Development of a Safe and High-Throughput Continuous Manufacturing Approach to 4-(2-Hydroxyethyl)thiomorpholine 1,1-Dioxide

Article abstract of DOI:10.1021/acs.oprd.8b00100

Continuous processing enabled the highly energetic double conjugate addition of ethanolamine to divinylsulfone to prepare 2 kg of 4-(2-hydroxyethyl)thiomorpholine 1,1-dioxide, as an intermediate in the synthesis of HIV Maturation Inhibitor BMS-955176. In situ IR was employed to monitor the steady state of the transformation for increased robustness via appearance of the thiomorpholine dioxide moiety and disappearance of the divinylsulfone. Surprisingly, a series of oligomers formed as intermediates, which converted to product with extended aging or heating, consistent with computational predictions. By running this process in flow, the highly exothermic reaction could be safely executed in an equal volume of water as the only solvent, despite an adiabatic temperature rise of 142 °C, leading to a streamlined and efficient process.

Full text of DOI:10.1021/acs.oprd.8b00100

Article
pubs.acs.org/OPRD
Cite This: Org. Process Res. Dev. XXXX, XXX, XXX−XXX
Development of a Safe and High-Throughput Continuous
Manufacturing Approach to 4‑(2-Hydroxyethyl)thiomorpholine
1,1-Dioxide
Neil A. Strotman,*,† Yichen Tan,† Kyle W. Powers,‡ Maxime Soumeillant,† and Simon W. Leung‡
†Chemical and Synthetic Development, Global Product Development, Bristol-Myers Squibb, One Squibb Drive, New Brunswick,
New Jersey 08903, United States
‡Research and Development External Manufacturing, Global Product Development, Bristol-Myers Squibb, One Squibb Drive, New
Brunswick, New Jersey 08903, United States
S
* Supporting Information
ABSTRACT: Continuous processing enabled the highly energetic double conjugate addition of ethanolamine to divinylsulfone
to prepare 2 kg of 4-(2-hydroxyethyl)thiomorpholine 1,1-dioxide, as an intermediate in the synthesis of HIV Maturation
Inhibitor BMS-955176. In situ IR was employed to monitor the steady state of the transformation for increased robustness via
appearance of the thiomorpholine dioxide moiety and disappearance of the divinylsulfone. Surprisingly, a series of oligomers
formed as intermediates, which converted to product with extended aging or heating, consistent with computational predictions.
By running this process in flow, the highly exothermic reaction could be safely executed in an equal volume of water as the only
solvent, despite an adiabatic temperature rise of 142 °C, leading to a streamlined and efficient process.
While this batch process was suitable on small scale and
could be controlled by slow addition of DVS, it did not appear
amenable to large scale production due to excessively long
addition times as the surface area to volume ratio decreased. As
an alternative, a continuous process was investigated, where the
heat generated in the reaction could be readily dissipated due to
the high surface area and heat transfer rate, minimizing the
chance of an uncontrolled reaction, overpressurization, and
thermal degradation or polymerization of DVS. Moreover, a
continuous process would minimize exposure to DVS, a toxic
alkylating agent.6,7
INTRODUCTION
■
The development of HIV MI Inhibitor BMS-955176 required
the preparation of kilogram quantities of side chain 4-(2-
hydroxyethyl)thiomorpholine 1,1-dioxide [CAS: 26475-62-7]
(1) (Scheme 1).1 Synthesis from the commodity feedstocks
Scheme 1. Thiomorpholine Dioxide 1 as a Key Building
Block for BMS-955176
RESULTS AND DISCUSSION
■
A continuous process was initially investigated on a small scale
in a 2 mL glass chip (Uniqsis) containing a static mixer, which
was submerged in a 22 °C water bath. Using two syringe
pumps, a solution of EA in water (6 vol) was combined with
divinyl sulfone (1.04 equiv), with a residence time of 1 min.
