Olefin Hydroformylation by Sol-gel Entrapped Rhodium Catalysts Bearing Hydrolysable Ligands

Complexos de rodio preparados in situ a partir de [Rh(OMe)(COD)]2, [Rh(CO)2(acac)], [Rh(cod)(acac)] ou [Rh(cod)(PPh3)2]+BPh 4- com ligantes como HS(CH2)3Si(OMe)3, Ph2P(CH2)2S(CH 2)3Si(OMe)3 ou Ph2P(CH2)2Si(OMe) 3 foram imobilizados em matrizes de silica, inorgânicas ou hibridas, pelo processo sol-gel. As matrizes inorgânicas foram preparadas apenas com tetrametilortossilicato enquanto que para as hibridas foram utilizados 1,4-bis(trietoxissilil)benzeno ou 1,2-bis(trietoxissilil)etano como agentes de co-condensacao. O sistema baseado em [Rh(CO)2(acac)]/ Ph2P(CH2)2S(CH 2)3Si(OMe)3 foi ativo na hidroformilacao de 1-hexeno e de 1-octadeceno sem lixiviacao de rodio. Tambem pode ser usado na ausencia de solvente, como observado na hidroformilacao do 1-deceno. Apesar do melhor sistema obtido apresentar uma matriz microporosa, nao foi possivel estabelecer uma correlacao direta entre composicao da matriz, grau de condensacao da mesma e propriedades da superficie.


Introduction
Hydroformylation of olefins is the largest scale homogeneous catalytic reaction. 1Homogeneous catalysis, however, presents several drawbacks, in particular the recovering of the catalyst at the end of the process, warranting a search for two-phase or immobilized catalysts.In the last years, the sol-gel method has been applied as an alternative to immobilize soluble catalysts, using either inorganic 2 or hybrid 3 matrices.Recently, we reported the use of this approach to prepare rhenium-, molybdenumand vanadium-based epoxidation catalysts, [4][5][6] as well as ruthenium-and rhodium-based catalysts for the hydrogenation and hydroformylation of olefins, respectively. 7,8Depending on the characteristics of the matrix, leaching-free systems could be prepared even when the transition metal complex was just physically entrapped inside the porous system.In the case of rhodium complexes containing a diphosphine and no ligand bearing a hydrolysable group, only a microporous matrix would lead to recyclable catalysts. 8An interesting and robust rhodium catalyst system containing a xanthene-based hydrolysable ligand was previously described, 9 but the matrix was not characterized.We wish to report here some results concerning the immobilization of rhodium complexes in silica matrices prepared by the sol-gel method using ligands bearing hydrolysable groups and their use in the hydroformylation of olefins.

Catalyst preparation
In a typical preparation, 5 mg (~10 mmol) of the rhodium precursor and a corresponding equivalent of the ligand ([P]:[Rh]=2:1) were added to a 50 mL Schlenk flask containing 6 mL of THF.The solution was kept under stirring for 15 min, followed by the addition of 2 mL (11.10 mmol) of deionized water, 2 mL (13.56 mmol) of TMOS (tetramethylorthosilicate), 1 mL (~3.4 mmol) of the co-condensation agent (in the case of hybrid matrices), methanol (ca. 1 mL, amount needed to obtain a clean solution) and 4 drops of a 3wt.% solution of (AcO) 2 Sn(Bu) 2 in polydimethylsiloxane (Dow Corning).The solution was stirred for 15 min and allowed to stand until gelation (from one night to 8 days, depending on the case).The gel thus obtained was dried under vacuum, washed with CH 2 Cl 2 in a Soxhlet and dried again under vacuum at room temperature.The resulting yellow-orange materials were stored under air at room temperature.

Catalytic experiments
All catalytic experiments were performed in a 100 mL stainless steel Parr reactor.The reaction temperature was kept at 80 o C and the solution was stirred at 300 rpm.In a typical experiment, 2-5 µmol of the rhodium complex (~250 mg of the catalyst), 0.28 g (2-5 mmol) of 1-hexene ([Rh]/[olefin] ~1/1000), 0.1 g of cyclooctane (internal standard) and 30 mL of THF (solvent) were employed.The reactor was first purged with H 2 , and then pressurized at 50 bar (CO/H 2 = 1/1).For recycling experiments, the catalyst was separated by filtration, washed in a Soxhlet with CH 2 Cl 2 , dried under vacuum and used in a new run.All procedures were performed under air.GC analyses were carried out in an HP5890 series II gas chromatograph, equipped with an HP5 capillary column (50 m x 0.2 mm) and a flame ionization detector.Products were quantified using calibration curves obtained with standard solutions.

Catalyst characterization
Nitrogen adsorption isotherms were determined at -196 o C with a Micromeritics ASAP 2010 automated porosimeter.All calculations were performed using the associated Micromeritics software.Samples were degassed at 80 o C for a minimum of 24 h prior to measurements.
TEM images were obtained on a Zeiss CEM-902 apparatus equipped with a CCD-Proscan camera and a high speed/slow scan system controller.The samples were suspended in iso-propanol and dispersed on carbon-coated copper grids.

