Continuous processing to control a potentially hazardous process: Conversion of aryl 1,1-dimethylpropargyl ethers to 2,2-dimethylchromenes (2,2-dimethyl-2H-1-benzopyrans)

Article abstract of DOI:10.1021/op0100504

The thermal Claisen rearrangement of 4-cyanophenyl 1,1-dimethylpropargyl ether (4) to 6-cyano-2,2-dimethylchromene (5), (6-cyano-2,2-dimethyl-2H-1-benzopyran), which is used in the synthesis of a potassium channel activator drug candidate, BMS-180448, created a significant process development issue. The resulting large heat release in this conversion posed not only a safety risk but could also cause product degradation if done in a batch-wise manner. The solution was to exploit the high surface-to-volume ratio of a plug-flow reactor that would maximize the heat transfer, thereby permitting tight and responsive temperature with better reaction control. In the course of successfully testing the plug-flow concept on "micro"-flow scale (gram quantity) and "kilo"-flow scale (～10 kg), a generalized mathematical model capable of predicting the reaction performance based on the physical properties of any given plug-flow reactor was generated. The model provides requisite information to design and operate a plug-flow reactor of any size for this reaction. This model would optimize reaction conditions for an acquired reactor system capable of producing ～7 kg/h of the dimethylchromene. Application of plug-flow reactor technology enabled production of high quality 2,2-dimethylchromenes in good yield (>98 mol %) without the use of solvents and with virtually no waste streams.

Full text of DOI:10.1021/op0100504

Organic Process Research & Development 2001, 5, 636−645
Continuous Processing to Control a Potentially Hazardous Process: Conversion
of Aryl 1,1-Dimethylpropargyl Ethers to 2,2-Dimethylchromenes
(2,2-Dimethyl-2H-1-Benzopyrans)
Ricardo J. Bogaert-Alvarez,† Paul Demena, Gus Kodersha, Robert E. Polomski,* N. Soundararajan,* and
Steve S. Y. Wang
Process Research and DeVelopment, Pharmaceutical Research Institute, Bristol-Myers Squibb Company,
1 Squibb DriVe, New Brunswick, New Jersey 08903, and Lafayette College, Chemical Engineering Department,
Easton, PennsylVania 18042, U.S.A.
Abstract:
The thermal Claisen rearrangement of 4-cyanophenyl 1,1-
dimethylpropargyl ether (4) to 6-cyano-2,2-dimethylchromene
(5), (6-cyano-2,2-dimethyl-2H-1-benzopyran), which is used in
the synthesis of a potassium channel activator drug candidate,
BMS-180448, created a significant process development issue.
The resulting large heat release in this conversion posed not
only a safety risk but could also cause product degradation if
done in a batch-wise manner. The solution was to exploit the
high surface-to-volume ratio of a plug-flow reactor that would
maximize the heat transfer, thereby permitting tight and
responsive temperature with better reaction control. In the
course of successfully testing the plug-flow concept on “micro”-
flow scale (gram quantity) and “kilo”-flow scale (∼10 kg), a
generalized mathematical model capable of predicting the
Figure 1. Benzopyran-based potassium channel activators.
reaction performance based on the physical properties of any
given plug-flow reactor was generated. The model provides
Scheme 1
requisite information to design and operate a plug-flow reactor
of any size for this reaction. This model would optimize reaction
conditions for an acquired reactor system capable of producing
∼7 kg/h of the dimethylchromene. Application of plug-flow
reactor technology enabled production of high quality 2,2-
dimethylchromenes in good yield (>98 mol %) without the use
The impetus for designing processes for large-scale
production of these compounds stems from the large number
of naturally occurring 2,2-dimethylchromene (2,2-dimethyl-
2H-1-benzopyran) derivatives 3 and the considerable interest
in them due to their biological activity in plants and animals.1
A series of benzopyran-based potassium channel activators2
(Figure 1) has generated intense interest in the synthesis of
2,2-dimethylchromenes, particularly those bearing an electron-
withdrawing group.
of solvents and with virtually no waste streams.
Introduction
The variety and relatively small annual production of
many compounds produced in the pharmaceutical and fine
chemical industries lead to a flexible production capability
based on batch or semibatch processing. However, some
chemical processes or reactions would be hazardous or
unnecessarily risky if performed in this manner. The prepara-
tion of 2,2-dimethylchromene (2,2-dimethyl-2H-1-benzopy-
ran) derivatives 3 necessitated consideration of an alternative
processing approach due to a high heat of reaction that could
lead to an uncontrollable temperature rise from batch-wise
preparation.
