Continuous Synthesis of γ-Valerolactone in a Trickle-Bed Reactor over

Feb 14, 2017 - Various Ni-based catalysts were tested in the continuous liquid phase hydrogenation of levulinic acid (LA) to γ-valerolactone (GVL) in...
0 downloads 0 Views 1MB Size
Download PDF
Subscriber access provided by University of Newcastle, Australia

Article

Continuous Synthesis of #–Valerolactone in a Trickle-bed Reactor over Supported Nickel Catalysts Konstantin Hengst, D. A. J. Michel Ligthart, Dmitry E. Doronkin, Karin M. Walter, Wolfgang Kleist, Emiel J. M. Hensen, and Jan-Dierk Grunwaldt Ind. Eng. Chem. Res., Just Accepted Manuscript • DOI: 10.1021/acs.iecr.6b03493 • Publication Date (Web): 14 Feb 2017 Downloaded from http://pubs.acs.org on February 15, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Industrial & Engineering Chemistry Research is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

Continuous Synthesis of γ– γ–Valerolactone in a TrickleTrickle-bed Reactor over Supported Nickel Catalysts Konstantin Hengsta,b, D. A. J. Michel Ligthartc, Dmitry E. Doronkina,b, Karin M. Waltera,b, Wolfgang Kleista,b, Emiel J. M. Hensenc, Jan-Dierk Grunwaldt*,a,b a

Institute for Chemical Technology and Polymer Chemistry, Karlsruhe Institute of Technology, Engesserstr. 20, D76131 Karlsruhe, Germany. E-mail: [email protected]; Fax: +49 7211 608 44805; Tel: +49 7211 608 42120

b

Institute of Catalysis Research and Technology, Karlsruhe Institute of Technology, Hermann-von-Helmholtz Platz 1, D-76344 Eggenstein-Leopoldshafen, Germany.

c

Inorganic Materials Chemistry, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, Netherlands. KEYWORDS: γ-Valerolactone; Nickel Catalysts, Levulinic Acid, Continuous Hydrogenation, Biobased Intermediates ABSTRACT: Various Ni-based catalysts were tested in the continuous liquid phase hydrogenation of levulinic acid (LA) to γ-valerolactone (GVL) in a trickle-bed reactor using water as solvent with the aim to develop an economic and environmentally friendly way to GVL synthesis. For this purpose, various synthesis methods were used to prepare Ni-based catalysts, which were first screened in batch reactors. Characterization by X-ray diffraction, temperature-programmed reduction, electron microscopy, hydrogen chemisorption and X-ray absorption spectroscopy showed that slow precipitation using urea resulted in a good Ni dispersion. The dispersion also improved at lower Ni loading and smaller Ni particles mostly showed an enhanced catalytic performance for the synthesis of GVL. 5 wt.% Ni/Al2O3 prepared by wet impregnation showed the highest specific activity for the hydrogenation of LA to GVL (90 % LA conversion and 75 % GVL yields) featuring an average Ni particle size of 6 nm. Some deactivation of the catalysts was observed, probably due to transformation of γ-Al2O3 to boehmite and sintering of the Ni particles. In addition, re-oxidation of Ni particles may additionally lead to deactivation as concluded by comparison with screening studies in batch reactors.

1 Introduction γ-Valerolactone (GVL) is a versatile platform molecule, which can be directly used as a fragrance, green solvent, monomer for the production of plastics or as a gasoline blending component. In addition, several bio-based fuels or fuel additives (2-methyltetrahydrofuran, aromatics and alkenes of different chain lengths) can be produced from GVL.1-6 Two possible reaction pathways exist for the synthesis of GVL from levulinic acid (LA).7 In the presence of acid functionalities, LA can be first dehydrated to angelica lactone, which is subsequently hydrogenated to GVL over transition metal catalysts. Alternatively, the keto group of LA can be hydrogenated to γ–hydroxyvaleric acid, which spontaneously condensates to GVL.7 Different transition metal based catalysts, both homogeneous and heterogeneous, have been developed for the hydrogenation of LA to GVL.8, 9 Raney-nickel catalysts were used for the GVL synthesis by Sabatier and Mailhe 10 and Christian et al. 11. Since 2000, various supported Ru, Pd and Pt catalysts have been reported for the hydrogenation of LA to GVL.7, 12-17 In most studies, reaction temperatures between 70 and 250 °C, hydrogen pressures up to 55 bar and various solvents (alcohols, water, ethers) were used in batch operation mode. Ru-based catalysts showed

