Case studies in process screening and optimization utilizing a new automated reactor - Analysis system

Article abstract of DOI:10.1021/op0001077

The application of a new automated workstation for conducting solution-phase reactions in parallel during chemical process development is described. The fully automated reactor - analysis system is capable of reagent/reactant addition, temperature and agitation control, and sample extraction/quench/injection for online HPLC analysis. Both precision and accuracy of the fluidic delivery and sampling systems were measured through the use of standard ultraviolet chromophore solutions in conjunction with HPLC analysis. Agitation efficiency was evaluated using a phase-transfer catalysis reaction known to be rate-dependent on stirring frequency. Chemistry cases examined included several single and multistep reactions studied to identify the best reagent/solvent combinations (screening) and optimal levels of continuous variables such as temperature, concentration, and stoichiometry (optimization). A multistep route to a chiral substituted imidazolidin-2-one was developed from initial route scouting through reaction condition screening, optimization, and scale-up. Reactions performed included alkylation, reduction, oxidation, reductive amination, displacement, and cyclization.

Full text of DOI:10.1021/op0001077

Organic Process Research & Development 2001, 5, 283−293
Case Studies in Process Screening and Optimization Utilizing a New Automated
Reactor-Analysis System
Owen W. Gooding,* Thomas Lindberg, Wendy Miller, Edward Munyak, and Lanchi Vo
Argonaut Technologies, 887 Industrial Road, Suite G, San Carlos, California, 94070, U.S.A.
Abstract:
medicinal chemists now have multiple choices for parallel
synthesis equipment ranging from simple manual reactor
blocks to fully automated platforms.2 More recently, synthesis
instruments designed specifically for the activities of process
chemists have become commercially available.3
The application of a new automated workstation for conducting
solution-phase reactions in parallel during chemical process
development is described. The fully automated reactor-analysis
system is capable of reagent/reactant addition, temperature and
agitation control, and sample extraction/quench/injection for
online HPLC analysis. Both precision and accuracy of the fluidic
delivery and sampling systems were measured through the use
of standard ultraviolet chromophore solutions in conjunction
with HPLC analysis. Agitation efficiency was evaluated using
a phase-transfer catalysis reaction known to be rate-dependent
on stirring frequency. Chemistry cases examined included
several single and multistep reactions studied to identify the
best reagent/solvent combinations (screening) and optimal levels
of continuous variables such as temperature, concentration, and
stoichiometry (optimization). A multistep route to a chiral
substituted imidazolidin-2-one was developed from initial route
scouting through reaction condition screening, optimization, and
scale-up. Reactions performed included alkylation, reduction,
oxidation, reductive amination, displacement, and cyclization.
The development of new chemical processes can be
divided into four phases: route scouting, screening, optimi-
zation, and validation. In Route Scouting completely different
pathways leading to a target structure are investigated. In
Process Screening the general route has been established and
reaction parameters, that is, temperature, reagent, solvent,
and so forth, are examined over broad ranges to determine
the most important variables and most effective reagents.
Once the reactions are completely defined from these studies,
Process Optimization is initiated to thoroughly study all
parameters, including reproducibility, workup technique, and
product isolation. Last, in Process Validation the reactions
are tested at larger scale to mimic plant conditions and further
define process ranges and limits. We have developed an
automated workstation (Surveyor) specifically designed for
conducting process screening and optimization during chemi-
cal process development. The fully automated reactor-
analysis system is capable of reagent/reactant addition,
temperature and agitation control, and sample extraction/
quench/injection for online HPLC analysis. Table 1 sum-
marizes the main instrument features and capabilities. This
paper describes the system tests and chemistry case studies
in process screening and optimization conducted during the
course of instrument development and validation.
Introduction
Chemical development in the pharmaceutical industry can
be defined to include everything between initial compound
discovery and routine batch or continuous production at full
manufacturing scale. As a part of chemical development, a
synthetic route used by discovery chemists may be developed
and optimized, or completely new routes, taken from
literature precedent (or new invention), may be investigated
for applicability. As there is usually immediate demand for
material, both of these activities often occur simultaneously.
Historically, the race to discover the most facile, economic,
and environmentally acceptable method before regulatory
deadlines expire has placed extreme pressure on the develop-
ment chemist. This has become particularly acute in the
pharmaceutical arena where combinatorial chemistry, parallel
synthesis, and high-throughput screening have accelerated
the promotion of compounds into clinical development.
Automation has played a key role in streamlining many
routine laboratory operations. Chemists have benefited from
tools such as autosamplers coupled with analytical instru-
mentation, robotic liquid handlers, and automated reactor
devices for synthesis.1 Combinatorial chemistry has been the
main driving force for the development of high-throughput
synthesis technologies in drug discovery. As a result,
Results and Discussion
Reagent Delivery and Sampling. Precise and accurate
delivery of starting materials and reagents is of obvious
importance in any reaction, automated or otherwise. During
the course of instrument development, delivery and sampling
were tested and validated prior to conducting chemical
reactions. Both precision and accuracy of the fluidic delivery
and reactor sampling systems were measured through the
use of standard solutions of biphenyl and anthracene in
conjunction with HPLC analysis. To test the delivery system
in isolation, a standard biphenyl solution was placed in a
Surveyor reagent position, and 100 µL aliquots were
(2) Hird, N. W. Drug DiscoVery Today 1999, 4, 265-274.
(3) (a) Wagner, R. W.; Li, F., Du, H.; Lindsey, J. S. Org. Process Res. DeV.
1999, 3, 28-37. (b) Armitage, M. A.; Smith, G. E.; Veal, K. T. Org. Process
Res. DeV.. 1999, 3, 189-195. (c) Emiabata-Smith, D. F.; Crookes, D. L.;
Owen, M. R. Org. Process Res. DeV. 1999, 3, 281-288. (d) Gooding, O.
W. Presented in part at the 3rd International Symposium, The Evolution of
a Revolution: Laboratory Automation in Process Research & Development,
Boston, 2000.
(1) For a very recent review, see: Harre, M.; Tilstam, U.; Weinmann, H. Org.
Process Res. DeV. 1999, 3, 304-318.
