Optical properties and chemical ordering of Ag-Pt nanoalloys : a computational study

A series of core@shell and layered ordered phases of AgPt bimetallic nanoparticles has been studied with Ag:Pt = 3:1 and 6:1 atomic compositions and sizes from 116 to 201 atoms. The elementary chemical order has been established by using a recent method (TOP), which assigns energy according to different topological degrees of freedom. The TOP lowest-energy structures, confirmed by density functional calculations, are then studied by time dependent density functional theory in order to calculate optical properties. The present study shows that for AgPt nanoparticles with core@shell structure the optical properties are sensitive to both the Pt concentration and system size. Spectral trends related to chemical order have also been identified. * Corresponding authors’ emails: stener@units.it; konstantin.neyman@icrea.cat

Once the structures of the systems and their chemical ordering are assessed, it is possible to calculate their optical properties. In this respect the best compromise between accuracy and computational expenditures is offered by the Time Dependent Density Functional Theory (TDDFT) formalism. In the present work we employ a recent scheme to solve the TDDFT equations representing the induced density over an auxiliary basis set of Slater Type Orbitals (STO) 6 -8 which has proven to be very efficient to treat quite large metal particles containing up to several hundred atoms. 9 In the photoabsorption experiments it has been found that while pure Ag clusters display strong Surface Plasmon Resonance (SPR), such feature disappears in bimetallic Ag-Pt nanoclusters, unless a very large cluster size (around 10 nm) is reached. 10 The role of Pt in the SPR suppression in AgPt clusters has been already considered in a previous work. 11 The present work deals with AgPt nanoalloys of different size and composition mixing silver and platinum atoms. Specifically, three different types of nanoparticles composed of 116, 140, or 201 atoms have been studied. For each size, an Ag-rich situation has been investigated, with Ag:Pt proportions of about 3:1 and 6:1 considering NPs with a truncated octahedron geometry. In addition, particles with a layered structure and Ag:Pt composition ratio ca. 4:1 have been considered as well, since a recent experiment has demonstrated the existence of AgPt multilayer nanoparticles with a L1 1 phase coated by a monoatomic Ag skin. 12 AgPt nanoparticles are used in several catalytic processes; 13 as an example, H 2 O 2 production by direct H 2 oxidation synthesis (H 2 + O 2 → H 2 O 2 ) is a promising alternative to 4 composition and crystalline lattice, the energy difference between any two chosen homotops is considered to depend only on the respective mutual positions of A and B atoms:  with the Perdew-Burke-Ernzerhof (PBE) 18 exchange-correlation functional. The PBE exchangecorrelation functional was found to be one of the most appropriate among common functionals to describe transition metal bulks and surfaces. 19 -21 The interaction between valence and core electrons was treated within the PAW (Projector Augmented Wave) approach. In order to moderate the computational cost, a 275 eV energy cutoff for the plane-wave basis sets was used.
The one-electron levels were smeared by 0.1 eV using the first-order method of Methfessel and Paxton, 22 and the converged energies were extrapolated to zero smearing. All calculations were performed only at the Γ-point in the reciprocal space, and the full set of atoms was allowed to locally relax during the geometry optimization. These electronic structure calculations represent the most challenging, computationally speaking, part of the method. To reduce computational costs, the minimal separation between NPs was chosen to be > 0.7 nm, a typical safe value which guarantees negligible interaction between adjacent NPs in neighboring supercells. 23 Optical properties: the TDDFT complex polarizability The ADF (Amsterdam Density Functional) program has been used to calculate the optical properties at the TDDFT level. The complex polarizability method has been employed to solve the TDDFT equations as described previously, 6

Structures
The i According to the obtained results, for different studied nanoparticles there is only a quite moderate variation upon the specific descriptor values. In particular, the energy gain due to the formation of heteroatomic bonds is rather small for these nanomaterials, and so it plays a secondary role in the determination of the NP ordering. In all cases, the stability of Pt atoms is highest inside the NP and lowest in the NP surface sites -corners, edges, and terraces. In general, the less the site is coordinated, the higher is the energy of a Pt atom in that position. This effect leads to the core@shell-like structure of the lowest-energy AgPt homotops, which have the surface shell enriched with Ag and the core composed mostly of Pt. In Figure 1 the optimized structures of all lowest-energy homotops under scrutiny (putative global minima) are shown.
Studies on bimetallic nanoparticles 2,5 composed of PdAu and PdAg revealed similar trends in the chemical ordering, defining that the coinage metal atoms Au and Ag preferentially occupy positions with lower coordination numbers in core@shell structures.