The O-alkylation of phenols with 3-chloro-3-methylbutyne
(1) followed by thermal rearrangement of the resulting aryl
(1) Chromenes, Chromanones, and Chromones. In The Chemistry of Hetero-
cyclic Compounds; Ellis, G. P., Ed.; John Wiley and Sons: New York,
1977; Chapter 2.
(2) Ashwood, V. A.; Buckingham, R. E.; Cassidy, F.; Evans, J. M.; Faruk, E.
A.; Hamilton, T. C.; Nash, D. J.; Stemp, G.; Willcocks, K. J. Med. Chem.
1986, 29, 2194-2201; Fozard, J. R.; Menninger, K.; Cook, N. S.; Blarer,
S.; Quast, U. Br. J. Pharmacol. 1990, 99, 6P; Buckle, D. R.; Eggleston, D.
S.; Pinto, I. L.; Smityh, D. G.; Tedder, J. M. Bioorg. Med. Chem. Lett.
1992, 2, 1161-1164; Bergmann, R.; Eiermann, V.; Gericke, R. J. Med.
Chem. 1990, 33, 2759-2767; Atwal, K. S.; Grover, G. J.; Ahmed, S. Z.;
Ferrara, F. N.; Harper, T. W.; Kim, K. S.; Sleph, P. G.; Dzwonczyk, S.;
Russell, A. D.; Moreland, S.; McCullough, J. R.; Normandin, D. E. J. Med.
Chem. 1993, 36, 3971-3974.
* All correspondence should be addressed to: Robert E. Polomski and
N. Soundararajan, Process Research and Development, Pharmaceutical
Research Institute, Bristol-Myers Squibb Company, 1 Squibb Drive, New
Brunswick, NJ 08903. E-mail: Robert.Polomski@bms.com. E-mail:
Nachimuthu.Soundararajan@bms.com. Telephone: Polomski: 732-519-2748,
N. Soundararajan: 732-519-3710. Fax: 732-519-1922.
† Lafayette College.
636
•
Vol. 5, No. 6, 2001 / Organic Process Research & Development
10.1021/op0100504 CCC: $20.00 © 2001 American Chemical Society and The Royal Society of Chemistry
Published on Web 10/16/2001

Scheme 2
1,1-dimethylpropargyl ether, 2, remains a convenient method
for the preparation of 2,2-dimethylchromenes 3 (Scheme 1).3
This method is based on the work of Iwai and Ide4 who
demonstrated that thermal rearrangement of simple aryl
propargyl ethers yielded chromenes. Subsequent work at
Bristol-Myers Squibb demonstrated that the conversion of
1 to 2 is catalyzed by copper I and II salts and that the
trifluroacetate corresponding to 1 (generated in situ) is an
effective alkylating agent in the presence of copper (Scheme
2).5
reaction on neat 4. Although showing promise, these experi-
ments demonstrated the difficulties of controlling the ex-
tremely energetic reaction.
Hazard Lab Evaluation. A hazards evaluation utilizing
differential scanning calorimetry (DSC), accelerating rate
calorimetry (ARC), and the reactive system screening tool
(RSST) provided information to guide further experimenta-
tion, resulting in the development of a strategy to enable
safe and consistent operations on pilot and (potential)
commercial scale.
DSC testing (Figure 2) showed two broad exotherms with
onset temperatures of 105 and 240 °C, respectively. The
adiabatic ARC test (Figure 3) showed the temperature and
pressure history of the thermal conversion of 4 under
adiabatic mode from which the heat of reaction and the
reaction kinetics were quantified. On the basis of the ARC
data, no thermal reaction would be expected to occur below
the onset temperature of ∼105 °C. However, in the absence
of sufficient temperature control, the heat generated in the
desired thermal reaction could propel the system temperature
above the onset of a secondary decomposition reaction. The
intramolecular nature of the reaction suggested first-order
kinetics. The data treatment of the calorimetric data empiri-
cally confirmed this observation. The heat rate versus
temperature plot (Figure 4) allowed calculation of the first-
order rate constant according to the kinetic expression:
Results and Discussion
An ongoing project required the preparation of a sub-
stantial amount of 6-cyano-2,2-dimethyl-[2H]-chromene (5).5
With the previously described improvements to the prepara-
tion of the propargyl ether 4, efforts shifted to the conversion
of 4 to 5. In the past, the thermal Claisen rearrangement-
cyclization of 4 had been performed by heating in o-
dichlorobenzene,5 N,N-diethylaniline,6 or peanut oil.7 On a
small scale, the reaction proceeded well, but with further
laboratory scale-up, an uncontrolled temperature excursion
occurred when the substrate and solvent were mixed at room
temperature and then heated, due to the rapid and highly
exothermic reaction. A semibatch approach, by portion-wise
addition of 4 into hot (∼180 °C) N,N-diethylamine, provided
yields of 80-85% after crystallization. However, diethy-
lamine removal required dilution with toluene and extraction
with aqueous acid. Handling the toxic amine solvent (the
OSHA exposure limit is a time weighed average of 5 ppm
over 8 h) on a large scale and the relatively lengthy cleanup
procedures discouraged consideration of this approach for
further scale-up efforts. Varying success in incorporating
other “high” boiling solvents (dodecane, ethylene glycol, 1,3-
dimethyl-2-imidazolidone, 2-benzylpyridine, or 2,4,6-colli-
dine) for the conversion prompted a small (2-5 g), laboratory-
scale assessment of the thermal rearrangement-cyclization
dC(4)
dt
-32000/
) kC(4) ) -4.6 × 1014 exp
C(4)
(
)
RT
in which:
C(4) ) the concentration of reactant 4 (mol/mL)
t ) time (min)
k ) rate constant (min-1
)
R ) gas constant, 1.987 (cal/mol K)
T ) reaction temperature (K).
