Article pubs.acs.org/JPCC
Fe−Ni Nanoparticles: A Multiscale First-Principles Study to Predict Geometry, Structure, and Catalytic Activity Juhani Teeriniemi,*,† Marko Melander,‡ Saana Lipasti,† Richard Hatz,† and Kari Laasonen† †
School of Chemical Technology, Aalto University, P.O. Box 16100, FI-00076, Aalto, Finland Department of Energy Conversion and Storage, Technical University of Denmark, DK-4000 Roskilde, Denmark
‡
S Supporting Information *
ABSTRACT: Nanoparticles of iron and nickel are promising candidates as nanosized soft magnetic materials and as catalysts for carbon nanotube synthesis and CO methanation, among others. To understand geometry- and size-dependent properties of these nanoparticles, phase diagram of Fe/Ni alloy nanoparticles was calculated by density functional theory and cluster expansion method. Ground state convex is presented for face-centered cubic (FCC), body-centered cubic (BCC), and icosahedral (ICO) particles. Previous experimental observations were explained by using multiscale model for particles with realistic size (diameter ