Relief vent sizing for a Grignard reaction

Ming-Huei Yue, John J. Sharkey* and Joseph C. Leung?

Merck & Co Inc, 216 E. Lincoln Avenue, PO Box 2000, RYlOl-10, Rahway,
NJ 07065, USA
* Merck & Co Inc, 1 Merck Drive, PO Box 100, WS-EX, Whitehouse Station,
NJ 08889, USA
t Fauske & Associates Inc, 16WO70 West 83rd Street, Burr Ridge, IL 60521,
USA
Received 23 February 1994; revised 23 March I994
The formation of a Grignard reagent is highly exothermic. In addition, the Grignard reagent
reacts violently with many compounds, such as water, to generate heat and possibly result in
the generation of volatile and flammable products. As a result, two potential upset scenarios
during the manufacturing of a Grignard reagent were identified through a hazard and operability
study. Bench-scale studies of these scenarios using the vent-sizing package (VSP) were
employed to evaluate the consequences of these scenarios. Based on the VSP test results,
two additional upset scenarios were identified and further evaluated by the VSP. Sizing of the
emergency relief device for these upset scenarios was then conducted by utilizing pseudo-
one-component tempered system analysis. The recommended emergency relief system was
installed prior to start-up and was found to provide adequate protection after one of the
identified scenarios (runaway reaction) occurred. The details and potential causes of this
incident are also discussed.

Keywords: Grignard reaction; emergency relief venting; runaway reactions

Grignard reagents are widely employed during the with common species, several runaway scenarios and
production of pharmaceuticals and other fine chemi- subsequent explosions or emergency releases may
calsl. A desired intermediate or product is formed by occur during the production and use of a Grignard
the coupling of a Grignard reagent with a variety of reagent. In this paper, several of these potential
compounds. Typically, a magnesium Grignard reagent runaway scenarios and the corresponding emergency
is employed in the absence of water; its formation as relief system design are discussed. In addition, a
shown below is usually highly exothermic2: runaway incident which occurred during the production
of a Grignard reagent will also be addressed.
Initrator
R-X+Mg -R-Mg-X
Initial plant configuration/upset scenario
+heat evolved (-80 kcal mol-l)
definition
where R denotes an alkyl or acyl group, X denotes During the batch production of one of Merck’s
the halides, and Mg denotes the magnesium chips. A products, a Grignard reagent was produced in the
common solvent used for this reaction is tetrahydro- plant and employed immediately to form the desired
furan (THF). intermediate. A schematic illustration of the initial
In general, Grignard reagents can also react equipment employed in the production of this Grignard
readily and exothermically with many common species, reagent is shown in Figure 1. During production, a
such as water, oxygen and carbon dioxidel. The constant nitrogen sweep was employed to minimize
reaction between a Grignard reagent and water, which the potential reaction between the Grignard reagent
is one of the commonly used heating/cooling media, and possible moisture and/or oxygen in the vapour
is shown as follows: above the Grignard batch. The Mg chips and THF
were loaded into the Grignard reactor first. After
R-Mg-X+H,G-+R-H+Mg(OH)X
heating the resultant Mg/THF mixture to the reflux
+heat evolved (20-40 kcal mall’)
temperature of -66”C, a Karl Fischer (KF) test was
In addition to significant heat releases, the reaction conducted to check whether the water content of the
product R-H may be a flammable gas if the alkyl or mixture was within the specification of 0.05%. A
acyl group has only a few carbon atoms. mixture of alkyl halide (R-X) and THF was prepared
Because of the significant heat release during the in a feed tank and checked for a KF less than 0.06%.
production of a Grignard reagent and its reactivity After adding about 0.3 gallons of the initiator,
095c-4230/94:05041~06
fQ 1994 Butterworth-Heinemann Ltd

J. Loss Prev. Process Ind., 1994, Volume 7, Number 5 413

Relief vent sizing for a Grignard reaction: M.-H. Yue et al.

2” Vent Line With can be extrapolated directly to a full-size vessel for

sizing of an emergency relief system. Details on the
Reflux Condenser principles, operation and evaluation of the VSP are
available elsewhere3,4.

