## Introduction

Pt-based alloys have been studied as one of the most promising catalysts for oxygen reduction reaction (ORR) in proton exchange membrane fuel cells (PEMFCs)1,2,3. Compared with pure Pt, which is a good ORR catalyst, the introduction of a transition metal (Ni, Co, etc.) provides tunability for the Pt lattice constant (strain effect due to size mismatch in a fcc random alloy or compound for example) and valence band (redistribution of Pt 5d states) in the form of alloys (intermetallic interaction) or core–shell structures (strain effect due to size mismatch of the core and the shell). In principle, the change of chemical environment (coordination species, interatomic distance, etc.) of Pt results in the redistribution of 5d densities of states (centroid shift and band width change), which directly determines the binding strength of the adsorbate species through bonding and anti-bonding interactions with the catalyst. The configuration of the Pt d-band plays an important role in determining the densities of states in the vicinity of the Fermi level, both occupied and unoccupied, hence its catalytic activity, which normally exhibits a volcano curve behavior towards the adsorbate binding strength4. The tuning of the Pt valence band to optimize the reactivity and selectivity is the most effective means towards Pt-based catalysts engineering. The distribution of the Pt 5d character in the valence band of an alloy, however, cannot be easily determined by conventional laboratory techniques. Ultraviolet photoelectron spectroscopy (UPS), for example, can detect the total valence band of Pt-Ni alloys; it has little elemental sensitivity to distinguish between the Pt and Ni character in the valence band and is surface sensitive.

In the past decade, valence-to-core X-ray emission spectroscopy (VTC-XES) has evolved to become a synchrotron technique for valence band detection with elemental specificity, as sophisticated high energy resolution X-ray crystal analyzers emerged. The elemental specificity is achieved by tuning the excitation energy across the absorption edges (core levels) of interest and tracking the inelastically scattered X-rays (just below the edge) and the fluorescence X-rays (when the core hole is switched on) using high energy resolution X-ray optics and area sensitive detectors. This is the so-called high energy resolution fluorescence detected (HERFD) setup, which uses either a cylindrically bent crystal with a dispersive geometry (von Hamos design)5,6,7,8 or a spherically bent crystal analyzer (SBCA) in a Rowland circle9,10. This development was propelled by the interest in resonant inelastic X-ray scattering (RIXS). RIXS is a concerted phenomenon of absorption and emission, has maximum intensity at threshold and can circumvent core–hole lifetime broadening; all these however require high energy resolution detection to be observed experimentally in the hard X-ray regime. As a result of the evolution of the bright and tunable incident X-ray and high energy resolution detection, the valence band of heavy metal elements can now be experimentally determined from VTC-XES. This has been recently demonstrated by studies of metal–ligand interactions for Ru complexes by Biasin et al.11, Levin et al.12, and Mn complexes by Hall et al.13. In this study, VTC-XES is conducted at the Pt L3-edge with great signal/noise (compared with 1s excitation on transition metals) due to high energy resolution and the dipole-allowed 5d5/2,3/2→2p3/2 transition.

The X-ray emission process with the excitation energy scanned across the Pt L3-edge is illustrated in Fig. 1 in three energy regions: (a) RIXS, ω resulting from an intermediate state and final state differentiated by the excitation of a valence electron into the conduction band via energy loss (ΔE2 = Ω—ω), by the incident X-ray, Ω; the process is enhanced by the proximity of the Pt 2p3/2 to 5d5/2,3/2 dipole transition; (b) VTC above threshold (E0, the point of inflection of the rising absorption edge) and near resonance (within the WL); and (c) VTC above resonance (non-resonant XES). Figure 1a shows the RIXS region where the emission energy exhibits a dispersion. The energy loss ΔE2 (Pt 5d band maximum) is a constant as the excitation energy (Ω) scans across the adsorption edge until it reaches the threshold then the core hole is turned on and resonant X-ray emission (fluorescence) takes place. In Fig. 1b where the excitation energy is above the threshold but not high enough to excite the core electron into the continuum, the VTC emission, as the name suggests, originates from a Pt 5d electron in the valence band combining with a 2p3/2 core hole via dipole transition, emitting a fluorescence photon. In Fig. 1c, at a few eV above the resonance (whiteline maximum), the excited core electron has enough kinetic energy to escape the Pt atom, and the VTC emission (ωAR) is still tracking the Pt 5d in the valence band but losing the advantage of enhanced intensity and core–hole lifetime broadening suppression. The absorption and emission processes are no longer concerted.

