Precise 3D structure determination of Cu single atoms on an α-Al₂O₃(0001) surface by polarization-dependent total reflection fluorescence X-ray absorption fine structure and first-principles calculations

Abstract

The valence state and three-dimensional (3D) structure of vacuum-deposited Cu species on an Al-terminated (1 × 1) α-Al₂O₃(0001) surface have been studied using polarization-dependent total reflection fluorescence X-ray absorption fine structure (PTRF-XAFS) and density functional theory (DFT) calculations. The Cu species were atomically dispersed on the α-Al₂O₃(0001) surface in its low-coverage region (0.07 ML on Cu/O basis) with a Cu–O distance of 0.194 ± 0.003 nm and were present in the form of Cu(I). The adsorption site for the Cu single atoms was determined to be a three-fold hollow oxygen site corresponding to the imaginary Al position in the next Al layer over the α-Al₂O₃(0001) surface, accompanied by an upward shift of the nearest neighbor oxygen atoms and expansion of the oxygen–oxygen distance. A similar local structure and surface relaxation mechanism have been reported in our previous PTRF-EXAFS study on Ni single atoms deposited on the α-Al₂O₃(0001) surface (K. Ijima et al., Chem. Phys. Lett., 2004, 384, 134–138), indicating important implications for stabilizing single metal atoms on the Al₂O₃ surface.

---

1. Introduction

Single-atom catalysts (SACs), which consist of isolated single metal atoms dispersed on a powdery oxide support, have attracted much attention due to their high catalytic activity and cost-savings on precious metal atoms, compared with those of conventional nanoparticle catalysts.^1–3 For instance, the Cu single-atom sites anchored to γ-Al₂O₃ oxides are highly active and stable in continuous conversion of methane-steam to methanol at 200 °C, comparable to most of the state-of-the-art Cu–zeolite catalysts.⁴ The Cu single atoms immobilized on TiO₂ also show high activity and excellent stability for photocatalytic hydrogen evolution.⁵ Since the catalytic performances of such SACs significantly depend on the coordination environment around the active metal site, i.e., metal–oxygen bond distance, metal–oxygen coordination number, coordination symmetry, and valence state of the active metal atoms, etc., unraveling their precise 3D structures is essential to understanding the stabilization mechanism of the single metal atoms and their structure–activity relationship for the development of more active SACs.

X-ray absorption fine structure (XAFS) is one of the most powerful techniques to investigate the structure of the SACs because it provides information about the active metal atom on the valence state through X-ray absorption near edge structure (XANES) analysis and the local structure through extended X-ray absorption fine structure (EXAFS) analysis. However, it is difficult to determine their precise 3D structure by using conventional XAFS techniques because the bonding information is averaged over all directions and provides a 1D radial distribution around the X-ray absorbing atom. Moreover, the powdery oxide-support surfaces are generally ill-defined, exposing various defects and crystal faces, and this often causes formation of non-uniform single metal species, making it difficult to accurately determine their structures. On the other hand, when a well-defined single-crystal oxide surface is used as a support, detailed information about the 3D structure and metal–support interaction for uniform single-metal species can be obtained using the polarization dependence shown in eqn (1),⁶

where χ_pol(k) is the polarization-dependent EXAFS signal as a function of wavenumber (k), θ_i is the angle between the electric vector of the incident X-rays and the i-th bond direction, and χ_i(k) is the EXAFS oscillation originating from the i-th bond. The problems of a low concentration of surface metal species can be overcome by applying the surface-sensitive polarization-dependent total reflection fluorescence (PTRF)-XAFS technique,^6–8 which has been used to determine the 3D structures of single metal atoms^9–14 and metal nanoclusters^15–18 on TiO₂(110) and α-Al₂O₃(0001) surfaces, based on EXAFS simulation of real-space model structures using the FEFF code.¹⁹ For instance, the Cu single atoms were formed on a TiO₂(110) surface premodified with o-mercaptobenzoic acid (o-MBA), via forming bonds with the sulfur atom of the o-MBA molecule and the oxygen atom in the TiO₂ lattice, and their 3D structure was precisely determined at less than 0.003 nm resolution in the Cu–S and Cu–O bond distances (0.219 and 0.185 nm, respectively) and at less than 3° precision in the Cu–S and Cu–O bond angles to the surface normal (45° and 43°, respectively).¹² Our previous polarization-dependent total reflection fluorescence (PTRF)-EXAFS studies on single metal atoms deposited on bare oxide single-crystal surfaces have given an important concept for atomic dispersion: a single metal atom is stabilized on an oxide surface through metal–oxygen interactions, not through metal–cation bonding, because no EXAFS oscillation corresponding to a metal–cation bond with a bond distance less than 0.27 nm has been observed.⁷