Water intrusion at 30 gpm

During the VSP testing, water was injected into
1. Initiator a laboratory-prepared Grignard reagent, which was
2. R-XITHF (7%) produced following the same procedures as in the
plant. The addition resulted in immediate temperature
and pressure increases, which suggested that the
reaction between water and the Grignard reagent was
instantaneous. This observation was similar to that for
other water-reactive chemicals, such as thionyl chloride
> or chlorosulfonic acid, which were found to react
Mg/THF
KF<0.05%
instantaneously with water.
Bnne or Hot Water
From the observed adiabatic temperature rise and
(P
1 Grignarcl Reactor p the concentration of the Grignard reagent, the heat
[
of reaction was estimated to be -20 kcal mol-i of
water. Since the reaction is instantaneous, the batch
MA WP
60 psig temperature rise rate during the intrusion is expected
to be proportional to the water intrusion rate. For a
Figure 1 Equipment configuration before HAZOPNSP studies
water intrusion rate of 30 gpm, the estimated batch
temperature rise rate was -500°C mini. The corres-
ponding required vent size was found to be approxi-
approximately 7% of the overall R-X/THF charge was
mately 14” (diameter).
dropped into the reactor with the agitator off. Initiation
of the Grignard reaction was confirmed by the shift
Cooling oil intrusion at 30 gpm
supervisor observing, through the sight glass, that
Since the vent size estimated from the water intrusion
violent boiling occurred inside the Grignard reactor.
scenario was too large, an oil cooling system with
After initiation, the remaining 93% of the R-X/THF
a hydrocarbon-based heat transfer Buid was also
mixture was charged over the next 2 h and the resultant
considered for the condenser to eliminate the water
reaction mixture was aged for 1 h. During the R-X/
intrusion scenario. Compatibility testing between the
THF addition and depending on the batch temperature,
oil and the Grignard reagent was conducted using the
the jacket service on the Grignard reactor can be
VSP in a fashion similar to the water intrusion scenario.
switched between hot water, none or brine.
Addition of the oil to the Grignard reagent under
Prior to the start-up of a new production schedule,
reflux conditions did not result in any significant
a hazard and operability study (HAZOP) was conduc-
temperature or pressure increase. The resultant mixture
ted on the Grignard production operations. Due
was then heated to higher temperatures and exothermic
to the reactivity hazards associated with Grignard
activity was detected between -105 and 180°C. During
reagents, two potential upset scenarios were identified
the exotherm, the maximum temperature rise rate was
for bench-scale evaluation. These two scenarios were:
-15°C mini, which was significantly smaller than the
l water (brine) intrusion resulting from one condenser value determined for the water intrusion scenario and
tube rupture with an estimated intrusion rate of thus resulted in a much smaller vent size.
30 gallons per min (gpm)
l addition of the R-X/THF solution at twice the Twice the normal R-XITHF addition rate
normal addition rate. During simulation of the fast R-X/THF addition, no
significant heat release (i.e. no Grignard initiation)
Water intrusion from the jacket of the Grignard
was observed. Further investigation of the VSP test
reactor was considered, but eliminated since a vessel
conditions revealed that no Grignard initiation
integrity check showed no corrosion or deterioration
occurred, primarily due to the high water content
of the stainless steel vessel during Grignard production.
(KF) of the solvent THF. To compensate for this
higher water level, the amount of initiator can be
Runaway scenarios/bench-scale studies increased or the THF can be dried to a KF much less
Simulation of the potential upset scenarios identified than -0.05%. By using five times the normal amount
in the HAZOP was conducted in the laboratory of initiator, initiation of the Grignard reaction was
utilizing the vent-sizing package (VSP)3 to provide observed.
data for sizing the emergency relief device to handle During the subsequent simulation of fast R-X/
two-phase discharge flow. The VSP is an apparatus THF addition, only a small heat release was observed,
designed to provide runaway reaction test data that which was not considered typical of a Grignard

414 J. Loss Prev. Process Ind., 1994, Volume 7, Number 5

 Relief vent sizing for a Grignard reaction: M.-H. Yue et al.

