Preparation of grignard reagents: FTIR and calorimetric investigation for safe scale-up

Article abstract of DOI:10.1021/op9901801

Preparation of Grignard reagents from organic halides and magnesium pose potential safety hazards on scale-up due to their high exothermic potential which can lead to overpressurization, discharge of contents, or explosion. One of the main challenges arises in ensuring the reaction has initiated before excessive accumulation of organic halide occurs or that the reaction does not stall and then reinitiate. Specifically, in production-scale equipment, it is sometimes difficult to ascertain whether initiation has occurred at all and whether it is safe to proceed. By using in situ infrared technology (FTIR), we have developed a method for safer scale-up of Grignard chemistry that can definitively identify that initiation has occurred. The process would involve adding approximately 5% of the organic halide charge and waiting for the initiation to occur using an in situ FTIR probe. FTIR spectroscopy can be used to monitor the accumulation of the halide and reveal when initiation occurs by the resulting decrease in the infrared absorbance. Once it has been determined that the organic halide has reacted as a result of the initiation, it is safe to proceed with the remaining halide charge. The organic halide concentration can then be continuously monitored after initiation to ensure the reaction does not stall or to halt the feed if it does stall. Further, it was shown that IR can be used to quantify the amount of water that is present in THF which is needed to confirm that the THF is dry. The IR results along with reaction calorimetry and ventsizing data are discussed.

Full text of DOI:10.1021/op9901801

Organic Process Research & Development 1999, 3, 319−329
Articles
Preparation of Grignard Reagents: FTIR and Calorimetric Investigation for Safe
Scale-Up
David J. am Ende,1 Pamela J. Clifford, David M. DeAntonis, Charles SantaMaria, and Steven J. Brenek
Pfizer, Inc., Central Research, Process Research and DeVelopment, Eastern Point Road,
Groton, Connecticut 06340, U.S.A.
Abstract:
reinitiated and a runaway reaction ensued resulting in the
release of approximately 30% of the batch through the open
vent lines and emergency relief system.
This paper is intended to provide feasibility for the general
application of IR technology to safely monitor large-scale
Grignard reactions in addition to providing thermodynamic
and kinetic results for the preparation of a Grignard reagent.
Preparation of Grignard reagents from organic halides and
magnesium pose potential safety hazards on scale-up due to
their high exothermic potential which can lead to overpressur-
ization, discharge of contents, or explosion. One of the main
challenges arises in ensuring the reaction has initiated before
excessive accumulation of organic halide occurs or that the
reaction does not stall and then reinitiate. Specifically, in
production-scale equipment, it is sometimes difficult to ascertain
whether initiation has occurred at all and whether it is safe to
proceed. By using in situ infrared technology (FTIR), we have
developed a method for safer scale-up of Grignard chemistry
that can definitively identify that initiation has occurred. The
process would involve adding approximately 5% of the organic
halide charge and waiting for the initiation to occur using an
in situ FTIR probe. FTIR spectroscopy can be used to monitor
the accumulation of the halide and reveal when initiation occurs
by the resulting decrease in the infrared absorbance. Once it
has been determined that the organic halide has reacted as a
result of the initiation, it is safe to proceed with the remaining
halide charge. The organic halide concentration can then be
continuously monitored after initiation to ensure the reaction
does not stall or to halt the feed if it does stall. Further, it was
shown that IR can be used to quantify the amount of water
that is present in THF which is needed to confirm that the THF
is dry. The IR results along with reaction calorimetry and vent-
sizing data are discussed.
Background Chemistry
Grignard reagents are usually prepared by introducing an
organic halide into a stirred suspension of magnesium
turnings in ether (diethyl ether or THF). The reaction occurs
at the surface of the magnesium and for this reason is
considered to be hetereogeneous and has been shown to
involve radical intermediates.3,4 After the Grignard reagent
forms on the reactive magnesium site it dissolves away from
the metal surface into the ether, exposing another fresh site
for reaction. Following the initiation period, mass transfer
of the alkyl halide to the reactive magnesium sites is
considered to be the rate-limiting step.2 The magnesium of
the Grignard reagent is two electron pairs short of an octet
but the oxygen of the ether can donate these electron pairs
to the magnesium of R-Mg-X. This coordination explains
the fact that ether solvents are generally required for the
formation of Grignard reagents because of their ability to
solubilize the Grignard reagent as it is formed. If this did
not occur, the metal would eventually be covered and the
reaction would cease.3 The actual mechanism and specifically
how the magnesium is inserted into the substrate is still not
completely understood.4
Introduction
Grignard reagents are extremely important in organic
synthesis because they provide the chemist a relatively easy
route to carbon-carbon bond formation. On production-
scale, engineers are faced with significant safety challenges.