In this work, we have investigated the valence state and 3D structure of vacuum-deposited Cu on an α-Al₂O₃(0001) surface in its low-coverage region by using the PTRF-XAFS technique and density functional theory (DFT) calculations, to unveil the metal–oxide interface structure and formation of single-atom species. The Cu/α-Al₂O₃(0001) system has been extensively studied using X-ray photoelectron spectroscopy (XPS),^20–22 scanning force microscopy (SFM),²³ and first-principles calculations.^24–29 The XPS studies indicated formation of Cu(I) at a Cu coverage of less than ∼0.35 ML (on a Cu/O basis) due to interaction of oxygen atoms of the α-Al₂O₃(0001) surface.²² First-principles calculations predicted that the Cu atoms are adsorbed preferentially on the three-fold hollow oxygen sites.^25–27,29 However, the precise 3D structure of the Cu species has not yet been directly determined experimentally. We carried out EXAFS simulations using real-space model structures obtained by the DFT calculations and determined the precise 3D structure of the Cu species by modifying the DFT-optimized structures so that the simulated EXAFS spectra reproduced the observed EXAFS spectra.

2. Experimental

2.1 Sample preparation

As-received α-Al₂O₃(0001) single crystals (15 × 15 × 0.5 mm³, Shinkosha Co., Ltd) were annealed in air at 1323 K for 3 h to obtain an atomically flat α-Al₂O₃(0001) surface.³⁰ The sample was then introduced into an ultra-high vacuum (UHV) chamber and heated at 673 K for 1 h. Afterwards the α-Al₂O₃(0001) surface was exposed to oxygen plasma (10 W, 10 min) for further surface cleaning, generated by a compact electron cyclotron resonance (ECR) plasma source (Aura-Wave, SAIREM). We have found that the O₂ plasma treatment efficiently removed the residual carbon species and gave a sharper low-energy electron diffraction (LEED) pattern with a hexagonal (1 × 1) periodicity, as shown in Fig. S2. Cu was then vacuum-deposited on the α-Al₂O₃(0001) surface at room temperature via resistive heating of a Cu wire (99.999% purity, Nilaco Co., Japan) in the UHV chamber. The Cu coverage was estimated to be 0.07 ML from the XPS peak intensity ratio of Cu 2p_3/2 to O 1s, where 1 ML was defined as 1.5 × 10¹⁵ cm⁻² with correspondence to the density of the surface O atoms of the α-Al₂O₃(0001) substrate. The Cu/α-Al₂O₃(0001) sample was then transferred to the compact UHV cell (operando PTRF-XAFS cell)¹⁸ for the PTRF-XAFS measurements.

2.2 PTRF-XAFS measurements

PTRF-XAFS measurements were conducted at the BL-9A beamline of the photon factory at the Institute of Materials Structure Science (KEK-IMSS-PF, Tsukuba, Japan) after the PTRF-XAFS cell was fixed on a 6-axis goniometer placed in the hutch of the beamline. The storage ring energy and ring current were 2.5 GeV and 450 mA, respectively. X-rays were monochromatized using Si(111) double crystals and focused onto the sample with double-bent conical mirrors. The beam size on the sample was regulated via a molybdenum pinhole with a diameter of 0.4 mm, which was attached in front of the ion chamber upstream (I₀), to reduce undesirable irradiation outside of the sample surface. The total reflection conditions were adjusted using the 6-axis goniometer. The PTRF-XAFS measurements were carried out at room temperature under the UHV conditions for two different orientations of the sample surface relative to the electric vector of the incident X-rays, i.e., parallel to the surface (s-polarization) and perpendicular to the surface (p-polarization). The Cu Kα fluorescence was detected by a 7-element silicon drift detector (XSDD50-07, TechnoAP, Japan). EXAFS analysis was performed using the REX 2000 (Rigaku Co., Japan) package.³¹ The EXAFS oscillations were extracted using a spline smoothing method and normalized with respect to the edge height with the energy dependence expressed in a Victoreen equation. The 3D structure was determined via an iterative method using the FEFF8.04 code¹⁹ with a Hedin–Lundqvist potential based on a real-space model structure. The goodness of fit between the observed (χ_obs(k)) and calculated (χ_cal(k)) EXAFS oscillations was evaluated using eqn (2),where N and ε(k) are the number of data points and the error of the observed spectrum, respectively. The structure model was adopted when R² values were lower than 1 for all orientations (s- and p-polarizations). Details of the PTRF-EXAFS analysis are provided in the SI.