formation reaction. The addition was terminated after

charging -41% of the total expected R-X/THF charge.
Shortly after the addition was terminated, the resultant
mixture started to runaway and the sample temperature
increased from -65 to -220°C. This test demonstrated
that a far more serious upset scenario, i.e. a delayed
Grignard initiation leading to accumulation of reac-
tants, can occur during Grignard production. Since
the delayed initiation scenario would be far more
serious than this fast addition scenario, VSP simulation
of the fast addition scenario was not continued.
The phenomenon of the delayed Grignard
initiation is explained in the following way. The Mg
Temperature [OC)
chips were typically covered with an oxide layer in
the initial stage of Grignard reaction. During an -3.L -2.8 -2.6 -2.4 -2.2 -2.0
initiation or induction period, Grignard reagent was -1000/T
formed and reacted with the surface oxide layer to
Figure2 Self-heat rate plots from the upset scenarios of
generate fresh and active Mg surface5s6. In addition, delayed initiation leading to reactant accumulation: (0) 100%
the inhibitors5 such as water also reacted with the R-X accumulated; (+) -41% R-X accumulated
Grignard reagent and thus prolonged the induction/
initiation period. The initiator was present in order to
remove inhibitors and surface Mg oxide to facilitate reaction. In addition, the reflux cooling effect was
the Grignard formation reaction. After the induction ignored during the VSP simulation; therefore, the
period, the rate of Grignard formation has been shown observed self-heat rates from the VSP represented a
in the literature2,5 to be limited by the mass transfer conservative estimate of the reaction rate and were
of R-X to the active Mg surface. used to estimate the vent size for the scenario of
delayed initiation leading to reactant accumulation.
Delayed initiation of the Grignard reaction leading to
reactant accumulation Vent-sizing analysis
A VSP test with 100% of the total R-X charge was The logarithm of vapour pressure versus inverse
conducted to simulate the delayed initiation of the temperature (-1000/T) plots for the water intrusion,
Grignard reaction leading to - 100% R-X accumu- oil intrusion and delayed initiation scenarios, shown
lation. After 100% of total R-X charge was added, in Figure 3, were almost parallel to the saturated
the sample temperature increased from ambient tem- vapour pressure plot of the THF. In addition, the
perature to greater than 300°C at which temperature post-test residual pressures at ambient temperatures
the test cell ruptured due to excessive THF vapour were 4 psig, -2 psig, and 2 psig for the water intrusion,
pressure. In previous testing with -41% R-X accumu- oil intrusion and delayed initiation scenarios with 41%
lation, the sample temperature increased from -65°C R-X/THF charge. The residual pressure for the delayed
to -220°C. initiation with 100% R-X/THF charge was not available
The observed self-heat plots below -220°C from
both delayed initiation tests are shown in Figure 2. In
the temperature region roughly between 100 and
170°C the observed self-heat rates show a small
dependence on the temperature, which is characteristic
of a mass-transfer-limited reaction. For a mass-transfer-
limited reaction, the reaction rate is proportional to
the R-X concentration in the bulk liquid (CR_x), the
mass transfer coefficient (K,), and the available fresh
Mg surface (a):
rate = (K,a)C,.x
However, estimation of the reaction rate or self-heat
rate under the plant upset condition is complex since 10”” ““‘1”“““‘1”“““‘1”‘~
100 150 200
the fresh Mg surface area, the mass transfer coefficient
Temperature (“Cl
and the R-X concentration vary as the batch tempera-
ture and conversion increase. In the VW simulation, -3.0 -2.8 -2.6 -2.4 -2.2 -2.0
the test utilized identical Mg chips, identical THF -1000/T
concentration, similar accumulation of R-X and similar
Figure3 Pressure vs. temperature plots from four VSP upset
agitation effects since the mass transfer was probably scenario simulations: (0) 100% R-X accumulated; (+) -41%
enhanced by the fast rate of the Grignard formation R-X accumulated; (*) water intrusion; (x) oil intrusion

J. Loss Prev. Process Ind., 1994, Volume 7, Number 5 415

Relief vent sizing for a Grignard reaction: M.-H. Yue et al.

