Activity and stability studies of H-transfer reduction reactions of aldehydes and ketones over aluminium isopropoxide heterogenised catalysts

Aluminium isopropoxide Al(OiPr)3 immobilised on various mesoporous supports (SiO2, TiO2 and γ-Al2O3) was tested for H-transfer reductions of various aldehydes and ketones in the presence of 2-propanol as a sacrificial agent. The heterogenised catalysts were characterised by N2 physisorption, XRD, SEM-EDX, FTIR and ICP-OES. The characterisation results show a successful grafting of the homogeneous aluminium isopropoxide catalyst, covalently bound to the solid surface, with high dispersion over the mesoporous supports. The heterogenised catalysts show an excellent catalytic activity with high selectivity towards the desired alcohol product, with performances that are comparable with those of the homogeneous Al(OiPr)3 catalyst. Al(OiPr)3 grafted on SiO2 shows higher activity compared to γ-Al2O3 and TiO2 supported catalysts. The catalysts remain very active after 5 cycles of reuse and no leached Al(OiPr)3 was found in the reaction mixture, hence showing an excellent stability. The work reported here shows that it is possible to effectively immobilise catalytic functions, usually working in the homogeneous phase, over solid supports, with the resulting heterogenised catalysts keeping the same catalytic activity of the homogeneous counterpart and excellent stability, and with the advantage of being able to recycle and reuse them, without loss of catalytic materials.


Introduction
H-transfer reduction of aldehydes and ketones through the Meerwein-Ponndorf-Verley (MPV) reaction is a well-known route for the production of the corresponding alcohols. Typically, a secondary alcohol, such as 2-propanol, serves as the hydride donor. The reaction involves a hydride transfer from a secondary alcohol to a carbonyl substrate via a six-membered transition state, initiated by the activation of the carbonyl groups by coordination to the Lewis acidic aluminium centre. 1,2 In unsaturated aldehydes and ketones, the double bond is untouched and no saturated counterparts are observed during the MPV reduction, which makes the reaction chemoselective towards the reduction of the carbonyl group only, which is a key advantage of such reactions. As a result, the MPV reduction provides a practical method for synthesising unsaturated alcohols, many of which are crucial raw materials for the synthesis of ne chemicals.
Traditionally, aluminium alkoxides, such as aluminium isopropoxide and other aluminium complexes, have been used to homogeneously catalyse the reaction. [3][4][5][6] Aluminium precatalysts derived from alkyl aluminium complexes have shown high activity for MPV reduction of aldehydes and ketones in toluene. 7 In addition to aluminium alkoxide complexes, other metal complexes, such as lanthanum alkoxides, 8 have been reported to be active catalysts for MPV reduction of carbonyl compounds. However, most of these catalysts are homogeneous and are therefore difficult to separate from the reaction mixture, which makes their reuse challenging, if not unfeasible. In this context, the use of solid catalysts would certainly be advantageous from a practical perspective; hence the search for reusable solid catalysts, with similar effectiveness as the homogeneous ones, has oen been pursued in recent decades. Lopez et al. reported the reduction of 4-tert-butylcyclohexanone using mixed oxides obtained from hydrotalcites, NaBEA zeolites, KF/alumina and La 2 O 3 . 9 The ndings suggest that basic sites are the catalyst active sites. 9 Other solid catalysts such as BEA zeolites, 10 magnesium phosphates 11,12 and zirconia 13 have been previously reported. However, these catalysts suffer from poor selectivity. The use of supported homogenous catalysts has emerged as an alternative to the inadequate separation of homogenous catalysts and the poor selectivity of conventional heterogeneous solid catalysts. Supported metal alkoxides such as zirconium isopropoxide, 14,15 boron isopropoxide, 16 indium isopropoxide, 17 lanthanum alkoxides 9 and ceric alkoxide 18 have been reported to be active catalysts for MPV reduction. MCM-41 was employed by Anwander et al. as a support material for the graing of aluminium isopropoxide. 19 It was discovered that the hybrid system was particularly active in the MPV reduction of 4-tert-butylcyclohexanone. The immobilised catalyst demonstrated excellent activity with 88% conversion of 4-tert-butylcyclohexanone aer 5 hours and >99% conversion aer 24 hours. However, the sacricial alcohol must be thoroughly dried to achieve good catalytic yield. 19 Despite the application of a range of aluminium alkoxide catalysts in MPV reduction, the active aluminium species are still poorly understood. In the crystal form of the aluminium isopropoxide catalyst, the six-coordinated aluminium centre is surrounded by three bridging Al(O i Pr) 3 groups, but in solution a variety of species forms and their simple interconversions makes it difficult to predict the active aluminium sites. 20 As previously mentioned, the use of homogeneous catalysts has also some drawbacks in terms of catalyst separation, reuse and recycling, which is oen unfeasible. In order to overcome these practical issues associated with homogeneous catalysis, the heterogenisation of aluminium isopropoxide over solid supports is appealing since it makes the catalytic function insoluble, easy to recycle and reuse, with the possibility to be engineered also in continuous xed-bed reactors. 21,22 Whilst SiO 2 -based materials have previously been investigated as potential supports, 19,23 the use of other types of supports has remained largely unexplored.
In this study, we investigate H-transfer reduction reactions of carbonyl compounds (aldehydes and ketones) catalysed by aluminium isopropoxide immobilised over mesoporous solid supports. The use of high surface area supports for the immobilisation is expected to enhance the dispersion of the aluminium isopropoxide catalyst. The inuence of the type of support studied, namely SiO 2 , TiO 2 and g-Al 2 O 3 , was investigated. Several aldehydes and ketones were used in the reaction screening and the results on catalytic performances are compared with those of the same aluminium isopropoxide catalyst in homogeneous solution.

