Extending Ru-BINAP catalyst life and separating products from catalyst using membrane recycling

Article abstract of DOI:10.1021/op900056s

Organic solvent nanofiltration (OSN) is a new membrane technology with many applications in pharmaceutical and fine chemicals development and manufacture, from laboratory through production scales. One application of particular industrial relevance is the ability to recover and recycle homogeneous catalysts, in particular asymmetric, organometallic, homogeneous transition metal catalysts. Industrial application of this group of catalysts is often limited due to the cost of applying these catalysts to single reactions and the subsequent product yield losses associated with removal of the catalyst from solution. OSN provides a technique that can maximise the value of the catalyst through catalyst recycle from one reaction batch to the next, whilst minimising the concentration of catalyst present in the reaction product. This contribution describes the development and demonstration of OSN catalyst recycle on an example catalytic system, the homogeneous, asymmetric hydrogenation of dimethyl itaconate (DMI) to dimethyl methylsuccinate (DMMS) with Ru-BINAP. The first step of the development process was to demonstrate that Ru-BINAP was sufficiently stable that the process concept, i.e., catalyst recycling, was feasible. A dilute solution (0.8 wt %) of DMI in methanol with a high catalyst loading (S/C = 500) was used, and 14 successive reactions were carried out with constant reaction rate and without yield or enantiomeric excess (ee) reduction. In industrial practice, the goal is always to operate a reaction at the highest, practical S/C ratio, and thus the next stage of development was to optimise the S/C ratio. The limiting, practical S/C ratio was found to be 7000. The process was then run with the optimised dilute reaction system, which allowed 10 reaction cycles to be carried out with only 20% addition of the initial (reaction 1) mass of catalyst to reactions 2-10 to maintain reaction rate, conversion, and ee. This leads to a 5 times increase in the overall S/C ratio. In addition, this optimisation reduced the amount of metal present in the product solution from 130 μg Ru per gram of product to <6 μg Ru per gram of product. The process was then scaled up to an industrially relevant substrate concentration (20 wt % DMI in methanol), and it was found that both the reaction (in terms of reaction rate, conversion, and yield) and the ability to recycle the catalyst were unaffected by the scale-up in concentration.

Full text of DOI:10.1021/op900056s

Organic Process Research & Development 2009, 13, 863–869
Extending Ru-BINAP Catalyst Life and Separating Products from Catalyst Using
Membrane Recycling
Dinesh Nair,†,‡ Hau-To Wong,§ Shejiao Han,§,⊥ Ivo F. J. Vankelecom,¶ Lloyd S. White,0 Andrew G. Livingston,§,† and
Andrew T. Boam*,†
Membrane Extraction Technology Ltd., Unit 8 Wharfside, Rosemont Road, Wembley HA0 4PE, U.K., Technical DeVelopment,
GlaxoSmithKline, Temple Hill, Dartford, Kent DA1 5AH, U.K., Department of Chemical Engineering and Chemical
Technology, Imperial College of Science, Technology and Medicine, London SW7 2BY, U.K., Merpro Limited,
Brent AVenue, Montrose, Angus DD10 9PB, U.K., Centre for Surface Chemistry and Catalysis, Katholieke UniVersiteit
LeuVen, Kasteelpark Arenberg 23, 3001 LeuVen, Belgium, and W.R. Grace and Co., 7500 Grace DriVe, Columbia,
Maryland 21044, U.S.A.
Abstract:
in methanol), and it was found that both the reaction (in terms of
reaction rate, conversion, and yield) and the ability to recycle the
catalyst were unaffected by the scale-up in concentration.
