Non-precious Metal Catalysts for Acetic Acid Reduction

Acetic acid (AA) hydroconversion was studied over various monometallic (Fe, Co, Ni, Cu, Zn, Pt) and bimetallic (doped with In as second, guest metal) catalysts loaded on a highly mesoporous, fumed silica support. The transformations were investigated in a fixed bed, flow-through reactor in temperature range of 240-380°C using hydrogen flow at 21bar total pressure. The catalyst precursors were activated in H2 flow at 21bar and 450°C as routine pre-treatment. Catalytic performances of the studied metal catalysts have nothing in common. Activities and the yields of main products contrast strikingly. Diversity of catalytic behaviour reflects the complexity of the surface reaction network. Contrary to the highly ethanol selective Co or Cu forms, over Ni catalyst mainly methane was produced. Indium doping can completely eliminate the hydrodecarbonylation activity and turn to the consecutive reduction route resulting in high ethanol yield. Metallic phases of different peculiarities can offer promising contacts for upgrading various oxygenates obtainable from biomass degradation.

TiO 2 , SiO 2 , Al 2 O 3 and Fe 2 O 3 and Pt powder catalyst without support. Hydrodecarbonylation and decomposition reactions dominated on Pt powder, Pt/SiO 2 and Pt/Al 2 O 3 similarly to that we found for Ni/Al 2 O 3 catalyst [13]. Vannice et al. [16] obtained the best ethanol selectivity using a titania-supported Pt catalyst. Efficient hydrogenation of long chain fatty acids to alcohols was recently reported [17] also using Pt/TiO 2 catalysts compared to other supports (SiO 2 , Al 2 O 3 , CeO 2 , ZrO 2 ). Pt/ TiO 2 catalysts were found the most active for AA hydrogenation and provided the highest ethanol selectivity (up to 70 %) [16]. We confirmed the results with titania (mainly anatase) support [13], as reported in ref. 16. Alcala et al. used fumed silica support (CAB-O-SIL (Cabot Corp.)) for supporting Pt [18] and could not produce ethanol selectively in accordance with results of Vannice et al. [16] who applied a quite different SiO 2 (Davison, Grade 57). A recent study [19] concerns the performance of various commercial supported noble metal catalysts (Ru/C, Pd/C, Pt/C, Ru/Al 2 O 3 , Pd/Al 2 O 3 and Pt/Al 2 O 3 ) in aqueous phase hydrogenation of AA. Technically aqueous-phase hydrogenation of AA over transition metals is more important. Among commercial catalysts Ru/C seemed to be the most efficient [20][21]. Use of computational methods DFT calculations suggest directions for designing new transition metal catalysts [20][21]. However, these catalysts still proved to be as poor as the supported Pt catalysts used by Vannice et al. [16]. However, the addition of Sn to Pt/SiO 2 catalysts inhibits the decomposition of AA to H 2 O, CO and CH 4 and ethanol was selectively produced [18] similarly to our findings when In was added to other metals and supports [12,13]. Using indium co-catalyst side reactions could be completely eliminated regardless of the support TiO 2 or Al 2 O 3 [12].
The support can play important role as examples demonstrated above. For example it can simple enhance the specific surface area of the active metal and/or affect the activity and the selectivity on various ways. For example, formation of alloys or intermetallic compounds, change in the reducibility, stabilization of the metal nanoparticles, homogeneous distribution inside the pore structure of the support, control the migration of metal atoms to bigger aggregates can be important. The Cu and Ni hydrotreating agents, especially with indium doping were found highly efficient in H 2 activating processes for step by step reduction of carboxylic acids. Pioneering research applying indium doping is continued in the present study. The support from one component (Al 2 O 3 ) of earlier profoundly investigated crystalline aluminosilicates (zeolites) for the second one (SiO 2 ) is changed. Picked up different types of silicas and compared with commercial alumina support, extra-special fumed silica (commercial CAB-O-SIL M-5 similarly to ref. 18) can be concluded to be a quite inert and highly mesoporous support for loading various hydrogenating metals.