The reaction stream was collected into D2O and immediately
divinylsulfone (DVS) and 1,2-ethanolamine (EA) through
double conjugate addition, which had been previously
employed on gram scale, appeared to be the most
straightforward, efficient, and atom economical approach
(Scheme 2).2 However, a process hazard evaluation prior to
implementation on larger scale revealed an exothermic reaction
(exp ΔH = 33.9 kcal/mol) with a large adiabatic temperature
rise of 79.2 °C (Scheme 2), prompting our team to seek out a
safe, scalable method of implementation.3−5
1
analyzed by H NMR, which showed full conversion of the
EA.8,9 Interestingly, although the EA and >95% of the divinyl
sulfone were completely consumed, an unidentified species was
observed at a level of 10−20%. Fortunately, in the reaction
stream, this species completely converted to 1 within 12 h at 20
°C, or within 10 min at 80 °C. Based on the reactivity and
preliminary NMR analysis, it seemed possible that the
unknown species could be an open-chain monomer (2),
dimer (3), or higher order oligomer of 1 (trimer 4, tetramer 5,
etc.), prompting further investigation by LC-MS, NMR, and
computational methods (Scheme 3).
Scheme 2. Batch Preparation of 1 from EA and DVS
Received: April 2, 2018
© XXXX American Chemical Society
A
DOI: 10.1021/acs.oprd.8b00100
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 3. Proposed Intermediates in Conversion of EA and DVS to 1
To study the formation of suspected oligomers and
rearrangement to product 1, DVS (1.05 equiv) was added
over 20 min to a solution of EA in 3 vol water with the internal
temperature maintained below 7 °C to minimize conversion to
1. Once addition was complete, the solution was heated to 20
°C (t = 0 min) and the progress of the reaction was monitored
1
through H NMR analysis of samples (30-fold dilution into
1
D2O). Based on H NMR of the reaction mixture at 2 h, the
structure is most consistent with trimer 4, with an ∼36:3 acyclic
aliphatic/vinyl proton ratio (32:3 theoretical) (Scheme 3).
However, this is not distinguishable from a mixture of both
higher and lower order oligomers, which could lead to the same
ratio.10
LC-MS analysis of the reaction mixture at 2 h showed many
adducts of EA and DVS, either in an n:n or n:n+1 ratio,
favoring the DVS. No peaks were observed corresponding to
structures with greater EA content, possibly due to the excess of
Figure 1. Conversion to 1 vs time post-DVS addition at 20 °C in a vial
(3.28 M) and NMR tube (0.106 M).
DVS used. Table 1 shows the exact masses of the oligomers
Table 1. Oligomers of 1 Observed by LC-MS
exact
mass
RT
exact
mass
RT
vs 3.28 M (3 vol water) in EA proceeded ∼3-fold faster from
∼50−90% conversion. This behavior is consistent with a
competition between an intramolecular reaction leading to 1
and an intermolecular reaction leading to oligomers 2, 3, 4, 5,
etc.12 As expected, the intermolecular process appears slower at
lower concentrations, allowing the intramolecular process to
dominate, leading to overall faster formation of 1 (Figure 1).
This transformation was also investigated computationally to
understand the reversibility of the oligomer formation. DFT
calculations (B3LYP, 6-31G*) revealed that formation of open-
chain monomer 2 from EA and DVS is favorable by ∼4 kcal
and that further conversion to product 1 is favored by an
additional ∼19 kcal (Table 2). Open-chain dimer (3), trimer
(4), tetramer (5), etc. are favored by ∼6 kcal relative to the
starting materials and are viable intermediates. However, the
oligomers are all much higher in energy than 1 (17−18 kcal),
explaining why they convert to product, with sufficient time and
temperature, in an irreversible manner. As experimental support
for these calculations, the reaction enthalpy (EA + DVS to 1)
was predicted computationally as ΔH = −39.9 kcal, which
aligns well with our previously measured value of ΔH = −33.9
kcal/mol.13
oligomer (n:n)
(min) oligomer (n:n+1)
(min)
EA1DVS1 (1)
EA1DVS1 (2)
EA2DVS2 (3)
EA3DVS3 (4)
EA4DVS4 (5)
EA5DVS5
179.1
179.1
358.1
537.2
716.2
895.3
1074.4
0.55
0.65
0.84
1.03
1.43
2.46
2.95
EA1DVS2
EA2DVS3
EA3DVS4
EA4DVS5
EA5DVS6
EA6DVS7
297.1
476.1
0.82
1.86
2.79
3.32
3.77
4.14
655.2
834.3
1013.3
1192.4
EA6DVS6
detected, along with the corresponding LC retention times.11
Additional oligomeric peaks appeared at later retention times,
but due to instrument limitations, m/z > 1200 were not
detected.
