Structure and Role of a Ga-Promoter in Ni-Based Catalysts for the Selective Hydrogenation of CO2 to Methanol

Supported, bimetallic catalysts have shown great promise for the selective hydrogenation of CO2 to methanol. In this study, we decipher the catalytically active structure of Ni–Ga-based catalysts. To this end, model Ni–Ga-based catalysts, with varying Ni:Ga ratios, were prepared by a surface organometallic chemistry approach. In situ differential pair distribution function (d-PDF) analysis revealed that catalyst activation in H2 leads to the formation of nanoparticles based on a Ni–Ga face-centered cubic (fcc) alloy along with a small quantity of GaOx. Structure refinements of the d-PDF data enabled us to determine the amount of both alloyed Ga and GaOx species. In situ X-ray absorption spectroscopy experiments confirmed the presence of alloyed Ga and GaOx and indicated that alloying with Ga affects the electronic structure of metallic Ni (viz., Niδ−). Both the Ni:Ga ratio in the alloy and the quantity of GaOx are found to minimize methanation and to determine the methanol formation rate and the resulting methanol selectivity. The highest formation rate and methanol selectivity are found for a Ni–Ga alloy having a Ni:Ga ratio of ∼75:25 along with a small quantity of oxidized Ga species (0.14 molNi–1). Furthermore, operando infrared spectroscopy experiments indicate that GaOx species play a role in the stabilization of formate surface intermediates, which are subsequently further hydrogenated to methoxy species and ultimately to methanol. Notably, operando XAS shows that alloying between Ni and Ga is maintained under reaction conditions and is key to attaining a high methanol selectivity (by minimizing CO and CH4 formation), while oxidized Ga species enhance the methanol formation rate.


CO hydrogenation tests
In addition to the NixGa(100-x)/SiO2, we also tested 100 mg of commercial Cu/ZnO/Al2O3 methanol synthesis catalyst (Alfa Aesar, #45776-36).For this test, the catalyst was loaded into the reactor under ambient air conditions and the in-situ catalyst activation was done by heating to 300°C (10°/min) in 1 bar 5% H2/N2 (50 ml/min) and holding at 300°C for 1h.All catalytic performance results from this work, as well as selected results from literature, can be found in Tables S1 and S2.
The catalytic performance parameters were calculated as follows: Total off-gas flow rate (mol min -1 ): ) Molar formation rate of product i per mol of (Ni+Ga) in the catalyst (mmoli mol(Ni+Ga) -1 s -1 ):

Selectivity of product i (%):
= where  , (mol min -1 ) is the total gas flow rate at the inlet of the reactor,  , (-) is the concentration of N2 in the inlet gas,  , (-) is the concentration of N2 in the off-gas,   (-) is the concentration of product i ∈{CH3OH, CO, CH4} in the off-gas, MWi (g mol -1 ) is the molecular weight of species i, mcat (g) is the catalyst mass,   (-) and   (-) are the weight fractions of Ni and Ga in the catalyst as measured by ICP, and  , (-) is the concentration of CO2 in the inlet gas.

S12
Table S1.a) wNi + wGa + wSiO2 = 100 wt% and ni denote the number of moles of species i. (c) Mean values over the first 180 minutes TOS (= 5 GC samples).
The value in parentheses denotes the uncertainty of the least significant digit of the mean.

S13
Table S2.Selected Ni, Cu, and Pd based catalysts used for the CO2 hydrogenation to methanol.b) In this study, only the CO-free selectivity was reported.

