A calorimetric investigation to safely scale-up a curtius rearrangement of acryloyl azide

Article abstract of DOI:10.1021/op970115w

N-Vinyl-O-benzyl urethane was prepared for use as a starting material in a multistep synthesis of a drug candidate. After unsuccessful attempts to employ non-Curtius options to prepare this intermediate, we decided to assess the thermodynamic and kinetic parameters of the acyl azide Curtius rearrangement and trapping of the intermediate isocyanate. (Prepared in analogy to the literature method of Hegedus et al.: Wieber, G. M.; Hegedus, L. S.; Akermark, B.; Michalson, E. T. J. Org. Chem. 1989, 54, 4649-4653.) Reaction calorimetry was utilized to show that the rate of the Curtius rearrangement was completely dose-rate controlled, and that under carefully defined conditions this reaction could be safely performed on limited scale. In addition to addressing the thermal hazard issues, profiling the reaction with reaction calorimetry and interfaced technologies resulted in significant process improvements.

Full text of DOI:10.1021/op970115w

Organic Process Research & Development 1998, 2, 382−392
A Calorimetric Investigation To Safely Scale-Up a Curtius Rearrangement of
Acryloyl Azide
David J. am Ende,* Keith M. DeVries, Pamela J. Clifford, and Steven J. Brenek
Central Research DiVision, Process Research & DeVelopment, Pfizer Inc., Eastern Point Road,
Groton, Connecticut 06340
Scheme 1
Abstract:
N-Vinyl-O-benzyl urethane was prepared for use as a starting
material in a multistep synthesis of a drug candidate. After
unsuccessful attempts to employ non-Curtius options to prepare
this intermediate, we decided to assess the thermodynamic and
kinetic parameters of the acyl azide Curtius rearrangement and
trapping of the intermediate isocyanate. (Prepared in analogy
to the literature method of Hegedus et al.: Wieber, G. M.;
Hegedus, L. S.; Akermark, B.; Michalson, E. T. J. Org. Chem.
1989, 54, 4649-4653.) Reaction calorimetry was utilized to
show that the rate of the Curtius rearrangement was completely
dose-rate controlled, and that under carefully defined conditions
this reaction could be safely performed on limited scale. In
addition to addressing the thermal hazard issues, profiling the
reaction with reaction calorimetry and interfaced technologies
resulted in significant process improvements.
Scheme 2
Introduction
N-Vinyl urethanes and amides, as exemplified by N-vinyl-
O-benzyl urethane (1), are very reactive intermediates, both
being electrophilic at the R carbon1 and nucleophilic at the
â center.2 This equilibrium (Scheme 1) lies almost exclu-
sively toward 1. The N-vinyl compounds are often masked
as the N-acyl-N,O-acetal 2 or the corresponding biscarbamate
3. The N-acyl-N,O-acetals represented by 2 have been
prepared by electrochemical oxidation of the corresponding
amides or urethanes3 as well as by oxidative decarboxylation
of the corresponding N-protected amino acids.4 The biscar-
bamates 3 are prepared by acid-catalyzed exchange of the
corresponding primary carbamate with an aldehyde.5 Due
to cost and safety concerns, we initially attempted to prepare
1 by acid-catalyzed exchange between acetaldehyde diethyl
* To whom correspondence should be addressed: Pfizer, Inc., Process R&D,
Bldg. 156, Box 5, Eastern Point Rd., Groton, CT 06340. Phone 860-441-6504.
Fax 860-441-5445. E-mail david_j_amende@groton.pfizer.com.
(1) For example, see: (a) Weinreb, S. M. Acc. Chem. Res. 1985, 18, 16-21.
(b) Weinreb, S. M. Heterocycles 1979, 12, 949-975. (c) Krow, G.;
Rodebaugh, R.; Carmosin, R.; Figures, W.; Pannella, H.; DeVicaris, G.;
Grippi, M. J. Am. Chem. Soc. 1973, 95, 5273-5280. (d) Shono, T.; Kise,
N.; Ohi, S.; Tsubata, K. Tetrahedron Lett. 1988, 29, 231-234. (e) Harding,
K. E.; Coleman, M. T.; Liu, L. T. Tetrahedron Lett. 1991, 32, 3795-3798.
(2) For example, see: (a) Shono, T.; Matsumura, Y.; Tsubata, K.; Sugihara,
Y.; Yamane, S.; Kanazawa, T.; Aoki, T. J. Am. Chem. Soc. 1982, 104,
6697-6703. (b) Katrizky, A. R.; Rachwal, B.; Rachwal, S. J. Org. Chem.
1995, 60, 3993-4001.
ketal and O-benzyl urethane followed by loss of ethanol to
provide the N-vinyl derivative (Scheme 2). Although we
were able to obtain small amounts of 2, the major product
which was the result of further disproportionation was the
biscarbamate 3. In Scheme 2, DCE and TsOH refer to
dichloroethane and para-toluene sulphonic acid, respectively.
All attempts to “crack” 2 or 3 with heat or base to prepare
pure 1 were unsuccessful. The stability of the N-vinyl
derivatives increases with the degree of substitution on the
nitrogen as well as the degree of branching on the nitrogen
substituents. Thus, 1 is quite unstable due to its lack of
substitution. In addition, attempts to use either 2 or 3 to
(3) See refs 1d and 2a and references therein.
(4) Harding, K. E.; Liu, L. T.; Farrar, D. G. Synth. Commun. 1991, 21, 1409-
1417.
(5) For example, see: (a) Giger, R.; Ben-Ishai, D. Isr. J. Chem. 1967, 5, 253-
259. (b) Zaugg, H. E.; Martin, W. B. Org. React. 1965, 14, 52.
382
•
Vol. 2, No. 6, 1998 / Organic Process Research & Development
10.1021/op970115w CCC: $15.00 © 1998 American Chemical Society and Royal Society of Chemistry
Published on Web 09/25/1998

Scheme 3
generate 1 in situ for the subsequent reaction gave inferior
results.
