RSC Advances

PAPER

aVITO (Flemish Institute for Technological

Technology, Boeretang 200, B-2400 Mol, B

be; Fax: +32 14 32 1186; Tel: +32 14 33 56bUniversity of Antwerp, Organic Synthesis, G

Belgium

† Present address: Janssen PMP, Turnhou

Cite this: RSC Adv., 2013, 3, 21501

Received 4th July 2013Accepted 9th September 2013

DOI: 10.1039/c3ra44860f

www.rsc.org/advances

This journal is ª The Royal Society of

Separation of metathesis catalysts and reactionproducts in flow reactors using organic solventnanofiltration

Dominic Ormerod,*a Bas Bongers,a Wim Porto-Carrero,a Saly Giegas,a Glenn Vijt,a

Nicolas Lefevre,b Dirk Lauwers,†a Wilfried Brustena and Anita Buekenhoudta

Organic solvent nanofiltration (OSN), a relatively new low energy separation technology, has been used to

reduce metal contamination of ring closing metathesis reaction products. The catalysts used were readily

available commercial Hoveyda–Grubbs and Umicore M series catalysts. These reactions were performed in a

flow reactor with in-line membrane separation, and high catalyst retention can be achieved. In the flow

reactor set up a beneficial effect on catalyst life-time on changing from solvents such as dichloromethane

to environmentally more benign acetone, which reduces initiation rates, was demonstrated.

Introduction

With the emergence during the last few decades of stable cata-lysts, olen metathesis has become increasingly popular, a factrecognized in 2005 by the award of the Nobel Prize in chemistryto Yves Chauvin, Robert H. Grubbs and Richard H. Schrock.Indeed, the popularity is such that several catalysts are nowcommercially available and a number of industrial examples ofring closing metathesis of high value molecules can be found inthe literature.1 In spite of the fact that these catalysts are widelyused, the complete removal of residual ruthenium species aerreaction is not only highly desirable but it can also be somewhatproblematic. This being especially the case for transformationsused within the pharmaceutical industry in which daily exposureto metal residues in drug substances must not exceed 100 mg fororal administration and 10 mg for parenteral administration.2

This translates into less than 10 ppm in oral formulations and 1ppm in parenterally administered formulations. As a result, overthe years several methods have been developed to remove Rufrom post reaction mixtures. These methods have recently beenreviewed by Vougioukalakis.3 More recent examples publishedaer the review includes immobilisation on magnetically sepa-rable iron oxide particles4 and the inclusion of a quaternaryammonium group in the N-heterocyclic carbene ligand.5 Unfor-tunately many of these methods are either destructive to thecatalyst or render them in a formwhichmakes recovery and reuseeconomically impractical.

Research), Separation and Conversion

elgium. E-mail: dominic.ormerod@vito.

50

roenenborgerlaan 171, B-2020 Antwerp,

tseweg 30, B-2340 Beerse, Belgium.

Chemistry 2013

The use of membranes not only offers the possibility ofseparating these catalysts from reaction components withoutthe need to deactivate them. It is also a technique that can alsobe incorporated into ow reactors,6 which in themselves cangive a substantial improvement in chemical manufacturingsustainability.7 Membrane separations are also low energy andthus oen considered green separations.8 The prerequisitebeing the membranes are stable in organic solvents. Organicsolvent nanoltration (OSN) is a pressure driven ltrationprocess capable of separating molecules in the molecularweight range of 200–1000 Da. This emerging technology hasbecome more popular within the last 10 to 15 years with thedevelopment of membranes that are stable in organic solvents,some of which are now commercially available. For an overviewof membranes and their uses in organic solvents we refer to thereview by Vankelecom and coworkers.9

Because of this, numerous groups have applied the emergingtechnology that OSN is, to ruthenium catalysed metathesis reac-tions.10 However, the overwhelming majority of these studiesmake use of molecular weight enlarged11 catalysts in order tofacilitate their separation via OSN.Whereas this approach is bothviable and interesting; the present lack of commercial availabilityof these enlarged catalysts does present an obstacle to theirimplementation on an industrial scale. Therefore, from anindustrial view point it would be better to use a readily availablecommercial catalyst whilst still maintaining the objective of highpartition of the catalyst from reaction products.

Two of the parameters generally used to characterizemembrane performance are solute rejection and permeate ux.12

The rejection of a solute across a membrane is a measure of itspartitioning. Thus, high rejection will lead to good partitioningand consequently easy separation. As rejection tends to zero,then the ability to separate the solute via OSN also becomes less

RSC Adv., 2013, 3, 21501–21510 | 21501

RSC Advances Paper

and less. Solute rejection (R) is calculated from the equationbelow, where Cp is the concentration of the solute in the permeateand Cr is the concentration of the solute in the retentate.