XAS and X-ray total scattering experiments
In situ/ operando XAS and X-ray total scattering experiments were performed at the beamlines BM31 and ID15A of the European Synchrotron Radiation Facility, respectively.The setup in both cases consisted of a custom-made capillary cell reactor which was connected to a manifold of mass flow controllers (Bronkhorst EL-FLOW series, max.30 ml/min, Pmax = 35 bar) and a backpressure regulator (Bronkhorst, EL-PRESS series, Pmax = 35 bar).The catalyst bed consisted of ca.2-3 mg of catalyst placed between two quartz wool plugs inside a quartz capillary (Hilgenberg, 1 mm OD, 0.02 mm wall thickness).The capillary was heated via a hot air blower from below.The temperature was carefully calibrated between 50 -800 °C prior to the experiments by placing a thermocouple in a capillary filled with quartz wool.The off-gas was analyzed via a compact gas chromatograph (Global Analyzer Solutions, Compact GC 4.0 ) equipped with TCD and FID detectors and a sampling rate of ca.1/7 min -1 .The limits of detection of the GC are approximately < 10 ppm for methanol and CH4 and < 500 ppm for CO.A typical in situ/operando experiment consisted of in situ activation, pressurizing and CO2 hydrogenation steps which are detailed in Figure S18 C and D for XAS and total scattering experiments, respectively.The difference between the two sets of experiments is that in the XAS experiments, the catalyst was cooled down to 50°C after in situ activation to collect EXAFS data, whereas in the total scattering experiments the catalyst was only cooled down to 230°C.XAS data were measured in transmission mode, via ion chamber detectors placed before and after the sample (Figure S18 A) and X-ray total scattering measurements were collected via an area detector (Pilatus3 CdTe 2M) placed behind the sample (Figure S18 B).

Collection of X-ray total scattering data and pair distribution function (PDF) analysis
X-ray total scattering data of Ni65Ga35/SiO2, Ni70Ga30/SiO2, Ni75Ga25/SiO2, and Ni100/SiO2 were collected continuously at an incident X-ray energy of 90.0 keV (0.138 Å) up to Qmax,instr = 30 Å -1 at a rate of 1 measurement/2.62minutes.For Ni65Ga35/SiO2-XAS, total scattering data was collected at 68.5 keV (0.181 Å) up to Qmax,instr = 26 Å -1 at a rate of 1 measurement/4.65 minutes.Total scattering data of the pristine silica support was measured under in situ activation and reaction conditions and used as background to calculate the d-PDF data of the NixGa(100-x)/SiO2 (see Figure S22).In addition, total scattering data of the CeO2 NIST reference materials were obtained to determine the experimental resolution parameters Qdamp and Qbroad.
To obtain the differential PDF (d-PDF) data from the total scattering patterns of NixGa(100-x), we performed background subtraction, data normalization and Fourier transform using the PDFgetX3 software (v 2.2.1). 7,8The total scattering data was processed within the range Qmin = 1.5 Å -1 and Qmax = 23 Å -1 with rpoly = 1.0, which is approximately the r-limit for the maximum frequency in the F(Q) correction polynomial.
Modelling of the d-PDF was performed in PDFGui (v 1.0), 9,10 using a random alloy model structure generated from fcc-Ni (Inorganic Crystal Structure Database, ICSD # 8688) in which x % Ni atoms (x as in NixGa(100-x)) were substituted by Ga.The fitted parameters included the scale factor, the cubic lattice parameter, an isotropic atomic displacement factor (same for Ni and Ga), the "delta2" atomic motion correlation factor, and the coherent particle (crystallite) size.The Qdamp and Qbroad parameters were set to the values obtained from the fitting of the CeO2 reference measurements (i.e.Qdamp= 0.0180 Å -1 and Qbroad= 0.0159 Å -1 ).We note here that we also tested modelling the d-PDF using a α'-Ni3Ga model structure (Pm-3m space group, Ni at Wyckoff position 3c with atomic coordinates 0,0.5,0.5 and Ga at Wyckoff position 1a and coordinates 0,0,0, see ICSD #103856) which resulted in fitted parameters equal (within the error) to the ones obtained with the random alloy.The d-PDF were fitted between 1.7 -25 Å (respective fit and lattice parameters are shown in Figures 2 B and 2 D) however, boxcar fittings between 2-7, 5-10, 8-13, and 11-16 Å were also performed to evaluate whether the structures in the different r ranges of the d-PDF are comparable.
The Ni:Ga ratios in the alloys (Tables 1 and S5) were obtained using the refined cell parameters obtained from the fitting of the d-PDF data and the slope of Vegard's law.The validity of Vegard's law (linear correlation between the lattice parameter and the Ni:Ga ratio) was verified for fcc NiyGa(100-y) alloys by using reported literature values (Table S4, data collected at room temperature, 25°C).After extracting the slope of Vegard's law, the obtained value was applied to the data collected for Ni100/SiO2 at 230°C.Note that this approach also accounts for potential differences in the cell parameters between bulk Ni and nanostructure Ni (see Figure S29).The Gaalloyed and GaOx amounts normalized by the Ni contents of the catalysts reported in Table 1 were obtained via the following equations, considering that all Ni in the catalysts is alloyed and that GaOx (see XAS analysis) is the difference between the total (determined by ICP) and alloyed Ga contents: ) is the Ga:Ni ratio in the alloy obtained from the d-PDF analysis, and is the ratio between the total amounts of Ga and Ni determined by ICP (where we assume   =  , ).

