Effects of morphology on hydrogen spillover on CeO2 supports
CeO2-NRs and CeO2-NCs were prepared using a hydrothermal method while varying the reaction conditions. In contrast, the CeO2-NPs were commercially available and so were simply purchased. TEM images of CeO2 samples having different morphologies are shown in Figures 1a-c. The average thickness and length of the CeO2-NRs were determined to be approximately 15 and 200 nm, respectively. Three types of lattice fringes, related to the (111), (200) and (220) planes of the material, were observed in the HR-TEM image acquired from this specimen (Figure S1). These analyses indicated that the CeO2-NRs preferentially grew in the [110] direction and had exposed (110) facets. The CeO2-NCs displayed a cubic structure with edge lengths in the range of 50-100 nm, and were found to have only (100) exposed facets based on HR-TEM images (Figure S2).30, 37 The CeO2-NPs comprised edge-truncated particles with low-index facets, such as (100), (110) and (111).31
The XRD diffraction patterns acquired from each of the CeO2 samples contained distinct peaks related to a fluorite-like structure (Figure S3). The lattice constants calculated from the most intense (111) diffraction peaks increased in the order of CeO2-NCs < CeO2-NPs < CeO2-NRs. These data suggest a higher degree of dispersion of Ce3+ ions in the CeO2-NRs, presumably because of the larger atomic radius of Ce3+ compared with Ce4+ (Ce3+: 0.114 nm, Ce4+: 0.097 nm). The data obtained from H2 temperature-programmed-reduction (H2-TPR) experiments exhibited primary reduction peaks at 456, 510 and 537 °C for the CeO2-NRs, CeO2-NPs and CeO2-NCs, respectively (Figure 1d), indicating that the CeO2-NRs promoted oxygen migration to the greatest extent and had the highest degree of surface reducibility. A similar trend was observed in the case of the UV-vis diffuse reflectance spectra (Figure 1e). Specifically, the absorption edge position shifted to a longer wavelength in the order of CeO2-NCs < CeO2-NPs < CeO2-NRs. This effect is ascribed to the presence of surface oxygen vacancies.38 The reducibility of each CeO2 sample was also assessed by X-ray photoelectron spectroscopy (XPS) data, as shown in Figures 1f and g. The Ce 3d spectra could be deconvoluted into several peaks attributed to Ce3+ and Ce4+ species, and the results of quantitative analyses are summarized in Table 1. The fraction of Ce3+ in the material (Ce3+/(Ce3++Ce4+)) was found to increase in the order of CeO2-NCs < CeO2-NPs < CeO2-NRs. The O 1s spectra could be fitted with three peaks, ascribed to lattice oxygen (Olat), weakly surface adsorbed oxygen with a low coordination value (Oads) and surface adsorbed water molecules (Owater). The Oads was derived from active oxygen species adsorbed on exposed surface oxygen vacancies. The ratio of the Oads amount relative to the total oxygen atoms (Oads/(Oads +Olat)) on the CeO2-NRs was higher than that for the CeO2-NPs and CeO2-NCs, indicating a correlation with the proportion of Ce3+ in these materials.
The hydrogen spillover ability of each CeO2 support was assessed by performing H/D exchange reactions. In these experiments, the Pd/CeO2 was first exposed to a D2 atmosphere at 300 °C to generate metallic Pd NPs on the CeO2 and introduce O−D bonds on the CeO2 surfaces. Subsequently, each sample was first cooled to −50 °C and then heated to 200 °C under H2. The yield of HD from each trial was assessed by monitoring the signal obtained at m/z = 3 using mass spectroscopy. The associated surface reaction was 2Had + Olat−D → HD (g) + Olat−H, where Had is an adsorbed H atom, Olat is lattice oxygen in the CeO2 and HD (g) is gaseous HD generated via hydrogen spillover.39 The Pd/CeO2-NRs produced the most intense HD peak in the temperature range of 0 to 150 °C (Figure 1h), exceeding those produced by the Pd/CeO2-NPs and Pd/CeO2-NCs by factors of 1.7 and 4.4, respectively.
Evidently, the CeO2-NRs (which possessed exposed (110) facets) represented the most readily reducible CeO2 support and so was the most likely to produce Ce3+ ions and oxygen vacancies. Additionally, more hydrogen atoms were consumed on the surface of the CeO2-NRs during the H/D exchange reaction. These results suggest that the CeO2-NR material was the most likely to promote hydrogen spillover because this phenomenon involved simultaneous proton–electron transfer associated with the reversible reduction and oxidation of the metal oxide (that is, Ce4+ + e− ↔ Ce3+).
Hydrogen spillover-assisted synthesis and characterization of HEA catalysts
Three types of CoNiCuZnPd/CeO2 specimens were synthesized, utilizing a conventional impregnation method. In this process, all the metal precursors were deposited on each CeO2 support followed by reduction under a hydrogen atmosphere at 600 °C. This combination of elements was selected based on the moderately different reduction potentials for these metals (E0(Co2+/Co0) = −0.28 V, E0(Ni2+/Ni0) = −0.26 V, E0(Cu2+/Cu0) = +0.34 V, E0(Zn2+/Zn0) = −0.76 V and E0(Pd2+/Pd0) = +0.99 V, all vs. a normal hydrogen electrode). The formation of a solid solution phase from a multimetallic system requires a specific difference in atomic size (d < 6.6 %) and in configurational enthalpy (−11.6 < ∆Hmix < 3.2 kJ/mol). The present CoNiCuZnPd combination satisfied these criteria (d = 3.8 %, ∆Hmix = −8.2 kJ/mol) and so the formation of an HEA without any segregation was expected.
The reduction of supported metal cations was monitored by H2-TPR analyses of as-deposited monometallic and quinary specimens prior to exposure to H2. As shown in Figure 2a, the CeO2-NRs in combination with single metal cations generated broad reduction peaks. The order of the reduction temperatures for these elements also corresponded to the order of their reduction potentials. Interestingly, the quinary CoNiCuZnPd/CeO2-NR specimen produced a single sharp reduction peak at approximately 200 °C, which was almost equivalent to the temperature at which the monometallic Pd/CeO2-NR specimen generated peaks. These data establish that all the metal cations were reduced simultaneously on the CeO2-NR surface to form a quinary solid solution alloy. The hydrogen spillover phenomenon occurring at the Pd nuclei evidently accelerated the reduction of Co2+, Ni2+, Cu2+ and Zn2+ ions, despite the different redox potentials of these elements. A sharp reduction peak was also observed in the case of the CoNiCuZnPd/CeO2-NPs, although the CoNiCuZnPd/CeO2-NCs provided a relatively broad reduction peak ranging from 130 to 220 °C (Figure S4). These results indicate that minimal hydrogen spillover took place on CeO2-NCs compared with CeO2-NRs and CeO2-NPs, leading to a sequential rather than simultaneous reduction of the deposited metal precursors.
High-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) images and corresponding energy-dispersive X-ray (EDX) elemental maps are provided in Figures S5-7. The presence of metallic NPs is difficult to distinguish in the case of the CoNiCuZnPd/CeO2-NRs STEM image because of the poor Z-contrast between the CoNiCuZnPd alloy and the CeO2 support. However, the EDX map acquired over a specific region clearly shows all elements and indicates the formation of highly dispersed quinary CoNiCuZnPd HEA nanoclusters having diameters of less than 1 nm. Quinary NPs can be observed to some extent on the CeO2-NP support although these NPs are seen to be relatively large (about 10 nm) and the position of each element evidently varies slightly within each particle. In contrast, the CoNiCuZnPd/CeO2-NCs image does not clearly show the formation of well-defined quinary NPs. Instead, small CoNi NPs approximately 5 nm in size together with large aggregated CuZnPd NPs approximately 20 nm in size are seen in the EDX map. These results are consistent with the H2-TPR data and suggest that the CeO2-NRs and CeO2-NPs (but not the CeO2-NCs) are suitable supports for the simultaneous rapid reduction of metal precursors and the formation of multi-element NPs assisted by the hydrogen spillover effect.
The reduction sequences were examined in more detail by acquiring in situ X-ray absorption fine structure (XAFS) data under a H2/He atmosphere between room temperature and 500 °C. The Cu and Pd K-edge X-ray absorption near-edge structure (XANES) spectra of the CoNiCuZnPd/CeO2-NRs specimen obtained at 200 °C can be seen to be similar to those of the corresponding metallic foil reference materials, suggesting complete reduction below 200 °C (Figure 2b). In the case of Co, Ni and Zn, whose reduction potentials are relatively low compared with those for Cu and Pd, the white line intensities in the XANES spectra were decreased even at 200 °C, suggesting that the precursors were partially reduced, from M2+ to M0. Note also that the shape of the spectra are almost identical to those of the corresponding foils after heating at 500 °C. A similar trend was observed in the case of the CeO2-NP support (Figure S8). In contrast, the shape of the Pd K-edge XANES spectrum obtained from the CoNiCuZnPd/CeO2-NC sample was closer to that of metallic Pd rather than Pd2+ even at 200 °C, while that for Co, Ni, Cu and Zn did not begin to change until approximately 300 °C (Figure S9).
Fourier-transform extended X-ray absorption fine structure (FT-EXAFS) analyses were also conducted to further assess the structural transformations occurring during the reduction sequence (Figure 2c). The as-deposited CeO2-NR specimen produced sharp singlet peaks attributable to M-O bonds at approximately 1.9 Å in all K-edge regions. The peaks due to Cu-O and Pd-O bonds disappeared at 200 °C while a new peak related to M-M bonds with lengths of 2.0 to 2.5 Å appeared. In the case of Co, Ni and Zn, the intensities of the peaks associated with M-O bonds began to decrease at 200 °C. These data provide clear evidence for the enhanced reduction of the base metals as a consequence of hydrogen spillover occurring on the CeO2-NRs. Interestingly, the peaks due to M-M bonds for all elements indicated different bond lengths from those in the corresponding metallic foils after heating at 500 °C. As an example, the Pd K-edge FT-EXAFS spectrum showed evidence for a metallic bond length of approximately 2.1 Å, which is shorter than the standard value of 2.5 Å. This outcome suggests that the Pd atoms were surrounded by elements having smaller atomic radii, such as Co, Ni, Cu and Zn, rather than by other Pd atoms. It is notable that the Pd K-edge FT-EXAFS spectrum acquired at 200 °C exhibited two peaks due to Pd-M (M = Co, Ni, Cu and Zn) and Pd-Pd bonds at 2.1 and 2.5 Å, respectively, and that the Pd-Pd bonds gradually disappeared with increasing temperature and were completely gone above 400 °C. This result indicates that Pd nuclei were generated during the early stage of the reduction process, following which the other precursors were reduced to form the quinary HEA. Similarly, in the case of the CeO2-NPs support, the Co, Ni and Zn (along with the Pd and Cu) produced metallic bonds even at 200 °C under a H2 atmosphere, demonstrating the contribution of the hydrogen spillover effect (Figure S10). However, the split of metallic bonds and the remain of Pd-Pd bonds even at 500 °C in Cu and Pd K-edge spectra, respectively, indicate that the coordination of these atoms was not homogeneous within each particle. Thus, partially phase-segregated NPs were formed on the CeO2-NPs support rather than a well-mixed solid solution phase, as confirmed by the STEM observations and elemental mapping.
The FT-EXAFS spectra acquired from the CoNiCuZnPd/CeO2-NCs were quite different from those obtained from the CoNiCuZnPd/CeO2-NRs (Figure S11). In the case of the base metals, the M-O bonds remained unchanged at 200 °C, although the Pd was completely reduced. The Pd K-edge FT-EXAFS spectra demonstrated the presence of both Pd-M and Pd-Pd bonds at 200 °C, and the intensity of the signal related to Pd-Pd bonds decreased with increasing temperature. However, this signal was still observed even at 400 °C. Moreover, the lengths of the Co-M bonds (2.2 Å) and Ni-M bonds (2.1 Å) at 400 °C were the same as those in the corresponding foils. These results suggest that all the elements were not homogeneously mixed and that Co and Ni were not involved in the formation of the alloy phase, in good agreement with the HAADF-STEM images of the CoNiCuZnPd/CeO2-NCs.
Further insights into the local coordination spheres in these materials were obtained by determining the coordination numbers (CNs) and interatomic distances based on curve fitting analyses of the Pd K-edge FT-EXAFS spectra (Figures 3a and S12). The data for each of the CoNiCuZnPd/CeO2 specimens show the presence of Pd-Pd bonds at approximately 2.7 Å together with neighboring Pd-M bonds (M = Co, Ni, Cu and Zn) at 2.6 Å (Table S1). Note that, because of the similar ionic radii of these 3d elements, the different Pd-M bonds could not be distinguished from one another, and so these bonds were treated as equivalent during fitting.4, 12 The total CN values (CNPd-Pd + CNPd-M) were determined to be 5.6, 11.4 and 10.9 for the CeO2-NRs, CeO2-NPs and CeO2-NCs, respectively. Assuming that the present clusters each had a fcc cuboctahedral lattice structure, the diameter of the quinary CoNiCuZnPd alloy cluster having a CN of 5.6 can be estimated to be 0.79 nm, consisting of 13 atoms, according to the relationship between the CN and the diameter of a Pd cluster.40. In addition, the Pd-Pd interatomic distance in the CoNiCuZnPd/CeO2-NRs (2.70 Å) was found to be much shorter than that in the Pd foil (2.74 Å). This shrinkage provides further evidence for the presence of sub-nanosized clusters.41 Overall, such data suggest that CeO2-NRs can serve as a promising platform for the synthesis of sub-nanometer HEA clusters based on the significant hydrogen spillover effect on this support.
Interestingly, the Pd K-edge XANES spectrum of the CoNiCuZnPd/CeO2-NRs showed a unique featureless shape in the post-edge region without the two characteristic peaks at 24,390 and 24,415 eV attributed to electron transition from the 1s to 5p and 4f orbitals, respectively. These peaks are typically obtained from metallic Pd.42 Hence, DFT-based simulations of Pd K-edge XANES spectra were conducted to elucidate the structure of the quinary HEA supported on the CeO2-NRs. The amplitudes of the oscillations in the post-edge region were also evaluated, employing several sets of models, and bulk Pd, bulk HEA and Pd13 cluster models produced fluctuating spectra (Figures 3b and S13). Conversely, the HEA13(Co2Ni2Cu3Zn3Pd3) cluster model reproduced a unique flat shape without obvious oscillations, further confirming that the CoNiCuZnPd alloy supported on the CeO2-NRs comprised a well-mixed solid solution phase with sub-nanometer clusters.
Based on the H2-TPR and in situ XAFS results discussed above, a mechanism for the formation of HEA clusters on the CeO2-NRs assisted by hydrogen spillover can be proposed (Figure 4). In this process, the supported Pd2+ cations are initially reduced to form Pd0 nuclei under a H2 atmosphere. Following this, H2 molecules are heterolytically dissociated at Pd/CeO2 interfaces to form Pd-H and Ce-O-H species, accompanied by the reduction of neighboring Ce4+ ions to Ce3+ (step 1). Electron transfer from Ce3+ ions to adjacent Ce4+ ions promotes the simultaneous migration of H+ to O2− anions attached to these contiguous Ce4+ ions (step 2). In this manner, H atoms and electrons rapidly diffuse over the CeO2 surface, leading to the simultaneous reduction of other deposited precursor cations with large reduction potentials (step 3) to fabricate sub-nanometer HEA clusters (step 4).
Catalytic performance of CoNiCuZnPd/CeO2 during NO reduction
The reduction of nitric oxide (NO) has long been studied because NO emissions can negatively affect the environment, such as by causing acid rain and photochemical smog, causing adverse effects on human health.43, 44, 45 H2 is a more promising reductant for NO than NH3 or urea because the latter compounds are produced from H2 through the Haber-Bosch process.46, 47 In the present work, catalytic NO reduction with H2 was conducted using the quinary alloy on the various CeO2 supports. The effect of temperature on the catalytic activity between 50 and 200 °C is summarized in Figure 5.
The CoNiCuZnPd/CeO2-NRs exhibited the highest activity at lower temperatures, giving an NO conversion of 100% at 100 °C. This temperature was considerably lower than those required when using NPs (150 °C) and NCs (200 °C) as supports. This high level of activity can be ascribed to the formation of only sub-nanometer quinary HEA clusters on the CeO2-NRs. It should also be noted that the catalytic activity of the CoNiCuRuPd/CeO2-NR sample was substantially higher than that of Pd/CeO2-NRs prepared by the same method. Furthermore, other non-noble metal catalysts (that is, M/CeO2-NRs where M = Co, Ni, Cu or Zn) exhibited essentially zero NO reduction activity at 200 °C, as shown in Figure S14. Similarly, the CoNiCuZnPd/CeO2-NPs provided improved activity in comparison with the corresponding monometallic Pd/CeO2-NPs catalyst at low temperature, confirming the partial formation of a quinary HEA, albeit with relatively large cluster sizes on the nanometer scale. Conversely, the CoNiCuZnPd/CeO2-NCs exhibited lower activity than the Pd/CeO2-NCs, presumably because of the absence of HEA NPs. These results clearly suggest that well-mixed sub-nanometer HEA clusters were formed on the CeO2-NRs and functioned to promote NO reduction due to the synergistic effects of multiple elements (that is, the so-called cocktail effect).