2.3 Density functional theory (DFT) calculations

DFT calculations were performed using VASP 5.4^32–35 with the projector augmented wave (PAW) and plane-wave basis set methods.³⁶ The kinetic energy cutoff for plane-wave expansion was set at 500 eV, and the exchange–correlation interaction between electrons was described by the Perdew–Burke–Ernzerhof functional (PBE).^37,38 The calculated lattice constants for the bulk α-Al₂O₃ (using 4 × 4 × 2 k-points, a = b = 4.778 Å and c = 13.034 Å) are found to be in good agreement with the experimental values (a = b = 4.758 Å and c = 12.990 Å).³⁹ The supercell was obtained by truncating the bulk α-Al₂O₃ structure, leaving one Al layer exposed with an oxygen layer right beneath it. This surface is known as the single Al layer terminated surface and is accepted to be the most stable unreconstructed (1 × 1) α-Al₂O₃(0001) surface observed in the UHV conditions.^40–42 The formation of this surface was confirmed in this study by the hexagonal (1 × 1) periodicity of the LEED pattern (Fig. S2). The Al-terminated (1 × 1) surface was modeled as a twelve-Al-layer slab with a (3 × 3) supercell, which consists of 108 Al atoms and 162 O atoms. The vacuum layer was set at 20 Å. The bottom three Al layers and two O layers were fixed throughout the calculations, while the upper nine Al layers, four O layers, and the Cu atom were relaxed using a conjugate gradient technique until the residual forces on the atoms were less than 0.005 eV Å⁻¹. The k-point mesh was set as 3 × 3 × 1. The effects of van der Waals forces were considered by the vdw-D3 method.⁴³ The dipole correction along the [0001] direction was included. The adsorption energy (E_ad) of single Cu atoms on the α-Al₂O₃(0001) was calculated as

E_ad = E_Al2O3+Cu − E_Al2O3 − E_Cu (3)

where E_Al2O3, E_Cu, and E_Al2O3+Cu are the total electron energies of α-Al₂O₃(0001), gaseous Cu atom, and α-Al₂O₃(0001) after Cu atom adsorption, respectively.

3. Results and discussion

3.1 Valence state of the Cu species on α-Al₂O₃(0001)

Fig. 1a and b show the PTRF-XANES spectra of the vacuum-deposited Cu species on the α-Al₂O₃(0001) surface and the XANES spectra of Cu foil, Cu₂O, and CuO as the references, respectively. The XANES spectra of the Cu/α-Al₂O₃(0001) showed different features from that of Cu foil, which suggested formation of non-metallic Cu species. The mid-edge features observed at approximately 8983 eV for the s- and p-polarizations could be assigned to the 1s to 4pπ* transition, as determined from previous XANES results and ab initio calculations.⁴⁴ The inflection points of the edge appeared at around 8980.4 eV for both polarization directions. Considering that the inflection points of Cu foil, Cu₂O, and CuO were at 8979.5, 8980.6, and 8984.7 eV, respectively, the oxidation state of the Cu species on the α-Al₂O₃(0001) was monovalent (Cu(I)). Kelber et al. demonstrated that Cu was oxidized to Cu(I) when Cu was sputter-deposited on the α-Al₂O₃(0001) surface at a coverage of less than ∼0.35 ML (on a Cu/O basis), as evidenced by the Cu(LMM) line shape data obtained via X-ray photoelectron spectroscopy (XPS) measurements.²² A previous first-principles study also indicated that Cu(I) was formed at coverages less than 1/3 (on a Cu/O basis) by interacting with surface oxygen atoms of the α-Al₂O₃(0001).²⁵

Fig. 1 (a) Cu K-edge PTRF-XANES spectra of Cu/α-Al₂O₃(0001) in s- and p-polarization directions. (b) Cu K-edge XANES spectra of reference compounds.

3.2 Three-dimensional (3D) structure of the Cu species on α-Al₂O₃(0001)

Fig. 2a and b show the PTRF-EXAFS spectra (χ_s-pol(k) and χ_p-pol(k)) of the Cu/α-Al₂O₃(0001) and the EXAFS spectra of the Cu reference compounds, respectively. In Fig. 2a, a slight polarization dependence was found around 50 nm⁻¹ between the two PTRF-EXAFS spectra. The envelopes of the PTRF-EXAFS oscillations are quickly damped in the higher k region, compared to that for Cu foil, which indicates that the nearest neighbor atom of Cu is not Cu, but a lighter atom such as O and Al. The curve fitting results in Table S1 showed no contribution from Cu–Cu interaction, which suggests that no Cu aggregation occurred, while Cu–O interaction (0.194 ± 0.003 nm) was dominant. The coordination number for the Cu–O interaction in the average PTRF-EXAFS spectrum χ_avg(k), which was calculated as (2 × χ_s-pol(k) + χ_p-pol(k))/3, was estimated to be 2.5 ± 0.8, which means that the Cu single atoms form bonds with two or three oxygen atoms on the Al₂O₃(0001) surface.

Fig. 2 (a) Cu K-edge PTRF-EXAFS spectra (χ_s-pol(k) and χ_p-pol(k)) of the Cu species on the α-Al₂O₃(0001) surface. (b) Cu K-edge EXAFS spectra of reference compounds.