(3) Hlubucek, J.; Ritchie, E.; Taylor, W. C. Tetrahedron Lett. 1969, 1369-
1370; Hlubucek, J.; Ritchie, E.; Taylor, W. C. Aust. J. Chem. 1971, 24,
2347-2354.
Assuming a specific heat for the ether derivative 4 of 0.61
cal/g °C, the heat of reaction was calculated to be 41.2 kcal/
mol. This value translates to an adiabatic temperature rise
of ∼375 °C during the course of the reaction. RSST
(4) Iwai, I.; Ide, J. Chem. Pharm. Bull. 1963, 11, 1042-1049.
(5) Godfrey, J. D., Jr.; Mueller, R. H.; Sedergran, T. C.; Soundararajan, N.;
Colandrea, V. C. 208th ACS National Meeting; American Chemical
Society: Washington, DC, August 21-25, 1994.
(6) Harfenist, M.; Thom, E. J. Org. Chem. 1972, 37, 841-848.
Vol. 5, No. 6, 2001 / Organic Process Research & Development
•
637

Figure 2. DSC thermograph of compound 4 (1 °C/min, 4.5 mg in aluminum crucible).
Figure 3. ARC temperature and pressure vs time plot.
evaluation of neat material showed thermal stability behavior
with a self-heat rate of less than 2 °C/min for temperatures
below 145 °C, consistent with adiabatic studies in the ARC
which showed similar self-heating rates (<1-1.5 °C/min)
at 150 °C. For these low self-heating rates, an adequately
jacketed batch reactor should control the reaction up to
∼140-150 °C. However, at >150 °C the rapid acceleration
of the reaction rate and hence the accelerating self-heating
creates the potential for an uncontrolled reaction. In one test,
the temperature of the reaction rose from 180 to 445 °C
within 33 seconds.
Scale-Up Studies. On the basis of these data, safety and
operating concerns had to be addressed prior to further
experimentation and scale-up. The concerns centered around
maintaining control of the reaction temperature while driving
the reaction to completion. Although temperature control
increases with lower operating temperatures, lower operating
temperatures prolong the time to reach reaction completion
(∼27 h at 140 °C vs ∼10 h at 150 °C). This encourages
using an operating temperature that is as high as can be safely
controlled while, at the same time, controlling the risk of
the accelerating heat generation at the higher temperatures.
638
•
Vol. 5, No. 6, 2001 / Organic Process Research & Development

Figure 5. RSST temperature vs time plot (5% compound 4 in
diethylaniline, heat rate ) 0.25 °C/min).
surface (S) for the volume (V) of reaction decreases inversely
as the diameter of the reaction vessel increases:
Figure 4. ARC self-heat rate data (O factor ) 2.15, Ti bomb).
ΠDL
Π(D/2)2 L
S/V )
) 4/D
For a batch system, selection of the operating temperature
requires that sufficient spare heat-exchange capability be
designed in the reactor to control potential thermal excursions
and avoid a “runaway” reaction. Typically, batch operations
use a solvent system to maintain a fixed boiling point of the
mixture at the desired reaction control point and to remove
the heat of reaction in the condenser system. Keeping the
overall boiling point within the desired temperature control
range using solvents typically available in pharmaceutical/
fine chemical manufacturing (boiling point typically below
120 °C) requires careful maintenance of the solvent/substrate
ratio. The operating temperature range would have to be
optimized in conjunction with the heat-transfer capabilities
of the reactor and the condenser to prevent loss of solvent.
