Overcoming Pd-TiO2Deactivation during H2Production from Photoreforming Using Cu@Pd Nanoparticles Supported on TiO2

Depósito de investigación de la Universidad de Sevilla https://idus.us.es/ "This document is the Accepted Manuscript version of a Published Work that appeared in final form in ACS APPLIED NANO MATERIALS, copyright © American Chemical Society after peer review and technical editing by the publisher. To access the final edited and published work see https://doi.org/10.1021/acsanm.1c00345 ." 1 Overcoming Pd/TiO2 Deactivation during H2 Production from Photoreforming by Using Cu@Pd Nanoparticles Supported on TiO2 F. Plateroa, A. López-Martína, A. Caballeroa, T.C. Rojasa, M. Nolanb, G. Colóna * a Instituto de Ciencia de Materiales de Sevilla. Centro Mixto Universidad de Sevilla- CSIC. Américo Vespucio s/n. 41092 Sevilla. Spain b Tyndall National Institute, University College Cork, Lee Maltings, Dyke Parade, Cork, T12 R5CP Ireland ABSTRACT Different Cu@Pd/TiO2 systems have been prepared by a two-step synthesis to obtain a bimetallic co-catalyst for H2 photoreforming reaction. We find that the tailored deposition of Pd covering the Cu nanoclusters by a galvanic replacement process results in the formation of a core@shell structure. The photocatalytic H2 production after 18 h is 350 mmol/g on [email protected]/TiO2 bimetallic system, being higher than on the monometallic ones, with H2 production of 250 mmol/g for Pd supported TiO2. Surface characterization from HAADF-STEM, H2-TPR, CO-FTIR and XPS gives clear evidence of the formation of a core@shell structure. With a Pd loading of 0.2-0.3 at% we propose full coverage of the Cu nanoparticles with Pd. Long time photoreforming runs show the enhanced performance of supported Cu@Pd with respect to bare palladium leading to a more stable catalyst and ultimately higher H2 production. KEYWORDS: Copper; Palladium; Core-Shell; TiO2; H2 Photoreforming; DFT. * Corresponding author: [email protected] 2 INTRODUCTION In recent years, H2 production by the alcohol photoreforming reaction has gained significant interest as a new topic in the field of photocatalysis.1,2 and is an alternative to water splitting for hydrogen production. Taking into consideration the current socioeconomical and environmental situation, it is now accepted that hydrogen must play the key role in the future energy system since it is considered the ultimate clean energy carrier. Despite significant work over 40 years on water splitting, the actual efficiencies reported for a wide variety of photocatalytic systems are still far from a technologically feasible situation for solar hydrogen production.2 To date, the best performance has been obtained by TiO2 based systems using UV light. In spite of that, this is of great importance for solar applications. Although UV region constitutes a small fraction of solar spectrum, the high efficiency of this UV photogenerated carriers justifies the studies at lab scale and would be the basis for pilot plant scaling-up.3 However, some important factors make its application at present far from practical situation. The rapid recombination process of the photogenerated charge pairs and the occurrence of backward reactions are considered critical issues that hinder high efficiency. Traditionally, many catalytic systems incorporate metal ions as charge trapping sites which can avoid the electron–hole recombination processes.4 The most active metals are noble metals (Pt, Au, Ag, Pd, Rh)5,6,7. It has been found that the addition of these noble metals could have different effects on the photocatalytic activity which is also strongly affected not only by the nature of metal but also by other parameters including sample history and features of the metal.8 Mul et al argued that the understanding of metallic particle sizes and its dispersion, the specific functionality of each co-catalyst is crucial for the development of active systems for hydrogen production.8 3 In this sense, Au and Ag show a surface plasmon resonance (SPR), which can absorb certain ranges of visible light that could improve the photocatalytic performance of TiO2.4,9 On the other hand, in the case of Pt, the enhancement in the photocatalytic activity is due to the Schottky junction formation at the interface between Pt and TiO2, resulting in a larger barrier height with respect to Au and Ag nanoparticles.10,11 This way the improvement in the photocatalytic activity could be due by the hindering of the electron–hole (e−–h+) recombination rate. Alternatively, transition metal doping with Cu, Fe, Co or Ni have been proposed as a promising cheaper option in comparison to noble metals. However, the efficiencies showed are still low in comparison to noble metals.3,12,13 Moreover, as reported in many other catalytic processes, significant activity improvement is usually obtained by formulations containing bimetallic components.14,15 This is due to their higher tunability with respect to monometallic systems. Thus, according to metals features, bimetallic active sites can show different structures, metal(s) support interactions, etc. Moreover, the structure of bimetallic clusters can varied from bimetallic hetero-structures, to core–shells, alloyed clusters, changing each case the type of metal interaction. In addition, the modification of a metal oxide such as TiO2 with bimetallic nanoparticles could be also a feasible strategy for enhancing hydrogen generation, at the same time minimizing the use of expensive noble metals. Among the improvements achieved by the bimetallic formulation, we may highlight the formation of unique interfaces at metal–support and metal–metal heterojunction that may increase the stability and enhance the interfacial charge transfer kinetics.16 In the case of photoreforming reaction, an improved co-catalyst stability would certainly affect to the final H2 production. 