While there was no detectable EA present at the end of DVS
addition at <7 °C, the conversion of oligomers of 1 to cyclic
monomer 1 at 20 °C proceeded over the course of a day from
52% to >99% conversion (Figure 1). Additionally, the t = 0
NMR sample taken at the end of the DVS addition, which was
far more dilute, was tracked over time to monitor the
conversion of oligomers to 1 (Figure 1). Interestingly, the
sample that was ∼30-fold more dilute (0.106 M (183 vol water)
B
DOI: 10.1021/acs.oprd.8b00100
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
With a firm understanding of the rapid rate of consumption
of EA and the slower rate of conversion of intermediates (2−5,
etc.) to 1, we were ready to move ahead with development of a
continuous process. As a next step, the reaction of DVS with
EA was conducted on an intermediate scale in a 12 mL
continuous reactor, which produced 150 g of 1 over ∼1 h
(Figure 2). This system was equipped with a DiComp IR flow
cell, capable of monitoring the disappearance of DVS and the
appearance of 1. Strong IR bands at 1390 cm−1 (divinyl
sulfone) and 1195 cm−1 (1) were readily monitored in this
system, despite the use of water as a solvent, which can
suppress IR signals even as a trace component. As shown in
Figure 3, IR monitoring of this continuous process initially
showed no signal when primed with only water and then
showed a peak for divinylsufone shortly after pump B was
started, followed by a peak for 1, along with concomitant
disappearance of divinyl sulfone, after pump A was started. IR
showed that a steady state was established rapidly and
maintained throughout, based on the consistent levels of
DVS and 1. After a 12 h age to promote oligomer conversion,
the aqueous stream of 1 was analyzed by 1H NMR and showed
high purity of 1.
Isolation of 1 was complicated by its extremely high water
solubility (Table 3), leading us to study the impact of solvent,
pH, and salt addition on an extractive workup. Among the
water immiscible solvents examined, solubility of 1 was highest
in DCM, but we also explored EtOAc as a greener alternative.16
Using a solution of 1 in 2.7 vol of aqueous 25% K3PO4,
extraction with 3 × 11 vol DCM resulted in 29%, 8%, and 2%
of 1 remaining in the aqueous layer after each of the
extractions. EtOAc was much less efficient, and extraction
with 3 × 11 vol resulted in 69%, 52%, and 42% of 1 remaining
in the aqueous layer after each of the extractions. Subsequent
experiments with half as much DCM demonstrated that a pH
of 10−11 (achieved by adding K2CO3 to a reaction stream) was
optimal for extraction, and that the addition of KCl to make a
15 wt % aq solution gave ∼10% better recovery per extraction
than without salt.
Table 2. Relative Energies of Intermediates en Route to 1
a
structure
calculated relative energy (ΔG, kcal)
EA + DVS
0
2
1
3
4
5
−4.0
−23.4
−5.6
−5.8
−6.4
a
Energies of oligomers are normalized based upon one EA and one
DVS.