The hazard associated with this reaction is due to the
thermally unstable azide solution. The acryloyl azide has a
heat of formation of +230 J/mol (calculated from CHETAH)
and, as will be discussed later, can decompose at low
temperatures to liberate an equivalent of nitrogen gas and
vinyl isocyanate. Positive heats of formation provide an
indication for potential thermal instability. Further, a
potential pressure hazard exists when handling the azide due
to its propensity to release gas and heat. The thermal stability
issue was minimized by preparing and storing the azide
solution at 0 °C and then using it as a stock solution in
toluene in the Curtius chemistry. As a safety precaution,
this azide should not be isolated from solution, as it could
be dangerously explosiVe! DSC, Radex, and ARC tests were
performed to characterize the acryloyl azide-toluene solu-
tion. The results describing the onset temperature, decom-
positions energy, and gas evolution regarding the thermal
stability of the azide are discussed in the Thermal Analysis
section. The azide-toluene solution is reacted with benzyl
alcohol (BnOH) in the Curtius rearrangement to produce 1,
the desired urethane. Three reaction calorimetry (RC-1)
experiments were performed for the Curtius rearrangement
which involved the addition of acrylolyl azide to benzyl
alcohol shown in eq 2 of Scheme 3. The selectivity results
from each run are shown at the bottom of Scheme 3. For
the RC-1 runs, we prepared the acryloyl azide from 100 mL
of acryloyl chloride as the stoichiometric limiting reagent
The preferred method in the literature for preparing
compound 1 was that reported by Hegedus et al.6 Curtius
chemistry involves potentially serious reaction hazards
associated with the use of sodium azide, the formation of
an acryloyl azide, and the isocyanate intermediate. The
following discussion describes our assessment of the Curtius
rearrangement (Scheme 3, step 2) using the Mettler RC-1
reaction calorimeter with interfaced technologies. Thermo-
dynamic and kinetic data obtained from this assessment
permitted us to safely scale-up the reaction.
Discussion
Chemistry. The initial step (eq 1, Scheme 3) was
formation of the acryloyl azide from acryloyl chloride, using
a slight modification of the literature procedure. To deter-
mine the point of reaction completion, a pilot reaction was
carried out in toluene-d8 and D2O. Aliquots of the reaction
mixture, sampled by 1H NMR, showed that the reaction was
complete in 2 h. After workup,7 the toluene solution was
shown to be stable for days with storage at 5 °C.
(6) Wieber, G. M.; Hegedus, L. S.; Akermark, B.; Michalson, E. T. J. Org.
Chem. 1989, 54, 4649-4653.
(7) In analogy to the literature preparation, we did one sodium carbonate wash
followed by four water washes. We did not do the silver nitrate test since
the chloride ion in the tap water complicated this assay.
Vol. 2, No. 6, 1998 / Organic Process Research & Development
•
383

according to eq 1 of Scheme 3.
an issue. With an addition time of 60 min12 and no hold
time, the selectivity improved significantly, with a ratio of
∼15:1 of the desired urethane relative to the undesired aminal
(i.e., 1:4 in Scheme 3, run 3).13
The Curtius rearrangement is shown in eq 2 of Scheme 3
which involves the formation of the vinyl isocyanate from
the acryloyl azide and in situ trapping with benzyl alcohol
using a small amount of pyridine and hydroquinone. The
literature procedure called for adding the acryloyl azide
solution to 105 °C benzyl alcohol. On larger scale, we saw
varying amounts of the aminal side product 4 (eq 3, Scheme
3). The amount of aminal byproduct formed decreased with
increasing number of aqueous washes of the acryloyl azide
starting material. However, in place of numerous water
washes, we found that, by adding finely ground potassium
carbonate to the reaction mixture, this side product was
greatly minimized, as shown by the results of runs 1 and 2.
The formation of the undesired O-benzyl aminal (4) is an
acid-catalyzed process, and the potassium carbonate served
to neutralize any acid which was present.8
Initially it was assumed that using a stoichiometric amount
of benzyl alcohol (1.0 equiv) would minimize the amount
of undesired 4 that was formed. Analysis of the headspace
by GC/MS, however, showed that, toward the end of the
azide addition of run 1, the rate of vinyl isocyanate being
liberated as a gas increased substantially (see off-gas rate
and composition section).9,10 This observation is consistent
with a change in selectivity from successful in situ trapping
of the isocyanate by the benzyl alcohol during the early
stages of the addition to unsuccessful trapping and release
of the vinyl isocyanate toward the end of the addition.
Evidently, toward the end of the addition, there was not
enough benzyl alcohol remaining for trapping of the isocy-
anate, as it had been consumed to produced the undesired
aminal, which consumes 2 equiv of the benzyl alcohol. The
addition of potassium carbonate allowed us to increase the
amount of benzyl alcohol to 2.0 equiv without increasing
the formation of 4. This greatly reduced the amount of vinyl
isocyanate being lost in the headspace.11
The RC1 showed that the desired reaction was completely
dose-rate controlled, with the gas evolution and heat libera-
tion ending immediately upon the completion of the dose.
This and the FTIR results (see FTIR results section) assured
us that no acyl azide accumulated over the course of the
addition and that by cooling the reaction immediately after
the azide was added, the formation of 4 was minimized.
Calorimetric Principle. In general, heat flow calorimetry
is accomplished by way of an energy balance around the
reaction calorimeter. This enables the determination of the
net heat flow produced from the reactor as a result of a
combination of reagents or other thermal events. Math-
ematically, the energy balance around the reactor is described
by eq 1 for a semibatch reaction under nonrefluxing
conditions:
dTr dm
qr(t) ) UA(Tr - Tj) + mrCp
+
doseCpdose(Tr - Tdose) (1)
r dt
dt
where q(t) is measured in watts. Through controlled
temperature ramping and the use of a calibration heater, the
reaction calorimeter returns a heat capacity (Cpr) and heat
transfer coefficient (U) for the initial and final reactor
contents.14 During a semibatch reaction, the energy balance
must also account for the sensible heat of the feed, which is
the last term of eq 1. This term is zero for a purely batch
reaction or when Tr ) Tdose
.
