Reactive metal-support interaction in the Cu-In2O3 system: intermetallic compound formation and its consequences for CO2-selective methanol steam reforming

ABSTRACT The reactive metal-support interaction in the Cu-In2O3 system and its implications on the CO2 selectivity in methanol steam reforming (MSR) have been assessed using nanosized Cu particles on a powdered cubic In2O3 support. Reduction in hydrogen at 300 °C resulted in the formation of metallic Cu particles on In2O3. This system already represents a highly CO2-selective MSR catalyst with ~93% selectivity, but only 56% methanol conversion and a maximum H2 formation rate of 1.3 µmol gCu−1 s−1. After reduction at 400 °C, the system enters an In2O3-supported intermetallic compound state with Cu2In as the majority phase. Cu2In exhibits markedly different self-activating properties at equally pronounced CO2 selectivities between 92% and 94%. A methanol conversion improvement from roughly 64% to 84% accompanied by an increase in the maximum hydrogen formation rate from 1.8 to 3.8 µmol gCu−1 s−1 has been observed from the first to the fourth consecutive runs. The presented results directly show the prospective properties of a new class of Cu-based intermetallic materials, beneficially combining the MSR properties of the catalyst’s constituents Cu and In2O3. In essence, the results also open up the pathway to in-depth development of potentially CO2-selective bulk intermetallic Cu-In compounds with well-defined stoichiometry in MSR.


Introduction
Methanol steam reforming (MSR) remains one of the most important reactions in hydrogen economy to access large amounts of hydrogen that can subsequently be used as a renewable energy carrier [1]. The most crucial parameter steering the reaction to high H 2 yields is efficient water activation, which is a prerequisite for high CO 2 selectivities [2]. Among many materials that can be used to selectively catalyze this reaction in the desired direction [3], intermetallic compounds have arisen as a promising material class [4][5][6][7][8][9][10]. Pd-based intermetallic compounds have been particularly scrutinized in this respect [11][12][13][14], but Cu-Zr materials have also been a focus of research [2,15,16]. As the common denominator to explain the catalytic selectivity patterns, a bifunctional synergistic action of intermetallic compound and supporting oxide, sharing the sites of methanol and water activation, respectively, has been proposed [11,17]. Different synthesis pathways have been reported to gain access to these materials, including reactive metal-support interaction starting from small metal particles on oxide supports (i.e. reduction in hydrogen), yielding intermetallic compound particles distributed on the oxide support and a defined interface between them [5]. ZnPd-ZnO [12], Ga 2 Pd-Ga 2 O 3 [13] and InPd-In 2 O 3 [11] are prime examples of such systems. A second pathway is related to the in situ decomposition of nanosized or bulk intermetallic compounds in the MSR mixture, yielding both bifunctionally operating interfaces. In this respect, oxidative segregation effects either yield small patches of oxide on the intermetallic compound surface (as observed, for example, for ZnPd/ZnO) [17,18] or a complete corrosive decomposition of the structure into a dispersed metal/oxide system (as prevalent, for example, for Cu 51 Zr 14 ) [15,16].
Explaining the catalytic action of such systems necessarily involves a deep understanding of the individual constituents of the material and, in particular, the oxide phase. Depending on the intrinsic catalytic properties, the oxide could beneficially or detrimentally contribute to the catalytic properties of the entire system. As for the latter, selectivity-spoiling effects by the (inverse) water-gas shift equilibrium (e.g. for Ga 2 O 3 ) could be observed [11]. However, In 2 O 3 is the archetypical beneficially acting oxide entity, as it is not only a highly CO 2 -selective (yet comparably inactive) catalytic material in MSR by itself [19] but has recently also been reported to act as a superior catalyst for the methanol synthesis by CO 2 hydrogenation [20].