DFT calculations were conducted to elucidate possible adsorption sites of the Cu single atoms on the α-Al₂O₃(0001) surface. Fig. 3 shows the optimized structure of the Al-terminated α-Al₂O₃(0001) surface. As illustrated in Fig. 3b, the outmost Al atoms (Al₁) largely relax inward by 88%, compared with those in the bulk-terminated surface, and are practically co-planar with surface oxygen atoms (O₁), which agrees with previous calculation results.^45,46 Based on the curve fitting results showing that the Cu–O coordination number in the average PTRF-EXAFS spectrum was 2.5 ± 0.8, we considered the bridge and hollow oxygen sites on the α-Al₂O₃(0001) surface as candidates of adsorption sites for the Cu single atoms. The α-Al₂O₃(0001) surface has two types of oxygen hollow sites (H₁ and H₂) and two types of oxygen bridge sites (B₁ and B₂), as illustrated in Fig. 3a. The O₁–O₁ distances of the equilateral oxygen triangles consisting of H₁ and H₂ sites were 0.261 and 0.274 nm, respectively. B₁ and B₂ are the sites bridging the two oxygen atoms of the H₁ and H₂ sites, respectively. We put a Cu atom on each of the corresponding positions and calculated the adsorption energies after the structure optimization, which are listed in Table 1. The Cu atoms on the B₁ and B₂ sites moved to the H₁ site after the structure optimization, and we found that the H₁ and H₂ sites were energetically stable. The adsorption energies at the H₁ and H₂ sites were calculated to be −2.54 and −2.14 eV, indicating that the H₁ site is preferred. The H₁ site corresponds to the imaginary Al position in the next Al layer over the α-Al₂O₃(0001) surface. Previous DFT calculations have also indicated that the H₁ site is most stable for adsorption of the Cu single atoms.^25–27 Fig. 4a and b show the optimized structures for the Cu single atoms adsorbed on the H₁ and H₂ sites, respectively.

Fig. 3 Portion of the Al terminated α-Al₂O₃(0001) surface showing the adsorption sites. The atomic positions are determined by DFT calculation. (a) top view and (b) side view. Dark blue, blue and light blue spheres represent surface Al atoms (Al₁), Al atoms in the 2nd and 3rd Al layer (Al₂ and Al₃), respectively. Red and orange spheres represent surface O atoms (O₁) and O atoms in the 2nd O layer (O₂), respectively.

Table 1 Adsorption energy of Cu single atoms on the α-Al₂O₃(0001) surface

	H1	H2	B1	B2
Adsorption energy/eV	−2.54	−2.14	Relaxed to H1	Relaxed to H1

---

Fig. 4 DFT-optimized structures for single Cu atoms on (a) H₁ site and (b) H₂ site of the α-Al₂O₃(0001) surface. Brown spheres: Cu atoms. The dark blue, blue and light blue spheres represent the surface Al atoms (Al₁), the Al atoms in the 2nd and 3rd Al layer (Al₂ and Al₃), respectively. The red and orange spheres represent the surface O atoms (O₁) and the O atoms in the 2nd O layer (O₂), respectively.

