We have used X-ray absorption spectroscopy and quantum chemical density functional theory calculations to identify critical features in the electronic structure of different sites in alloys that govern the local chemical reactivity. The measurements led to a simple model relating local geometric features of a site in an alloy to its electronic structure and chemical reactivity. The central feature of the model is that the formation of alloys does not lead to significant charge transfer between the constituent metal elements in the alloys, and that the local electronic structure and chemical reactivity can be predicted based on physical characteristics of constituent metal elements in their unalloyed form.
Alloys are an important class of materials that often exhibit unique physical and chemical characteristics, such as exceptional hardness1 or superior chemical reactivity.2 As such, alloys show promise for many applications including heterogeneous catalysis3–7 and electrocatalysis.8–11 While the potential for the utilization of alloys in heterogeneous catalysis and electrocatalysis is significant, predictive models relating the geometric structure of alloys to their chemical reactivity are lacking and desirable alloy catalysts are typically identified through trial-and-error experimental approaches.
In this letter, we propose a model, developed based on X-ray absorption spectroscopy measurements and density functional theory (DFT) calculations, allowing us to predict the impact of alloying on the electronic structure of different sites in alloys and on their chemical reactivity. The advantage of the proposed model is that the chemical reactivity of various sites in complex alloy materials can be predicted based only on properties of the constituent metals in their unalloyed forms. The proposed model has the potential to assist in rapid screening and identification of alloys for different catalytic and electrocatalytic reactions.
The chemical reactivity of metal surfaces (including alloys) can be understood in terms of the Hammer–Nørskov -band model,12–17 which relates the electronic structure of a site on a metal, more specifically the center of the -band (i.e., the average energy of -states projected on the site), to the local chemical reactivity. Metals with a higher -band center with respect to the Fermi level bind most adsorbates more strongly than metals with a lower -band center.12,13,16
For pure monatomic metals, the relative positions of the -band center can be easily estimated and thus their relative chemical reactivity can be predicted.14 The only parameter in the model is the -band center which is governed by the filling and the shape (mainly width) of the -band. The -band filling is related to the number of occupied -orbitals. The width of the -band [i.e., the width of the projected density of states (DOS)] is a function of the geometry [the distance between atoms, in Fig. 1(b)] and the spatial extent of the -orbitals [ in Fig. 1(b)].18 For any pure metal, the -band filling, the distance between atoms, and the spatial extent of -orbitals are known and therefore the relative positions of the center of -band and the relative chemical reactivity can be easily estimated.18
Mechanisms by which the center of the local -band can change include (a) charge transfer from one element to another and (b) a change in the overlap between -states and the valence orbitals of neighboring atoms leading to a change in the -band width.
Mechanisms by which the center of the local -band can change include (a) charge transfer from one element to another and (b) a change in the overlap between -states and the valence orbitals of neighboring atoms leading to a change in the -band width.
Predicting the position of the -band center for various sites in alloys and therefore the chemical reactivity of the sites is more challenging. When two different elements interact with each other in an alloy, the center of the -band localized on an atom can be affected by: (i) transfer of electronic charge from or to the -states localized on the atom without a change in the width or shape of -band (i.e., change in local -band filling), as shown in Fig. 1(a). This is basically a rigid band model;19 (ii) change in the width of -band due to hybridization between local -orbital and the valence orbitals of neighboring atoms without significant charge transfer, as shown in Fig. 1(b). This mechanism is consistent with the rectangular band model18 and it has been proposed to play the critical role by Kitchin et al.;20 and (iii) combination of charge transfer and the change in the width of -band. The mechanisms (i)–(iii) are dominant as long as the shape of the -band defined in terms of skewness and kurtosis is not significantly affected by alloying.