R ¼�1� Cp

Cr

�� 100

The ux of the various solvent and solute mixtures wasdetermined by measuring the permeate volume (V) per unit time(t) where A is the effective membrane area using the equation:

J ¼ V

At

The mechanism of solute rejection in OSN is not simply sizeexclusion;13 solubility parameters of the solvent, solute andmembrane surface also play an important role. A fact that canbe used to advantage, especially when applied to ow reactors, areactor conguration that is becoming increasingly popular inboth industry and academia.14 In this paper we investigate thepossibility of using readily available commercial catalyst in acontinuous ow reactor in which there is in-line separation ofcatalyst from reaction product. In such a reactor the idealsituation would be the combination of a membrane that giveshigh retention of the catalyst and as low as possible retention ofthe reaction product. Also the long term catalyst stability willbecome more important than in a batch reaction as this needsto remain stable during the addition of substrate. A time spanthat can easily exceed the time a catalyst is stable when usedunder batch conditions. Furthermore, if the catalyst can besuccessfully separated from reaction products without deacti-vating it before isolation of the reaction product; then concep-tually at least the catalyst turnover number can be increased.

Results and discussionReaction, catalyst and membrane choice

In order to determine the best combination of catalyst,membrane and solvent usable in a ow reactor, a suitablereaction is required. For this purpose the ring closing metath-esis of diethyldiallyl malonate (dedam) was chosen, a reactionoen used to characterise olen metathesis catalysts.15 Cata-lysts used in this study were primarily the second generationUmicore M2 and M51 catalysts, the rst generation Hoveyda–Grubbs and the latent catalyst Umicore M41 (Scheme 1).

In the initial experiments the emphasis was placed oncatalyst removal. Two initial experiments were performed in asemi-continuous, reaction–lter–ll,16 mode shown schemati-cally in Fig. 1a. Thereaer all reactions were performed in amore continuous manner shown schematically in Fig. 1b.

Catalyst retention experiments were performed using tightnanoltration membranes, namely the polymeric Duramem�-200 (ref. 17) an Inopor 0.9 nm TiO2 (ref. 18) ceramic membraneand an Inopor 1 nm TiO2 surface modied19 in house with C8

alkane groups.20 The molecular weight cut-off21 of thesemembranes as reported by the manufacturers is 200 Da, 450 Daand 1500 Da respectively.

21502 | RSC Adv., 2013, 3, 21501–21510

Reactions in semi continuous mode

The rst experiments performed were carried out using thereaction–lter–ll mode and using as solvent, the oen used formetathesis, dichloromethane. The experiment was carried outat 25 �C and atmospheric pressure, using as membrane eitherthe 0.9 nm TiO2 or the Duramem-200. Despite the fact that thering closing metathesis of dedam can be carried out atconcentrations of 100 mmol and higher, these experimentswere performed at low concentration, 10 mmol, in order toavoid effects of concentration polarisation on the membrane.12

Which is the formation of a high concentration boundary layerat the surface of themembrane that can result in changes in uxand solute rejection. The quantity of catalyst used in theseexperiments was 1 mol%, reaction time was typically 4 hours. Ascan be seen in Fig. 2 the rst run of the reaction worked effi-ciently in both cases. Thereaer a signicant drop in the reac-tion conversion is noted. Indeed, a maximum of four reactionscould be performed under these conditions subsequently noconversion was observed. Furthermore, it should be noted thatwith either the ceramic or polymeric membranes, rejection ofall precatalysts was $99.95%. However, the observed rejectionof Ru species, as measured by ICP-AES, during the ltrationprocess was far lower (Table 1).

Notable is themuch lower rejection of both the Ru species andreagents and product found within the reaction mixture over theDuramem membrane than the ceramic membrane. At rst sitethis may seem somewhat surprising considering the suppliersmolecular weight cut-off information that would indicate thepolymeric membrane to be the tighter of the two. However, soluterejection in OSN is also related to the solubility parameters ofthe membrane, solvent and solute. This higher rejection over themore open ceramic membrane can be explained in terms of thelow affinity of the solutes within the reactionmixture to the highlypolar ceramic membrane surface.13

Reactions in ow reactor

Performing the reaction as above with a separate reaction andltration step inevitably implies long term stability of the catalyst.A more efficient approach would be to allow in-line separation ofcatalyst from reaction products. Using a membrane reactor asshown schematically in Fig. 1b the reaction was performed in theltration system feed tank and continually during the processpassed over the membrane. There is also a simultaneous addi-tion of a solution of dedam from the dialtration tank to thereaction vessel via constant volume dialtration,22 a process inwhich an equivalent volume of solvent (in this case dedamsolution) is added to the system as permeates through themembrane. The permeate from the membrane can either becompletely returned to the OSN system via the dialtration tankor, on completion of the reaction, kept separate from the reten-tate by diverting it away from the dialtration tank.

There are however, problems associated with this approach.Because the reaction is occurring within the pressurized part ofthe system, typically 10 bar when using ceramic membranesand 20 bar with the polymeric membranes, the only manner inwhich ethene can escape is via permeation through the

This journal is ª The Royal Society of Chemistry 2013

Scheme 1 Model reaction and catalysts used.

Fig. 1 Schematic of themembrane reactor set-up (a) reaction–filter–fill were post reaction nanofiltration is performed on the reactionmixture and (b) flow reactor set-up were reaction and separation can occur simultaneously. P1 is a circulation pump and P2 is a diafiltration pump. Everything within the dotted line box is in thepressure loop and thus under pressure during use.

This journal is ª The Royal Society of Chemistry 2013 RSC Adv., 2013, 3, 21501–21510 | 21503

Paper RSC Advances

Fig. 2 Dedam conversion in reaction–filter–fill mode using as membranes either0.9 nm TiO2 or Duramem-200.