The experimental results demonstrating that intermolecular
reaction of monomer 2 with EA or DVS leading to oligomers is
competitive with intramolecular reaction of 2 to give product 1
were highly surprising. A possible explanation is that the
geometric constraints of the intramolecular process make the
transition state more strained. While conjugate addition to form
open-chain species could be achieved through a six-membered
ring chair transition state involving concerted transfer of the
nitrogen to carbon and proton to oxygen, a similar intra-
molecular reaction involves a highly strained bicyclic [2.2.2]
transition state structure, which would be higher in energy
(Scheme 4).14 While there have been previous reports on
Scheme 4. Proposed Transition State Structures for Intra-
and Intermolecular Reaction of 2
With optimized workup conditions in hand, the end of
reaction stream was concentrated by distilling half of the water,
K2CO3 was charged to adjust the pH, and KCl was charged to
adjust the ionic strength to minimize losses to the aqueous. The
aqueous mixture was extracted with DCM, resulting in ∼12%
loss to the aqueous layer following the third extraction. The
DCM was exchanged for MTBE through distillation, and the
process stream was cooled to 0 °C to crystallize 1 in 74% yield.
The yield and purity were comparable to material made
previously through an otherwise identical small scale batch
mode reaction using slow addition of DVS.
copolymerization of DVS with various amines or diamines, this
is the first evidence for the reversibility of this process, and
preferential monomer formation.15
Figure 2. Equipment design for intermediate scale continuous process.
C
DOI: 10.1021/acs.oprd.8b00100
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 3. Monitoring of continuous process by ReactIR.
Table 3. Solubility of 1 in Water-Immiscible Solvents
suitable for a continuous process where heat exchange with the
cooling medium is very rapid.5 A continuous reaction involving
EA, 3 volumes water, and 1.04 equiv DVS was run on a 150 g
scale and showed comparable behavior by ReactIR and a
solvent
water
solubility of 1 (mg/mL)
>1250
62.3
37.6
18.9
16.8
13.8
13.5
16.8
3.6
a
methyl acetate
EtOAc
1
comparable reaction profile by H NMR.
This process was next transferred to larger equipment for kg-
scale production (Figure 4). To ensure safe, robust operation,
the continuous reactor was equipped with (1) check valves to
ensure back-mixing would not occur, (2) a ReactIR flow cell to
monitor the steady state of the process in real time, (3)
temperature and pressure monitors at several points throughout
the equipment, (4) mechanical pressure relief valves set below
the pressure rating of system, (5) a 30 psi automatic cutoff on
the pumps to avoid overpressurization in the event of a clog,
and (6) a 5 psi backpressure regulator to promote uniform
flow.
n-propyl acetate
isopropyl acetate
n-butyl acetate
isobutyl acetate
2-methyl THF
MTBE
DCM
96.0
71.7
3.4
DCE
toluene
a
Forms a 25 wt % solution in water.
The system was primed with water, before introducing neat
DVS (22 mL/min) through one pump and an aq solution of
EA (51 mL/min) through a second pump, giving a residence
time of 1.6 min in the 110 mL reactor. Once the IR indicated
that steady state had been reached, the flow path was diverted
from waste to a jacketed 20 L reactor for collection, maintained
at 20 °C. While a jacket for the continuous reactor was set at 20
°C, the temperature reached a steady state of ∼40 °C
immediately after the static mixer (∼10 mL volume) due to
However, distillation of water from 6 to 3 volumes was an
energy intensive and time-consuming operation, which was
required to make the extraction into DCM efficient.17
modification was investigated whereby the continuous process
could be run in 3 volumes of water (relative to EA), rather than
the 6 volumes required for the batch reaction, to completely
remove this distillation unit operation. While the estimated
adiabatic temperature rise of 142 °C with this modification was
deemed too high to be safe in batch mode, this would still be
A
Figure 4. Equipment for plant scale production of 1.
D
DOI: 10.1021/acs.oprd.8b00100
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
the exothermicity of this process, although the stream had
returned to 20 °C by the time it exited the continuous reactor.
Over the course of 2 h, DVS and an aqueous EA solution
were pumped through the continuous reactor. The reaction
mixture was collected in a 20 L reactor where it was aged for 12
h at 20 °C prior to workup and isolation, allowing for full
conversion of intermediates to 1. The solution was weighed and
the aqueous phase was extracted twice with DCM (1.40 L).