XAS measurements 3.2.1. Collection of XAS data
Ni and Ga K-edge XAS scans were collected consecutively, using XANES and EXAFS macros, covering the energy ranges specified below.The data collected during the in situ activation step (heating up from room temperature to 600°C in 1 bar H2) were collected using the XANES macro, whereas the EXAFS macro was used for isothermal measurements.

XAS data analysis
The energy scale at the Ni and Ga K-edges spectra were calibrated by setting the absorption edge positions (set at the maximum of the first derivative of µ(E)) of the reference samples Ni-foil and Zn foil to the known values of 8333.0 eV and 9659.0 eV respectively.
Linear combination fittings of the normalized XANES (μ(E)) were performed using the Athena/Demeter v 0.9.26 software 11 between -20 and +50 eV around the edge position ("E0"), constraining the LCF weights to values between 0 and 1 and their sum to 1.
In the EXAFS fittings we included the first metal-metal spheres, as well as a Ga-O sphere.One limitation of EXAFS is the difficulty in distinguishing Ga and Ni scattering owing to their similar atomic numbers 12 and similar interatomic distances.However, by fitting simultaneously the Ni and Ga K-edge data of the same catalyst (see below) and applying a set of constraints to the fitting parameters, we aimed to distinguished between the overlapping Ni-Ni, Ni-Ga and Ga-Ga spheres (combined fitting).Additionally, we performed a fitting in which we did not apply any constraints between the fitting parameters of the Ni and Ga K-edge data and only distinguished between Ni-M and Ga-M spheres (where M = Ni,Ga) (independent fitting).We performed EXAFS fittings of the catalysts Ni65Ga35/SiO2-XAS and Ni75Ga25/SiO2-XAS after in situ activation (and after cooling down to 50 °C in 1 bar H2), of reacted Ni65Ga35/SiO2-XAS after ca.4h TOS under CO2 hydrogenation conditions (after cooling down to 50 °C in 1 bar CO2:N2:H2 = 1:1:3), and of ex-situ activated Ni100/SiO2 (room temperature).