The literature describes catalysts for AA hydrodeoxygenation (HDO) comprising of one or more noble metals of Group VIII dispersed on oxides of elements of Group III/A or IV/A. However, over monometallic forms hydrogenolysis of C-C bonds is a characteristic reaction. In such cases, hydrodecarbonylation may be the main reaction route hardly producing alcohols. Non precious metals of Group VIII have been more rarely studied. Among cheaper non noble metal hydrogenating agents (Cu and Ni) the adjacent cobalt in the periodic table is also highly efficient in some important H 2 activating processes. Cressely et al. studied acetic acid hydrogenation at atmospheric pressure over Fe, Co or Cu dispersed on an inert silica support [22], and Co loaded on various supports or using bimetallic combinations of the mentioned metals [23]. Their results showed that Cu/ SiO 2 was active and quite selective for the production of ethanol, acetaldehyde and ethyl acetate through consecutive reduction of acetic acid, whereas Fe/SiO 2 resulted in the production of acetone and CO 2 following the ketonization reaction route. However, over Co/SiO 2 , dominantly decomposition products (CH 4 and CO 2 ) were obtained. The high selectivity of Cu/SiO 2 for ethanol, acetaldehyde and ethyl acetate production was confirmed one and half decade later [24] and in our recent studies. However, a patent is of an age with ref. 19 and 20 claims that under super-atmospheric pressure over cobalt-containing catalysts high ethanol production is dominant [25]. Present work aimed to clarify discrepancies found in the literature as for the behaviour of cobalt and to compare succeeding metals (Fe, Co, Ni, Cu, Zn) adjacent in the periodic table supported on fumed silica in carboxylic acid hydroconversion using acetic acid as model compound. For comparison with noble metals from VIII group platinum was also applied from the Pt group in the periodic table under nickel. Applying our novel, solid state indium doping method, amalgamation with nascent indium, twelve different catalysts can be obtained, an ensemble of which can provide wide facilities in carboxylic group reduction.

Characterization
Nitrogen adsorption measurements were carried out at -196°C using Quantochrome Autosorb 1C instrument. Before measurements, the samples were outgassed under vacuum at 350 °C for 24 h. The specific surface area was determined by the BET method. The pore size distribution was calculated from desorption branch of the isotherms by the BJH method.
The XRD patterns of the catalysts were recorded by Philips PW 1810 diffractometer applying monochromatized Cu α radiation (40kV, 35mA) at elevated temperatures in hydrogen flow using a high temperature XRD cell (HT-XRD).
The reducibility and the extent of reduction of the samples were examined by temperature-programmed H 2 -reduction (H 2 -TPR) using a flow-through quartz microreactor. About 30mg catalyst was pretreated in a flow of 30 ml min -1 nitrogen at 350°C for 1h. The pretreated sample then cooled to room temperature in the same N 2 flow before was contacted with a 30 cm 3 min −1 flow of 9.7%H 2 /N 2 mixture. The reactor temperature was ramped up at a rate of 10 °C min −1 to 800°C and maintained for 1h at latter temperature. The effluent gas was passed through a liquid nitrogen trap and a thermal conductivity detector (TCD). Data were recorded and processed by computer. Calculation of the corresponding hydrogen consumptions based on the peak areas was carried out by using the calibration value determined with the H 2 -TPR of CuO reference material.

Investigation of catalytic properties
The catalytic hydrogenation of AA (96% Molar Chemicals Kft, Hungary) was carried out in a high-pressure fixed bed flowthrough reactor (16 mm inner diameter and 120 mm length) at 21bar total pressure in the temperature ranges of 240-380°C. Space time was 1g AA /g cat.• h WHSV, applying 2 g of catalysts and 2 g/h mass flow of AA. The reactor effluent was cooled to room temperature and the liquid and gas phase products were separated. The liquid was analyzed by a gas chromatograph (GC, Shimadzu 2010) equipped with flame ionization detector and a Restek Rt-U-BOND capillary column. The gas was analyzed by an on-line GC (HP 5890) equipped with thermal conductivity detector and Carboxen 1006 PLOT capillary column. The activity and the selectivity of the catalysts were characterized by product distributions represented by stacked area graphs. In this representation, the distance between two neighboring curves gives the yield of the specified product in mole percent.