Organic solvent nanofiltration (OSN) is a new membrane technol-
ogy with many applications in pharmaceutical and fine chemicals
development and manufacture, from laboratory through produc-
tion scales. One application of particular industrial relevance is
the ability to recover and recycle homogeneous catalysts, in
particular asymmetric, organometallic, homogeneous transition
metal catalysts. Industrial application of this group of catalysts is
often limited due to the cost of applying these catalysts to single
reactions and the subsequent product yield losses associated with
removal of the catalyst from solution. OSN provides a technique
that can maximise the value of the catalyst through catalyst recycle
from one reaction batch to the next, whilst minimising the
concentration of catalyst present in the reaction product. This
contribution describes the development and demonstration of OSN
catalyst recycle on an example catalytic system, the homogeneous,
asymmetric hydrogenation of dimethyl itaconate (DMI) to di-
methyl methylsuccinate (DMMS) with Ru-BINAP. The first step
of the development process was to demonstrate that Ru-BINAP
was sufficiently stable that the process concept, i.e., catalyst
recycling, was feasible. A dilute solution (0.8 wt %) of DMI in
methanol with a high catalyst loading (S/C ) 500) was used, and
14 successive reactions were carried out with constant reaction
rate and without yield or enantiomeric excess (ee) reduction. In
industrial practice, the goal is always to operate a reaction at the
highest, practical S/C ratio, and thus the next stage of development
was to optimise the S/C ratio. The limiting, practical S/C ratio
was found to be 7000. The process was then run with the optimised
dilute reaction system, which allowed 10 reaction cycles to be
carried out with only 20% addition of the initial (reaction 1) mass
of catalyst to reactions 2-10 to maintain reaction rate, conversion,
and ee. This leads to a 5 times increase in the overall S/C ratio. In
addition, this optimisation reduced the amount of metal present
in the product solution from 130 µg Ru per gram of product to
<6 µg Ru per gram of product. The process was then scaled up
to an industrially relevant substrate concentration (20 wt % DMI
1. Introduction
Many homogeneous organic syntheses are catalysed by
transition metals complexed with organic ligands.1-4 Many
transition metal catalysts (TMCs) are expensive, and the organic
ligands can be difficult to synthesise, which may lead to the
ligands being more valuable than the metal, especially if they
are chirally directing.5 A major drawback of using TMCs is
that they must be separated from the product at the end of the
reaction. Generally, a pure (metal-free) product is only obtained
with energy intensive and/or waste-generating downstream
processing that destroys the catalyst and reduces the final
product yield.6,7 Even after this extensive, destructive postre-
action processing8-10 only a fraction of the metal value can
typically be recovered by returning separation residues to the
catalyst suppliers, while the ligands are irreversibly lost. A
nondestructive separation method to recover the catalyst in an
active form could remove these disadvantages of using TMCs
by providing a product stream with reduced metal content and
allowing the catalyst value to be maximised through its reuse.
Such technology would be particularly useful in regulated
industries where metal content in products is tightly controlled.
For example, in pharmaceutical products the daily dosage of
transition metals must be minimised to prevent any unwanted
(1) Davies, S. G. Organotransition Metal Chemistry Applications to
Organic Synthesis; Pergamon: Oxford, 1982.
(2) Parshall, G. W.; Ittel, S. D. Homogeneous Catalysis. The Applications
and Chemistry of Catalysis by Soluble Transition Metal Complexes,
2nd ed.; John Wiley & Sons: New York, 1992.
(3) Akutagawa, S. Appl. Catal., A 1995, 128, 171–207.
(4) Hegedus, L. S. Transition Metals in the Synthesis of Complex Organic
Molecules, 2nd ed.; W. H. Freeman; New York, 1999.
(5) Kagan, H. B.; Sasaki, M. In Optically ActiVe Phosphines; Preparation,
Uses and Chiroptical Properties; Hartley, F. R., Ed.; The Chemistry
of Organophosphorous Compounds, Vol. 1; Wiley: New York, 1990.
(6) Heck, R. F. Palladium Reagents in Organic Syntheses; Academic Press:
New York, 1985.
* Corresponding author. Telephone: +44 (0)20 8902 0040. Fax: +44 (0)20
8902 0001. E-mail: atb@membrane-extraction-technology.com.
† Membrane Extraction Technology Ltd.
‡ GlaxoSmithKline.
(7) Jo¨dicke, G.; Zenklusen, O.; Weidenhaupt, A.; Hungerbu¨hler, K. J. J.
Cleaner Prod. 1999, 7, 159–166.
§ Imperial College of Science, Technology and Medicine.
⊥ Merpro Limited.
(8) Trost, B. M. Science 1991, 254, 1471–1477.
(9) Sheldon, R. A. CHEMTECH 1994, 38–48.
¶ Katholieke Universiteit Leuven.
(10) Curzons, A. D.; Constable, D. J. C.; Mortimer, D. N.; Cunningham,
V. L. Green Chem. 2001, 3, 1–6.
0 W.R. Grace and Co.
10.1021/op900056s CCC: $40.75  2009 American Chemical Society
Published on Web 08/11/2009
Vol. 13, No. 5, 2009 / Organic Process Research & Development
•
863

side effects. In particular, the platinoid group metals (Pt, Pd,
Ru, Rh, etc.) have particularly low concentration specifications,
10 µg or less of metal per g product, due to their toxicity.11
Organic solvent nanofiltration (OSN) is a membrane technology
that can be used to achieve this process concept.
OSN is a membrane technology capable of retaining large
homogeneous TMCs, usually MWTMC > 500 Da, from postre-
action organic solvent mixtures,12 while allowing smaller
product molecules to permeate through the membrane. As many
TMCs are significantly larger than the reactants they act on and
the product molecules they generate, it is possible to use OSN
to retain the catalyst molecule, whilst allowing the product and
unused reactants to permeate through the membranes.
Several authors have studied the use of homogeneous TMC
recovery and recycle using membrane-coupled catalytic systems,
with the first reported study by Gosser et al.13 published in 1977.