Constraints applied to the fitted radial distances (r), coordination numbers (CN) and Debye
Waller factors (σ 2 ) of the combined fitting: , where α = CGa/CNi which is the Ga:Ni (atomic) ratio estimated from PDF analysis (iv) The constraints listed above assume that Ni and Ga are part of a random fcc-NiyGa(100y) alloy [where y:(100-y) is the Ni:Ga ratio in the alloy] containing atomic Ni and Ga fractions of CNi and CGa, respectively.A similar approach has been used in the literature to fit EXAFS of ceria-based solid solutions. 13This approach assumes that Ni and Ga are equally coordinated (CN(Ni-Ni) + CN(Ni-Ga) = CN(Ga-Ni) + CN(Ga-Ga)) and that, on average, a fraction of CGa atoms in the (Ni, Ga) alloy are occupied by Ga.
Continuous Cauchy Wavelet Transform (CCWT) analysis of the k 2 weighted Ga EXAFS were performed using a freely available MATLAB script by Muñoz, Argoul and Farges. 14The same kranges as reported above for the EXAFS fittings were used and the CCWT was calculated in the range 0.5 Å -4 Å (Δr = 0.0175 Å).The Cauchy order of the WT was 50.Table S3.Parameters obtained from fitting the d-PDF of in situ activated NixGa(100-x)/SiO2 (data acquired at 230°C, Figure 2B) to a random fcc Ni-Ga alloy structure.All of the d-PDF were normalized by a factor of 10 and thus only the ratios of errors are significant but not the absolute values as they scale with the d-PDF normalization.a) Atomic motion correlation factor denoted as delta2 in PDFGui was used. 9Qdamp and Qbroad were obtained from the fitting of a CeO2 reference material.Figure S26.Off-gas composition measured via a GC during the operando X-ray total scattering -CO2 hydrogenation experiment using Ni65Ga35/SiO2-XAS (PDF data shown in Figure S25).Conditions: 20 bar CO2:N2:H2 = 1:1:3, 230°C.Table S4.Literature data of cell parameters for Ni and Ni-Ga fcc alloys used to derive the slope of Vegard's law and available in the ICSD.

S35
Table S7.EXAFS fitting results of the reference materials used to determine the amplitude reduction factor 0 2 at the Ni (Ni-foil) and Ga K-edge (α'-Ni3Ga and β-Ga2O3).The errors on the reported values as estimated by the software Artemis are given in compact (crystallographic) notation, i.e., 2.52(1) corresponds to 2.52±0.01.Parameters without errors were fixed.Abbreviations: Rw: weighted R factor, ∆0: the edge energy shift, CN: coordination number, σ 2 : Debye-Waller factor, r: interatomic distance.Ga-Otet and Ga-Ooct denote Ga which is tetrahedrally and octahedrally coordinated by O.

S36
Table S8.Independent EXAFS fitting results.a) The Debye-Waller factor of Ga-O was set to the value obtained from β-Ga2O3 for tetrahedrally coordinated Ga-O (σ 2 (Ga-Otet)).M = Ni/Ga for Ni65Ga35/SiO2 and Ni75Ga25/SiO2, and M = Ni for Ni100/SiO2 Table S9.Combined Ni and Ga K-edge EXAFS fitting results.Notation of the errors and abbreviations are as in Table S7.a) The Debye-Waller factor of Ga-O was set to the value obtained from β-Ga2O3 for tetrahedrally coordinated Ga-O (σ 2 (Ga-Otet)).

Figure S5 .
Figure S5.Representative HAADF-STEM images of as-prepared Ni75Ga25/SiO2 and the corresponding particle size distribution.

Figure
Figure S11.Representative HAADF-STEM/EDX image of spent Ni70Ga30/SiO2 and the corresponding EDX spectrum.

FigureFigure
Figure S12.Representative HAADF-STEM images of as-prepared Ni65Ga35/SiO2 and the corresponding particle size distribution.

Figure
Figure S15.Representative HAADF-STEM/EDX image of spent Ni65Ga35/SiO2 and the corresponding EDX spectrum.

Figure
Figure S17.(A) Methanol selectivity as a function of CO2 conversion.(B) Methanol formation rate as a function of time-on-stream under CO2 hydrogenation conditions with linear fits (black dotted lines) indicating the deactivation trends.(C) Methanol selectivity as a function of TOS.(D) Slope of the linear curves fitted to the methanol formation rates in (B) as a function of xICP in NixGa(100-x)/SiO2.Conditions: 25 bar CO2:N2:H2 = 1:1:3, GHSV = 60 L h -1 gcat -1 , 230°C.