4 Many examples have been reported in the literature describing improvements using bimetallic co-catalysts compared to monometallic ones. Thus, Murcia et al proposed bimetallic Cu-Ag and Cu-Pt co-catalytic systems showing higher H2 productivities, where the presence of Cu aids the control of the extent of the oxidation reaction.17 Similarly, bimetallic Pd-Ni samples have been reported to be more efficient than either monometallic systems. These authors reported a synergetic effect with the presence of Pd nanoparticles and Ni0-NiO clusters on TiO2.18 An enhanced rate of hydrogen evolution was obtained for Au-Ni co-modified TiO2, revealing a synergetic effect of the two metals.19 The evidence so far is that the role of the metal is not only to avoid the recombination of charge-carriers by trapping electrons4, but also to act as a recombination center for atomic hydrogen (Hꞏ) from water oxidation or alcohol reforming to produce hydrogen. In the present paper we present an interesting approach to copper based catalysts by incorporating increasing amounts of palladium into a bimetallic or core-shell structure supported on TiO2. The proposed preparation method through redox displacement assure the selective location of these palladium atoms intimately joined to copper. In fact, redox displacement has been reported as a very effective method for the preparation of coreshell materials.20 Thus core-shell structuration of Pd clusters supported on TiO2 induces a notable improvement of the photoreforming behaviour of monometallic Pd/TiO2 system to produce hydrogen. EXPERIMENTAL Catalysts preparation and characterisation The TiO2 system used in the present work was commercial Evonik P25. Then metal loading was performed through sequential deposition procedure (Scheme 1). At a first 5 step, copper was deposited by photodeposition method using a close immersion-well reactor. In this case, 0.5 g of TiO 2 was suspended in 350 mL of isopropanol-water (1:10 v/v) containing the stoichiometric amount of Cu 2+ precursor for a nominal loading of 2 at%. Photodeposition was achieved by illumination with UV lamp (125 W medium pressure Hg lamp) under N 2 flow (20 mL/min) for 1 hour at 18ºC ± 1ºC. After that, lamp was switched off and the stoichiometric amount of Pd 2+ was added to the above suspension (from 0.05 at% to 1.0 at%). This mixture was stirring in the dark for 1 hour at the same temperature and N 2 flow. At this moment, Pd 2+ ions were interchanged with deposited Cu 0 atoms through galvanic replacement process. 20 Therefore, initial Cu 2+ content would vary as a function of Pd 2+ incorporated. The obtained systems were filtered, thoroughly washed with distilled water and finally dried at 90 ºC. Bimetallic doped systems are expected to have 2 at% loading and were denoted as Cu@Pd y , being y the Pd at%. Monometallic Pd systems were prepared by photodeposition method as in the case of monometallic Cu. Scheme 1: Multistep preparation synthetic route for core-shell structuration through galvanic replacement. 6 BET surface area measurements were carried out by N2 adsorption using a Micromeritics 2000 instrument. X-ray diffraction (XRD) patterns were obtained using a Siemens D-501 diffractometer with Ni filter and graphite monochromator. The X-ray source was Cu K radiation. From the line broadening of corresponding X-ray diffraction peaks, we have calculated the anatase fraction as well the mean crystallite size by Rietveld fitting using HighScore-Plus software. Inductively coupled plasma-optical emission spectroscopy (ICP-OES, iCAP 7200 Duo, ThermoFisher Scientific) was applied to determine the metal loadings in the samples. Samples were previously dissolved by means of acid digestion with a H2SO4 and H2O2 mixture. Diffuse reflectance spectra were obtained on a UV–vis scanning spectrophotometer Agilent Cary 300, equipped with an integrating sphere, using Spectralon® as reference material. UV-vis spectra were performed in the diffuse reflectance mode (R) and transformed to a magnitude proportional to the extinction coefficient (K) through the Kubelka-Munk function, F(R). For the sake of comparison, all spectra were arbitrary normalized in intensity to 1. Band gap values were obtained from the plot of the modified Kubelka-Munk function (F(R)E)1/2) versus the energy of the absorbed light E. The morphology of samples was followed by means of TEM microscopy (JEOL 2100Plus microscope). The samples were dispersed in ethanol using an ultrasonicator and dropped on a nickel grid. High-angle annular dark field scanning TEM (HAADF-STEM) images and energy-dispersive X-ray (EDX) elemental maps were performed using a TALOS 7 F200S microscope from FEI company, working at 200 kV with 0.25 nm resolution, with a Super-X: 2SDD EDX and HAADF-STEM detectors. The samples were deposited on a holey carbon support film on 400 mesh Au grid, without using any solvent to avoid contamination during the maps acquisition. Field emission scanning electron microscopy (SEM-FEG) was performed by using a Hitachi S4800 microscope. In this case, the samples were dispersed in ethanol using an ultrasonicator and dropped on a nickel grid. XPS data were recorded on pellets prepared by slightly pressing the powdered materials which were outgassed in the prechamber of the instrument at room temperature up to a pressure < 2ꞏ10-8 torr to remove chemisorbed water from their surfaces. Spectra were recorded using a Leybold-Heraeus LHS-10 spectrometer, working with constant pass energy of 50 eV. The spectrometer main chamber, working at a pressure <2ꞏ10−9 Torr, is equipped with an EA-200 MCD hemispherical electron analyser with a dual X-ray source working with Al Kα (h = 1486.6 eV) at 120 W and 30 mA. C1s signal (284.6 eV) was used as internal energy reference in all the experiments. Temperature programmed reduction (H2-TPR) was performed using a Chemstar instrument (Quantachrome). About 80 mg of the catalysts previously degassed at 120 °C for 30 min under Ar flow. The TPR spectra were collected in 10 mL/min mixture of 5% H2/Ar from 35 ºC to 700 °C with a heating rate of 10 °C/min. CO adsorption on the fresh catalysts (pretreated in He at 120 ºC for 30 min) was performed in a Harricks praying mantis cell. The spectra were recorded on a Nicolet FT- IR spectrometer at 50 ºC with a resolution of 4 cm−1. After pretreatment, samples were exposed to CO/He flow for 10 min. Then, spectra were recorded after evacuated with He for 10 min. The contribution from CO in the gas phase was in all cases eliminated by subtracting the corresponding spectra obtained from the bare TiO2 support. 