Results and Discussion
This study reveals that for the hydroconversion of carboxylic acids the role of novel bimetallic phases in the catalysts can be decisive, but non-precious monometallic forms can be rivals. Furthermore, the properties of the supports seem to be also quite significant under formation of metallic particles and under the reaction conditions. The variety of the most suitable catalysts for hydroconversion of bioacids is expanded in this study. Further improvement can be attained by optimization of the support and the catalyst preparation technique. First hand experiences showed, that highly mesoporous fumed silica is advantageous of overriding. Nature of silica surface excludes side reactions (e.g. mono-or bimolecular dehydration) which are characteristic for alumina and other supports having significant acidity. Esterification of the reactant acetic acid and primary product ethanol results in ethyl acetate as secondary product. The rate of this reaction can be also increased by the increasing number of acidic sites on the surface of the support (compare alumina and silica supports in Fig. 1). Beside the surface properties of the silica supports their pore structure, morphological feature was turned out a success. Loading fumed silica support with copper much higher activity can be attained than over the alumina supported catalyst (compare in Fig. 1), although the specific surface areas are practically equal (~200m 2 /g). This study applies the quite inert fumed silica support to compare the properties of various metal particles so clearly as possible besides limited support influences. This fumed silica support is a loose mesoporous material as reflected by the nitrogen ad-and desorption isotherm and the pore size distribution shown in Fig. 2. The sample has 200 m 2 /g specific surface area and extremely wide pore size distribution around 27 nm. This feature can determine fundamentally the formation, consequently the quality of active metallic particles and the mass transport properties in the catalysts highly influencing the catalytic performances under reaction conditions. The isotherms of all derivatives (precursors, mono-or bimetal loaded forms, too) are practically overlapping resulting in nearly the same specific surface area. This means that, there is no significant influence of the various mono-or bimetal particle formation for the accessibility of full volume inside the catalyst particles because of these three-dimensional, quite capacious pore structures exclude the pore blocking effect. However this is true only for the mesoporous supports. Fig. 2 represents the porosity of the cobalt catalysts. Decreased nitrogen adsorption, consequently lower specific surface area (instead of 563 m 2 /g 298 m 2 /g) can be obtained alone on the microporous silica gel following the cobalt loading because of pore blocking. Among mesoporous supports the fumed silica is a most opened pore structure (see pore size distributions in Fig. 2) which seems to be highly advantageous from catalytic point of view (Fig. 1).
One reason of the lower activity of alumina supported catalysts compared with silica based samples can be the difficult reduction of precursor oxides bound solidly on the alumina surface compared to the more inert silica. The H 2 -TPR curves recorded on alumina and the highly mesoporous silica reflect important differences induced by diverse supports.
Complete reduction of all precursors was found in the measured temperature range on the base of mass balances of added metal oxides and hydrogen. Applying identical linear heating up program for varied oxide precursors in hydrogen flow the obtained H 2 -TPR curves demonstrate that presence of alumina can affect intensely the reduction of supported oxides (e.g. NiO). Contrary, NiO loading on SiO 2 (1) can be completely reduced into pure nickel phase at low temperature below 300 °C and in a very narrow range without appearance of two possible oxidation steps. Cobalt shows highly similar behaviour as nickel (only formation of Co/SiO 2 (1) is shown in Fig. 3). Just the opposite is the behaviour of In 2 O 3 , the precursor of modifying guest metal (containing larger particles than 100 nm) which simply admixed to SiO 2 (1) powder by grinding can be fully reduced only up to 700 o C. The three possible oxidation steps can be detected clearly appearing long-drawnout in a wide temperature range. The two metal oxide phases (precursors of the host and guest metals, Ni and In) being present together in the highly mesoporous fumed silica give an overlapped, merged reduction profile. This is not a simple summation, but the result of a bilateral action, where reduction of NiO is noticeably slowed down, while reduction of In 2 O 3 is picked up speed, seems to be significantly