Optical properties
For each of the six AgPt clusters, a TDDFT photoabsorption spectrum was calculated (see Figure   2). The charge of the clusters has been chosen in order to obtain a closed@shell electronic structure. In some instances, a closed@shell electronic structure can be obtained for different values of the cluster charge. For example, [Ag 99 Pt 17 ] + and [Ag 99 Pt 17 ]are both closed shell and the effect of the charge on the optical photoabsorption spectrum is considered in Figure S1 of the Supporting Information (SI). The effect is quite modest, although not so small as in pure gold or silver clusters. 28 The general shape of the photoabsorption remains very similar for the two charge states, with a regular intensity increasing from 2 eV up to 5 eV, and with an evident feature around 4.5 eV in both cases. Then the intensity is modulated but remains constant in average up to 7 eV when it starts to decrease.  17 ] + ) we notice that when the Pt concentration increases the intensity is increased up to 6 eV, while from 6 eV to 7 eV the intensity is quite comparable and above 7 eV the intensity is lower at higher Pt concentration.
A very useful tool to discuss the spectra and their trends in nanoalloys is the analysis in terms of fragments 29 as reported in Figure 5 for all studied systems. In fact for the present nanoalloys it is very natural to define two fragments, each one consisting of the atoms of the same element. With this tool the photoabsorption profile is split in four components, giving the weight of the fragment in initial state and final state, so in the figure are displayed these four contributions, namely, Ag→Ag, Ag→Pt, Pt→Ag, and Pt→Pt using different colors. The fragment analysis for [Ag 87 Pt 29 ] -, shows that up to 5 eV Pt→Ag is the most important contribution, while Ag→Ag becomes the leading one at higher energies. The other contributions are less relevant, in particular Pt→Pt one is weak but rather constant over the whole considered energetic interval, while Ag→Pt one is weak but follows the shape of Ag→Ag, although strongly reduced. The fragment analysis for this system reveals that the silver absorption plays a major role at high energy (>5 eV), however at lower energy the importance of the Pt→Ag contributions suggests that the absorption can be described as an inter-band transition from occupied valence Pt states to virtual Ag empty states. In summary, Pt doping can be described as an effect of increasing the availability of electrons, which can be easily promoted to silver by light absorption. is presented in Figure 6: in the 4.32 eV peak we observe three strong spots on the diagonal (the diagonal corresponds to an energy difference between virtual and occupied orbitals equal to the excitation energy) together with strong extra-diagonal spots parallel to the occupied orbitals energy axis, corresponding to lower energy configurations with final states close to the LUMO. 8 This shape is typically plasmonic, since the presence of extra-diagonal spots is an indication of a collective behavior. ICM-OS analysis of the peak at 5.16 eV provides a much simpler picture: almost all the intensity is carried by a single strong spot on the diagonal, ruling out the presence of a plasmon. This finding is quite surprising, since the 5.16 eV peak is of pure silver nature according to previous fragment analysis, so it would have been rather natural to ascribe plasmonic behavior to such silver peak. On the other hand, the fragment nature of the 4.16 eV plasmonic peak is essentially Pt→Ag, with low silver participation in the initial states. Also this finding is interesting, since plasmonic behavior is expected to be limited to silver contributions.
In order to extrapolate Pt doping effect to zero, we have calculated a series of pure silver clusters ( Figure S2) obtained by substituting Pt atoms in the present bimetallic clusters with Ag atoms, imposing charges required for closed@shell electronic structures. The spectra are characterized by an intense plasmonic peak just above 3 eV, which becomes narrower and stronger as the particle size increases. Comparing these spectral features with those of the bimetallic clusters we observe that in pure Ag clusters the photoabsorption starts suddenly around 3 eV, while in bimetallic clusters the photoabsorpion starts smoothly around 1 eV.
Moreover, the spectral features well defined in Ag clusters become unresolved and just sketched<?> peaks over a smooth background in AgPt. In general, the presence of Pt reduces the plasmonic behavior and makes all the spectral features much less prominent. the lower-energy feature displays strong plasmonic character, which is lost in the feature at higher energy. It is interesting that for the lager clusters the low-energy peak is still plasmonic but its intensity is much lower than for pure Ag clusters (see Figure S2). This effect can be explained by noticing negative contributions in ICM-OS plots near the diagonal: such contributions give rise to a destructive interference strongly reducing the plasmon intensity.
Summarizing the results of the above analysis we can rationalize the plasmonic peak in the AgPt nanoalloys as follows: when Ag nanoparticles are doped by a small amount of Pt the plasmon keeps its Ag nature only as concerns the final states, which remain on silver. On the other hand, the initial states are mainly ascribed to Pt. This happens because Pt has an excess of electronic charge compared to Ag. This rationalization in terms of electronic structure is consistent with the Partial Density of States (PDOS) plot in Figure 7, where the energy scale is shifted with respect to Fermi energy taken as the zero energy. All NPs considered in this work show an intense PDOS in the range -6 eV -2 eV, which corresponds to the 4d silver band. As the energy increases the Ag PDOS smoothly decreases and it becomes irregular but constant in average above the Fermi level. Differently, the Pt contribution is rather uniform and low on the Ag 158 Pt 43 with the layers perpendicular to the (111) direction (L1 1 arrangement), the latter modeling the experimentally observed structure. 12 Two homotops of these NPs featuring typical core@shell ordering have been also considered as references (see structures of these four models in Figure 8). DFT energy shows that for the particle Ag 161 Pt 40 the L1 0 layered structure is less stable than the core@shell one by 1.7 eV, while for the similarly large particle Ag 158 Pt 43 the L1 1 layered structure is more stable than the core@shell one by 1.6 eV. Although this finding is consistent with the experiment, indicating a preference for the L1 1 ordered phase, the energy difference is so tiny (around 8 meV per atom) that it is more reasonable to assume very similar stability of the L1 1 , L1 0 and core@shell structures. This is supported by another stability indicator, so-called excess energy 31  (Ag 151 Pt 50 , core@shell) and -37 (Ag 172 Pt 29 , core@shell). Therefore, it is very hard to discriminate such ordered phases by simple energetic analysis, and other properties such as optical ones could be useful to help in this respect. Figure 9 displays photoabsorption spectra of these four models, calculated with charges giving closed-shell electronic structures. For the core@shell ordering spectra in panels a) and b), as already mentioned,the partial dipole contributions are very similar to each other due to the spherical shape of the systems. Going to the L1 0 and L1 1 layered structures -panels c) and d) -we observe that the profile is much less smooth than that in the corresponding core@shell structures. Moreover, the Z component of the electric dipole contribution for the L1 0 structure displays features, which are out-of-phase with respect to the X and Y components. Both effects can be ascribed to the presence of layers which reduce the symmetry and increase anisotropy. It is worth to directly compare the spectra of core@shell and layered structures -panels e) and f)-in order to easier identify the effect of the chemical order. For both NPs the effect is quite modest. At low energy (below<above?> 3 eV) the layered structure promotes a weak increase of the intensity, a slightly more pronounced increase is apparent between 4 and 6 eV for both NPs. Above 6 eV chemical ordering has almost no effect for Ag 158 Pt 43 , while for Ag 161 Pt 40 a slight decrease is observed. The rather modest effect of the chemical ordering is not surprising since in both structures the Pt atoms are confined inside the particles covered by a complete Ag skin. This suggests that the photoabsorption is localized on the surface, which is well established for plasmons. This is also consistent with a previous computational work 11 on AgPt clusters, where the effect of Pt doping on optical properties was found more pronounced when Pt atoms were set on the surface.
Partial DOS plots of the core@shell and layered Ag 158 Pt 43 and Ag 161 Pt 40 models are shown in Figure S5. Although the global shape closely resembles that in Figure 7, Pt partial DOS in the layered models is strongly enhanced between 0 and 2 eV below the Fermi energy. This is consistent with the weak absorption enhancement at low energy. The fragment analysis ( Figure   S6) agrees with previous data. Only the Ag→Ag partial contribution deserves a remark: while for the core@shell orderings it remains very low up to 4 eV, this contribution is more intense and structured between 2 and 4 eV for the layered orderings. Also the ICM-OS analysis ( Figure S7) of the core-shell and layered Ag 158 Pt 43 particles taken at two energies (4.65 and 5.54 eV) corresponding to the most salient spectral features agrees with the previous analysis: only the peak at lower energy (4.65 eV) displays a plasmonic behavior. It is worth noting, however, that the plasmonic behavior is enhanced in the layered model compared to the core-shell one.