Fig. 3 Dedam conversion after 16 hours reaction time in a flow reactor indichloromethane at room temperature using as membrane 0.9 nm TiO2. HG-I isthe Hoveyda–Grubbs 1st generation catalyst. Catalyst load was 1 mol%.

RSC Advances Paper

membrane. Consequently, this will lead to higher concentra-tions of ethene within the reaction mixture23 than found in astandard batch reaction. These are also oen carried out atreux in order to remove the ethene from the reaction. Asethene is a known inhibitor24 of the Ru based metathesis cata-lysts; it is likely this will affect the catalyst activity.

Because the ceramic membranes had given higher rejectionthan the polymeric membranes in the previous experiments,these were the membranes initially used in the ow reactor. Asolution of dedam at the same concentration as previously usedin the reaction–lter–ll mode was prepared some of which wasbrought into the reaction vessel along with the catalyst theremaining solution was placed in the dialtration tank. Catalyst,1 mol%, was added and the system brought immediately under10 bar pressure and circulated over the membrane. The perme-ated solution was returned to the dialtration tank to allowunreacted starting material to be reintroduced to the reactionvessel thus effectively increasing residence time. The intentionbeing that on completion of the reaction the permeate can bediverted away from the dialtration tank and product washed outof the reactor by the addition of fresh solvent.

However, despite extended reaction times of 16 hours andindependent of the catalyst used (Fig. 3) conversion of dedamwas never as high as the rst run in the semi-continuous mode.Note also the particularly low conversion of dedam with thelatent catalyst Umicore M41. A tentative explanation for this is,on the surface of these ceramic membranes is absorbed a thinlayer of water25 that can be difficult to remove. This surfacewater was reacting with the trichloro(phenyl)silane used withthis catalyst as a chemical activator.26 This produced, in situ,HCl which further reacted with the M41 catalyst resulting in itsbreakdown.

Table 1 Summary of the OSN characteristics from the reaction–filter–fill reaction m

Entry Membrane Catalyst Solvent

Rejection (%)

PrecatalystPosRu

1 0.9 nm TiO2 HG-1 CH2Cl2 $99.95 802 DM-200 HG-1 CH2Cl2 $99.95 35

21504 | RSC Adv., 2013, 3, 21501–21510

In order to determine if this reduced conversion was dueto the higher concentration of ethene within the reactionmixture, the membrane was removed and the reaction repeatedunder the same conditions as used with the membrane. Thus,in this experiment there was no way in which the ethene couldbe removed from the reaction mixture. As expected theconversion of dedam in this reaction was far lower than previ-ously being only 9%. Rejection of all the second generationcatalysts was high. It should be noted that as ICP-AES will showthe presence of all Ru species and cannot distinguish betweencatalyst or catalyst degradation products; as Rabiller-Baudryet al.10e had successfully used UV-vis spectroscopy to determinethe rejection of Ru metathesis catalyst; this technique was alsoapplied here. As can be seen in Table 2 both techniques givevery similar results. Relative standard deviation (RSD) for thesetechniques is 2% and 3% respectively.

By way of comparison, the reaction was also performed usingas membrane the Duramem-200 (Table 2, entry 5). Mostnoticeable in this experiment was the signicant and rapid dropin rejection of the catalyst from an initial 90% to 60% (Fig. 6).A rejection prole that is not conducive with the goal ofproducing reaction products with low metal contamination.

Effects of solvent on catalysts and OSN performance

The results so far would tend to indicate that using ceramicmembranes rejection values for the catalyst of greater than 90%can be achieved. The ideal situation would be as high aspossible rejection of the catalyst and as low as possible for thereaction product. Though the high rejection of the catalyst ispositive, the rejection of both dedam and the cyclopentene

ode

Permeability (Lm�2 h�1 bar�1)t reactionspecies Dedam Product

53 48 10–21 10 0.4

This journal is ª The Royal Society of Chemistry 2013

Table 2 Summary of the OSN characteristics of the membrane flow reactor. M2, M51 and M41 refer to Umicore M2, Umicore M51 and Umicore M41 catalysts

Entry Membr. Catalyst Solvent

Rejection (%)

Permeability(Lm�2 h�1 bar�1)

Catalyst

Dedam ProductUV-vis ICP-AES

1 0.9 nm TiO2 HG-I CH2Cl2 — 80 90 87 0.42 0.9 nm TiO2 M2 CH2Cl2 >99.95 94–83 70–78 73–83 0.863 0.9 nm TiO2 M51 CH2Cl2 >99.95–94 99–97 73–80 70–77 1.84 0.9 nm TiO2 M41 CH2Cl2 >99.95–99 99 88 83 0.25 Dm-200 M2 CH2Cl2 90–60 87–56 38–20 20–30 3.16 0.9 nm TiO2 M2 Acetone 99–93 99–88 40–57 57–64 0.37 Dm-200 M2 Acetone 91 99 98 97 0.88 1 nm C8 M2 Acetone 87 85–78 35–38 30–40 59 0.9 nm TiO2 HG-I Acetone 95–88 97–91 46–45 37–32 0.310 1 nm C8 M51 Toluene >99.95 98–85 40 35 0.6411 1 nm C8 M51 CH2Cl2 83 96 65–60 60–43 212 1 nm C8 M51 Acetone 92–99 90–96 55–59 45–40 2

Paper RSC Advances

reaction product though not prohibitively high, would be pref-erably lower than that observed in dichloromethane. Onemanner of changing the rejection prole of the reaction is tochange the solvent.