Loss of solvent (through insufficient condensing capacity,
loss of coolant, excessive reflux) would accelerate the
reaction (due to increased boiling point of the batch), leading
to further solvent loss. Should the rate of heat generation
exceed control limitations, a rapid and excessive pressure
buildup in the reactor (from solvent evaporation and possible
product decomposition) could cause the material to entrain
from the reactor in a potentially violent manner. The thermal
potential for compound 4 with various solvents was tested
with the reactive system screening tool (RSST). Figure 5
shows the runaway potential with only 5% of 4 in diethyl-
aniline. In the absence of controls, this reaction mixture could
begin to accelerate at ∼140-150 °C and reach a maximum
temperature of ∼215 °C.
thus, further hindering the ability to control the reaction.
Application of a Plug-Flow Reactor. The nature of this
reaction makes it an ideal candidate for further scale-up in
a plug-flow reactor (PFR). The “plug flow” nature of the
reactor stems from the “flat” (constant) cross-sectional
velocity profile due to the turbulent nature of the flow or
induced by the use of static mixers. A PFR improves the
efficiency of heat transfer by maintaining a high surface-to-
volume ratio as the operation increases in scale. Scale-up
could mean simply using multiple identical tubular reactors
as were used in a pilot campaign within a large heat
exchanger. Although the material is processed continuously,
only a relatively small portion of the overall material is
committed to the reaction at a given time, limiting the extent
of problems, should difficulties arise with the reaction.
Should the process-monitoring equipment indicate a loss of
control of the desired steady-state operating conditions,
adjustments can be made to the flow rate of the reactant and
the flow rate and temperature of the heat-transfer fluid to
reestablish control or to quickly bring the operation to a safe
halt by allowing the residual material in the reactor to expend
its energy in a controlled fashion.
Lab Results-Experimental. The feasibility of the con-
tinuous plug-flow approach was tested at bench scale using
a 0.0625 in. i.d. × 10 ft stainless steel coil immersed in an
oil bath at 220 °C. Neat 4 was pumped through the coil using
a syringe pump at a rate such that the liquid residence time
(reactor volume/volumetric flow rate) varied from 3.5 to 17.7
min over the various trials. In all of these trials, compound
4 underwent almost quantitative conversion to 5 of good
quality (HPLC purity >96%). The outlet temperature from
the steel coil was close to the heating bath temperature (220
Increasing the operating scale of the batch reaction
exacerbates the control problem. Since most of the heat
transfer is through the wall of the reactor, the available
Vol. 5, No. 6, 2001 / Organic Process Research & Development
•
639

Table 1. Conversion of 2 to 3 at Bench-Scale Plug-Flow
Reactor
Mathematical Modeling
The PFR demonstrated its utility for handling this highly
exothermic reaction on small scale. However, any further
scale-up would require changes in PFR capacity (diameter,
length, flow rates) to handle the increased volume of material
to be processed. In turn, this necessitates a means to predict
the performance of the reactor system to (1) ensure safe
operating conditions, (2) set conditions for optimal perfor-
mance within the constraints of the heat-transfer capabilities
of a given PFR, and (3) avoid operating conditions that could
compromise product quality. The hazard analysis confirmed
the first-order nature of the reaction through data treatment
that gave an excellent fit to the first-order expression allowing
derivation of the rate constant and its temperature depen-
dence. The resulting kinetic rate expression enabled the
generation of a mathematical model reaction system capable
of predicting the extent of conversion and the temperature
profile throughout a given PFR. Substitution of the appropri-
ate physical and operating parameters for a given PFR into
the model allows prediction of conversion and extent of the
temperature rise within the reactor.
product(s)
temp residence
yield
2, R )
(°C) time (min)
3, R )
3, R ) (%)a ref
4-CN
4-OCH3
220
240
3.5
11.7
5.5
6-CN
98
78
80
96
78
5
8
9
9
10
6-OCH3
6-COCH3
6-NO2
6-I
7-CF3
7-NO2
8-CHO
4-COCH3 240
4-NO2
4-I
3-CF3
3-NO2
2-CHO
240
220
260
230
240
5.5
5.5
17.7
5.5
5.5
5-CF3
5-NO2 82b
74c none
86b 11
8
a Yields refer to purified (crystallization or column chromatography) product.
b Refers to the combined yields of the regio-isomers (the ratio of 5 to 7 isomer
is ca. 5 to 3). The components were separated by column chromatography.
c Besides the main component (oil) an unknown species (oil) with a molecular
weight of 414 was isolated.