In this study, the VTC emission of Pt-Ni alloys is studied with HERFD detection at the Sector 20-ID beamline of the Advanced Photon Source (APS). A Si (311) SBCA using the (933) reflection (~ 1 eV resolution) is coupled to a PILATUS 100 K-S 2D detector (DECTRIS Ltd., Switzerland) in a Rowland Circle, allowing for the simultaneous collection of the X-ray emission from the entire Pt 5d band. The incident (excitation) X-rays from the Si (111) monochromators provides an energy resolution of 1.4 × 10–4 near the Pt L3-edge. Pt-Ni alloys including Pt3Ni, PtNi, and PtNi3 were synthesized by vacuum arc-melt method14. The crystal structure, local structure, and electronic structure have been reported using X-ray diffraction, normal mode X-ray absorption spectroscopy, ultraviolet photoelectron spectroscopy (UPS), and density functional theory calculation (DFT). Charge transfer from Ni to Pt (by filling the Pt 5d holes) upon alloying has been established14,15. Experiment with elemental sensitivity, however, is still in need to provide direct evidence of how the Pt valence band, especially the Pt 5d states, changes upon alloying, which in turn determines its catalytic behavior. Herein, with the state-of-the-art VTC-XES, we report the observation of chemical shifts and both broadening and narrowing of the Pt 5d band in the Pt-Ni alloy valence band due to competing strain and ligand effects.

## Results and discussion

Figure 2 displays the 2D and 3D plots of excitation vs emission X-ray energies across the Pt L3-edge with the intensity color coded. In the emission panel, the energy of the emitted X-ray (collected by the area sensitive detector) is very close to the excitation energy which will also appear in the spectrum as an elastically scattered (ES) peak, as shown in the right column of Fig. 2. A clear trend of increasing intensity of the ES (diagonal) is observed as Pt becomes more diluted in Ni. This is because Ni mainly contributes to elastic X-ray scattering instead of absorption at X-ray energies near the Pt L3-edge. In PtNi3, the most dilute case, the ES signal is strong enough to partially overlap with the VTC emission, while it is least intense in Pt3Ni. Note that the Kapton tape as the sealing material for the Pt foil and alloy samples also contributes to the ES peak. Also note that the quasi-Pt 5d3/2 and 5d5/2 band features are not resolved in the VTC emission spectra because the energy separation of the 5d spin–orbit derived features in the valence band is smaller than the energy resolution.

In the left panel of Fig. 2, ωR is the emission energy from excitation at whiteline (WL) maximum, ωNR is from excitation near the resonance within the WL envelope and ωAR is beyond the WL. They are marked with vertical bars and tabulated in Table 1. The WL exhibits a trend consistent with Ni → Pt charge transfer upon alloying. The VTC emission energies near the resonance ωNR and above resonance ωAR result in the energy difference ΔE1 = ωNR – ωAR (1.0, 0.7, 0.5, 0.4, and 0.3 eV for PtO2, Pt, Pt3Ni, PtNi, and PtNi3, respectively), also shown in Table 1. This difference is always positive as expected from the energy difference between the adiabatic and the sudden regime in core hole screening; that is that the addition of the excited electron in the unoccupied Pt 5d states above the Fermi level leads to a better screened core hole and a larger X-ray emission energy. Recall that both ωNR and ωAR are the X-ray emission from the recombination of a valence electron of Pt 5d character and a 2p3/2 core hole. ΔE1 hence arises from the existence of the excited electron in the vicinity of the Pt atom. There is a clear trend in the chemical systematic in that the more Pt 5d holes there are, the larger the ΔE1. This is reasonable since more available 5d holes will lead to poorer screening, hence a larger chemical shift.