8 Photocatalytic runs Photocatalytic H2 production tests were performed in a flow-reactor system (Figure S1). The powder photocatalysts were suspended in water-isopropanol solution (1:9 v/v). The reaction media was continuously thermostated at 18ºC ± 1ºC to prevent any temperature effect. The catalyst suspension (0.5 g/L) was firstly degassed with an N2 stream (80 mL/min) for 30 min to remove dissolved and headspace oxygen. After that the N2 flow was settled at 15 mL/min and stabilised for 30 min. This nitrogen flow was used to displace the hydrogen produced from the photoreactor headspace towards the GC measuring system. Then, the lamp (125 W medium pressure Hg lamp) was switched on and the effluent gases were analysed to quantify H2 production by gas chromatography (Agilent 7890B GC) using a thermal conductivity detector. The photonic efficiency for the H2 evolution reaction has been determined from the reaction rate and the flux of incoming photons (calculated for the irradiation wavelengths of 365 nm)13. In order to determine the intermediates originated during isopropanol photoreforming, we have also performed the qualitative analysis of the liquid phase by GC-MS (Agilent 7000D). Computational Methodology All density functional theory computations presented are carried out in a periodic plane wave basis set using the VASP.5.4 code.21 The cut-off for the kinetic energy is 400 eV and projector augmented wave all-electron, frozen core potentials22 are used, with the following valance electron configurations: Ti 4s2 3d2, O 2s2 2p4, Cu 4s1 3d10, Pd 4d10 and the one-electron hydrogen PAW potential. The Perdew-Wang 91 approximation to the exchange-correlation functional is used.23 The bulk lattice parameters for rutile were computed as a=b=4.614 Å and c=2.962 Å The TiO2 substrate is the (110) rutile surface, 15 ca. 2 nm to 4 nm as Pd is incorporated. In addition to this size growth, there is increased aggregation of the particles for higher Pd content (Figure 2.c). Indeed, [email protected] sample mainly shows this aggregate morphology. The core–shell structure of the proposed Cu@Pd NPs was verified by STEM-EDS. Figure 3 shows a HAADF-STEM image of [email protected]. From selected EDS elemental mapping along corresponding line-scan plot it is evident that Cu nanoparticles are confined in the center of the nanoparticle while Pd appears uniformly distributed throughout the entire particle. Thus, the EDS line-scan profile clearly evidences the core–shell structure of Cu@Pd nanoparticles. Moreover, in this side view of the nanoparticle, it is also evidenced that Cu is specifically in contact with Ti. We have observed a similar core-shell structuration for [email protected]/TiO2 sample (not shown), denoting that the preparation procedure assure such morphology. Isolated Cu or Pd particles were not detected. In order to assess any possible change of metal site morphology during the reaction, we have observed by TEM the [email protected] and Pd1.0 samples after 5 hours of reaction (Figure S5). It is worthy to note that Pd1.0 clearly suffers an important cluster size growth, showing an average particle size of 5.6 nm. However, [email protected] seems to loss the aggregated morphology, maintaining at the same time the cluster size (ca. 4.1 nm in this case). By comparing the average metal size for fresh and used samples, it can be clearly stated that Pd1.0 sample shows an important size growth in the first 5 hours of reaction (5.6 nm vs 4.0 nm for used and fresh samples respectively). The behavior of bimetallic system is completely different. In this case, the initial Pd aggregates seems to diffuse on TiO2 support surface keeping the metal cluster size. In this sense, Spanu et al observed from operando XAS study the dissolution/redeposition process in a CuNi/TiO2 catalyst under UV illumination.37 These authors argued that partially oxidized Ni and Cu present in the catalyst would be dissolved during reaction that would be further reduced/redeposited as 16 a bimetallic NiCu phase at the TiO2 surface under illumination. This complex mechanism would explain the aggregation observed in Pd monometallic systems and the stability of bimetallic ones at long reaction times. We also mention that attending on examining the evolution of the Pd and Cu content from ICP chemical analysis, for Pd contents higher than 0.4, the amount of copper appears to be stable. Therefore at higher Pd loadings, copper can still reduce Pd but without being displaced. The complete coverage of copper clusters by palladium at this Pd content would block further displacement of copper. Thus, palladium reduction and deposition over metal clusters would proceed without copper leaching. Upon this situation, the electron transfer from copper core to palladium in solution would take place through the metallic palladium shell or even through TiO2 support. We would consider that Cu photodeposition take place over defective points of TiO2. At this defective structural points electron transfer would be favoured and would allow Pd2+ reduction and deposition. We have also measured the UV−vis DRS to study the optical properties of photocatalysts. Figure S6 shows the diffuse reflectance spectra for different Cu@Pd/TiO2 systems. The