Also, dichloromethane is frequently used as solvent inmetathesis reactions as this is a polar solvent that allowsdissociation of a ruthenium ligand hence forming the 14 elec-tron intermediate that is susceptible to trapping by an olenicsubstrate.27 It is non-coordinating thus, allowing coordinationof the reactive olen, which leads to high catalyst initiationrates. However, initiation and decomposition rates are alsolinked.27a In a batch reactor situation this is less of a problem asall starting material is present from the onset of the reaction. Ina ow reactor where starting material is added slowly overtimethis has the potential to be more problematic. As such thequestion arises as to whether a change of solvent could beadvantageous for both catalyst and ltration characteristics.

With this in mind, it was decided to change from dichloro-methane to acetone as solvent. Acetone was chosen because thissolvent was predicted to give slow catalyst initiation,28 whichwith its ability to coordinate with the catalyst metal centre was

Fig. 4 Dedam conversion after 16 hours in acetone using Umicore M2 ascatalyst.

This journal is ª The Royal Society of Chemistry 2013

intended to increase its stability and thus longevity. Further-more, its somewhat lower Hildebrand solubility parameter (19.7as oppose to 20.2 (ref. 29)) will lead to a different rejectionprole. A further advantage of using acetone is the lower envi-ronmental impact of this solvent as compared to the moreclassic metathesis solvents such as dichloromethane ortoluene.30

The experiment was performed under the same conditionsas previously using 1 mol% of Umicore M2 as catalyst and the0.9 nm TiO2 membrane. Reaction conversion was increasedfrom 45% to 83% (Fig. 3 & 4). Not only does the catalyst rejectionremain high but the rejection of both starting material andreaction product are somewhat lower (Fig. 5). The use of othermembranes under similar conditions also leads to dedamconversions greater than 80%. Whereas, the rejection of the M2catalyst in dichloromethane decreased rapidly over the Dura-mem-200 membrane, in acetone the catalyst rejection isremarkably stable (Fig. 6). Unfortunately in acetone the rejec-tion of both dedam and the cyclopentene product is also high.

Rejection of Umicore M2 over the surface modied TiO2

membrane, though lower than 90% is surprisingly high when

Fig. 5 Rejection profile for dedam, product and catalyst Umicore M2 in acetoneover the ceramic 0.9 nm TiO2 membrane.

RSC Adv., 2013, 3, 21501–21510 | 21505

Fig. 6 Reaction rejection profile over polymeric membrane Dm-200 in acetoneand dichloromethane.

RSC Advances Paper

the molecular weight cut off data is taken into consideration(Table 2, entry 8).

Further conrmation of the positive effect of the solvent onthe catalyst is given by examining the conversion of dedam perunit time (Fig. 7) for the Umicore M2 catalyst in the ow reactorusing the 0.9 nm TiO2 membrane. In dichloromethane thisreaches a maximum of approximately 30% aer 120 minutes. Itthen starts to fall off which would appear to be due to thedegradation of the catalyst resulting in less catalyst beingpresent within the system. In acetone this process is retarded tosome extent, conversion reaching a maximum of 60% aer360 minutes at which it is levelling off.

In order to ascertain whether the apparent positive effect ofacetone on catalyst rejection and reaction performanceextended to the less stable rst generation Ru metathesiscatalysts, the experiment was repeated using the Hoveyda–Grubbs 1st generation catalyst. Again an increase in both cata-lyst rejection and reaction conversion was observed (Table 2,entries 1 & 9). Also an increase in conversion was noted from65% to 83% in acetone.

The fact that the solvent–solute–membrane interactions areabsolutely fundamental in determining the ux and soluterejection characteristics can be seen by examining the ceramic0.9 nm TiO2 membrane performance (Table 2, entries 2 & 6).

Fig. 7 Conversion–time plot for the flow reactor using as catalyst Umicore M2 indichloromethane and acetone and membrane the ceramic 0.9 nm TiO2.

21506 | RSC Adv., 2013, 3, 21501–21510

Molecular weight cut-off supplied by the manufactures is 450Da. Therefore, the observed rejection in dichloromethane ofboth dedam and the cyclopentene reaction product which havemolecular mass far below 450 Da (240.3 g mol�1 & 212.2 g mol�1

respectively) based on this information is higher than might beexpected. An explanation for this is due to the highly polarnature of the surface of this membrane. As dedam and reactionproduct are of low polarity there is very little affinity of thesemolecules for the membrane, thus retarding their transportthrough it. Changing the solvent to acetone which also has alower Hildebrand solubility parameter leads to a reduction inthe affinity of the solvent for the membrane which consequentlyresults in lower ux. The concomitant lowering of the rejectionof dedam and reaction product was explained by Darvishma-nesh31 et al. by means of, as the solvent affinity for themembrane lowers a competitive permeation of the solvent andsolute(s) occurs which in extreme cases can result in negativerejections.