10.1021/op0001077 CCC: $20.00 © 2001 American Chemical Society and The Royal Society of Chemistry
Published on Web 03/13/2001
Vol. 5, No. 3, 2001 / Organic Process Research & Development
•
283

Table 1. Surveyor features and capabilities
number of reactors
reactor volume
10 independent vessels
57 mL, 15-40 mL (working volume), inertness is maintained
by a nitrogen blanket delivered to the headspace
vertical oscillation, adjustable, two zones
range: -99 to +150 °C; ramping rates are controllable
control accuracy: (2 °C between -40 and 150 °C
rate: ambient to +150 °C (30 min)
agitation
reactor temperature settings
heating/cooling rates
ambient to -40 °C (30 min)
reagent/solvent additions
automated additions of up to nine different bulk liquids from reservoirs
via the fluidic system or up to 176a solutions or slurries from sample
vials via robotic syringe; metering rates are controllable
manual additions of solutions, slurries, and solids
min: 100 µL, max: 25 mL (per aliquot), increment: 10 µL
automated sampling: up to 176a total samples per run
adjustable volume: 100 µL to 1 mL
liquid delivery (fluidics)
sampling/quench/analysis
automated quench, mix, and dilute
automated HPLC injection
adjustable sampling phase selection
drag and drop programming
computer interface
real-time data tracking
complete log file generation
import/export (Excel, text, etc.)
a The total number of vial positions available for both reagents and samples.
Table 2. Sampling precision test
Table 3. Experimental parameters for the chromophore test
sample volume
area
RSDc
(%)
accuracyd
(%)
RV#
biphenyl (µL)
anthracene (µL)
(µL)
averagea
SDb
1
2
3
4
5
6
7
8
9
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
2500
0
2500
0
2500
0
2500
0
100
300
500
506
1470
2367
21
22
74
4.23
1.49
3.15
1.34
1.05
3.69
a Average peak area for 15 samples in mAU. b Standard deviation. c Relative
SD. d Percent error determined from a calibration curve.
2500
0
10
automatically delivered to each of five HPLC vials containing
a fixed volume of acetonitrile. Offline HPLC analysis showed
that an average of 99.98 ( 3.27 µL had been delivered as
determined by comparison with a standard calibration curve.
This was well within the goal of (15 µL at 200 µL volume.
Likewise, the sampling system was tested by a manual addi-
tion of a solution containing biphenyl to one of the reactors
which was subsequently sampled at three different volume
settings (100, 300, and 500 µL) 15 times each via the robotic
needle and deposited into collection vials. The aliquots were
then automatically diluted and analyzed by HPLC (Table
2). The precision of sampling was determined directly from
the HPLC area count data to have relative standard deviations
of 4.23, 1.49, and 3.69% for the 100, 300, and 500 µL
samples, respectively. The accuracy of these sample volumes
was determined by comparison with a calibration curve and
ranged from 1.05 to 3.69% error. It is important to note that
the error associated with the dilution and HPLC analysis is
also included in the figure. Precise (reproducible volumes)
sampling is of crucial importance because this allows direct
comparison of HPLC area counts among different samples
without need of an internal standard or the use of peak ratios.
For example, in a 10-reaction screening experiment employ-
ing 10 different solvents, the sample giving the most area
for the product peak in the HPLC would also be the reaction
with the most product present.
With verification that the delivery and sampling compo-
nents of the system were performing within specification, a
self-test of the entire system working together (chromophore
test) was developed. In this test, two Surveyor reservoirs were
charged with standard solutions of biphenyl (50 mg/mL) and
anthracene (25 mg/mL). Automatic deliveries of anthracene,
biphenyl, or both were then made to all 10 reaction vessels
(RVs) as defined in Table 3. The volume of biphenyl was
progressively increased across RVs 1-10 while the volume
of anthracene was held constant and added only to alternating
(odd) vessels. In this way accuracy, precision, and cross
contamination of all system components were checked
simultaneously. Each RV was sampled three times. A plot
of the HPLC area counts derived from the biphenyl peaks
gave a straight line showing that it had been accurately and
precisely delivered and that the sampling volumes were
uniform. The anthracene points gave a horizontal line
showing that reagent delivery and sampling were uniform
with an average relative standard deviation of 3.1%. No cross
contamination of anthracene into even RVs was detected
(Figure 1).
Agitation. Agitation efficiency is an important reaction
variable that must be considered during chemical develop-
284
•
Vol. 5, No. 3, 2001 / Organic Process Research & Development

Figure 1. Plot of reaction vessel number (RV#) vs HPLC area for anthracene and biphenyl in the chromophore test.
Scheme 1. Alkylation of phenylacetonitrile (PAN)
deprotonation; therefore, the better the mixing, the larger the
surface area of biphasic contact, the faster the rate. By
measuring the rate of disappearance of PAN (V0) in a round-
bottomed flask at five different stirring speeds (RPM), a
series of lines with slope equal to V0 was obtained (Figure
2). Plotting the log of V0 vs the log of the RPM (Figure 3)
then produced a calibration curve. This curve was a
polynomial described by eq 1. Solving eq 1 for RPM
provided eq 2. With eq 2 in hand, the rate of this reaction
was then measured under various instrument agitation
settings, and a corresponding “effective RPM” was calcu-
lated.
ment and scale-up. During progression from discovery
through the process development cycle, agitation methodol-
ogy is often changed from magnetic to mechanical paddles
and ultimately to reactor impellers. These graduations can
be accompanied by inconsistent results with some reaction
types being more sensitive than others. It is therefore
desirable to have very efficient (though realistic) agitation
in any reactor system used for chemical development,
particularly for optimization and validation. After considering
the alternatives of magnetic spinning bars and mechanical
paddle stirrers, we settled on the technique of magnetically
coupled vertical oscillation that had proven effective in other
automated reactors.4 This technique was more easily imple-
mented than other commonly used forms; however, valida-
tion of its performance relative to standard methods was
required.
A phase-transfer catalysis (PTC) reaction known to be
rate-dependent almost exclusively on stirring frequency5 was
used to compare agitation by vertical oscillation with the
other standard laboratory method of mechanical stirring
paddles. The reaction chosen was the alkylation of phenyl-
acetonitrile (PAN) with ethyl bromide in the presence of
tetrabutylammonium bromide (TBAB) as shown in Scheme
1.
log V0 ) 3.9694(log RPM)2 - 18.731(log RPM) + 18.088 (1)
RPM ) antilog {(18.731 + [(-18.731)2 -
4(3.9694)(18.088 - log V0)]0.5}/2(3.9694) (2)
Thus, a quantitative method for evaluating the efficiency of
vertical oscillation (or any other type of agitation) relative
to a standard mechanical paddle in a round-bottomed flask
was developed. The instrument’s performance at two dif-
ferent vertical oscillation settings and three different volumes
is shown in Table 4. At a rate of 1 stroke/second, the effective
RPM stirring rate of 1011, 796, and 591 was observed at a
volume of 20, 30, and 40 mL, respectively. Cycle times of
as little as 0.5 strokes per second are possible so that even
more vigorous agitation may be achieved on Surveyor. For
comparison purposes the reaction rate was also measured
when magnetic stirring was employed. A stir plate with a
commonly used setting of 3.5 (provides a good vortex) gave
a calculated effective RPM of 464. A setting of 7.0
(maximum setting before decoupling of the magnet occurred)
corresponded to 879 rpm. These results suggest that vertical
oscillation provides more effective mixing than a standard
magnetic stirring plate is capable of. It is important to note
that this is only a comparison of agitating a biphasic mixture
and will not necessarily be valid for all types of mixtures.