the highest catalytic activity leading to quantitative LA conversion under mild reaction conditions.9 The main disadvantage of noble metal catalysts concerns their high cost and, therefore, the development of nonnoble metal based catalysts is desirable. Recently, Ni- and Cu-based catalysts have been studied for LA hydrogenation to GVL.18-25 Bimetallic Ni-MoOx/C 19 and Ni-Cu/Al2O3 23 tested at T = 250 °C and PH2 = 50 - 60 bar resulted in GVL yields (YGVL) of >90 %. Shimizu et al.19 tested Ni on different supports for LA hydrogenation at much milder conditions (T = 140 °C and PH2 = 8 bar) and found a maximum LA conversion (XLA) of 38 % with a 100% selectivity to GVL. Our group recently reported the synthesis of GVL over Ni/Al2O3: Different solvents, reaction temperatures (T) and H2 pressures (PH2) were screened; XLA = 100 % and YGVL =92 % were achieved at T = 200 °C and PH2 = 50 bar under solvent free conditions.26 Typically, for non-noble metal catalysts higher reaction temperatures (T >200 °C) are necessary to obtain similar conversion levels (XLA >80 %) as for noble metal catalysts. With a view on industrial applications of biobased LA, the continuous production of GVL has many advantages compared to batch processes. Up to now, only a few studies focused on the continuous hydrogenation of LA to GVL.12,

ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 12

21, 22, 24, 25, 27

The group of Poliakoff tested the reaction at T = 200 °C and P = 100 bar in supercritical CO2 over 5 wt.% Ru/SiO2 and achieved quantitative conversion of LA to GVL.27 Tukacs et al. tested Ru/C, Pd/C and Raney-Ni at T = 100 °C and PH2 = 100 bar. The addition of a phosphine ligand resulted in a positive effect on XLA (from 83 to 98 %) for Ru/C.24 The groups of Chang 12 (Ru/C, Pd/C, Pt/C), Chary 25 (Cu/Al2O3) and Rao 21, 22 (Ni on different supports) investigated the vapour phase hydrocyclization of LA to GVL at T >250 °C. Ru/C emerged as the most active and stable catalyst, but the sustainability of these studies was limited due to the use of 1,4-dioxane as solvent. Lower reaction temperatures in combination with cheap catalysts and water as a green solvent would in principle provide an economic and environmental friendly process for GVL synthesis from LA.

the sample was filtered off, washed, dried at 110 °C for 1 h and calcined at 600 °C for 5 h. A Ni+Pt/Al2O3_urea catalyst was prepared similarly by adding 0.0744 g of tetraammine platinum(II) nitrate to the Ni(II) nitrate solution. For 15Ni/Al2O3_fsp, 2.975 g Ni(NO3)2·6H2O and 24.995 g aluminum nitrate were dissolved in 120 ml methanol. This mixture was pumped at a 5 mL/min flow rate into the nozzle and dispersed with an O2 flow (5 L/min) in a 10 % CH4/O2 flame.28 The sample was collected on a filter 30 cm above the flame and calcined at 600 °C for 2 h. Note that particle production and collection on the filter must be performed in a well-closed fume hood under appropriate safety measures. All catalysts were pre-reduced in a 10 % H2/N2 flow (20 L/h) at 600 °C for 2 h (5 °C/min) in a tubular furnace before catalytic tests in the batch mode or the continuous mode (trickle-bed reactor).

In this work, we report on the continuous liquid phase hydrogenation of LA to GVL using supported non-noble metal catalysts. We investigated the liquid phase hydrogenation in a trickle-bed reactor over supported Ni nanoparticles using water as solvent, which was found to be an attractive reaction medium under batch operation in a previous study. For this purpose, first a set of Ni based catalysts was synthesised using different preparation methods, characterized with respect to Ni particle size and prescreened in a batch reactor. Selected catalysts were then investigated under continuous flow conditions to study their time-on-stream behavior and possible deactivation.