The combined DCM extracts were concentrated to 700 mL
and cooled to 30 °C. MTBE (2.25 L) was added over 2 h, with
crystallization initiating after ∼550 mL of MTBE had been
introduced. The slurry was cooled to 0 °C over ∼2 h and was
aged for an additional 2 h at 0 °C before collecting the solid by
filtration. The reactor was rinsed with 675 mL of MTBE at 0
°C, and this rinse was used to wash the filter cake. The cake was
air-dried for 1 h and then was further dried in a vacuum
18
1
assayed by H NMR showing a 98.5% in process yield of 1.
K2CO3 and KCl were added to the reaction stream, which was
extracted three times with DCM. After a solvent swap to
MTBE and seeding with 0.5% 1, a seed bed formed and the
reaction mixture was cooled to −10 °C. The crystals were
washed with MTBE and dried to give 2.45 kg of 1 in 71% yield.
This white, crystalline powder had 100% potency and 99.8%
purity by 1H NMR. Although this was a short duration
demonstration, it showed that by using very small equipment
(110 mL continuous reactor) which maximizes heat transfer
and minimizes safety hazards and plant footprint, 1 could be
prepared at a rapid rate of ∼1.6 kg/h, which could support any
future material demands.
1
chamber at 30 °C to yield 212 g of product (77.6% yield). H
NMR (500 MHz, D2O) δ 3.64 (t, J = 5.9 Hz, 2H), 3.20 (d, J =
4.7 Hz, 4H), 3.08 (d, J = 4.7 Hz, 4H), 2.68 (t, J = 5.9 Hz, 2H).
Reaction of Ethanolamine and Divinylsulfone at Low
Temperature and Conversion to 1. EA (400 mg) and water
(1.2 mL) were combined in an 8 mL vial which was cooled in
an ice bath. DVS (690 uL, 812 mg, 1.05 equiv) was added
dropwise so that the temperature in the vial did not exceed 7
°C. After the addition was complete, the vial was transferred to
a bath at 20 °C. Periodically (between 5 min and 32 h) 25 μL
samples were diluted in 750 μL of D2O and were analyzed by
1H NMR within 5 min. The mixture of oligomers that formed
1
CONCLUSION
was most consistent with trimer 4 by H NMR, although LC-
■
MS showed masses for monomer through hexamers. A sample
Our team has developed a safe, scalable, continuous
manufacturing process for 1, which circumvents the problem
of a highly exothermic reaction not amenable to batch scale up.
The highly concentrated reaction conditions allowed for a
rapid, high-throughput process and a streamlined isolation
procedure. Careful monitoring throughout production, includ-
ing ReactIR for reaction progress, and temperature and
pressure to ensure safe operation, were critical to the successful
implementation of this process. Additionally, it was found that
the initial product of EA and DVS is not 1 or even ring-opened
intermediate 2, but is rather a complex mixture of oligomers
which eventually are transformed to 1. NMR, MS, and
computational methods were helpful in understanding the
kinetics and thermodynamics of this fluxional system to ensure
consistent product quality and yield.
taken 2 h after the temperature was increased to 20 °C
1
(mixture with 1): H NMR (500 MHz, D2O) δ 6.84 (dd, J =
16.0, 10.3 Hz, 1H), 6.37 (d, J = 16.5 Hz, 1H), 6.28 (d, J = 9.8
Hz, 1H), 3.62 (m, 6H), 3.45−3.38 (m, 10H), 3.04 (t, J = 6.3
Hz, 8H), 2.98 (t, J = 7.0 Hz, 2H), 2.64 (m, 6H). COSY (500
MHz, D2O) δ 3.62−2.64, 3.42−2.98, 3.42−3.04. See
Supporting Information for tentative assignments and compar-
ison to 1, EA, and DVS spectra. Since these samples could react
on a relatively short time scale, they were not suitable for 13C
NMR analysis. LC-MS (ESI) method (Table 1): Ascentis
Express C18, 2.7 μm, 4.6 mm × 50 mm, 1.5 mL/min, 220 nm,
40 °C, solvent A: 0.05% TFA in MeCN/water (5:95), solvent
B: 0.05% TFA in MeCN/water (95:5), 0 to 10% B at 6 min,
10% to 100% B at 10 min.