A summary of the subsequent application of this technique is
provided in Vankelecom et al.14 and Vandezande et al.15 These
studies have typically been carried out at dilute, “non-industrial”
substrate loadings, with the aim of understanding the maximum
turnover number (TON) for a particular catalytic system. A
different perspective on the technology is provided by Nair et
al.,16 who have investigated the final product purity attainable
using OSN recycle technology for a Heck coupling reaction.
In many industrial applications, achieving a sufficient purity
(i.e., low metal level in the product) is at least as important as
the ability to recover and reuse the catalyst.
To date, catalyst recycle using OSN membranes has only
been reported in the literature with systems using dilute substrate
concentrations (typically <1 wt %) and the influence of higher,
more industrially relevant substrate concentrations (>5 wt %)
on this process has not been investigated, and it remains unclear
whether scale-up of OSN processes by increasing substrate
concentration affects process performance.
In this report we demonstrate the increase in catalytic
productivity possible through TMC recycle in a batch-operated
OSN cell, allowing the TMC to be reused for multiple
consecutive reactions. The enantioselective hydrogenation of
dimethyl itaconate (DMI) to dimethyl methylsuccinate (DMMS),
catalysed by Ru-BINAP, was selected as the model reaction.
The reaction scheme and TMC are shown in Figure 1. Excellent
performance in asymmetric hydrogenation reactions has been
observed using Ru-BINAP,17,18 and it is known to catalyse
Figure 1. DMI hydrogenation catalysed by Ru-BINAP.
industrially important reactions;19,2,3,20 thus, we chose to use it
for this study. The usefulness of the OSN technology for TMC
recycle and obtaining a high purity product will be examined
at high and low TMC loadings at 0.8 wt % DMI concentrations.
The process will then be scaled up with a low TMC loading to
20 wt % DMI, to demonstrate that the process can be a useful
tool at industrially relevant substrate loadings.
2. Experimental Section
2.1. Chemicals. Methanol (99+ %), dimethyl itaconate
(97%) and dimethyl methylsuccinate (98%) were all obtained
from Aldrich Chemical Co. (Poole, Dorset, U.K.). (R,R)-Ru-
BINAP was obtained from Strem Chemicals (Kehl, Germany).
All preparations were carried out using standard Schlenk
techniques in a dry nitrogen atmosphere. The solvent was dried
by standard methods prior to use, to exclude moisture. All other
starting materials were used as supplied.
2.2. Analytical Techniques. Concentrations of dimethyl
itaconate (DMI) and dimethyl methylsuccinate (DMMS) were
determined using a Shimadzu GC-14A gas chromatograph with
a flame ionisation detector (FID) equipped with a BP1 capillary
column (length 25 m, 0.32 mm i.d.) (SGE, Australia). The oven
temperature was held at 50 °C for 2.5 min and then increased
to 300 °C at a rate of 25 °C min-1. The coefficient of variation
of the assay was 1.2% for DMI concentration fixed at 0.8 wt
%. A 0.1 mL amount of organic sample was evaporated and
then dissolved in 10 mL of a 10 wt % aqueous nitric acid
solution for ICP spectroscopy. Concentrations of Ru and P
were measured with a Jobin Yvon Ultima spectrometer.
Coefficients of variation for Ru and P, measured at initial
experimental concentrations, were 0.69% and 0.61%
respectively. Enantiomeric excess was measured using a
Unicam Crystal 200 HPLC system. The column used was a
CHIRALCEL OD-H, 250 mm long and 4.6 mm diameter,
containing 5 µm packing. The mobile phase, 9:1 hexane:
IPA, was pumped through the column at a flow rate of 500
µL min-1, and a diode array detector at a fixed wavelength
of 200 nm was used to detect the elluents. The major product
from the reaction is the S enantiomer, and the ee of the S
enantiomer is calculated as ([S] - [R])/([S] + [R]).
(11) Note for Guidance on Specification Limits for Residues of Metal
Catalysts. European Agency for the Evaluation of Medicinal Products
(EMEA), London, 2002; http://www.emea.europa.eu/pdfs/human/swp/
444600en.pdf.
(12) Scarpello, J. T.; Nair, D.; Freitas dos Santos, L. M.; White, L. S.;
Livingston, A. G. J. Membr. Sci. 2002, 203, 71–85.
(13) Gosser, L. W.; Knoth, W. H.; Parshall, G. W. J. Mol. Catal. 1977, 2,
253–263.
(14) Vankelecom, I. F. J. Chem. ReV. 2002, 102, 3779–3810.
(15) Vandezande, P.; Gevers, L. E. M.; Vankelecom, I. F. J. Chem. Soc.
ReV. 2008, 37, 365–405.
2.3. OSN Membrane and Filtration/Reaction Cells.
Samples of Davison Membranes STARMEM 122 (STARMEM
is a trademark of W.R. Grace), an integrally skinned, asym-
(16) Nair, D.; Scarpello, J. T.; Vankelecom, I. F. J.; Freitas dos Santos,
L. M.; White, L. S.; Kloetzing, R. J.; Welton, T.; Livingston, A. G.
Green Chem. 2002, 4, 319–324.