For example, an incident taken from the literature illustrates
the consequence of excessive accumulation of the alkyl
halide.2 An impurity in one of the feed drums caused the
Grignard reaction to stall at the midpoint of the process.
Alkyl halide continued to be added, at which point the batch
Example: Droloxifene
An example of Grignard chemistry being used on scale
within Pfizer is in the production of droloxifene.5 Droloxifene
is a phase II candidate for the treatment of osteoporosis being
co-developed by Pfizer and Klinge-Pharma. Step 1 of
Scheme 1 refers to the preparation of the Grignard reagent
from a slurry of magnesium chips, THF, and the aryl halide
(3) Louden, G. M. Organic Chemistry; Addison-Wesley: Reading, MA, 1984.
(4) Handbook of Grignard Reagents; Silverman, G. S., Rakita, P., Eds.; Marcel
Dekker: New York, 1996.
(5) Schickaneder, H.; Loser, R.; Grill, H. U.S. Patent 5,047,431, assigned to
Klinge Pharma GmbH & Co., September 10, 1991.
(1) To whom correspondence should be addressed. P.O. Box 8013, Pfizer, Inc,
Groton, CT 06340. E-mail: david_j_amende@groton.pfizer.com.
(2) Yue, Sharkey, Leung, Relief vent-sizing for a Grignard reaction, J. Loss
PreV. Process Ind. 1994, 7, 413-418.
10.1021/op9901801 CCC: $18.00 © 1999 American Chemical Society and The Royal Society of Chemistry
Published on Web 07/16/1999
Vol. 3, No. 5, 1999 / Organic Process Research & Development
•
319

Scheme 1. Droloxifene synthesis
referred to as THP-ether.6 Once the Grignard reagent is
formed it is later reacted or condensed with a substrate FCA-
ketone, 3, shown in step 2, to form a carbinol 4. Finally, the
reaction is quenched to eliminate the magnesium complex
to form aqueous salts of magnesium. Further treatment with
acid leads to elimination and removal of the THP-protecting
group. The desired isomer is isolated with further work up.
runaway reaction can result which could lead to a release or
an explosion due to overpressurization. Before this chemistry
can be scaled up for production, many safety issues need to
be addressed. For example the following are a few primary
concerns:
• Accurate detection of initation is necessary to (1) signal
the operators that it is safe to proceed and (2) prevent
excessive accumulation of organic halide prior to initiation.
• Relief vent sizing for the scale-up equipment must be
adequate to cope with runaway scenarios for the particular
process.
Process Safety
Preparation of Grignard reagents are generally accompa-
nied by an initiation period which is the chief hazard
associated with their manufacture. In addition, the reaction
is sensitive to water and impurities which can result in
delayed initiation periods. If the organic halide concentration
reaches critical levels and then initiates, a serious potential
• The condenser must be adequately sized to handle the
maximum reflux rate.
• Adequate safety interlocks must be in place to prevent
overcharging of either the organic halide or the substrate.
Water intrusion from all sources (solvents and condenser
etc.) must be avoided for two reasons: (1) the reaction will
fail to initiate if the system has too much moisture and (2)
(6) THP-ether is 2-(3-bromophenoxy)tetrahydro-2H-pyran and is prepared from
3-bromophenol and 3,4-dihydro-2H-pyran. Chauez, F.; Godinez, R. Synth.
Commun. 1992, 22, 159-164.
320
•
Vol. 3, No. 5, 1999 / Organic Process Research & Development

Figure 1. Infrared spectra obtained during the preparation of Grignard reagent, a precursor in the synthesis of droloxifene shown
in Scheme 1, step 1.
water leaks after initation can result in a violent hydrolysis
reaction with exothermic energy in the range of 20-40 kcal/
mol.
stretching vibration in the IR as a shift of about 34
wavenumbers from 1590 cm-1 down to 1556 cm-1 as seen
in Figure 3. The formation of Grignard reagent, 2, is observed
without interference of solvent IR bands and can be seen
“growing in” as a peak in Figure 1 at 1556 cm-1. For online
monitoring purposes the starting material and product are
clearly resolved by IR.
Application of in Situ FTIR
Regarding safe process control for Grignard chemistry,
we needed an online method to monitor the concentration
of organic halide to prevent excessive accumulation of
starting material 1. We applied in situ FTIR to the droloxifene
Grignard chemistry shown in Scheme 1 with particular
emphasis on Step 1.