29
Si MAS NMR spectra were recorded on a Bruker AC 300 spectrometer using zirconia rotors and the following conditions: delay between each scan = 15 s; acquisition time = 0.1 s.Typically, 5 000 scans were accumulated.Solution 31 P NMR spectra were recorded on a Gemini 300 P instrument at 121.5 MHz, in CDCl 3 .

Results and Discussion
In order to determine the effects of the nature of the rhodium complex on the properties of the resulting matrix, four different precursors, viz.
as well as three ligands bearing hydrolysable groups, viz.HS(CH 2 ) 3 Si(OMe) 3 , Ph 2 P(CH 2 ) 2 Si(OMe) 3  and Ph 2 P(CH 2 ) 2 S(CH 2 ) 3 Si(OMe) 3 were employed.The acaccontaining complexes were expected to lead to mesoporous matrices, which would facilitate diffusion of the substrates. 8he ligand HS(CH 2 ) 3 Si(OMe) 3 would lead to a dimer whose anchoring to silica has been reported to give an almost leachless system. 16Two co-condensation agents were also tested, 1,4-bis(triethoxysilyl)benzene and 1,2bis(triethoxysilyl)ethane, aiming to reduce the degree of 3Dcross-linking, decreasing the rigidity and improving the swelling properties of the resulting matrices.The expected immobilized rhodium species are depicted in Scheme 1.The structures of systems I+XANTPHOS, and III, IV are proposed on the basis of their 31 P NMR spectra obtained before gelation.For systems II and III, the free acac -ligand would deprotonate a silanol group form the surface, producing a ≡SiO -anion. 9he same anion could be formed in system IV via a reaction between BPh 4 -and a silanol group.It must be kept in mind, however, that the ionic species depicted in Scheme 1 would be present only in a fresh catalyst: under hydroformylation conditions, the cationic species are neutralized 17 and systems II to IV would eventually be of the type [RhH(CO) x L 2 ], where L = is a neutral ligand bearing P or S as donor atoms.
Systems based on inorganic matrices are labeled with the index a; b and c are related to hybrid matrices based on 1,4-bis(triethoxysilyl)benzene and 1,2bis(triethoxysilyl)ethane, respectively.All systems were characterized by 29 Si MAS NMR and, whenever possible, also by nitrogen adsorption/desorption isotherms.

Catalyst characterization
The nitrogen adsorption/desorption isotherms of catalysts Ia, IIb and IIIa are of type I (IUPAC classification), 18 typical of microporous systems (Figure 1).A small hysteresis is clearly observed only for catalyst IIb, suggesting that the pores are mainly smooth and cylindrical, with a low contribution of mesopores. 19The Horvath-Kawazoe differential pore volume plots for these catalysts are shown in Figure 2. Catalyst IVa is characterized by a type IV isotherm (mesoporous) with a small contribution of micropores: the volume adsorbed at the lowest relative pressure represents ~20% of the total pore volume (Figure 1).Its BJH adsorption pore size distribution is shown in Figure 2. BET surface areas, pore volumes and average pore size determined from the isotherms, along with the final rhodium loading, are presented in Table 1.The amount of immobilized rhodium was always smaller than the added one, with the excess being washed away during Soxhlet extractions, which were repeated until a clean solution was obtained.
We failed to obtain isotherms for catalysts Ib and Ic.They were, therefore, characterized by transmission electron microscopy.Taking into account that these materials were able to entrap a relatively large amount of rhodium without leaching in the catalytic experiments (vide infra), their TEM micrographs, shown in Figure 3, suggest a dense structure with packed microporous domains.
29 MAS Si NMR spectra could provide a relationship between the condensation degree of TMOS and the porosity of the corresponding material.Figure 4 shows the spectra obtained for three different samples: an inorganic matrix (catalyst Ia); a hybrid matrix containing 1,4-      concentration of such ligands relatively to TMOS (molar ratios: 1:1400 for system IIb; 1:700 for systems I, III and IV).Taking into account that the molar ratio [TMOS]/[cocondensation agent] was kept constant (=4) in all preparations, a Σ Q i /Σ T i ~ 2/1 was expected.However, such ratios varied from 2.3 (Ib) to 3.8 (IIb), suggesting that some amount of co-condensation agent was not incorporated into the matrix and was washed away during the Soxhlet treatment.
Comparing the Q 4 /Q 3 ratio for systems I, the addition of a co-condensation agent appears to have little effect on the condensation degree of TMOS.The Q 4 /Q 2 ratio, however, is strongly affected, with a high increase in the relative concentration of Q 2 sites, in particular for system Ib.For the hybrid matrices, T 2 sites predominate in the case of systems Ic and IIb; system Ib, however, presents a T 2 /T 1 ratio close to one.T 3 sites are always minority.These results show that the overall 3D-cross-linking degree is indeed strongly affected by the addition of a cocondensation agent, but not always to the same extent.The highest degree of TMOS condensation is observed for systems IIIa and IVa, the only ones presenting Q 4 /Q 3 ratios higher than 1.
Nevertheless, neither the condensation degree of TMOS nor the presence of a co-condensation agent can be directly correlated with surface area, pore diameter or pore volume.It seems that the nature of the rhodium complex plays an important role in determining the final characteristics of the matrix.Evidences are the changes in Q 4 /Q 2 and T 2 /T 1 ratios when system Ia and Ib, respectively, were prepared in presence of the ligand XANTPHOS (Table 2).On the other hand, in a previous work 8 it was observed that neutral rhodium complexes bearing an acac -ligand, or even cationic rhodium complexes always led to mesoporous matrices.It should be noted, however, the absence of a hydrolysable ligand in those systems.Therefore, the microporosity observed for systems IIb and IVa was a little surprising.Also surprising were the results obtained for systems IIIa and IVa: upon reaction between the precursor complex and the ligand, the same species should be immobilized (Scheme 1) but the matrices are strongly different.This result clearly implies that the nature of free ligand plays an important role in determining the final characteristics of the matrix.The only similarity was the very small amount of entrapped rhodium species.Whether this fact is related to the high condensation degree of TMOS or to a possible cleavage of the hydrolysable ligand cannot be decided yet.However, a high condensation degree might force the hydrolysable ligand to the outside surface, which could account for the high leaching of rhodium in the absence of a chelating ligand.