We also note that the emission centroid at the resonance ωR (11,563.9, 11,563.3, 11,563.2, 11,563.1, and 11,562.8 eV for PtO2, Pt, Pt3Ni, PtNi, and PtNi3, respectively, see Table 1) is slightly different from ωNR. This is because of the interference of the RIXS signal, which spans along the energy transfer direction (diagonal in Fig. 2) and associated chemical effect (5d hole population), resulting in the elongation and asymmetry of the emission pattern near the resonance. The elongation effect is most evident in PtO2 where Pt 5d electrons are partially transferred to oxygen. Correspondingly, more 5d holes are available to accommodate the intermediate states in the RIXS process, hence the XES pattern is elongated by the RIXS to the highest extent (see Figure S1). In comparison, Pt attracts electrons from Ni in the Pt-Ni alloys to fill the d band, thus the elongation becomes less severe than in pure Pt.

To obtain the precise Pt 5d distribution in the valence band, the XES spectra excited across the Pt L3-edge WL for each sample are extracted (Fig. 3). Figure 3a–e show the VTC-XES for the five compounds as the excitation energy scans across the WL. One can clearly see the dispersion of the ES (dash green curve), the RIXS (dash blue curve) and the emergence of the fluorescence (solid blue line). Five XES recorded at the excitation energies of 11,560.8–11,562.8 eV are fitted with a Gaussian profile (FWHM = 3.0, 3.1, 3.1, 3.0, and 2.8 eV for PtO2, Pt, Pt3Ni, PtNi, and PtNi3, respectively, determined from the ES signal at the excitation energy of ~ 11,555 eV for each sample) and the RIXS signal (Pt d band). The results are shown in Table S2, Figure S4, S5, and summarized in Fig. 3f where the green peak is the elastic peak (~ 11,562 eV), and the blue band is the Pt 5d component of the VB of the alloy defined by a peak maximum and a band width. To check the validity of these values, we consider the ES width and the Pt d band in quadrature ($$6.5=\sqrt{{\Delta }_{5d}^{2}+{\Delta }_{\mathrm{E}\mathrm{S}}^{2}}$$ ), The observed 6.5 eV width yields a Pt 5d band width Δ5d of 5.7 eV, which is consistent with that observed in UPS and predicted by theory14.

The energy transfer (3.3, 2.5, 2.8, 3.0, and 3.1 eV for PtO2, Pt, Pt3Ni, PtNi, and PtNi3, respectively), ΔE2 = Ω – ω, represents the energy loss to the excitation of the Pt 5d valence electrons to the narrow unoccupied d states just above the Fermi level; in other words, it is the binding energy of the centroid of the Pt 5d states below the Fermi level. For each sample, ΔE2 is a constant. In the case of PtO2, a relatively large ΔE2 of 3.3 eV is expected because of the bandgap (> 1.0 eV)16,17,18. In the cases of Pt metal and alloys, ΔE2 shifts away from the Fermi level as Pt becomes more diluted in Ni. The same trend is observed in the DFT calculation results of these alloys in our previous study, attributing to the redistribution of Pt 5d states upon alloying with Ni14, and also in good agreement with the theoretical study by Matanovic et al., where the d-band centers of Pt were calculated to be − 2.01, − 2.17, − 2.31 and − 2.56 eV for Pt, Pt3Ni, PtNi, and PtNi3, respectively19. Similar behavior is also observed in the Au d band in Au-Cu20 and Au-Pt alloys21. The most exciting observation is in the full width at half maximum (FWHM): 5.8, 6.5, 6.8, 6.6, and 6.3 eV for PtO2, Pt, Pt3Ni, PtNi, and PtNi3, respectively. The results are summarized in Table 1. The difference is significantly above uncertainty. Note that these excitation energies were selected because they are in the RIXS region with an enhanced cross-section and suppressed core–hole lifetime broadening, hence it will provide higher chemical sensitivity compared with the VTC emission at and above the WL. The reason for this is further explained in Figure S1 which shows that the VTC becomes broader as the excitation energy moves across the threshold. For PtO2, a smaller width is observed because it is Pt (IV) with depletion of Pt 5d electrons compared to Pt metal, and there is little Pt–Pt interaction in the nearest neighbors hence a narrow Pt d band.