The time-dependent IR profiles measured during and after
initiation are discussed next. During the initial dosing of 1,
wavenumbers 1590 and 1556 cm-1 corresponding to R-X
and R-Mg-X, respectively, were monitored and profiled in
Figure 4. The mass of the aryl halide-THF solution charged
to the reactor is shown on the left axis, while the IR
absorbance for R-X and R-Mg-X is shown on the right
axis. Absorbance for R-X increases immediately with the
actual reactor charge as observed by the parallel profiles.
Initiation occurred after about 15 min as seen by the decrease
in absorbance of R-X at (1590 cm-1) at 0.25 h and the
corresponding increase in Grignard reagent, 2, (1556 cm-1).
Figure 4 is precisely the type of profile that is needed by
the plant engineer to confirm initiation has occurred.
The in situ IR data for R-X and RMgX are profiled for
the entire experiment as shown in Figure 5. After initiation
which occurred near 0.25 h, the remainder of the alkyl halide
was dosed in linearly over 1 h. During this time, the R-X
levels continue to rise as the magnesium is being consumed
and the reaction rate diminishes. Eventually there is no
reactive magnesium left because the alkyl halide was added
in excess. From Figure 5, the absorbance for R-X reached
a maximum near 0.045 relative absorbance at the end of the
addition before leveling off which signified the completion
of the reaction. This also indicated that the dose rate of R-X
was slightly faster than the reaction rate. Because R-X was
added in excess, relative to magnesium, the concentration
of R-X did not return to zero but rather leveled off near
Detection of the Initiation
The Grignard reaction (Scheme 1, Step 1) for droloxifene
was performed in a 1-L reaction calorimeter equipped with
an in situ FTIR probe. Figure 1 shows an IR spectra
“waterfall” plot obtained over the course of the reaction. At
the beginning of the experiment only 5 wt % of the aryl
halide (R-X) in THF solution was charged into the reaction
vessel over 3 min. In this experiment we wanted to monitor
the initiation period in situ after rapidly dosing in a fraction
of the alkyl halide. Typically, Grignard reactions take several
minutes or longer to initiate depending on the water content
of the solvent and other factors such as the condition of the
magnesium surface, impurities in the feed, and agitation rate.
Before discussing the temporal profiles, it is necessary to
review the infrared frequency assignments for the main
reaction components.
The aryl halide referred to as THP-ether, 1 of Scheme
1, is distinguishable by IR from THF due to the ν(CdC)
aromatic stretching absorbance peak which occurs at 1590
cm-1 as seen in Figure 2. THF has a strong absorbing
asymmetric ether ν(C-O) stretch near 1069 cm-1 and a
symmetric ν(C-O) stretch near 913 cm-1. The insertion of
magnesium to form 2 influences the aromatic ν(CdC)
Vol. 3, No. 5, 1999 / Organic Process Research & Development
•
321

Figure 2. IR spectra for R-X (THP-ether) and THF. The aromatic ring of R-X provides distinguishing absorbances due to the
CdC stretch at 1590 and 1475 cm-1. THF has asymmetric and symmetric C-O stretch frequencies at 1069 and 913 cm-1, respectively.
The R-X spectrum is autoscaled for comparison with the THF spectrum.
Figure 3. Aryl halide, R-X, is easily distinguishable from the Grignard reagent R-Mg-X via of shift of 34 wavenumbers due to
the insertion of magnesium and its influence on the aromatic carbon-carbon stretch frequency.
0.035 relative absorbance. On production scale, the dose rate
would be reduced to keep pace with the Grignard formation
and maintain a concentration slightly lower than the 5%
accumulation level (near 0.03 relative absorbance). On line
monitoring by IR also provides confidence that the Grignard
reagent was successfully generated by the increase in
absorbance of wavenumber 1556 cm-1, clearly seen in Figure
5.
The ether, or in the present work THF, plays a role in
the mechanism mentioned above. Namely, the oxygen of the
ether coordinates with magnesium in R-Mg-X and es-
sentially solubilizes the species away from the magnesium
surface. By IR it was evident that THF was involved
“chemically” as the Grignard reagent was being formed due
to the marked decrease in absorbance for the THF band as
seen in Figure 6. The absorbance for the strong asymmetric
C-O stretch at 1069 cm-1 of the THF decreased as a result
of its “consumption” or coordination with the Mg in the
R-Mg-X. This is consistent with the simplified stoichi-
ometry shown in Scheme 2.