Catalytic activity
All systems were tested in the hydroformylation of 1-hexene and the results are shown in Table 3. PPh 3 ([PPh 3 ]/ [Rh] = 5/1) was added to systems I since the precursor complex is not active in hydroformylation. 20When analyzing systems I, the importance of the matrix composition is clear as the catalyst based on the inorganic matrix (Ia) did not present any activity; for a hybrid matrix based on 1,2-bis(triethoxysilyl)ethane (Ic) the activity was very low but the system prepared with the co-condensation agent bis(triethoxysilyl)benzene (Ib) presented a reasonable activity and could be used in at least three runs without any rhodium leaching.The lack of activity observed for system Ia might be due to a non-interaction between the rhodium complex and the added PPh 3 .Therefore, a new gel was prepared by addition of the chelating ligand XANTPHOS ([XANTPHOS]/[Rh] = 2/1) to the initial solution containing [Rh(OMe)(cod)] 2 and HS(CH 2 ) 3 Si(OMe) 3 .For comparison, a sample of system Ib was prepared in the same way.Both systems were active although the observed n/i ratios (ratio between normal and branched aldehydes) were rather lower than those reported for monomeric rhodium catalysts containing the ligand XANTPHOS (n/i > 50). 21lthough system IVa presented the highest n/i ratio observed in this work, its catalytic activity was very low.This may be due to mass transfer limitations arising from a very  small pore diameter of the matrix (Figure 2).In contrast, the activity of system IIIa was due to leached rhodium, not retained by a matrix characterized by a large distribution of mesopores (Figure 2).System IIb turned out to be more active than all other systems prepared in this work.It was active in at least four runs, with a total TON of 1510, albeit some deactivation from run to run.When tested in the hydroformylation of 1-decene in the absence of a solvent ([Rh]/[olefin] = 1/4540), a TON of 2450 was observed after 24h (n/i = 1.5;H/I = 2.7).It was also active in the hydroformylation of 1-octadecene: using the same conditions employed for 1-hexene, a TON = 530 was obtained (close to the 570 value observed for 1-hexene in the first run, Table 3), with n/i = 2.3 and H/I = 1.7.Thus, in spite of a microporous matrix, higher linear substrates could be hydroformylated.
Taking into account that a chelating hydrolysable ligand was employed, the location of the rhodium complex (inside the porous system or on the external surface of the matrix) is a question that could be raised.Therefore, system IIb was tested in the hydroformylation of limonene, a voluminous substrate (volume ~0,435 nm 3 ). 22Within 24h no reaction was observed.In the same experimental conditions, with the system [RhCl(CO) 2 ] 2 /BINAP in solution a 21% conversion to hydroformylation products was obtained in 4h.These results strongly suggest that the complex is indeed inside the porous system of the matrix and that the deactivation observed in successive runs is due to a degradation of the complex upon manipulation after each reaction (always performed under air).

Conclusions
Although it seems that the nature of the rhodium complex plays an important role in determining the final characteristics of the matrix, no direct correlation between rhodium complex, matrix composition, condensation degree and surface properties could be found.Therefore, a straightforward way to prepare a desired matrix with any metal complex remains to be found.The system based on [Rh(CO) 2 (acac)]/ Ph 2 P(CH 2 ) 2 S(CH 2 ) 3 Si(OMe) 3 was active in the hydroformylation of 1-decene in the absence of a solvent, without any rhodium leaching.In spite of its microporous matrix, this catalyst was also active in the hydroformylation of 1-octadecene.The lack of any activity in the hydroformylation of limonene suggests that the complex is indeed entrapped inside the porous system of the matrix.

Figure 2 .
Figure 2. Pore size distribution for some systems.

Table 1 .
Composition and surface characteristics of the catalysts

Table 2 .
Concentration of silicon sites as obtained by29Si MAS NMR