Preparation of 4-(2-Hydroxyethyl)thiomorpholine
1,1-Dioxide [CAS: 26475-62-7] (1) on Plant Scale.
Equipment Design. The plant scale continuous flow reaction
was performed with an insulated tubing system composed of 1/
4″ (0.035″ wall) Hastelloy C-276 Swagelok tubing. Each feed
solution was filtered using Upchurch Scientific inlet solvent
filters of 10 μm in porosity and was delivered to the continuous
reactor via an Encynova Model 2−4 high precision liquid
metering pump. Sentry tube-in-tube countercurrent heat
exchangers (1/4″ inner tube diameter; 125 mL heat transfer
volume) provided cooling to each feed stream prior to mixing
as well as the resulting reaction mixture for additional residence
time. The two feed streams combined in a jacketed 1/4″
Hastelloy C-276 Kenics KM static mixer (9.25″, 27 mixing
elements). Cooling was provided using one portable air-cooled
Budzar chiller unit; thus temperature control of the system was
near isothermal. A Mettler Toledo Dicomp flow cell was
installed at the continuous reactor outlet. Flow and pressure
throughout the system was regulated using a downstream
TESCOM 0−25 psi pressure regulator. The temperature was
recorded with Omega Type T thermocouples in four locations;
each stream prior to mixing (2), combined stream after mixing,
and combined stream after residence time (reactor outlet).
Endress and Hauser Promass 83A meters were used to record
the mass flow exiting each feed pump.
EXPERIMENTAL SECTION
■
Preparation of 4-(2-Hydroxyethyl)thiomorpholine
1,1-Dioxide [CAS: 26475-62-7] (1) on Lab Scale. Equip-
ment Design. Lab scale continuous flow reactions were
performed using two HPLC pumps, a 2 mL (3-way, Uniqsis)
glass chip for mixing, and a 10 mL (1/8 ″) Hastelloy coil. The
glass chip and metal coil were submerged in a water batch held
at 20 °C. A Mettler Toledo React IR instrument equipped with
a DiComp in-line flow cell was used for real-time monitoring.
Process. The continuous reactor was primed with water, and
then neat DVS (203.6 g, 173.3 mL, 1.03 equiv) was pumped
through at a rate of 3.0 mL/min. Once the DVS signal was
observed by ReactIR at the exit of the continuous reactor, an
aqueous solution of EA (102.2 g in 303.2 mL of water) was
introduced at 7.0 mL/min giving a ratio of 1.03:1 DVS/EA
(residence time = 1.2 min). Once the DVS signal disappeared
and the product reached a steady state, collection of the process
stream in a reaction vessel was begun and continued for ∼1 h.
The process stream (554.1 g) was transferred to a 5 L reactor
and, after stirring for 16 h, was diluted with DCM (1.40 L).
Aqueous K2CO3 (11 mL, 20 wt % solution) was added to
adjust the aqueous layer above pH 10. Solid KCl (101 g) was
charged to the biphasic mixture which was stirred for 35 min
until fully dissolved. The lower organic phase was reserved, and
E
DOI: 10.1021/acs.oprd.8b00100
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Protack, T.; Lin, Z.; Terry, B.; Samanta, H.; Zhang, S.; Li, Z.; Beno, B.
R.; Huang, X. S.; Rahematpura, H.; Parker, D. D.; Haskell, R.; Jenkins,
S.; Santone, K. S.; Cockett, M. I.; Krystal, M.; Meanwell, N. A.;
Hanumegowda, U.; Dicker, I. B. Identification and Characterization of
BMS-955176, a Second-Generation HIV-1 Maturation Inhibitor with
Improved Potency, Antiviral Spectrum, and Gag Polymorphic
Coverage. ACS Med. Chem. Lett. 2016, 7, 568−572. (b) Ortiz, A.;
Soumeillant, M.; Savage, S.; Strotman, N. A.; Haley, M.; Benkovics, T.;
Nye, T.; Xu, Z.; Tan, Y.; Ayers, S.; Gao, Q.; Kiau, S. Synthesis of HIV-
Maturation Inhibitor BMS-955176 from Betulin by an Enabling
Oxidation Strategy. J. Org. Chem. 2017, 82, 4958−4963.