strong absorption band below 380 nm (giving a band gap value of 3.2 eV) would correspond to the ligand to metal O2−→Ti4+ charge transfer (LMCT) transitions in the TiO2 support. In addition, new absorption bands can be observed in the visible range for all systems. As reported in the literature, CuOx generally exhibits absorption bands in the visible range at 400-600 nm and above 700 nm, assigned to both Cu+ and Cu2+ species.38,39 Cu2.0/TiO2 shows a small shoulder at 450 nm and broad band at around 780 nm. The absorption edge for Cu2O nanoparticles corresponding to an interband transition has been located at ca. 441 nm.40 Moreover, the absorption at wavelengths than 650 nm would be consistent with the intraband absorption of Cu2O NPs.40,41 This later band would extent 17 to 700-800 nm range and can be understood in terms of d–d transition of copper ions. Metallic copper is reported to show a plasmon resonance absorption between 450-600 nm which strongly depends on the cluster size.41,42 Therefore the likely coexistence of Cu0 and Cu+ species can be derived from the UV-vis absorption spectrum. The occurrence of Pd ions in the bimetallic systems, even at low content, results in an intense absorption in the visible region. Indeed, the absorption at around 460 nm has been associated to the d-d transitions of Pd2+ ions from the energy states of –O–Pd–O– surface species to the TiO2 conduction band.43 Furthermore, metallic Pd also shows a plasmon absorption (surface plasmon resonance effect) at around 500 nm.44 However, a further increase of the Pd loading leads to an excessive absorption which is caused by the induced light scattering phenomena which restricts a clear interpretation of the present species. To further clarify the nature of the surface Cu and Pd species we have performed FTIR analysis of CO chemisorption and TPR analysis (Figure S7). The infrared spectrum of irreversibly adsorbed carbon monoxide for Cu2.0 TiO 2 sample presents intense bands located at 2110 and 2095 cm−1 and an additional smaller band at 2055 cm-1 (Figure S7.a). It has been reported that CO adsorption onto copper showing different oxidation states leads to absorption bands located between 2050 and 2220 cm−1 according to the crystal plane and to the nature of the support.45 Previous literature studies of CO adsorption yield the following assignments: bands in the range 2220−2150 cm−1 have been associated to Cu2+–CO, whereas bands in the range 2160−2080 cm−1 are attributed to Cu +–CO. Metallic copper carbonyls, which can be single-site or nanostructures species are expected to be located below 2130 cm−1.46,47 Due to the overlapping of different Cu species, the interpretation of CO adsorption by FTIR is quite complex. Nevertheless, as a general trend, it is widely accepted that that peaks >2100 cm−1 could be ascribed to CO adsorption on oxidized copper species Cun+, while peaks <2100 cm−1 are due to CO 18 adsorption on metallic copper.48 Therefore, at the first glance, since no signal has been observed at wavenumber higher than 2110 cm−1, the first important statement would be the lack of Cu2+ species. This fact will be discussed later. Moreover, considering the above assignments from other works, we could propose that more intense bands at ca. 2110 cm−1 with its shoulder at 2095 cm−1 observed in the present work could be assigned to edge sites or Cu+ species and linearly adsorbed CO onto metallic copper at edge or corner sites, respectively. In addition, the small band at 2055 cm-1 could be ascribed to the adsorption onto Cu (111) facet of small metallic copper particles. On the other hand, monometallic Pd/TiO2 system shows two absorption bands for irreversibly adsorbed CO. A broad band centered at around 1980 cm−1 (with poorly resolved shoulders at 1985 and 1950 cm−1), corresponds to bridged multicentered carbon monoxide.49 An additional small wide band at 2065 cm−1 has been associated to carbon monoxide bonded to surface palladium atoms in linear forms.49,50 The band at about 1980 cm–1 is notably stronger than that at 2065 cm–1 and this is due to the larger particle size of Pd which favours multibridged CO adsorption. It has been discussed that CO adsorption on Pd2+ and Pd+ results in bands at 2173 cm–1 and 2120 cm–1.51 In our case the absence of bands at such wavenumbers would clearly demonstrate that only metallic Pd is present, as it would be expected from photodeposition method. Thus, considering these assignments for monometallic systems, describing the CO IR absorption bands for bimetallic one becomes quite complex. For low Pd content we observe the following evolution in the FTIR bands: i) the 2095 cm−1 band trends to disappear as Pd content increases; ii) the overall intensity of the originally intense band 2110-2095 cm-1 progressively diminishes and is shifted to 2105 cm-1; iii) the band at 2055 cm−1 remains almost unaltered for low Pd content; iv) a new broad band at 1985 cm−1 starts to develop. This evolution clearly shows that the addition 19 of Pd will diminish the bands corresponding to CO adsorption at Cu+,Cu0. Moreover, such diminution seems to be more evident for the band at 2095 cm−1, that has been formerly assigned to metallic copper at defective sites. In addition, a progressive shift toward lower wavenumber would indicate that the remaining copper is strongly affected by the presence of palladium. From the evolution of the bands formerly assigned to copper (Cu+ and/or Cu0) we would envisage two important processes: i) the loss of copper from the surface; ii) covering of copper by Pd. Thus, it can be concluded that Pd2+ tends to replace Cu+ and Cu0 located at defective sites in the cluster. Moreover, as the Pd content increases, the band at 2055 cm−1 previously associated to linearly bonded CO seems to be also affected. Additionally, the occurrence of a new wide band at 1985 cm−1 for Pd contents higher than 0.1 mol% would indicate