2.2 Catalyst characterization

2 Experimental section 2.1 Catalyst preparation A set of Ni-based catalysts was prepared using wet impregnation (wi), precipitation with urea (urea), precipitation with NaOH (p) and flame spray pyrolysis (FSP). The catalysts are denoted xNi/Al2O3_m with x being the metal loading and m the synthesis method. Nickel(II) nitrate (Ni(NO3)2·6H2O, Sigma-Aldrich) and tetraammineplatinum(II) nitrate (Sigma-Aldrich) were used as metal precursors. γ-Al2O3 (Carl Roth, used as-received) and Siral70 (amorphous silica-alumina with 70 wt. % silica, Sasol) were used as support materials and aluminum nitrate (Sigma-Aldrich) served as a precursor for FSP synthesis. Typically, 15Ni/Al2O3_wi was synthesized by suspending 8.5 g γ-Al2O3 in a 30 mL aqueous solution containing 7.427 g of Ni(NO3)2·6H2O and removal of the excess of water in a rotary evaporator. The impregnated sample was calcined at 600 °C for 5 h after drying at 110 °C for 1 h. 15Ni/Al2O3_p was prepared by suspending 8.5 g γ-Al2O3 in an aqueous solution containing 7.427 g Ni(NO3)2·6H2O (100 ml, stirred for 1 h). Subsequently, 465 mL 0.1 M NaOH solution were added at room temperature until a pH of 9 was reached. For 15Ni/Al2O3_urea, 4.25 g γ-Al2O3 were suspended in a 30 mL aqueous solution containing 3.714 g Ni(NO3)2·6H2O and a 50 ml aqueous solution with 4.93 g urea (molar ratio urea/precursor = 6.3). The mixture was diluted to a total volume of 550 mL, the pH was adjusted to 2 by nitric acid, and thereafter the mixture was gradually heated to 90 °C for 18 h. After precipitation,

Inductively coupled plasma optical emission spectrometry: The Ni content of the catalysts was validated by inductively coupled plasma optical emission spectrometry (ICP-OES) using an Agilent 720/725-ES spectrometer. The catalysts were dissolved in 5 M H2SO4 solution in a microwave at 600 W for 2 h, afterwards diluted with distilled water and finally injected into the plasma. Nitrogen physisorption and temperature-programmed reduction (TPR): The surface area of the catalysts was determined by N2 physisorption (Belsorp II mini, BEL Japan Inc.) at -196 °C using 10 points in the range of P/P0 = 0.05 – 0.3. All samples were outgassed at 300 °C for 2 h prior to physisorption. The reduction behavior of the catalysts was investigated by TPR using a ChemBET TPR/TPD analyzer (Quantachrome). Calcined catalysts (100 mg) were loaded into a quartz reactor, placed into the tube furnace and heated to 900 °C (5 °C/min) in a 5 % H2/Ar mixture (73 mL/min). The H2 consumption was measured with a thermal conductivity detector (TCD) as function of the temperature. X-ray diffraction (XRD): Powder XRD patterns were collected with a PANalytical X`Pert PRO diffractometer with Cu Kα radiation (Cu Kα1 = 1.54060 Å and Cu Kα2 = 1.54443 Å and Ni filter). The scan was recorded in a 2 θ range of 20 – 80° (0.017° step width, 0.51 s data acquisition per point). Crystalline phases were determined using the Cambridge Structural Database (CSD) of the Cambridge Crystallographic Data Center (CCDC). The Scherrer equation was used to estimate the mean particle diameter (average of all Ni reflexes) and LaB6 was used as standard to correct for the instrumental line broadening. Chemisorption: H2 chemisorption was carried out at 40°C (Micromeritics ASAP 2020C) assuming an adsorption stoichiometry of one H per surface Ni .29 For this purpose the double isotherm method was employed with an intermediate vacuum treatment of 1 h (sample reduced at 600 °C at 5 °C/min for 2 h, evacuated for 4.5 h). Scanning transmission electron microscopy (STEM): A powdered sample was ultrasonically dispersed in ethanol and one drop of the suspension was dried on a gold grid covered with holey carbon film. The catalyst speci-