This dependence may be explained by a mechanism where
deprotonation of PAN occurs only at the aqueous/organic
interface while the catalyst functions only to transport the
anion from the interface into the bulk organic phase where
alkylation rapidly occurs. The rate-determining step is the
(4) This technique is used in the Quest 210, Quest 205, and First Mate
Synthesizers (Argonaut Technologies).
(5) Solaro, R.; D’Antone, S.; Chiellini, E. J. Org. Chem. 1980, 45, 4179-
4183.
Vol. 5, No. 3, 2001 / Organic Process Research & Development
•
285

Figure 2. Plot of the reaction time vs concentration of PAN at various stirring rates in a standard round-bottomed flask.
Figure 3. Plot of the log of stirring RPM vs log of the initial
Figure 4. Plot of internal temperature vs time for 10 different
rate of disappearance of PAN in a standard round-bottomed
heating ramp rates.
flask.
Table 4. Calculated effective RPM rates for Surveyor,
magnetic stir plate, and shaker
agitation type, setting
log V0 calcd log RPM calcd RPM
Surveyor, 30 mL 1 s/stroke -2.8451
Surveyor, 30 mL 3 s/stroke -3.3754
Surveyor, 40 mL 1 s/stroke -3.3325
Surveyor, 20 mL 1 s/stroke -2.3778
2.901
2.759
2.772
3.001
2.666
2.944
2.784
796
574
591
1011
464
879
749
Corning stir plate, 3.5
Corning stir plate, 7.0
-3.6345
-2.6535
Gyrotory shaker, 200 rpm -2.957
Temperature Control. Temperature control and monitor-
ing are of paramount importance during reaction develop-
ment. The Surveyor is equipped with resistive heaters on
each RV that are controlled by thermocouples inside each
RV. Cooling is achieved by passing liquid nitrogen through
a manifold in which there is a separate cryo-valve for each
reactor that regulates flow in a jacket surrounding each
reactor. Liquid nitrogen and heaters work in concert to
control subambient temperature settings. Actual internal
temperature is continuously measured, plotted in real time,
and written to a log file for later reference. To demonstrate
the effectiveness of a heating ramp, each of 10 RVs
containing THF was programmed to heat to reflux (70 °C
set point) over 30, 60, 90, 120, ..., 270 min, and the real-
time plot generated by the Surveyor software is shown in
Figure 4. A very linear response was observed, even over
the longest ramp time. Refluxing is accomplished through
the use of a heat exchanger through which a coolant is
Figure 5. Plot of internal temperature vs time for 10 vessels
with a fixed cooling ramp rate of 30 min.
circulated. Likewise, to test the effectiveness of cooling, all
10 RVs containing THF were cooled to -40 °C at a constant
rate (Figure 5). The observed internal temperatures were
linear and reasonably consistent among all 10 vessels. As a
final test of cooling, the RVs were cooled to 10 different
temperatures ranging from 20 to -70 °C and held for a
period of 1 h (Figure 6). Again, ramps were linear, and the
desired internal temperatures were held to within 2 °C of
the set point even at -70 °C. It should be noted that these
plots were created directly by the control software and are
accessible in real time during an experiment.
Chemistry. Case 1. Our targets for chemistry validation
centered on materials generally useful for combinatorial
chemistry, that is, resins and linkers, or parallel synthesis,
286
•
Vol. 5, No. 3, 2001 / Organic Process Research & Development

peaks is shown for RV1 (Figure 7) and the complete data
set is summarized in Table 6.
On the basis of considering this process step by itself, it
appeared prudent to select KOH in either DMF or THF for
highest conversion in shortest time. However, when the
second step in the process (ester hydrolysis) was considered,
a different conclusion was reached. From a throughput
standpoint it was desirable to omit isolation of the intermedi-
ate ester by combining the two reactions into a “one-pot”
operation. The second step involving ester hydrolysis (sa-
ponification) is normally carried out in polar protic solvents
such as water or alcohol. Both DMF and THF were expected
to be inferior to EtOH for ester hydrolysis. An additional
consideration was that DMF and THF are more expensive
than EtOH. Therefore, EtOH appeared to be a good solvent
for attempting a one-pot synthesis that would avoid a solvent
exchange. Examination of the data set in Table 6 showed
that the best base for use in ethanol was KOt-Bu that gave
a respectable 91% conversion. To test this hypothesis a scale-
up run (1 kg) was then conducted using KOt-Bu in EtOH
followed by in situ hydrolysis affording the acid in 84%
isolated yield. The improved process was discovered using
a combination of automated screening and general process
development knowledge. Improvements included: (1) elimi-
nating one isolation/drying step; (2) conversion improvement
by selecting a different solvent-base combination; (3) raw
materials cost savings and waste minimization; (4) cycle time
improvement for both steps; (5) yield increase from 66 to
84%.
Case 2. A novel multistep route to a chiral imidazolidin-
2-one was developed from initial route scouting through
reaction condition screening, optimization, and scale-up. Our
retrosynthetic approach is shown in Scheme 3. Imidazolidin-
2-ones have been used as chiral auxiliaries in asymmetric
aldol condensations,9 alkylations,10 Michael additions,11 and
Diels-Alder reactions.12 They are also intermediates in the
synthesis of a class of HIV protease inhibitors.13 The
enantioselectivities achieved in alkylation reactions are
similar to their analogue oxazolidinones; however, imida-
zolidin-2-one adducts are usually crystalline and are less
susceptible to nucleophilic ring-opening, offering advantages
in downstream purification (diastereomeric upgrade) and
auxiliary recycling. Substituted imidadzolidin-2-ones have
been prepared by fusion of 1,2-amino alcohols with urea14
and by reaction of diamines with phosgene.15 In this new
and general route, chirality is derived from readily available
and relatively inexpensive amino acids 1 (Scheme 4).
Carboxyl reduction followed by protection at nitrogen would
Figure 6. Plot of internal temperature vs time for cooling 10
vessels to different set points.