catalysed by nickel formed prior to that. The same precursors reduced in contact with γ-Al 2 O 3 support give similar pictures, but shifted to higher temperatures. These different reduction processes on silica and alumina supports should produce active metallic species in some degree different from catalytic point of view. The HT-XRD patterns given in Fig. 4-6 represent also the formation of silica supported active metal particles. The upper temperatures in Fig. 4 are those at which a complete reduction was observed and there may be a correlation with the standard reduction potentials of the investigated metals. The step by step reduction of added oxides of host (Cu, Ni, Co, Fe) and guest (In) metals to the active hydrogenating mono-or bimetallic phases is demonstrated in pure forms (Fig. 4-5) and in those mixtures (Fig. 6). The reductive treatment of the 9Cu/ or 9Ni/ and 9Co/SiO 2 (1) catalyst precursors generates small Cu, Ni or Co metal particles (Fig. 4) completely reducing the CuO, NiO or Co 3 O 4 phases up to 450°C which is the routine pre-treatment temperature before catalytic measurements. CuO is already completely reduced up to 200 o C and in this case there is no any difference using alumina or silica supports. The average size of Cu particles (calculated by the Scherrer equation) is 20 nm formed on alumina. Significantly higher value (35 nm) was obtained on SiO 2 (1). The higher porosity of silica means smaller diffusional resistance, which can result in aggregation of copper atoms into somewhat bigger particles in the more porous fumed silica. The average size of metal particles formed at higher temperature than copper was found around 20 nm. Iron-oxide seemed to be less reducible. This course is completed only up to 550°C.  Contrary to dynamical H 2 -TPR measurement, in the HT-XRD cell In 2 O 3 heated up step by step in H 2 flow the reduction is almost completed already at 450°C (see Fig. 5). At 550°C In 2 O 3 is fully reduced to liquid In 0 (melting point is 156.6°C) and cooling down to room temperature the presence solid indium metal particles can be detected. So complete formation of bimetallic phases (e.g. Ni 2 In or Cu 2 In ) can be controlled by disappearance of indium diffraction lines from the HT-XRD patterns at R.T.
Formation of bimetallic catalyst can be demonstrated in Fig. 6. All admixed In 2 O 3 seem to be converted to indium atoms around 450°C in hydrogen flow indicated by disappearance of its diffraction lines from the HT-XRD patterns and that can form immediately new Ni 2 In (or Cu 2 In [9-10]) bimetallic phases. Copper particles are ready for alloying already at 200°C and can catalyse the reduction of the guest metal indium which can immediately form the bimetallic particles around 350°C. In silica support NiO can be reduced at a shade lower temperature (350-400°C) than In 2 O 3 being also ready for alloying with nascent indium atoms.
The formed indium (as metal phase can be detectable by XRD only below 156.6 °C, the melting temperature of indium metal) can be consumed in the reaction attaining the Ni 2 In stoichiometry with the host metal. Excess indium cannot be detected (not shown) excluding the formation of intermetallic compounds with higher Ni-content /Ni 4 In, Ni 3 In/, furthermore applying more amount of In 2 O 3 NiIn or Ni 2 In 3 were also not formed. In the case of Cu or Ni host metals, formation of Cu 2 In and Ni 2 In phases were clearly detected in numerous studies and the selective catalytic hydroconversion of carboxylic acids to alcohols suggest that these phases are unequivocally responsible for the favourable catalytic effects. Summarized, Cu and Ni amalgamation with indium only to Cu 2 In and Ni 2 In intermetallic compounds can be attained among various opportunities. Contrary to Cu and Ni, Co and Fe host metals in presence of indium liquid phase are not able for simple solid state amalgamation, for formation of intermetallic compounds in full extension of the metal particles under the investigated conditions. CoIn and FeIn intermetallic compounds cannot be detected by XRD technique and the unused indium phase can be observed cooling down to room temperature. Naturally, alloying of the upper layer can be possible, which can be proven catalytically. The product distributions characteristic for silica supported monometallic catalysts are compared in Fig. 7. The diverse catalytic features of the samples need different reaction temperature ranges to attain high level of the acetic acid hydroconversion under otherwise identical experimental conditions. Strikingly different selectivities are attached with the different activities. One of the dominant reaction routes is the step-bystep reduction of acetic acid through acetaldehyde to ethyl alcohol producing water as by-product.