Another possible reason for the improved performance of the HEA sub-nanoclusters is the presence of under-coordinated surface atoms, which would be expected to drastically affect the capacity of the material to adsorb reactants.48, 49 The adsorption characteristics of the monometallic and quinary alloy catalysts on the CeO2-NRs were therefore assessed using temperature-programmed desorption (TPD) following NO adsorption in conjunction with Fourier transform infrared spectroscopy (Figures 6a and b). The Pd/CeO2-NRs sample generated a peak attributed to linear-stretching vibrations of NO (νNO) adsorbed on top of Pd atoms.50 In addition, the intensities of the νNO peaks obtained from the monometallic Pd/CeO2-NRs specimens gradually decayed with increasing temperature, and almost NO molecules were desorbed from the Pd at 100 °C. In contrast, the CoNiCuZnPd/CeO2-NRs continued to produce this peak even at 100 °C. These results suggest that the adsorption sites provided by the CoNiCuZnPdCeO2-NRs strengthened the interaction of the material with NO molecules via electron enrichment.
DFT calculations were also performed to provide support for these experimental results. The adsorption energy values (Ead) for NO and H on fcc on-top, fcc hollow and hcp hollow sites were evaluated in this manner. In preparation for these calculations, a CoNiCuZnPd HEA model slab was constructed by randomly populating each constituent atom in the fcc (111) facets of repeating slabs. A total of 25 different elemental distribution patterns were employed for the fcc on-top calculations together with 30 different patterns for the fcc hollow and hcp hollow models so as to investigate the effects of the HEA configuration (Tables S2-4). The Ead values for NO and H adsorbed at fcc on-top, fcc hollow and hcp hollow sites on CoNiCuZnPd HEA slabs and the corresponding monometallic slabs are summarized in Figures 6c, 6d and S15.
For fcc on-top adsorption, each element was found to more strongly adsorb NO molecules in the case of the HEA slab rather than the corresponding monometallic slab. However, the H Ead values were identical between the HEA and monometallic models. Similar trends were observed for fcc and hcp hollow sites. The NO Ead on the hollow sites associated with the HEA slab was also higher than the average value for the monometallic models, while the H Ead on these same sites was almost unchanged from the average value for the monometallic slabs. These experimental and theoretical results demonstrate that the formation of HEA improved NO adsorption more than H adsorption. The strengthening of the interactions between NO molecules and the metal surface would be expected to weaken N-O bonds and thus promote dissociation of these bonds. This effect is attributed to greater electron donation and back-donation based on d-5σ and d-2π* interactions, respectively.50 Thus, the NO reduction ability of the CoNiCuZnPd/CeO2-NRs was improved.
Reversible structural change of HEA sub-nanoclusters
The stability of the present catalyst under oxidative/reductive conditions is a key factor affecting the application of this material to automotive emissions control. This stability was therefore assessed by monitoring changes in local structure using in situ XAFS. The experimental protocol and the resulting spectra are presented in Figures S16-22. Figure 7a shows the Pd K-edge XANES spectra obtained at 400 °C under either an oxidative (8000 ppm NO/He) or reductive (20,000 ppm H2/He) atmosphere. The data initially acquired from the Pd/CeO2-NRs indicate Pd in the metallic state but the edge position is shifted to a higher energy under NO, accompanied by a gradual change in the shape in the post-edge region. These variations are indicative of oxidation of Pd. Following the introduction of H2, the edge position returned to the initial lower energy value and two characteristic peaks related to metallic Pd appeared. These changes were also observed even after successive oxidation and reduction.
