Safety evaluation of an unexpected incident with a nitro compound

Article abstract of DOI:10.1021/op700128r

An unexpected highly exothermic reaction followed by explosion has been studied by means of reaction calorimetry and adiabatic calorimetry. The investigation of the process shows that the incident took place by a local adiabatic temperature rise in the reactor vessel due to the presence of potassium carbonate, which activated the decomposition of a nitro compound with an unusual low temperature onset The study is completed with an unproved and safer procedure consisting of changing the order of chemicals addition as an alternative approach.

Full text of DOI:10.1021/op700128r

Organic Process Research & Development 2007, 11, 1131–1134
Communications to the Editor
Safety Evaluation of an Unexpected Incident with a Nitro Compound
Rafael Barbas,‡ Manuel Botija,† Helena Camps,† Anna Portell,‡ Rafel Prohens,*,‡ and Cristina Puigjaner*,‡
Moehs Ibérica S.A., César Martinell i Brunet 12, 08191 Rubí, Barcelona, Spain, and Plataforma de Polimorfisme i
Calorimetria, SerVeis Cientificotècnics, UniVersitat de Barcelona, Josep Samitier 1-5, 08028 Barcelona, Spain
Abstract:
An unexpected highly exothermic reaction followed by explosion
has been studied by means of reaction calorimetry and adiabatic
calorimetry. The investigation of the process shows that the
incident took place by a local adiabatic temperature rise in the
reactor vessel due to the presence of potassium carbonate, which
Figure 1. Synthesis of 2,3-dimethyl-4-(2,2,2,-trifluoroethoxy)
pyridine N-oxide (2).
activated the decomposition of a nitro compound with an unusual
low temperature onset. The study is completed with an improved
and safer procedure consisting of changing the order of chemicals
crucible was used, and a heating rate of 10 °C/min was applied
addition as an alternative approach.
in the DSC experiment. Compound 1 melts at 101 °C and shows
a high severe decomposition beginning at 237 °C, with an
associated heat of 1797 J/g, according to the results obtained
directly from the DSC analysis (Figure 3). An adiabatic
temperature rise of 998 °C has been calculated according to
the equation ∆Tadiabatic ) ∆H/Cp (a heat capacity of 1.8 J/g °C
Introduction
In the past, Moehs Ibérica S.A. has produced Lansoprazol
(a proton pump inhibitor indicated for gastric ulcer treatment)
by a multistep synthesis from 1. One of the intermediates is
2,3-dimethyl-4-(2,2,2-trifluoroethoxy)pyridine N-oxide (2), ob-
has been estimated).
This result suggested the possibility that the other chemicals
tained by condensation of 2,3-dimethyl-4-nitropyridine N-oxide
involved in the reaction mixture might affect the decomposition
(1) with trifluoroethanol in MEK/water using triethylbenzy-
onset of 1, since this value (237 °C) was too high to allow for
lammonium chloride as a phase transfer catalyst (Figure 1).
the quick runaway observed. It has been described that pure
During the production of one batch, an incident occurred
organic nitro compounds decompose at high temperatures,
unexpectedly. The incident happened as follows. The vessel,
exhibiting large exotherms, the reaction being violent or
at room temperature, was first charged with 450 kg of 2,
explosive in most cases. Moreover, reactants, solvents, and
3-dimethyl-4-nitropyridine N-oxide (1), then 720 kg of potas-
impurities may affect organic nitro compounds’ thermal stabil-
ity, lowering their temperature of decomposition.2
sium carbonate and 15 kg of triethylbenzylammonium chloride
(TEBA) were added without stirring, and finally 15 L of water
was added. The incident started 10 min after the vessel was
closed for alternate vacuum/nitrogen purge. A high increase of
the internal temperature associated with fast gas evolution was
detected (24 h after the incident, the temperature was 330 °C),
generating a sudden pressure rise that ended up with the
explosive breakage of the reactor (Figure 2). In view of the
unexpected incident, an investigation was set up to clarify
the origin of the severe system thermal runaway.
Therefore, we investigated the thermal stability of 1 in the
presence of the other chemicals. Since DSC analysis is known
to give results with a relative uncertainty regarding the
decomposition onset as a result of the small sample size and
high heating rates,3 it was important to carry out screening
experiments on a larger sample size. This was done in an
accelerating rate calorimeter (ARC), which offers the possibility
to prepare reaction mixtures, gives a more accurate determi-
nation of the onset temperature of the decomposition and makes
it possible to determine the pressure rise during decomposition,
which finally is what provokes the reactor breakage and
explosion. It was observed that the mixture of 1 with triethyl-
benzylammonium chloride and potassium carbonate, analyzed
by ARC, showed an slow exotherm from 72 °C, which became
extremely fast at 89 °C with an adiabatic temperature rise higher
Results and Discussion
First, we evaluated the reactivity of 1 by differential scanning
calorimetry to determine its explosive potential.1 A sealed
* To whom correspondence should be addressed. Telephone: +34 93 4034656.
E-mail: cris@sct.ub.es and rafel@sct.ub.es.
† Moehs Ibérica S.A.
‡ Plataforma de Polimorfisme i Calorimetria, Serveis Cientificotècnics,
Universitat de Barcelona.
(2) Gustin, J-L. Org. Process Res. DeV. 1998, 2, 27–33.
(3) Steinbach, J. Safety Assessment for Chemical Processes, Wiley-VCH:
Weinheim, 1999; pp 29–26.
(4) Townsend, I. J. Therm. Anal. 1991, 37, 2031–2066.
(1) King, R.; Hirst, R. Safety in the Process Industries, 2nd ed.; Elsevier:
Oxford, 2002.
10.1021/op700128r CCC: $37.00
Published on Web 10/09/2007
2007 American Chemical Society
Vol. 11, No. 6, 2007 / Organic Process Research & Development
•
1131