To develop reliable models that would allow us to predict how the chemical reactivity of a metal site changes as another element is introduced to form an alloy, it is imperative to unearth which mechanism [(i)–(iii) above] is responsible for the change in the -band center of local sites in alloys. To address this question, a number of Pt (Pt/Cu, Pt/Ru, and Pt/Sn) and Ni (Ni/Sn, Ni/Au, Ni/Ag) alloy catalysts were synthesized21 and their electronic structure in the neighborhood of Fermi level were measured. The alloys were selected to sample a broad range of electronically different elements with (Ni and Cu) metals interacting with (Ag and Ru), (Au and Pt), and (Sn) metals. Since we arrived at identical conclusions for both families of alloys, we discuss in detail the results for the Pt alloys. The results for the Ni alloys are shown in Ref. 21.
To establish that our synthesis resulted in the formation of alloys, we used extended X-ray absorption fine structure (EXAFS) spectroscopy, which probes the average local geometric environment of an element in a material. In Fig. 2, we show the measured -edge EXAFS spectra for Pt, Pt/Cu, and Pt/Sn. Figure 2 shows that there is a significant difference in the spectra of pure Pt and the Pt alloys which is indicative of the formation of the alloy materials. The measured alloy spectra were also compared to the best-fit, simulated EXAFS spectra.22 The fitting parameters for the best-fit spectra are shown in Ref. 21 (Table S1). Since the best-fit spectra required nearest neighbor Pt–Pt and Pt–M (where M is either Cu or Sn) scattering paths, it was concluded that alloys, defined as partial or complete Pt–M solutions, were formed. The measured bond distances for different alloys agreed very well with previously identified crystal structures of PtRu, , , and .23,24
Measured and fitted EXAFS spectra for pure Pt and (a) Pt–Sn alloy, (b) Pt–Cu alloys.
Measured and fitted EXAFS spectra for pure Pt and (a) Pt–Sn alloy, (b) Pt–Cu alloys.
To measure the electronic structure of the alloy catalyst, we used X-ray absorption near edge structure (XANES) spectroscopy. This technique allows us to measure electronic states above the Fermi level localized on a particular element in the material. For example, the -edge XANES spectra probes the unoccupied and DOS localized on Pt atoms in the alloy materials by measuring the photoabsorption cross-section associated with the excitation of Pt core electrons to the unoccupied Pt and states. Since the dipole transition probability to the -states is much larger than that to the -states,25 the spectra are essentially dominated by the -states.25
The -edge XANES spectra for pure Pt and the Pt alloys are shown in Fig. 3(a). The figure shows that the formation of the alloys affects the -edge: there are shifts in the positions of the edge onsets and changes in the width and height of the post-edge peak. To identify the origin of the observed changes in the -edge, we used DFT to calculate oscillator strengths,21 the computational equivalent of XANES spectra, for the electron transition from the Pt core states to the Pt and states above the Fermi level for pure Pt and the Pt alloys. The calculated oscillator strengths for the alloy crystal structures identified in our EXAFS measurements are shown in Fig. 3(a). The figure shows that there is an excellent agreement between calculated oscillator strengths and measured XANES spectra, i.e., the main features of the spectra for different alloys, including the edge onset and the peak width are reproduced in our calculations. We found that the initial (ground) state electronic structures, used to calculate the oscillator strengths in Fig. 3(a), reproduced the measured spectra better than the electronic structures calculated for a partially filled Pt core hole (Ref. 21, Fig. S8). This has been observed before and it is attributed to the electron shielding effect in Pt, which greatly reduces the interaction between the core hole and the promoted electron in the excited state.26,27 The agreement between the measured spectra and the spectra calculated using the ground state electronic structure indicates that the measured -edge XANES spectra maps well on the ground state Pt DOS of -symmetry, therefore allowing us to experimentally measure the ground state electronic structure. Further analysis of the calculated changes in the local -DOS in response to alloying showed that the changes in XANES spectra due to alloying were the result of the formation of new electronic states (orbitals), through the hybridization of the Pt -states and the valence orbitals of the neighboring atoms in the alloys, as shown in Fig. 3(b).28
In addition to providing information about the shape of -band, the XANES -edge also allows for a quantification of the relative number of electronic states above the Fermi level localized on the atom in different environments. The number of localized -states is proportional to the integrated area under the measured normalized spectra.25,29–31 In the inset in Fig. 3(a), we show measured integrated areas, i.e., the relative numbers of -states above the Fermi level localized on Pt atoms for pure Pt and the Pt alloys. The figure shows that the number of -states localized on Pt atoms does not change significantly in response to alloying. This indicates that the transfer of -holes (or -electrons) between Pt atoms and the atoms of the other elements in the alloys is very small, i.e., the local potential of elements in the alloys is sufficient to preserve the local charge (local -band filling) on the atoms.
