Supported Intermetallic PdZn Nanoparticles as Bifunctional Catalysts for the Direct Synthesis of Dimethyl Ether from CO‐Rich Synthesis Gas

Abstract The single‐step syngas‐to‐dimethyl ether (STD) process entails economic and technical advantages over the current industrial two‐step process. Pd/ZnO‐based catalysts have recently emerged as interesting alternatives to currently used Cu/ZnO/Al2O3 catalysts, but the nature of the active site(s), the reaction mechanism, and the role of Pd and ZnO in the solid catalyst are not well established. Now, Zn‐stabilized Pd colloids with a size of 2 nm served as the key building blocks for the methanol active component in bifunctional Pd/ZnO‐γ‐Al2O3 catalysts. The catalysts were characterized by combining high‐pressure operando X‐ray absorption spectroscopy and DFT calculations. The enhanced stability, longevity, and high dimethyl ether selectivity observed makes Pd/ZnO‐γ‐Al2O3 an effective alternative system for the STD process compared to Cu/ZnO/γ‐Al2O3.


Table of Contents
Experimental procedures 3 M a t e r i a l s 3 Preparation of nanoparticle precursors 3 Preparation of bifunctional STD catalysts 3 C h a r a c t e r i z a t i o n 3 C a t a l y t i c t e s t i n g 4 In situ and operando X A S e x p e r i m e n t s 4 D F T c a l c u l a t i o n s 5

Results and Discussions 6
Elemental compositions and specific surface areas (Table S1) 6 CO conversions and C-based product selectivities in the direct synthesis of DME ( Table S2) 6 Results of EXAFS analysis (Table S3) 6 Segregation energies for Figure S8 compared to the ordered intermetallic L1 0 structure (Table S4) 7 TEM images of calcined and spent Pd/Zn-based STD catalysts ( Figure S1) 8 Carbon-based product selectivities in the STD process for Pd/Zn (8) Characterization: For analysis of the NPs, a transmission electron microscope (TEM, FEI Tecnai F20 ST TEM, operating voltage 200 kV, equipped with a field emission gun and an EDAX EDS X-ray spectrometer (Si(Li) detecting unit, super ultra-thin window, active area 30 mm 2 , resolution 135 eV (at 5.9 keV)) was used. Additionally, an aberration (image) corrected FEI Titan 80-300 (operated at 300 kV) equipped with an EDAX s-UTW EDX detector was employed for high resolution transmission electron microscopy (HRTEM) images of the used catalyst. For sample preparation, a droplet of the dispersed sample was deposited on amorphous carbon-coated Cu grids (400 mesh) and typically dried in air. For HRTEM of the spent catalysts, the samples were handled in an anoxic glove box and transferred to the TEM using a Gatan 648 vacuum transfer holder to avoid oxidation of the catalyst particles. The mean particle diameter was calculated based on TEM images by measuring the size of at least 100 particles. The analysis of the catalyst composition was carried out with a DSM 982 Gemini SEM (Zeiss corp., Germany, equipped with a Schottky-type thermal field emission cathode) coupled with EDX.
Powder x-ray diffraction (XRD) was carried out with a PANalytical X'Pert Pro X-ray diffractometer (Bragg-Brentano geometry with Cu Kα radiation and a Ni filter). The range between 5° and 120° was measured within 1 h. The diffraction patterns were compared to reference compounds from the Joint Committee of Powder Diffraction Standards (JCPDS) database. Reflections of the calcined and spent catalysts were assigned to the following JCPDS references: PdO (JCPDS 00-043-1024), Pd (JCPDS 01-087-0638), ZnO (JCPDS 01-089-0510) and -Al 2 O 3 (JCPDS 00-010-0425) and the L1 0 PdZn phase (JCPDS 00-006-0620) (Fig. 1b). Crystallite sizes of the PdZn phase (L hkl ) were calculated for the reflection at 41.2° (2θ) (i.e. (111)) using the Scherrer equation where L hkl is the particle size perpendicular to the (hkl) crystal plane,  the full width at half maximum,  the Bragg angle, and  (1.541 Å) the wavelength of the incident X-rays. LaB 6 (NIST) was used as a standard to account for the contribution of the instrumental broadening. To record XRD patterns of the used catalysts after the STD reaction, the samples were removed from the reactor in the anoxic glove box and deposited on the XRD sample holder which was eventually covered by Kapton foil. The broad peak at low Bragg angles (approx. 12-30° range) in the diffractograms of the spent catalysts (i.e., Pd/Zn (8) Temperature-programmed ammonia desorption (NH 3 -TPD) was conducted in an Autochem 2950 device (Micromeritics). Prior to these experiments, the bifunctional catalysts were reduced in 2 vol.-% H 2 in Ar at 250 °C (heating rate 2 °C/min). Ammonia (1.221 ± 0.024 vol.-% NH 3 in He, 30 mL/min) was subsequently loaded after cooling to 100 °C. Physisorbed ammonia was initially removed in a He flow (30 mL/min) at 120 °C, and then ammonia desorption was confirmed from 120 to 750 °C (heating rate 4 °C/min) in He (30 mL/min) using a Cirrus 2 mass spectrometer (MKS).