Preparation of supported catalysts
Aluminium isopropoxide Al(O i Pr) 3 was graed onto the various supports according to a method previously published in the literature. 24 The support (SiO 2 , TiO 2 , g-Al 2 O 3 ) was dried for 4 hours at a temperature of 250°C prior to graing. The heterogenised homogeneous catalyst was made by mixing 5 mmol of aluminium isopropoxide in 25 ml of dry n-hexane with 2 g of the support (SiO 2 , TiO 2 , g-Al 2 O 3 ). The mixture was reuxed at 69°C for 12 hours while being agitated at 500 to 700 rpm. The suspension was ltered in a N 2 atmosphere, washed with nhexane three times, and dried in an inert environment. Al(O i Pr) 3

Characterisation of materials
The surface area of the supports and graed catalysts were analysed using a Micromeritics surface area analyser. Testing vials were rst heated at 100°C for 2 hours and purged under N 2 gas. All samples were heated to 300°C under vacuum for 6 hours. About 0.1-0.2 g of sample was used. Nitrogen adsorption and desorption isotherms were recorded at liquid nitrogen temperature of −196°C. Specic surface area was calculated using Brunauer-Emmett-Teller (BET) method while pore size analysis was conducted by the Barrett-Joyner-Halenda (BJH) method using the adsorption branch. The crystalline phases of the support and prepared catalyst were analysed using X-ray diffraction measurements. The XRD patterns were collected using Philips X'Pert Xray diffractometer operated at 40 kV and 40 mA with a Cuka1 X-ray source (l = 0.154 6 nm) in a 2q range of 20°to 80°with 0.02°step size. To identify the relevant phases, the XRD pattern were analysed using JADE 6 (Material Data Inc., Livermore, CA) to compare with standard structures in the International Centre for Diffraction Data (ICDD) database. Surface morphology and EDX analysis of the samples was performed using Quanta 250. Samples were prepared by rst dissolving in ethanol and sprinkled onto carbon tape stuck to an aluminium stub. To make the samples conductive, the samples were dried under light and coated with platinum using Cressington Platinum (Pt) Sputter Coater for about 50 s (approximately 10 nm thickness). The actual quantity of aluminium accessible in the catalysts was determined using inductively coupled plasma optical emission spectrometry (ICP-OES, Plasma Quant PQ 9000). Prior to ICP analysis, the catalysts were typically microwave-digested in acid solution (HCl, H 2 SO 4 and HNO 3 ). An acid solution containing aluminium was used as standard reference. The FTIR spectra of the samples were collected using a Bruker Vertex 7.0 Fourier transform infrared (FTIR) spectrometer with a scanning wavenumber (ranging from 400 to 4000 cm −1 ) and a spectral resolution of 4 cm −1 .

