Synthesis of γ-valerolactone by hydrogenation of levulinic acid over supported nickel catalysts

Accepted Manuscript Title: Synthesis of ␥-Valerolactone by Hydrogenation of Levulinic Acid over Supported Nickel Catalysts Author: Konstantin Hengst Martin Schubert Hudson W.P. Carvalho Changbo Lu Wolfgang Kleist Jan-Dierk Grunwaldt PII: DOI: Reference:

S0926-860X(15)00299-9 http://dx.doi.org/doi:10.1016/j.apcata.2015.05.007 APCATA 15379

To appear in:

Applied Catalysis A: General

Received date: Revised date: Accepted date:

16-2-2015 8-5-2015 9-5-2015

Please cite this article as: K. Hengst, M. Schubert, H.W.P. Carvalho, C. Lu, W. Kleist, J.-D. Grunwaldt, Synthesis of rmgamma-Valerolactone by Hydrogenation of Levulinic Acid over Supported Nickel Catalysts, Applied Catalysis A, General (2015), http://dx.doi.org/10.1016/j.apcata.2015.05.007 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ce pt

ed

M an

us

cr

i

*Graphical Abstract (for review)

water

LA conversion

solvent free

Ac

alcohols

GVL selectivity

Page 1 of 32

*Highlights (for review)

Highlights Ni particles supported on alumina for the hydrogenation of levulinic acid (LA)



High conversion to γ-valerolactone under solvent free conditions



Side reactions of LA with alcohols lower at higher reaction pressure



High yields and 100 % selectivity in aqueous medium

Ac

ce pt

ed

M

an

us

cr

ip t



Page 2 of 32

Synthesis of γ-Valerolactone by Hydrogenation of Levulinic Acid over Supported Nickel Catalysts

Wolfgang Kleist1,2, Jan-Dierk Grunwaldt*,1,2 1

ip t

Konstantin Hengst1,2, Martin Schubert1,2, Hudson W. P. Carvalho1, Changbo Lu1,

cr

Institute for Chemical Technology Chemistry and Polymer Chemistry, Karlsruhe Institute of Technology, Engesserstraße 20, D-76131 Karlsruhe, Germany

2

us

Institute of Catalysis Research and Technology, Karlsruhe Institute of Technology,

*

Ac ce p

te

d

M

an

Hermann von Helmholtz Platz 1, D-76344 Eggenstein-Leopoldshafen, Germany

Corresponding author

Phone: +49 721 608 42120

; Fax: +49 721 608 44820

E-mail: [email protected] (J.-D. Grunwaldt)

1 Page 3 of 32

Abstract Ni/Al2O3 catalysts were tested for the hydrogenation of levulinic acid (LA) to γvalerolactone (GVL) as an important bio-based platform molecule for chemical products

ip t

based on renewable feedstocks. The catalysts were prepared by wet impregnation, incipient wetness impregnation, precipitation, and flame spray pyrolysis; both the

cr

influence of different solvents (monovalent alcohols and water) as well as solvent free reaction conditions were screened in batch autoclaves. Whereas alcohols led to a

us

number of side reactions that could only be suppressed by high hydrogen pressures (>20 bar), water as solvent resulted in a GVL selectivity of 100 %. The GVL yields

an

reached 57 %. Further improvement was achieved without any solvent, whereby the GVL yield increased to 92 % at 100 % LA conversion. Reuse of the Ni catalysts resulted

M

in a significant drop in activity. The catalysts were thoroughly characterized by temperature programmed reduction (TPR), X-ray diffraction (XRD), linear combination

d

analysis of X-ray absorption near edge structure (XANES) spectra and extended X-ray

te

absorption fine structure (EXAFS). The results indicated that incorporated Ni2+ as

Ac ce p

present in flame-derived catalysts was less active for GVL synthesis compared to surface Ni particles as present in the wet impregnated catalyst.

Keywords: Hydrogenation; γ-valerolactone; levulinic acid; Ni/Al2O3; solvent effect

2 Page 4 of 32

1. Introduction γ-Valerolactone (GVL) is a versatile lactone which can be directly used as fragrance, green solvent, monomer for biomass-derived plastics or gasoline blender. Furthermore,

ip t

GVL is utilized to produce various bio-based products like aromatics (fuel additives), short chain alkenes (jet fuel) or long chain alkanes (diesel fuel).[1-3] GVL can be

cr

synthesized from levulinic acid (LA) either by hydrogenation of LA to γ-hydroxyvaleric acid, which spontaneously condensates to GVL, or by dehydration of LA to angelica

us

lactone, which is subsequently hydrogenated to GVL.[4] The condensation of LA to GVL occurs mainly in the presence of acid functionalities. LA as starting reactant is attractive

an

since it can be easily generated from lignocellulosic feedstocks by selective chemical transformations.[5-7] In contrast to the complex production processes of other platform

M

molecules (oxidation or biochemical processes like fermentation), only de- and rehydration reactions as well as a deformylation reaction (5-hydroxymethylfurfural to LA)

te

d

are necessary to produce LA.[8, 9]

Various transition metal based catalysts, both homogeneous and heterogeneous, have

Ac ce p

been developed for the hydrogenation of LA to GVL.[1, 9-11] Already in 1909, the hydrogenation of LA to GVL was reported by Sabatier and Mailhe [12] using a Raneynickel catalyst in gas phase at 250 °C. Also Christian et al. [13] used a Raney-nickel catalyst at 220 °C for the hydrogenation of LA to GVL (GVL yields of 94 %) and Schütte and Thomas [14] investigated the GVL synthesis using platinum oxide as catalyst and diethyl ether as solvent (GVL yields of 87 %). Since 2000, the hydrogenation of LA to GVL has received renewed attention using supported Ru, Pd and Pt based catalysts in both continuous and discontinuous reaction modes at reaction temperatures between 130 and 220 °C and hydrogen pressures up to 55 bar. I n most studies, various alcohols, water or different ethers were used as solvents and only a few investigations on solvent free hydrogenation of LA to GVL have been reported.[3, 4, 11, 15, 16] For instance, Al3 Page 5 of 32

Shaal et al. [17] tested 5 wt.% Ru supported on activated carbon, Al2O3, TiO2 and SiO2 as catalysts for the GVL synthesis. The reactions were conducted in autoclaves using different solvents. Nearly quantitative conversion of LA to GVL was achieved after 2.5 h using butanol as solvent, a hydrogen pressure of 20 bar and a reaction temperature of

ip t

130 °C. The groups of Manzer [18] and Poliakoff [19] i nvestigated the hydrogenation of LA to GVL in supercritical CO2 using 5 wt.% Ru supported on Al2O3 (250 bar, 150 °C,

cr

2 h) and 5 wt.% Ru supported on SiO2 (100 bar, 200 °C), respectively. In both studies

us

quantitative conversion of LA to GVL was achieved. The use of Pd or Pt on acid supports like zeolites led to the hydrogenation of γ-valerolactone to 1,4-pentanediol

an

which partly further condensates to 2-methyltetrahydrofuran, thus decreasing the GVL selectivity.[11, 20-22]

M

The main disadvantage of noble metal based catalysts is their high costs and therefore the development of non-noble metal based catalysts is important. In the last few years

d

non-noble metals like Ni or Cu were studied as hydrogenation catalysts for the GVL

te

synthesis.[3, 23-31] Manzer [3] investigated 5 % Ni supported on carbon in 1,4-dioxane

Ac ce p

as solvent at a reaction temperature of 150 °C, a react ion time of 2 h and 55 bar hydrogen pressure. Compared to other metals (Ir, Rh, Pd, Ru, Pt, Re), which were tested under the same conditions, Ni showed both the lowest conversion of LA (5 %) and the lowest GVL selectivity (20 %). The group of Rao [28, 29] investigated Ni supported on different supports in a vapor phase hydrocyclization of LA to GVL at temperatures above 250 °C under atmospheric pressure and obtained GVL yields of 80 % with 30 wt.% Ni on SiO2. Bimetallic catalysts like Ni-Cu/Al2O3 [30] or Ni-MoOx/C [26] were studied at high reaction temperatures (250 °C) and H 2 pressures (50 to 65 bar) and with both catalysts GVL yields of over 90 % were achieved. Up to now, little attention has been paid to the solvent and the preparation method for non noble metal based catalysts. Hence, we report here on the liquid phase 4 Page 6 of 32

hydrogenation of LA to GVL in different solvents using supported Ni/Al2O3 catalysts prepared by different preparation methods. The focus is laid on supported nickel catalysts as cheaper alternative to the expensive noble metals and Raney-Nickel. Especially, solvent free conditions or water as solvent would enhance the sustainability

ip t

of the GVL synthesis. All catalysts were well characterized before and after the reaction

reduction (TPR) and X-ray absorption spectroscopy (XAS).

us

2. Experimental section

cr

by specific surface area, powder X-ray diffraction (XRD), temperature-programmed

an

2.1 Catalyst preparation

Four different Ni catalysts were prepared using wet impregnation (WI), incipient wetness

M

impregnation (IWI), precipitation (P) and flame spray pyrolysis (FSP). Each catalyst has a metal content of 15 wt% nickel. Nickel(II) nitrate (Sigma-Aldrich) was used as

d

precursor, γ-Al2O3 (Carl Roth) as support for impregnation and precipitation and

te

aluminum nitrate (Sigma-Aldrich) as precursor for the flame spray pyrolysis.

Ac ce p

Wet impregnation was performed by suspending of 8.5 g uncalcined γ-Al2O3 in 30 mL of an aqueous Ni(II) nitrate solution (7.427 g Ni(II) nitrate·6H2O) and the excess water was removed in a rotary evaporator. The impregnated catalyst was dried for 1 h at 110 ° and afterwards calcined for 5 h at 600 °C. The Ni/Al2O3_iwi catalyst was prepared by suspending 8.5 g of the uncalcined γ-Al2O3 thoroughly with 8.5 mL of an aqueous Ni(II) nitrate solution (7.427 g Ni(II) nitrate·6H2O). The volume of the aqueous Ni(II) nitrate solution was equal to the pore volume of the γAl2O3 (pore volume 1 cm3/g). The IWI catalyst was dried for 1 h at 110 °C and afterwards calcined for 5 hours at 450 °C. Apart from t his rather simple method we used additional preparation methods, which were applied after the solvent screening tests.

5 Page 7 of 32

For preparation of 15 wt.% Ni/Al2O3 by precipitation, 8.5 g of the uncalcined γ-Al2O3 was suspended in 100 mL of an aqueous Ni(II) nitrate solution (7.427 g Ni(II) nitrate·6H2O) and stirred for one hour. Thereafter, 465 mL of a 0.1 M NaOH solution was added at room temperature to the mixture until a pH of 9 was reached. The catalyst was filtered

ip t

off, washed with distilled water until the pH of the filtrate was 7, dried for 1 h at 110 °C

cr

and afterwards calcined for 5 h at 600 °C.

For flame spray pyrolysis 2.975 g Ni(II) nitrate and 24.995 g aluminum nitrate were

us

dissolved in 120 mL methanol, pumped with 5 mL/min into the center of a methane/oxygen flame and dispersed with 5 L/min oxygen to form a fine spray. The

an

catalyst was collected on a filter 30 cm above the flame and calcined at 600 °C for 2 h. The set up used for the flame spray pyrolysis was similar to the one described in [32].

M

Note that particle production and collection on the filter need to be performed in a fume hood and appropriate safety measures need to be undertaken not to be exposed to any

d

nanoparticles. Ni/Al2O3_wi, Ni/Al2O3_p and Ni/Al2O3_fsp were reduced in a tubular

te

furnace under 10 % H2/N2 with a total flow of 20 nL/h at 600 °C for 2 h (he ating rate of 5

Ac ce p

K/min) while the Ni/Al2O3_iwi catalyst was only reduced at 450 °C.

2.2 Catalyst characterization

2.2.1 Physisorption measurements The specific surface area of the catalysts was determined by N2 physisorption (Belsorp II mini, BEL Japan Inc.) at -196 °C. All samples were o utgassed for 2 h at 300 °C prior to recording the adsorption isotherms. The BET surface area was determined using 10 points in the range of p/p0 = 0.05 – 0.3.

6 Page 8 of 32

2.2.2 Temperature programmed reduction The reduction behavior of the catalysts was investigated by temperature programmed reduction and performed in an in-house built set-up. A calcined sample of the catalyst

ip t

(100 mg) was filled in a quartz glass tube, put in a tube furnace and heated up to 800 °C (5 K/min) in 5 % H 2/Ar with a total flow of 96 mL/min. The H2 concentration of the

cr

influent and effluent gas was measured with a thermal conductivity detector and

us

afterwards the H2 consumption was plotted against the temperature. 2.2.3 X-ray diffraction

an

Powder X-ray diffraction (XRD) patterns were collected using a PANalytical X`Pert PRO diffractometer with Cu Kα radiation (Ni filter; Cu Kα1 = 1.54060 Å and Cu Kα2 =

M

1.54443 Å and). The scan was recorded in a 2θ range of 20 – 80° with 0.017° step width and 0.51 s data acquisition per data point. Crystalline phases were determined

d

using the Cambridge structural data base (CSD), Cambridge crystallographic center, as

te

a reference data base. The Scherrer equation was used to estimate the mean particle

Ac ce p

diameter of the nickel particles and their dispersion.[33] A LaB6 standard was used calibrate and correct the instrumental line broadening. 2.2.4 In situ X-ray diffraction

In situ X-ray diffraction patterns were recorded using a Bruker D8 Advance with Cu Kα radiation (Cu Kα1 = 1.5406 Å and Cu Kα2 = 1.54439 Å). The calcined catalyst was loaded in a in situ XRD cell (Anton Paar, XRK900) and afterwards heated stepwise in 50 °C steps in 5 % H 2/N2 with a total flow of 100 mL/min to 650 °C. At each temperature a scan was recorded in a 2θ range of 20 – 80° (0.016 ° step width, 1 s data acqu isition time, total 1 hour). For analysis of the crystalline phase, the reflections were assigned using references from the International Centre for Diffraction Data (ICDD).

7 Page 9 of 32

2.2.5 Inductively coupled plasma optical emission spectrometry The nickel content of the catalysts was validated by inductively coupled plasma optical emission spectrometry (ICP-OES) using an Agilent 720/725-ES spectrometer. The

ip t

catalyst was dissolved in HNO3, HCl, HF and 30 % H2O2 in a microwave at 600 W for 2 h, afterwards diluted with distilled water and finally injected in the plasma.

cr

2.2.6 Transmission electron microscopy

us

Transmission electron microscopy was conducted on a FEI Tecnai 20 electron microscope at an acceleration voltage of 200 kV with a LaB6 filament. Typically, a small

an

amount of sample was ground and suspended in ethanol, sonicated and dispersed over a Cu grid with a carbon film.

M

2.2.7 X-ray absorption spectroscopy

X-ray absorption spectra at the Ni K edge (8333 eV) in terms of X-ray absorption near

d

edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) were

te

recorded in transmission mode at the bending magnet beamline ANKA-XAS of the

Ac ce p

ANKA synchrotron radiation facility (Karlsruhe, Germany). The energy of the X-rays were scanned by a Si(111) double crystal monochromator (DCM). Higher harmonics were rejected by detuning the DCM to 60% of its maximum intensity using a D-Mostab (Struck, Germany). XAS spectra were recorded on pelletized samples (pressed after dilution in cellulose) using a beam of 1 mm x 7 mm (height x width). Data reduction and analysis were carried out using Athena and Artemis software of the IFEFFIT package.[34] The spectra were energy calibrated using a reference Ni foil (E0 = 8333 eV), normalized and background subtracted using the autobkg algorithm in Athena software. In addition, a least-square linear combination analysis fitting of the normalized XANES spectra was performed in the energy range -20 to 90 eV relative to the edge position to identify the fraction of mixed phases. As potential components the XANES 8 Page 10 of 32

spectra of NiO, Ni foil, NiAl2O4 and highly dispersed NiO particles obtained by wet impregnation were used. For analysis of the extended X-ray absorption fine structure (EXAFS) only the single

ip t

scattering paths were considered. The corresponding theoretical backscattering amplitudes and phases were calculated by the FEFF 6.0 code.[35, 36] The theoretical

cr

single scattering paths were adjusted to the experimental ones by a least square method in R-space to obtain the coordination number (N), bond distances (R), and

us

mean square deviation of interatomic distances (σ2). The amplitude reduction factor (S0) was obtained by refining a Ni reference foil and used for the other samples. The abso-

an

lute misfit between theory and experiment is expressed by the ρ factor (cf. ref. [37]).

M

2.3 Catalytic tests

Screening of catalysts in aqueous phase or under solvent free conditions were carried

d

out in custom-made batch autoclaves (max. temperature 350 °C, max. pressure

te

200 bar, volume 65 mL). Catalytic tests in other solvents (monovalent alcohols, acetic acid, decane) were performed in a Parr batch autoclave allowing higher temperature

Ac ce p

and pressure (max. temperature 500 °C, max. pressure 35 0 bar, volume 100 mL). After charging the reactor with 10 mL of the reaction mixture (solvent and LA) and the 0.1 g of catalyst, the reactor was purged with nitrogen and pressurized with hydrogen. The ratio between solvents and LA was 20:1 (mLsolvent/gLA) and the typical molar ratio of Ni and LA was 1:30. The turnover frequency was calculated with the starting amount of LA in mol (nLA), the LA conversion (XLA), the amount of Ni in mol (nNi) and the dispersion (D).

TOF =


 ∙  

  

∙

The dispersion was obtained from the mean particle size calculated using the Scherrer equation according to ref. [33].

9 Page 11 of 32

The custom made autoclave was heated with heating sleeve and heating plate and was magnetically stirred whereas for the Parr batch autoclave a tube furnace was used and the mixture was stirred using a mechanical stirrer. The starting point of the reaction was defined as the time, when the desired temperature was reached (usually after 20 to 30

ip t

minutes). After the corresponding reaction time the reactor was quenched in an ice bath, depressurized, flushed with nitrogen and finally the product was separated from

cr

the catalyst by centrifugation.

us

2.4 Product analysis

The product mixtures of catalytic tests in aqueous medium or neat LA were diluted with

an

distilled water and afterwards analyzed by high performance liquid chromatography (HPLC) using a HPLC apparatus from Merck-Hitachi containing a BioRad organic acid

M

column (Aminex HPX 87H), a refraction index detector and a UV detector (wavelength:

d

254 nm). A 0.004 M sulfuric acid solution was used as mobile phase with a flow rate of

te

0.5 mL/min. The column was operated at a temperature of 40 °C with a pressure of 45 bar. The product mixtures of the catalytic tests with organic solvents were analyzed

Ac ce p

using gas chromatography (Shimadzu GC2010 plus; polar Restek column (Rxi®624Sil)) and tetradecane (TD) as internal standard to calculate the LA and GVL concentrations.

3. Results and discussion 3.1 Catalyst characterization

3.1.1 Standard characterization The Ni contents, BET surface areas, reduction peaks and crystal phases of the four catalysts are shown in Table 1. The Ni content was similar for all catalysts and in good accordance with the theoretical values. The specific BET surface areas of Ni/Al2O3_wi, 10 Page 12 of 32

Ni/Al2O3_p and Ni/Al2O3_iwi were 96 m2/g, 122 m2/g and 98 m2/g, respectively, which is slightly lower than the surface area of pure γ-Al2O3 (145 m2/g). The surface area of the Ni/Al2O3_fsp catalyst (31 m2/g) was significantly lower than that of Ni/Al2O3_wi, Ni/Al2O3_iwi and Ni/Al2O3_p, but in the same range like typical values for nickel

ip t

aluminates.[38]

cr

Table 1: Analysis of the Ni catalysts by ICP-OES, BET surface area TPR and XRD.

Catalyst

Ni content by ICP-OES (wt.%]

Specific surface Area [m2/g]

Reduction Peak Maximum [°C]

Ni/Al2O3_wia

14.0

96

380 and 700

Ni

50

Ni/Al2O3_pa

14.0

122

380 and 700

Ni

50

Ni/Al2O3_fspa

13.4

31

800

no Ni

-e

Ni/Al2O3_iwib

14.8

98

530

Ni, NiO

7 (NiO)

Particle size [nm]d

M

an

us

Crystal Phase [XRD]c

a: reduction at 600 °C, b: reduction at 450 °C, c: after reduction, d: estimated with Scherrer equation,e: no metallic Ni phase.

d

The structure of the four 15 wt% Ni/γ-Al2O3 catalysts was further investigated by powder

te

X-ray diffraction. The XRD patterns of the Ni/γ-Al2O3 catalysts prepared by wet

Ac ce p

impregnation, incipient wetness impregnation and precipitation showed a NiO phase in the calcined catalysts and metallic Ni in the reduced catalysts. Ni/Al2O3_iwi was reduced at 450 °C and the NiO phase was still found a fter the reduction. The XRD patterns of the Ni/Al2O3_wi, Ni/Al2O3_iwi and Ni/Al2O3_p catalysts after the reaction were similar to the XRD patterns of the reduced catalysts before the reaction. The Ni or NiO particle size of the reduced Ni/Al2O3_wi, Ni/Al2O3_p and Ni/Al2O3_iwi catalysts was estimated using the Scherrer equation and is about 50 nm for Ni/Al2O3_wi and Ni/Al2O3_p (reflections of Ni) and about 7 nm for Ni/Al2O3_iwi (reflections of NiO). The additional existence of small Ni particles which are too small to be detected by XRD besides these large Ni particles, are not contributing in this case (see also TEM results).