Catalytic testing:
The performance of the bifunctional catalysts was evaluated using a continuously operated laboratory scale plant with a plug flow reactor. Inlet gases (obtained from Air Liquide: CO (N37), H 2 (6.0), Ar(6.0), N 2 (6.0)) were dosed using individual mass flow controllers (MFC, Bronkhorst). The reaction mixture was diluted with inert gas (70 vol.-% Ar/N 2 ) to avoid the formation of hotspots in the reactor. Analysis of outlet gases was carried out with a gas chromatograph (GC, Hewlett Packard 6890) equipped with two columns (RESTEK RT®-U-Bond, RESTEK RT-M sieve) and two detectors (thermal conductivity detector, flame ionization detector). 2 g calcined catalyst were placed in the fixed bed reactor between two layers of silicon carbide followed by in situ activation (activation procedure: 1) heating (17 °C/h) in 2 vol.-% H 2 in Ar (100 mL NTP /min) to 200 °C which was maintained for 1 h, 2) heating (17 °C/h) to 240 °C which was maintained for 1 h, and 3) heating (10 °C/h) in pure H 2 (50 mL NTP /min) to 250 °C which was maintained for 1 h). After applying the reaction pressure (50 bar), the reactant gases were fed into the reactor with a total flow of 50 mL NTP /min and a composition of Ar : N 2 : H 2 : CO = 5 : 2 : 1.5 : 1.5. The temperature was increased every 6 h by 10 °C within 1h until the final temperature of 300 °C was reached. A maximum deviation of the reaction temperature by ± 0.2 K over the entire length of the catalyst bed was ensured by monitoring the temperature profile with a thermocouple in a guiding sleeve in the middle of the reactor. Stability with time on stream (TOS) was tested at 250°C over a period of 270 h and 260 h for Pd/Zn (8)--Al 2 O 3 and for the Cu/ZnO/-Al 2 O 3 -based reference catalyst, respectively. CO conversion and product selectivities were evaluated according to the following equations: where is the molar CO flow at the reactor inlet, the molar CO flow at reactor outlet and X CO the CO conversion (%).
where S i is the selectivity of component i, and and molar flow of component i at reactor outlet and at reactor inlet, respectively, and ξ i is the number of carbon atoms in a molecule of component i.