As we saw for R-X to R-Mg-X in Figure 3 which has
an IR shift of 34 cm-1 from 1590 to 1556 cm-1 due to the
insertion of magnesium, it appears that a similar shift
occurred for the uncoordinated to the coordinated THF from
1069 to 1034, a shift of 35 cm-1. Recall that wavenumber
1069 cm-1 was assigned the asymmetric C-O stretch for
THF. Similarly, the symmetric C-O stretch of THF is visible
at 911 cm-1 with an analagous shift of 35 cm-1 to 876 cm-1
as seen in Figure 6. The shifts are visible as shoulder peaks
growing as seen in Figure 6. The shoulder peaks at 1034
322
•
Vol. 3, No. 5, 1999 / Organic Process Research & Development

Figure 4. IR monitoring during the initial phase of the reaction. Left axis is the reaction mass during the addition of the R-X/
THF feed solution into the reactor. R-X solution (5 wt %) was added over 3 min to monitor the accumulation and initiation by IR.
The relative IR absorbance values are shown on the right axis for R-X and R-Mg-X at 1590 and 1556 cm-1, respectively.
Figure 5. IR and reaction mass flow profiles for the duration of the Grignard preparation experiment. R-X was added in excess
and reaches a maximum at the end of the dose. Because the magnesium is completely reacted, some R-X remains at the end of the
experiment.
and 876 cm-1 are the new C-O stretch frequencies for the
bound THF whose oxygen is coordinated with the Mg of
the R-Mg-X.
cm-1 with the same kinetic profile as seen with the formation
of Grignard reagent. Figure 7 provides further evidence that
THF is chemically involved in the formation of the Grignard
reagent and that this in situ IR method is able to distinguish
and monitor its kinetics.
As clarification, the FTIR probe used in our study “sees”
only what is in solution. Solid magnesium turnings stirred
in THF are not visible by IR until the Mg becomes dissolved
and incorporated as Grignard reagent. Thus, initially, before
any organic halide is added, the IR sees 100% THF. As the
organic halide-THF solution is charged to the reactor, the
concentration of the THF decreases, and thus the IR
absorbance decreases. If no reaction was occurring, a linear
decrease in the THF absorbance would be expected. Figure
7 shows IR profiles for the bound and unbound THF.
Unbound THF is seen to decrease in IR absorbance at 1069
cm-1 in a nonlinear fashion as it coordinates with magnesium
and becomes bound. The bound THF is seen to form at 1034
Water in THF
The induction time for initiation of the Grignard reagent
is profoundly affected by water present in the solvent.
Excessive water levels, can tremendously increase time of
initiation or prevent it completely. For this reason careful
control of water content is essential prior to start of the
reaction. Calibration standards of THF with different levels
of water were prepared and tested by IR. We sought a method
that could be used as in situ technique in place of the current
method of pulling samples and performing Karl Fischer
Vol. 3, No. 5, 1999 / Organic Process Research & Development
•
323

Figure 6. THF plays a role in the mechanism by coordinating with the magnesium of R-Mg-X. This coordination enables the
R-Mg-X to be solubilized by the THF and removed from the magnesium surface. The IR spectra is consistent with uncoordinated
THF having a C-O stretch at 1069 cm-1 being shifted downward by 35 cm-1 as THF becomes coordinated to the Grignard reagent.
Evidently the influence of magnesium on the C-O stretch causes this shift in absorbance.
Scheme 2. THF coordinates with the magnesium of the
Grignard reagenta
a The coordinated THF is distinguishable by IR from the unbound THF due
to the effect of Mg on the C-O stretch frequency of the ether.
testing. Figure 8 shows overlayed spectra of varying water
concentrations in THF. Water is distinguishable from THF
at wavenumbers 3500 and 1648 cm-1. The sensitivity of this
initial testing was good enough to accurately measure below
0.05% water which is typically the dryness specification we
Figure 7. Bound vs unbound THF. THF coordinated to Mg
use for scale-ups of Grignard reactions. A quantitative
has C-O stretch frequency at 1034 cm-1, while unbound or
method which resulted in a linear calibration curve was
generated on the basis of the spectral data between 0 and
2% water which provided accurate quantitation to within
(0.015 absolute error. Higher sensitivities with the IR probe
can be obtained by increasing the number of scans taken
per spectrum and thus a further reduction in the absolute
error is feasible. The spectra shown in Figure 8 were acquired
using 256 scans with a resolution of 8 cm-1 per spectrum.
Scan numbers as high as 1024 are not unreasonable to
increase the signal-to-noise by a factor of 2.