These solvent–solute–membrane interactions also explainthe rejection prole seen with the polymeric membrane Dura-mem-200 (Table 2, entries 5 & 7). The surface layer of thismembrane is not as polar as the ceramic membrane. This lowerpolarity manifests itself in dichloromethane as a high solventmembrane interaction, one where the solvent is causing themembrane to swell, increasing the interstitial space betweenthe polymer chains thus leading to high ux and low rejectionof all the solutes in the mixture. By contrast, the membranesolvent interaction with acetone is far lower thus resulting inlower ux and high retention. With amolecular weight cut off of200 Da this tight membrane would be expected to give highrejection in solvents that do not cause swelling.

If however, swelling effects are removed by using a ceramicmembrane that has a lower surface polarity than the TiO2,ceramic membranes are resistant to swelling. Then the inter-actions between the membrane surface and solvent and solutebecome somewhat clearer. This can be done using the ceramicmembranes that have the surface modied with alkyl groups, inthis case a C8 alkane.32 This modication reduces the polarity ofthe surface of the membrane to a level very similar to theDuramem series. This modication was carried out on a 1 nmTiO2 membrane resulting in a membrane with molecularweight cut off of 1500 Da when using either polyethylene glycolsdissolved in water or styrene oligomers dissolved in acetone astest mixture. With such an open membrane it would bereasonable to expect very low retention of all solutes within thesolution. However, these membranes displayed far higherrejections not only of the M2 but also the M51 in acetone,dichloromethane and toluene. Indeed, toluene appears to givethe highest rejection of the catalyst, unfortunately this is alsocoupled to low ux due to the low affinity of the solvent for themembrane. Also whereas, the rejection of the catalyst remainsstable and high in this solvent the rejection of dedam and thecyclopentene are very variable. As seemingly further conrma-tion of the positive effect of acetone, conversions in dichloro-methane or toluene were lower than those observed in acetone.

Attempts at reducing the catalyst load in the reaction(Table 3) invariably gave higher turnover numbers (TON) but at

This journal is ª The Royal Society of Chemistry 2013

Table 3 Comparison of TON and product purity on reduction of the catalyst load

Entry Membrane Solvent Catalyst Loading (mol%) Conversion (%) Product purity (%) TON

1 1 nm C8 Acetone M2 0.25 22 0.4 692 1 nm C8 Acetone M2 0.5 76 60 1293 1 nm C8 Acetone M2 1 84 81 79

Paper RSC Advances

the expense of both reaction conversion and product purity.Product purity was calculated at the end of the reaction usingthe equation below. In this case it is clear that TON is aninappropriate indicator of the catalyst efficiency.

Purity ¼ moles product

moles productþmoles dedam

The observed increase in TON is however, minimal.

Product isolation

The reaction in the ow reactor set up is performed with totalreturn of the permeate until reaction has gone to completion orthe catalyst is no longer active. At the end of this time in order towash out the reaction product from the reactor a dialtrationwithfresh solvent needs to be executed, this time without returningthe permeate to the OSN loop. This is carried out as a constantvolume dialtration using fresh solvent. However, as the rejectionof the cyclopentene is never zero this will require a number ofdialtration volumes to achieve; the lower the cyclopentenerejection the fewer this will be. Furthermore, though rejections of>90% for the catalyst can be achieved, the more dialtrationvolumes required to wash out the product themore of the catalystwill be lost into the permeate. The percentage of the original loadof solute in the retentate can be calculated using the equationbelow, where %Ri is the percentage of solute i in the retentate,N is the number of dialtration volumes and ri is the averagerejection of solute i during the process. Average rejections used inthis calculation are listed in Table 4.

%Ri ¼ 100 � e�N(1�ri)

The results of this simulation are shown graphically belowfor each membrane with both dichloromethane and acetone assolvents. The aim was to simulate washing out 95% of thecyclopentene into the permeate from which the number of

Table 4 Table of average rejections used in simulation calculations

Process average rejection

0.9 nm TiO2 Dm-2001 nm C8 modiedTiO2

CH2Cl2 Acetone CH2Cl2 Acetone CH2Cl2 Acetone

Catalyst 90 95 69.5 91 83.4 93.7Product 78 55 24.7 97.4 51.3 41

This journal is ª The Royal Society of Chemistry 2013

dialtration volumes required to do this and the quantity of thecatalyst remaining in the retentate can be ascertained. For somesolvent–membrane combinations such as acetone and theDuramem-200, the target of 95% cyclopentene in the permeate

Fig. 8 Simulation of retentate composition in dichloromethane and acetoneover the three membranes used in this study (a) 0.9 nm TiO2, (b) Duramem-200and (c) 1 nm C8 modified TiO2.

RSC Adv., 2013, 3, 21501–21510 | 21507

RSC Advances Paper

was not achieved within the 10 dialtration volumes set as theupper limit. For the 0.9 nm TiO2 and the Duramem-200membrane the catalyst used with this simulation was the M2whereas with the 1 nm C8 modied membrane the catalyst usedis the M51.

