The thermal runaway of a hydrogen-transfer reaction

Article abstract of DOI:10.1021/op000108z

A thermal runaway is reported which occurred during the conversion of an unsaturated alcohol to the saturated ketone via internal hydrogen transfer. An unexpected and extremely rapid rise in temperature and pressure was observed when the reaction was run

Full text of DOI:10.1021/op000108z

Organic Process Research & Development 2002, 6, 304−307
The Thermal Runaway of a Hydrogen-Transfer Reaction
Andrew V. West* and Chris P. Newman
Process Research and DeVelopment, Quest International, Ashford, Kent TN24 OLT, United Kingdom
Abstract:
Scheme 1. Transformation of the TCD Alcohol E to TCD
Ketone A by internal hydrogen transfer
A thermal runaway is reported which occurred during the
conversion of an unsaturated alcohol to the saturated ketone
via internal hydrogen transfer. An unexpected and extremely
rapid rise in temperature and pressure was observed when the
reaction was run at production scale. The vessel was able to
withstand the pressure generated, and no injury or damage
resulted. The possible causes of this incident are considered
here. It should be noted that whilst the rate of a hydrogenation
reaction can generally be controlled by the rate of addition of
hydrogen, this may not be the case where the possibility of
internal hydrogen transfer exists, whether by design or unin-
tentionally. Unsaturated alcohols are perhaps less obvious,
although they are very good examples of molecules where such
a possibility exists. Our pre-scale-up reaction hazard assessment
protocol now considers this point specifically.
This chemistry had been carried out many times previ-
ously, both at laboratory- and small-production scale (up to
2
00 kg) without incident. The scale-up of the reaction had
been rigorously assessed: indeed trials had been carried out
at greater than 210 °C specifically looking for decomposition
or hazardous side reactions. All previous work on this
reaction suggested that the overall rate is governed by the
rate of dehydrogenation of the alcohol, which is slow and
requires elevated temperature. However, when the reaction
was carried out on a larger scale, a highly exothermic event
became apparent at 145 °C. The contents of the reactor self-
heated extremely rapidly, over approximately 1-2 min, to
a temperature in excess of 200 °C (the exact temperature is
not known, as it was above the upper limit of the sensors
being used). This was accompanied by pressurisation of the
reactor to 8.2 bar. Analysis of the contents of the reactor,
by gas chromatography after the event, showed the reaction
products to be as expected and the degree of conversion to
be 90%.
Introduction
We bring to the attention of the readers of this journal
the potential hazards associated with the scale-up of hydrogen-
transfer reactions. This stems from an incident that occurred
some while ago involving such a reaction.
For commercial reasons a large amount of an intermediate
was required in a short time scale. The decision was taken
to satisfy demand for this material by producing a large batch
utilising a process that had already been run on a small
production scale. An improved process was to be developed
for future use.
As a result of this incident, the process concerned was
modified to preclude uncontrolled internal hydrogen transfer,
by derivatisation of the alcohol functional group. It is not
possible to discuss details of these modifications here for
commercial reasons. This modified process has since been
run safely.
The incident related to the transformation of tricyclo-
2,6
2,6
[
5.2.1.0 ]dec-4-en-8-ol (TCD Alcohol E) to tricyclo[5.2.1.0 ]-
decan-8-one (TCD Ketone A) via an internal hydrogen
transfer (Scheme 1). The reaction was carried out in the
presence of a nickel catalyst at elevated temperature. In both
the laboratory and small-production facility (at 200 kg scale)
at Quest this reaction was very slow, requiring a relatively
high loading of catalyst (10% w/w), temperatures in excess
of 170 °C, and approximately 30-40 h to go to completion.
When the reaction was carried out at 1.3 tonne scale, an
extremely rapid temperature rise was observed (after several
hours of heating with no significant rise in the measured
temperature of the reaction mixture, although there is a
question as to whether the heating system was functioning
correctly during this period), accompanied by pressurisation
of the reactor. The reactor being used was a pressure vessel
Discussion
At the time of this event, the necessary resources and
equipment to fully investigate it to today’s standards did not
exist within Quest, although some calculations were made
1
using the approach suggested by van Krevelen. This uses
group contributions to estimate the Gibbs free energy of
formation of organic compounds and hence can be used to
estimate a value for the Gibbs free energy of a reaction. This
work suggested that the hydrogen-transfer reaction is thermo-
dynamically favoured in the temperature range of interest
and that elevated temperature favours it further. Whilst this
approach can be used with the overall hydrogen-transfer
reaction, it is now felt that the lack of an entropy term for
hydrogen precludes its use for looking at the separate
dehydrogenation or hydrogenation steps.
(chosen to reduce the emission of volatiles) and thus was
able to withstand the pressure generated. Neither injury to
personnel nor material release was caused by this incident.
Since this incident a dedicated group has been established
within Quest to investigate chemical reaction hazards and
*
To whom correspondence should be addressed. E-mail: andy.west@
questintl.com.
(1) Van Krevelen, D. W.; Chermin, H. A. G. Chem. Eng. Sci. 1951, 2, 66.
3
04
•
Vol. 6, No. 3, 2002 / Organic Process Research & Development
10.1021/op000108z CCC: $22.00 © 2002 American Chemical Society
Published on Web 04/04/2002