Figure S18 .
Figure S18.Schematic of the setups for in-situ/operando (A) XAS at BM31 and (B) total scattering experiments at ID15A; (C) and (D) specify the gas and temperature treatment steps during in situ/operando (C) XAS and (D) total scattering experiments where 1. Refers to the activation, 2. To the pressurizing and 3. To the reaction step.

Figure
Figure S19.X-ray total scattering patterns of NixGa(100-x)/SiO2 and SiO2.(A) shows the as-prepared, airexposed state (data collected at 50°C in 1 atm N2) and (B) after in situ activation in H2 (data collected at 230°C in H2).

Figure
Figure S20.D-PDF of Ni65Ga35/SiO2 and Ni100/SiO2.Solid lines show the as-prepared, air-exposed state (data collected at 50°C in 1 atm N2) and (B) after in situ activation in H2 (data collected at 230°C in H2).

Figure
Figure S21.(A) SiO2 subtracted X-ray total scattering patterns and a simulated fcc-Ni75Ga25 alloy pattern shown for reference.The dotted lines mark the positions of the Bragg peaks corresponding to Ni100/SiO2.(B) shows the respective reduced structure functions F(Q).Data were collected at 230 °C in 1 bar H2 after in situ activation.

Figure S22 .
Figure S22.Approach to yield the differential PDF (d-PDF) at the example of Ni100/SiO2.The d-PDF can be calculated using two approaches, i) subtraction in G(r), i.e. subtraction of the PDF of SiO2 from the PDF of Ni100, and by ii) subtraction in I(Q): i.e. subtraction of the total scattering pattern of SiO2 from the total scattering pattern of Ni100, the result of which is then normalized and Fourier transformed to obtain the d-PDF.All of the d-PDFs reported in this work were obtained via the latter approach, as this resulted in d-PDFs with fewer noise ripples in the low r region (see r < 1.2 Å in the d-PDFs).

Figure
Figure S27.Particle (crystallite) size obtained from the d-PDF fitting of activated Ni65Ga35/SiO2, Ni70Ga30/SiO2, Ni75Ga25/SiO2, and Ni100/SiO2, compared to the particle size obtained from TEM analysis of the as-prepared materials.The error bars represent the standard deviation.

Figure
Figure S29.Vegard's law for NiyGa(100-y) bulk alloys (blue dashed curve, F(x)) derived from the cell parameters reported in TableS4.Vegard's law for the NiyGa(100-y) nano alloys (red dashed curve, F'(x)) is obtained by applying a vertical shift of Δa = 0.004 Å to F(x).
Figure S30.Simulated X-ray diffraction patterns of a fcc-Ni3Ga random alloy (top) and the intermetallic α'-Ni3Ga (bottom) (Pm-3m space group) with their respective unit cells.Below the red dotted line, the intensities are magnified by a factor of 8.

Figure S32 .
Figure S32.Ga K-edge XANES of the reference materials used for the LCF analysis.

Figure S35 .
Figure S35.Ni and Ga K-edges XANES of (A and B) Ni65Ga35/SiO2-XAS and (C-D) Ni75Ga25/SiO2-XAS exposed to air during in situ activation (temperature ramp from 50°C to 600°C (10°/min) in 1 bar H2).The arrows show the direction of the changes.

Figure S37 .
Figure S37.EXAFS experimental data and fittings (k and r space) of the reference materials used to determine  0

Figure S40 .
Figure S40.EXAFS experimental data and fittings (k and r space) of in situ activated Ni100/SiO2, measured exsitu (data collected at room temperature, N2 atmosphere).

Figure S44 .
Figure S44.Continuous Cauchy wavelet transform of the k 2 weighted Ga K edge EXAFS of air-exposed (A-B, data collected at 50°C in 1bar H2), and in situ activated (C-D, data collected at 50°C in 1 bar H2) Ni75Ga25/SiO2-XAS.B and D show an inset in the r-region 0.5-2 Å (Ga-O).In Ga-M in A), M stands for Ni, Ga and/or Si.
Table of contents .