Immobilization of Chiral Ferrocenyl Ligands on Silica Gel and their Testing in Pd-catalyzed Allylic Substitution and Rh-catalyzed Hydrogenation

Five different silica gels containing two chiral ferrocenyl ligands were prepared by various synthetic routes and tested in an enantioselective Pd(0)-catalyzed allylic substitution and Rh-catalyzed hydrogenation. All the prepared anchored ligands were characterized by porosimetry data, DRIFTS spectra, thermal data and AAS. The aim of the work was to compare the influence of the carrier, surface properties and immobilization strategy on the performance of the catalyst.


Introduction
Heterogenization of the catalytic systems is still a very frequent goal in the development of highly efficient synthesis. Although none of the immobilized catalysts has been so far exploited in large scale production, mostly because of the lower performance when comparing with the homogeneous analogues or high costs, the possibility of the regenerating and recycling of the expensive catalysts is very attractive for the both industrial and academical research [1].
The development of new materials suitable for the anchoring of the homogeneous catalysts and novel synthetic strategies is one of the most efficient ways for improving the properties of the immobilized catalysts. Silica-based materials are still number one in the world of carriers. This is especially true since the discovery of the MCM family by the researchers of Mobil Corporation [2,3]. Excellent sol-gel-template method for the preparation of uniform materials with excellent properties was subsequently modified and developed in many directions. Two of them should be mentioned especially for their exploitation in this work: Tanev´s method using neutral template [ 4 ] and Macquarrie´s and Mann´s copolymerization using two types of silanes [5 , 6].
Although a few papers from Pugin [7], Toma [8 , 9] and Johnson [10] deal with the anchoring of ferrocenyl ligands, the immobilization of such a large molecule (as well as the others) is often problematic because of the limited mass transport and high steric hindrance. Thus the immobilization by the sol-gel-template method would be a very comfortable way for the heterogenization of the ligands. We have focused on the exploitation of the various silica gels as the carriers and comparing the efficiency and the simplicity of the immobilizing strategies.

Immobilizations of the ferrocenyl ligands
Homogeneous ligands (R,pS)-2, (R,pS)-3, precursors for the immobilization, were readily derived from (R,pS)-1 by the following simple and fast reactions (Scheme 1): Two methods were chosen for the immobilization: grafting and sol-gel-template copolymerization. The grafting was based on reaction of the functional groups on the silica gel surface with the appropriate ferrocenyl derivative. Several strategies were proposed for the heterogenization based on an alkylation-like synthesis. (Scheme 2).
An attempt based on the alkylation of the aminopropylated Keiselgel K100 (pore size diameter 100 Å) was unsuccesful. The amide preferred to hydrolyze under basic conditions and to decompose under non-basic conditions, higher temperatures and longer reaction times. The absence of the desired reaction is also probably caused by the problems with the mass transport and decreased reactivity of the functional groups on the silica gel.