(2) Ford-Moore, A. H.; Lidstone, A. G.; Waters, W. A. Syntheses of
some β-Chlorinated Amines. J. Chem. Soc. 1946, 0, 819−820.
(3) The Tad was measured by heat flow calorimetry on an RC1 with 6
L of water per kg of ethanolamine.
(4) With a large TAD, there is a risk of a batch reaction mixture
overpressuring or thermal decomposition.
(5) Divinyl sulfone exhibits strongly exothermic decomposition
behavior with an onset of 160 °C based on DSC with a predicted
adiabatic decomposition temperature in 24 (ADT24) of 120 °C. An
advanced reactive system screening tool (ARSST) was used to scan for
thermal decomposition of the reaction mixture, but we did not see any
decomposition up to 150 °C at the end of the scan.
(6) Divinylsulfone has been classified as an acute toxin hazardous
chemical by the European Chemicals Agency: https://echa.europa.eu/
substance-information/-/substanceinfo/100.000.962. Divinylsulfone
has been deemed a hypertoxic chemical in China based on the
following numeric cut-off criteria: Oral LD50 ≤ 5mg/kg bodyweight,
or Dermal LD50 ≤ 50mg/kg bodyweight, or Inhalation LC50 ≤ 0.5
mg/L (vapor): http://www.chinasafety.gov.cn/newpage/Contents/
Channel_5330/2015/0309/247023/content_247023.htm.
(7) Flow can minimize exposure to toxic reagents by limiting the
volume of reactor spills/leaks and avoiding over-pressurization events.
For a review of handling hazardous reagents in flow, see: Movsisyan,
M.; Delbeke, E. I. P.; Berton, J. K. E. T; Battilocchio, C.; Ley, S. V.;
Stevens, C. V. Taming Hazardous Chemistry by Continuous Flow
Technology. Chem. Soc. Rev. 2016, 45, 4892−4928.
(8) When ethanolamine and divinyl sulfone were combined at 1/
40th of the reaction concentration, complete consumption was
observed within 2 min, verifying the rapid rate for this bimolecular
reaction.
Process. The temperature control was set to 10 °C, and the
continuous reactor was flushed with water via both feed pumps
until the process temperature stabilized. Neat DVS was
pumped through the reactor to a waste container via pump B
at a rate of 22 mL/min (9.96 M). Once the IR flow cell
detected the presence of DVS, a 4.10 M EA solution (1496 g of
EA in 4.5 L of water) was pumped via pump A at a rate of 51
mL/min (5% excess of DVS). After the IR flow cell detected
the formation of product, the reaction stream was redirected
from waste to an inerted 20 L glass reactor held at 25 °C. The
flow rates remained constant throughout operation and resulted
in a residence time of 1.6 min. The reaction stream was
collected over ∼1.5 h, aged for 15 h at room temperature to
allow complete conversion to 1, and discharged into drums
(6.90 kg). Analysis of the concentration of product in this
stream by 1H NMR vs an internal standard showed a 98.5% in-
process yield. In a 50 L reactor, the reaction stream was
combined with DCM (24.3 kg) and heated to 30 °C. Aqueous
K2CO3 (178 g, 20 wt % solution) was added to adjust the
aqueous layer to pH 11.3 (target pH > 10). Solid KCl (600 g)
was charged to the biphasic mixture which was stirred for 2 h
until fully dissolved. The lower organic phase was discharged to
a drum, and the aqueous phase was extracted with DCM (2 ×
24.0 kg). The combined DCM extracts were concentrated to
∼10 L through distillation where a KF of 450 ppm was
achieved (target KF < 500 ppm). MTBE (3.90 kg) was added
at a rate of ∼150 g/min followed by product seed (189g, 0.5 wt
%). The remaining MTBE (18.8 kg) was added over 2 h,
followed by cooling from 30 to −10 °C over 1.5 h. The product
was isolated in a Buchner funnel and was washed with MTBE
(6.86 kg). The material was tray dried at 40 °C and ∼100 mbar
over 24 h to give 1 as a white crystalline solid (2.45 kg, 71%
yield). The potency by QNMR was 100%, and the purity by
1
NMR was >99.8%. Mp = 75.2 °C. H NMR (400 MHz,
DMSO-d6) δ 4.51 (t, J = 4.0 Hz, 1H), 3.52−3.47 (m, 2H),
3.07−3.04 (m, 4H), 2.95−2.93 (m, 4H), 2.57 (t, J = 4.0 Hz,
2H). 13C {1H} NMR (100 MHz, DMSO-d6) δ 58.77, 57.73,
50.65, 50.32. HRMS (ESI-MS) calcd for C6H14NO3S: 180.0694
(M + H)+], found 180.0698.