11 Page 13 of 32

The XRD patterns of the Ni/Al2O3_wi, Ni/Al2O3_p, Ni/Al2O3_fsp and Ni/Al2O3_iwi catalysts in the calcined and reduced form are shown in Figure S1 of the Electronic Supporting Information (ESI). The X-ray diffraction patterns of the calcined, the one reduced at 600 °C and the used Ni/Al 2O3_fsp catalysts were similar. Only reflections

ip t

stemming from γ-Al2O3 and not from Ni containing phases were found (Figure S1). Also

cr

possible nickel aluminum spinel formation could not be identified using X-ray diffraction. The reduction behavior of Ni/Al2O3_wi, Ni/Al2O3_p Ni/Al2O3_iwi and Ni/Al2O3_fsp was

us

further investigated by TPR and gave similar results for the catalysts prepared by wet impregnation and precipitation. For both catalysts the reduction started at 300 °C and

an

the H2 consumption featured a broad peak at 380 °C. A second p eak of the H2 consumption was observed for both catalysts at 700 °C which was more defined for the wet

M

impregnated catalyst. The two peaks of the H2 consumption may result from a broad particle size distribution. Large NiO particles are reported to be reduced at lower tem-

d

peratures (300 °C – 400 °C) while higher temperature s (600 °C – 750 °C) are necessary

te

for the reduction of small NiO particles [39]. The reduction of Ni/Al2O3_iwi started at 350

Ac ce p

°C and the H 2 consumption featured a broad peak with a maximum at 530 °C. Also for the catalyst prepared by flame spray pyrolysis reduction started at 300 °C and the H2 consumption increased continuously to 800 °C. The contin uously increasing H2 consumption might be caused by the structure of the FSP catalyst. Due to the high temperature (> 900 °C) during the flame spray pyrolysis a spinel structure (NiAl2O4) or NiO, which is enclosed in alumina and thus more difficult to reduce, might be formed. The TPR profile of the FSP catalyst was similar to those reported for Ni aluminum spinels.[40, 41] In addition, transmission electron microscopic analysis was conducted on the reduced Ni/Al2O3_wi and Ni/Al2O3_p catalysts to gain insight into the existence of small and large Ni particles as assumed from XRD and TPR analysis. Also TEM images of the 12 Page 14 of 32

Ni/Al2O3_fsp catalyst were recorded as the Ni phase is invisible in XRD and to unravel the Ni particle size. The corresponding TEM images of the three catalysts can be found in the ESI (Figures S2, S3 and S4). The TEM-images confirm the existence of small and large Ni particles on the Ni/Al2O3_wi and Ni/Al2O3_p catalysts. Also small Ni particles on

ip t

the Ni/Al2O3_fsp catalyst as assumed from XAS measurements (section 3.1.2) could be verified. The particle size supports the data from XRD and as only a limited number of

cr

images are available and the smaller Ni particles due to re-oxidation give a rather bad

us

contrast we used the average particle size determined by XRD for calculation of the TOF (chapter 3.3).

an

To investigate the reduction behavior of the catalyst prepared by wet impregnation in more detail and to determine the optimal reduction temperature we followed the reduc-

M

tion via in situ XRD. Similar to the TPR results reduction of NiO to Ni started at 300 °C. The XRD pattern collected at 300 °C showed reflections of both NiO and small contribu-

d

tions of metallic Ni. The broad temperature range of reduction obtained with TPR could

te

be proven by the in situ collected XRD patterns. All XRD patterns between 300 °C and

Ac ce p

550 °C showed reflections of NiO and metallic Ni (cf. ESI, Figure S5). At 600 °C the XRD pattern featured only a Ni phase indicating a fully reduced catalyst. 3.1.2 X-ray absorption spectroscopy (EXAFS and XANES) Figure 1a shows the near edge X-ray absorption spectra at Ni K edge for the reduced 15 wt.% Ni catalysts prepared by wet impregnation (wi), precipitation (p) and flame spray pyrolysis (fsp). According to XRD analysis (Figure S5) this treatment at 600 °C under H2 led to a completely reduced Ni/Al2O3_wi catalyst. Due to the similar structural data the same is expected for Ni/Al2O3_p. On the other hand, the XAS analysis unravels that these catalysts were partially oxidized after air exposure. Due to the lower white line intensity and edge position, one may infer that the Ni/Al2O3_wi and Ni/Al2O3_p catalysts were more reduced than the Ni/Al2O3_fsp catalyst. This ex situ XAS analysis reflected the state of the catalyst as they were loaded in the batch reactor. Re-oxidation

13 Page 15 of 32

te

d

M

an

us

cr

ip t

may partially occur for the Ni/Al2O3_fsp after the reduction procedure as well well.

Ac ce p

Figure 1: Figure 1a a) X-ray ray absorption near edge spectra at the Ni K-edge K edge of the reduced Ni/Al2O3_wi, Ni/Al2O3_p and Ni/Al2O3_fsp catalysts; b) Experimental and refined Fourier transformed k2 weighted EXAFS spectra.

The he refinement of EXAFS data provided further information on the local chemical environment of Ni. Figure 1b shows the experimental and refined Fourier transformed k2 weighted EXAFS spectra of Ni/Al2O3_wi and Ni/Al2O3_p. The backscattering contributions of the oxygen and Ni neighbours of the Ni/Al2O3_p catalyst is higher than those of Ni/Al2O3_wi. This is also reflected by the lower coordination numbers (Table 2) and indicates that Ni/Al2O3_wi has smaller particles than Ni/Al2O3_p, although they appear to be very similar both from TPR and XRD analysis (Table 1). According to the

Page 16 of 32

fit results in Table 2 Ni/Al2O3_wi and Ni/Al2O3_p contain mainly metallic species, although some oxygen neighbours were also found around Ni (more details on the

ip t

refinement and individual scattering paths in the ESI, cf. Figure S6).

Table 2: Structural parameters of the Ni absorber extracted from the EXAFS spectra of the calcined and activated Ni/Al2O3 catalyst.

Ni/Al2O3_fsp

1.2 ± 0.4a

2.04 ± 0.04a

2nd

Ni

8.0 ± 0.9a

2.49§a

3rd

Ni

4.0 ± 0.4c

3.49 ± 0.02a

4th

Ni

17.5 ± 6.1a

4.34 ± 0.01a

1st

O

1.0 ± 0.4a

2.02 ± 0.03a

6.0 ± 3.5a

2nd

Ni

10.0 ± 0.7a

2.49§a

6.4 ± 0.5a

3rd

Ni

5.0 ± 0.4c

3.49 ± 0.02a

8.7 ± 1.8a

4th

Ni

20.0 ± 2.0c

4.34 ± 0.01a

8.5 ± 0.9a

1st

O

3.8f

9.2 ± 2.7a

an

9.6 ± 2.8a

7.5±0.7

2.7

5.1 ± 2.1a -1.6±4.1

2.1

rd

3

Al

3

2.80 ± 0.03

16.1 ± 5.9a

4th

Al

3f

3.29 ± 0.02a

3.6 ± 2.1a

5th

Al

6f

3.70± 0.02a

9.2± 3.2a

1st

Ni

12f

2.48§a

5.9± 0.1a

6.7±0.3

a

-0.7±0.9

Ni

a

6.2 ± 0.6

st

1

O

f

f

6

2.5

6.7 ± 0.9a

2.02 ± 0.03a

Ac ce p

NiO

6.2 ± 1.6a 7.9 ± 1.0

ρ [%]

us

O

2

2

∆E0[eV]

1st

nd

Ni foil

σ 2x 10-3[Å2]

cr

R[Å]

M

Ni/Al2O3_p

N

d

Ni/Al2O3_wi

Shell Atom

te

Sample

§a

a

2.49

6.4 ± 0.5 a

a

2.08 ± 0.01

3.8 ± 1.5

1.5 1.2

rd

S0 =0.80, §= fitted uncertainty lower than 1%, a= fitted, f=fixed and c=constraint (3 shell=2nd shell/2 and 4th shell=2nd shell x 2). Structural parameters: N=number of neighboring atoms, r= interatomic distance, σ 2= mean square deviation of interatomic distances, ρ= misfit between the experimental data and the theory.