For process control purposes, concentration of 1 would
be monitored during the entire step 1 process and not just
for the initiation period. Further, IR technology has reached
a stage where it could be used as part of a process control
strategy.7 For instance, in the early stage of the reaction this
technology would clearly identify that initiation has occurred
free THF has a C-O stretch frequency at 1069 cm-1. The
relative IR absorbance for the THF during the formation of
the Grignard reagent is shown on the left axis. The reaction
mass profile is shown on the right axis. After the initiation
period, the aryl bromide was added to the reactor linearly over
a period of 1 h.
and that it is safe to proceed with further charging of organic
halide. In the middle and latter stage of reaction IR
monitoring could be used to provide an output signal for an
alarm or to automatically deactivate a feed valve in the event
the reaction began to stall and the concentration of R-X
began to rise unexpectedly. Had this technology been
available as an online monitoring strategy, it would have
likely prevented the release incident reported by Yue et al.2
In general, the IR technology is a valuable tool for
chemical process development and we have recently installed
a process unit into a plant environment. Several issues that
(7) Mettler-Toledo, Inc, Millersville, MD 21108, manufactures the RC-1 reaction
calorimeter.
324
•
Vol. 3, No. 5, 1999 / Organic Process Research & Development

Figure 8. Water content in THF as measured by IR. Water is distinguishable from THF at wavenumbers 500 and 1648 cm-1. A
linear quantitative method is obtained by using peak height to single point baseline method. The method produced an absolute
error of 0.015.
Figure 9. Heat flow calorimetry data. Left axis refers to the temperature and heat flow associated with the Grignard reaction. The
right axis refers to the dose of the R-X/THF feed solution. The heat flow due to initiation is visible at 0.3 h.
we considered for implementing this technology into the
plant are the following:
Experimental Section
Instrumentation. Experiments were performed using a
Mettler Toledo RC-1 reaction calorimeter7 consisting of a
1-L conical shaped reactor (SV01) with a glass agitator and
Brinkmann-style reflux condenser. The reactor cover and
condenser were insulated with glass wool. FTIR data were
collected in situ using Applied Systems ReactIR 1000
system8 with DiComp probe during the calorimetry experi-
ments. A VSP9 vent-sizing package was used for the vent-
sizing experiment.
(1) A “light pipe” is required for transmission of the IR
signal from the source to the detector. The length of the light
pipe is currently limited to about 4 feet which constrains
the location of the IR detector assembly to be placed in close
proximity to the sampling probe.
(2) Since moisture within the light pipe can interfere with
the signal, the light pipe should be continuously purged with
a dry source of nitrogen or air.
(3) Other utilities needed for plant operation are liquid
nitrogen for the high sensitivity (MCT) detector, cooling
water to maintain constant temperature of the IR processor,
and, of course, electrical utilities.
Procedure. A weighed amount of magnesium was stirred
into a known volume of THF. THF was used new as supplied
by JT Baker. Karl Fischer analyses were performed to
determine moisture content in the THF. The magnesium-
In general, the potential benefits of this in situ IR
technology in terms of safety and productivity, in our view,
clearly outweigh the initial challenges of implementing these
units into the production facility.
(8) ASI Applied Systems, a Mettler-Toledo Company, Millersville, MD, 21108
(manufactures the ReactIR and Dicomp probe).
(9) Fauske and Associates of Burr Ridge, IL. The VSP2 is based on DIERS
(Design Institute for Emergency Relief Systems) research results.
Vol. 3, No. 5, 1999 / Organic Process Research & Development
•
325

Figure 10. Heatflow calorimetry obtained for the preparation of Grignard reagent and subsequent coupling with FCAsketone of
Scheme 1. The dose of R-X/THF was uninterrupted. The reaction initiated after 0.33 h (20 mins) with 30 wt % of the R-X added.
An extremely high heat flow (550 W/L) was generated due to this excessive accumulation of R-X. The coupling with FCA-ketone
was much less exothermic with heat never near 50 °C.
THF slurry was brought to reflux (about 67 °C) in a 1-L
reactor. A weighed amount of the alkyl halide (THP-ether)
was dissolved in THF and placed into a dosing bottle which
was positioned on a balance. After calibrations and heat
capacity measurements were performed for calorimetry, the
first 5 wt % of the alkyl halide-THF solution was dosed in
linearly from a balance over a 3-min time interval via a
control loop from the RC-1. IR scans were acquired every
minute for the first 27 scans and then every 3 min thereafter.