(17) Akutagawa, S. In Asymmetric Hydrogenation with Ru-BINAP: Chiral-
ity in Industry; Collins, A. N., Sheldrake, G. N., Crosby, J., Eds.;
Wiley: London, 1992.
(19) James, B. R. Homogeneous Hydrogenation; Wiley InterScience: New
York, 1992.
(20) Stinson, S. C. Chem. Eng. News 1998, 76 (38), 83–104.
(18) De Smet, K.; Aerts, S.; Cuelemans, E.; Vankelecom, I. F. J.; Jacobs,
P. A. Chem. Commun. 2001, 39 (10), 597–598.
864
•
Vol. 13, No. 5, 2009 / Organic Process Research & Development

the reaction (as shown in Figure 2a). Regular sampling was
carried out until a 95-100% conversion of DMI had been
obtained. The pressure was then increased to 30 bar, and the
OSN permeation valve was opened (as shown in Figure 2b) to
allow the solvent and DMMS to permeate across the membrane
under the applied pressure into a collection receptacle. Once
40 mL (80% of the reaction volume) had permeated, the cell
was depressurised. The time, t, to collect the permeate volume,
Vp, was recorded, and the permeate was sampled for analysis.
A fresh 40 mL feed solution was prepared under inert
conditions and fed to the OSN reactor to replenish the consumed
substrate and restore the original 50 mL reaction volume (Figure
2c), again under inert conditions. The cell was then closed and
allowed to reach reaction temperature, and the hydrogen valve
was opened to reinitiate the reaction. This TMC recycle
procedure was repeated multiple times. A single membrane disk
was used in each recycle system, since the feed solution or
retentate would always be in contact with the membrane until
the end of a sequence, thus preventing damage to the membrane
through drying out.
2.5. Detailed Reaction Compositions. Five different reac-
tion solutions were used during this development work. The
initial and fresh feeds for Systems 1-5, with substrate/catalyst
(S/C) ratios, are given below. In all cases, reactions were
conducted at 30 °C. In System 2, 80% of the postreaction
volume was not filtered but poured away and instead diluted
with the fresh feed. In all other systems the TMC was recycled
by the procedure outlined in section 2.4.
Reaction System 1 (initial S/C ) 500): Initial Feed ) 0.337
g (2.13 mmol) DMI, 0.0034 g (0.00427 mmol) Ru-BINAP, 50
mL MeOH; Fresh Feed ) 0.337 g (2.13 mmol) DMI, 40 mL
MeOH. TMC recycled by OSN.
Reaction System 2 (initial S/C ) 500): Initial Feed ) 0.337
g (2.13 mmol) DMI, 0.0034 g (0.00427 mmol) Ru-BINAP, 50
mL MeOH; Fresh Feed ) 0.337 g (2.13 mmol) DMI, 40 mL
MeOH. TMC not recycled by OSN.
Reaction System 3 (initial S/C ) 7000): Initial Feed ) 0.337
g (2.13 mmol) DMI, 0.00024 g (0.00031 mmol) Ru-BINAP,
50 mL MeOH; Fresh Feed ) 0.337 g (2.13 mmol) DMI, 40
mL MeOH. TMC recycled by OSN.
Reaction System 4 (initial S/C ) 7000): 0.337 g (2.13 mmol)
DMI, 0.00024 g (0.00031 mmol) Ru-BINAP, 50 mL MeOH;
Fresh Feed ) 0.337 g (2.13 mmol) DMI, 0.000048 g (0.000061
mmol) Ru-BINAP, 40 mL MeOH. TMC recycled by OSN.
Reaction System 5 (initial S/C ) 4000): 16 g (94.9 mmol)
DMI, 0.02 g (0.0252 mmol) Ru-BINAP, 64 g MeOH; Fresh
Feed ) 16 g (94.9 mmol) DMI, 0.006 g (0.0076 mmol) Ru-
BINAP, 64 g MeOH. TMC recycled by OSN.
2.6. Experimental Calculations. For each reaction-separa-
tion cycle, all calculations were based on volume permeated,
Vp, permeation time, t, and measurement of Ru, P, and DMMS
concentrations in each permeate and in the final retentate.