Reaction studies
In a 50 ml round bottom ask equipped with reux condenser, thermometer and a magnetic stirrer, the heterogenised catalyst (200-220 mg) was added to the reaction mixture containing 1.4 mmol of the carbonyl compound of interest and 60 mmol of 2-propanol. Anhydrous 2-propanol was used as previous studies suggest that drying of the 2-propanol could signicantly improve catalytic activity of aluminium isopropoxide catalyst in MPV reduction. 19 An excess of 2-propanol was also necessary to shi the equilibrium reaction to the desired product. The mixture was stirred at 750 rpm and heated to reux. The mixture was analysed using Agilent 7820A gas chromatography system equipped with FID detector and a HP-5 methylpolysiloxane column (30 m × 320 mm × 0.25 mm). Products were identied by their retention time and compared with authentic samples. Yield was calculated using eqn (S1) of the ESI. † At the end of each round of reaction, the spent catalyst was recovered by ltration. The reuse of the catalyst was tested choosing the reduction of propionaldehyde as benchmark reaction. Aer the reaction, the spent catalysts were washed several times with 2-propanol, dried at 80°C for 6 hours and subsequently tested again for the reduction of fresh propionaldehyde through 5 rounds of reaction under the same conditions.
To assess for leaching, 200 mg of the heterogenised catalyst, Al(O i Pr) 3 -SiO 2 , Al(O i Pr) 3 -TiO 2 and Al(O i Pr) 3 -Al 2 O 3 , was reuxed in 60 mmol 2-propanol at 80°C for 4 hours. The solution was ltered, and the ltrate was evaluated for reduction of propionaldehyde.