As copper is a recurrent metal constituent of many prospective MSR catalysts (despite the ongoing discussion about the nature of the active site, if either strain phenomena or ionic copper species are responsible for the catalytic action) [3], it is straightforward to combine Cu and In 2 O 3 to potentially create a new MSR catalyst with superior activity and selectivity. Our interest in the system Cu-In 2 O 3 is also fueled by recent reports on the beneficial influence of Cu on In 2 O 3 in the formation of methanol by hydrogen/water reduction of CO 2 [21]. Whether In 2 O 3 is able to stabilize the potentially active and selective copper species (thereby suppressing the notorious sintering of Cu particles, for example, deactivating Cu particles on ZnO and significantly decreasing the lifetime of these catalysts) or rather leads to specific intermetallic Cu-In particles supported on In 2 O 3 is in the focus of this article. The Cu-In phase diagram is well known and features a variety of different intermetallic compounds with variable stoichiometries [22]. As the region around 30 at% In exhibits compounds with high melting points, chances are high that single-phase oxidesupported intermetallic compounds can be accessed by reactive metal-support interaction. Additionally, a high temperature Cu 3 In phase, which undergoes eutectoid decomposition upon cooling, was reported to catalyze the oxidative dehydrogenation of methanol to formaldehyde. This proves that the obtained eutectic cast with the overall stoichiometry Cu 3 In is able to activate methanol, even though the resulting phases are not identified by the authors in [23].
This article focuses on two selected research areas, whose comprehension is necessary to develop a knowledge-driven synthesis routine and enable a full understanding of the operation of the Cu-In 2 O 3 system in MSR. First, the reactive metal-support interaction between Cu and In 2 O 3 , that is, the reduction in hydrogen and the corresponding intermetallic compound formation, will be assessed. As the removal of reaction-induced water is a key parameter in the efficient formation of intermetallic compounds, we will perform experiments under recirculating batch and quasi-flowing conditions. This potentially enables us to access different oxide-supported single-phase intermetallic Cu-In compounds, whose catalytic properties can then be assessed. In the second part, we test the properties of these oxide-supported intermetallic Cu-In compound particles in MSR and relate the results to previously tested intermetallic materials, especially systems based on Pd. An integral part of the characterization will be devoted to synchrotron-based in situ X-ray diffraction (XRD) measurements to follow the structural transitions in the course of reactive metal-support interaction.

Preparation of the Cu/c-In 2 O 3 catalyst
Bixbyite-type cubic indium oxide (c-In 2 O 3 , space group Ia 3, a = 10.118 Å [24]; Alfa Aesar, 99.99% metals basis) was suspended in deionized water and an aqueous solution of Cu(II) acetate (Cu(OOCCH 3 ) 2 ·H 2 O; Merck) was added dropwise under stirring. The solvent was subsequently removed under vacuum at 60°C and the resulting turquoise powder calcined in air at 400°C for 2 h, yielding the gray CuO/c-In 2 O 3 catalyst with a nominal loading of 8.0 wt% CuO in the calcined state (corresponding to 6.5 wt% metallic Cu after pre-reduction).

Catalytic measurements
Catalytic testing was performed in a recirculating batch reactor connected to a quadrupole mass spectrometer (QMS) arranged in cross-beam geometry, equipped with a secondary electron multiplier (Balzers QMG 311). This setup is specialized for the measurement of small sample amounts (approximately 100 mg) and conversions (reactor volume = 13.8 ml). The reactor and the sample holder are made of quartz glass and can be heated up to 1100°C in a Linn High Term furnace. The temperature is monitored with a K-type thermocouple (NiCr-Ni).
For the MSR measurements,~28 mbar of a mixture with a ratio of methanol:water = 1:2 (v/v) in the gas phase is prepared to avoid water depletion during the reaction. One MSR cycle consists of three steps: (1) pre-oxidation at 400°C in 1 bar pure O 2 for 1 h (termed O400), (2) prereduction at 300°C in 1 bar pure H 2 for 1 h (termed H300) and (3) MSR reaction. The latter is carried out by the addition of Ar to the MSR mixture (for correction of the signals considering the thermal expansion and the capillary leak to the QMS) and balancing the pressure to 1 bar with He (for improvement of the recirculation efficiency and the thermal conductivity) at 100°C. Starting from this temperature, a heating ramp with 5°C min −1 is applied, and the gas phase composition is continuously monitored by the QMS.