Conclusions
The present study shows that for AgPt nanoclusters with core@shell structure, the optical properties are sensitive to both Pt concentration and the size. For low Pt concentration (Ag:Pt = 6:1) the size effect is smooth, while a less regular trend is identified when Pt concentration is higher (Ag:Pt = 3:1). On the other hand, for constant cluster size, the Pt concentration is found to shift intensity from the low energy part (which is depleted up to around 6 eV) to the high energy region (above 7 eV) where intensity increases with Pt concentration. Such findings have been rationalized in terms of recently developed tools (fragment analysis and <to spell-out again in the Conclusions>ICM-OS). Fragment analysis suggests that the low energy part of the spectrum is dominated by the Pt→Ag contribution, while at higher energy Ag→Ag contribution plays a major role. ICM-OS analysis allows identifying collective behaviors and therefore to follow the evolution of the plasmon for different models.
Special attention has been devoted to clusters with ordered phase (in particular L1 1 one), which have been recently found in experiments for AgPt nanoparticles with size below 2.5 nm.
In this case rather modest differences with respect to core@shell structures have been found in the optical properties. This has been rationalized by the fact that the photoabsorption takes place mainly on the surface which is made of Ag atoms. In fact, also the studied ordered phases are covered by a monatomic silver skin due to the energetic destabilization of Pt atoms in all surface positions compared to inner ones. Although the effect of the chemical order on the optical properties is modest, some trends have been identified, especially at low energy. Therefore it would be very desirable to have new photoabsorption experimental data on different phases to verify, if such trends can be detected by experiments and therefore are useful to assess the structures of these elusive systems.