Scheme 2. Synthesis of PS-Indole-CHO resin
that is, scavengers and polymeric reagents. Several single
and multistep reactions were studied to identify the best route
(scouting), reagent/solvent combinations (screening), or
optimal levels of continuous variables such as temperature,
concentration, and stoichiometry (optimization). The indole
linker6 is a key intermediate in the synthesis of PS-Indole-
CHO,7 a polystyrene resin used in combinatorial synthesis
when employing reductive amination/acylation methodology
(Scheme 2). Literature conditions for the alkylation of indole-
3-carboxaldehyde with ethyl bromoacetate (K2CO3, DMF,
60 °C) gave 50% conversion at best in our hands at the
bench. We chose to screen three solvents (DMF, THF, EtOH)
and three bases, K2CO3 (2.5 equiv) KOH (2.5 equiv), and
KOt-Bu (1.5 equiv) in this reaction for a total of nine
reactions leaving one RV for a replicate. Operations were
conducted as described in Table 5. All 10 RVs were manually
charged with solid indole-3-carboxaldehyde followed by the
solid bases: K2CO3 (RVs 1, 4, 7) or KOH (RVs 2, 5, 8).
Solvents were then added automatically followed by the
KO-tBu solution in THF automatically. Agitation was
initiated, and ethyl bromoacetate (1.0 equiv) was added neat
over a 10-min period. The reactions were sampled, and a
second charge of bromide (1.0 equiv) was added. The
reactions were sampled again at 30 and 240 min. For ease
of visualization graphs were constructed of HPLC area versus
sample time for each RV.8 A representative plot for observed
(9) Roder, H.; Helmchen, G.; Peters, K.; von Schnering, H.-G. Angew. Chem.,
Int. Ed. Engl. 1984, 96, 895-896.
(10) (a) Cardillo, G.; D’Amico, A.; Orena, M.; Sandri, S. J. Org. Chem. 1988,
53, 2354-2356. (b) Konigsberger, K.; Prasad, K.; Repic, O.; Blacklock, T.
J. Tetrahedron: Asymmetry 1997, 8, 2347-2354.
(11) Melnyk, O.; Stephan, E.; Pourcelot, P.; Cresson, P. Tetrahedron 1992, 48,
841-850.
(12) Roos, G. H. P.; Jensen, K. N. Tetrahedron: Asymmetry 1992, 3, 1553-
1554.
(6) Estep, K. E.; Neipp, C. E.; Stephens Stramiello, L. M.; Adam, M. D.; Allen,
M. P.; Robinson, S.; Roskampet, E. J. J. Org. Chem. 1998, 63, 5300.
(7) PS-Indole-CHO is commercially available from Argonaut Technologies.
(8) This feature is now a part of Surveyor software and is done automatically.
(13) Tung, R. D.; Salituro, F. G.; Deininger, D. D. International Patent WO 97/
27180, 1997.
(14) Close, J. W. J. Org. Chem. 1950, 15, 1131-1134.
(15) See ref 10b.
Vol. 5, No. 3, 2001 / Organic Process Research & Development
•
287

Table 5. Operations used in the indole alkylation solvent-base screening experiment
action
RV-1
RV-2
RV-3
RV-4
RV-5
RV-6
RV-7
RV-8
RV-9
RV-10
add quench to sample vials
manual add, indole (g)
manual add, KOH (g)
manual add, K2CO3 (g)
add THF (mL)
×3
×3
×3
×3
×3
×3
×3
×3
×3
×3
2.00
2.00
1.94
2.00
2.00
2.00
1.94
2.00
2.00
2.00
1.94
2.00
2.00
4.79
4.79
20
4.79
20
20
20
20
add EtOH (mL)
20
20
add DMF (mL)
add KOt-Bu (mL)
add bromide, 1 equiv (mL)
wait 5 min, sample all
add bromide, 1 equiv (mL)
wait 30 min, sample all
wait 240 min, sample all
20
20
20
12.61
1.54
12.61
1.54
12.61
1.54
12.61
1.54
1.54
1.54
1.54
1.54
1.54
1.54
1.54
1.54
1.54
1.54
1.54
1.54
1.54
1.54
1.54
1.54
Scheme 3
and ultimately provides the urea carbonyl upon cyclization,
obviating the need for phosgene (Scheme 6). In this work
we employed L-phenylalanine (Phe) as the amino acid and
benzylamine17 as the nitrogen nucleophile (R1 ) R2 )
benzyl) for route-scouting and screening studies. Each route
was evaluated and compared from a scale-up standpoint.
In the first stage the synthesis of pivotal intermediate 2
was studied by a one-pot lithium aluminum hydride (LAH)
reduction and Boc protection. The in situ N-protection of
the polar amino alcohol intermediate facilitated product
isolation from the spent aluminum salts. A literature proce-
dure18 describing the preparation of Boc-phenylglycinol used
LAH (2 equiv), THF (25 mL/g phenylglycine), Boc2O (1.1
equiv), and DMAP (cat) to afford a 53% isolated yield. In
our lab, a manual run with Phe in place of phenylglycine
gave desired product in 43% isolated yield (average of 2
runs) along with 10-15% of a side product identified as
bis-Boc-phenylalaninol. A third run conducted with the
omission of DMAP led to a similar yield with no formation
of bis-Boc byproduct. Therefore, addition of DMAP was
found unnecessary and was not used in the automated run.
An automated run was then conducted in which the LAH
charge and the total concentration were varied (Table 7).
During the carboxyl reduction step each reaction suspension
was sampled at 30, 90, and 270 min. The samples were
quenched into aqueous NaOH, diluted with dioxane, allowed
to settle, and injected onto an HPLC. This was found to be
a clean conversion of Phe to the intermediate amino alcohol
that reached a steady state after 90 min. Continuing on to
the protection step, each reaction was automatically quenched
with aqueous NaOH and treated with a DCM solution of
di-tert-butyl dicarbonate. The reaction suspensions were
again sampled over time, and complete conversion of the
intermediate amino alcohol to Boc amine 2 was observed
after 120 min. The overall process was found to be very
Figure 7. Plot of sample time vs HPLC area for RV-1 in the
alkylation of indole-3-carboxaldehyde.