Development of the Model. For a PFR, the material
balance equation becomes:
Table 2: Conversion of 4 to 5 in a Laboratory Plug-Flow
Reactor
wt. input 4
wt. output 5
yield
HPLC
dC(4)
dz
-32.08
RT
) -kC(4) ) -4.6 × 1014 exp
C(4)
batch no.
(kg)
(kg)
(m%)
purity (%)
u
(
)
001
002
003
004
005
13.20
10.70
10.90
10.67
6.82
12.02
10.48
10.30
10.43
6.52
91.0
97.9
94.7
97.7
95.6
95.92
96.70
97.93
97.86
97.79
where time (t) is replaced by z/u such that
z ) distance along the axial length of the PFR (cm)
u ) fluid velocity (cm/min)
°C). Since the steel coil was long enough and immersed in
a hot oil bath, the heat generated from the reaction was well
dissipated to the bath by the time the product exited the
reactor. The limited extent of each trial provided sufficient
qualitative information on the extent of the reaction and yield
to encourage further development efforts. The generality and
usefulness of this method was further demonstrated by
applying this technique for the conversion of other aryl 1,1-
dimethylpropargyl ethers5 2 to the corresponding 2,2-
dimethylchromenes 35,8-10 (Table 1).
Kilo Lab Results-Experimental. To provide quantities
of material for subsequent process development, 4 was
converted to 5 in a stainless steel PFR, which was a 0.125
in. i.d. × 25 ft tubing coil immersed in a constant-temperature
bath maintained at 195-200 °C. A flow of 20 mL/min
afforded the reaction mixture a residence time of ∼4 min.
Conversion reached completion in each of the five batches.
Although yields varied; the losses can be attributed to
residual material in the reactor and physical handling losses
(Table 2).
Because heat generation may exceed the heat-ttransfer
capabilities of the reactor, the reaction must be modeled as
a nonisothermal system. Hence, to predict the conversion of
4 through the reactor, the model has to take into account the
effect of the change in temperature throughout the reactor
due to the extent of conversion and the degree of heat transfer
along the reactor length. With the assumption of constant
reaction mixture density and heat capacity, the energy
balance that provides this information is given by:
dT
FuC
) -∆HkC(4) - Ua(T - Tc)
pdz
In this expression,
F ) density (1.01 g/mL)
Cp ) heat capacity, (0.6 cal/(g °C)
-∆H ) heat of reaction (41 200 cal/g mol)
U ) heat-transfer coefficient (cal/(cm2 min °C)
a ) specific area (S/V) ) 4/D
(7) Mark, C. P.; Reber, J.-L.; Peham, M.; Schultz, R.; Berthold, R.; Collect,
A.; Malthete, J. 9th International Conference on Organic Synthesis,
University of Montreal: Montreal, Quebec, Canada, June 28-July 2, 1992.
(8) Naidu, V.; Rao, K. Ind. J. Chem. 1979, 17B, 73-75. Also see, Ahulwalia,
V. K.; Mekherjee, I.; Mukherjee, K. Ind. J. Chem. 1984, 23B, 1124-1125.
(9) North, J. T.; Kronental, D. R.; Pullockaran, A. J.; Real, S. D.; Chen, H. Y.
J. Org. Chem. 1995, 60, 3397-3400.
(10) Ding, C. Z.; Rovnyak, G. C.; Misra, R. N.; Grover, G. J.; Miller, A. V.;
Ahmed, S. Z.; Kelly, Y.; Normadin, D. E.; Sleph, P. G.; Atwal, K. S. J.
Med. Chem. 1999, 42 (18), 3711-3717.
This model requires a simultaneous solution of the two
first-order, ordinary differential equations. Because the
equations are coupled by the dependent variables (the
temperature variable appears in the material balance expres-
sion, and the concentration variable appears in the energy
640
•
Vol. 5, No. 6, 2001 / Organic Process Research & Development

balance) a simple analytical solution cannot be provided. An
approximation, such as the Runge-Kutta method,12 can solve
systems of such equations by an iterative, numerical tech-
nique. An adaptive step fifth-order Runge-Kutta algorithm,13
part of the Mathcad 8 Professional software,14 was used to
solve the equations defining the model.