ACS Paragon Plus Environment

Page 3 of 12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

mens were examined in a FEI Titan 80-300 aberration corrected electron microscope operated at 300 kV. STEM images were acquired by a Fischione model 3000 HAADF STEM detector and energy dispersive X-ray spectra (EDX) were acquired by an EDAX SUTW EDX detector. X-ray absorption spectroscopy (XAS): The local structure of Ni was characterized using XAS at the Ni K edge (8333 eV). X-ray absorption spectra (X-ray absorption near edge spectra, XANES, and extended X-ray absorption fine structure, EXAFS) were recorded at the XAS beamline of ANKA (Karlsruhe, Germany). The catalyst (sieve fraction: 100-200 μm, ca. 5 mg) was placed in an in situ microreactor (quartz, 1 mm diameter, 20 μm wall thickness) heated by a hot air blower (Gas Blower GSB-1300, FMB Oxford). 30 The spectra were energy-calibrated, background subtracted and normalized using the ATHENA program of the IFFEFIT package 31. During TPR-XANES measurements, the catalyst was heated in a 5 % H2/He flow (50 mL/min) to 600 °C (5 °C/min). Linear combination analysis (LCA) of the TPR-XANES spectra was performed using Ni foil, NiO (Fluka, purum), and NiAl2O4 spinel reference spectra in the range 8325 - 8360 eV. 2.3 Catalyst testing Custom-made batch autoclaves (Tmax = 350 °C, Pmax = 200 bar, V = 65 mL) and a continuous trickle-bed reactor (Tmax = 700 °C, Pmax = 100 bar) were used for the catalytic tests. For catalytic screening in batch mode, the reactor was charged with 10 mL of the reaction mixture (10 wt. % LA in H2O) and 0.1 g of the pre-reduced and air-stored Ni catalyst (fine powder, < 50 µm). The reactor was purged with N2 and pressurized with 50 bar of H2. To allow comparison among the different catalysts, the same reaction conditions were used for all catalysts. Note that the simple design of the screening reactors is not suited for kinetic studies. The magnetically stirred autoclave was heated with heating sleeve and plate. The starting point of the reaction was defined as the time, when the desired temperature was reached (usually after 20 to 30 min). After 4 h, the reactor was quenched in ice water, depressurized, flushed with nitrogen and, finally, the product was separated from the catalyst by filtration. The catalytic activity

measurements in the trickle-bed reactor (TU Eindhoven) were performed as follows. The reactor was a vertically placed stainless-steel tube with an inner diameter of 4 mm and a length of 200 mm, which was heated by a tube furnace. The liquid feed was administered in downward flow mode by a high pressure piston pump (Hewlett Packard series 1050). Gases were delivered by thermal mass flow controllers (Bronkhorst). Usually, 0.5 g of catalyst (60 - 100 µm sieve fraction, diluted with SiC of same sieve fraction to improve plug flow conditions and isothermicity) were loaded into the reactor (1 ml, ca. 10 cm bed length, between two glass wool plugs). The performance of the trickle-bed reactor for the applied reaction conditions was verified according to criteria reviewed by Mary et al.32 and Hickman et al.33 The criteria were only fulfilled for the smallest size revealing that a compromise between particle size and mass transport had to be taken. A further decrease in the catalyst particle size was not possible due to blockage of the reactor. Although SiC has been used as dilutent to improve isothermicity of the reactor, it is limited since considerable wall effects are present. Potential hot spots could influence the reactions performed in the trickle-bed reactor and lead to deviations in the behavior compared to batch reactions. Further details on the reaction engineering aspects including the calculations can be found in the ESI. Prior to reaction, the set-up was flushed with N2 followed by reduction of the catalyst at 600 °C (10 °C/min) in a 250 mL/min H2 flow for 1 h. After cooling the reactor to the desired reaction temperature, the gas flow was changed to a 50 mL/min H2 flow. The reaction pressure was set using a back pressure controller (Tescom). The liquid flow consisting of 10 wt.% levulinic acid in water was set at 5 g/h. The typical reaction conditions were T = 200 °C and PH2 = 50 bar. Effluent products were collected every 60 min for analysis. The weight hourly space velocity (WHSV) based on LA feed rate  =



 

was varied by adjusting the amount of catalyst.