Since the atomic coordinates of Cu, O, and Al atoms were obtained from the DFT-optimized structures in Fig. 4, we conducted theoretical Cu K-edge EXAFS simulations using the FEFF code to confirm if the simulations reproduced the observed EXAFS spectra. In the simulations, we considered the surrounding atoms around Cu which were located within 0.3 nm because the contribution from the more distant atoms is small and negligible in the EXAFS oscillations. Fig. 5a and b show the simulated EXAFS spectra obtained for the Cu single atoms on the H₁ and H₂ sites, respectively, and their comparisons with the observed EXAFS spectra. Three O₁ atoms and one Al₂ (or Al₃) atom were included in the EXAFS simulations, and their atomic coordinates are listed in Table S2. The R² values for the s- and p-polarizations in the H₁ site model were 1.6 and 10.3, respectively, and those in the H₂ site model were 2.6 and 17.5, respectively, showing that the H₁ site provides a better fit although the corresponding R² values are larger than 1. The primary distinctions between the H₁ and H₂ site models pertain to the Cu–Al and Cu–O bond distances, which are 0.267 and 0.249 nm for the Cu–Al bond, and 0.207 and 0.214 nm for the Cu–O bond, respectively. The shorter Cu–Al bond distance in the H₂ site model made a larger contribution to the total EXAFS oscillations, resulting in lowering of the degree of the fit between the observed and simulated spectra, as indicated by the curve fitting results in Table S1 showing that the Cu–O oscillations are dominant and contribution from the Al atoms should be small in the observed spectra. Moreover, the longer Cu–O bond distance in the H₂ site model contributes to the further reduced degree of the fit, as evidenced by the larger discrepancy with the experimentally determined Cu–O bond distance (0.194 ± 0.003 nm). Considering the EXAFS simulation results and the adsorption energies in Table 1, we concluded that the H₁ site model is most plausible. However, the simulation based on the H₁ model did not adequately reproduce the observed spectra, as indicated by the substantial R² values. One reason for this discrepancy is the Cu–O₁ bond distance (0.207 nm), which is slightly longer than the 0.194 ± 0.003 nm obtained by the curve fitting results. A secondary rationale pertains to the position of the Al₃ atom, which is located less than 0.27 nm from the Cu atom. This is in contrast to our previous PTRF-EXAFS results on single metal atoms on an oxide surface showing no observable EXAFS oscillation corresponding to metal–cation bond with a bond distance less than 0.27 nm, as already mentioned. And it is imperative to optimize the Cu–O₁ bond angle to the surface normal (in the [0001] direction) to accurately reproduce the amplitudes of the EXAFS oscillation in both the s- and p-polarizations, as demonstrated in eqn (1). Then we allowed the positions of the Cu, O₁, and Al₃ atoms to move from the original positions by keeping the three-fold symmetry and fixing the Cu–O₁ bond at 0.194 nm. We systematically changed the distances of the Cu and Al₃ atoms from the O₁ triangle plane, which are defined as variables d₁ and d₂, respectively, and simulated the EXAFS oscillations (see Fig. S4). Then we calculated the R² values for s- and p-polarization EXAFS spectra as functions of d₁ and d₂. The acceptable ranges of d₁ and d₂ were determined as those of the region where the R² values for both the s- and p-polarizations were less than 1. Consequently, that region was successfully found, where d₁ lay within the range of 0.116 nm ± 0.001 nm, and d₂ lay within the range of 0.156 nm ± 0.002 nm. Thus, we concluded that the H₁ site is the most reasonable both from the EXAFS simulations and DFT calculations. A previous PTRF-EXAFS study on vacuum-deposited Ni atoms on α-Al₂O₃(0001) surface showed that the Ni atoms were atomically dispersed and stabilized on the H₁ site, accompanied that the O₁ triangle shifted upwards and was expanded.⁹ We then tested a similar model structure as shown in Fig. 6a, where the Cu single atom on the H₁ site is located at d₁ = 0.116 nm and d₂ = 0.156 nm. The Cu–O₁ and Cu–Al₃ bond distances are 0.194 nm and 0.272 nm, respectively. This model structure well reproduced the observed EXAFS oscillations for both the s- and p-polarizations, as indicated in Fig. 6b. For Fig. 6a, we assumed that the surface atoms such as Cu and three-fold O₁ atoms moved from their original positions while the Al₃ atom in the bulk kept its position (see Fig. 4a for the original structure). Three-fold O₁ atoms moved upward by 0.032 nm and the O₁–O₁ distance was expanded from 0.261 nm to 0.269 nm. The Cu–Al₃ distance was extended from 0.268 nm to 0.272 nm. The upward shift of the oxygen triangle can be understood as “surface lifting”, which is the surface structural change induced by adsorption of atoms or molecules, particularly the upward displacement of surface atoms. A previous surface X-ray diffraction (SXRD) study of the Cu overlayer on TiO₂(110) suggested a significant upward displacement (0.065 nm) of the topmost oxygen atoms after vacuum deposition of Cu.⁴⁷ In our case, the chemical bond formation between Cu and the triangle oxygen atoms of the α-Al₂O₃(0001) surface induced the upward displacement (0.032 nm) of the oxygen atoms. The lateral expansion of the oxygen triangle may be attributed to the propensity of three-coordinate Cu(I) complexes to adopt a trigonal planar structure rather than a trigonal pyramidal one, as evidenced by formation of the trigonal planar Cu(I) complexes such as [Cu(CN)₃]²⁻,⁴⁸ [Cu(C₆H₇N)₃]⁺,⁴⁹ and [Cu(etu)₃]⁺ (tris(ethylenethiourea)copper(I)).⁵⁰ It is worth considering the reasons why the structure obtained from the DFT calculations (the H₁ site model in Fig. 5(a)) had to be modified to reproduce the observed PTRF-EXAFS spectra using the theoretical EXAFS simulation. The discrepancy in the local structure around the Cu single atoms, as determined by zero-temperature DFT calculations and room-temperature PTRF-EXAFS measurements, may be attributed to the presence of large, anharmonic vibrations in the Cu–O and Cu–Al bonds.^51,52 Given that such vibrations were not taken into account in our zero-temperature DFT calculations, the local surface relaxation observed in our study at room temperature could not be accurately predicted.

Fig. 5 The simulated EXAFS spectra (red dotted lines) obtained by assuming (a) H₁ site and (b) H₂ site for the Cu single atoms and their comparison with the observed EXAFS spectra (black solid lines). The brown spheres are the Cu atoms. The light blue and blue spheres represent surface the Al atoms in the 3rd and 2nd Al layer (Al₃ and Al₂), respectively. The red spheres show the surface O atoms (O₁).