Table 6. Screening of bases and solvents in the alkylation of
indole-3-carboxaldehyde
RV#
base
solvent SMa producta % conversionb
7
4
5
1
3
6
10
9
2
8
K2CO3
K2CO3
KOH
K2CO3
KOt-Bu
KOt-Bu
KOt-Bu
KOt-Bu
KOH
THF
8225
7146
5557
2189
813
632
253
235
216
0
1112
2107
4643
7821
6932
6126
6719
6574
10768
7735
11.9
22.8
45.5
78.1
89.5
90.6
96.4
96.5
98.0
100.0
EtOH
EtOH
DMF
DMF
EtOH
THF
THF
DMF
THF
KOH
a HPLC area (mAU) for starting material (SM) or product. b Calculated using
the formula % conversion ) product peak area/total peak area.
give the pivotal intermediate 2. Oxidation to the aldehyde 3
followed by reductive amination would give penultimate
intermediate 5. Alternatively, 5 could be accessible from 2
by a sequence of mesylation, to give 4, followed bydisplace-
ment.16 In either route, the alkoxycarbonyl group (Boc) serves
as a protecting group through the early stages of the synthesis
(17) An ultimate goal of this work is to develop an insoluble, solid-supported
version of this chiral auxiliary. In this case, benzylamine serves as a model
for aminomethyl polystyrene resin.
(16) This is similar to the sequence described previously; see ref 13.
(18) Correa, A.; Denis, J.-N.; Greene, A. E. Synth. Commun. 1991, 21, 1.
288
•
Vol. 5, No. 3, 2001 / Organic Process Research & Development

Scheme 4. Two routes to the penultimate imidazolidinone intermediate 5
Table 7. Optimization of LAH equivalents and
concentration in the conversion of amino acid (1) to N-Boc
amino alcohol (2)
Table 8. Comparison of original literature conditions to the
optimized conditions for the conversion of amino acid (1) to
N-Boc amino alcohol (2)
HPLC areaa
literature conditions
RV#
LAH (equiv)
concentration (mL/g)
Step 1
LAH (equiv)
THF (mL/g)
time
Step 2
1
2
2.0
30
30
30
30
30
20
20
20
20
20
7005
6360
5902
3999
3893
5560
5594
4376
4406
3293
1.75
1.50
1.25
1.00
2.00
1.75
1.50
1.25
1.00
2
25
6 h
reflux
Boc2O
DMAP
time
1.2
yes
6 h
reflux
3
4
5
temp
temp
6
7
total time
isolated yield
12 h
43%
8
9
10
optimized conditions
a HPLC area counts in mAU associated with product, no other peaks were
observed.
Step 1
Step 2
LAH (equiv)
THF (mL/g)
time
2
30
2 hr
60
Boc2O
DMAP
time
1.0
No
2 h
25
sensitive to concentration as well as stoichiometry. The
highest concentration of product, as determined by HPLC
area counts, was found in the reaction where 2.0 equiv of
LAH was used under the more dilute conditions of 30 mL
of THF/g Phe (RV1). Each reaction step was determined to
require ∼2 h rather than the 6 h for each step reported in
the literature. A scale-up run was then conducted at the bench
using the optimized conditions affording a 63% isolated yield
(vs 43% unoptimized) with the reaction time being cut by a
factor of 3 (Table 8). This scale-up run provided material
for subsequent route-scouting studies.
Displacement Route. The mesylate 4 was prepared from
alcohol 2 in quantitative yield under standard conditions,19
and the displacement reaction was studied with butylamine
and benzylamine. Treatment of mesylate 4 with 20 equiv20
of either amine in dioxane at 60 °C led to clean conversion
to the desired amine (90% isolated). Efforts to reduce the
amine charge to 3 equiv gave a second product that was
isolated and identified as the oxazolidinone 7. Precedence
for this reaction was found in the literature21 where it was
shown that treatment of analogue tosyl derivatives of 4 with
DIEA gave oxazolidinone 7 in excellent yield (Scheme 5).
Because it appeared that suppression of this side reaction
temp
temp
total time
isolated yield
4 h
63%
could only be realized through the use of a large excess of
amine, this route was ultimately abandoned.
Oxidation/Reductive Amination Route. The second
route to penultimate intermediate 5 involved oxidation of
alcohol 2 to the aldehyde 3 followed by reductive amination.
Preliminary screening of oxidation procedures included the
Swern, Moffatt, pyridine-SO3, and Marko procedures. In
summary, Swern oxidation was reliable, affording product
in 75-85% conversion based on HPLC. The Moffatt
procedure gave clean oxidation, but the product was con-
taminated with diisopropylurea that presented purification
problems. Marko oxidation (CuCl-phenanthroline (cat.)
DBAD, K2CO3, air22 gave low conversion for this substrate.
Pyridine-SO3-TEA gave mixed results, probably due to the
inconsistent purity of commercially available material.
Ultimately, the Swern procedure was adopted to produce
material needed for downstream studies despite the draw-
backs of low-temperature requirements and dimethyl sulfide
formation.
(19) Gooding, O. W.; Bansel, R. P. Synth. Commun. 1995, 25, 1155.
(20) Twenty equivalents of amine was used for a similar substrate in the Vertex
patent, see ref 13.
(21) Curran, T. P.; Pollastri, M. P.; Abelleira, S. M.; Messier, R. J.; McCollum,
T. A.; Rowe, C. G. Tetrahedron Lett. 1994, 35, 5409.
(22) Marko, I. E.; Gautier, A.; Chelle-Regnaut, I.; Giles, P. R.; Tsukazaki, M.;
Urch, C. J. J. Org. Chem. 1998, 63, 7576.
Vol. 5, No. 3, 2001 / Organic Process Research & Development
•
289

Scheme 6
are two points plotted for each reaction temperature; the first
was from the 60-min sample and the second from the 240-
min sample. It is interesting that the reaction continued to
proceed after warming to 25 °C as Swern oxidations are
normally worked up immediately after warming.
The reductive amination reaction (3 to 5) was studied
using sodium triacetoxyborohydride reagent in a procedure
known to be very general and reliable.23 An automated run
was conducted in which four solvents: DCM, THF, DCE,
and ACN were screened. The charges of benzylamine (1.2
equiv) and reducing agent (1.9 equiv) were held constant,
and the reaction was sampled over time. All reactions went
to completion within 1 h with no differences detected.
Because this reaction was found not to be sensitive to solvent,
we standardized on DCM for further studies.
Figure 8. Plot of temperature vs % conversion for the
automated Swern oxidation experiment.
Scheme 5. Decomposition pathway leading from mesylate
4 to oxazolidinone 7
Because Boc-aminoaldehydes are known to be unstable
to storage and can be difficult to isolate in pure form, it was
desirable to seek a one-pot procedure for converting 2 to 5
to avoid isolating this material.
A scale-up run was conducted using the optimal condi-
tions for each step. Thus, alcohol (1.0 equiv), oxalyl chloride
(1.5 equiv), DMSO (3 equiv), and DIEA (6 equiv) in DCM
at -60 °C followed by warming to 25 °C for 1 h gave clean
oxidation by HPLC. Direct in situ treatment with benzy-
lamine (1.2 equiv) and NaBH(OAc)3 (1.9 equiv) provided a
62% isolated yield of pure material at the 10-g scale. This
provided sufficient material to study the cyclization reaction
while further reaction screening/optimization experiments on
this step were continued.