Table 1: Basic characterization (elemental analysis, BET surface area and maxima of reduction during TPR) of the Ni-based catalysts. BET Surface Area [m2/g]b

Reduction Peak

after Reactiona [wt.%]

4.7

4.8

108

600

15Ni/Al2O3_wi

14.0

13.4

96

380 and 700

15Ni/Al2O3_p

14.0

12.9

122

380

15Ni/Al2O3_urea

11.1

9.7

127

380c and 700

15Ni/Al2O3_fsp

13.4

n.d.

31

800

15Ni+0.75Pt/Al2O3_urea

14.5 (0.05 Pt)

13.2

115

700d

15Ni/Siral70_wi

14.1

n.d.

242

Catalyst

Ni [wt.%]

5Ni/Al2O3_wi

Elemental Analysis

Maximum [°C]

650 o

a = continuous hydrogenation of LA in water (4 h TOS); b = after reduction at 600 C for 2 h; c = main peak at 380 °C; d = starts at 450 °C; n.d. = not determined

ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The turnover frequency (TOF) was calculated from mol LA (nLA), percent LA conversion (XLA), mol Ni (nNi) and dispersion (D) obtained by hydrogen chemisorption: TOF =

  ∙  



∙

.

Page 4 of 12

cles and a fairly homogeneous particle size distribution as a result of the low Ni content or Pt. The high reduction temperature of 15Ni/Al2O3_fsp may originate from NiAl2O4 spinel or NiO, embedded in alumina that might both be formed during flame-spray pyrolysis.36, 37 Therefore, this catalyst was only tested in batch operation.

2.4 Product analysis Product mixtures were diluted with demineralized water and then analyzed by high performance liquid chromatography (HPLC, Merck-Hitachi) containing a BioRad organic acid column (Aminex HPX 87H), a refraction index detector and a UV detector (wavelength: 254 nm). A 0.004 M H2SO4 solution was used as the mobile phase with a 0.5 mL/min flowrate. The column was operated at 50 °C and 50 bar. Product concentrations were calculated using calibration curves achieved with solutions of known concentrations. In case of GVL selectivity 15Ni/Al2O3_wi > 15Ni/Al2O3_urea, as supported by XANES and EXAFS spectra after in situ TPR (cf. Fig. S3). This also supports the hypothesis that some of the chemisorption results underestimate the dispersion.

Fig. 3: XANES spectra at the Ni K edge of (i) calcined 5Ni/Al2O3_wi (b), 15Ni/Al2O3_wi (c) and 15Ni/Al2O3_urea (d) catalysts in comparison with NiO (e) and NiAl2O4 (a) reference spectra and (ii) of 5Ni/Al2O3_wi (a), 15 Ni/Al2O3_wi (b) and 15Ni/Al2O3_urea (c) catalysts after in situ TPR in comparison with Ni reference spectrum (d).

Fig. 5: Linear combination analysis (LCA) results from in situ TPR-XANES spectra at the Ni K edge of 5Ni/Al2O3_wi (a), 15Ni/Al2O3_wi (b) and 15Ni/Al2O3_urea (c) using the reference spectra of bulk NiO, NiAl2O4 and Ni metal. TPR conditions: 5% H2/He flow (5 °C/min).

3.2 Catalyst screening in batch mode In the first series of experiments, the catalysts were tested at T = 200 °C and PH2 = 50 bar in batch autoclaves using water as a solvent. The results are displayed in Fig. 6. All catalysts were reduced and stored under air atmosphere at room temperature before reaction. 2

Fig. 4: (i) k -weighted EXAFS spectra of calcined 5Ni/Al2O3_wi, 15Ni/Al2O3_wi and 15Ni/Al2O3_urea catalysts in comparison with NiO and NiAl2O4 reference spectra; (ii) Fourier transformed EXAFS spectra of calcined 5Ni/Al2O3_wi, 15Ni/Al2O3_wi and 15Ni/Al2O3_urea catalysts in comparison with NiO and NiAl2O4 reference spectra.