In order to shed more insight into the structure of the partially reduced fsp catalyst, linear combination fitting of the Ni K edge XANES spectrum with the reduced Ni/Al2O3_wi catalyst (mainly metallic particles, small), a NiAl2O4 reference and NiO was 15 Page 17 of 32

conducted (Figure 2).. The linear combination analysis unravelled that the Ni/Al2O3_fsp catalyst could be best described as a combination of 47% NiAl2O4 and air exposed Ni Niparticles (fraction of 53 %, reduced Ni/Al2O3_wi catalyst). Linear combinations based on a combination of NiO and metallic N Ni resulted in poor agreement (cf. ESI, Figure S7). S7

te

d

M

an

us

cr

as a spinel-like phase and as metallic Ni after reduction at 600 °C.

ip t

Hence, the fsp-derived catalyst seems to be best described by a mixture of Ni present

Ac ce p

Figure 2: X-ray ray absorption near edge spectra at the Ni K-edge K edge and linear combination analysis of Ni/Al2O3_fsp catalyst after reduction.

This is in line with previous revious studies showing show that catalysts prepared by flame spray pyrolysis are more difficult to reduce than impregnated ones, since the metallic ions can be within the oxide lattice. [42] According to powder X-ray diffraction (only the presence of an Al2O3 phase is visible) Ni is highly dispersed. Hence, a structural model was built to refine the EXAFS data based on the following assumptions: (i) one first coordination shell with Ni in a spinel-like phase in octahedral coordination and the the outer shells with Ni in the lattice of Al2O3 (ii) a metallic Ni phase. The Fourier transformed EXAFS spectrum and corresponding refinement ement are presented in Figure S8. S8. The results suggest that at least some of the Ni atoms were incorporated into the Al2O3 lattice during the preparation. Therefore it could be that surface Ni was reduced forming the metallic

Page 18 of 32

particles whereas the incorporated fraction was converted into a spinel-like phase. The spinel-like phase was similar to those reported by Fu et. al. [43]. In this study the authors annealed Ni acetate impregnated on γ-Al2O3. While the sample treated at 500°C presented mainly NiO features, the spectra for t he samples treated at 750 °C

ip t

and 950 °C were similar to the one obtained for Ni/A l2O3_fsp in the present study. However, their Ni-O spinel-like bond length was around 1.83 Å, whereas in the present

cr

study we have found 2.02 Å (Table 2), indicating that only a spinel-like phase is formed

us

in case of Ni/Al2O3_fsp.

In summary, the local structure of nickel was strongly influenced by the preparation

an

method, which therefore may play an important role on the catalytic activity (see comparison of catalysts in section 3.3). Both precipitation and wet impregnation resulted

M

in small metallic Ni particles covered by oxygen. However, the impregnation method led to smaller particles. On the other hand, the flame made material was composed of a

d

mixture of metallic Ni and Ni in an Al2O3 matrix. Together with the particle size obtained

te

from the XRD patterns calculated with the Scherrer equation, the analysis of the TPR

Ac ce p

and TEM images, these result support the hypothesis that beside few large Ni particles (50 nm) also small Ni particles (2 – 4 nm) exist. Compared to XRD, which only detects particles above 5 nm, with EXAFS all Ni particles are probed in a global manner (not locally like in TEM). H2-chemisorption measurements on the Ni/Al2O3_fsp catalysts were performed to determine the particle size of the small particles. Due to the spinel formation, the reduction temperature to achieve full reduction is very high. For the chemisorption measurements we reduced the catalyst at 600 °C as we did before the catalytic tests. After the reduction at 600 °C the cataly st is not completely reduced. As the exact degree of reduction is unknown, the obtained dispersion is too low.

17 Page 19 of 32

3.2 Conversion of LA to GVL in different solvents First, the influence of different solvents on the conversion of LA to GVL was investigated using Ni/Al2O3_iwi and Ni/Al2O3_wi as catalysts. Different monovalent alcohols (C1 – C5), acetic acid ,DMF and water were used as solvents. Using acetic acid only traces of

ip t

GVL were found in the reaction mixture. Using DMF a GVL yield of only 3 % was achieved at a LA conversion of 24 %. Hence, these solvents were unsuited for the

LA conversion GVL yield

M

80

an

15 wt% Ni/γ-Al2O3

100

us

yield obtained with different alcohols and water as solvent.

cr

hydrogenation of LA to GVL (cf. Table S1). Figure 3 shows LA conversion and GVL

d

[%]

60

te

40

Ac ce p

20

0

water

methanol

ethanol

propanol

butanol

pentanol

Figure 3: LA conversion and GVL yield using different solvents and Ni/Al2O3_iwi catalyst; reaction conditions: pH2: 10 bar, Treaction: 150 °C, t reaction: 6 hours, nNi/nLA: 0.03, mcatalyst: 0.1 g.

Quantitative conversion of LA was obtained using methanol, 1-butanol and 1-pentanol as solvent. The conversions of LA in ethanol and 2-propanol were 75 % and 87 %, respectively, while only 2 % of LA conversion was observed in aqueous medium. In contrast to the high LA conversion, GVL yields were comparatively low in the presence of alcohols. The highest GVL yield of 34 % was achieved with 2-propanol. 1-Butanol and 1-pentanol resulted in a GVL yield of only 19 %. With methanol and ethanol the GVL yield was lower than 10 %. The GVL yield using water as solvent was 2 % which is 18 Page 20 of 32

much lower compared to catalytic tests with alcohols. However, the selectivity was 100 % in the presence of water. A selectivity of 100 % to GVL was also found when Pd [22] Cu-ZrO2 [27] and Ni-MoOx [26] catalysts were used in water, whereas Ru/C as catalyst [17] resulted in a lower GVL selectivity of 86 %. The low selectivity towards

ip t

GVL in the presents of alcohols can be explained by esterification of LA to the corresponding levulinic acid esters (cf. Figure S9) in good agreement with the

cr

literature.[27] The gas chromatograms (Figure 9, ESI) indicate, that the corresponding

us

levulinic acid esters are the main side products (except using methanol and propanol as solvent). The levulinic ester yields were also estimated with the effective carbon number

an

(ECN) method and the obtained levulinic acid ester yields can be found in Table S2 in the ESI. Also Palkovits et al. [17] described the formation of levulinic acid esters, thus

M

lowering the GVL yields. Note, that the levulinic acid ester formation decreases the GVL yields much more if Ni catalysts are used because noble metal based catalysts also

d

catalyze the GVL formation starting from levulinic acid esters and therefore in those

te

cases often alcohols as solvents are used.[17, 44] A longer reaction time for the

Ac ce p

hydrogenation of LA to GVL in presence of an alcohol did not increase the GVL yield, also not once 100 % LA is converted. Note that no 4-hydroxypentanoic acid was found, also not by NMR (not shown). The highest GVL selectivity over Ni/Al2O3_iwi in the presence of an alcohol was obtained for 2-propanol (39 %). Additionally, the effect of the H2 pressure was investigated using 2-propanol and water as solvents. Figure 4 shows LA conversion, GVL yields and selectivities at hydrogen pressures between 5 and 50 bar with 2-propanol as solvent. Both LA conversion and GVL yield increased from 70 % (conversion) and 9% (yield) at 5 bar H2 pressure to 87 % (conversion) and 34 % (yield) at 10 bar H2 pressure. Further increase of the pressure led to a decrease of the LA conversion as well as the GVL yield to 45 % and 25 % (at 30 bar) and to 22 % and 20 % (at 50 bar), respectively. 19 Page 21 of 32

Interestingly, the selectivity towards GVL increased over the whole pressure range. Obviously, the side reaction of LA with propanol to the corresponding levulinic acid ester was suppressed at elevated hydrogen pressure. Figure 5 depicts the LA conversion and GVL yields with water as solvent at different hydrogen pressures. Compared to the test

ip t

in 2-propanol, the reactions were carried out at 200 °C with a reaction time of 4 hours

100 LA conversion GVL yield GVL selectivity

an

80

15 wt% Ni/γ-Al2O3

us

cr

and the Ni/Al2O3_wi catalyst (reduced at 600 °C) was used.