(9) While divinylsulfone and 1 could be observed by HPLC, the
broad peak shapes and inconsistent retention times made NMR a
superior method for analysis.
ASSOCIATED CONTENT
* Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/acs.oprd.8b00100.
■
1
(10) See Supporting Information for interpretation of H NMR
S
(11) Product 1 has a much weaker chromaphore than oligomers of 1
bearing a vinylsulfone moiety, resulting in the low UV response.
(12) The faster conversion to 1 under more dilute conditions could
also be explained by a change in the solvent composition from 50%
water to 99% water, in accordance with a change in the dielectric
constant of the system.
(13) This difference between the experimental and computational
prediction of ΔH could be explained by aqueous solvation which was
not taken into account in the gas phase calculations.
Spectroscopic data and computational details (PDF)
AUTHOR INFORMATION
Corresponding Author
*E-mail: neil.strotman@bms.com.
■
ORCID
(14) Using computations we were not able to identify a transition
state for intramolecular cyclization of 2 to 1.
Neil A. Strotman: 0000-0002-5350-8735
(15) (a) Gao, C.; Yan, D.; Zhu, X.; Huang, W. Preparation of water-
soluble hperbranched poly(sulfone-amine)s by polyaddition to N-
ethylethylenediamine to divinylsulfone. Polymer 2001, 42, 7603−7610.
(b) Gao, C.; Yan, D. Polyaddition of B2 and BB′2 Type Monomers to
A2 Type Monomer. 1. Synthesis of Highly Branched Copoly(sulfone-
amine)s. Macromolecules 2001, 34, 156−161. (c) Yan, D.; Gao, C.
Hyperbranched Polymers Made from A2 and BB′2 Type Monomers.
1. Polyaddition of 1-(2-Aminoethyl)piperazine to Divinyl Sulfone.
Macromolecules 2000, 33, 7693−7699.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank Dr. Susanne Kiau, Dr. Antonio Ramirez, Dr. Robert
Waltermire, Dr. Srinivas Tummala, and Dr. Michael Cassidy for
helpful conversations and support of this work.
REFERENCES
(16) Although the solubility was higher in methyl acetate, it was not
investigated for development due to its high water solubility and
susceptibility to hydrolysis.
■
(1) (a) Regueiro-Ren, A.; Liu, Z.; Chen, Y.; Sin, N.; Sit, S.-Y.;
Swidorski, J. J.; Chen, J.; Venables, B. L.; Zhu, J.; Nowicka-Sans, B.;
F
DOI: 10.1021/acs.oprd.8b00100
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
(17) The reaction was also investigated in alcoholic solvents (MeOH,
IPA, n-BuOH), which gave fast consumption of the ethanolamine, but
much slower conversion of oligomers to 1.
(18) The yield was based on the mass of reaction stream collected
and excluded any ethanolamine stream sent to waste prior to achieving
steady state. For a larger scale, a several hour production run, any
losses during this startup transition would be negligible in the context
of the entire manufacture.
G
DOI: 10.1021/acs.oprd.8b00100
Org. Process Res. Dev. XXXX, XXX, XXX−XXX