Next, complementary to the TPR data (section 3.1.1.), the reduction behavior was studied by XANES for three selected catalysts. The results of the LCA of the TPR-XANES spectra are shown in Fig. 5 (for XANES data, cf. Fig. S2). For both Ni/Al2O3_wi samples, the NiO fraction was not

Most of the Ni catalysts converted LA to GVL with a selectivity of 100%. As an exception, 15Ni/Al2O3_fsp showed a maximum yield of GVL 82 % and 4-hydroxyvaleric acid as a minor by-product at complete LA conversion, whereas 15Ni/Al2O3_p was able to convert only 9% of LA with a GVL yield of nearly 3 %. 15Ni/Al2O3_wi converted 57% of LA with 100% selectivity to GVL. Unfortunately, LA conversions of 15Ni/Al2O3_urea and 15Ni+0.75Pt/Al2O3_urea were much lower, but resulted in 100% selectivity to GVL. This lower catalytic activity but high selectivity might have been a consequence of the fact that these catalysts contained oxidized Ni particles (small particles that were more prone to oxidation than larger ones). Therefore, an

6

ACS Paragon Plus Environment

Page 7 of 12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

influence of the average Ni particle size on the catalyst performance could not be deduced in these batch experiments. The dependence on the structure of the Ni particles is more complex, since the conversion of LA over 5Ni/Al2O3_wi was lower than for 15Ni/Al2O3_urea and the latter one contained some larger particles. Furthermore, we tested 15Ni/Siral70_wi to study the effect of support acidity. Siral70 contains Brønsted acid sides and is therefore more acidic than γ-Al2O3. This might possibly open up other reaction pathways via α- or β-angelica lactones as intermediate products.7 As Fig. 5 shows, 15Ni/Siral70_wi performed similar to 15Ni/Al2O3_urea or 15Ni+0.75Pt/Al2O3_urea that featured a comparable average Ni particle size. Thus, we did not further investigate the performance of this catalyst in this study.

towards γ-valerolactone was lower at higher T because of the additional formation of 4-hydroxyvaleric acid. From 180 °C to 200 °C, XLA and YGVL increased almost by a factor of two, whereas a further temperature raise to 220 °C only led to an increase of 15% in XLA and YGVL. A strong increase in catalytic performance at >180 °C has also been reported earlier by our group for LA hydrogenation in batch mode using various Ni-based catalysts.26 For other non-noble metal catalysts, temperatures higher than 200 °C were required to reach similar levels of XLA as for noble metal catalysts (>80 %).19, 23 For further reactions, we used a reaction temperature of T = 200 °C to elucidate the impact of other reaction parameters.

Fig. 6: Conversion of levulinic acid XLA (black bars) and yield of γ-valerolactone YGVL (grey bars) after hydrogenation of levulinic acid in a batch autoclave over various catalysts. Reaction conditions: T= 200 °C, pH2 = 50 bar, TOS = 4 h, nNi/nLA = 0.03, mcat = 0.1 g. 3.3 Catalytic tests in continuous mode 3.3.1 Influence of temperature and hydrogen pressure on the conversion of levulinic acid over 15Ni/Al2O3_wi After catalyst screening in batch mode, five of the catalysts were tested in the continuous LA hydrogenation using a trickle-bed reactor. The catalysts were reduced in situ prior to the experiment and not exposed to air afterwards (cf. section 2.3). In a first step, the reaction temperature and the hydrogen pressure PH2 were optimized in continuous mode over 15Ni/Al2O3_wi based on the good performance in the screening tests. The effect of the temperature on the conversion XLA and yield of γvalerolactone (YGVL) for this catalyst is depicted in Fig. 7. With increasing reaction temperature, both XLA (up to 100 %) and YGVL increased (up to 89 % at 220 °C and 1 h on stream (TOS)). In this temperature range, XLA and YGVL were rather constant for at least 2 h on stream before they gradually decreased. At 180 °C, the selectivity towards γvalerolactone SGVL was nearly 100 % whereas it was slightly lower (between 80 % and 90 %) and constant for at least 4 h on stream at 200 °C and 220 °C. The selectivity

Fig. 7: XLA (closed symbols) and YGVL (open symbols) during continuous hydrogenation of levulinic acid in a trickle bed reactor over 15Ni/Al2O3_wi at different reaction temperatures -1 (T). Reaction conditions: pH2 = 50 bar, WHSV = 1 h , mcat= 0.5 g, LA/H2O = 5 g/h, H2 = 50 mL/min.