due to test cell rupture. Based on these observations, according to the recommendations before start-up. A
the situation during emergency relief can be regarded schematic illustration of the modified configuration is
as a tempered (vapour) system4r7. In addition, the shown in Figure 4. The service fluid in the reflux
main contributor to the observed pressure was the condenser was changed to cooling oil. A safety valve/
vapour pressure of THF. The pseudo-one-component rupture disk combination (PSVIRD) with a set pressure
analysis’s (o-method) was then used in the vent- of 43.5 psig was used to handle standard process upset
sizing calculations and the equations are given in the scenarios, such as excessive feed and high jacket
Appendix _ temperature.
From the tempered system calculations, the relief A second 3” rupture disk with a set pressure of
vent diameter for the water intrusion scenario was 62.3 psig was also installed to handle the upset scenario
found to be -14”. Since the estimated vent size was of delayed Grignard initiation and oil intrusion. During
significantly larger than the largest nozzle on the the emergency release from the delayed initiation
Grignard reactor, this scenario was eliminated by scenario, the expected vent flow would be three-phase,
switching to an oil cooling system. For the oil intrusion including THF vapour, R-X/GrignardffHF liquid mix-
scenario, the vent diameter was estimated to be -1.5”, ture, and Mg chips. This vent flow was different from
which was much smaller than the available nozzle on the homogeneous two-phase flow assumed during the
the Grignard reactor. vent-sizing analysis. Consequently, the benefits of
For the scenario of delayed initiation leading to early release through the 2” condenser vent line fitted
reactant accumulation, the relief vent diameters were with a flame arrester and PSV/RD vent line were
estimated at two conditions, with -41% unreacted conservatively ignored due to their relatively small
R-XITHF and with 100% unreacted R-X/THP. The cross-sectional area and/or high flow resistance.
relief vent diameter estimated in both cases was found
to be -3”. Calculation results for the case of -41% The Grignard release incident
unreacted R-X/THF are listed in Table I.
After all the pre-start-up activities and recommended
physical changes to the Grignard reagent production
Revised factory set-up
facility were completed, the first batch of Grignard
After evaluation of the four different potential process reagent was scheduled to be made. Temperature and
upset scenarios, the plant equipment was modified pressure excursion occurred during this first batch and
a release of -30% of the batch through the vent lines
was observed.
Table 1 Tempered system vent sizing calculation for delayed
The events leading to this incident were found to
initiation scenario with -41% R-XTTHF charge
be as follows. The THF and Mg chips were charged
Plant parameters to the Grignard reactor, heated to reflux, cooled to
Volume of plant vessel (VI 1.1355 m3 25°C and checked for KF. The KF was 0.001% versus
Maximum allowable pressure 0.515 MPa
a maximum specification of 0.05%. The alkyl halide
of plant vessel
Charge to plant vessel (mJ 363 kg
Volume of vessel charge (V,,r,) 0.3528 m3
Bulk density of vessel charge 1029 kg rnds
Discharge pressure (P& 0.101 MPa
Relief device set pressure (P,,) 0.536 MPa
Maximum pressure 0.581 MPa
accumulation 1
(P turn-around
I h
I
VSP test data MA WP
Relief set temperature (T.,,) 128.5% 150 psig
RD RD/PSV
Temperature rise rate at relief 25°C min-’ PM Pset
set temperature 62.3 psig 43.5 psig
w
Turn-around temperature 132.1”C 1. Initiator Cooling Oil
I..-
I ~t”,“-.r,u”dl 2. R-X/THF (7%)
Temoerature rise rate at turn- 30°C min-’ 3. R-X/THF (93%)
around temoerature KF<O.OG%
Experimental 7;7 4 232 650 N m-*
( ‘) T-t
w-method
w evaluated at stagnation 8.75
condition
Critical (choked) pressure ratio 0.845 Mgi-fHF
(nJ KF<0.05%
Mass flux (G) 3700 kg m-2 s-’ Brine or Hot Water
Assumed specific heat CC,) 0.5
Flow reduction factor for an RD
and tailpipe UD = 200 (F)
Ideal vent diameter
0.5