Figure 4. Temperature and pressure as a function of time of a
mixture of 1 with triethylbenzylammonium chloride and potas-
sium carbonate, analyzed by ARC.
ate in water generated enough heat to locally increase the
temperature of the reaction mass and to trigger the decomposi-
tion of the nitro compound 1, generating the observed thermal
runaway.
The synthesis under study has been repeated many times at
large scale in industry, but the incident occurred only once. So
this suggests that the heat transfer is not usually easy, but
depending on the particular conditions of charging (i.e., the way
solids were deposited), it could occur.
Figure 2. View of the facilities after the incident.
To eliminate the potential secondary reaction of the mixture
K2CO3/H2O/TEBA as the cause of the incident, we performed
an adiabatic calorimetry experiment (data not shown) in the
same molar rates as in the original methodology. We did not
detect any exotherm neither a significant pressure increase.
We decided to develop a safer process, and an alternative
method that avoided the initial heat evolution was designed.
The process was divided in two steps: in a first step, the reaction
mass was prepared by adding successively at room temperature
all of the reagents except 1 into the reaction vessel and letting
the system release all of the evolved heat. In a second step, the
nitro compound 1 was added and then heated up to 85 °C. To
check the safety of this procedure, reaction calorimetry and
adiabatic calorimetry experiments5 were performed with the
following results.
Figure 3. DSC curve of 1 carried out at a heating rate of 10
°C/min.
Reaction Calorimetry (RC). The heat flow analysis6
applied to this dynamic process shows an exothermic event
during heating up to 85 °C (Figure 5). The obtained data are
shown in Table 1.
than 364 °C and an internal pressure rise higher than 21 bar
(Figure 4). So, a dramatic lowering of the decomposition onset
was observed.
The adiabatic temperature rise is calculated to be 18 °C
according to the equation ∆Tadiabatic ) ∆H/mCp while the MTSR
is 103 °C in the worst case (total accumulation). This moderate
value demonstrates that safe conditions can be obtained if 1 is
added to the reaction mixture maintained at room temperature
once the previous exothermic (mixture and dissolution) events
have ceased.
Adiabatic Calorimetry (ARC). To define the scenario of
criticality, an ARC experiment of the reaction mass was
performed in the following manner. Potassium carbonate and
water were introduced into the calorimetric bomb, and a
Since the onset temperature of the decomposition observed
by ARC is above the operating temperature of the loading
process (room temperature), we decided to study the cause of
the thermal runaway activation. Our hypothesis was that an
initial heating of the reaction mass activated the severe
decomposition of compound 1. The order of addition of
reactants suggested the possibility of a local temperature increase
due to the dissolution of potassium carbonate in water. So we
tried to determine the temperature rise provoked by this
phenomenon of dissolution in a 1-L automatic laboratory reactor
in the absence of any other reactants. A local temperature rise
from 22.3 to 81.5 °C was observed by a temperature sensor
located on the carbonate surface when water was added. These
data show that the exothermic dissolution of potassium carbon-
(5) (a) Landau, R. N.; Blackmond, D. G. Chem. Eng. Prog. 1994, 90 (11),
43–48. (b) Landau, R. N. Thermochim. Acta 1996, 289, 101–126.
(6) Coates, C. F. Thermochima Acta 1985, 85, 369–372.
1132
•
Vol. 11, No. 6, 2007 / Organic Process Research & Development