In the next step, the hydrogen pressure was varied for 15Ni/Al2O3_wi. At 50 bar (cf. Fig. 7), the conversion of levulinic acid was similar to 30 bar for 3 h on stream as depicted in (Fig. S5). After 4 h on stream, the catalyst deactivated to some extent. This effect was strongly influenced by PH2 as the catalyst performance was more stable at higher PH2. The selectivity towards γ-valerolactone was also higher at higher PH2 (Fig. S5). Note that it cannot be excluded that some mass transport limitations occurred, although the smallest sieve fraction and SiC as dilutent were used (cf. ESI). Overall, we established that T = 200 °C and PH2 = 50 bar were the optimal reaction conditions for further catalyst tests in continuous mode.

7

ACS Paragon Plus Environment

Industrial & Engineering Chemistry Research

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

3.3.2 Effect of the average Ni particle size on continuous hydrogenation of levulinic acid Fig. 8 shows XLA and YGVL as a function of time on stream for the different Ni catalysts. 15Ni/Al2O3_p with the largest Ni particles (65 nm) showed the lowest XLA and YGVL (Fig. 8c). However, this catalyst exhibited the highest GVL selectivity (100 %) of all investigated catalysts. A strong deactivation was found: LA conversion decreased steadily from 47 % after 1 h to 9 % after 4 h TOS. The catalysts 5Ni/Al2O3_wi, 15Ni/Al2O3_wi and 15Ni/Al2O3_urea showed a similar catalytic activity for LA hydrogenation (Fig. 8 a, b, d). GVL selectivities were always between 80 and 95 % (not 100%, probably due to the high conversion, also observed in batch reactors, cf. Ref.26) and the corresponding LA conversions and GVL yields were constant (or slightly increasing) during the first 2 h TOS. After 3 h and 4 h, TOS both the LA conversion and GVL yields decreased. The drop in catalytic activity was more pronounced for 15Ni/Al2O3_wi (featuring the largest Ni particle size among those three catalysts) compared to 5Ni/Al2O3_wi and 15Ni/Al2O3_urea. The average Ni particle sizes of 5Ni/Al2O3_wi and 15Ni/Al2O3_urea were similar, but also some larger Ni particles were present in 15Ni/Al2O3_urea due to the higher loading (cf. Table 2 and Fig. 2). The same amount of catalyst was used for LA hydrogenation. The catalytic activity was nearly stable over 4 h TOS using 5Ni/Al2O3_wi. Strikingly, the 15Ni+0.75Pt/Al2O3_urea catalyst showed the highest LA conversion (100 % for 4 h TOS) of all catalysts and no visible catalyst deactivation probably due to its high conversion. The GVL yields were low (43 %) after 1 h TOS but increased to the level of catalysts 5Ni/Al2O3_wi, 15Ni/Al2O3_wi and 15Ni/Al2O3_urea after 2 h TOS and remained constant afterwards. According to these results, smaller Ni particles showed higher catalytic activity compared to larger ones. 15Ni/Al2O3_wi and 15Ni/Al2O3_urea showed good catalytic performance compared to 15Ni/Al2O3_p either due to their higher Ni surface area or because of the higher reactivity of low-coordinated Ni sites.38 However, they were less active than 15Ni+0.75Pt/Al2O3_urea, which featured higher dispersion according to chemisorption results (cf. Table 2). Note, however, that noble metals often give a higher catalytic activity, although with lower selectivity.12, 24 Catalysts 15Ni/Al2O3_p, 15Ni/Al2O3_wi and 15Ni/Al2O3_urea were significantly deactivated during the reaction. As leaching was hardly observed by elemental analysis after reaction, both sintering or phase transformations may be responsible for this observation.39 For example, the Ni content of 15Ni/Al2O3 catalysts after 4 h on stream in continuous mode was just ~1 wt.% lower in comparison to their fresh counterparts. This amount of leached Ni did not reflect the origin of deactivation. Furthermore, no leaching was detected for 5Ni/Al2O3_wi. This may be related to the stronger interaction between small Ni particles (cf. Table 2 and Fig. 8) and γ-Al2O3 (Fig. S1). The possibility of sintering was investigated by re-