80

60

[%]

M

60

40

d

40

20

Ac ce p

te

20

0

5

100

0 10

30

50

reaction pressure [bar]

Figure 4: Variation of the H2 pressure with propanol as solvent and Ni/Al2O3_iwi catalyst; reaction conditions: Treaction: 150 °C, t reaction: 6 hours, nNi/nLA: 0.03, mcatalyst: 0.1 g.

20 Page 22 of 32

15 wt% Ni/γ-Al2O3

100

80

LA conversion GVL yield

[%]

ip t

60

cr

40

us

20

0 10

30

50

an

reaction pressure [bar]

M

Figure 5: Variation of the H2 pressure with water as solvent and Ni/Al2O3_wi catalyst; reaction conditions: Treaction: 200 °C, t reaction: 4 hours, nNi/nLA: 0.03, mcatalyst: 0.1 g.

A higher reaction temperature of 200 °C and a reactio n time of 4 hours led to an in-

d

creased LA conversion (20 %) compared to 2 % LA conversion at 150 °C and a reaction

te

time of 6 hours (cf. Figure 3). Also the higher reduction temperature of the Ni/Al2O3_wi catalyst (600 °C) might affect the catalytic activity. The GVL selectivity was 100 % at

Ac ce p

both reaction temperatures. In contrast to the tests with 2-propanol both the LA conversion and GVL yield increased with an elevated H2 pressure when water was used as solvent. The LA conversion rose to 37 % at 30 bar and the highest GVL yield (57 %) was obtained with a hydrogen pressure of 50 bar. With respect to the behavior in water a similar trend was found on Pd catalysts [22] and the LA conversions reported therein are similar to those presented in our study. Catalyst stability is an important issue, especially under hydrothermal conditions. Analyzing the Ni/Al2O3_wi catalyst after the reactions in water, we observed a partial phase change of the γ–Al2O3 support to boehmite.

21 Page 23 of 32

3.3 Solvent free conversion of LA to GVL To prevent side reactions with alcohols to levulinic acid esters and to potentially increase the LA conversion compared to the use of water as solvent, the hydrogenation of LA to GVL was additionally conducted in the absence of solvents. GVL is reported as

ip t

a good solvent for the LA hydrogenation to GVL.[45, 46] Therefore, the formed GVL in the solvent-free hydrogenation of LA might have a positive effect on the catalytic

cr

activity. For the catalytic tests with neat LA, 10 g LA and 50 bar H2 pressure were used

us

for all experiments. Hydrogen consumption was very high under these conditions which resulted in a high pressure drop. Figure 6 shows LA conversion and GVL yields after a

an

reaction time of 4 hours at different reaction temperatures. In order to obtain

M

comparable data, Ni/Al2O3_wi was used in all parameter optimization experiments.

15 wt% Ni/γ-Al2O3

100 90

d

80

LA conversion GVL yield

te

60 50

Ac ce p

[%]

70

40 30 20 10

0

120

140

160

180

200

reaction temperature [°C]

Figure 6: Variation of the reaction temperature in a solvent free reaction with the Ni/Al2O3_wi catalyst; reaction conditions: treaction: 4 hours, nNi/nLA: 0.03, mcatalyst: 1 g, pH2: 50 bar.

At 120 °C a LA conversion of only 2 % was obtained which increased at higher reaction temperatures. LA conversions of 44 % and 92 % achieved at 180 °C and 200 °C, respectively. The GVL selectivity was 100 % at all temperatures. 22 Page 24 of 32

Also for other transition metal catalysts or those containing Raney-Nickel, reaction temperatures above 200 °C were required to achieve LA co nversions in the same range (> 80 %) as reported for noble metal catalyst. [13, 22, 26, 30] A reaction temperature of 200 °C as reported here is still quite low for non-no ble metal catalysts at comparable

ip t

catalyst/reactant ratio to obtain a LA conversion over 90 %. The influence of the catalyst

15 wt% Ni/γ-Al2O3

LA conversion GVL yield

us

100

an

80

60

40

M

[wt.%]

cr

amount on the LA conversion is depicted in Figure 7.

0 0

te

d

20

0,6

1,2

1,8

2,4

3

Ac ce p

catalyst concentration [mol% Ni]

Figure 7: Variation of the Ni concentration in a solvent free reaction with a Ni/Al2O3_wi catalyst; reaction conditions: Treaction: 200 °C, t reaction: 4 hours, pH2: 50 bar, mLA: 10 g.

The LA conversion and GVL selectivity were about 2 % in the absence of any catalyst (blank test) and increased from 11 % (0.6 mol % Ni) to 92 % (3.0 mol % Ni) with increasing nickel content. In the literature, generally higher reaction temperatures were required to obtain comparable LA conversion and GVL yield when similar Ni concentrations (1 to 3 mol % Ni [26, 30]) were applied. Finally, the reaction time was varied and the obtained GVL yields are depicted in Figure 8. The GVL yields increased continuously with longer reaction times. The starting point of the reaction (t = 0h) was defined after reaching the desired temperature, when a GVL 23 Page 25 of 32

yield of 5 % was found. After 30 minutes and 1 hour the GVL yield increased to 20 % and 33 %, respectively, and after 4 hours 92 % GVL yield were achieved. The GVL selectivity was 100 % in all cases.

100

ip t

15 wt% Ni/γ-Al2O3

90

cr

80

us

60 50 40

an

GVL yield [wt %]

70

30

M

20 10 0 1

2

3

4

d

0

te

reaction time [h]

Ac ce p

Figure 8: Variation of the reaction time in a solvent free reaction with a Ni/Al2O3_wi catalyst; reaction conditions: Treaction: 200 °C, n Ni/nLA: 0.03, mcatalyst: 1 g, pH2: 50 bar.

Although the catalyst does not show a significant deactivation with increasing reaction time (Fig. 8), recycling experiments unraveled a drop in the catalytic activity (Table 3). Table 3: Recycling experiments with Ni/Al2O3_wi; reaction conditions: Run 1: Treaction: 200 °C, nNi/nLA: 0.03, mcatalyst: 1.5 g, mLA: 15 g, pH2: 50 bar; treaction: 4 hours Run 2: Treaction: 200 °C, nNi/nLA: 0.03, mcatalyst: 1 g, mLA: 10 g, pH2: 50 bar; treaction: 4 hours. Run

Treatment after 1st run

LA conversion [%]

GVL yield [%]

1st

fresh reduced

68

68

2nd

washed with acetone

4

4

2nd

washed with acetone, calcined (550 °C), reduced (600 °C)

38

38

24 Page 26 of 32

Washing of the catalyst with acetone before the second run resulted in a complete loss of activity. Washing, calcination and reduction of the catalyst before the second run increased the activity compared to only washing the catalyst. However, the LA conversion and GVL yields were still 44 % lower compared to the 1st run with the fresh

ip t

catalyst. The characterization data of the spend catalysts showed no significant changes in their structure which could explain the loss of activity. Also ICP-OES

cr

analysis of the catalyst after the reaction and the second run showed no metal leaching

us

which might explain the lower catalytic activity. The reason for this deactivation is presently studied in our laboratory in more detail and continuous-flow experiments in a

an

fixed-bed reactor are presently conducted. This has the advantage that the catalyst is not exposed to air and / or other solvents after the reaction which may also have an

M

influence on the reaction.