The intermediate compound acetaldehyde can be detected over all catalyst samples confirmed this reaction route. On monometallic Cu/SiO 2 (1) and Co/SiO 2 (1) catalysts formed almost selectively ethanol. Each of the studied metals (Co, Ni, Cu, Pt, Fe and Zn) can catalyse the hydrogen attack to C-O bonds reducing carboxylic acids however on Ni, Pt, Fe and Zn AA hydrodecarbonylation to CH 4 , CO and H 2 O seems to be significant. Activity of Fe/SiO 2 (1) and Zn/SiO 2 (1) is very low compared to other studied samples. Furthermore platinum is one of the most expensive metals. Among the tested six metals only Co, Cu and Ni seems to be promising. The ethanol selective cobalt and copper species formed on fumed silica are found also active in hydrodecarbonylation resulting in the formation undesired by-products: methane, carbon monoxide and water, but only at higher reaction temperature than a complete AA reduction to ethanol has been attained. Over the monometallic Co and Cu catalysts the extension of rival competing, parallel reaction routes can simply be controlled by the reaction condition. The highly active nickel catalyst prefers the AA hydrodecarbonylation in the investigated wide reaction temperature range. However, this unwanted reaction can be completely eliminated with modification of the active metallic surface (see Fig 8B) [12,26].
The indium doping of nickel changed the monometallic particles to bimetallic ones, the active surface results in higher yield of acetaldehyde as intermediate compound and ethanol as main product. Indium admission can eliminate completely the hydrodecarbonylation reaction similarly to all former findings and the activity is still improved given so efficient catalyst which can be exceeded only by use of nickel as main, host metal which was proved most active also in our former studies. New findings, that this high activity and selectivity can be attained with a more simpler monometallic catalysts, Co/SiO 2 (1) (compare Fig. 7b and Fig. 8B). What is more the monometallic Co/SiO 2 (1) exceeds the properties of bimetallic Ni 2 In/ SiO 2 (1) sample at lower reaction temperature. Interestingly, indium admission to cobalt catalysts results in lower activity (see CoIn/SiO 2 (1) in Fig. 8A). Important observation is that in the case of Co-catalysts after indium doping the complete improvement of selectivity is attached with a significant decrease of the activity, i.e. full conversion can be attained only at higher temperature. One of the possible explanations (simple geometric effect) is that cobalt atoms on the surface are much more diluted with indium because cobalt can form only CoIn 2 or CoIn 3 phases as intermetallic compounds contrary to existence of Cu 2 In or Ni 2 In.
On fumed silica supported Ni and Co catalysts indium doping shows inverse effect regarding the change of catalytic activity (compare Fig. 8A and B). Adding indium guest metal to Cu/ SiO 2 (1) the activity was hardly influenced and the effect for the selectivity was also slight. It can be concluded that selective hydroconversion of acetic acid to ethanol can be achieved over different catalysts. Fig. 9 reflects that the texture of the supports can be very important. It is still easy to accept that microporous silicagel cannot be suited for catalyst support because of its extremely high diffusional resistance. However, the great differences observed using different mesoporous silicas manufactured with different methods seem to be quite strange.
SiO 2 (3) having narrower pores can be characterized with significantly higher specific surface area than SiO 2 (1). At an early stage both show similar high activity however over Co/ SiO 2 (3) a fast activity decrease can be observed. Here also a mass transport problem can be supposed to explain the fast activity loss at increasing time-on-stream (see Fig. 9B). Important finding that different metal catalysts prepared with fumed silica, SiO 2 (1) were proven stable under reaction conditions resulting in high AA conversion.

Conclusions
In former studies highly efficient Ni 2 In bimetallic catalysts were found to be outstandingly efficient for selective reduction of aliphatic carboxylic acids to alcohols. This study reveals that although the role of the bimetallic phases is decisive however the properties of the supports seem to be also quite significant.
Highly mesoporous fumed silica was proven to be advantageous of overriding. Nature of silica surface excludes side reactions (e.g. mono-or bimolecular dehydration) which are characteristic for alumina. Beside the surface properties of the silica supports their pore structure, morphological feature was turned out an important factor. Loading fumed silica support with the less active hydrogenating metal, copper the properties of the more active nickel catalysts can be explicitly approached. Using cobalt on this special silica support a highly active and selective catalyst can be obtained which can replace expensive precious metals. The variety of the most suitable catalysts for hydroconversion of bioacids is further expanded and additional improvement can be attained by optimization of the support and the catalyst preparation technique.

Acknowledgement
Just before my retirement, many thanks are due to my alma mater, Budapest Technical University for supporting my walk of life with fundamental knowledge.