Cyclization. Although the literature precedent13 for the
conversion of 5 to 6 used a deprotection/phosgene treatment
to conduct cyclization, we envisioned that the final cycliza-
tion reaction (Scheme 6) could be facilitated either thermally,
under base catalysis, or Lewis acid catalysis. An initial broad
screen showed that bases were most effective. A Surveyor
run was conducted in which two solvents (THF and toluene),
three bases (potassium t-butoxide, NaHMDS, DIEA), at two
different stoichiometries (2.5 and 5 equiv) were examined
as shown in Table 9. Samples were taken at 20 and 60 min
for HPLC analysis. It was found that K-OtBu and NaHMDS
were effective at both charge levels in THF but not toluene.
DIEA was not effective in either solvent. A control experi-
ment with no base in toluene at 110 °C gave partial con-
version (thermal conditions). From a manufacturing stand-
point the case of 2.5 equiv of K-OtBu in THF for 1 h
(complete conversion) was the method of choice for scale-
up. These conditions were subsequently applied on a 6-g
scale, affording the imidazolidin-2-one in 90% isolated yield.
An initial screening experiment on the Surveyor examined
the effect of varying temperature and stoichiometry on the
Swern oxidation conversion when using DCM as a solvent.
Not unexpectedly, this revealed that temperature was the
most important variable. As for stoichiometry, conversions
did not suffer until oxalyl chloride, DMSO, and DIEA
charges dropped below 1.5, 3.0, and 6.0 equiv, respectively.
A second experiment held solvent and stoichiometry constant
at these levels and looked at five different temperatures, -78,
-60, -40, -20, and 0 °C for the reaction. The oxidations
were done at the prescribed temperatures. After warming to
25 °C samples were taken at 60 and 240 min. As shown in
Figure 8, conversions remained between 73 and 85% until
the reaction temperature was above -60 °C. Note that there
(23) Abdel-Magid, A. F.; Carson, K. G.; Harris, B. D.; Maryanoff, C. A.; Shah,
R. D. J. Org. Chem. 1996, 61, 3849.
290
•
Vol. 5, No. 3, 2001 / Organic Process Research & Development

Table 9. Screening study of the cyclization of 5 to 6
% conversiona
20 min 60 min
chromophore test was programmed as described in Table 3.
HPLC analysis: (mobile phase, acetonitrile/water, gradient,
60-80% acetonitrile in 4 min; flow rate, 2 mL/min; UV
detector 250 nm); retention time of anthracene ) 2.45 min
and biphenyl ) 1.96 min.
RV#
solvent
base
equiv
1
2
3
4
5
6
7
8
THF
toluene
THF
toluene
THF
THF
KOt-Bu
KOt-Bu
KOt-Bu
KOt-Bu
NaHMDS
DIEA
2.5
5
2.5
5
2.5
2.5
5
88.6
0.0
95.5
0.7
82.8
3.9
88.5
6.8
97.1
21.1
100.0
28.6
100.0
2.8
General Methods for the Agitation Efficiency Experi-
ments. The round-bottomed flask experiments were per-
formed in a 100-mL three-necked flask fitted with a standard
Teflon stirring paddle and a mechanical stirring motor.
Revolutions per minute (RPM) were measured with an
optical tachometer. A stock solution (20 mL) containing 4.85
M phenylacetonitrile (PAN), 5.79 M bromoethane (RBr), and
0.115 M tetrabutylammonium bromide (TBAB) was prepared
by combining 11.2 mL of PAN, 8.64 mL of RBr, and 0.0741
g of TBAB. This solution was charged to the flask followed
by 40 mL of 50 wt % aqueous NaOH. The flask was heated
to 55 °C, and stirring at the specified RPM was initiated.
Sampling was accomplished by stopping the agitation,
removing 0.200 mL of organic layer via syringe, and
quenching into 1.0 mL of of 0.1 N aqueous hydrochloric
acid using a rinse of 0.80 mL of THF. Samples were further
diluted with 0.60 mL of diethyl, ether and the upper layer
was analyzed by HPLC. Samples were taken at 0, 4, 8, 12,
16, 20, 30, 40, 50, 60, 90, and 120 min. Data from these
runs were processed using Excel24 to determine V0 at each
RPM and to plot log RPM versus log V0. The automated
run was conducted in an analogous manner with the
appropriate volume of stock solution delivered automatically
and the NaOH solution added manually because of its high
viscosity. Samples were automatically withdrawn, quenched,
and rinsed in a manner analogous to the manual runs. The
total reaction volume and agitation settings were as described
in Table 4. Effective RPMs were calculated using eq 2.
HPLC analysis: (mobile phase, acetonitrile/water, gradient
40-75% acetonitrile in 3.5 min, flow rate 2 mL/min; UV
detection 254 nm); retention time of PAN ) 1.57 min,
ethylated PAN ) 2.20 min.
General Methods for Indole Linker Screening Experi-
ments. The automated run was programmed as described in
Table 5. Solids were manually added through a powder
funnel inserted into the top port of the vessels. HPLC
analysis: (mobile phase, acetonitrile/water, gradient 10-80%
acetonitrile in 6 min, flow rate 2 mL/min; UV detection 254
nm); retention time of indole-3-carboxaldehyde ) 2.67 min,
ethyl 3-formyl-indol-1-yl-acetate ) 3.29 min, 3-formyl-indol-
1-yl-acetic acid ) 1.78 min.
3-Formyl-indole-1-yl-acetic Acid. A dry 22-L flask was
charged with indole-3-carboxaldehyde (1.000 kg, 6.889 mol)
followed by ethanol (10 L). Agitation was initated and
potassium tert-butoxide solution (1.65 M in THF, 4.6 L) was
added over 10 min during which time the temperature rose
to 37 °C. After 20 min the flask was charged with ethyl
bromoacetate (1.265 kg, 7.58 mol, 1.1 equiv) at a rate of 20
g min-1. The reaction was allowed to proceed for a further
20 min, at which time the flask was charged with potassium
hydroxide (463 g, 8.26 mmol, 1.2 equiv) portionwise over
20 min. The solution was heated and solvent distilled as water
THF
THF
NaHMDS
DIEA
97.9
5.7
5
9
10
THF
toluene
KOt-Bu
none
5
0
54.9
36.5
94.7
38.6
a Calculated using the formula % conversion ) product peak area/total peak
area.