2.1 inch
4 Grignard Reactor

Vent diameter with RD and 3.0 inch MAWP

tailpipe 60 psig

Figure 4 Equipment configuration after HAZOPNSP studies

416 J. Loss Prev. Process Ind., 1994, Volume 7, Number 5


(R-X)iTHF mixture was prepared in a feed tank and
also checked for KF. The KF was 0.09% versus a
maximum specification of 0.06%. About 0.3 gallons
of initiator were added to the Mg/THF mixture and
the resultant mixture was heated to reflux. After
shutting off the agitator and the hot water in the
jacket, about 7% of the R-X/THF was charged to the
Grignard reactor. Initiation of the Grignard formation
reaction was confirmed by noting significant foaming
and boiling in the Grignard reactor and no drop in
the batch temperature. Addition of the remaining
R-X/THF was thus completed over the following 2 h.
At the midpoint of the addition, the reflux rate slowed
to a trickle and hot water was put on the jacket in
accordance with the batch sheet instructions. After
the addition was completed and during the following
ageing period, the batch temperature increased sud-
denly from -75 to -104°C in -4 min, accompanied
by a pressure increase. Significant amounts of batch
contents were then released through the open vent
lines and the emergency relief system. The batch was
eventually cooled to -10°C with brine. During the
release, THF vapour and liquid were discharged
resulting in a possible flammable atmosphere. Any
ignition was prevented by the use of explosion-proof
equipment in the area.
During the release, the PSV/RD popped open
and the separate 3” RD was partially open. The telltale
gauge located close to the flame arrester showed
88 psig. After the release, a significant amount of
unreacted Mg chips was found outside the building
containing the Grignard reactor. In addition, Mg chips
and liquid mixtures were found in the vent line before
the flame arrester and inside the condenser. A large
solid residue was also found inside the Grignard
reactor, which took several days to break-up and
remove. Liquid samples taken out of the reactor,
condenser and vent lines all showed the presence of
the Grignard reagent.
Incident investigation and observations
Several potential causes of the incident were evaluated
to determine the cause and to prevent further occur-
rence of a similar event. The water intrusion scenario
was eliminated due to the use of an oil system in the
condenser and the presence of Grignard reagent after
the incident. Mischarge of the raw materials was ruled
out through material balances. Delayed initiation of
the Grignard reaction leading to accumulation
excess R-X was also ruled out since initiation was
confirmed by the shift supervisor and shift engineer.
Investigation after the incident indicated that the
drums containing the R-X feed contained different
levels of water. Several small-scale laboratory exper-
iments (use test) were then conducted and the results
are listed in Table 2.
These results agree with the literature’ which
indicates that the Grignard reaction initiates well
with small water concentrations, initiates slowly with
moderate water contents, and does not initiate with
J. Loss Prev. Process Ind., 1994, Volume
Relief vent sizing for a Grignard reaction: M.-H. Yue et al.
Table 2 Results of small-scale laboratory tests of R-X
R-X drum(s) KF Observation during use test of R-X
2 and 3 0.09 Grignard reaction did not initiate
7 0.06 Reaction initiated slowly and ran
well
Mixture of 2 and 6 0.04 Reaction initiated slowly, but
stopped halfway
4 to 6 0.03 Reaction initiated and ran well
2 (after drying) 0.02 Reaction required secondary
initiation procedure
excessive water. However, in further testing conducted
on the R-X from drum 2 with high KF, rather
unexpected behaviour was observed. A mixture of
R-X from drums 2 and 6, with a water content about
0.04%, initiated well but stopped roughly halfway
during the addition of the R-X/THF. This suggested
that other impurities in the R-X from drum 2 could
also inhibit the Grignard reaction. From the literature9,
species other than water, such as methanol and ethanol,
were also found to inhibit the Grignard reaction. This
inhibition effect, by unidentified impurities, was further
confirmed by an additional test in which the R-X from
drum 2 was molecular-sieve-dried to a water content
of 0.02%) much less than the specification of 0.06%.
A secondary initiation procedure was required to
initiate the Grignard reaction after the first initiator
was added.
From these findings, the final scenario developed
was that, after initiation, the Grignard reaction stopped
due to the off-specification R-X/THF mixture, which
eventually led to accumulation of R-X. Re-initiation
of the Grignard reaction occurred at a later stage and
caused a significant heat release and pressure increase
as evidenced by the VSP simulation.
Post-incident changes
After a detailed investigation of the Grignard incident,
the most probable cause was found to be a loss of
Grignard initiation due to the presence of excess water
and other unidentified impurities. To prevent future
occurrence of a similar incident, action has been taken
and can be categorized as follows:
Accurate detection of Grignard initiation. Procedures
of were included to provide a better indication of
proper initiation of the Grignard reaction. This
includes accurate metering of a 7% initial R-X/
THF charge, an additional 10 min ageing period
after the 7% charge, and eliminating the jacket hot
water service during this chargelageing period.
Zdentijkation and correction of stopped Grignard
reaction after initiation. Procedure and hardware
changes were employed to record the reflux rate
and reflux temperature during the remaining 93%
R-X!THF addition. If the reflux rate drops signifi-
cantly, the addition must be stopped while several
7, Number 5 417
Relief vent sizing for a Grignard reaction: M.-H. Yue et al.