The formation rates are obtained by differentiation of the partial pressure versus reaction time graphs, which is acquired by the application of the calibration. Then, the conversion to μmol using the ideal gas law and normalization to the total copper mass determined by differential pulse voltammetry (DPP; see sections 2.4 and 3.1) yields the formation rates in μmol gCu −1 s −1 .
The apparent activation energy for the CO 2 evolution (E a ) was calculated by Arrhenius fitting of the formation rate versus the temperature plot at the beginning of the rate increase, thus, excluding influences of the reaction products on the equilibrium. To enhance the comparability of the apparent activation energy between different catalytic materials, the pre-exponential factor A is fixed to a value of 10 8 µmol g Cu −1 s −1 to avoid any influence of statistical fluctuations on the fitting procedure. These values are in good agreement with pre-exponential factors reported by Baetzold and Somorjai for diffusion-controlled surface reactions [25].

Structural characterization
The ex situ XRD measurements were performed in the transmission mode utilizing a Stadi P diffractometer (STOE & Cie GmbH, Darmstadt, Germany). This setup is equipped with a MYTHEN2 DCS4 detector (DECTRIS Ltd., Switzerland) and a Mo X-ray tube (GE Sensing & Inspection Technologies GmbH, Ahrensburg, Germany) with a heating current of 40 mA and an acceleration voltage of 50 kV. A curved Ge(111) crystal selects the Mo K α1 radiation with a wavelength of 0.7093 Å. The analysis of the diffractograms was carried out with the software WinX POW using reference data from the ICDD database [26,27]. The in situ high-temperature synchrotron XRD experiments in H 2 were performed at the beamline 12.2.2, Advanced Light Source (ALS), Lawrence Berkeley National Lab, CA, USA. The in situ diffraction patterns were collected in the angle-dispersive transmission mode with a focused 25-keV monochromatic beam (λ = 0.4984 Å, 30 μm spot size). The sample powder was heated in a 0.7-mm quartz capillary under a continuous gas flow (10 ml min −1 pure H 2 ) injected through a 0.5-mm tungsten tube. The capillary is heated at 10°C min −1 to 500°C in an infrared-heated SiC tube furnace as described in [28,29]. Diffraction patterns were recorded by a PerkinElmer flat panel detector (XRD 1621, dark image and strain correction) every 35 s during the heating cycle.
Transmission electron microscopy (TEM) measurements were carried out using a FEI Tecnai F20 S-TWIN analytical (high-resolution) transmission electron microscope (200 kV), equipped with an Apollo XLTW SDD Xray detector (for collecting energy-dispersive X-ray (EDX) data).

DPP
The total copper content of the CuO/c-In 2 O 3 catalyst was determined by DPP) in 0.05 M sodium acetate as supporting electrolyte. Measurements were performed on a CV 50 W potentiostat (BAS) controlled by the software CV 50 (BAS) in combination with an EG&G 303a electrode stand and a classical threeelectrode configuration with a hanging mercury drop electrode (HMDE; drop size M) as working electrode, a platinum wire auxiliary electrode and a Ag/AgCl/3 M KCl reference electrode. A pulse amplitude of 50 mV with a duration of 50 ms, a sampling time of 20 ms, a pulse period of 200 ms and a scan rate of 10 mV s −1 was applied during each voltammetric scan. The Cu 2+ /Cu reduction peak occurs at 0.050 V versus Ag/AgCl/KCl (3 M) under these measurement conditions. Removal of oxygen from the solution is essential and was achieved by purging at least for 10 min with nitrogen 5.0 before the first recording of the voltammogram of the pure supporting electrolyte and at least 2 min after each sample or standard addition. 3.2 Reactive metal-support interaction between Cu and In 2 O 3 : monitoring the intermetallic compound formation between Cu and In during reduction in hydrogen

Results and discussion
The formation of intermetallic Cu-In compounds was investigated at first by ex situ powder XRD after different pre-reduction temperatures in the recirculating batch reactor to directly relate the structural consequences during pre-reduction to the subsequent catalytic measurements in MSR. A measure of 1 bar of pure H 2 for pre-reduction was used for 1 h, whereby the temperature was increased stepwise from 100°C to 500°C. Higher reduction temperatures were not used, since at 500°C, the In 2 O 3 support already starts to be reduced to metallic In. A fresh calcined sample was used for each treatment. The label indicating the last treatment performed on a sample consists of a letter for the respective gas (A for air, O for O 2 , H for H 2 , MSR for the MSR mixture) and a value for the temperature (i.e. H100 denotes pre-reduction in pure H 2 at 100°C for 1 h).