The FT-EXAFS spectra provided further evidence for the above scenario (Figures 7b and c). The intensity of the initial main peak due to Pd-Pd bonds at approximately 2.5 Å decreased under NO, accompanied by an increase in the peak related to Pd-O bonds at approximately 1.6 Å. Following the introduction of H2, the latter peak disappeared and the Pd-Pd peak was recovered. These results are clear evidence for the reversible redox capabilities of Pd species on monometallic Pd/CeO2 catalysts under oxidative/reductive conditions, as has been reported previously.51
Interestingly, the CoNiCuZnPd/CeO2-NR specimen exhibited different regeneration properties from those for monometallic Pd/CeO2-NRs (Figures 7d, e and f). The initially flat XANES spectrum transitioned to a shape typical of monometallic Pd upon the introduction of NO. In the case of the FT-EXAFS spectra, the peak attributed to short Pd-M bonds at approximately 2.1 Å disappeared and a peak related to long Pd-Pd bonds gradually appeared. These data clearly suggest that oxidation of Pd species was suppressed even upon exposure to NO, with the other metals acting as sacrificial species to prevent this oxidation. After switching to a H2 flow, the XANES spectrum regained the flat shape characteristic of the HEA. The FT-EXAFS spectrum also recovered the original HEA structure with decreasing and increasing intensity of peaks associated with long Pd-Pd bonds and short Pd-M bonds, respectively.
Linear combination fitting (LCF) of these Pd K-edge XANES spectra provided further information regarding the oxidation state of each metal during the redox sequence. As shown in Figure 7g, the Pd0 fraction (Pd0 / (Pd0 + Pd2+)) in the Pd/CeO2-NRs was initially 0.90 but subsequently decreased to 0.37 after the introduction of NO for 30 min, and then returned to its initial value under a reductive H2 atmosphere. Conversely, oxidation of Pd species in the CoNiCuZnPd/CeO2-NRs was suppressed and a fraction of 0.78 was observed even after a 30 min exposure to NO. Moreover, the Pd in this material was almost entirely reduced to the metallic state upon the introduction of H2. It should also be noted that the Pd species in the CoNiCuZnPd/CeO2-NRs sample exhibited almost complete reversibility of this process following successive oxidation and reduction.
The fraction of M0 species among all the elements in the CoNiCuZnPd/CeO2-NRs was evaluated based on applying the LCF technique to XANES spectra. As summarized in Figure 7h, each of the base metals was oxidized under NO but returned to the metallic form under H2. It should be noted that Co and Ni were oxidized completely while Cu and Zn were only partially oxidized during exposure to NO. Figure 7i show the HEA proportions determined by the LCF analyses of Pd K-edge XANES spectra of Pd foil and the original CoNiCuZnPd/CeO2-NRs. Under NO, this proportion decreased from 1 to 0.31 but, following the introduction of H2, the HEA structure was recovered.
From these results, differences in the mechanisms responsible for the structural reversibility of the monometallic and quinary catalysts under oxidative/reductive conditions can be proposed (Figure 8). Monometallic Pd NPs on the CeO2-NRs, which were determined to have diameters of approximately 2 nm based on curve fitting of FT-EXAFS data, were easily oxidized and dispersed on the CeO2. These NPs were therefore transformed into PdO clusters and single atom Pdδ+ species under oxidative NO conditions, and subsequently transitioned back to NPs by reduction with H2.52, 53 In the case of the CoNiCuZnPd/CeO2-NRs, the randomly distributed Pd atoms formed core nuclei encapsulated by non-noble metal oxide shells (that is, Pd@MOx; M = Co, Ni, Cu and Zn) via the sacrificial oxidation of the base metals under NO. These oxide shells provided steric hindrance to interactions between the Pd atoms and NO molecules and also inhibited the diffusion of Pd atoms into the CeO2. It is notable that the morphology of the 13-atom nanoclusters changed from spherical cuboctahedral to raft-like, and that this change limited the exposure of Pd atoms. These structures could be reversibly changed back to well-mixed HEA nanoclusters under a H2 atmosphere. This proposed model based on Pd@MOx clusters having raft-like morphologies can be confirmed based on a Pd3@(Co3Ni2Cu3Zn2)O6 model, as illustrated in Figure S23. The simulated Pd K-edge XANES spectrum contained two characteristic peaks due to monometallic Pd0, in sharp contrast to the spectrum obtained for metallic HEA clusters (that is, the HEA13 clusters).