Fig. 6 (a) Proposed structure for the Cu single atom at H₁ site with the in-plane expansion and upward shift of the oxygen atoms. Circles with broken lines indicate the oxygen atom positions before such relaxation, shown in Fig. 4a. (b) Comparison between the observed and simulated PTRF-EXAFS spectra for the Cu/α-Al₂O₃(0001) (black solid and red dotted lines, respectively) using the structure model shown in (a). The brown spheres are the Cu atoms. The dark blue, blue and light blue spheres represent the surface Al atoms (Al₁), the Al atoms in the 2nd and 3rd Al layer (Al₂ and Al₃), respectively. The red and orange spheres show the surface O atoms (O₁) and the O atoms in the 2nd O layer (O₂), respectively.

The coverage of the H₁ adsorption sites of the α-Al₂O₃(0001) surface that can stabilize the Cu single atoms is 0.33 ML, which may represent the maximum coverage of the Cu single atoms. However, as the Cu loading increases, the limited availability of these specific sites may result in aggregation of vacuum-deposited Cu atoms through their diffusion and collision even at a lower coverage than 0.33 ML. Therefore, the conclusions of the present study are strictly applicable to the low-coverage regime.

3.3 Effect of surface hydroxyl groups on the Cu structure

It has been reported that the hydroxyl groups on the α-Al₂O₃(0001) surface affect the local structure and energetic stability of the deposited Cu atoms.^20–22 The analysis of the XPS O 1s peak of the α-Al₂O₃(0001) surface suggested that the OH coverage was 0.11 ML (see Fig. S5), which may not be negligible. We then performed additional DFT calculations to investigate the effect of the surface hydroxyl groups on the structure and stability of the Cu single atoms. The origin of the hydroxyl groups may come from dissociative adsorption of residual water molecules in the UHV chamber since additional water exposure results in increase in their coverage.^20,21 Therefore, we considered the water-dissociated α-Al₂O₃(0001) surface, where dissociated OH and H are absorbed at Al₁ and the nearest O₁ atoms, respectively (the OH coverage is 0.07 ML). Three energetically stable sites for the Cu atoms were found, which are the H₁ site, the T₁ site consisting of H₁ and O of Al₁–OH, and the B₃ site bridging O₁ and O of Al₁–OH. The corresponding optimized structures and adsorption energies are shown in Fig. S6. The Cu atoms placed on the T₁ and B₃ sites were more stable than those on the H₁ site probably because they interact with the hydroxyl oxygen. However, they were unlikely to form or would be minor species for the following reasons: (1) the Cu atoms on the T₁ sites interact with four oxygen atoms, which may contradict the curve-fitting results in Table S1 showing that that the Cu–O coordination number is 2.5 ± 0.8. (2) The Cu–Al bond distance in the B₃ site model (0.260 nm) is shorter than that in the H₁ site model (0.267 nm). This is in contrast to the results of the present study, which indicate that the Cu–O oscillations are dominant and the contribution from the Al atoms should be small. Therefore, we concluded that the Cu single atoms on the H₁ sites were the dominant Cu species under the applied experimental conditions. In our experiments, the coverages of the OH groups on the α-Al₂O₃(0001) surface and of the evaporated Cu atoms were low at 0.11 ML and 0.07 ML, respectively. Since diffusion of the evaporated Cu atoms on the α-Al₂O₃(0001) surface was limited, the probability of them encountering OH groups was very low. Consequently, most of the Cu single atoms might be stabilized on the H₁ sites.

4. Conclusions

This work experimentally investigated the valence state and 3D structure of the low-coverage Cu species deposited on α-Al₂O₃(0001) at room temperature under the UHV conditions. The valence state of the Cu species was determined to be monovalent (Cu(I)) according to the PTRF-XANES spectra. The Cu species was found to be atomically dispersed on the α-Al₂O₃(0001) surface, as determined by the absence of Cu–Cu interaction in the curve fitting analysis of the PTRF-XAFS spectra. DFT calculations showed that the three-fold hollow oxygen site above the third layer Al atoms was energetically stable for adsorption of the Cu single atom. Local structure modification around the Cu single atom in the DFT-optimized structure was conducted so that the simulated EXAFS spectra reproduce the observed spectra well in the two polarization directions. We proposed local surface relaxation around the Cu atom, where the three-fold oxygen triangle moved upward by 0.032 nm and the oxygen–oxygen distance in the oxygen triangle expanded from 0.261 nm to 0.269 nm. Such relaxation suggests a strong metal–support interaction that can effectively anchor the Cu single atom, thereby suppressing thermal diffusion and aggregation even under reaction conditions. In addition, the modified Cu–O coordination geometry may influence the local electronic structure of the Cu center, which is expected to be relevant for elementary steps involving charge transfer or bond activation in catalytic reactions. Although catalytic measurements are beyond the scope of the present work, the identified adsorption site and relaxation mechanism provide an atomistic basis for understanding the stability and functionality of Cu single-atom catalysts on oxide surfaces.