Despite the fact that none of these catalysts are rejected by themembranes to the extent of the enlarged catalysts reportedpreviously10 they are still of interest. Especially for the ceramicmembranes with acetone as solvent which aer washing out thecyclopentene there remains with both membranes 70% of thecatalyst within the retentate. Whereas this is not enough topreclude further purication to completely remove all rutheniumspecies. It does have the advantage that these catalysts are alreadycommercial and if the enlarged catalyst do at some time alsobecome commercial their synthesis will probably be longer thanthe catalysts used here and therefore result in a higherpurchasing cost. Therefore the compromise between lower cata-lyst cost and the possibility that extra purication steps beingrequired to completely remove all residual Ru which, in practicecould occur in downstream purication steps within a synthesis,is at least worth consideration (Fig. 8).

Conclusion

In this study we have demonstrated the feasibility of usingmembranes in ow reactors to separate organometallic speciesfrom reaction mixtures and highlighted some of the specicproblems related to metathesis in a pressurized reaction vessel.The benecial effects of changing from solvents that gives highinitiation rates such as dichloromethane to acetone that has alower environmental impact have also been demonstrated.Whereas, the rejection of the commercial catalysts used in thiswork does not achieve the extremely high rejection of molecu-larly enlarged catalysts previously reported. The simulationindicates that not only should the focus be placed on highcatalyst rejection but the whole reaction system must beconsidered. This suggests that rejections that might otherwisebe considered non ideal could be economically viable due tothe ready availability and lower cost of the catalyst used.

ExperimentalMaterials

The solvents used in this study were dichloromethane, acetoneand toluene, all were technical grade purchased from VWR andused without prior purication. Commercially availablemembranes selected for this study were Duramem�membranes, purchased from Evonik MET Ltd. (London, UK)and Inopor� ceramic membranes from Inopor (Veilsdorf,Germany). The C8 modied ceramic membranes were Inopormembranes modied in house to have C8 alkyl chains on themembranes top layer. Hoveyda–Grubbs 1st generation catalystand diethyldiallyl malonate were purchased from Sigma-Aldrich(Belgium), Umicore M series catalysts, M2, M41 and M51 weredonated by Umicore AG & Co. KG (Hanau, Germany). Allmembrane experiments were performed in an in house madecross-ow nanoltration unit on lab scale.

21508 | RSC Adv., 2013, 3, 21501–21510

Membrane pre-treatment

Duramemmembranes are supplied in a “dry” form in at sheets.A disk of membrane with effective membrane area of 44 cm2 wascut out from the sheet and placed in a cross owltration cell andto the system was added 500 ml of the solvent used in theexperiment. The membrane was saturated in solvent by bringingthe system under 10 bar pressure with N2 until the rst drops ofpermeate are observed. Thereaer the pressure was released andthe membrane allowed to equilibrate overnight. The membranewas then conditioned by bringing the system under 20 bar pres-sure and allowing approximately 50 ml of solvent to permeatethrough the membrane to remove the preservatives. Pure solventwas then permeated through the membrane until a stable uxwas obtained. Ceramic membranes can be used directly withoutthe need for the pretreatment procedure.

Analysis

Ring closing metathesis of diethyldiallyl malonate (dedam) todiethyl cyclopent-3-ene-1,1-dicarboxylate (cyclopentene) wasanalysed using a 20 minute GC method on an InterscienceTrace GC, gas chromatograph. Samples were analysed using aame ionisation detector with an Agilent DB5-MS column ofdimensions 60 m � 0.25 mm � 0.25 mm. The injector temper-ature was constant at 280 �C and an injection volume of 1.0 mlusing a split injector mode of 50 : 1. The column temperatureprogram started at 100 �C and was held for 1 minute beforebeing increased to 310 �C at a rate of 20 �Cmin�1 and held for 5minutes. The detector temperature was set at 320 �C, detectionbeing enabled by amake up ow of 30 ml min�1 with an air owof 350 ml min�1 and hydrogen ow of 35 ml min�1. Analysis ofruthenium species was carried out by UV-vis spectrophotometryand ICP-AES. UV-vis spectrophotometry was performed on aHach DR3900 spectrophotometer with the aid of calibrationcurves the concentration in solution can be calculated from theabsorption. For each catalyst the maximum absorption is:Hoveyda–Grubbs 1st generation 362 nm, Umicore M2 396 nm,Umicore M51 373 nm and Umicore M41 344 nm. Inductivelycoupled plasma atomic emission spectroscopy (ICP-AES) anal-ysis was performed as follows. Aer evaporation of the solvent,the residue was digested with aqua regia and diluted withdeionized water to the required concentration range. Thesamples were then analysed for metal content.

Reactions in semi-continuous reaction–lter–ll mode

To the ltration apparatus tted with a membrane (if required,preconditioned) was added a solution of dedam in dichloro-methane at a concentration of 10 mM. The circulation pumpwas switched on and the mixture circulated for approximately 2minutes and a sample taken for GC analysis. Hoveyda–Grubbs1st generation catalyst (1 mol% compared to the quantity ofdedam used) was added and the solution circulated within theapparatus at atmospheric pressure and 25 �C for 4 hours.Samples were taken on a regular basis for GC analysis. To all GCsamples was added DMSO. Aer 4 hours reaction time thesystem was brought under pressure (10 bar with ceramic

This journal is ª The Royal Society of Chemistry 2013

Paper RSC Advances

membranes, 20 bar with polymeric membranes) ux wascontinually monitored. Once 500 ml solution had permeatedacross the membrane samples of retentate and permeate weretaken for GC, UV-vis and ICP-AES analysis. 500 ml of a 10 mMsolution of dedam in dichloromethane was added and thesolution circulated at 25 �C and atmospheric pressure for 4hours with regular sampling for GC analysis. No fresh catalystwas added to the solution. Filtration procedure was repeated asbefore. The above procedure was repeated until no furthercatalyst activity was observed.