Characterisation of supports and heterogenised catalysts
The N 2 adsorption-desorption isotherms for the SiO 2 support and Al(O i Pr) 3 -SiO 2 with corresponding pore size distribution (PSD) are shown in Fig. 1 show a type IV isotherm. The adsorption isotherm can be categorised into three parts: the monolayer multiple adsorptions of N 2 on the surface of the mesoporous material, capillary condensation of the N 2 into the mesopores and then saturation. The adsorption and desorption isotherms for both samples are not superimposed. A phenomenon referred to as hysteresis, which is linked to capillary condensation occurring in the mesopores, is an indication that the materials are mesoporous. The capillary condensation for untreated SiO 2 occurs at relative pressure p/p°= 0.50-0.82. Aer graing with aluminium isopropoxide, the capillary condensation is observed to shis towards lower region of the relative pressure. This suggests changes in the mesoporous structure of the support aer incorporation with the homogeneous aluminium isopropoxide catalyst as suggested by the data in Table 1. As expected, the graing of aluminium isopropoxide decreases the surface area of the SiO 2 and g-Al 2 O 3 supports as well as the pore volume and pore diameter ( Table 1). The Brunauer-Emmett-Teller (BET) surface area, S BET and pore volume, V p of Al(O i Pr) 3 -SiO 2 (S BET = 408 m 2 g −1 and V p = 0.4 cm 3 g −1 ) are slightly lower than those of SiO 2 (S BET = 434 m 2 g −1 and V p = 0.73 cm 3 g −1 ). Similarly, the S BET and V p of Al(O i Pr) 3 -Al 2 O 3 (S BET = 212 m 2 g −1 and V p = 0.35 cm 3 g −1 ) are lower than those of g-Al 2 O 3 (S BET = 240 m 2 g −1 and V p = 0.74 cm 3 g −1 ). The N 2 adsorption-desorption isotherms for the TiO 2 support and Al(O i Pr) 3 -TiO 2 with the corresponding pore size distribution (PSD) are shown in Fig. 2(a) and (b), respectively. TiO 2 and Al(O i Pr) 3 -TiO 2 show a type III isotherm, an indication that the material is mesoporous with weak adsorbate-adsorbent interaction. 25 The graing of aluminium isopropoxide again decrease the surface area of the TiO 2 support as well as the pore volume and pore diameter (Table 1), but the material still retains its characteristics of ordered mesoporous material. The observed decrease in surface area, pore volume and pore   diameter may be due to plugging of several pores of the mesoporous material by the aluminium isopropoxide catalyst. 23 ICP-OES analysis was used to assess how much aluminium was graed across the three support samples. The Al content over the silica and alumina support was found to be almost similar. The result show 6.45 wt% aluminium content for the alumina support while Al(O i Pr) 3 -SiO 2 shows a slightly higher content of 6.61 wt%. The slightly lower percentage of Al found in the Al(O i Pr) 3 -Al 2 O 3 sample can be attributed to limited accessibility to the surface hydroxyl groups in Al(O i Pr) 3 or due to variation in physicochemical properties between the two mesoporous supports (Table 1). Al(O i Pr) 3 -TiO 2 has the lowest Al content, 5.99 wt%, in comparison to the other samples. The lower Al loading may be attributed to the smaller surface area and pore volume than those found in Al(O i Pr) 3 -SiO 2 and Al(O i Pr) 3 -Al 2 O 3 catalysts (Table 1). A previous study indicates that smaller surface area could lead to lower catalyst loading by restricting the graing to not more than a monolayer. 14 Al(O i Pr) 3 -SiO 2 , out of the three catalysts, exhibits the highest Al content, possibly indicating a higher proportion of silanol groups for the stabilisation of the Al species.
The XRD patterns of the SiO 2 support and the graed Al(O i Pr) 3 -SiO 2 are presented in Fig. 4. The pattern for pure SiO 2 support shows a characteristic peak of the amorphous silica broad at approximately 10-30°( Fig. 4(a)). Aer graing, Al(O i Pr) 3 -SiO 2 displays similar pattern as the support as shown in Fig. 4(b), which indicates that the homogeneous aluminium isopropoxide catalyst is highly dispersed on the support. The   intensity of the peak is also observed to weaken aer graing of the aluminium isopropoxide, which may be related to the integration of aluminium complexes into the channels of the SiO 2 substrate. Moreover, the observed decrease in intensity is an indication of reduced pore size aer graing, as suggested by the surface area and pore volume size (Table 1) only peaks from TiO 2 emerge and no peak from aluminium isopropoxide could be detected. This indicates that at appropriate loading, aluminium isopropoxide is highly dispersed on the surface of the TiO 2 support (Fig. 5(b)). The X-ray diffraction pattern of the g-Al 2 O 3 support shows peaks at 2q = 37.3°, 39.5°, 46.5°, 61.1°, 67.1°and 85.2°, which match closely with those on JCPDS le (46-1131), (Fig. 6(a)). Al(O i Pr) 3 -Al 2 O 3 shows a similar pattern to that of the g-Al 2 O 3 support as no new peaks were detected ( Fig. 6(b)).  27 The strong band at 1060 cm −1 is attributed to Si-O-Si/Si-O-Al asymmetric stretching. 28 The band at 970 cm −1 is assigned to Si-OH stretching. 29,30 Comparing the 970 cm −1 peak of the SiO 2 support to the 945 cm −1 peak of the graed Al(O i Pr) 3 -SiO 2 , we found that the latter shis toward a lower wave number. This may be explained by the formation of Al-O-Si bonds, which result from the interaction of aluminium and silicon atoms. 31 This also suggests that the Al species are covalently bound to the surface of the SiO 2 support.    The peaks observed at 3400 and 3394 cm −1 for TiO 2 and Al(O i Pr) 3 -TiO 2 are attributed to symmetric and asymmetric vibrations of Ti-OH. 32 The absorption band observed at 730 and 728 cm −1 are associated with Ti-O-Ti stretching vibrations. For pure titanium oxide, the contributions from the anatase titania are visible in the range of 400-800 cm −1 .
The FTIR spectra of the gamma alumina measured between 400 to 4000 cm −1 is shown in Fig. 9. Scanning electron microscopy was used to examine morphology and size distribution of the support and graed catalyst. Elemental dispersive X-ray spectroscopy (EDX-SEM) was used to examine the elemental distribution and content