To obtain the total heat liberated for a reaction or thermal
event occurring over an interval of time, the integral of eq
1 must be computed over the reaction limits according to
the following equation:
tf
The final parameter which we observed to reduce the
amount of 4 formed was the reaction hold time. Even in
the presence of potassium carbonate, if the azide addition
and subsequent hold times were long, formation of 4 became
∆Htotal
)
q (t) dt
(2)
∫
r
0
where∆Htotal is the net heat liberated over the time interval
between zero and some final time, tf.
The adiabatic temperature rise calculation provides an
estimate for the final temperature that could be reached if
the exothermic energy went to raising the temperature of
the desired reaction. This calculation obviously assumes that
no undesired reactions are encountered along the way. There
are different levels of sophistication for calculating adiabatic
temperature rise. For example, the temporal dependencies
of heat capacity, Cp, and reaction mass, mrxn, with extent of
reaction can be considered in its description.15 A simpler
(8) The original literature report has pyridine in it, presumably to neutralize
any acid which may be generated. Pydidine‚HCl, however, can also serve
as an acid catalyst.
(9) Vinyl isocyanate has been reported in the literature to boil at 38.5 °C. We
strongly recommend the use of a scrubber. See: Konig, K.-H.: Reitel, C.;
Mangold, D.; Feuerhead, K.-H.; Oeser, H.-G. AIChE J. 1979, 18, 319-
320.
(10) We also saw varying amounts of the bis-carbamate shown below as a solid,
which precipitated from the reaction mixture. This is consistent with no
benzyl alcohol being present at the end of the addition due to the formation
of 4, and the isocyanate being lost as a gas or being trapped by the
hydroquinone stabilizer. 1H NMR (400 MHz, CDCl3 + CD3OD): δ 7.15
(s, 4H), 6.55 (dd, J ) 15.9 and 8.9 Hz, 2H), 4.63 (d, J ) 15.9 Hz, 2H),
4.28 (d, J ) 8.9 Hz, 2H).
(12) Caution: this reaction is very exothermic. Very efficient cooling and internal
temperature control are necessary for this rate of addition.
(13) 1H NMR ratio.
(14) Specifically, during the temperature ramp, the heat flow, qr(t), is zero, while
the temperature rise, dTr/dt, remains a constant. During the action of the
precision heater, qr(t) is known while dTr/dt is zero. As a result, two
equations in two unknowns are solved simultaneously to solve for both the
heat transfer coefficent, U, and the heat capacity of the reactor contents,
Cpr. This sequence is usually performed before and after the thermal events
so that a linear interpolation can be calculated throughout the course of the
reaction.
(11) The other three gases observed by GC/MS were nitrogen, carbon dioxide,
and acetaldehyde. The carbon dioxide is a result of the potassium carbonate
used, and the N-vinyl portion of compound 1 is a latent source of
acetaldehyde.
(15) Landau, R. N.; Cutro, R. S. Chem. Eng. Prog. 1993, 66-71.
384
•
Vol. 2, No. 6, 1998 / Organic Process Research & Development

Table 1. Comparison of thermal data obtained from selected instrumentation for the decomposition of the acryloyl azide
solution (the solution was 36% acryloyl azide in toluene)
moles of gas per
test
∆H (J/g of soln)
∆H (J/g of azide)
onset temp (°C)
mole of azide
TMR at onset (h)
11
DSCa (4 °C/min)
-731
-282
-482
-
-2031 (-197 kJ/mol)
-783 (-76 kJ/mol)
-1340 (-130 kJ/mol)
-
40
190
ARC (Φ ) 1.74)
Radex
27 (∆Tad ) 157)
50-60
0.7
1
CHETAHb
-776
-2144 (-208 kJ/mol)
a Two exotherms were observed; see Figure 1. b Gas-phase heats of formation values were used to estimate the heat of reaction. No Benson group for the carbonyl
in OdCdN was available but was approximated as CdO-(2N).
Figure 1. DSC scan of a 36% solution of acryloyl azide in toluene. The initial decomposition exotherm (731 J/g) is presumably due
to the formation of nitrogen and vinyl isocyanate. The sample size was 8.2 mg, with a scan rate of 4 °C/min using gold-plated
stainless steel crucible. Onset of the first exotherm was shown to be less than 50 °C.
form is the one shown in eq 3:
min, the energy associated with the acryloyl azide decom-
position was measured to be -731 J/g of solution (or -2031
J/g of azide ) -197 kJ/mol of azide), with an onset
occurring in the neighborhood of 40 °C (see Figure 1). A
secondary exotherm was observed from the DSC at higher
temperatures but was of lower energy (-282 J/g of solu-
tion).18
Analysis of the Radex data (Figure 2) revealed that the
observed pressure rise corresponded closely to the volume
of nitrogen that was expected from the stoichiometry.19 The
number of moles of gas calculated from the pressure rise
∆Hrxnn
Tadiabatic ) Trxn
+
(3)
Cp(mrxn
)
where ∆Hrxn and n have units of kilojoules per mole and
moles, respectively, and Cp is the heat capacity of the
reaction mass. The adiabatic temperature rise is useful in
evaluating whether potential decomposition reactions could
be reached during a thermal runaway.
Thermal Analysis. Caution: In general, carbamoyl
azides are explosiVe compounds. InVestigators considering
work with these compounds should consult the literature and
take proper precautions.16,17
Thermal analysis was performed for both the acryloyl
azide solution (in toluene) and the isolated reaction products
1 and 4 to examine their thermal stability. The thermal
analysis data for the azide solution are listed in Table 1.