ensure the safety of our processes. The incident itself has
been the subject of a retrospective investigation in an attempt
to learn from it. The techniques applied to this problem are
described later. Unfortunately it has not been possible to
recreate this incident in the laboratory; therefore, it has not
been possible to determine the course the reaction took
experimentally. The reactor being used had previously been
cleaned to a mirror finish to assist in meeting clean-down
standards. It is thought possible that this process introduced
an unknown material that interacted with the catalyst used
later, thus altering its activity and possibly causing a lowering
of activation energy. However, we have not been able to
prove this experimentally.
Table 1. Predicted vapour pressure of TCD Ketone A
temperature [°C]
vapour pressure [bar]
60
0.012
0.078
0.44
1
1
2
00
50
00
1.61
250
300
4.44
10.02
11.60
19.92
28.72
3
3
3
10
50
80
Consideration of the predicted temperature rises as
estimated from theory may help distinguish between these
two possibilities. Whilst the temperature reached cannot be
used directly, as it was above the upper limit of the detectors,
we have used it to estimate the vapour pressure of the
product. The separate dehydrogenation-hydrogenation hy-
pothesis would result in the release of more heat in the
runaway phase than the concomitant hydrogen-transfer
reaction. We have estimated the vapour pressure of the
product of this reaction, the saturated ketone, at the predicted
maximum temperatures (Table 1). It can be seen from these
results that the estimated vapour pressure at the predicted
maximum temperature for the separate dehydrogenation-
hydrogenation case is significantly higher than the 8.2 bar
observed. The vapour pressure estimated at the maximum
temperature suggested for the combined hydrogen-transfer
reaction, however, is in good agreement with the pressure
observed.
The incident may be analysed from a theoretical view-
point. The heats of reaction have been estimated, by various
means, and the effects of this heat release on the pressure of
the system have been considered.
Two possible hypotheses have been put forward to
account for the rapid temperature rise. The reaction con-
cerned is the transfer of hydrogen within the molecule, with
the hydrogen liberated by the dehydrogenation of the alcohol
hydrogenating the alkene double bond. The first hypothesis
is that the dehydrogenation and hydrogenation reactions
became uncoupled (previous experience indicated that the
dehydrogenation had a higher activation energy than the
hydrogenation; this meant that once hydrogen had been
liberated, it rapidly reacted with the double bond). The
alternative hypothesis is that the combined hydrogen-transfer
ran away. These hypotheses are discussed below.
Whilst this is not conclusive evidence, we feel that it lends
credence to the combined hydrogen transfer hypothesis.
However, we have not been able to explain the runaway
unequivocally.
1. The Separate Dehydrogenation and Hydrogenation
Reactions. The production reactor was charged with the
starting materials, including catalyst, at room temperature
and was being heated to its operating temperature when the
incident occurred. The heating phase was slower than
expected, and it has been postulated that during this time
the unsaturated alcohol was dehydrogenating, at least par-
tially (the endothermic nature of the dehydrogenation ac-
counting for the slow heating rate). This is complicated
because during this phase there were also some problems
with the oil-heating system on the reactor. The resulting
hydrogen could have loaded the catalyst (the reactor did not
pressurise significantly during this phase). Once a temper-
ature was reached at which the rate of hydrogenation became
significant, sufficient heat was generated to increase the rate
of both reactions, leading to a thermal runaway.
Experimental Section
This reaction has been studied extensively in the labora-
tory, in an attempt to recreate the incident. We have not been
able to do so.
Differential Scanning Calorimetry (DSC) Studies. The
materials concerned were examined by DSC. Samples of the
materials concerned and mixtures thereof were subjected to
range of test regimes (varying temperature range, heating
rate). The effects of possible contaminants (including acid,
base, and iron filings) were also tested. The incident could
not be recreated in any of these tests. It is possible that the
lack of agitation, inherent in this technique, might prevent
the catalyst and TCD Alcohol E from interacting.
Autoclave Experiments. In an attempt to simulate the
agitated pressure vessel in which the incident occurred
experiments were conducted in small, laboratory-scale auto-
claves. TCD Alcohol E and catalyst, with and without various
possible contaminants, were charged to the autoclave and
heated whilst being agitated. Different heating regimes were
employed, including heating at full power and a slower
heating rate that simulated that observed on the production
scale. Again, it did not prove possible to recreate the incident.
It should be noted that it was not possible in the laboratory
to exactly recreate the agitation conditions experienced in
the full-scale reactor. Since this is a heterogeneous catalysis
This hypothesis relies on the activation energy of the
dehydrogenation reaction only being lowered. Had the
activation energy of the hydrogenation reaction been simi-
larly lowered there would not have been any accumulation
of hydrogen.
2. The Combined Hydrogen-Transfer Reaction. The
alternative hypothesis considered was that the rates of both
the dehydrogenation and hydrogenation reactions were
enhanced to the same degree. The reactions proceeded much
more rapidly than expected at approximately 145 °C. The
heat released was more than could be removed from the
reactor, causing a temperature rise, which caused a further
increase in rate, and thermal runaway ensued.
Vol. 6, No. 3, 2002 / Organic Process Research & Development
•
305