In situ and operando XAS experiments:
In situ X-ray absorption spectra (XAS) in terms of the X-ray absorption near-edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) were recorded at the CAT-ACT beamline at the Synchrotron Radiation Source at KIT (Karlsruhe) in fluorescence mode using a PIPS diode. [1] The X-ray energy was varied using a double crystal monochromator with a pair of Si (111) and Si (311) crystals to scan energies around Zn K and Pd K edges, respectively. Higher harmonics were rejected by a pair of Si or Pt/Ir coated mirrors for Zn K or Pd K edges. The incident beam was focused with the second mirror but no slits were available to shape the beam in front of the first ionization chamber. Due to this and the long optical path between the optics hutch and the experimental hutch, the resulting beam size on the sample at 24350 eV could not be precisely controlled and could exceed 1 mm x 1 mm (vertical x horizontal) and at Pd K edge approach 1.5 mm x 3 mm. The catalyst (approx. 6 mg, 80-160 μm sieve fraction) was placed in an in situ microreactor (quartz capillary with a plug flow reactor geometry, 1.5 mm diameter, 20 μm wall thickness, catalyst bed length approx. 5 mm) between two quartz wool plugs and mounted horizontally over a hot air blower (FMB Oxford GSB-1300). [2] The temperature was calibrated using readings of an external thermocouple at the top and the bottom of the capillary microreactor in a separate test without X-ray beam. He or 5 vol.% H 2 /He were dosed using individual mass flow controllers at a flow rate of 50 mL NTP /min, and the capillary microreactor was pressurized up to 20 bar by means of a Back Pressure Regulator (BPR, Bronkhorst). First, XANES and EXAFS spectra were recorded at 25 °C in He flow, afterwards the gas mixture was switched to 5 % H 2 /He (50 mL NTP /min, ambient pressure) and a temperature programmed reduction (TPR) was performed. For this purpose, XANES spectra were recorded continuously during heating to 250 °C at a ramp rate of 1 °C/min. After reaching 250 °C, the catalyst was kept for 60 min in 5% H 2 /He, after which another set of XANES and EXAFS spectra were recorded (5 % H 2 /He, 250 °C). The reduction experiments were repeated twice while recording either spectra at Pd K edge (the sieved catalyst was used as received) or at Zn K edge (the catalyst was diluted by mechanically mixing with -Al 2 O 3 (catalyst/-Al 2 O 3 ratio 1:1) prior to pressing and sieving to diminish the incident beam self-absorption effect). After the TPR experiment at the Pd K edge, the reaction conditions of the direct DME synthesis were applied to the in situ microreactor (15 vol.-% H 2 , 15 vol.-% CO in He, 250 °C, 18 mL NTP /min, 20 bar) and maintained for one hour. The product formation was qualitatively monitored thanks to the elevated pressure by a Pfeiffer Vacuum OmniStar GSD320 mass spectrometer (m/z 45 and 31 were observed for DME and methanol, respectively). The possibility to actually observe the reaction products and to prove that the catalyst actually produced DME at the same time makes the study the first operando (i.e., performed on a working material) [3] study of DME catalysts. During in situ low pressure (1 bar) measurements hardly any methanol or DME could be detected (Fig. S6c); the signals of the products appeared only after pressurizing to 5 bar with methanol prevailing over DME (we neglect difference in ionization probabilities due to similar chemical nature of methanol and DME). Only after applying 20 bar, DME became the main reaction product as demonstrated by the MS data. Thus, performing measurements at high pressure is a prerequisite for an operando study of methanol or DME synthesis catalysts.
During the test under reaction conditions, XANES spectra were continuously recorded, an additional EXAFS spectrum at the Pd K edge was recorded after 1 h TOS when concentrations of DME and methanol were stabilized. The spectra were normalized, and background subtracted to extract EXAFS using the ATHENA program from the IFFEFIT software package. [4] The spectra of the catalyst measured at Zn K edge were corrected for incident beam self-absorption using the Fluo algorithm built in the ATHENA program. [4] The k 1 -, k 2 -, and k 3 -weighted EXAFS functions were Fourier transformed in the k range of 3.0 -11 Å −1 and multiplied by a Hanning window with sill size of 1 Å -1 (not corrected for phase shift). The structural models were based on bulk Pd metal (Inorganic Crystal Structure Database, ICSD collection code 52251) and PdZn alloy (ICSD collection code 105752). The coordination numbers and interatomic distances were received by fitting one coordination shell for the Pd-Zn neighbors and a second coordination shell for the Pd-Pd neighbors. The structure refinement was performed using ARTEMIS software (IFFEFIT). [4] For this purpose, the corresponding theoretical backscattering amplitudes and phases were calculated by FEFF 6.0. [5] The theoretical data were then adjusted to the experimental spectra by a least square method in R space between 1.5 and 2.9 Å -1 . First, the amplitude reduction factors (S 0 2 = 0.58) were calculated using the Pd foil reference spectrum, and then the coordination numbers (CN), interatomic distances (d), energy shift (E 0 ) and mean square deviation of interatomic distances ( 2 ) were refined. The absolute misfit between theory and experiment was expressed by . Fitting was done in iterative way several times constraining CNs of Zn or Pd until reaching the lowest . The number of fitting parameters was always lower than the number of independent points.