Based on the DSC measurement using a scan rate of 4 °C/
(16) Bretherick’s Reactive Chemical Hazards Database, Butterworth-Heinemann
Ltd., 1996, warns that azides of low molecular weight (more than 25%
nitrogen) should not be isolated from solution, as the concentrated material
is likely to be dangerously explosive and the concentration of such solutions
should be kept to <10%. Failure to ensure that solutions are free of residual
acyl azide before distillation or heat treatment has led to a violent explosion.
(17) See also: Lieber, E.; Minnus, R.; Rao, C. Chem. ReV. 1965, 65, 377-384.
(18) DSCs were obtained using sealed gold-plated stainless steel crucibles rated
to 2000 psi.
Vol. 2, No. 6, 1998 / Organic Process Research & Development
•
385

Figure 2. Pressure profile obtained from Radex experiment of a 36% solution of acryloyl azide in toluene. The rise in pressure
closely corresponds to that predicted for 1 stoichiometric equiv of nitrogen liberated. The vapor pressure of toluene over the same
temperature range is shown for reference. The sample size was 0.660 g of solution with a scan rate of 2 °C/min.
was equivalent to 1 mol of nitrogen per mole of azide. These
data suggest that the stoichiometry of the initial exotherm
shown in Figure 1 represented the formation of the iso-
cyanate with the liberation of nitrogen. As a check of the
hypothesis of the first exotherm, CHETAH20 was used to
estimate the heat of reaction for acryloyl azide decomposing
to 1 mol of nitrogen and 1 mol of vinyl isocyanate. In this
exercise, CHETAH returned a gas-phase heat of reaction (of
decomposition) of -2144 J/g of azide (-208 kJ/mol), while
the DSC measured an energy of -2031 J/g of azide (-197
kJ/mol); thus, very good agreement was obtained.21 The
second exotherm shown in the DSC scan of Figure 1 was
presumed to be that of the decomposition of vinyl iso-
cyanate.
Accelerating rate calorimetry (ARC), which is an adiabatic
experiment, was performed to establish a more sensitive self-
heat temperature for the azide solution. The ARC returned
a self-heat (onset) temperature of 27 °C when corrected for
the Φ factor (in this case, Φ was 1.74).22 The self-heat data
are shown in Figure 3 which exhibit rates exceeding 100
°C/min. The Φ-corrected onset temperature is obtained by
extrapolation of the self-heat rate curve until it intersects with
Figure 3. Accelerating rate calorimetry (ARC), self-heat rate
vs temperature plot for a 36% acryloyl azide in toluene solution.
Data were obtained in a 6.1 g thin-walled (0.02 in.) titatium
bomb loaded with 2.48 g of sample, resulting in a Φ of 1.74
with a heat step of 5 °C and sensitivity of 0.02 °C/min. The
Φ-adjusted slope sensitivity was 0.0115 °C/min. As shown
above, the extrapolated self-heat curve results in an onset for
self-heating at 27.2 °C.
(19) 0.26 g (0.0027 mol) of acyl azide was present in the toluene with an
approximate headspace of 5 mL. The stoichiometry was 1 mol of N2 per
mole of azide. At 80 °C, this corresponded to a change in pressure of about
230 psi above the vapor pressure of toluene at 80 °C in Figure 2.
(20) CHETAH version 7.0 is the ASTM computer program for chemical
thermodynamic and energy release evaluation (NIST special database 16).
(21) Hf(gas) of the acryloyl azide was computed with CHETAH to be +55.1 kcal/
mol. The ∆Hf(gas) for the vinyl isocyanate was computed to be +5.4 kcal/
mol. ∆Hf(gas) for N2 is zero. (+5.4 - 55.1) kcal/mol × 4.18 J/cal × (mol/
97 g) ) -2144 J/g azide.
(22) The Φ factor is the ratio of the combined thermal capacity (mCp) of the
sample and container to that of the sample; thus, Φ of 1.0 implies that the
thermal capacity of the container is negligible relative to the sample. Φ )
1 + (mCp bomb/mCp sample).
the Φ-corrected instrument sensitivity obtained by dividing
0.02 °C/min by 1.74 (0.0115 °C/min). The temperature at
which these lines intersect is the corrected onset tem-
perature, which is shown graphically in Figure 3. Time to
maximum rate (TMR) data are shown in Figure 4 which,
when extrapolated to room temperature, 25 °C, indicate a
TMR of about 11 h. The data shown in Figure 4 have been
corrected for Φ factor. For example, under adiabatic
conditions and initially held at 25 °C, it would take
approximately 11 h for the acryloyl azide solution to
386
•
Vol. 2, No. 6, 1998 / Organic Process Research & Development

tion was calculated to be +102 kJ/mol of azide using the
initial portion of the self-heat rate data shown in Figure 3.
DSC scan of the products 1 and 4 and of the reaction
mixture upon completion of the Curtius rearrangement
showed no exotherms. This finding was important for
assessing the safety of the overall process because it confined
the safety issues to the thermal stability of the acryloyl azide
feed and to controlling the rate of conversion in the reactor.
Furthermore, in an adequately vented vessel under ambient
to subambient temperatures, the acryloyl azide solution (in
toluene) can be safely handled and transferred from a charge
tank to the reaction vessel. It was also learned by DSC tests
that decomposed azide solutions no longer pose thermal
stability or pressure hazards.
Reaction Calorimetry Results. Isothermal reaction
calorimetry (RC-1) provided the instantaneous heat flow data
measured during the Curtius rearrangement reactions. The
azide-toluene solution was dosed for 2.5 h in run 1 and for
3.0 h in run 2. The heat flow profiles for runs 1 and 2 are
shown in Figure 5. The measured exotherm from the RC-1
energy balance was -202 and -203 kJ/mol of azide for runs
1 and 2, respectively. The energy balance is as follows:
Figure 4. ARC time to maximum rate (TMR) curve attained
from the data shown in Figure 3 for 35% acrylolyl azide in
toluene. The TMR curve has been corrected for Φ. As seen
from the curve, a TMR of about 3 h at 34 °C is obtained, and,
if extrapolated, a TMR of approximately 11 h is predicted for
36% acyrloyl azide solution at 25 °C under adiabatic conditions.