The parameter screening of the hydrogentation of LA to GVL under solventless

d

conditions using the 3 mol % Ni/Al2O3_wi catalyst unraveled that the highest LA

te

conversion (92 %) and GVL yield (92 %) were obtained at a reaction temperature of

Ac ce p

200 °C. For comparison, catalysts prepared by precipita tion and flame spray pyrolysis were tested under these optimized conditions (Figure 9). LA conversion and GVL yield were much lower with Ni/Al2O3_p (25 %) and Ni/Al2O3_fsp (45 %) compared to Ni/Al2O3_wi (92 %). The lower catalytic activity of Ni/Al2O3_fsp may be caused by a partial incorporation of Ni into the framework structure of the support, as evidenced by TPR and XAS (cf. section 3.1). The significantly lower LA conversion of Ni/Al2O3_p compared to the Ni/Al2O3_wi catalyst is surprising since it featured similar textural properties (see section. 3.1). Both the peaks in the TPR and the reflections for NiO and Ni in the calcined and the reduced catalyst were similar. The surface area of the precipitated catalyst was slightly higher.

25 Page 27 of 32

15 wt% Ni/γ-Al2O3

100

GVL yield LA conversion

80

[%]

ip t

60

cr

40

0 precipitation

flame spray pyrolysis

an

wet impregnation

us

20

Figure 9: Comparison of different catalyst preparation methods in a solvent free reaction; reaction conditions: Treaction: 200 °C, t reaction: 4 hours, nNi/nLA: 0.03, mcatalyst: 1 g, pH2: 50 bar.

M

Also the XANES analysis showed strong similarities. Only EXAFS analysis showed significantly higher backscattering contributions for the precipitated catalyst (Figure S6)

d

and thus also the data fitting resulted in smaller Ni-Ni coordination numbers for

te

Ni/Al2O3_wi than Ni/Al2O3_p. Hence, the difference in the catalytic activity may result from Ni particles with smaller size (< 5 nm) which are invisible for XRD but extracted

Ac ce p

from EXAFS. In contrast to Ni/Al2O3_wi, the Ni particles were not detected by XRD. Although only 50% of the Ni sites are probably available in the flame-made catalyst, as LC-analysis gave an almost 1:1 ratio of Ni particles and Ni present in spinel-like structures, it is remarkable that it is significantly better than the catalyst prepared by precipitation and reaches almost half of the conversion of the wet impregnated catalyst. The TOF and initial rates of the catalysts prepared by wet impregnated and precipitated were calculated at ~ 20 % LA conversion (Table 4).

26 Page 28 of 32

Table 4: TOF and initial rates of catalysts Ni/Al2O3_wi and Ni/Al2O3_p at iso-conversion. Catalyst

LA conversion [%]

treaction [min]

TOF [h-1]

Initial rate [mmol/min]

20

30

674

0.57

Ni/Al2O3_p

25

240

105

0.09

ip t

Ni/Al2O3_wi

Note that the same particle size was used in the TOF calculation for both catalysts and

cr

was based on the XRD-particle size, which may disregard smaller particles. The TOF

us

and initial rate of the WI catalyst are about six times higher compared to that of the precipitated catalyst and may be attributed to the smaller particles as seen by EXAFS

an

on the Ni/Al2O3_wi catalyst. This confirmed the statement, that smaller Ni particles are more active for the LA hydrogenation to GVL. To our knowledge the highest TON for

M

nickel catalysts reported in literature is 4950 at 250 °C after 24 h.[26] Hence, the Ni/Al2O3_wi catalyst shows very interesting turnover rates at similar LA conversions and

d

GVL yields (> 90 %) but a 50 °C lower reaction temper ature. This indicates that Ni

te

particles with a size < 5nm lead to an increased catalytic activity and thus to much

Ac ce p

higher LA conversions and GVL yields. The higher activity of smaller Ni particles was also found for other hydrogenation reactions, e.g., for the hydrogenation of benzene with supported Ni catalysts, described by Molina and Poncelet [47].

4. Conclusions

Four different 15 wt.% Ni/γ-Al2O3 catalysts were prepared using different preparation methods that lead to different textural and catalytic properties in the hydrogenation of LA to GVL. In the first part of the study solvents and reactions conditions were varied to optimize the hydrogenation of LA to GVL over Ni-based catalysts. For this purpose a standard IWI catalyst was used. Among all alcohols, 2-propanol gave highest GVL yield (34 % at 87 % LA conversion). Due to side reactions to the corresponding levulinic acid esters, GVL selectivity was relatively low (40 %) at 10 bar H2 pressure. By increasing 27 Page 29 of 32

the hydrogen pressure both LA conversion and GVL yields were decreasing, but the GVL selectivity increased (91 % at 50 bar). Using water as solvent the GVL selectivity was 100 % but higher reaction temperatures were required. Under optimized conditions (reaction temperature: 200 °C, reaction time 4 h, p H2 = 50 bar, nNi : nLA = 0.3, reduction

ip t

temperature 600 °C) GVL yields of 57 % were achieved. This could be further improved

cr

by optimizing the catalyst preparation and the solvent.

Most attractive over the Ni catalysts in this study has been hydrogenation of LA to GVL

us

under solvent-free conditions. In all cases, the GVL selectivity was 100 % and the best GVL yields (92 %) were obtained at 200 °C with 3 mol % Ni after 4 h. The catalytic acti-

an

vity of the GVL synthesis strongly dependent on the reaction temperature. Among the different catalysts Ni/Al2O3_wi showed the highest catalytic activity (GVL yield 92 %) of

M

all applied catalysts with high reaction rates. The comparison with a flame derived and a catalyst prepared by precipitation indicated that the difference in catalytic activity can

d

probably be explained by the different Ni particle size and incorporation of Ni into the

te

lattice. Combined TPR, XRD and EXAFS analysis unraveled that in the flame made

Ac ce p

catalysts Ni was partially incorporated into the lattice and that in the catalyst prepared by wet impregnation both small particles (active, as evidenced by EXAFS and TEM, invisible for XRD) and larger particles (reduced at low temperatures in TPR and detected by XRD) are present. Hence, smaller Ni particles seem to be beneficial for the hydrogenation of LA to GVL.

Acknowledgments

We acknowledge financial support from the European Institute of Innovation and Technology under the KIC InnoEnergy SYNCON project. Moreover, we thank Hermann Köhler for the ICP-OES measurements, Arno van Hoof for recording the TEM images and Angela Beilmann for the physisorption measurements. Finally, the authors wish to

28 Page 30 of 32

thank ANKA (Karlsruhe, Germany) for providing beamtime and Dr. Stefan Mangold for support during the XAS experiments at ANKA-XAS.