HMS HMS
For that reason we have decided to separate the functional group further from the surface by preparing the functionalized ligands 5 (by the modification of the amorphous silica gel with the N- [3-(triethoxysilyl)propyl]amide of 5-bromovaleric acid), 6 (by the reaction of 5-bromovaleroyl chloride with aminopropylated K100) and 7 (by sol-gel-template method of the N- [3-(triethoxysilyl)propyl]amide of 5-bromovaleric acid and tetraethoxysilane). The anchoring method based on the alkylation of the amino group of the ligand (R,pS)-1 with the bromoalkyl group in 5, 6, or 7 was not successful even after prolonged reaction times, increasing the temperature or performing the reaction with the assistance of microwaves (Scheme 2).
With these results in mind we have decided to use a more reactive triethoxysilyl moiety for the immobilization, which is very suitable for the recombination with the hydroxyl groups on the surface of the silica gel (Scheme 3) and thus we have immobilized compound (R,pS)-3 on the amorphous silica gel to form modified silica gel (R,pS)-11 (route A). Porosimetry showed surface area of 467 m 2 /g, pore volume of 0.76 cm 3 /g and average pore diameter 63 Å with C = 69,0 (indication of a relatively polar surface). Thermal analysis (20 Kmin -1 ) gave the following weight losses: up to 150 °C The necessity of the post-protection of the residual silanol groups is still an unanswered question. On one hand they are potential partners for the coordination with the metal, on the other hand their protection (e.g. with trimethylsilyl groups) leads to a change in the hydrophobic properties. In order to examine this problem we took half of the modified silica gel (R,pS)-11 and performed the reaction with HMDS to prepare silica gel with blocked silanol groups (R,pS)-12 (route B). Analytical data confirmed the successful protection of free silanol groups: decrease of the surface area: (347 m 2 /g), decrease of the pore volume (0.46 cm 3 /g), decrease of the average pore diameter (53 Å) and decrease of the polarity of the surface (C = 16,2). DRIFTS spectrum was similar as in the previous case, but signal for free silanol groups was missing and signals for -CH 3 appeared (2846, 2902, 2960). Thermal analysis showed weight losses at 150 °C (0.4 % -water and solvents), 180 -350 °C (1.6 %), 350 -750 °C (4.4 %, organic material). Loading of the ligand was specified as 50 µmol/g.
Having the compound (R,pS)-3 in our hands we have immobilized it by an alternative way based on the sol-gel-neutral template method [ 11 ] (route C). A mixture of tetraethoxysilane and the corresponding triethoxysilyl derivative (R,pS)-3 in the ratio of 9:1 was stirred in an1:1 ethanol/water mixture in the presence of dodecylamine as structure determining agent to directly give modified silica gel (R,pS)-13 with partial mesoporous structure. Surface area is 407 m 2 /g. HMS prepared by this method consists of two different types of pores: framework and textural. Pore volume of the pores formed by the template mechanism (framework pores) is 0.17 cm 3 /g. and average pore diameter is 14 Å. In addition another significant amount of textural pores with amorphous character was formed. They might be cracks in the structure, spaces between particles or, more likely, an amorphous silica gel formed simultaneously [ 12 ]. Given the size of the organic silane, it is also likely that the ferrocenyl units are present predominantly in this pore system. This pore system has a pore volume of 0.175 cm 3 /g and an average pore diameter of 65 Ǻ. Loss of weight was observed at 180 °C (4.4 %water and solvents), 180 -290 °C (4.7 %), 290 -750 °C (17.3 % -organic material AAS determination (Fe) showed unusual loading of 478 µmol/g, which is significantly higher than with the previous route, but approximately half that expected from the starting gel composition. DRIFT spectrum (ν, cm -1 ) indicates the presence of isolated silanol groups (3731, 3644, 3411), aryl C-H (3056, 3066, 3095), aliphatic C-H (2977, 2933, 2908, 2850), urea moiety (1644, 1635, 1519), aryl C-H def (1536, 1560), aliphatic C-H def (1477, 1434) and SiO def (968). 13 C CPMAS NMR gave the following resonances: 133-145 aryl-P; 78 Cp; 57 N-CH 3 ; 51 N-CH 2 ; 32 and 36 CH 2 CH 2 CH2 and CH-Cp; 24 CH 3 -CH; 17 CH 2 -Si. Additional peaks at 21 and 64 are likely to derive from OEt groups attached to the surface.
The presence of two types of pores led us to the decision to modify pure HMS with 3 in order to prepare silica gel with only well-ordered mesoporous architecture (R,pS)-14 (route D). This led to a material with a well defined pore structure, typical of a HMS material. The total pore volume was 1.022 cm 3 g -1 and the surface area 1092 m 2 g -1 . Average pore diameter was 25.1Å. Very little textural porosity was present (ca. 5% of the total). This indicates that the internal pore surface is not well covered with silanes, as this would lead to a drastic reduction in pore volume. This is not surprising, since such large silanes would be expected to bind preferentially to the external surface of such materials. Thermal analysis (<150 o C -2.6% loss of water + solvent; 215-450 o C, gradual loss followed by more rapid loss between 450 o C and 650 o C (6.2% in total; 650-800 o C, 1.1%) indicates that the loading of organics on the surface is 120 µmol g -1 . This is significantly higher than most of the samples, with the exception of (R,pS)-13. CPMAS 13 C-NMR spectra were weak, but appeared broadly consistent with the expected structure. DRIFT spectrum run at 150 o C indicated the following bands ((ν, cm     (EtO) 4 Si, dodecylamine One goal was to have another ferrocenyl ligand immobilized on different surfaces. The second one was to compare the catalytic performance of the bidentate ligand with the catalytic properties of the tridentate ligand (based on BPPFA -already published by Toma [8]). The question was whether there is in the case of tridentate ligand some kind of competition between the coordinating atoms that negatively affects the catalytic properties.
The precursor for the immobilisation (R,pS)-17 was used as a crude product without further purification. It was anchored to the amorphous silica gel to form (R,pS)-18 with TMS-protected residual silanol groups. Porosimetry indicated that the structure of the materials changed little upon functionalisation (surface area 373m 2 g -1 , pore volume 0.53 cm 3 g -1 ; average pore diameter 55Å) other than a slight reduction in surface area and pore volume due to pore filling. Atomic absorption spectrometry indicated a loading of 44 µmol g -1 , Thermal analysis with weight losses at <150 o C of 0.5% (indicating good hydrophobisation) and 200-800 o C of 5.7% (organics + TMS). Consistent with the successful hydrophobisation of the surface indicated by thermal analysis, the infra-red spectrum did not contain isolated silanols (compare with (R,pS)-14 where this is a strong peak). A weak and broad silanol stretch was evident at 3687cm -1 and 3502 cm -1 with C-H str at 3102 and 3065 cm -1 (aryl) and 2961, 2902 and 2843 cm -1 (aliphatic). Amide bands were evident at 1648, 1636 and 1518 cm -1 with C-H def at 1436 and 1410 cm -1 . Again MAS NMR was too weak to be useful.
Finally, triethoxysilyl derivative (R,pS)-17 was involved in the SGT method with TEOS to produce (R,pS)-19. AAS showed loading 50 µmol/g. The other analyses were not unfortunately performed.