To summarize, Fig. 3(a) provides experimental evidence that for all studied Pt alloys, irrespective of their composition: (i) the formation of the alloys results in a change in the width of the -band localized on Pt and (ii) there is no significant charge transfer to or from the Pt -states. We obtained identical conclusions based on similar experimental analysis for various Ni alloys, shown in Fig. S5 of Ref. 21.
There are a number of important consequences of these observations. One critical consequence is that since upon the formation of alloys the local -charge on an atom in the alloys is preserved, the local -band center and the local chemical reactivity are functions of only the local -band width (measured with respect to the Fermi level). The width is governed by the spatial extent of the hybridizing valence orbitals and the geometry (bond distance between atoms in the alloy, ). This means that for different alloys with similar geometries, the width of the -band, and therefore the -band center, is a unique function of only the spatial extent of the -orbitals of constituent metal elements.
For example, the distance between Pt atoms and metal atoms in different -M/Pt alloys is approximately equal for a given crystal structure32 because the geometry is governed to a large degree by the -states, which are in the case of metals very similar to each other. The width of the -band projected on geometrically similar Pt sites (and therefore the center of -band and the chemical reactivity of the Pt sites) should be only a function of the spatial extent of the -orbitals of the metals directly interacting with the Pt atom.
We have tested the predicted relationships using DFT to calculate adsorption energies of various adsorbates on geometrically identical Pt sites on a number of model -M/Pt skin alloys. The skin alloys were modeled as a Pt fcc slab terminated with the Pt(111) surface, with metals replacing Pt in the subsurface layer of the slab. Identical models were used previously by Kitchin et al.33 We note that the -M/Pt skin alloys exhibit higher tolerance to CO and lower overpotential losses associated with the oxygen reduction reaction compared to the pure Pt when used as electrodes for low temperature proton exchange membrane fuel cells. Figure 4(a) shows that, as predicted by the analysis above, the adsorption energies for various adsorbates on the Pt surface sites scale linearly not only with the center of -band, as shown previously, but also with , where is the spatial extent of the -orbitals of the metals which form the -M/Pt skin alloys. The values for the spatial extent of -orbitals were obtained from Ref. 18. Similar relationships have been found for adsorption on geometrically identical sites on -M/Pt and -M/Pt alloys.21 The scaling of the adsorption energies with is not accidental. Muffin–Tin orbital theory (atomic sphere approximation)18 predicts that for a fixed geometry, the -band width projected on a given atom in an alloy is proportional to .
The relationships presented in Fig. 4 were obtained for geometrically identical Pt sites on various Pt skin alloys. However, it is important to stress that our experiments show that the preserved local -charge concept is also valid even if there is a change in the local geometry. For example, while the bond distances between Pt and Cu or Sn in the respective Pt alloys are different, the measured local charge on Pt is preserved (see Fig. 3). The impact of the local geometry on the width of the -band can also be estimated based on tight binding approximation, showing that the width of the -band is proportional to for and hybridization and for hybridization, where is the length of the bond between the atoms in the alloy.18 The effect of geometry on the -band center has been studied before and these relationships have been supported by extensive DFT calculations.20,33
Electronic properties of Pt alloys. (a) Measured -edge XANES spectra and calculated oscillator strengths for different Pt alloys. Insert shows calculated areas under the measured XANES spectra. These are plotted for different cutoff energies since it is not clear where the near edge ends. We find that, independent of the cutoff energy, the integrated spectra give almost identical values for different alloys. (b) DOS projected on different elements in the alloys: Pt states are solid blue, and metal (Sn and Cu) and states are dashed and dashed-dotted lines, respectively. Cu states are solid green and purple for PtCu and alloy, respectively.