The ex situ collected XRD patterns of the pre-reduction series at five temperatures from 100°C to 500°C are displayed in Figure 1. Starting at 100°C, Cu remains fully oxidized (as CuO), whereas reduction at 200°C already yields metallic Cu on the yet unchanged c-In 2 O 3 support. The H300 sample appears structurally very similar to the one treated at 200°C, but a further increase to 400°C leads to the formation of intermetallic Cu-In compounds. According to the Cu-In phase diagram, Cu 2 In is the intermetallic phase with the highest melting point. The most prominent reflections in the H400 XRD pattern can be assigned to such a hexagonal Cu 2 In structure with the space group P6 3 /mmc (a = 4.2943 Å and c = 5.2328 Å) [30]. Cu 4 In could also be present, as it also displays its most intense reflection in the same 2θ region. However, the other reflections do not clearly match Cu 4 In. Metallic copper vanishes completely, which further corroborates the formation of intermetallic Cu-In compounds. Most of the Cu-In intermetallic reflections are somewhat shifted from their ideal position, which can be explained by the still ongoing phase formation and the homogeneity range, allowing for the incorporation of some amounts of Cu/In into the parent intermetallic compound lattice resulting in deviations from the ideal lattice parameters.
After pre-reduction at 500°C, c-In 2 O 3 starts to become partially reduced to metallic In, and the reflections of the intermetallic compounds change significantly in their position, as well as in their relative intensity. The shift in the diffraction angle could once more be the result of a phase width of a distinct intermetallic Cu-In compound with a concomitant change of the lattice parameters. In summary, we can conclude that after a reduction treatment at 300°C, metallic Cu, but no bulk intermetallic Cu-In, compound is present, whereas after reduction at 400°C, the metallic Cu state disappears and is replaced by Cu 2 In.
We mention already at this stage that a pre-reduction at 300°C and 400°C conveniently allows to assess the potential influence of intermetallic compound formation on the catalytic behavior in MSR as a test reaction. Two states can, hence, directly be accessed: Cu/c-In 2 O 3 (300°C) and Cu 2 In/c-In 2 O 3 (400°C). An obvious prerequisite, however, is the bulk structural stability of the phases during the catalytic reaction. This has been assessed by ex situ XRD and is shown accordingly in Figure 2. Starting from the calcined state (A400, with an additional O400 treatment prior to MSR, ensuring that the sample is completely oxidized and removing residual surface impurities), the catalyst is reduced to a mixture of Cu 2 O and Cu by the MSR mixture in the course of reaction (diffractograms after two MSR cycles each consisting of O400 and MSR up to 350°C (termed MSR350) are being depicted in Panel A of Figure 2). Introducing an additional pre-reduction step at 300°C in H 2 before MSR changes the initial state to fully reduced Cu on c-In 2 O 3 and no significant changes in the diffractogram are visible after two complete MSR cycles (each consisting of O400, H300 and MSR350, see  Further insight into the reactive metal-support interaction between Cu and cubic In 2 O 3 is given by transmission electron microscopy analysis (Figure 3). The bright field overview image in Panel A clearly shows the decoration of an In 2 O 3 grain with several small particles with rounded outlines and rather large particle sizes between 30 and 80 nm after a H400 reduction step. Two selected particles have been marked, which are correspondingly also visible in the high-angle annular dark-field (HAADF) image and the Cu-K EDX map of Panel E. Further structural analysis of the two particles in the red and the blue boxes indicate the intermetallic compound formation between Cu and In, following reduction of c-In 2 O 3 and subsequent interdiffusion of In and Cu. The observed fringes are clearly located at the individual particles and do not extend into the support grains. Lattice fringes of the hexagonal intermetallic Cu 2 In compound measured at~2 Å and~3 Å are clearly visible (space group P6 3 /mmc, a = 4.2943 Å and c = 5.2328 Å, d theor (110) = 2.14 Å, d theor (101) = 3.03 Å) [30]. The combined structural analysis, hence, indicates