Summary
We have developed a new automated reactor-analysis
system for use during chemical process development. The
fully automated system is capable of reagent/solvent addition,
temperature and agitation control, and sample extraction/
quench/injection for online HPLC analysis. Both precision
and accuracy of the fluidic delivery and sampling compo-
nents were measured through the use of standard solutions
of biphenyl and anthracene in conjunction with HPLC
analysis. Agitation efficiency was evaluated using a phase-
transfer catalysis reaction known to be rate-dependent on
stirring frequency. The instrument’s performance at various
agitation settings was determined and compared to mechan-
ical and magnetic stirring. The efficiency of agitation by
vertical oscillation compared well with both magnetic and
mechanical stirring with effective stirring rates of >1000
rpm being measured. Chemistry cases examined included
several single-step and multistep reactions studied to identify
the best reagent/solvent combinations (screening) and optimal
levels of continuous variables such as temperature, concen-
tration, and stoichiometry (optimization). The system was
shown to be a useful tool for performing process screening
and process optimization during chemical development.
Experimental Section
Solvents and reagents were used as received from
commercial suppliers. HPLC was performed on a HP-1050
or HP-1100 system equipped with a Platinum EPS C18 100A
3-µm rocket column from Alltech (unless otherwise speci-
fied) and UV detector. All NMR experiments were done on
a Varian 300 spectrometer and are reported in ppm.
Elemental analyses were obtained from Desert Analytics,
Tucson, AZ.
General Procedure for Accuracy and Precision Test-
ing. Stock solutions of biphenyl (50 g/L) and anthracene (25
g/L) were prepared and diluted using standard volumetric
techniques. Delivery and sampling precision was calculated
by taking the standard deviation of the HPLC peak area
counts for all replicates. Accuracy was determined by
comparison with a calibration curve constructed by plotting
concentration versus peak area for a set of serial dilutions
centered around the concentration range of the samples. The
(24) Excel is a registered trade name of Microsoft Corp.
Vol. 5, No. 3, 2001 / Organic Process Research & Development
•
291

was added to make up the lost volume (4 L). When the head
temperature reached 93 °C, the flask was cooled to 20 °C
and washed twice with 60:40 ethyl acetate/hexanes to remove
unreacted starting material. The aqueous layer was then
heated to 45 °C with agitation and treated with aqueous
hydrochloric acid (0.746 L, 12 M, 1.3 equiv) added at a rate
of 20 mL per min. The mixture was cooled to 3 °C and aged
30 min; the product was collected by filtration, washing the
cake with water (1.5 L). The product was then vacuum-dried
to constant weight (50 °C, 0.1 bar) affording 1.183 kg
(84.6%) as an orange solid. HPLC purity >98%, physical
properties were identical to those previously reported.6
General Methods for LAH Reduction Optimization
Experiment. The automated run was programmed as de-
scribed in Table 7. A 1.00 M suspension of lithium aluminum
hydride (LAH) was prepared from crushed pellets by
agitation in a round-bottomed flask while heating at 60 °C.
Sample quench vials were automatically loaded with 3 mL
of dioxane. The solid L-Phe (1.00 g, 6.05 mmol) was added
manually at the beginning of the experiment followed by
the appropriate volume of THF automatically. Agitation was
started, and the appropriate volume of LAH suspension was
added manually via syringe with a 16-gauge needle. After
being held at an internal temperature of 60 °C for 2 h, the
vessels were cooled to 20 °C and quenched with 2 M NaOH
solution (1.6 mL/g LAH used) followed by water (2.0 mL/g
LAH used). DCM was added (12 mL to RVs 1-5 and 8
mL to RVs 6-10) followed by di-tert-butyl dicarbonate (1.32
g, 6.05 mmol). Samples were taken (0.300 mL from RVs
1-5 and 0.200 mL from RVs 6-10). Sample size was
proportional to total volume so that HPLC peak areas could
be compared directly. Final results are given in Table 7.
HPLC analysis: (mobile phase, acetonitrile with 0.1% TFA/
water, gradient 0-100% acetonitrile with 0.1% TFA in 6
min, flow rate 2 mL/min; UV detection 214 nm); retention
time of Phe ) 1.00 min, phenylalaninol ) 2.39 min, N-(tert-
butoxycarbonyl)phenylalaninol ) 4.17 min.
(S)-N-(tert-Butoxycarbonyl)phenylalaninol (2). A dry
3-L flask was flushed with nitrogen and charged with THF
(1.8 L). Agitation was started, and crushed pellets of lithium
aluminum hydride (34.2 g, 900 mmol) were added over 5
min. The flask was charged with solid (S)-(-)-phenylalanine
(61.0 g, 0.369 mol) portionwise over 1 h. The reaction was
heated to 62 °C and held for 2 h prior to cooling to 20 °C
in an ice bath. The reaction was quenched by the addition
of aqueous NaOH (68.4 mL, 2 M) followed by water (55
mL). CAUTION: exothermic, hydrogen gas eVolution! When
the addition was complete, the flask was allowed to cool to
5 °C, and a solution di-tert-butyl dicarbonate (98.2 g, 450
mmol) in DCM (0.300 L) was added at a rate of 20 mL
min-1, during which time the temperature rose to 29 °C. The
bath was removed, and the suspension was agitated at 25
°C for 1 h. The suspension was treated with diatomaceous
earth (100 g) and filtered through a sintered glass funnel.
The cake was rinsed twice with THF (0.5 L), and the
combined organics were charged to a 3-L flask and concen-
trated by distillation while adding 500 mL of heptane to make
up the volume loss. When the head temperature reached 80
°C, the heat was turned off, and the flask allowed to naturally
cool to 20 °C. The thick mixture was further diluted with
hexanes (400 mL) and cooled to 5 °C. The product was
collected by filtration and vacuum-dried to constant weight
at 25 °C, affording 58.0 g (63%) as a white crystalline solid.
HPLC purity >99%. Physical properties were identical to
authentic commercial material.25
General Methods for Oxidation Experiment. At the
beginning of the experiment sample vials were automatically
loaded with aqueous sodium bicarbonate (0.30 mL) and
acetonitrile (1.40 mL) to quench the samples prior to HPLC
analysis. Each of five RVs were dried by heating dry THF
under reflux, draining hot, and blowing dry with nitrogen
purge. The program added DCM (13.6 mL) followed by
oxalyl chloride solution (50% in DCM, 0.69 mL, 2.98 mmol).