Page 8 of 12

cording XRD patterns of the catalysts after the continuous reaction in the trickle bed reactor (see Fig. 1). Significant changes were visible compared to the freshly reduced catalysts. Reflections corresponding to γ-Al2O3, boehmite and Ni were clearly visible (Fig. 1 (ii)). The intensities of the latter phases increased for several catalysts. This indicated that larger Ni particles were present and, hence, sintering occurred (Table 2). Sintering could be triggered by the transformation of γ-Al2O3 to boehmite which weakens the metal-support interactions as also reported by Li et al.. 39

Fig. 8: XLA (closed symbols) and YGVL (open symbols) during continuous hydrogenation of levulinic acid in a trickle bed reactor over Ni catalysts: 5Ni/Al2O3_wi (a), 15Ni/Al2O3_wi (b), 15Ni/Al2O3_p (c), 15Ni/Al2O3_urea (d) and 15Ni+0.75 Pt/Al2O3_urea (e). Reaction conditions: T = 200 °C, PH2 = 50 -1 bar, WHSV = 1 h , mcat = 0.5 g , LA/H2O = 5 g/h, H2 = 50 mL/min.

Considering the low Ni content of 5Ni/Al2O3_wi, this catalyst showed the best performance among the undoped Ni-based catalysts for the hydrogenation of LA. In order to compare the reaction rate in terms of turnover frequency (TOF), the WHSV was increased using 5Ni/Al2O3_wi and the TOF was calculated at low LA con-

8

ACS Paragon Plus Environment

Page 9 of 12

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Industrial & Engineering Chemistry Research

versions (WHSV: 1.75 and 2.5 h-1). After 1 h on stream, XLA = 24 % and XLA = 42 % were found at WHSV values of 2.5 h-1 and 1.75 h-1, corresponding to TOF values of 1.9·10-3 s-1 and 2.4·10-3 s-1, respectively (overall, TOF ~2·10-3 s-1). Notably, catalyst deactivation was much more pronounced at higher WHSV. A comparison of these TOFs to literature values revealed that the performance of 5Ni/Al2O3_wi was similar to values reported for Cu/Al2O3 25 (5·10-3 s-1 at 265 °C), but one order of magnitude lower than for various Ni/H-ZSM-5 catalysts (5·10-2 s-1 with lower Ni loading and 15·10-2 s-1 at 250 °C, respectively).22 3.3.3 Catalyst stability and comparison with literature To further evaluate the catalyst stability, we measured 5Ni/Al2O3_wi for longer time on stream, i.e. 10 h (Fig. 9) instead of 4 h (Fig. 8). XLA and YGVL were higher in comparison to the former test for 4 h on stream (Fig. 8a). Between 5 h and 8 h TOS, LA conversions and GVL yields decreased steadily to the level obtained after 1 h TOS (80 % LA conversion and 70 % GVL yield). After 9 h and 10 h TOS the deactivation of the catalyst was significantly stronger. A comparison with literature showed a similar deactivation as Cu/Al2O3 25 and Ni/H-ZSM-5 22 catalysts with longer TOS.

4 Conclusions A set of Ni/Al2O3 catalysts with average Ni particle sizes in the range of 6 – 65 nm were tested in the continuous liquid phase hydrogenation of LA to GVL in a trickle-bed reactor in water as solvent. The size and oxidation state of the Ni particles were found to be crucial factors for a good catalytic performance. The different Ni particles supported on Al2O3 were prepared by wet impregnation, precipitation and flame spray pyrolysis, respectively. In particular, slow precipitation using urea resulted in small Ni particles after reduction in hydrogen. Catalysts containing smaller Ni particles (