the presence of metallic Pd species multicentere bonded with CO. Therefore, it is clear that the incorporation of Pd2+ into the surface is taking place as metallic palladium due to a redox replacement process. As in the case of monometallic Pd systems the greater intensity shown by the band at 1985 cm−1 would indicate the larger size of Pd clusters. From these considerations, it can be envisaged that palladium incorporation would take place intimately at copper clusters, in good agreement with STEM analysis (Figure 3). The copper-palladium interaction can be examined by analyzing the reduction behavior of the metal clusters by H2 temperature programmed reduction (Figure S7.b). The TPR profile of the reference Cu2.0 sample exhibits a three step H2-consumption at 130 ºC, 150 ºC and 168 ºC corresponding to reduction of Cu2+/+ species to metallic copper. From the H2 consumption and taking into account the consumption for a CuO reference, we calculate that the metal loading corresponds to 2.1 at%, which is close to the nominal value from ICP. This would indicate that whole Cu2+ in the solution has been photodeposited, which is also in agreement with ICP analysis (Table S3), and that a 2+ 20 oxidation state is present. This could appear to contradict the previous information drawn from FTIR. Indeed, from CO adsorption it was stated that no Cu2+ is present in Cu/TiO2. This could be explained by the easy reducibility of Cu2+ species. Thus it could be said that Cu2+ would be rapidly reduced to Cu+/Cu0 in the presence of CO. In fact, the lower reduction temperatures would point out the easy reduction of any Cu2+ species to Cu0, probably due to the high dispersion (which include isolated copper ions, weak magnetic associates, and small two- and three-dimensional clusters) as well the soft interaction with TiO2 support. Indeed, we have previously shown from DFT studies that the reduction of CuO (111) with H2 is favourable.52 In the case of the Pd1.0 sample, two significant effects can be observed. Firstly, a negative H2 consumption peak at around 77 °C is ascribed to H2 release from palladium hydride β-PdHx.53 This palladium hydride was formed during H2 purge and TCD stabilization prior to TPR analysis. Thus, H2 could be absorbed by metastable Pd crystallite particles at ambient temperature. A second effect is observed at around 340 ºC which would be either assigned to the hydrogen consumption by a spillover from Pd to the support material,54 or to the reduction of oxidised PdOx species in intimate contact with the support. For bimetallic systems, the former reduction peak of copper shows two important effects. The H2 consumption peak is shifted to lower temperatures (below 100 ºC), and this shift is more notable as the Pd content increased. Besides, H2-TPR peaks show markedly reduced intensities. These two effects would indicate out that reduction process of Cu species is taking place at lower temperature, probably due to the strong interaction with Pd. Such a promotion in the CuOx redox potential should be the result of a SMSI effect between Pd and CuOx that facilitated the diffusion of dissociated hydrogen to Cu clusters.55 Thus, the formation of bimetallic clusters can be envisaged. Secondly, the 21 lower intensities for H2-TPR peak could be explained by considering the negative peak exhibited by Pd corresponding to the H2 desorption from -PdHx species. Given that these two effects, Cu reduction uptake and Pd desorption release, are observed at similar temperatures, they could be partially balanced leading to a significantly supressed H2- TPR peak. Moreover, Batista et al investigated the reduction of monometallic palladium, copper and bimetallic Pd–Cu catalysts and found that in the presence of palladium, Cu2+ would be reduced to Cu0 at lower temperatures through a spillover mechanism.56 Therefore, Cu would be partially reduced before TPR experiment leading to a lower H2 consumption. The progressive diminution of the copper content would also lead to a lower H2-TPR peak. In addition to this small reduction peak, further H2-uptake is found at around 340 ºC which is unaltered with Pd content. This, second reduction peaks was already present in Pd1.0 TiO2 sample so the occurrence might be associated to the presence of Pd in the bimetallic system. As in the case of monometallic system, this peak could be associated to the reduction of a small fraction of PdOx as well as reduction of the TiO2 support through spillover process. Surface features of Cu@Pd/TiO2 systems have been also studied by XPS. In Figure 4 (upper row) we show the Cu 2p and Pd 3d XPS peaks for all systems. Firstly, it can be said the all bimetallic systems show the presence of both doping ions denoting that the replacement of Cu by Pd has been successful. Thus, as we increased the amount of Pd it is evident that Cu content is progressively decreased (Table S3). It is worthy to note that the evolution of surface composition attained from XPS is similar to that obtained from chemical analysis. To evaluate the oxidation states of Cu and Pd species deposited on the TiO2 support, we measured the Cu 2p3/2 and Pd 3d5/2 electronic transitions, as well as the Cu LMM Auger 22 transitions. Considering the Cu 2p level (Figure 4.a, upper row), there is a clear doublet located at 934.2 and 952.1 eV. From the binding energy value and the absence of the shake-up satellite we can state that copper species are Cu0 and/or Cu+, which agrees with CO-FTIR discussion. Similarly, the easy reducibility of former Cu2+ species, would favour the occurrence of reduced Cu+/Cu0 species during XPS analysis. Attending to the Cu LMM auger transition (Figure 4.b, upper row) it is possible to assess that a mixture of Cu0 and Cu+ is always present.57 From the relative intensities of Cu LMM peaks at kinetic energies ca. 917 and 919 eV (from this KE, modified Auger parameters of ca. 1849 end 1851 eV can