IR spectra were collected using 64 scans with a resolution
of 8 cm-1. The remainder of the alkyl halide solution was
added over 1 h.
flow rate and where ∆Treflux ) Tout - Tin. Tout and Tin refer
to the temperature of the condenser coolant outlet and inlet,
respectively. (MCp)reflux is the mass and heat capacity of the
condenser cooling fluid.
Ethylene glycol was used as the coolant through the reflux
condenser and ∆Treflux was measured by the calorimeter via
inlet and outlet coolant thermocouples. To compute
(MCp)reflux we used a 20-W calibration heater at reflux to
measure the condenser response i.e., ∆Treflux. By integrating
the ∆Treflux curve in response to the 20-W calibration heater,
10
we were able to calculate (MCp)reflux
.
Because the mass
flow rate of the coolant fluid was steady through the
condenser, dm/dt is a constant flow rate designated as M,
thus:
Heat Flow Calorimetry Theory
Heat flow calorimetry was performed under reflux condi-
tions using an energy balance around the reactor and reflux
condenser. The energy balance is made up of the following
terms:
(MCp)reflux
)
Q dt/ (∆T - ∆TBL) dt ≡ J/s-K ≡ W/K (2)
∫
∫
c
c
where Qc is the known calibration heater wattage, ∆Tc is
the condenser temperature response, and ∆TBL is the baseline
condenser response when the calibration heater is off.
Alternatively, one could use an accurate measure of the flow
rate and the condenser coolants heat capacity to calculate
Qr ) Qflow + Qaccum + Qdose + Qreflux
where:
Qr ) heat of reaction or other thermal event with units
of Watts or J/s
(MCp)reflux
.
Given the collective terms of the energy balance, the heat
of reaction can be computed by integrating over the time
interval of the thermal event, namely:
Qflow ) heat flow through the vessel jacket ) UoA(Tr -
Ta) where Ta is the average jacket temperature, Uo is the
overall heat transfer coefficient, W/m2-K, and A is the wetted
heat transfer area of the vessel.
Qaccum ) heat flow accumulation of the reaction mass and
the inserts ) (mrCpr + MiCpi)dTr/dt where mCp refer to the
mass and heat capacity of the reaction mass and inserts,
respectively.
Qdose ) heat flow due to sensible heat of the feed )
dm/dt Cpd (Tr - Tdose) where dm/dt and Cpd refer to the mass
flow rate and heat capacity of the material being dosed to
the reactor, respectively.
∆Hreaction
)
Q dt ≡ J
(3)
∫
r
Heat Flow Calorimetry Results and Discussion
The heat flow profile for step 1 Scheme 1 is illustrated
in Figure 9 and was obtained in the same experiment that
produced the IR data shown in Figure 5. The dosing of R-X
was carried out in two stages with the first being the initial
Qreflux ) heat removed by the reflux condenser )
(dm/dt Cp)reflux ∆Treflux ) (MCp)reflux ∆Treflux for condenser
(10) Lerena, P., Reaction Calorimetry at Reflux, 8th RC-1 users forum, Lugano,
Switzerland, 1997.
326
•
Vol. 3, No. 5, 1999 / Organic Process Research & Development

Table 1. Summary of reaction calorimetry data
heat capacity of
reaction mass (J/kg-K)
maximum
heat flow
heat released
(kcal/mol)
∆Tadiabatic
(°C)
operation
preparation of Grignard
1. dose in 5% THP-ether
2. dose in remaining THP-ether
preparation of Grignard (uninterrupted feed)
coupling with FCA-Ketone
quench
1987
1836
1635
1490
140 @ reflux
240 @ reflux
550 @ reflux
50 @ reflux
89 (of R-X)
87 (of Mg)
83 (of Mg)
41 (of FCA)
30 (of FCA)
22
154
150
43
1700-2500
600 (of short duration)
15-20
Figure 11. Heat flow calorimetry for the quenching of the Grignard reaction with a 25% solution of ammonium chloride. The
quench occurred after coupling with the FCA-ketone andthe temperature was lowered to 25 °C. The adiabatic temperature rise
for the initial part of the dose was approximately 20 °C above the process temperature.