Reaction performance was assessed in terms of turnover
number (TON). TON, a measure of catalyst efficiency, was
calculated as:
Figure 2. TMC recycle sequence “React”-“Filter”-“Re-fill”
using OSN (shaded valves are closed).
metric OSN membrane with a polyimide active surface, were
provided by Membrane Extraction Technology Ltd. (London,
U.K.). STARMEM 122 has a nominal molecular weight cutoff
(MWCO) of 220 Da (manufacturer values, using toluene as a
solvent and based on 90% rejection of n-alkanes). Prior to use,
each new membrane disk was conditioned (prepared for use)
by permeating pure methanol under 30 bar pressure until a
constant flux was obtained. This was to ensure complete
removal of the preserving agent (lube oil) and minimise the
influence of flux decline, due to irreversible membrane compac-
tion, on the results.21
Reaction and filtration of the dilute (0.8 wt %) DMI systems
were performed in a Sepa ST cell (GE Osmonics, Minnetonka,
MN, U.S.A.) set up in an oil bath controlled to within 0.1 °C
of the required reaction temperature. The stainless steel OSN
cell was mixed at 900 rpm with a PTFE stirrer bar. Membrane
disks of 49 mm diameter were mounted in the pressure cell
using a FEP-coated ring-seal. The cell was connected to a
hydrogen cylinder (to provide a necessary reaction substrate
and the pressure for nanofiltration) and filled with a degassed
feed solution. The available membrane area for filtration, Am,
was 1.69 × 10-3 m2.
Reaction and filtration for the concentrated (20 wt %) DMI
system was performed in a METcell (Membrane Extraction
Technology Ltd., London, U.K., www.membrane-extraction-
technology.com). The temperature was controlled by switching
the hot plate section of a magnetic stirrer/hot plate on and off,
based on readings from a thermocouple inserted into the reaction
mixture. The temperature was controlled to within 0.1 °C of
the required reaction temperature. The METcell was mixed at
450 rpm with a dedicated PTFE stirrer disk using a magnetic
stirrer. Membrane disks of 90 mm diameter were mounted in
the pressure cell using a FEP-coated ring-seal. The cell was
connected to a hydrogen cylinder (to provide a necessary
reaction substrate and the pressure for nanofiltration) and filled
with a degassed feed solution. The available membrane area
for filtration, Am, was 5.1 × 10-3 m2.
2.4. TMC Recycle Procedure. The TMC recycle scheme
is shown in Figure 2 and can be considered as three stages per
cycle: react, filter, and refill. First, the reaction feed solution
was placed into the cell (containing a conditioned membrane
disk) under inert conditions. The reaction solution was then
allowed to reach the reaction temperature, and the hydrogen
valve was opened to pressurise the cell to 10 bar and initiate
total moles DMI converted over all cycles
TON )
(
)
total moles Ru fed over all cycles
(1)
(21) White, L. S.; Nitsch, A. R. J. Membr. Sci. 2000, 179, 267–274.
Vol. 13, No. 5, 2009 / Organic Process Research & Development
•
865

The membrane performance was assessed in terms of solute
rejection (eq 2) and solvent flux (eq 3). The rejection of species
i, reji, for a given membrane is a measure of its separation
performance and is defined by:
Cpi
reji ) 1 -
× 100%
(2)
(
)
Cri
where Cpi and Cri are the concentration of solute i in the
permeate and retentate respectively, as measured at the end of
a batch filtration. A rejection of 100% corresponds to perfect
retention of the species and 0% to no separation.
The flux, J (measured in L m-2 h-1), was calculated using:
Figure 3. DMI conversion in Reaction System 1 for the 1st
and 14th reactions.
Vp
J )
(3)
(
)
Amt
where Am is the area of the membrane disk used in the filtration
(Vp and t are defined above). The flux is the volume of solution
permeated per unit membrane area per unit time at a given trans-
membrane pressure.
The mass balance error (∆MB) was calculated for Ru, P,
and DMMS at the end of a sequence of reactions and
separations using the following equation:
Xiin + Xigen - Xiperm - Xriet
∆MB )
× 100%
Xiin + Xigen
(
)
(4)
where Xiin, Xgi en, Xpi erm, and Xriet are the total moles of species i
respectively input, generated, permeated, and retained during
DMMS
DMMS
Figure 4. Rejection of TMC species and product purity in
Reaction System 1.
the course of the reactions. Clearly, X
) 0, and only X
in
gen
* 0. ∆MB was within (7% for all species in every system.
As it is not possible to measure directly the concentration
of intact, active Ru-BINAP, the concentrations of Ru and P
are measured as indicators of the catalyst-related species present
in a particular solution. It should be noted that this technique
does not distinguish between active catalyst, deactivated catalyst,
and catalyst degradation products. Figure 4 shows Ru and P
rejection and product purity (defined as mass of Ru per unit
mass of product) after each filtration in Reaction System 1. Ru
rejection decreased from 98.8 to 76.8% during the sequence of
filtrations, while P rejection decreased from 99.6 to 96.0%. It
is thought that under the reaction conditions the TMC fragments
into separate metal and ligand species, with the resulting metal
species more easily able to permeate the membrane than the
ligand-related species. As P, rather than the whole ligand group,
was detected using ICP, it was not possible to determine whether
the BINAP ligand had degraded or was still intact. It is expected
that if the catalyst remains intact, then the molar ratio of P and
Ru in the permeate should be the same as the fresh feed of
Ru-BINAP (i.e., Ru:P ) 1:2). From the change in Ru and P
permeate concentrations over time (Figure 5), it was evident
that they were not permeating in the ratio expected for intact
Ru-BINAP and in fact the final permeate Ru:P ratio at 1:1.3
was significantly less than the expected 1:2. This was confirmed
through both the mass balances and the measured Ru:P ratio
of 1:8.1 in the retentate, which should also be 1:2 if the Ru-
BINAP was permeating intact. To test if any of the Ru species
3. Results and Discussion
3.1. Proof of Concept (Reaction System 1). The first aim
of this work was to demonstrate that the TMC catalyst recycle
concept was feasible when using Ru-BINAP. This initial work
was carried out with Reaction System 1.