Electronic support information

ip t

In the Electronic Supporting Information additional data are given on the textural properties of the catalysts (XRD patterns, TEM images, Fourier transformed EXAFS

cr

spectra), details on EXAFS refinements, additional catalytic data (LA hydrogenation in other solvents), examples for GC-analysis and the calculation of levulinic acid ester

us

yields.

an

References

Ac ce p

te

d

M

[1] D.M. Alonso, S.G. Wettstein, J.A. Dumesic, Green Chem., 15 (2013) 584-595. [2] C. Jaimes, R. Dobreva-Schué, O. Giani-Beaune, F. Schué, W. Amass, A. Amass, Polym. Int., 48 (1999) 23-32. [3] L.E. Manzer, Appl. Catal. A, 272 (2004) 249-256. [4] Z.P. Yan, L. Lin, S.J. Liu, Energy Fuels, 23 (2009) 3853-3858. [5] D.W. Rackemann, W.O.S. Doherty, Biofuels Bioprod. Biorefin., 5 (2011) 198-214. [6] M.J. Climent, A. Corma, S. Iborra, Green Chem., 16 (2014) 516-547. [7] J. Han, S.M. Sen, D.M. Alonso, J.A. Dumesic, C.T. Maravelias, Green Chem., 16 (2014) 653-661. [8] G.P. T. Werpy, J. Holladay, J. White, A. Manheim, M. Gerber, K.Ibsen, L. Lumberg, S.Kelley, Top Value Added Chemicals From Biomass. Volume I: Results of Screening for Potential Canditates from Sugars and Synthesis Gas, U.S. Department of Energy, 2004. [9] K. Hengst, M. Schubert, W. Kleist, J.-D. Grunwaldt, Hydrodeoxygenation of LignocelluloseDerived Platform Molecules, in: R. Rinaldi (Ed.) Catalytic Hydrogenation for Biomass Valorization, Royal Society of Chemistry (2015), 125-150. [10] W.R.H. Wright, R. Palkovits, ChemSusChem, 5 (2012) 1657-1667. [11] P.P. Upare, J.M. Lee, D.W. Hwang, S.B. Halligudi, Y.K. Hwang, J.S. Chang, J. Ind. Eng. Chem., 17 (2011) 287-292. [12] P. Sabatier, A. Mailhe, Ann. Chin. Physi., 8 (1909). [13] R.V. Christian, H.D. Brown, R.M. Hixon, J. Am. Chem. Soc., 69 (1947) 1961-1963. [14] H. A. Schuette, R.W. Thomas, J. Am. Chem. Soc., 52 (1930) 3010 - 30121. [15] J. Zhang, A Study on Hydrogenation of Levulinic Acid to gamma-Valerolactone in different solvents, University of Chemical Technology, University of Chemical Technology, Beijing, 2010. [16] W. Luo, U. Deka, A.M. Beale, E.R.H. van Eck, P.C.A. Bruijnincx, B.M. Weckhuysen, J. Catal., 301 (2013) 175-186. [17] M.G. Al-Shaal, W.R.H. Wright, R. Palkovits, Green Chem., 14 (2012) 1260-1263. [18] L.E. Manzer, K.W. Hutcherson, (E. I. Du Pont De Nemours & Co., USA), Production of 5methyl-dihydro-furan-2-one from Levulinic Acid in Supercritical Media 2004. [19] R.A. Bourne, J.G. Stevens, J. Ke, M. Poliakoff, Chem. Commun., (2007) 4632-4634. [20] J.C. Serrano-Ruiz, R.M. West, J.A. Durnesic, Catalytic Conversion of Renewable Biomass Resources to Fuels and Chemicals, in: J.M. Prausnitz, M.F. Doherty, R.A. Segalman (Eds.) Annual Review of Chemical and Biomolecular Engineering, Vol 1, Annual Reviews, Palo Alto, (2010), 79-100. [21] K. Yan, C. Jarvis, T. Lafleur, Y. Qiao, X. Xie, RSC Adv., 3 (2013) 25865-25871.

29 Page 31 of 32

Ac ce p

te

d

M

an

us

cr

ip t

[22] K. Yan, T. Lafleur, G. Wu, J. Liao, C. Ceng, X. Xie, Appl. Catal. A, 468 (2013) 52-58. [23] A. P. Dunlop, J.W. Madden, (Quaker Oats Co.), Process of Preparing gammaValerolactone, US 2786852, 1957. [24] Z. Yang, Y.-B. Huang, Q.-X. Guo, Y. Fu, Chem. Commun., 49 (2013) 5328-5330. [25] X. Du, Y. Liu, J. Wang, Y. Cao, K. Fan, Chinese J. Catal., 34 (2013) 993-1001. [26] K.-I. Shimizu, S. Kanno, K. Kon, Green Chem., 16 (2014) 3899-3903. [27] A.M. Hengne, C.V. Rode, Green Chem., 14 (2012) 1064-1072. [28] V. Mohan, C. Raghavendra, C.V. Pramod, B.D. Raju, K.S. Rama Rao, RSC Adv., 4 (2014) 9660-9668. [29] V. Mohan, V. Venkateshwarlu, C.V. Pramod, B.D. Raju, K.S.R. Rao, Catal. Sci. Tech., 4 (2014) 1253-1259. [30] I. Obregón, E. Corro, U. Izquierdo, J. Requies, P.L. Arias, Chinese J. Catal., 35 (2014) 656662. [31] J.M. Tukacs, R.V. Jones, F. Darvas, G. Dibo, G. Lezsak, L.T. Mika, RSC Adv., 3 (2013) 16283-16287. [32] M. Høj, K. Linde, T.K. Hansen, M. Brorson, A.D. Jensen, J.-D. Grunwaldt, Appl. Catal. A, 397 (2011) 201-208. [33] G. Bergeret, P. Gallezot, Particle Size and Dispersion Measurements, in: G. Ertl, H. Knözinger, F. Schuth, J. Weitkamp (Eds.) Handbook of Heterogeneous Catalysis, Wiley-VCH Verlag GmbH & Co. KGaA (2008), 738-765. [34] B. Ravel, M. Newville, J. Synchrotron Radiat., 12 (2005) 537-541. [35] J.J. Rehr, R.C. Albers, Rev. Mod. Phys., 72 (2000) 621. [36] J.J.K. J.J. Rehr, M.P. Prange, A.P. Sorini, Y. Takimoto, F.D. Vila, C. R. Phys., 10 (2009) 548-559. [37] S. Calvin, E.E. Carpenter, B. Ravel, V.G. Harris, S.A. Morrison, Phys. Rev., 66 (2002) 224405. [38] Z. Boukha, C. Jiménez-González, B. de Rivas, J.R. González-Velasco, J.I. Gutiérrez-Ortiz, R. López-Fonseca, Appl. Catal. B, 158–159 (2014) 190-201. [39] D.A.J.M. Ligthart, J.A.Z. Pieterse, E.J.M. Hensen, Appl. Catal. A, 405 (2011) 108-119. [40] N. Salhi, A. Boulahouache, C. Petit, A. Kiennemann, C. Rabia, Int. J. Hydrogen Energy, 36 (2011) 11433-11439. [41] R. López-Fonseca, C. Jiménez-González, B. de Rivas, J.I. Gutiérrez-Ortiz, Appl. Catal. A, 437–438 (2012) 53-62. [42] J.-D. Grunwaldt, B. Kimmerle, S. Hannemann, A. Baiker, P. Boye, C.G. Schroer, J. Mater. Chem., 17 (2007) 2603-2606. [43] M. Meng, P. Lin, Y. Fu, Spectrosc. Lett., 34 (2001) 83-92. [44] K. Kon, W. Onodera, K.-i. Shimizu, Catal. Sci. Tech., 4 (2014) 3227-3234. [45] L. Qi, I.T. Horvath, ACS Catal., 2 (2012) 2247-2249. [46] S.G. Wettstein, D.M. Alonso, Y.X. Chong, J.A. Dumesic, Energy Environ. Sci., 5 (2012) 8199-8203. [47] R. Molina, G. Poncelet, J. Catal., 199 (2001) 162-170.

30 Page 32 of 32