be attained), which correspond to Cu0 and Cu+ respectively, we could estimate the Cu0/Cu+ ratio in the different systems (Table S5). Monometallic Cu/TiO2 systems shows a mixture of Cu0 and Cu+, with Cu0 being the predominant oxidation state. As Pd is incorporated the amount of Cu0 significantly increases. On the other hand, the considering Pd 3d peak, it can be noticed that it is composed by three peaks located at ca. 334.5 and 336.5 eV which correspond to Pd0 and Pd2+ respectively (Figure 4.c, upper row). The first important point that can be drawn is that for lower palladium loadings the binding energy values for Pd 3d is notably positively shifted with respect to bare Pd sample (ca. 335.5 eV for Pd0.2Cu sample vs 334.6 eV for Pd1.0 sample). This clearly denotes a possible electron transfer from Pd to Cu, and it confirms that for this lower palladium content systems, the electronic states of Pd were altered by the electronic interactions between Pd and Cu clusters. Similar positive shift in the Pd 3d band has been already shown for Pd-Co as well as for Pd-Cu systems.58,59 By deconvoluting this peak it is possible to estimate the reduction degree of Pd species on the surface (Table S5. For the monometallic Pd system, practically the 80% consists on Pd0. However, when Pd is incorporated with Cu, the amount of oxidized Pd increases. This trend can be related to the evolution of Cu species. As mentioned, the originally 23 photodeposited copper consists of a mixture of Cu0 and Cu+. However, as Pd is incorporated the fraction of Cu+ gradually diminishes. This could mean that palladium replacement primary affects Cu+ which are not able to reduce completely palladium ions. Then, as palladium addition is higher, copper mostly remains in the metallic state, while at the same time the reduction degree of palladium is also higher. The higher reduced fraction of Pd together with the lower amount of copper would explain the lower H2-TPR peak observed for Cu@Pd samples. From these considerations, bimetallic Cu@Pd system with intermediate compositions would exhibit a combination of metallic Cu and Pd which are present with some fraction of oxidised species. From combined interpretation of CO adsorption, TPR and XPS results it is possible to predict a strong interaction between incorporated palladium and copper which would support the idea that palladium is being deposited over copper clusters forming a core-shell structure at higher Pd content. This core-shell structuration have been also stated from TEM images. If we represent the calculated Pd/Cu ratio from XPS versus Pd/Cu obtained from ICP (Figure 4, lower row), it is possible to note a tentative linear evolution for low palladium contents. However, for higher Pd loadings, Pd/Cu from XPS analysis appear above the theoretical Pd/Cu from ICP. This behaviour could indicate that, for higher contents, palladium is covering Cu clusters which remains buried. This fact has been previously stated from TEM images (Figure 2) for Pd1.0Cu sample for which aggregate clusters can be envisaged. DFT Studies of Hydrogen Adsorption on Cu and Cu@Pd/TiO2 systems. DFT simulations of Cu, Pd and Cu@Pd modified model rutile TiO2 (110) have been performed to explore the atomistic details of the activity towards H2 production. Figure 5 shows the atomic structure of our model systems, composed of a 1D periodic rod-like 24 structure of the metal (Cu or Pd) supported on a perfect rutile (110) surface; a similar model was used by Koga et al in their investigation of CO oxidation on Au-modified TiO2.60,61 This structure is then modified as follows: (1) substituting a small concentration (2.5%) of Pd in the outermost Cu layer; (2) a core-shell bimetallic structure in which the outer layer of Cu is replaced by Pd. While there are clearly many ways to prepare these Cu@Pd models, those we use herein are representative of small Pd concentrations and a supported core-shell Cu@Pd structure. In these model structures, the metal binds to the rutile (110) surface through formation of new interfacial Cu-O bonds, with typical Cu-O distances of 1.95 – 2.10 Å; for comparison Cu-O distances in CuO and Cu2O are 1.95 and 1.86 Å. As a result of depositing the metal on the rutile substrate, there are distortions in the metal rod upon relaxation. Typically, there are strong distortions in the Cu-Cu distances. In the interface region, these cluster around 2.45 – 2.58 Å (bulk Cu-Cu distances are 2.56 Å) or Cu-Cu bonds are broken, with typical Cu-Cu distances of 2.63 – 2.90 Å. In the regions away from the interface, typical Cu-Cu distances are 2.48 – 2.61 Å, which are moderately distorted from the bulk Cu-Cu distance. Clearly the interaction with oxygen on TiO2 has a strong impact on the interface structure. Introduction of a single Pd atom into the surface layer is 2 eV more stable than introducing Pd in the interface region. This has a notable effect on the local Cu-Pd distances, which are 2.55 – 2.69 Å and are elongated when compared to the corresponding Cu-Cu distances. In the interface region, the Cu-O distances are 1.98 – 2.25 Å. In the Pd-shell- Cu-core structure, Pd-Pd distances in the outermost layer are 2.60 – 2.75 Å, typical of the longer Pd-Pd distances compared to Cu-Cu distances. The Cu-Pd distances range from 2.45 – 2.61 Å, while the Cu-O distances in the interface are 2.05 – 2.26 Å. 31 promote HER appears to be sensitive to the composition of the supported Cu@Pd bimetallic structure. These DFT results show that the incorporation of Pd into supported Cu structures can help promote the HER and that the coverage of hydrogen at the supported metal also plays an important role, as previously discussed for HER on metal phosphides and nitrides.71,72 Indeed, it has been found that for PtCu co-catalyst on TiO2 it has been observed a synergistic activity enhancement and a significantly higher activity toward photocatalytic H2 