Fourteen consecutive reaction-separation cycles were at-
tempted, with each reaction taking approximately 2.5 h to reach
complete conversion of DMI to DMMS. The conversion profiles
over time for the first and fourteenth reactions demonstrated
that no reaction rate decline was occurring during this sequence
(Figure 3), and a cumulative TON of 6950 was reached.
Calculations of DMMS rejection for System 1 showed that it
increased from 49% for the first filtration to 70% after five
filtrations, and ultimately reached 73% for the last eight
filtrations. This is most likely caused by irreversible compaction
of the membrane during the first six filtration cycles. After these
six filtrations, the compaction is complete, and no further change
in DMMS rejection is observed. At steady state, with a DMMS
rejection of 73%, it is expected that almost two reaction
equivalents of DMMS will be held up in the reactor at steady
state, while only one reaction equivalent is permeated. After
the ninth reaction in the sequence, it was noted that the
postreaction concentration of DMMS became constant at a value
of 1.9 equiv of DMMS, as expected. During the filtrations, the
permeate flux averaged 35.9 L m-2 h-1.
866
•
Vol. 13, No. 5, 2009 / Organic Process Research & Development

Figure 6. DMI conversion and cumulative TON for Reaction
System 2.
Figure 5. Permeate concentrations of TMC species in Reaction
System 1.
in the permeate was intact catalyst or demonstrated any catalytic
activity, fresh DMI/methanol solution was added back to
samples of the permeate, hydrogen pressure applied, and the
conversion monitored. No conversion was obtained, which
suggests that the permeated TMC species was no longer
catalytically active.
Metal contamination of the product (product purity) only
reaches steady state when both DMMS and the metal permeate
are at their respective steady state values. Since metal perme-
ation rates were increasing over time and DMMS concentration
increased over the first nine reactions, a steady state value was
not reached. Product purity for the first two reactions was close
to 120 µg Ru per g DMMS, which improved to 65 µg Ru per
g DMMS (due to increasing DMMS permeation) and then
worsened to 120-140 µg Ru per g DMMS as Ru permeation
across the membrane increased. It should be noted that, without
filtration, 1309 µg Ru per g DMMS would be expected in the
postreaction mixture, showing that even with this nonoptimised
reaction/filtration system there is significant benefit in product
purity using OSN.
Figure 7. DMI conversion and cumulative TON for Reaction
System 3.
reaction, with an S/C ratio of 12,500 started to slow down
significantly and achieved only 80% conversion in 10 h. At an
S/C ratio between 5000 and 12,500, the reaction appeared to
become controlled by the limiting catalyst concentration. A
cumulative TON (based on the original TMC loading and the
total DMMS produced in the 3 reactions) of 1400 was achieved
in a total of 15 h. The product purity in the produced DMMS
streams from reactions 1, 2, and 3 by calculation would be
1,309, 262, and 65 µg Ru per g DMMS, respectively. The
product purity was improving due to dilution as TMC was lost
from the system in each unfiltered permeate volume.
By the final reaction, the working S/C ratio had increased
from an initial value of 500 to 5000, due to TMC losses. This
increase in S/C ratio was not observed to be reaction rate
limiting.
This data demonstrated that membrane recycle of the TMC
was important to achieve more than a single repeated reaction.
It also suggested that the reaction could be achieved without
reaction rate decline at an S/C ratio between 5000 and 12,500.
3.3. Verification of S/C ) 7000 as Limiting Catalyst
Concentration (Reaction System 3). An S/C ratio of 7000
was chosen because it not only fulfilled the requirement of being
close to the rate-limiting S/C ratio, but it also represented a
system that could match the cumulative TON of the initial S/C
) 500 catalyst recycle system with a single reaction (instead
of 14 reactions). The results of testing this S/C ratio are shown
in Figure 7. The initial reaction took 3.5 h (an hour longer than
with the S/C ) 500 system) to reach 100% conversion,
indicating that the catalyst loading was indeed rate limiting.
The second reaction reached 88% conversion in just under 10 h,
giving a cumulative TON of 13,100 in just under 14 h. The
significant reaction rate decline during the second reaction
suggested that TMC losses during filtration and/or catalyst
deactivation were sufficient to increase the working S/C ratio
above the reaction rate-limiting value.