The variables that can be manipulated to control the
system response include the diameter and length of the PFR,
the volumetric flow rate of the reactant, and the temperature
of the heat-transfer medium. [For simplicity in this analysis,
a constant temperature of the heat-transfer medium will be
assumed. Depending on the system, (the flow rate and heat
capacity of the heat-transfer fluid), this temperature may vary
along the length of the reactor in co-current or counter-current
operation. An energy balance on the heat-transfer fluid can
be included in the model and solved simultaneously with
the continuity equations for the reaction.]. A critical param-
eter, the overall heat-transfer coefficient, cannot be inde-
pendently set as a control variable. The fluid properties of
the reaction mixture and the heat-transfer fluid, the thickness
and conductivity of the wall separating the two fluids, and
the incorporation of baffles and fins used to improve mixing
and heat-transfer influence this parameter. The overall heat-
transfer coefficient must be determined for a given reactor
system.
To simplify and generalize the analysis, dimensionless
variables convert the differential equations to the following:
dX
djz
1 - ht
ht
) -khX exp - ꢀ
[
(
)
]
dht
djz
1 - ht
) âX exp - ꢀ
-Uh(ht - tc)
[
(
)
]
ht
where
X ) unreacted fraction of 4 [C(4)/C0(4)]
C0(4) ) the initial concentration of 4
jz ) z/L ) ratio of the distance (z) in the reactor to
the reactor length (L)
An early consideration of running this reaction in a batch-
recycle manner highlights the risk of insufficient completion
of this reaction in the PFR. The reactant within a batch
reactor (storage vessel) achieves partial conversion passing
through a heated reactor. Prior to the return to the storage
vessel, the material passes through a heat exchanger thereby,
returning to the safe holding temperature. During the
operation, a potential exists of introducing heated material
into the hold vessel in the event of loss of cooling to the
heat exchanger or circumstances that might upset the heat
balance in the system. Until sufficient material is converted,
a risk remains that the reaction could proceed in an
uncontrolled manner in the recycle vessel. Therefore, if the
operation strategy for a given PFR dictates incomplete
conversion, the reactor should be operated to convert the
majority of material in a single pass. This conversion should
ensure that any conversion of the material in the hold tank
would limit the temperature rise well below the boiling point
of the mixture and the onset of decomposition. As a
precaution, additional heat removal capabilities should be
designed for the hold vessel.
kh ) k0L/u ) dimensionless reaction constant
k0 ) A0 exp(- ꢀ)
ht ) T/T0 )
ratio of the reaction temperature to the inlet temperature
ꢀ ) E/RT0
â ) -∆Hk0C0L/FCpT0u
Uh ) ULa/FCpu ) dimensionless heat-transfer coefficient
tc ) T/Tc )
ratio of the reaction temperature to the jacket temperature
subjected to the initial conditions:
jz ) 0,X ) 1,ht ) 1
Figures 6 and 7 provide temperature profiles and conver-
sions along a PFR for three different regimes of operation.
The most desirable situation would be the isothermal case
wherein the operating conditions are set so that heat is
removed as fast as it is generated, and the reaction mixture
is maintained at the temperature of the heating fluid. As
shown in Figure 7, for a properly designed reactor, the
reaction reaches completion at the end of the reactor. As
long as the temperature of the reaction mixture does not reach
the regime where thermal decomposition begins, a less
desirable, but operable, case occurs where the heat generated
exceeds the capability for heat removal. With careful
consideration of reactor performance and good temperature
control, this can be a means to improve reactor throughput.
As the corresponding conversion curve in Figure 7 shows,
this particular reactor is over-designed as the reaction reaches
completion within the first 25% of the length of the reactor.
Capacity can be increased by increasing the flow rate, or a
Discussion of Modeled Reactor Performance
Application of the model to the reaction system implies
turbulent flow or more specifically a flat velocity profile
across the reactor diameter. Turbulent flow indicates good
mixing and hence a constant temperature profile. For the
nature of this reaction system, laminar flow should be
avoided for this very reason. Static mixers incorporated in
the reactor improve the cross-sectional mixing and enhance
the heat transfer. This discussion demonstrates the use of
the model as a guide for selecting appropriate reaction
conditions (temperature, reactor length, reactor diameter, and
flow rate) to ensure adequate control of a thermally sensitive
reaction.
(11) Evans, J. M.; Stemp, G.; Tedder, J. M. PCT Int. Appl. 1991.
(12) Carnahan, B.; Luther, H. A.; Wilkes, J. O. Applied Numerical Methods,
John Wiley and Sons: New York, 1969.
(13) Press, H. E.; Flannery, S. A.; Teukolsky, B P.; Vetterling, W. T. Numerical
Recipes in C; Cambridge University Press, New York, 1992.
(14) Mathcad 8 Professional; MathSoft, Inc.: Cambridge, MA, 1998.
Vol. 5, No. 6, 2001 / Organic Process Research & Development
•
641

Figure 6. Temperature profile throughout reactor for different modes of operation.