system, agitation will have an effect on reaction rate;
however, it is considered unlikely that this alone explains
the extreme difference in behaviour observed in this instance.
Unfortunately, we were unable to recreate the runaway
reaction in the laboratory. The intention, had the reaction
been reproducible, was to study it using the techniques of
reaction calorimetry, adiabatic calorimetry, and in situ FTIR
analysis. The experimental data generated would have
allowed us to differentiate between the two proposed
mechanisms. However, this did not prove possible.
Reaction calorimetry is used to study a desired synthetic
reaction. It gives data not only on the overall heat liberated,
or absorbed, by a reaction but also on the rate of heat
production or uptake. In this case the overall heat released
would have been the same if the reaction had proceeded by
either separate dehydrogenation/hydrogenation or combined
hydrogen-transfer, but the profile of the heat flow would have
differed markedly. The separate dehydrogenation/hydrogena-
tion reaction would have had distinct endothermic and
exothermic portions, whilst the combined hydrogen-transfer
reaction would have been exothermic throughout.
Table 2. Estimated parameters used for vapour pressure
prediction
parameter
estimated valuea
critical temperature
critical pressure
acentric factor
386 °C
30.8 bar
0.39937
a
All parameters were estimated in Predict.
a thermal runaway, assuming that no further reactions occur
at elevated temperatures. If a further reaction does occur,
for example the exothermic decomposition of a product, the
maximum temperature and pressure reached may be greater
than that predicted from the heat of reaction alone.
Prediction of the maximum temperature reached can be
simplified by treating the reaction mass as an adiabatic
system. This is a valid assumption for large production scale
reactors. In this case the extremely rapid temperature rise
also supports this assumption, as the reactor heating/cooling
system would not have had the capacity to remove sufficient
heat in the time scale required.
A value of 2000 J/kg/K has been used for the specific
heat capacity of the reaction mixture (estimated value). The
predicted adiabatic temperature rises for the two hypotheses
can then be compared. In the first hypothesis the dehydro-
genation reaction occurs during the heating phase so that
the runaway is caused purely by hydrogenation of the double
bond. In this case the predicted adiabatic temperature rise is
approximately 400 K. This figure can be amended to 360 K
because the reaction went to 90% completion. Similarly, the
predicted adiabatic temperature rise for the combined
hydrogen-transfer scenario is approximately 164 K (at 90%
conversion).
Adiabatic calorimetry is applied to the study of runaway
reactions under heat-loss conditions similar to those on plant,
that is essentially adiabatic. It yields data on the rates of
temperature and pressure increase and also the final tem-
perature and pressure attained by a runaway. Such data would
have helped differentiate between the two proposed mech-
anisms.
In situ FTIR can be used to monitor the progress of a
reaction. In this instance it would have been used to
determine whether any unsaturated ketone was present in
the reaction mixture. The presence of such material would
suggest that the dehydrogenation and hydrogenation reactions
had proceeded separately.
The maximum temperature of the reaction mass after the
runaway is thus approximately 505 °C for the separate
dehydrogenation/hydrogenation case or approximately 309
Theoretical Considerations
°
C for the combined hydrogen-transfer case.
The vapour pressure of the system (effectively TCD
The fact that we could not recreate the incident practically
led us to consider a theoretical approach. Various methods
were used to derive an estimate of the heats of reaction of
the dehydrogenation and hydrogenation steps. These included
the use of heats of formation of structurally related com-
Ketone A at this point in time) can then be estimated using
a prediction package such as PREDICT. This package offers
6
a variety of methods for estimating vapour pressure, includ-
ing the Pitzer, Riedel, and GomezThdos methods. These
methods were evaluated for TCD Ketone A, and the Pitzer
method was chosen as the most representative, although it
is interesting to note that there was relatively little variation
between the data generated by each model. It should be stated
here that these figures are only approximations and that their
derivation required the estimation of other parameters such
as the critical temperature and pressure (Table 2). Therefore
they should be treated with caution.
2
pounds published in the literature and also the computer-
3
based predictive package, CHETAH (this package takes the
approach of using Benson groups4 to estimate the thermo-
,5
dynamic properties of a molecule). The heats of reaction
-
1
derived were considered, and values of +60 kJ mol for
-1
the dehydrogenation and -120 kJ mol for the hydrogena-
tion were taken as being representative. This means that a
-
1
value of -60 kJ mol was used for the heat of reaction of
the combined hydrogen-transfer reaction.
These estimated heats of reaction can be used to predict
the maximum temperature the reaction mass will achieve in
The data generated by the Pitzer model is presented in
Table 1 for selected temperatures. It can be seen that in the
region of 310 °C, the maximum temperature attainable for
the combined hydrogen-transfer hypothesis, the vapour
pressure of the system is approximately 11.6 bar. It was only
possible to predict the vapour pressure at temperatures up
(2) Pedley, J. B. Thermochemical Data and Structures of Organic Compounds;
Thermodynamics Research Centre: Texas, 1994; Vol. 1.
(
3) CHETAH, ver 4.4, The ASTM Chemical Thermodynamic and Energy
Release EValuation Program; American Society for Testing and Materi-
als: West Conshohocken, PA, 1990.
(
(
4) Benson, S. W.; Buss, J. H. J. Chem. Phys. 1958, 29, 546.
5) Benson, S. W.; Cruickshank, F. R.; Golden, D. M.; Haugen, G. R.; O’Neal,
H. E.; Rodgers, A. S.; Shaw, R.; Walsh, R. Chem. ReV. 1969, 69, 279.
(6) PREDICT, Prediction of Thermodynamics and Transport Properties;
Dragon Technology Inc., P.O. Box 16012, Golden, CO 80402, U.S.A.
306
•
Vol. 6, No. 3, 2002 / Organic Process Research & Development