3.2. Optimisation of S/C Ratio (Reaction System 2).
Having demonstrated the proof of concept for the process, the
next stage of work was to optimise the S/C ratio. Reaction
System 2 was used to investigate the effect of S/C ratio on
conversion. The work carried out for this optimisation was very
similar to the proof-of-concept work, except that rather than
filtering the postreaction mixture, 80% of postreaction liquid
was simply poured out of the filtration cell rather than carrying
out a filtration. That is, every reaction contained only 20% of
the catalyst present in the previous one, with the exception of
the initial reaction. The results of this experiment are shown in
Figure 6.
The initial reaction, with an S/C ratio of 500, reached
complete conversion of DMI after 2.5 h (as expected from the
proof-of-concept results). The second reaction, with an S/C ratio
of 2500 also took only 2.5 h to reach completion. This was
expected, since the proof-of-concept work demonstrated that
an S/C ratio of 5000 was not rate limiting. However, the third
Rejection of Ru was 98.4% for the filtration between the
initial and second reactions, which gave a product purity of
Vol. 13, No. 5, 2009 / Organic Process Research & Development
•
867

Figure 9. DMI conversion and ee for 20 wt % DMI reactions
(Reaction System 5).
Figure 8. Ru rejection and product purity for Reaction Sys-
tem 4.
During the filtrations, the permeate flux averaged 35.7 L m-2
h-1. A cumulative TON of 40,900 over the 10 reactions (taking
into account each new catalyst addition) was achieved in 35 h.
Ru rejection and product purity for the S/C ) 7000 recycle
system is shown in Figure 8. Ru rejection did not decrease
below 97.5% during the reaction sequence, although product
purity increased from 7.9 to 16 µg Ru per g DMMS. The
increase in Ru content in the product is due to a fraction of the
Ru added to each reaction accumulating in the system, until
steady state is reached. At steady-state, the mass of Ru fed per
batch equals the rate at which the Ru permeates. By assuming
100% conversion of DMI to DMMS, i.e. at no point is the rate-
limiting S/C ratio exceeded, and that steady-state permeation
rates of DMMS and Ru can be achieved, 18.4 µg Ru per g
DMMS would be expected in the product at steady state. From
the results in Figure 8, it can be seen that the product purity is
tending towards this value. If an additional 5-10 reactions had
been carried out, it is expected that the steady-state product
purity value would have been attained.
Again, during these 10 reactions it was observed that Ru
permeated faster than expected from stoichiometry, based on
the P permeation rate (mean Ru:P ratio during the 10 reactions
was 1:1.27, stoichiometry predicts Ru:P ) 1:2), suggesting that
degradation of the catalyst was still occurring.
3.5. Scale-Up to Industrially Relevant Substrate Con-
centration (Reaction System 5). To maximise productivity in
commercial applications, as high a substrate concentration as
practically feasible is used. For asymmetric hydrogenations, this
is typically in the range 10-40 wt % substrate. Thus, following
demonstration and optimisation of the OSN catalyst recycle
system with the dilute (0.8%) DMI solution, the process was
scaled up to the more industrially relevant substrate concentra-
tion of 20 wt % DMI.
7.1 µg Ru per g DMMS in the filtered product solution. In
comparison, without filtering the product solution and assuming
100% conversion, a product purity of only 89.1 µg Ru per g
DMMS would have been obtained. Certainly for pharmaceutical
applications, this Ru concentration from the unfiltered solution
would be regarded as too high, and further workup would be
required, whereas the product from the filtered solution would
not require additional workup.
3.4. Application of Optimised S/C Ratio in OSN Catalyst
Recycle Process (Reaction System 4). The rapid reaction rate
decline observed during the second reaction of the S/C ) 7000
verification test is avoidable only by maintaining the TMC
loading above the reaction rate-limiting concentration. This can
be achieved by adding a fraction of the initial catalyst mass to
the system with each successive batch of fresh reactant solution.
In this way, the TMC loading per batch can be kept at a
relatively low level, such that the membrane can still provide
low metal concentration in the product, whilst preventing
reaction rate decline and reaction stalling observed in cases
where the S/C ratio becomes rate limiting.
In approaching the problem of what fraction of the TMC
added to the initial reaction batch should be added to each
subsequent recycle batch, it seemed logical to simply replenish
the amount of TMC observed leaving the system through the
permeate in each reaction cycle. In the reaction verification work
at S/C ) 7000, only 8.6% of the Ru fed to the system was
detected in the permeate. Hence, 10% of the initial mass of
TMC was added to each subsequent reaction, operating with
an initial S/C of 7000. Although some improvement in reaction
rate was observed, the second reaction still suffered from
reaction rate decline, stalling at 95% conversion after 7 h (data
not shown). This suggested that some of the Ru species retained
by the membrane was also deactivated or degraded and needed
replacing by the additional catalyst.