Pd(0)-catalyzed allylic substitution
Six prepared heterogeneous ligands were examined in an enantioselective Pd(0)-catalyzed allylic substitution (Scheme 5). Each ligand was tested at two reaction times: 24 and 95 hours. Results are given in Table 1. The reaction was performed by the methods described in the literature [13] with 1 mol. % of Pd at 40°C with sodium hydride as the base. The results were compared with the homogeneous ligand (R,pS)-20 prepared by the reaction of 3 with phenylisocyanate. The homogeneous analogue gave the product in 49 % yield and 8 % enantiomeric purity. Heterogeneous ligands were better concerning the enantioselectivity but the yields of the product were lower.
An improvement was observed after protection of the silanol groups (yield by 8 % and ee by 2 %). It confirms that the free hydroxy groups have a tendency to coordinate the metal. This might be also the explanation for the opposite configuration of the major product obtained with silica gel with trimethylsilylated surface Silica gel formed by the SGT method (R,pS)-13 gave lower values of the yield and ee when compared with the ligand immobilized on amorphous silica gel (R,pS)-12.
All the results showed that the prolonging the reaction time leads to increased yields but decreased enantioselectivity. This might be caused by the damaging of the reaction centres. According to this the modified silica gels were not regenerated and recycled in another reaction.

Rh-catalyzed hydrogenation
As we were not very satisfied with the results of allylic substitution we have examined the immobilized ligand in another enantioselective reaction: Rh-catalyzed hydrogenation of (Z)-αacetamidocinnamic acid by the procedure described by Hayashi [14] (Scheme 5). The reaction was performed in degassed methanol at a pressure of 50 atm with 1 mol% of the catalyst. Results are given in Table 2. The homogeneous ligand (R,pS)-20 gave the product in 25 % yield. The best result with the immobilized catalytic systems was achieved with the one anchored to the amorphous silica gel with protected silanol groups (R,pS)-12. Unfortunately, the catalyst showed significantly lower activity when recycled. Prolongation of the reaction time to 50 h led to the complete conversion. Enantioselectivity was determined in this case by polarimetric measurement of the product, which was purified by the method described by Kagan [15], but ee's lower than 5% were observed. Presence of the free silanol groups proved to be deleterious for the activity of the catalyst also in this case. The results obtained with the modified HMS turned out to be disappointing as (R,pS)-13 gave the product only in 48 % and (R,pS)-14 seemed to be totally inactive. Since both materials should be of the same architecture, unless prepared by different methods, one can assume that ferrocenylphosphane ligands attached to the mesoporous silica are inactive and that the reaction in the case of (R,pS)-13 is catalyzed only by the catalyst attached to the amorphous silica with textural pores, that was formed along with the mesoporous material. Ligands (R,pS)-18 and (R,pS)-19 showed poor activity, though the performance of the modified amorphous silica was again higher compared to the mesoporous one.

Conclusions
We have successfully immobilized two different ferrocenylphosphane ligands on silica by the functionalization of the carrier with the ligand precursor or by the direct incorporation of the ferrocene unit in the silica-forming process. However, sol-gel-template process suffered from the diverse size of the applied silanes and resulting material consisted of templated silica together with the amorphous one, which was probably the sole active part of the modified carrier. Heterogeneous ligands were tested in an enantioselective Pd(0)-catalyzed allylic substitution and Rh-catalyzed hydrogenation. The obtained results fall in the long queue of the unsuccesful attempts to maintain the activity of the catalytic results after its immobilization on solid surface. Prepared ligands showed a low activity with regard to the yield and enantioselectivity. Increasing of the yield by the prolongation of the reaction time, led to lower enantioselectivities indicating that the active catalytic centres are irreversibly damaged during the reaction. This is also confirmed by the unsuccesful attempt to recycle the catalyst.

Acknowledgements
Our thanks are due to Dr. K. Gáplovská and Dr. Solčániová and their staff for microanalysis and 1 H, 13 C NMR spectra.

Experimental
General 1 H-NMR (δ, ppm) spectra of samples were obtained from CDCl 3 solutions on a Varian Gemini 2000 spectrometer operating at 300 MHz ( 1 H) and 75 MHz ( 13 C) frequency with tetramethylsilane as internal standard. IR and UV-VIS spectra of soluble samples were measure as CHCl 3 solutions on Perkin Elmer 781 spectrometer resp. Hewlett Packard 781 spectrometer. Optical rotations were measured on Perkin-Elmer 241 polarimeter. Diffuse reflectance FTIR (DRIFTS) spectrum of modified silica gel was measured as a mixture of 5% sample and 95% KBr in an environmental chamber at 150°C under vacuum on a Bruker Equinox FTIR. The porosimetry data were measured on Micromeritics ASAP2100 system using nitrogen as adsorbate at 77K. Samples were degassed at 120°C until a constant vacuum of 5 microns Hg was obtained. Solvents were purified and dried according to standard published methods. (R, pS)-PPFNMe was prepared according to described methods [16].