The fractions of ZnO and Zn 0 species in the in situ XANES spectra were determined by a Linear Combination Analysis (LCA) using spectra of the calcined catalyst corrected for the incident beam self-absorption (an ex situ spectrum of the calcined sample recorded in transmission mode confirmed ZnO structure) and Zn foil reference compounds in the fitting range 9652 -9702 eV. No additional constraints were used / required to achieve the linear combination fit. R-factor obtained during the fit was 0.0004 (0.04%), and the error bars were ±0.6%. On one hand, the obtained fit is rather good implying that Zn foil spectrum can be used to represent the spectrum of reduced Zn species, however, on the other hand, it has been shown that XANES spectra of bulk metals do not necessarily represent well spectra of the nanoparticles. [6] This is especially true if intermetallic compound is formed. Hence, we cannot be completely sure that Zn foil spectrum represents the reduced Zn species well and, therefore, in the text we use very conservative ±10% error bars for the LCA.
In addition, LCA was also performed on the Pd K edge XANES spectra to evaluate relative spectral changes during the reductive activation of the catalyst. The spectral changes were very small (Fig. 3a). As no spectra of reference compounds could be used, the first spectrum before the heating (in H 2 /He at 20 °C) and the spectrum of the activated catalyst were used as internal references instead.
DFT calculations: DFT calculations have been performed using the Vienna Ab Initio Simulation Package (VASP) [7] in connection with the Atomic Simulation Environment (ASE). [8] The Bayesian Error Estimation Functional with van der Waals correlations (BEEF-vdW), [9] that is performing well for adsorption energies and transition states on transition metal surfaces [10] has been used. The planewave cutoff was 500 eV. The (111) facet of PdZn surfaces in the L1 0 structure was modeled by periodic slabs consisting of 2 x 2 atoms in the x and y direction and four layers in the z-direction with 16 Å of vacuum between the slabs. Only the top two layers were allowed to relax while the other two layers where kept at their bulk position. The Brillouin zone was sampled using a 6 x 6 x 1 Monkhorst-Pack k-point grid. [11] Transition state searches were performed using the nudged elastic band (NEB) method. [12] Results and Discussion The total specific surface areas of 154 m 2 /g (Pd/Zn (8) (14)-γ-Al 2 O 3 ) of the bifunctional catalysts, which amounts to 111 m²/g and 70.5 m²/g, respectively. Hence, the difference in specific surface area is ascribed to the contribution of the supported Pd/Zn-based NPs. For γ-Al 2 O 3 , the weak to medium acidic sites are expected to exhibit a superior dehydration activity compared to strong acidic sites. [13] Table S2. CO conversion and carbon-based product selectivities in the direct synthesis of DME for the Pd/Zn catalysts (Reaction conditions: temperature range 250 °C to 300 °C, 50 bar, feed composition H 2 : CO : Ar : N 2 = 3 : 3 : 10 : 4, total flow 50 ml NTP /min) X Eq Equilibrium CO conversion under the chosen reaction conditions according to the reation equation: 3 CO + 3 H 2 ⇌ CH 3 OCH 3 + CO 2

Pd/Zn (8)-γ-Al 2 O 3
Reaction temperature (°C)  Table S3. Results of the EXAFS analysis of the spectra of the reduced and reacted Pd/Zn (8)-γ-Al 2 O 3 (fit in R space between 1.5 and 3.0 Å -1 , S 0 2 = 0.58).  Table S4. Segregation energies for the systems given in Figure S6 compared to the ordered intermetallic L1 0 structure ( Figure S6a). CO and OH induced segregation energies are also given along with CO and OH binding energies relative to CO and H 2 O-1/2H 2 gas phase, respectively.  In general, catalyst deactivation of conventional Cu/ZnO-based methanol catalysts has been ascribed to multiple mechanisms such as carbon deposition, [14] sintering, [15] and the formation of a dense ZnO overlayer on the Cu nanoparticles. [16] A Cu/Zn-based catalyst of the conventional methanol catalyst composition was prepared via a similar colloidal approach and used as a reference in the STD reaction. The catalytic tests revealed a decrease in CO conversion from originally 19% to 9% after testing for 260 h TOS. Copper has a low Hüttig temperature which is reflected by a relatively low melting point (i.e., bulk Cu 1083 °C), and therefore, Cu catalysts are operated typically at relatively low temperatures (240 to 260 °C). However, our Cu/ZnO-based reference catalyst revealed a constant deactivation even at 250°C with a decrease of 53% in CO conversion during the first 260 h TOS ( Figure S3).  Internal references (the first and the last spectra measured during the activation) are used for the Pd K edge data, ZnO and Zn 0 (Zn foil) spectra are used for the Zn K edge data.