Recognising that TMC degradation may be higher than that
implied by the concentration of permeated Ru alone, the amount
of TMC added to reaction 2 onwards was 20% of the initial
catalyst mass. By applying this strategy, 10 consecutive reac-
tions (including nine catalyst recycles) were conducted with
complete conversion and no reaction rate decline.
It was observed during initial tests that at this higher substrate
concentration a lower S/C ratio (S/C ) 4000 rather than S/C
) 7000 for the dilute system) was required to maintain the same
reaction batch time as was attained with the optimised dilute
system. Therefore, demonstration of the catalyst recycle system
at 20 wt % DMI was carried out at the higher catalyst loading
of S/C ) 4000.
The amount of catalyst added to the reaction system with
each batch of fresh substrate solution was reduced sequentially
from 30% initial catalyst mass, to 20% initial catalyst mass,
and finally to 10% initial catalyst mass. The goal was to verify
that the fraction of initial catalyst mass addition identified in
DMMS rejection increased, as observed during the S/C )
500 proof of concept work, from 64.3% to 89.5%. As in the
proof of concept work, this is believed to be due to irreversible
compaction of the membrane during the first five filtrations.
868
•
Vol. 13, No. 5, 2009 / Organic Process Research & Development

Figure 10. Steady-state mass balance for one reaction-filtration cycle for Reaction System 5 with 20% of initial catalyst loading
added per reaction-filtration cycle.
the dilute system optimisation was also appropriate for the 20
wt % DMI system. By bracketing the previously identified 20%
readdition rate, it would be possible to verify if this addition
rate still remained optimal.
sary. By reducing and optimising the catalyst loading (S/C )
7000), product purities <20 µg Ru per g DMMS could be
achieved, potentially removing the need for any further pos-
treaction processing, and the turnover number of the catalyst
was increased from 6950 (14 reactions at S/C ) 500) to 13,100
(two reactions at S/C ) 7000 with no readdition of catalyst).
The TON was further increased to >40,000 by feeding small
amounts of the TMC to each repeat reaction to maintain an
acceptable reaction rate with the recycled TMC. The optimal
solution for high TON and product purity required system
operation close to the S/C limiting ratio (working S/C ) 7000)
with a small addition of TMC to compensate for catalytic and
filtration losses (20% addition of initial catalyst charge). Clearly,
a completely stable catalyst that is present at a reaction rate-
limiting S/C ratio, and completely retained by a membrane,
should require no further addition of catalyst after the initial
loading. Nevertheless, in the system presented in this study the
requirement to add only 20% of the initial catalyst load to all
subsequent reactions is equivalent to a 5-fold increase in catalyst
life and a 5-fold reduction in catalyst quantity/cost—this
represents a significant cost saving in processes where TMCs
are used.
Figure 9 shows the evolution of DMI conversion and DMMS
ee over the course of 20 reactions, incorporating 19 catalyst
recycles. The desired end-point of the reaction during each cycle
was DMI conversion >95% and DMMS ee >80%. The first 10
reaction/recycles were carried out with the maximum readdition
of catalyst, 30% addition of the initial catalyst mass. During
these 10 reactions, the conversion averaged 96.5% and the ee
averaged 87.3%, with the conversion and ee stabilising at 99%
and 93%, respectively, by reactions 7-10. Given this successful
demonstration with 30% catalyst readdition, the readdition
fraction was reduced to 20% for reactions 11-19. With 20%
readdition, the conversion and ee averaged 98.8 and 85%,
respectively, which is comparable with the results for 30%
readdition. However, although the conversion is constant with
20% catalyst readdition, there does appear to be a trend of
reducing ee after reaction 15, perhaps due to degradation of
the chirally directing BINAP ligand. Figure 10 shows the steady-
state mass balance for each reaction-filtration cycle at these
operating conditions. Also, as reaction 20 (Figure 9) shows,
adding only 10% of catalyst has a significant affect on the
conversion in addition to the ee. As both conversion and ee
were both below the desired reaction end-point values, the
recycles were terminated at reaction 20. During these reaction/
filtrations, the permeate flux averaged 40 L m-2 h-1.
The overall productivity was improved by increasing the
DMI concentration by an order of magnitude from 0.8 wt % to
20 wt %, showing that application of this technology at
industrially relevant substrate concentrations is feasible.
Acknowledgment
S.H. and A.G.L. acknowledge the support of Engineering
and Physical Sciences Research Council, U.K., Grant Number
GR/R94725/01 during this work. H.-t.W. acknowledges the
support of the Croucher Foundation, Hong Kong.
4. Conclusions
It has been shown that applying OSN technology to recycle
a homogeneous TMC (Ru-BINAP) has substantially increased
the catalyst TON and lifetime for a hydrogenation reaction. The
membrane performance at high catalyst loadings (S/C ) 500)
allowed sufficient catalyst to permeate through the membrane
that further downstream product purification would be neces-
Received for review March 11, 2009.
OP900056S
Vol. 13, No. 5, 2009 / Organic Process Research & Development
•
869