evolution with respect to the monometallic Pt- or Cu/TiO2 catalysts.73 These authors proposed that the enhanced activity could be due to Pt–Cu electronic interactions, where Cu increases the electron density on Pt, favoring a more efficient electron transfer for H2 evolution. We note that a small amount of Pd incorporated into Cu promotes H adsorption at all coverages examined, while the Cu-core-Pd-shell structure with a significantly higher Pd content promotes HER at the moderate hydrogen coverages and is less favourable at the lowest and highest H coverages. Similarly, the pure Pd system appears to be more effective for promoting HER at high hydrogen coverages, while Cu is particularly effective at low H coverages. One origin for this is that as H is adsorbed at the Pd-rich structures, Pd atoms neighbouring the adsorption site become partially reduced and this provides higher energy electronic states that can interact with H states and prevent overstabilisation of H atom adsorption. CONCLUSIONS By a simple deposition method, we have controlled the incorporation of increasing amount of palladium onto Cu/TiO2 system. Thus, by redox displacement it is possible to substitute Pd by Cu which has been previously photodeposited into TiO2 support. This controlled sequential deposition guarantees the dispersion and a strong interaction 32 between Cu and Pd. Moreover, the obtained bimetallic clusters showed a small size with a narrow distribution with a Cu@Pd core-shell structure. DFT results on models of Cu@Pd bimetallic 1D rod supported on rutile TiO2 show that the preference is for Pd to incorporate in the shell of the structure forming a Cu core-Pd shell structure. Investigation of the adsorption free energy of hydrogen studies clearly point out that the incorporation of Pd into supported Cu structures can help promote the HER and that the coverage of hydrogen at the supported metal also plays an important role. This way, the H2 production by isopropanol photoreforming reaction is strongly affected by the incorporation of even very low amounts of palladium. At long-term reaction times, the efficiencies obtained for [email protected] system is notably enhanced with respect to former Cu and Pd monometallic references. The formation of a core-shell structure would stabilize both metals against deactivation at longer reaction times. This result would demonstrate that the formation of a bimetallic core-shell nanostructuration, would be a promising strategy for developing more stable and highly active systems for H 2 production. ACKNOWLEDGMENTS GC acknowledges the financial support by PCIN-2017-056 project from M-ERA.Net programme through Acciones Programación Conjunta (Ministerio de Ciencia e Innovación). MN acknowledges support from Science Foundation Ireland through the Horizon 2020 M-ERA.net cofund program, project RATOCAT, grant numbers 685451 and SFI 17/M-ERA/3418 and access to computational resources at Tyndall and through the Irish Center for High End Computing (ICHEC), www.ichec.ie. 33 ASSOCIATED CONTENT. Supporting Information Tables from different characterization analysis. Additional Figures: Flow reaction scheme, evolution of Cu and Pd loading measured from ICP analysis, structural and surface characterization, TEM analysis of fresh and used Cu@Pd samples, diffuse reflectance UV-vis spectra. 34 Figure 1. Photocatalytic activity for isopropanol photoreforming: a) H2 production evolution with time for mono- and bimetallic Cu@Pd doped TiO2 systems 18 hours of reaction; b) H2 production after 18 hours of reaction for mono- and bimetallic Cu@Pd doped TiO2 systems. 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0 100 200 300 400 Cu 2+ Cu 0 ,Cu + Pd 2+ Pd 0 (b) Cu 2+ Cu 0 ,Cu + Pd 2+ Pd 0 Pd/TiO 2 Cu@Pd/TiO 2 Hydrogen yield (mmol/g) Nominal Pd loading (at%) Cu 2+ Cu 0 ,Cu + Pd 2+ Pd 0 @ 18 hours 0 2 4 6 8 10 12 14 16 18 0 50 100 150 200 250 300 350 Pd 0.1 Pd 0.2 Pd 0.3 Pd 0.4 Pd 0.5 Pd 1.0 H2 (mmol/g) Time (hours) 0 2 4 6 8 10 12 14 16 18 Cu 2.0 Cu@Pd 0.1 Cu@Pd 0.2 Cu@Pd 0.3 Cu@Pd 0.4 Cu@Pd 0.5 Cu@Pd 1.0 Time (hours) (a) 35 Figure 2. TEM/SEM analysis for: a) Cu2.0/TiO2; b) Pd1.0/TiO2; and c) [email protected]/TiO2 samples. b 024681012 0.0 2.0k 4.0k 6.0k 8.0k 10.0k 12.0k Cu Ni Ti Counts Energy (KeV) Ti a 012345678910 0 5 10 15 20 25 30 Average: 4.0 nm S.D. 1.1 Number of particles Diamter (nm) c 02468101214 0.0 5.0k 10.0k 15.0k 20.0k Ni Cu Pd Ti Counts Energy (KeV) Ti 0123456789 0 5 10 15 20 25 Average: 4.7 nm S.D. 0.9 Number of particles Diameter (nm) 36 Figure 3. HAADF−STEM image and EDX maps of Pd, Cu and Ti for Cu@Pd 0.3 /TiO 2 sample. 0246810 0 1 2 3 4 5 6 7 8 9 10 Net Intensity (counts) Position (nm) Ti Cu Pd 37 Figure 4. Upper row: XPS analysis: a) Cu 2p spectra; b) Pd 3d spectra; c) Cu LMM Auger spectra for different Cu@Pd/TiO2 systems. Lower row: Evolution of Pd/Cu ratio from XPS with respect to Pd/Cu ratio obtained from ICP analysis. 960 950 940 930 920 1500 2000 2500 3000 3500 Cu 0 , Cu 1+ Intensity (counts) Binding Energy (eV) Cu 2.0 /TiO 2 Cu@Pd 0.05 /TiO 2 Cu@Pd 0.10 /TiO 2 Cu@Pd 0.20 /TiO 2 Cu@Pd 0.30 /TiO 2 Cu@Pd 0.40 /TiO 2 Cu@Pd 0.50 /TiO 2 Cu@Pd 1.0 /TiO 2 (a) 0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 Cu@Pd 0.3 Cu@Pd 0.4 Cu@Pd 1.0 Pd/Cu from XPS Pd/Cu from ICP Cu@Pd 0.5 355 350 345 340 335 330 325 0 100 200 300 400 (a) Pd 4+ Pd 2+ Pd 0 Intensity (counts) Binding Energy (eV) Pd 1.0 TiO 2 Cu@Pd 1.0 /TiO 2 Cu@Pd 0.5 /TiO 2 Cu@Pd 0.4 /TiO 2 Cu@Pd 0.3 /TiO 2 Cu@Pd 0.2 /TiO 2 Cu@Pd 0.1 /TiO 2 Cu@Pd 0.05 /TiO 2 880 900 920 940 600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600 2800 3000 (c) Cu + Intensity (counts) Kinetic energy (eV) Cu 2.0 /TiO 2 Cu@Pd 0.05 /TiO 2 Cu@Pd 0.1 /TiO 2 Cu@Pd 0.2 /TiO 2 Cu@Pd 0.3 /TiO 2 Cu@Pd 0.4 /TiO 2 Cu@Pd 0.5 /TiO 2 Cu@Pd 1.0 /TiO 2 Cu LMM Cu 0 38 Figure 5. Atomic structure of rutile TiO 2 (110) modified with (a): Cu 42 , (b): Pd 42 , (c): Cu 41 Pd and (d) core-shell Cu 24 Pd 18 . Colour coding: Ti – grey spheres, O – red spheres, Cu – brown spheres and Pd – blue spheres. 