Electronic properties of Pt alloys. (a) Measured -edge XANES spectra and calculated oscillator strengths for different Pt alloys. Insert shows calculated areas under the measured XANES spectra. These are plotted for different cutoff energies since it is not clear where the near edge ends. We find that, independent of the cutoff energy, the integrated spectra give almost identical values for different alloys. (b) DOS projected on different elements in the alloys: Pt states are solid blue, and metal (Sn and Cu) and states are dashed and dashed-dotted lines, respectively. Cu states are solid green and purple for PtCu and alloy, respectively.
(a) Binding energy of various adsorbates on -M/Pt skin alloys as a function of characteristic orbital length of metals. (b) Charge analysis of Pt on -M/Pt skin alloys.
(a) Binding energy of various adsorbates on -M/Pt skin alloys as a function of characteristic orbital length of metals. (b) Charge analysis of Pt on -M/Pt skin alloys.
Finally, it is important to address the ambiguity of the concept of charge transfer from one element in an alloy to another. In the analysis above, we have referred to the transfer of -electrons from one element to another in terms of a uniquely defined area under the measured or calculated oscillator strength. We established that there is no significant transfer of -electrons between constituent metal elements. This information was sufficient to design a model, which allows us to predict the impact of alloying on the local electronic structure and chemical reactivity of a catalytic site based on the ground states electronic structures. On the other hand, it is also possible to define the localized charge in terms of the partitioning of the ground state electron density among the atoms in the alloy. For example, the charge analysis of the DFT-calculated electron density of the model alloys of Bader and Mulliken34,35 shows that there are shifts in the electron density from the metal atoms to the Pt atoms as shown in Fig. 4(b). However, projections of the local -orbitals on the partitioned volumes show that the only contributions to the charge transfer are to the -states of Pt. It can be shown, using the Newns–Anderson model,12,36,37 that small changes in the filling of the -bands have only a negligible effect on the chemical reactivity of the surfaces (Ref. 21, Fig. S9).
In summary, the picture that emerges from these studies is that when two transition metals are brought together to form an alloy, there are no significant shifts in the filling of local -states (no charge transfer to or from the -states), and therefore the change in the position of the -band center and the local chemical reactivity of alloy sites is, to a large degree, governed only by the width of the local -band. The width of local -band is a function of the local geometry (more specifically the bond distance between the elements that form the alloy) and the spatial extent of valence orbitals of the atoms. Since the bond distance between constituent metal elements in alloys can be predicted, for example, by applying Vegard’s law,38 and since the spatial extent of the orbitals of elements in alloys is known, the position of the center of -band for various sites in alloys and therefore the chemical reactivity of these sites can also be predicted easily. It is also important to note that relaxations at surfaces could affect the bond distance; however, for a given family of alloys (, , or ) with similar crystal structures these changes are approximately the same.
We gratefully acknowledge the support of the U.S. Department of Energy DOE-BES, Division of Chemical Sciences (Grant No. FG-02-05ER15686), NSF (Grant Nos. CTS-CAREER 0543067 and NSF CBET 0756255), and ONR (Grant No. N00014010810122). S.L. also acknowledges the DuPont Young Professor grant by DuPont Corporation and the Camille Dreyfus Teacher-Scholar Award from the Camille & Henry Dreyfus Foundation.