Extrapolations such as this should, in general, be confirmed
with isothermal (adiabatic) experiments for confirmation as the
mechanism for decomposition may change at lower tempera-
tures.
∆Hreaction ) QRC-1 measured + Qdose + Qgas evolution + Qloss (5)
QRC-1 measured ) -202 kJ/mol of azide
Qdose ) mCp∆T ) -42 kJ/mol of azide
thermally run away to its maximum rate. The following
Qgas evolution ≈ 0
equation describes how TMR is corrected for the Φ factor:
Qlosses ≈ 0
TMR (T, °C) ) Φ(tTMR - tT)
(4)
The heat of reaction for the Curtius rearrangement is the
sum of several terms shown in eq 5. The first term, QRC-1
measured, is the heat flow measured by the RC-1, which includes
the heat removed through the jacket and any heat consumed
to raise the temperature of the reactor contents. To calculate
the heat of reaction, Qdose must be included to account for
the sensible heat of the feed. The loss of nitrogen has a
slight cooling effect as it escapes the reactor but is assumed
to be a relatively minor contribution, as it is not a latent
heat. Variant heat losses can occur from the reactor lid,
especially during nonisothermal operation. However, under
isothermal conditions, as was the case here, heat losses from
the reactor tend to have a less significant effect on the
measured heat flow. This is a result of the ability of the
calorimeter to remove the exothermic heat through the
cooling jacket while maintaining isothermal conditions.
Summing the significant contributions, the heat of reaction
was measured to be -244 kJ/mol of azide. Most of this
reaction heat can be attributed to the decomposition of the
azide to release vinyl isocyanate and nitrogen, which was
measured by DSC to be about -197 kJ/mol for the initial
exotherm. The coupling of the vinyl isocyanate with the
benzyl alcohol would presumably make up the difference,
contributing approximately -44 kJ/mol. Calculations using
CHETAH to estimate ∆Hrxn of the desired reaction gave
-353 kJ/mol for the gas-phase and -192 kJ/mol for the
liquid-phase reaction. The latter is in fairly good agreement
where (tTMR - tT) is the elapsed time between some
temperature, T, and the maximum rate, TMR. Figure 4 was
obtained by multiplying the raw TMR data by the Φ factor
of 1.74. The ARC experiment resulted in an adiabatic
temperature rise of 157 °C. When corrected for Φ and
estimating a heat capacity for the azide solution, the heat of
decomposition reaction from the ARC was calculated to be
-1340 J/g of azide (-130 kJ/mol). This value is thought
to be low because the self-heat rate exceeded 100 °C/min
and, consequently, resulted in a loss of adiabaticity in the
sample container. Thus, the true adiabatic temperature
probably was not reached.
Based on the thermal analysis, the propensity for the
acryloyl azide to liberate nitrogen, and the fact that it was
considered “plosive” 23 by CHETAH, solutions of this
material should be consumed relatively quickly after their
preparation. The acryloyl azide certainly poses a pressure
hazard, and storage containers must be adequately vented
to allow for the escape of nitrogen. The rate of nitrogen
liberation, resulting from decomposition of the azide, is
highly dependent on temperature. The rate is significantly
reduced at temperatures below 25 °C but is not entirely
suppressed, thus the need for adequate venting. Based on
the ARC data, the activation energy for the azide decomposi-
(23) A “plosive” determination implies that the material is expected to decompose
with violence.
Vol. 2, No. 6, 1998 / Organic Process Research & Development
•
387

flow and off-gas flow rate immediately subsided. Thus, if
there were a cooling failure, simply stopping the feed would
stop the exothermic event because there is no significant
accumulation of reagent. Dose-rate control refers to the
reaction rate being controlled by the rate of addition. For
this reason, rate constants are not reported here.
Heat flow profiles for dose-rate controlled reactions tend
to have square wave heat flow profiles. As a result, longer
addition times have lower maximum heat flows because the
energy is being spread over a longer period of time. In run
1, for instance, the addition time was 2.5 h, and the maximum
heat flow was measured to be about 70 W/L. When the
add time was increased to 3 h (run 2), the maximum heat
flow dropped to about 32 W/L, as can be seen in Figure 5.
Table 2 clarifies the conditions of each RC-1 experiment.
Run 2, for example, used 2 equiv of benzyl alcohol and thus
had a much larger initial volume than run 1, which reduced
the heat flow per unit volume. Run 3 had an add time of
about 1 h, but calorimetry was not obtained so as to eliminate
postreaction heating needed for calibrations and heat capacity
measurements. The estimated heat flow for run 3 with a 1
h add time was 100 W/L.24 The cooling needed to remove
100 W/L without refluxing is significant and would probably
pose difficulties on scale.25 In this case, however, the
reaction temperature is within 5-10 °C of reflux or the
boiling point of toluene (110 °C), which acted as a barrier
to temperature rise, tempering the reaction mass as heat of
vaporization and preventing further temperature rise. (Clearly
this will not be the case, and the temperature will rise if the
reactor is sealed rather than vented to atmosphere.) As a
result of reflux, higher heat loads such as that calculated for
run 3 (100 W/L) are feasible.