the presence of Cu 2 In (also on the basis of the in situ reduction studies displayed in Figure 4). On a qualitative basis, the intermetallic compound formation can also be directly inferred from combined HAADF/EDX imaging ( Figure  3, Panels D-G). The individual EDX maps of the Cu-K and In-K intensities clearly show the simultaneous presence of Cu and In at the same locations. The Cu particles can be seen clearly, but due to the underlying In 2 O 3 support, the In-K intensity is more homogeneously distributed. Interestingly, the Cu 2 In particles can also be seen in the HAADF image due to the different average atom number of Cu 2 In and In 2 O 3 . Generally, the HAADF intensity can be approximated as I HAADF ≈ t·ρ·Z 1.5 , where t is the sample thickness, ρ the material density and Z the average atom number. Hence, constant thickness and density provided, the HAADF intensity is dominated by Z, showing elemental contrast. Areas with higher average atom number, therefore, appear brighter in an HAADF image. Applying this concept to the Cu-In 2 O 3 system, we note that the average atom number Z of In 2 O 3 is 24.4, the one of Cu 29.0 and the one of Cu 2 In 35.6. Hence, the HAADF intensities I of the three components under question can be approximated by 1396t, 2042t and 858t for Cu, Cu 2 In and In 2 O 3 , respectively (assuming densities of 8.936 g cm −3 for Cu [34], 9.614 g cm −3 for Cu 2 In [30] and 7.117 g cm −3 for In 2 O 3 [24]). The intensity difference between the intermetallic compound and In 2 O 3 is the largest, directly validating the observed contrast.
Summarizing the reactive metal-support interaction between Cu and cubic In 2 O 3 , we note that the interaction, as observed for corresponding metaloxide systems, can be described as a multistep process [5]. The latter may involve reduction of the oxide support species (In 2 O 3 in this case), followed by diffusion of the reduced In 2 O 3 species (even In metal) into the Cu lattice and subsequent formation of the intermetallic compound. A tentative reaction of Cu, In 2 O 3 and H 2 to result in Cu 2 In is given by The thermodynamic driving force for intermetallic compound formation is, thus, the efficient removal of reaction-formed water, which shifts the equilibrium toward the products. Thus, it can be expected that a treatment in the recirculating batch reactor, where the water basically is not removed, merely causes a minor reduction of In 2 O 3 . Hence, also the intermetallic compound formation is apparently less efficient compared to treatments, where water is constantly removed. To assess the influence of water removal, in situ XRD measurements have been performed. In this capillary setup, quasi-flowing Figure 2. Comparison of the ex situ collected XRD patterns of the Cu/c-In 2 O 3 catalyst before and after MSR using different pre-treatments. The most intense reflections have been cut to increase the visibility of the other peaks. The diffractograms after MSR are depicted with an offset. The references were taken from the ICDD database [27] (for PDF numbers, see caption of Figure 1). Panel A shows the XRD patterns after calcination (A400), O400 and two MSR cycles without prereduction; Panel B compares the patterns after O400, H300 and two MSR cycles; and Panel C compares the patterns after O400, H400 and four MSR cycles.
conditions can be realized as discussed in the experimental section. These are collectively shown in Figure 4.
In the course of the experiment, qualitatively the same sequence of phase transformations can be observed, corroborating the data shown in Figure  1. Starting from oxidized Cu particles on c-In 2 O 3 , formation of metallic Cu sets in at around 240°C, before reduction of In 2 O 3 and intermixing of Cu and In is observed roughly between 370°C and 450°C. In this temperature region, a complex phase mixture of several intermetallic Cu-In compounds is observed, before finally at 460°C, Cu 2 In prevails as the sole intermetallic compound. The 2θ shift in the reflections of Cu 2 In and Cu in the temperature range of Cu 2 In formation suggests that the intermetallic compound is formed by interdiffusion of Cu and In. This finding directly corroborates the phase assignment in the ex situ XRD measurements, where around 400°C also Cu 2 In prevails as the main phase.

