
rate calorimeter and the thermal activity monitor to estimate stab-
ility temperature. Journal of Loss Prevention, 6 (2), 95–101.
Yu, Y. H., & Hasegawa, K. (1996). Derivation of the self-accelerating
decomposition temperature for self-reactive substances using iso-
thermal calorimetry. Journal of Hazardous Materials, 45, 193–205.