Conflicts of interest

There are no conflicts to declare.

Data availability

The data supporting this article have been included as part of the supplementary information (SI). Supplementary information is available. See DOI: https://doi.org/10.1039/d5cp03666f.

Acknowledgements

The XAFS measurements have been carried out under the approval of PF-PAC (PF advisory committee, PAC Proposal No. 2020G643 and 2022G576). This study was financially supported by the CREST Project (No. JPMJCR19R3) of the Japan Science and Technology Agency (JST), Kakenhi Grants-in-Aid (No. 21H01802, 24H00445, and 24KJ0008), and a Grant-in-Aid for Transformative Research Areas (A) (No. 22H05109) from the Japan Society for the Promotion of Science (JSPS). The computational work was partially supported by JSPS KAKENHI Grant Number 24H02219 (M. G.) in Transformative Research Areas (A) 24A202 Integrated Science of Synthesis by Chemical Structure Reprogramming (SReP). The computational results in this study were achieved through the MANABIYA (ACADEMIC) program conducted by the Institute for Chemical Reaction Design and Discovery (ICReDD), Hokkaido University, which was established by the World Premier International Research Initiative (WPI), MEXT, Japan.

References

1. X.-F. Yang, A. Wang, B. Qiao, J. Li, J. Liu and T. Zhang, Acc. Chem. Res., 2013, 46, 1740–1748.
2. L. Liu and A. Corma, Chem. Rev., 2018, 118, 4981–5079.
3. S. K. Kaiser, Z. Chen, D. Faust Akl, S. Mitchell and J. Pérez-Ramírez, Chem. Rev., 2020, 120, 11703–11809.
4. H. Zhang, X. Zhang, X. Shi, X. Liu, J. Liang, S. Li, S. Ma, X. Xiao, J. Guo, L. Zhang and M. Tan, Appl. Catal., B, 2025, 361, 124648.
5. Y. Zhang, J. Zhao, H. Wang, B. Xiao, W. Zhang, X. Zhao, T. Lv, M. Thangamuthu, J. Zhang, Y. Guo, J. Ma, L. Lin, J. Tang, R. Huang and Q. Liu, Nat. Commun., 2022, 13, 58.
6. K. Asakura, Polarization-dependent total reflection fluorescence extended X-ray absorption fine structure and its application to supported catalysis, in Catalysis, RSC Publishing, 2012, pp. 281–322.
7. S. Takakusagi, W.-J. Chun, H. Uehara, K. Asakura and Y. Iwasawa, Top. Catal., 2013, 56, 1477–1487.
8. S. Takakusagi, Y. Iwasawa and K. Asakura, Chem. Rec., 2019, 19, 1244–1255.
9. K. Ijima, Y. Koike, W.-J. Chun, Y. Saito, Y. Tanizawa, T. Shido, Y. Iwasawa, M. Nomura and K. Asakura, Chem. Phys. Lett., 2004, 384, 134–138.
10. Y. Koike, K. Ijima, W.-J. Chun, H. Ashima, T. Yamamoto, K. Fujikawa, S. Suzuki, Y. Iwasawa, M. Nomura and K. Asakura, Chem. Phys. Lett., 2006, 421, 27–30.
11. K. Asakura, S. Takakusagi, H. Ariga, W.-J. Chun, S. Suzuki, Y. Koike, H. Uehara, K. Miyazaki and Y. Iwasawa, Faraday Discuss., 2013, 162, 165–177.
12. S. Takakusagi, H. Nojima, H. Ariga, H. Uehara, K. Miyazaki, W.-J. Chun, Y. Iwasawa and K. Asakura, Phys. Chem. Chem. Phys., 2013, 15, 14080–14088.
13. C. R. Kim, H. Maeda, B. Lu, Y. Nakamura, Y. Lin, Y. Wakisaka, D. Kido, K. Asakura and S. Takakusagi, Chem. Lett., 2024, 53, upad045.
14. Y. Lin, K. Oshiro, J.-Y. Hasegawa, S. Takakusagi, W.-J. Chun, M. Tabuchi and K. Asakura, J. Phys. Chem. C, 2025, 129, 8193–8205.
15. Y. Tanizawa, T. Shido, W.-J. Chun, K. Asakura, M. Nomura and Y. Iwasawa, J. Phys. Chem. B, 2003, 107, 12917–12929.
16. Y. Koike, K. Fujikawa, S. Suzuki, W.-J. Chun, K. Ijima, M. Nomura, Y. Iwasawa and K. Asakura, J. Phys. Chem. C, 2008, 112, 4667–4675.
17. W.-J. Chun, K. Miyazaki, N. Watanabe, Y. Koike, S. Takakusagi, K. Fujikawa, M. Nomura, Y. Iwasawa and K. Asakura, J. Phys. Chem. C, 2013, 117, 252–257.