Figure 5. Heat flow profiles of Curtius rearrangement for runs
1 (top) and 2 (bottom). The acryloyl azide is added to the reactor
linearly and is the stoichiometric limiting reagent. It is apparent
from run 1 that a change in mechanism occurred at about 2 h
into the addition in which the vinyl isocyanate was no longer
being trapped. With the increase in benzyl alcohol and the
addition of carbonate in run 2, the trapping efficiency was
further improved. Run 3 (not shown) was further improved
with a shorter add time and immediate cooling after the
addition of azide was complete.
with experimental results and was estimated on the basis the
summation of terms shown in eq 6:
A thermal mass flow meter in combination with GC/MS
was used during the RC-1 experiments to measure the rate
of off-gas evolution. The two curves shown in Figure 6
represent the rate of nitrogen and other gaseous byproducts
evolved during the Curtius rearrangement. In the upper curve
of Figure 6 (run 1), it is evident that the off-gas rate increased
as the reaction progressed. This was shown to be due to
acetaldehyde and vinyl isocyanate being released in larger
fraction as the reaction progressed. It was presumed that
BnOH had been entirely consumed by side reaction (Scheme
3, eq 3), with the urethane forming the undesired aminal.
When BnOH is no longer present, the vinyl isocyanate cannot
be trapped and thus is liberated as a gas through the
condenser. To address this problem, a second equivalent of
benzyl alcohol and potassium carbonate was added in run 2
to more efficiently trap the vinyl isocyanate and to minimize
the formation of the aminal 4 side product. The increased
trapping efficiency which resulted from this modification was
evident in the off-gas profile of run 2 (Figure 6), as shown
∆Hrxn(liquid) ) (∆Hf,urethane - ∆Hvap,urethane +
∆Hf,N2) - (∆Hf,azide - ∆Hvap,azide +
∆Hf,BnOH - ∆Hvap,BnOH) (6)
where ∆Hf is the gas-phase heat of formation and ∆Hvap is
the latent heat of vaporization. It was assumed that
∆Hvap,urethane and ∆Hvap,azide were approximately on the
same order of magnitude and cancel each other. Thus,
∆Hrxn(liquid) ) ∆Hrxn(gas) - ∆Hvap,BnOH, where ∆Hvap,Bn-
OH is referenced to be +161 kJ/mol. Thus, given the
approximations made in the theoretical CHETAH calculation,
the estimate was found to be fairly reasonable compared to
the RC-1 experimental result (-192 vs -244 kJ/mol).
The adiabatic temperature rise calculated for the Curtius
rearrangement was 142 °C. This is interpreted for the desired
reaction such that there is sufficient exothermic energy to
raise the temperature of the reaction mixture above the
processing temperature by 142 °C in a purely adiabatic
system. This calculation does not recognize whether second-
ary exotherms are encountered along the way, however. In
this particular process, no thermal activity remains based on
DSC, once the azide has been added at 105 °C and the
urethane has been formed.
(24) In this case, the maximum heat flow was estimated from the following
equation: (Qmax/volume) ) (∆H/tadd)/volume ) (243 kJ/3600 s/0.67 L) ×
1000 J/kJ ) 100 W/L. This assumes a constant heat flow, or q(t) ) Qmax
) ∆H/tadd, which is an approximation for dose-rate controlled reactions
(i.e., square-wave heat flow profiles).
(25) Knowledge of the heat transfer coefficient and heat transfer area for the
large scale reactor would allow for a calculational check of the ∆T across
the jacket. Then a determination of feasibility for a given addition rate could
be made.
The reaction of acryloyl azide with benzyl alcohol was
dose-rate controlled. This was based on the observation at
the end of the azide addition which indicated that the heat
388
•
Vol. 2, No. 6, 1998 / Organic Process Research & Development

Table 2. Conditions of reaction calorimetry experiments (all three experiments were performed at 105 °C in RC-1)
run 1 run 2 run 3
azide-toluene solution
(36% acryloyl azide)
(1.23 mol of azide)
benzyl alcohol
toluene
hydroquinone
335 g
335 g
335 g
1 equiv (134 g)
242 g
6 g
2 equiv (268 g)
242 g
6 g
2 equiv (268 g)
242 g
6 g
pyridine
potassium carbonate
total mass
6 g
0 g
388 g
723 g
6 g
85 g
607 g
942 g
6 g
85 g
607 g
942 g
initial
final
add time for dosing the azide solution
hold time at 100 °C after the dosing
calorimetry measurements
measured heat of reaction without Qdose
measured heat of reaction with Qdose
maximum heat flow
2.5 h
3 h
yes
-202 kJ/mol
-244 kJ/mol
70 W/L
3.0 h
3 h
yes
-203 kJ/mol
-245 kJ/mol
32 W/L
1.0 h
1 h
no
100 W/L (calcd)
Figure 6. Off-gas evolution curves measured from runs 1 (top)
and 2 (bottom) using a thermal mass flow meter. The profiles
were consistent with the interpretation of the change in
mechanism observed in the heat flow data in Figure 3. In run
1, not only was nitrogen being liberated, but also vinyl
isocyanate was liberated through the condenser in increasing
fashion instead of being trapped to produce the urethane. In
run 2, it is apparent that the off-gas remained more constant
throughout the addition, which suggests that vinyl isocyanate
was not being liberated in significant quantities.
Figure 7. Instantaneous heat of Curtius rearrangement per
mole of gas liberated as a function of time. This plot was derived
by dividing the heat flow profiles shown in Figure 5 with the
gas flow profiles shown in Figure 6.
increased selectivity while the heat per mole of gas was fairly
steady at 250 kJ/mol of gas.
FTIR spectra obtained for the azide in toluene feed
solution showed a strong band absorbing at 2135 cm-1,
which corresponds to the azide R-N3 resonance structure.
During the Curtius rearrangement, this band was not apparent
during the reaction profile.26 This is significant as it indicated
qualitatively that no accumulation of the azide occurred
during the reaction. Further evidence of this was observed
from the RC-1 and mass flow meter profile where, at the
end of the azide addition, the heat flow and off-gas signals
return to zero, indicating no further reaction. FTIR profiles
of characteristic bands for the desired product also provided
by less vinyl isocyanate or other gases being produced.