Table 2. Vent-sizing parameters
5 wt % to induce initiation followed by the second stage to
add the remaining 95 wt %.
vessel parameters
Stage 1. The reactor contained 16.51 g (0.679 mols) of
magnesium turnings and 267.8 g of THF. After calibrations
were made, the temperature was raised to reflux. THP-ether
(5%) in THF (27 g) was fed to the reactor over a period of
3 min. The resulting heat liberated as measured by the energy
balance was -373 kJ/mol or -89 kcal/mol of R-X.11 The
maximum heat flow during this stage as seen in the graph
was about 45 W or 140 W/L removed by the reflux
condenser.
set pressure, Ps
vessel rating, MAWP
60 psig
100 psig maximum allowable
working pressure
500 gal (67 ft3)
3155 lb
vessel volume, V
charge, mo
specific heat, Cp
vapor density, Fg
liquid density, Ff
0.5 BTU/lb °F
0.7 lb/ft3
50 lb/ft3
Stage 2. Once initiation was confirmed, the remainder
of the R-X was added (510.7 g of solution) over 1 h. R-X
was added in stoichiometric excess of the magnesium. The
heat measured for stage 2 was -363 kJ/mol or -87 kcal/
mol of magnesium.12 The adiabatic temperature rise for the
combined events was computed to be 167 °C based on a
total of -246.8 kJ, a reaction mass of 1.133 kg, and a heat
capacity of 1836 J/kgK, and using the relation:
magnesium slurry or the feed solution. For the purpose of
this experiment, 18 volumes of THF were used per gram of
magnesium in the initial fill. For instance, if half the volume
of THF were used in the initial fill, the new ∆Tadiabatic
)
199 °C which is about 30 °C higher than the more dilute
case. During stage 2 the maximum heat flow was measured
to be about 80 W or 240 W/L of reaction being removed
through the condenser.
In a separate experiment13 the Grignard preparation was
performed with constant dosing, without pausing for the
initiation to occur. The heat flow profile is shown in Figure
10. In this case the R-X/THF solution was continuously
dosed over a 1-h period. The initiation occurred after about
20 min and approximately 30% accumulation of THP-ether.
The heat flow due to initiation at this point is extremely high,
measured at reflux at 550 W/L as seen in Figure 9. Clearly,
this is not the recommended approach to take on scale-up
due to the potential for overpressurization from rapid
∆Tadiabatic ) ∆H/mCp
(4)
Because ∆Tadiabatic depends on the mass, the value will
be correspondingly higher if less THF is used in the
(11) At this point, 0.037 mol of THP-ether had been added, and the heat
measured was -13.8 kJ.
(12) During stage 2, there was 0.679 - 0.037 ) 0.642 mol of magnesium left
to react, and the heat measured was -233 kJ.
(13) Magnesium (10 g, 0.41 mol) was used in 25 volumes of THF. 383 g of
R-X/THF (0.48 mol THP-ether) solution was added. FCA-ketone (100
g, 0.32 mol) dissolved in 220 mL of THF was used for coupling.
Vol. 3, No. 5, 1999 / Organic Process Research & Development
•
327

Figure 12. VSP results for 500-gal vessel with rupture disk setting of 60 psig. These data represent the worst case scenario of an
all-at-once addition of R-X/THF solution to a slurry of magnesium chips in THF with homogeneous vessel venting. The existing
6-in. vent will result in a slight but acceptable increase in pressure to about 74 psig for this scenario. The maximum allowable
working pressure for the vessel is 100 psig.
vaporization of solvent. The total heat of reaction for this
experiment was -83 kcal/mol of Mg with an adiabatic
temperature rise, ∆Tadiabatic, of 150 °C. The heat measured
during the coupling reaction with FCA-ketone was shown
to be -41 kcal/mol of FCA ketone (Figure 10). Following
the coupling, the reaction temperature was reduced to 25
°C, and a 25 wt % solution of ammonium chloride was added
as a quench (shown in Figure 11). The quench resulted in
an extremely high heat flow for the first fraction of the
quench with values in excess of 500 W/L. This heat is no
longer being removed by vaporization of solvent but rather
by the cooling jacket of the vessel. Because, the RC-1 is
not able to remove that rate of heat production, the
temperature in the reactor increased about 13 °C. The high
exotherm was short-lived. The calculated adiabatic temper-
ature rise for the quench was 15-20 °C depending on how
much of the quench solution was dosed. The calorimetry
data are summarized in Table 1 for all the experiments
described.
and 12.02 g of THF (8 volumes). The cell was equipped
with a standard thermocouple whose tip was 1/2 in. from the
base of the cell and with dual /16 in. o.d. fill lines. The
1
sample was heated to 64 °C. At 21.3 min the THP-ether/
THF solution was charged to the test cell via the auxiliary
fill line. The test cell was closed. The dosed solution was
made up of 16.48 g of THP-ether and 30.57 g of THF
solvent. After the solution was added to the test cell, the
temperature dropped to 38.6 °C, but the temperature then
immediately rose due to the exotherm to 172 °C. Thus, the
pseudo-adiabatic temperature rise for this experiment was
(172 °C - 38.6 °C) ) 133 °C. The pressure rose to 145
psig by 28.96 min and then slowly continued to rise to 170
psig.