Figure 5. Heat power (Qr) and reaction temperature (Tr) as a function of time associated with heating up to 85 °C.
Table 1. Results of reaction calorimetry investigation
heat of reaction (kJ/kg)
35.4
7
18
91
103
2.0
maximum heat flow (W/kg)
∆Tad (K)
heat accumulation when reaching 85 °C (%)
maximum temp of the synthesis reaction MTSR (°C)
heat capacity of reaction mass (kJ/kg K)
temperature increase was observed. Once the temperature
reached room temperature again, triethylbenzylammonium
chloride, MEK, 2,2,2-trifluoroethanol and 2,3-dimethyl-4-ni-
tropiridine N-oxide were added without stirring. The bomb was
introduced into the calorimeter, and the “heat-wait-seek” method
was applied.6 A high severity exothermic phenomenon starting
from 173 °C was detected, with a maximum self-heating rate
of 4.8 °C/min and an adiabatic temperature rise higher than
200 °C (Figure 6). A value for the heat of decomposition higher
than 360 J/g was calculated according to the equation ∆H )
∆Tadiabàtic · Cp (estimated value for Cp ) 1.8 kJ/kg K). The
pressure rise due to decomposition was higher than 86 bar, and
the time to maximum rate at the final reaction temperature of
85 °C was 1000 h, while at the maximum temperature of the
desired reaction (MTSR ) 103 °C) it was 300 h.
Finally, to classify the criticality of the process, the well
established methodology by Stoessel can be used.7 We must
consider the relative position of four temperature levels: process
temperature (85 °C), MTSR (103 °C), temperature at which
TMRad is 24 h (145 °C) and boiling point (85 °C). This allows
us to classify our process in a risk class 3 (Figure 7), that is,
after loss of control of the synthesis reaction, the boiling point
will be reached but the decomposition reaction cannot be
triggered. The safety of the process depends on the heat release
rate of the synthesis reaction at the boiling point.
Figure 6. Temperature and pressure as a function of time of a
mixture of potassium carbonate, water, triethylbenzylammo-
nium chloride, MEK, 2,2,2-trifluoroethanol and 1 analyzed by
ARC.
Conclusions
Our investigations have demonstrated that potassium carbon-
ate accelerates the decomposition of 1, increasing the probability
of runaway of the reaction mass in the original methodology.
The incident occurred because a local temperature increase took
place during water addition, leading the reaction mass to heat
up to a runaway activation temperature followed by explosion.
An improved process changing the order of addition of the
chemicals as an alternative approach has been safely designed,
evaluated and successfully scaled-up.
Experimental Section
Equipment Used. Differential scanning calorimetry equip-
ment: Mettler-Toledo DSC-30 differential scanning calorimeter;
high pressure crucibles with a capacity of 30 µL. Automatic
laboratory reactor: Mettler-Toledo LabMax equipped with a
temperature sensor and with a glass anchor stirrer. Reaction
(7) (a) Stoessel, F. Chem. Eng. Prog. 1993, 89 (10), 68–75. (b) Stoessel,
F.; Fierz, H.; Lerena, P.; Killé, G. Org. Process Res. DeV. 1997, 1,
428–434.
Vol. 11, No. 6, 2007 / Organic Process Research & Development
•
1133

Hazard Technology; ARCTC-HC-MCQ (Hastelloy) test cells
were used; sensitivity threshold was set to 0.02 °C/min.
General Procedure for RC-1 Experiment. In the reaction
vessel at 20 °C were charged potassium carbonate (128.0 g),
water (73.7 g), MEK (124.5 g), TEBA (2.6 g) and 2,2,2-
trifluoroethanol (157.4 g). The reaction mass was stirred at 150
rpm. Then, at 20 °C, 80.0 g of 1 were added slowly. Finally,
the reaction mass was pressurized with nitrogen (1.5 bar), and
it was heated up to 85 °C and stirred for 10 h.
General Procedure for ARC Experiments. Original Meth-
odology. 2,3-Dimethyl-4-nitropiridine N-oxide (2.21 g), TEBA
(0.08 g) and potassium carbonate (3.56 g) were introduced into
the calorimetric bomb at room temperature, without stirring.
The bomb was introduced into the calorimeter, and the “heat-
wait-seek” method was applied. ImproVed Methodology. Potas-
sium carbonate (1.37 g) and water (0.72 mL) were introduced
into the calorimetric bomb. Once the temperature reached room
temperature again, TEBA (27.2 mg), MEK (1.63 mL), 2,2,2-
trifluoroethanol (1.12 mL) and 2,3-dimethyl-4-nitropiridine
N-oxide (0.78 g) were added successively without stirring. The
bomb was introduced into the calorimeter, and the “heat-wait-
seek” method was applied.
Figure 7. Evaluation of the thermal risk of the improved
7
process, according to the criticality classes of scenarios.
calorimetry equipment: Mettler-Toledo RC-1 reaction calori-
meter was equipped with a 1-L medium pressure reactor (MP10)
and with a hastelloy anchor stirrer. Adiabatic calorimetry
equipment: accelerating rate calorimeter was from Thermal
Received for review June 8, 2007.
OP700128R
1134
•
Vol. 11, No. 6, 2007 / Organic Process Research & Development