18. B. Lu, D. Kido, Y. Sato, H. Xu, W.-J. Chun, K. Asakura and S. Takakusagi, J. Phys. Chem. C, 2021, 125, 12424–12432.
19. A. L. Ankudinov, B. Ravel, J. J. Rehr and S. D. Conradson, Phys. Rev. B: Condens. Matter Mater. Phys., 1998, 58, 7565–7576.
20. Q. Fu, T. Wagner and M. Rühle, Surf. Sci., 2006, 600, 4870–4877.
21. C. Niu, K. Shepherd, D. Martini, J. Tong, J. A. Kelber, D. R. Jennison and A. Bogicevic, Surf. Sci., 2000, 465, 163–176.
22. J. A. Kelber, C. Niu, K. Shepherd, D. R. Jennison and A. Bogicevic, Surf. Sci., 2000, 446, 76–88.
23. M. C. R. Jensen, K. Venkataramani, S. Helveg, B. S. Clausen, M. Reichling, F. Besenbacher and J. V. Lauritsen, J. Phys. Chem. C, 2008, 112, 16953–16960.
24. Z. Łodziana and J. K. Nørskov, J. Chem. Phys., 2001, 115, 11261–11267.
25. N. C. Hernández and J. F. Sanz, J. Phys. Chem. B, 2002, 106, 11495–11500.
26. N. C. Hernández and J. F. Sanz, Appl. Surf. Sci., 2004, 238, 228–232.
27. N. C. Hernández, J. Graciani, A. Márquez and J. F. Sanz, Surf. Sci., 2005, 575, 189–196.
28. J. F. Sanz and N. C. Hernández, Phys. Rev. Lett., 2005, 94, 016104.
29. H. Zhou, J. Zhu, H. Jin, Y. Li, B. Wang, S. Huang, W. Lin, Y. Li, W. Chen and Y. Zhang, Mol. Catal., 2023, 551, 113646.
30. T. Isono, T. Ikeda, R. Aoki, K. Yamazaki and T. Ogino, Surf. Sci., 2010, 604, 2055–2063.
31. T. Taguchi, T. Ozawa and H. Yashiro, Phys. Scr., 2005, 205.
32. G. Kresse and J. Hafner, Phys. Rev. B: Condens. Matter Mater. Phys., 1993, 47, 558–561.
33. G. Kresse and J. Hafner, Phys. Rev. B: Condens. Matter Mater. Phys., 1994, 49, 14251–14269.
34. G. Kresse and J. Furthmüller, Comput. Mater. Sci., 1996, 6, 15–50.
35. G. Kresse and J. Furthmüller, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 11169–11186.
36. P. E. Blöchl, Phys. Rev. B: Condens. Matter Mater. Phys., 1994, 50, 17953–17979.
37. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868.
38. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1997, 78, 1396.
39. P. Thompson, D. E. Cox and J. B. Hastings, J. Appl. Crystallogr., 1987, 20, 79–83.
40. J. Ahn and J. W. Rabalais, Surf. Sci., 1997, 388, 121–131.
41. E. A. Soares, M. A. Van Hove, C. F. Walters and K. F. McCarty, Phys. Rev. B: Condens. Matter Mater. Phys., 2002, 65, 195405.
42. P. Guénard, G. Renaud, A. Barbier and M. Gautier-Soyer, Surf. Rev. Lett., 1998, 05, 321–324.
43. S. Grimme, S. Ehrlich and L. Goerigk, J. Comput. Chem., 2011, 32, 1456–1465.
44. N. Kosugi, H. Kondoh, H. Tajima and H. Kuroda, Chem. Phys., 1989, 135, 149–160.
45. C. Rajesh, S. Nigam and C. Majumder, Phys. Chem. Chem. Phys., 2014, 16, 26561–26569.
46. T. Kurita, K. Uchida and A. Oshiyama, Phys. Rev. B: Condens. Matter Mater. Phys., 2010, 82, 155319.
47. G. Charlton, P. B. Howes, C. A. Muryn, H. Raza, N. Jones, J. S. G. Taylor, C. Norris, R. McGrath, D. Norman, T. S. Turner and G. Thornton, Phys. Rev. B: Condens. Matter Mater. Phys., 2000, 61, 16117–16120.
48. C. Kappenstein and R. P. Hugel, Inorg. Chem., 1978, 17, 1945–1949.
49. A. H. Lewin, R. J. Michl, P. Ganis and U. Lepore, J. Chem. Soc., Chem. Commun., 1972, 661–662.
50. M. S. Weininger, G. W. Hunt and E. L. Amma, J. Chem. Soc., Chem. Commun., 1972, 1140–1141.
51. E. A. Soares, M. A. Van Hove, C. F. Walters and K. F. McCarty, Phys. Rev. B: Condens. Matter Mater. Phys., 2002, 65, 195405.
52. N. M. Harrison, X.-G. Wang, J. Muscat and M. Scheffler, Faraday Discuss., 1999, 114, 305–312.

---