possible corrective actions, such as quench or re- References

initiation procedures, are evaluated.
Development of quench and secondary initiation
procedures. For a Grignard reaction which fails to
initiate, a secondary initiation procedure can be
employed. For a Grignard reaction which stops
after charging more than -10% R-X/THF or which
fails to re-initiate, a quench procedure can be
employed.
Modification of tfze drying procedure. Modifications
of the existing molecular sieve-drying procedure
were carried out to reduce water content in the
R-X feed.
After these actions were implemented, the remainder
of the schedule ran well without a similar incident. In
addition, no similar incident has occurred to date.
Another method to detect initiation of the Grig-
nard reaction is to start the reaction at about 10°C
less than the reflux temperature and add a small
portion of the R-X/THF charge. A noticeable increase
in the batch temperature is an indication of Grignard
reaction initiation.
1 Rakita, P. E. Chem. Eng. 1990, March, 110
2 Jones, M. C. PIantlOpm Prog. 1989, S(4), 200
3 Fauske, H. K. and Leung, J. C. C&m. Eng. Prog. 1985, 81(S),
39
4 Noronha, J. A. and Fisher, H. G. ‘Emergency Relief System
Design Using DIERS Technology’, AIChE, New York, 1992,
chapter VI
5 Hammerschmidt, W. H. and Richarz, W. Ind. Eng. Chem. Res.
1991, 30(l), 82
6 Hill, C. L., Sande, J. B. V. and Whitesides, G. M. J. Org. Chem.
1980, 45, 1020
7 Leung, J. C. In ‘Safety of Chemical Batch Reactors and Storage
Tanks’ (Eds A. Benuzzi and I. M. Zaldivar), ECSUEEUEAEC,
Brussels/Luxembourg, 1991. p. 285
8 Leung, J. C. and Fisher, H. G. J. Loss Prev. Process Znd. 1989,
2, 78
9 Tuulmets. A. and Heinoja, K. Org. Remtidy 1990, 27(1-2), 42
Appendix

Conclusions
Due to the reactivity of the Grignard reagent, several
runaway scenarios may occur during its production.
HAZOP studies and VSP simulation of the runaway
scenarios provided better understanding of the desired
and undesired process chemistry and resulted in safer
operations. In addition, the pseudo-one-component
tempered system analysis was found to provide a G = PtU - w)(l - 4 - w lwc}Yn
conservative emergency relief system design which 1
provided adequate relief venting when one of the
evaluated upset scenarios occurred.
J--
Pset .T
w(- >
77c
-1 +1

with w estimated from

Acknowledgements
The authors would like to express their thanks to the
following people at Merck & Co. Inc. for their
and PdisiP,,t less than 7)c, which is estimated from:
assistance during the preparation of this manuscript:
Mr R. Colucci, Mr J. Ella, MS L. Meisenbacher, MS vc = 0.55 + 0.217 lno - 0.046 (lnw)”
K. Murray and MS S. Stopya. + 0.004 (lnw)3

418 J. Loss Prev. Process Ind., 1994, Volume 7, Number 5