Instantaneous heat of Curtius rearrangement per mole of
gas liberated is shown in Figure 7 and was obtained by
dividing the heat flow profiles shown in Figure 5 with the
gas flow data of Figure 6. Notice in run 1 (Figure 7) that
the heat per mole of liberated gas initiates near 224 kJ/mol
but quickly tapers off toward the end of the addition. This
is consistent with less trapping and increased rate of gas
liberation (both N2 and vinyl isocyanate), which tend to
reduce the heat per volume of gas liberated in run 1.
Essentially, only nitrogen was liberated in run 2 due to more
efficient trapping of the vinyl isocyanate resulting in
(26) ASI’s DiComp probe is a composite of Zn/Se and diamond crystal with a
region of low energy in the 1900-2100 cm-1 range. Nevertheless, the azide
functional group was clearly visible in the acyl azide in toluene spectrum.
Vol. 2, No. 6, 1998 / Organic Process Research & Development
•
389

Figure 8. FTIR spectra (above) of Curtius rearrangement reaction (in situ) in run 2. A wavenumber of 1648 cm-1 can be assigned
to a characteristic peak of product 1. No unique peak was observed for the aminal byproduct 4, but wavenumber 1730 cm-1
included absorbance for both 1 and 4. Wavenumber 1027 cm-1 can be assigned to benzyl alcohol but with some overlap with 1.
Temporal profiles of assigned wavenumbers are also shown in the lower plot.
valuable information. Figure 8 shows that the product
urethane 1 band at 1648 cm-1 reached a maximum before it
declined, while a band characteristic of both product 1 and
aminal 4 reached a maximum and then leveled off (1730
cm-1). No unique peak was observed for the aminal
byproduct 4. Wavenumber 1027 cm-1 was assigned to
benzyl alcohol but with slight overlap with 1. Taking the
difference between the temporal profiles of 1648 and 1730
cm-1, it followed that the formation of 1 reached a maximum
at the end of the addition and that the undesired aminal side
product 4 continued to be produced with prolonged heating.
This observation was consistent with calorimetry and yield
results. Namely, the longer add times and excessive heating
caused 1 to react further with remaining BnOH to form the
undesired aminal 4. This undesired reaction was supressed
by faster add times and rapid cool-downs following the azide
addition. The improvement is noted by the selectivity of
run 3 (Scheme 3), with a 15:1 desired urethane relative to
undesired aminal ratio.
Conclusion
Run 3 was carried out with a total addition time of 1 h
and in the presence of potassium carbonate and 2 equiv of
benzyl alcohol. Under these conditions, the ratio of product
390
•
Vol. 2, No. 6, 1998 / Organic Process Research & Development

to aminal was ∼15:1. The extra equivalent of benzyl alcohol
and other process changes essentially eliminated the amount
of vinyl isocyanate being lost in the headspace. The amount
of undesired aminal formed was also shown to be minimized
with a short addition time, cooling immediately upon
completion of the addition, as well as the presence of
potassium carbonate to serve as an acid scavenger.
The stoichiometric limiting reagent of the Curtius re-
arrangement was the acryloyl azide, which was established
to be thermally unstable at temperatures below the synthesis
temperature. However, as we learned from this investigation,
the mechanism for thermal decomposition of the acyl azide
(and liberation of N2) was also the desired reaction for the
trapping reaction of the Curtius rearrangement, albeit under
carefully controlled conditions. Viewed another way, the
desired reaction (Curtius and trapping) involves a controlled
decomposition of the acyl azide, as it is introduced into the
reactor held at 105 °C to generate the vinyl isocyanate which
immediately reacts with the benzyl alcohol. Because the
final reaction products and byproducts were thermally stable
and the desired reaction was dose-rate controlled, the safe
operation of the process focused primarily on the storage
and handling of the acyl azide solution. Because the azide
in toluene solution poses a thermal and pressure hazard, it
is recommended that the inventory of this solution be
minimized and solutions be consumed quickly after prepara-
tion and that, if stored, they be maintained at ambient
temperatures or below and that they be properly vented to
allow for the escape of nitrogen. Further, vent-sizing
experiments should be performed for the vessel feed tank
and reactor to prevent overpressurization under potential
worst case scenarios.
The reaction calorimetry revealed that the rate of reaction
and the associated exotherm were strictly dependent on the
rate of azide addition. The combined technologies utilized
in this investigation greatly increased our understanding of
the process and accelerated the delivery of the safety
assessment to our proposed vendor.
Combination of reaction calorimetry and thermal analysis
with the interfaced technologies (FTIR, GC/MS, and thermal
mass flow meters) resulted in the following:
(i) increased clarity and understanding of the reaction
mechanisms;
(ii) confidence that the process could be safely scaled up;
(iii) a significant increase in the selectivity and yield of
the desired urethane; and
Acryloyl Azide. Following a slight modification of the
literature preparation of compound 1, acryloyl azide was
prepared by dropwise addition of 100 mL of acryloyl chloride
(1.23 mol, 1.0 equiv) dissolved in 150 mL of toluene to 83.8
g of sodium azide (1.29 mol, 1.05 equiv) dissolved in 300
mL of water27 at a rate to keep the temperature <0 °C. This
reaction was fairly exothermic, and the addition was done
over an hour using an ice/acetone cooling bath. The mixture
was Vigorously stirred at 0 °C for 3 h.28 The final pH was
∼8. An additional 100 mL of toluene was then added to
give a better separation of the organic and aqueous layers.
The layers were separated, and the organic layer was washed
with 100 mL of 10% sodium carbonate. This was followed
by 4 × 100 mL water29 washes and 2 × 100 mL brine
washes, and then the toluene solution containing the acyl
azide was filtered through 50 g of sodium sulfate to provide
a clear solution. This solution was stored at 5 °C overnight.30
1H NMR (400 MHz, toluene-d8): δ 6.48-6.54 (m, 1H),
6.10-6.19 (m, 1H), 5.91-5.98 (m, 1H).