The vessels were cooled to -78, -60, -40, -20, and 0 °C,
and a solution of DMSO (25% in DCM, 2.3 mL, 5.97 mmol)
was metered in over 2 min. The program then metered in a
solution of alcohol 4 (2.99 mL, 1.0 M, 1.99 mmol). The
temperature was held for 1 h, and a solution of DIEA in
DCM (50%, 4.17 mL, 11.94 mmol) was metered in over 2
min. The vessels were taken to 0 °C using a linear ramp
over 30 min and sampled. The vessels were further warmed
to 25 °C and sampled again at 60, and 240 min for on-line
analysis. Results for the 60- and 240-min points are shown
in Figure 8. HPLC analysis: (column, HP-Zorbax XDA-
C8; mobile phase, acetonitrile/water; gradient 40-70%
acetonitrile in 5 min; flow rate 1 mL/min; UV detection 214
nm); retention time of 2 ) 4.51 min, 3 ) 3.21 min.
General Methods for Reductive Amination Experi-
ment. Aldehyde 3 (0.40 g, 1.6 mmol) was charged to each
of four RVs followed by one of four solvents, DCM, THF,
1,2-dichloroethane, or acetonitrile (60 mL/g). Agitation was
started, and the reactions were treated with benzylamine (0.24
g, 2.2 mmol) and sodium triacetoxyborohydride (0.64 g, 3.0
mmol). The reactions were periodically sampled into vials
containing acetonitrile and analyzed by HPLC. All reactions
were found complete within 1 h. HPLC analysis: (mobile
phase, acetonitrile/water, gradient 40-70% acetonitrile in 5
min, flow rate 1 mL/min; UV detection 214 nm); retention
time 3 ) 3.21 min, 4 ) 4.32 min.
(S)-2-(tert-Butoxycarbonylamino)-3-phenyl-1-(N-ben-
zyl)propylamine (5). A dry 500-mL round-bottomed flask
was charged with DCM (50 mL) and DMSO (4.7 g, 60
mmol, 3.0 equiv). The flask was placed in a dry ice/acetone
bath, and neat oxalyl chloride (3.8 g, 30 mmol) was added
over 30 s via syringe. After 30 min a solution of alcohol 2
(5.00 g, 20 mmol) in DCM (50 mL) was added dropwise
over 15 min, and the solution was stirred for an additional
30 min. A solution of diisopropylethylamine (15.5 g, 120
mmol) in DCM (50 mL) was added dropwise over 15 min,
and the solution was stirred for an additional 30 min. A
solution of benzylamine (2.4 g, 22 mmol) in DCM (10 mL)
was added followed immediately by solid sodium triacetoxy-
borohydride (6.36 g, 30 mmol). After an additional 30 min
the flask was removed from the bath and allowed to warm
(25) (S)-2-(tert-Butoxycarbonylamino)-3-phenyl-1-propanol was obtained from
Aldrich.
292
•
Vol. 5, No. 3, 2001 / Organic Process Research & Development

to ambient temperature overnight. The solution was treated
with aqueous NaHCO3 with agitation for 10 min, and the
layers were separated. The organic phase was dried over
MgSO4, filtered, and concentrated under reduced pressure.
The residue was purified on a column of silica gel, eluting
with 20:80 ethyl acetate/DCM. Fractions containing product
were concentrated, affording 4.2 g of analytically pure
product (62%). 1H NMR(CDCl3): 7.35-7.16(m, 10H), 4.75-
(br s, 1H), 3.90(br s, 1H), 3.79(d, 1H, J ) 13.2), 3.75(d,
1H, J ) 13.2), 3.00-2.61(m, 5H), 1.41(s, 9H). 13C NMR-
(CDCl3): 155.85, 140.5, 138.3, 129.6, 129.5, 128.7, 128.6,
128.3, 127.2, 126.5, 79.4, 54.0, 51.6, 39.4, 28.6. Anal. Calcd
for C21H28N2O2: C, 74.08; H, 8.29; N, 8.23. Found: C,
73.80; H, 8.50; N, 8.81.
General Methods for Cyclization Experiment. The
automated run was programmed as described in Table 9. At
the beginning of the experiment sample vials were loaded
with aqueous HCl (0.30 mL) and THF (0.70 mL). Each of
10 RVs were manually charged with solid amine 5 (0.50 g,
1.47 mmol). The program then added the appropriate solvent
followed by the base (either 2.5 or 5 equiv). RVs 1-9 were
heated to 60 °C, and 10 was heated to 110 °C, and samples
were quenched into vials at 20 and 60 min for on-line
analysis. HPLC analysis: (mobile phase, acetonitrile/water,
gradient 20-70% acetonitrile in 7 min, flow rate 2 mL/min;
UV detection 214 nm); retention time of 6 ) 5.17 min.
(S)-1-Benzyl-4-benzylimidazolidin-2-one (6). A dry 250-
mL round-bottomed flask was charged with amine 5 (5.76
g, 16.9 mmol) and THF (50 mL). The resulting solution was
treated with a solution of potassium-tert-butoxide (31.9 mL,
51 mmol, 1.6 M), and the resulting yellow solution was
heated to 60 °C for 3 h. The solution was cooled to ambient
temperature and acidified with aqueous HCl (35 mL, 1 M)
and concentrated under reduced pressure. The aqueous
residue was extracted with ethyl acetate, and the organic
phase was washed twice with brine, dried over MgSO4, and
concentrated under reduced pressure to a viscous oil that
1
solidified on standing, affording 4.07 g (90%). H NMR-
(CDCl3): 7.32-7.07(m, 10H), 4.32(s, 2H), 3.84-3.80(m,
1H), 3.33(t, 1H, J ) 8 Hz), 2.99(t, 1H, J ) 8 Hz), 2.73(d,
2H, J ) 7 Hz). 13C NMR(CDCl3): 161.7, 137.2, 137.1,
129.2, 129.0, 128.8, 128.2, 127.7, 127.1, 51.45, 49.8, 47.7,
42.2. Anal. Calcd for C17H18N2O: C, 76.66; H, 6.81; N,
10.52. Found: C, 76.40; H, 6.86; N, 10.58.
Acknowledgment
We acknowledge the following Argonaut staff: Edward
Munyak, Steve Miller, Peter Wright (Engineering) and Dan
Bernstein, Joe Slater (Software Development) without whom
this work would not have been possible. We also thank the
Surveyor Product Development Consortium: Joel Hawkins
(Pfizer), Van Martin (Agouron), Michael O’Brien (Aventis),
Lori Spangler (Rohm and Haas), and Tony Zhang (Lilly)
for helpful discussions during the course of this work.
Received for review October 13, 2000.
OP0001077
Vol. 5, No. 3, 2001 / Organic Process Research & Development
•
293