Influence of the intermetallic compound formation on the CO 2 selectivity and performance in MSR
The influence of intermetallic compound formation starting from CuO on c-In 2 O 3 on the performance in MSR was investigated by starting the reaction from different initial states. The most important catalytic parameters of relevant cycles of the three initial states are summarized in Table 1. In Figure 5, the MSR reaction was conducted after calcination with an additional oxidation step at 400°C for 1 h in pure O 2 (O400). The formation of hydrogen and carbon dioxide starts at approximately 180°C, but no CO and only traces of CH 4 (below 0.05 mbar in total) are produced. As expected, hydrogen evolving in the beginning is almost entirely consumed again, reducing the initially present CuO to Cu 2 O and Cu (see also Figure 2, Panel A). The methanol conversion reaches a maximum of 90%, and the CO 2 selectivity amounts to 100% at any time because of the absence of CO formation. The decrease in the formation rates in all catalytic profiles is caused by the depletion of the reaction mixture, which is inherent to a batch reactor. The second cycle on this system was omitted, because it exhibits an almost identical behavior.
Starting MSR after pre-reduction in 1 bar pure H 2 for 1 h at 300°C yields the catalytic profiles depicted in Panels A and B of Figure 6, which represent two full catalytic cycles (O400, H300, MSR350) on the same sample without exposure to air. In contrast to the measurements without pre-reduction, hydrogen is formed as the main product of the reforming reaction with a rate that is three times as high at its maximum as the corresponding value for CO 2 at the same time, representing the ideal stoichiometry of MSR. Only traces of CO and CH 4 are being formed with their maximum rates amounting to approximately 1.5% (1.8% CO and 1.0% CH 4 in the second cycle) compared to H 2 . This is also reflected in the CO 2  [24,27,[34][35][36][37]. Note that the reaction scenario outlined in Equation (1) is not the only pathway that might lead to intermetallic compound formation. In terms of structural consequences of reactive metal-support interaction, the main question remains, if (i) first oxidized Cu is being reduced to metallic Cu, then In 2 O 3 reduction to metallic In takes place and subsequently In reacts with Cu to form Cu x In intermetalloids, (ii) if it proceeds as outlined in the reaction according to Equation (1) or (iii) CuO and c-In 2 O 3 are reduced simultaneously to form Cu x In phases. So far, the data suggest that reduction of both oxidized Cu and In 2 O 3 occurs, but thermodynamic considerations, especially also with respect to the relative stabilities of the participating phases, are needed.
selectivity, which exceeds 93% over the course of the entire measurement in both cycles. The methanol conversion, however, only reaches a final value of 56% (51% in the second cycle), suggesting a rather low activity of the catalyst.
Increasing the pre-reduction temperature to 400°C, that is, starting from Cu 2 In/c-In 2 O 3 , leads to the catalytic patterns of four complete MSR cycles (O400, H400, MSR350) depicted in Panels C-F of Figure 6, which look similar to the ones Figure 5. Catalytic profile of one MSR cycle after oxidation in pure oxygen at 400°C (O400) without further pre-reduction (starting from CuO/c-In 2 O 3 ). The formation rates (in µmol g Cu −1 s −1 ; green axis) of H 2 , CO, CO 2 and CH 4 ; the CO 2 selectivity (blue axis and trace); and the methanol conversion (blue axis, but orange trace) are plotted versus the reaction time. The temperature profile is depicted in red. Figure 6. Catalytic profiles of two MSR cycles with O400 and H300 (starting from Cu/c-In 2 O 3 ; Panels A and B) and four MSR cycles with O400 and H400 (starting from Cu 2 In/c-In 2 O 3 ; Panels C-F). In each graph, the formation rates (in µmol g Cu −1 s −1 , green axis) of H 2 , CO, CO 2 and CH 4 ; the CO 2 selectivity (blue axis and trace); and the methanol conversion (blue axis, but orange trace) are plotted versus the reaction time. The temperature profile is depicted in red.
conducted at beamline 12.2.2 in the framework of the AP program ALS-08865. We particularly acknowledge the group of H. Huppertz from the Institute of General, Inorganic and Theoretical Chemistry for access to the X-ray diffractometer.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This work was supported by the Austrian Science Fund [F4503-N16].