N-Vinyl-O-benzyl Urethane (1). The toluene solution
of the acyl azide was added dropwise to a 100 °C mixture
of 256 mL of benzyl alcohol (2.46 mol, 2.0 equiv), 6.1 g of
hydroquinone (0.055 mol, 0.045 equiv), 6.0 mL of anhydrous
pyridine (0.074 mol, 0.060 equiv), 85.0 g of potassium
carbonate (0.62 mol, 0.5 equiv), and 250 mL of toluene.31
This reaction is Very exothermic, and nitrogen is liberated!
The nitrogen evolution subsided immediately upon compe-
tion of the addition. After the potassium carbonate was
filtered off, the reaction mixture was concentrated in vacuo
to remove toluene and some of the benzyl alcohol. The
residue could be further purified by fractional distillation,
according to Hegedus,32 or purified by silica gel chroma-
tography.33 The final product is a crystalline solid which
melts at 43-44 °C. Our Discovery group routinely obtained
150 g of clean product (69% of theory).
1H NMR (400 MHz, CDCl3): δ 7.30-7.41 (m, 5H),
6.65-6.77 (m, 1H), 6.49 (bs, 1H), 5.12 (s, 2H), 4.41-4.52
(m,1H), 4.28-4.32 (m, 1H). TLC (silica gel, 80:20 hexane/
ethyl acetate): Rf 0.40. HPLC (2400:1600:8:4 water/
acetonitrile/triethylamine/acetic acid, 4.6 × 150 mm C8
Zorbax, 210 nm, 1 mL/min): tr ) 8.3 min.
Aminal Side Product 4. For characterization puroposes,
a purified sample of 4 was obtained by flash chromatography
using the system previously described.
(27) The original preparation used distilled water. We used Groton city water
with no noticeable difference.
(28) If the reaction was allowed to warm to room temperature and then stir
overnight, some decomposition of the azide was observed.
(29) Due the presence of chloride in the tap water used, we did not do the silver
nitrate test described by Hegedus.
(iv) a savings of 2-4 months of development time while
on critical path.
The acryloyl process was transferred to a vendor where
it was safely scaled to produce 1, N-vinyl-O-benzyl urethane.
(30) A solution of the acyl azide in toluene-d8 was shown to be stable under
these storage conditions. If the solution is stored at room temperature, some
nitrogen pressure is observed. The solution may get cloudy after overnight
storage.
(31) The original procedure did not add additional toluene here. We saw a more
controlled addition/exotherm with this modification.
(32) This requires a very efficient vacuum source, or else the longer contact
times on larger scale for the material in the pot results in some decomposi-
tion. Bp 120 °C at 0.5 mmHg.
(33) 10 g of silica gel/1 g of product theory. Gradient of hexane to 97.5:2.5
hexane/ethyl acetate. Conditions courtesy of G. T. Magnus-Aryitey, C. J.
Mularski, and R. T. Wester.
Experimental Section
Reagents. Acryloyl chloride was 96% pure and used as
received from Aldrich. Benzyl alcohol was 99.8% anhydrous
and used as received from Aldrich. Potassium carbonate
from J. T. Baker was finely ground with a mortar and pestle.
Toluene from Fisher (certified ACS) was used as received.
All other reagents (pyridine, hydroquinone, and sodium
azide) were also used as received.
Vol. 2, No. 6, 1998 / Organic Process Research & Development
•
391

1H NMR (300 MHz, CD3CN): δ 7.29-7.40 (m, 10H),
6.13 (bs, 1H), 5.01-5.17 (m, 3H), 4.55 (d, J ) 11.8 Hz,
1H), 4.45 (d, J ) 11.8 Hz, 1H), 1.31 (d, J ) 6.0 Hz, 3H).
TLC (silica gel, 80:20 hexane/ethyl acetate): Rf 0.35. HPLC
(2400:1600:8:4 water/acetonitrile/triethylamine/acetic acid,
4.6 × 150 mm C8 Zorbax, 210 nm, 1 mL/min): tr ) 24.2
min. (The N-vinyl compound has a 1.9 times higher
absorbance than the aminal at this wavelength.)
equipped with automatic sampling valves. Thermal analysis
was performed using DSC (Mettler 821e), Radex-Solo (Astra
Scientific), and accelerating rate calorimetry (ARC, Columbia
Scientific) instruments.
Nomenclature
A: heat transfer area, m2
qr(t): instantaneous heat transfer, W or J/s
mr: reaction mass, kg
Instrumentation. To investigate the preparation of 1 (N-
vinyl-O-benzyl urethane), the Curtius reactions were per-
formed in a fully automated reaction calorimeter (Mettler
RC-1) equipped with a 1 L (MP-10) glass reactor and
thermostated stainless steel head. Stirring was accomplished
with a stainless steel anchor impeller. The acyl azide in
toluene solution was pumped to the reactor at a linear rate
using a control loop between the RC-1 computer and an
analytical balance. An Applied Systems (ReactIR 1000)
reaction analysis system was utilized to collect in situ
compositional (FTIR) data. FTIR spectra were collected
using a diamond-Zn/Se composite (DiComp) attenuated total
reflectance (ATR) sampling probe inserted into the RC-1.
Off-gas flow rates were measured with an Omega FMA-
1800 series thermal mass flow meter, while off-gas compo-
sitional data were measured with a Hewlett-Packard 6890
gas chromatograph with mass-selective detector (MSD)
dmdose/dt: addition rate of dosed material, kg/s
Cpdose: heat capacity of added material, J/(kg K)
∆Hrxn: molar heat of reaction, kJ/mol
∆Htotal: total enthalpy of thermal event, kJ/kg
t: time, s
Tadiabatic: adiabatic temperature rise, °C
Tdose: temperature of added (dosed) material, °C
Tr: reaction temperature, °C
Tj: temperature of reactor jacket, °C
TMR: time to maximum rate, h
U: heat transfer coefficent, W/(m2 K)
V: volume of raction mass, L
Received for review September 1, 1997.
OP970115W
392
•
Vol. 2, No. 6, 1998 / Organic Process Research & Development