to 380 °C, but at this temperature the predicted vapour
pressure of the system is already in excess of 28 bar.
These data lead us to favour the combined hydrogen-
transfer hypothesis, although it is not proved conclusively.
dehydrogenate, supplying hydrogen to hydrogenate a
second group).
This incident also serves to reiterate the importance of
considering reaction hazards whenever changes to a process
(or the scales at which a process is operated) are made.
Conclusions
Nowadays, one would assess the safety of a reaction such
as this using various techniques, including differential
scanning calorimetry, reaction calorimetry, and adiabatic
calorimetry. However, it would not be plausible to study the
effect of each possible contaminant on the reaction; therefore,
a theoretical consideration of internal hydrogen transfer
remains important.
Whilst we have been unable to unequivocally explain this
incident, we feel that there are important learning points
which should be shared. This specific process was modified
and carried out in a manner that precluded internal hydrogen
transfer. Whilst hydrogenations are generally run in a
semi-batch manner, controlled to a large degree by the
addition rate of hydrogen, this cannot be guaranteed
where the possibility exists for hydrogen transfer within
the molecule, thus rendering a batch reaction possible
Acknowledgment
We acknowledge Will K o¨ nst, Markus Buter, and Clive
Brinklow for useful discussions, both during the investigation
of the incident and the preparation of this manuscript.
(with the dangers inherent to this mode of operation).
Unsaturated alcohols are good examples of molecules
where such a possibility exists. Our pre-scale-up reaction
hazard assessment protocol now considers this point
specifically (from a theoretical viewpoint, by looking for
the presence of a group within a molecule that may
Received for review November 3, 2000.
OP000108Z
Vol. 6, No. 3, 2002 / Organic Process Research & Development
•
307