DIERS calculations were performed for the scenario in a
500-gal reactor used to prepare the Grignard reagent. The
DIERS two-phase venting technology was used which
assumes a homogeneous-vessel venting behavior. The cal-
culations were performed by Fauske and Associates using
the input data shown in Table 2.
Vent-Sizing
The output for the DIERS homogeneous-vessel venting
prediction is graphically illustrated in Figure 12. The graph
shows the accumulated pressure for a 500-gal plant vessel
under the worst case scenario of a 100% accumulation of
R-X. In this worst case scenario a sudden and complete
charging of the THP-ether solution is made to the slurry of
magnesium turnings. The existing vent on this vessel is 6-in.
i.d. with a rupture disk pressure of 60 psig. The calculations
presented in Figure 12 show that if a runaway occurred, the
rupture disk would release at 60 psig but the pressure would
continue to rise another 14 psig to 74 psig which is acceptable
and well below the maximum allowable working pressure
of 100 psig. The data shows that a 4.4 in. diameter vent-
size vent would be the minimum vent-size to cope with this
worst case scenario that would not exceed the 100 psig
MAWP of the reactor. Thus the 6-in. vent currently in place
would be more than acceptable to cope with this runaway
Prior to scale-up to the plant, vent-sizing experiments were
performed to assess the consequences of worst case scenarios
in order to adequately design the relief system for the plant
vessels.
Two vent-sizing experiments were performed using a
VSP2 Vent Sizing Package: one for the preparation of the
Grignard reagent and the second for the coupling of the
Grignard reagent with FCA-ketone. The first experiment
was for a worst case scenario of an all-at-once addition of
the alkyl halide to the magnesium-THF slurry, and the
second experiment was for all at once addition of the FCA-
ketone. These tests were undertaken to indicate whether
existing rupture disks were adequate to cope with these
scenarios.
Test 1. The stainless steel test cell containing a large stir
bar was filled with 1.47 g (0.06 mol) of magnesium turnings
328
•
Vol. 3, No. 5, 1999 / Organic Process Research & Development

scenario without exceeding the 100 psig allowable pressure
in the tank. In summary, the vent-sizing experiments and
calculations are essential for the scale-up engineers to ensure
proper mitigation in the event of a process upset or
overpressurization due to a runaway reaction.
Online monitoring by IR can be used not only to ensure that
initiation has been achieved but also to identify an accumula-
tion of organic halide in the event the reaction stalled as
was cited in the events that led to the incident reported by
Yue et al.2 Infrared technology currently being used in our
Process Safety and Reaction Engineering Laboratory was
utilized for the initial feasibility studies. As part of this work,
it was also identified that this technology could be employed
to quantify water levels for these water-sensitive reactions
which could further reduce time and costs of current sampling
and analytical testing via K-F methods. The infrared
monitoring results described here were obtained at the R&D
stage and have not been employed on scale to date. However,
an FTIR unit was recently installed in our pilot plant
facilities, in part, on the basis of the success of this work.
Test 2. Test 2 involved the coupling reaction between
FCA-ketone and the Grignard reagent. Grignard reagent was
prepared from a solution containing 1.07 g of magnesium,
11.87 g of THP-ether, and 30.44 g of THF. The Grignard
reagent (40.2 g) was introduced to a stainless steel closed
test cell with dual 1/16 in. o.d. fill lines and 1-in. stir bar and
thermocouple. The sample cell was heated to 66 °C. Then
at 15.67 min 27.46 g of the ambient solution of 34% FCA-
ketone/THF was added which resulted in an immediate
exotherm up to 84 °C with an accompanied pressure rise of
14.5 psig. Unlike test 1, the consequence of this scenario is
not considered a significant hazard because of the relatively
low resulting pressure rise.
Acknowledgment
The authors acknowledge Joel Hawkins of Pfizer’s
Process Research for his helpful discussion and Dave
Johnson of Applied Systems for his technical support.
Conclusions
Because of the runaway reaction and release potential of
Grignard reactions we sought a general method to monitor
the concentration of organic halide for scale-up consider-
ations to increase safety and reduce the risk of incident.
Received for review November 23, 1998.
OP9901801
Vol. 3, No. 5, 1999 / Organic Process Research & Development
•
329