39 Figure 6. Atomic structures for those adsorption structures for adsorbed 1 hydrogen atom with free energies between -0.15 and 0.15 eV, referenced to half the total energy of a free H 2 molecule. (a), (b) Cu/TiO 2 , (c) Pd/TiO 2 , (d), (e) 1PdCu/TiO 2 and (f) core-shell Cu@Pd/TiO 2 . Colour coding is the same as Figure 5, with the addition of white spheres for hydrogen atoms. 40 Figure 7. Atomic structures for H adsorption on Cu/TiO 2 . (a), (b) 2 adsorbed hydrogen atoms, (c), (d) 4 H (e), (f) 5 H. Colour coding is the same as Figure 6. 47 48. Schumann, J.; Kröhnert, J.; Frei, E.; Schlögl, R.; Trunschke, A. IR-Spectroscopic Study on the Interface of Cu-Based Methanol Synthesis Catalysts: Evidence for the Formation of a ZnO Overlayer. Top. Catal. 2017, 60, 1735–1743. 49. Spezzati, G.; Su, Y.; Hofmann, J.P.; Benavidez, A.D.; De La Riva, A.T.; McCabe, J.; Datye, A.K.; Hensen, E.J.M. Atomically Dispersed Pd–O Species on CeO2(111) as Highly Active Sites for Low-Temperature CO Oxidation. ACS Catal. 2017, 7, 6887–6891. 50. Ström, L.; Ström, H.; Carlsson, P.A.; Skoglundh, M.; Härelind, H. Catalytically Active Pd–Ag Alloy Nanoparticles Synthesized in Microemulsion Template. Langmuir 2018, 34, 9754–9761. 51. Chen, P.; Khetan, A.; Yang, F.; Migunov, V.; Weide, P.; Stürmer, S.P.; Guo, P.; Kähler, K.; Xia, W.; Mayer, J.; Pitsch, H.; Simon, U.; Muhler, M. Experimental and Theoretical Understanding of Nitrogen-Doping-Induced Strong Metal–Support Interactions in Pd/TiO2 Catalysts for Nitrobenzene Hydrogenation. ACS Catal. 2017, 7, 1197–1206. 52. Maimaiti, Y.; Nolan, M.; Elliott, S.D. Reduction Mechanisms of the CuO(111) Surface Through Surface Oxygen Vacancy Formation and Hydrogen Adsorption. Phys. Chem. Chem. Phys. 2014, 16, 3036-3046. 53. Kim, Y.; Kim, J.; Kim, D.H. Investigation on the Enhanced Catalytic Activity of a Ni-Promoted Pd/C Catalyst for Formic Acid Dehydrogenation: Effects of Preparation Methods and Ni/Pd Ratios. RSC Adv. 2018, 8, 2441– 2448. 54. Zhu, Y.; Liu, D.; Meng, M. H2 Spillover Enhanced Hydrogenation Capability of TiO2 Used for Photocatalytic Splitting of Water: A Traditional Phenomenon for New Applications. Chem. Comm. 2014, 50, 6049–6051 55. Noronha, F.B.; Schmal, M.; Primet, M.; Frety, R. Characterization of Palladium-Copper Bimetallic Catalysts Supported on Silica and Niobia. Appl. Catal. 1991, 78, 125–139. 56. Batista, J.; Pintar, A.; Mandrino, D.; Jenko, M.; Martin, V. XPS and TPR Examinations of γ-Alumina- Supported Pd-Cu catalysts. Appl. Catal. A: Gen. 2001, 206, 113–124. 57. Espinós, J.P.; Morales, J.; Barranco, A.; Caballero, A.; Holgado, J.P.; González-Elipe, A.R. Interface Effects for Cu, CuO, and Cu2O Deposited on SiO2 and ZrO2. XPS Determination of the Valence State of Copper in Cu/SiO2 and Cu/ZrO2 Catalysts. J. Phys. Chem. B, 2002, 106, 6921–6929. 58. Ye, S.H.; Feng, J.X.; Li, G.R. Pd Nanoparticle/CoP Nanosheet Hybrids: Highly Electroactive and Durable Catalysts for Ethanol Electrooxidation. ACS Catal. 2016, 6, 7962–7969. 59. Ji, Y.; Ying, Y.; Pan, Y.; Li, M.; Guo, X.; Wu, Y.; Wen, Y.; Yang, H. Palladium Networks Decorated by Cuprous Oxide for Remarkably Enhanced Electrocatalytic Activity of Methanol Oxidation Reaction with High CO-Tolerance. J. Power Sources, 2016, 329, 115–122. 60. Koga, H.; Tada, K.; Okumura, M. DFT Study of CO Oxidation over Au/TiO2(1 1 0): The Extent of the Reactive Perimeter Zone. Chem. Phys. Lett. 2014, 610, 76-81 61. Koga, H.; Tada, K.; Okumura, M. Density Functional Theory Study of Active Oxygen at the Perimeter of Au/TiO2 Catalysts. J. Phys. Chem. C 2015, 119, 25907- 25916 62. Strayer, M.E.; Senftle, T.P.; Winterstein, J.P.; Vargas-Barbosa, N.M.; Sharma, R.; Rioux, R.M.; Janik, M.J. Mallouk; T.E. Charge Transfer Stabilization of Late Transition Metal Oxide Nanoparticles on a Layered Niobate Support. J. Am. Chem. Soc. 2015, 137, 16216–16224. 63. Chen, L.J.; Tang, Y.; Cui, L.; Ouyang, C.; Shi, S. Charge Transfer and Formation of Ce3+ upon Adsorption of Metal Atom M (M = Cu, Ag, Au) on CeO2 (100) Surface. J. Power Sources 2013, 234, 69-81. 48 64. Zhu, K.J.; Liu, J.; Yang, Y.J.; Xu, Y.X.; Teng, B.T.; Wen, X.D.; Fan, M. A Method to Explore the Quantitative Interactions Between Metal and Ceria for M/CeO2 Catalysts. Surf. Sci., 2018, 669, 79-86. 65. Maimaiti, Y.; Nolan, M.; Elliott, S.D. Phys. Chem. Chem. Phys. 2014, 16, 3036-3046. 66. Sharma, P.K.; Cortes, M.A.L.R.M.; Hamilton, J.W.J.; Han, Y.; Byrne, J.A.; Nolan, M. Catalysis Today 2019, 321, 9–17. 67. Jäger, M.O.J.; Morooka, E.V.; Federici Canova, F.; Himanen, L.; Foster, A.S. Machine Learning Hydrogen Adsorption on Nanoclusters Through Structural Descriptors, npj Computational Materials, 2018, 4, 37. 68. Hu, G.; Tang, Q.; Jiang, D.E. CoP for Hydrogen Evolution: Implications from Hydrogen Adsorption. Phys. Chem. Chem. Phys., 2016, 18, 23864–23871. 69. Bonde, J.; Moses, P.G.; Jaramillo, T.F.; Nørskov, J.K.; Chorkendorff, I. Hydrogen Evolution on Nano- Particulate Transition Metal Sulphides. Faraday Discuss. 2009, 140, 219–231. 70. Tsai, C.; Chan, K.; Abild-Pedersen, F.; Nørskov, J.K. Active Edge Sites in MoSe2 and WSe2 Catalysts for the Hydrogen Evolution Reaction: A Density Functional Study. Phys. Chem. Chem. Phys., 2014, 16, 13156-13164. 71. Abghoui, Y.; Skúlason, E. Hydrogen Evolution Reaction Catalyzed by Transition-Metal Nitrides. J. Phys. Chem. C 2017, 121, 24036–24045. 72. Partanen, L.; Hakala, M.; Laasonen, K. Hydrogen Adsorption Trends on Various Metal-Doped Ni2P Surfaces for Optimal Catalyst Design. Phys. Chem. Chem. Phys., 2019, 21, 184–191. 73. Shahvaranfard, F.; Ghigna, P.; Minguzzi, A.; Wierzbicka, E.; Schmuki, P.; Altomare, M. Dewetting of PtCu Nanoalloys on TiO2 Nanocavities Provides a Synergistic Photocatalytic Enhancement for Efficient H2 Evolution. ACS Appl. Mater. Interfaces 2020, 12, 38211–38221.