Figure 7. Conversion profile across reactor for various operating conditions.
much shorter reactor can be used to achieve the same
performance.
The undesirable case occurs if the reaction mixture
temperature exceeds the threshold of thermal decomposition.
In this case, the operating temperature may be too high or
the volumetric flow rate may be too low (or the combination
of both) for the specific reactor configuration. For this case,
Figure 7 shows that conversion is complete within the initial
part of the reactor. Milder operating conditions can be
employed to avoid the high-temperature excursion and to
642
•
Vol. 5, No. 6, 2001 / Organic Process Research & Development

Figure 8. Effect of jacket temperature on maximum temperature in reactor.
make better use of the overall capacity of the reactor. Ideally,
the reactor should be sized or the operating parameters set
to complete the thermal conversion by the time the material
exits the PFR. In cases where a PFR not specifically designed
for this application is used, total conversion might not be
possible on a single pass. As stated above, the first pass
should convert a majority of the material to ensure that the
partially converted material can be safely cooled and held
prior to final pass to complete the reaction.
Use of the mathematical simulation for (1) understanding
the behavior of the reaction, (2) helping design a reactor for
scaled-up processing, and (3) establishing an operation
strategy to maximize productivity while providing the desired
product in a safe, controlled, and consistent manner are
demonstrated in the following sections.
Simulation of a Small-Scale Reactor. The simulated
results in Figure 8 show the significant improvement in
conversion that is achieved with small (5-10 °C) increments
in the jacket temperature. Although Figure 8 is based on a
relatively small reactor (inside diameter ) 0.0625 in., with
a volumetric flow rate of 0.5 mL/min), a similar performance
could be expected with larger reactors. The figure demon-
strates that in a region of “lower” jacket temperature,
temperature control will not become an issue, although the
reaction may not proceed to completion for a fixed length
of reactor. Increasing the jacket temperature increases the
degree of conversion, but at some point concerns develop
about the possibility of a “hot-spot” in the reactor. When
the rate of reaction reaches a point that the heat generation
exceeds the ability to transfer the heat to the heat-transfer
fluid, the temperature rises and reaches a maximum at some
point in the reactor. Once the majority of the reaction is
complete, the heat-transfer process will bring the temperature
of the flowing reactant back toward the control temperature
downstream of the maximum point. Figure 8 indicates that
a temperature maximum of 550 °C was reached in the reactor
when 220 °C was used as the jacket temperature. Figure 8
clearly shows that operation above 200 °C gives a significant
liability for exceeding the thermal decomposition limit. The
real significance of Figure 8, however, lies in the extreme
sensitivity for a thermal excursion at a threshold of ∼195
°C. This threshold temperature will vary with reactor
systems, based on the diameter of the reactor, the overall
heat-transfer coefficient, and the volumetric flow rate.
Beyond that temperature, relatively small increases in the
reaction temperature can accelerate the reaction and heat
generation well beyond the temperature where thermal
degradation of the product sets in.
Mathematical Model for Large-Scale Unit. The heat-
transfer coefficient plays a determining role in the setting
of operating parameters. The model simulation provides
insight into the performance of a larger-diameter reactor
(length ) 35 ft, i.d. ) 0.375 in., flow rate ) 45 mL/min)
operating with a jacket temperature of 172 °C while varying
the heat-transfer coefficient from 0.1 to 2.0 cal/(cm2 min
K): values that can be considered a reasonable range for a
tubular heat exchanger. For the given reactor and operating
conditions, Figure 9 shows the extent of conversion and the
Vol. 5, No. 6, 2001 / Organic Process Research & Development
•
643

Figure 9. Maximum conversion and temperature for different heat-transfer coefficients.
maximum temperature liability. The sensitivity of the opera-
tion should the overall heat transfer coefficient be <0.8 cal/
(cm2 min K) is evident. The system shows fairly isothermal
behavior for heat-transfer coefficients >0.8 cal/(cm2 min K),
although significant material remains unreacted. Either a
longer reactor would be needed, or the material would require
an additional pass through the reactor. Clearly, should the
heat-transfer coefficient be <0.7 cal/(cm2min K), temperature
control can become problematic, especially if heat-transfer
performance diminishes (e.g., by fouling on either side of
the reactor). For this reactor configuration, a lower operating
temperature and higher flow rates, with possibly two cycles
through the reactor, would be necessary to avoid the
predicted elevated reaction temperature. A potential mode
for running this reaction with the low heat-transfer coefficient
would be to run the material in two passes while maximizing
the flow rate at a jacket temperature of ∼172 °C to achieve
at least a 50% conversion. Since the majority of the heat of
reaction would be dissipated and the converted material
would help serve as a heat sink, the second pass can be run
at a higher temperature to ensure complete conversion.