Reaction under continuous ow mode

To the ltration apparatus tted with a membrane (if required,preconditioned) was added 700ml of a solution of Dedam in thereaction solvent at a concentration of 10 mM. The circulationpump was switched on and the mixture circulated at atmo-spheric pressure until the internal temperature was 25 �C. Asample was taken for GC analysis. All GC samples were treatedwith DMSO. Connected to the ltration apparatus via a pumpand set up to perform constant volume dialtration was a300 ml solution of dedam in the reaction solvent at concen-tration of 10 mM (dialtration solution). Ruthenium catalyst(1 mol% compared to the quantity of dedam used) was added tothe ltration apparatus and samples were taken for GC analysis,UV-vis and ICP-AES. The system brought under pressure (10 barfor ceramic membranes, 20 bar for polymeric membranes) themembrane ux was continually monitored. Permeate fromthe membrane was added directly to the dialtration solution.The reaction was allowed to proceed in such a manner for16 hours with regular sampling of the ltration loop contents(retentate), the dialtration solution and the membranepermeate outlet (permeate) for GC. Also on a regular basissamples were taken of the retentate and dialtration solutionfor UV-vis and ICP-AES analysis.

Reactions with lower catalyst loading were performed asabove.

Acknowledgements

The authors would like to acknowledge Umicore AG & Co. KG(Hanau, Germany) for the generous donation of the Umicore Mseries catalysts used in this study.

References

1 (a) N. K. Yee, V. Farina, I. N. Houpis, N. Haddad, R. P. Frutos,F. Gallou, X. Wang, X. Wei, R. D. Simpson, X. Feng, V. Fuchs,Y. Xu, J. Tan, L. Zhang, J. Xu, L. L. Smith-keenan, J. Vitous,M. D. Ridges, E. M. Spinelli, M. Johnson, K. Donsbach,T. Nicola, M. Brenner, E. Winter and P. Kreye, J. Org.Chem., 2006, 71, 7133; (b) H. Wang, S. Goodman, Q. Dai,G. Stockdale and W. Clark, Org. Process Res. Dev., 2008, 12,226; (c) V. Farina, C. Shu, X. Zeng, X. Wei, Z. Han,N. K. Yee and C. H. Senanayake, Org. Process Res. Dev.,2009, 13, 250; (d) J. Kong, C. Chen, J. Balsells-Padros,Y. Cao, R. F. Dunn, S. Dolman, J. Janey, H. Li andM. Zucuto, J. Org. Chem., 2012, 77, 3820; (e)

This journal is ª The Royal Society of Chemistry 2013

J. Arumugasamy, K. Arunachalam, D. Bauer, A. Becker,C. A. Caillet, R. Glynn, G. M. Latham, J. Lim, J. Liu,B. A. Mayes, A. Moussa, E. Rosinovsky, A. E. Salanson,A. F. Soret, A. Stewart, J. Wang and X. Wu, Org. Process Res.Dev., 2013, 17, 811.

2 The European Agency for the Evaluation of MedicinalProducts, http://www.emea.europa.eu/pdfs/human/swp/444600en.pdf, 2007.

3 G. C. Vougioukalakis, Chem.–Eur. J., 2012, 18, 8868.4 M. J. Byrnes, A. M. Hilton, C. P. Woodward, W. R. Jacksonand A. J. Robinson, Green Chem., 2012, 14, 81.

5 K. Skowerski, C. Wierzbicka, G. Szczepaniak, Ł. Gułajski,M. Bieniek and K. Grela, Green Chem., 2012, 14, 3264.

6 M. Janssen, J. Wilting, C. Muller and D. Vogt, Angew. Chem.,Int. Ed., 2010, 49, 7738.

7 (a) C. Wiles and P. Watts, Green Chem., 2012, 14, 38; (b)S. G. Newman and K. F. Jensen, Green Chem., 2013, 1456.

8 A. T. Boam and A. Nozari, PharmaChem, 2005, 4, 57.9 P. Vandezande, L. E. M. Gevers and I. F. J. Vankelecom,Chem. Soc. Rev., 2008, 37, 365.

10 (a) A. Keraani, T. Renouard, C. Fischmeister, C. Bruneau andM. Rabiller-Baudry, ChemSusChem, 2008, 1, 927; (b)D. Schoeps, K. Buhr, M. Dijkstra, K. Ebert and H. Plenio,Chem.–Eur. J., 2009, 15, 2960; (c) P. van der Gryp,A. Barnard, J.-P. Cronje, D. de Vlieger, S. Marx andH. C. M. Vosloo, J. Membr. Sci., 2010, 353, 70; (d)A. Kajetanowicz, J. Czaban, G. R. Krishnan, M. Malinska,K. Wozniak, H. Siddique, L. G. Peeva, A. G. Livingston andK. Grela, ChemSusChem, 2013, 6, 182; (e) M. Rabiller-Baudry, G. Nasser, T. Renouard, D. Delaunay andM. Camus, Sep. Purif. Technol., 2013, 116, 46.