Catalytic H-transfer reaction activity of the heterogenised catalysts
The feasibility of using homogeneous aluminium isopropoxide for H-transfer reduction though MPV reaction of propionaldehyde has been previously reported. [33][34][35] In this work we describe the preparation and characterisation of the heterogenised aluminium isopropoxide over three supports (SiO 2 , TiO 2 and g-Al 2 O 3 ). Whilst SiO 2 -based supports have been previously screened for heterogenisation of metal alkoxide species, 17,24,36,37 the use of other supports remained unexplored. It is therefore of interest to test the general validity of metal alkoxide heterogenisation over supports for MPV reductions. The synthesised catalysts were initially evaluated for the reduction of propionaldehyde using 2-propanol as the hydride source. The results show a very good yield up to 90% and selectivity of 100% for Al(O i Pr) 3 -SiO 2 . Table S1 of the ESI † shows the catalytic yield and selectivity for other aldehydes and ketones over the heterogenised catalysts Al(O i Pr) 3 -SiO 2 , Al(O i Pr) 3 -TiO 2 and Al(O i Pr) 3 -Al 2 O 3 . As observed from the data, the corresponding unsaturated alcohols were formed by reducing the unsaturated aldehydes and ketones with high selectivity and no other reduction products were found. A series of other carbonyl compounds was tested and generally good catalytic yields to the corresponding alcohol for cinnamaldehyde, benzaldehyde, cyclohexanone and acetophenone were obtained. Aer 3 hours, benzaldehyde was highly reduced to benzyl alcohol with a yield ranging between 87.7-95.7% across the three heterogenised catalysts. Out of the studied carbonyl compounds, acetophenone has the lowest yield (<50%) while cinnamaldehyde was converted into its corresponding alcohol with a 55.4% yield aer 6 hours. The trend observed in reactivity across the carbonyl compounds could be due to steric hindrance effects. Smaller aldehyde molecules, such as propionaldehyde and benzaldehyde, could have more access to the catalyst active sites. Diffusion and mass transfer effects could also play a role in restricting more bulky molecules from reaching catalytic active sites. 14 It is noted that the unmodied SiO 2 , TiO 2 and g-Al 2 O 3 were tested and found to be totally inactive for the reduction reactions, demonstrating that the reaction is mainly catalysed by the Al(O i Pr) 3 immobilised over the surface.
In terms of comparing activity across different supports used for the catalyst preparation, the activity of the catalysts is in the order Al(O i Pr) 3 Table 1. The effectiveness of MPV reduction reactions is known to be affected by Lewis acid sites of the metal centre and its ligand exchange abilities. 3,4 However, these factors are signicantly affected by the type of ligand and coordination geometry at the metal centre. 19 Al species prepared by graing method on siliceous materials are known to contain more acid aluminium centres compared to those prepared by other conventional methods. 38 The presence of aluminium coordinated in the form of heavily distorted tetrahedrons is hypothesised to be responsible for the presence of more acidic aluminium centres, which enhance the catalytic activity of such materials. 19 Another important advantage of the graed catalyst compared to the homogeneous one is that the aluminium alkoxide molecules cannot self-assemble due to surface connement, and the support material, for example silica, is known to serve as an electron-withdrawing medium. 19 To compare the activity of the heterogenised catalysts with the homogeneous counterpart, the same amount of the aluminium isopropoxide (Al(O i Pr) 3 ) catalyst found in the heterogenised catalysts was used for the reduction of the studied aldehydes and ketones in homogeneous phase. As shown in Table 2, the turnover  24 This nding is particularly worth highlighting as it is usually reported that when graed over supports, homogeneous catalysts have a markedly reduced catalytic activity. 40 In our case, the catalytic activity remains essentially unaltered aer the homogeneous catalyst is immobilised, with possibility to achieve this over different supports, which suggests that it is possible to run such reactions effectively whilst being able to easily separate and recycle the catalyst. The excellent catalytic activity and high selectivity of the heterogenised catalysts could be attributed to the presence of well-dened, single-site catalytic centres over the solid support, which also minimise formation of byproducts. 41 Constraint effects could also play a role in the high performance of the supported catalysts. 42,43