For a somewhat longer reactor (length ) 100 ft), the
simulation results in Figure 10 demonstrate how flow rate
can be adjusted to establish a tradeoff between conversion
and thermal liability, thereby using this variable as part of
the operating strategy. For the operating (jacket) temperature
of 182 °C, Figure 10 demonstrates the interrelationship of
conversion with maximum reaction temperature as the flow
rate changes over a (∼20% flow range from 1500 to 2200
mL/min. As might be anticipated for high throughput
operation, temperature excursions are of little concern, but
residence time does not allow sufficient conversion. The flow
rate for this particular reactor configuration must slow to
less than 1700 mL/min before conversion increases signifi-
cantly, but temperature liability rapidly becomes problematic
with small decreases in flow rate. Confronted with this
configuration, the alternatives are: (1) higher jacket tem-
perature with a higher flow rate for maximum productivity
or (2) or lower jacket temperatures with lower flow rates
for greater limitation on temperature liability. In either case,
multiple reactor passes will need to be incorporated to
achieve full conversion.
Commentary
Theoretically, the conversion from the starting aryl
propargyl ether 4 via a Claisen condensation/ring closure
seemed simple: application of sufficient heat to initiate the
reaction with the potential for complete reaction in quantita-
tive yields. The reality of inducing the reaction at a high
temperature accompanied by the rapid onset of heat genera-
tion provided a challenge in developing a scheme to safely
scale up the process in a predictable, controllable, and
consistent manner. The application of known continuous
processing concepts not only reduced safety and quality risks
but provided a means for producing large volumes of material
with a minimal commitment of reactor volume. The positive
results demonstrated by the PFR performance at the “micro”-
and “kilo”-scales and the mathematical model capable of
644
•
Vol. 5, No. 6, 2001 / Organic Process Research & Development

Figure 10. Effect of flow rate on conversion and maximum temperature.
guiding further development efforts led to acquisition of a
large-scale (3/8 in. × 10 ft) jacketed reactor capable of
producing ∼7 kg/h of the dimethylchromene 5.
Conclusions
During process development of the conversion of aryl
propargyl ethers to chromenes, a strategy to control a
potentially hazardous reaction involves (1) maximizing the
primary means of controlling the reaction temperature (heat
transfer area) and (2) thoroughly understanding the sensitivity
of the reaction system to the variables that influence the
reaction performance. Use of a continuous PFR minimizes
the amount of material exposed to reaction conditions at any
given time, further limiting liability in the event of a
processing upset. Encouraging results of the preparation of
the dimethylchromene at “micro”- and “kilo”-scales enabled
and encouraged investment in a large-scale jacketed reactor
capable of producing ∼7 kg/h.
A wide range of parameters can be employed to improve
the control and safety of the operation. The PFR discussion
is specific to the reaction in a solvent-free manner: a worst-
case scenario from the extent of the potential thermal
excursion. However, from a production standpoint, safety
and efficiency might dictate operating the system diluted with
solvent, presumably directly from the process that prepared
the aryl propargyl ether. Solvent removal from the aryl
propargyl ether exposes the material to elevated temperatures
which (under an upset condition) might initiate the thermal
reaction in an uncontrolled manner. Further, the nature of
the aryl propargyl ether increases handling concerns since
the solvent-free material solidifies at 40 °C. Should the neat
material be stored, it would need to be in a cool area and
some means of gentle warming devised for melting the solid
for transfer prior to further processing. For safety and
practicality, the reaction mixture from the preparation of the
aryl propargyl ether can be pumped directly to the PFR and
allows the reaction to proceed at the elevated pressures
generated by the partial pressure of the solvent. The pressure
would not be extreme based on conventional practices in
the chemical industry and the dilution of the aryl propargyl
ether would provide a heat sink, thereby minimizing the
extent of the thermal excursion.
The mathematical model of the process generated during
the course of the development effort not only provides insight
into the performance of the reaction in a given PFR but also
allows development of the knowledge of reaction sensitivity
to operational parameters. Thereby, the model provides a
means for safely developing operational strategies and deter-
mining effects of process adjustments prior to confirmation
by actual operation. The dimensionless nature of the model
allows it to be applied to any PFR handling this reaction by
substitution of the appropriate physical (length, diameter, and
heat-transfer coefficient) and operating (flow rate, inlet
temperature, and heating medium temperature) parameters.
Received for review June 20, 2001.
OP0100504
Vol. 5, No. 6, 2001 / Organic Process Research & Development
•
645