11 M. Janssen, C. Muller and D. Vogt, Dalton Trans., 2010, 39,8403.

12 M. Mulder, Basic Principles of Membrane Technology andApplications, ed. M. Mulder, Kluwer Academic Publications,Dordrecht, 2nd edn, 2004.

13 (a) S. Darvishmanesh, J. Degreve and B. Van der Bruggen,Phys. Chem. Chem. Phys., 2010, 12, 13333; (b)H. J. Zwijnenberg, S. M. Dutczak, M. E. Boerrigter,M. A. Hempenius, M. W. J. Luiten-Olieman, N. E. Benes,M. Wessling and D. Stamatialis, J. Membr. Sci., 2012, 390–391, 211; (c) P. Marchetti, A. Butte and A. G. Livingston, J.Membr. Sci., 2013, 444, 101.

14 L. Protasova, M. Bulut, D. Ormerod, A. Buekenhoudt,J. Berton and C. V. Stevens, Org. Process Res. Dev., 2013, 17,760.

15 T. Ritter, A. Hejl, A. Wenzel, T. Funk and R. H. Grubbs,Organometallics, 2006, 25, 5740.

16 D. Nair, H.-T. Wong, S. Han, I. F. J. Vankelecom, L. S. White,A. G. Livingston and A. T. Boam, Org. Process Res. Dev., 2009,13, 863.

17 Purchased from Evonik MET Ltd. (London UK), Molecularweight cut-off based the rejection of styrene oligomersdissolved in acetone.

18 Purchased from Inopor (Veilsdorf, Germany), Molecularweight cut-off based on the rejection of polyethylene glycoldissolved in water.

RSC Adv., 2013, 3, 21501–21510 | 21509

RSC Advances Paper

19 Molecular weight cut-off based on rejection of polyethyleneglycol dissolved in water or styrene oligomers in acetone.

20 A. Buekenhoudt, K. Wyns, V. Meynen, B. Maes and P. Cool,Surface-modied inorganic matrix and method forpreparation thereof, WO2010106167 A1, 2012.

21 Y. H. See Toh, X. X. Loh, K. Li, A. Bismarck andA. G. Livingston, J. Membr. Sci., 2007, 291, 120.

22 (a) M. Mulder, Basic Principles of Membrane Technology andApplications, ed. M. Mulder, Kluwer Academic Publications,Dordrecht, 2nd edn, 2004; (b) I. Sereewatthanawut, F. Lim,Y. Bhole, D. Ormerod, A. Horvath, A. T. Boam andA. G. Livingston, Org. Process Res. Dev., 2010, 14, 600.

23 (a) A. Sahgal, H. M. La and W. Hayduk, Can. J. Chem. Eng.,1978, 56, 354; (b) L. Lee, H. Ou and H. Hsu, Fluid PhaseEquilib., 2005, 231, 221.

24 (a) J. Scholz, S. Loekman, N. Szesni, W. Hieringer, A. Gorling,M. Haumann and P. Wasserscheid, Adv. Synth. Catal., 2011,353, 2701; (b) M. Ulman and R. H. Grubbs, J. Org. Chem.,1999, 64, 7202.

25 A. Buekenhoudt, F. Bisignano, G. De Luca, P. Vandezande,M. Wouters and K. Verhulst, J. Membr. Sci., 2013, 439, 36.

21510 | RSC Adv., 2013, 3, 21501–21510

26 S. Monsaert, N. Ledoux, R. Drozdzak and F. Verpoort, J.Polym. Sci., Part A: Polym. Chem., 2010, 48, 302.

27 (a) M. S. Sanford, J. A. Love and R. H. Grubbs, J. Am. Chem.Soc., 2001, 123, 6543; (b) V. Thiel, M. Hendann,K.-J. Wannowius and H. Plenio, J. Am. Chem. Soc., 2012,134, 1104.

28 I. W. Ashworth, D. J. Nelson and J. M. Percy, Dalton Trans.,2013, 42, 4110.

29 A. F. M. Barton, Handbook of Solubility Parameters, CRCPress, Bosa Roca, 2nd edn, 1991.

30 R. K. Henderson, C. Jimenez-Gonzalez, D. J. C. Constable,S. R. Alston, G. G. A. Inglis, G. Fisher, J. Sherwood,S. P. Binks and A. D. Curzons, Green Chem., 2011, 13, 854.

31 S. Darvishmanesh, J. Degreve and B. Van der Bruggen, Phys.Chem. Chem. Phys., 2010, 12, 13333.

32 (a) P. Van Heetvelde, E. Beyers, K. Wyns, P. Adriaensens,B. Maes, S. Mullens, A. Buekenhoudt and V. Meynen,Chem. Commun., 2013, 49, 6998; (b) A. Buekenhoudt,K. Wyns, V. Meynen, B. Maes and P. Cool, Surface-modied inorganic matrix and method for preparationthereof, WO2010106167 A1, 2012.

This journal is ª The Royal Society of Chemistry 2013