Effect of support surface
Using the same amount of Al(O i Pr) 3 , the catalytic activity of the graed catalyst on SiO 2 , TiO 2 , and g-Al 2 O 3 was evaluated. A high catalytic performance was recorded for all the three catalysts in the reaction. Compared to Al(O i Pr) 3 -SiO 2 and Al(O i Pr) 3 -Al 2 O 3 catalysts, the Al(O i Pr) 3 -TiO 2 heterogeneous catalyst displays a lower yield. This could be due to the smaller surface area of the TiO 2 support, which can affect the aluminium isopropoxide dispersion. It was observed that the Al(O i Pr) 3 -SiO 2 catalyst had similar activity to Al(O i Pr) 3 -Al 2 O 3 catalyst for MPV reduction of unsaturated aldehydes and ketones. These ndings demonstrate that the slight differences in the structural characteristics of the two support materials, SiO 2 and g-Al 2 O 3 , have little impact on the catalytic activity, with Al(O i Pr) 3 -SiO 2 showing slightly higher activity than Al(O i Pr) 3 -Al 2 O 3 . Mesoporous materials with a large surface area, such as SiO 2 and g-Al 2 O 3 , offer a better support for the heterogenisation of metal alkoxides, such as Al(O i Pr) 3 . For most liquid phase processes, it is generally required for the support to have a relatively high surface area (usually >100 m 2 g −1 ) such that the active sites are well dispersed and easily accessible on the surface of the support. 44 Additionally, appropriate pore size is required to allow easy diffusion of reactants to the active sites of the catalyst. 45 As such, the pore structure of these support materials, SiO 2 and g-Al 2 O 3 , likely favours easier access of the substrate to the aluminium centres, thus facilitating the MPV reduction reaction.

Activity of reused catalysts
Aer the graed catalyst was reuxed in 2-propanol, the ltrate was found to be inactive for the MPV reduction of propionaldehyde, ruling out the possible presence of any leached materials in the reaction medium. The recovered catalyst was evaluated in a subsequent batch reaction aer being cleaned with 2-propanol following each round of reaction. Fig. 13 shows the activity of the three supported catalysts aer ve rounds of reaction, using propionaldehyde as a model reaction for testing catalyst stability. The information in Fig. 13 shows that the activities of Al(O i Pr) 3 -SiO 2 , Al(O i Pr) 3 -TiO 2 , and Al(O i Pr) 3 -Al 2 O 3 for the MPV reduction of propionaldehyde remain high even aer 5 rounds of reaction and the selectivity to 1-propanol in each case was close to 100%. This demonstrates that the heterogenised catalysts have an exceptional stability.  XRD pattern of the recycled catalysts were obtained aer 5 rounds of reuse as shown in Fig. 4(c), 5(c) and 6(c). The XRD pattern of all the three recycled catalysts (Al(O i Pr) 3 -SiO 2 , Al(O i Pr) 3 -TiO 2 , Al(O i Pr) 3 -Al 2 O 3 ) maintained characteristic peaks similar to the initial substrates. This clearly indicates that even aer rounds of reaction, the materials still retain their pore structure. The observed decrease in intensities of the peak may be due to plugging of the pores aer several rounds of reuse.

Conclusions
We demonstrated that heterogenised catalysts synthesised by graing of aluminium isopropoxide over mesoporous supports (SiO 2 , TiO 2 and g-Al 2 O 3 ) are efficient and stable catalysts for the MPV reduction of various aldehydes and ketones. The use of mesoporous supports improves the dispersion the Al(O i Pr) 3 , hence resulting in high catalytic activity and selectivity of the catalysts. Higher catalytic activity was observed in Al(O i Pr) 3 -SiO 2 compared to Al(O i Pr) 3 -TiO 2 , and Al(O i Pr) 3 -Al 2 O 3 . This may be due to differences in surface area and pore volume that could restrict the carbonyl compounds from accessing the catalyst active sites. All the heterogenised catalysts have an activity that is essentially the same of the homogeneous counterpart and show an excellent stability, displaying the ability to be recyclable up to 5 rounds of reaction with no signicant decrease in activity. In summary, the work clearly shows that it is possible to adopt strategies of homogeneous catalyst immobilisation to perform chemical reactions, usually performed in homogeneous phase, using solid catalysts that are able to keep the same activity as the homogeneous counterpart and with excellent stability, which allows their ease of separation and reuse, hence avoiding waste of valuable catalytic materials. The approach adopted here could be further explored to develop strategies for more complex organic syntheses involving drug molecules, natural products or a wide range of substrates, including sterically hindered carbonyl compounds.

Data availability
The data that support the ndings of this study are available within the article and the ESI. †

Conflicts of interest
There are no conicts to declare.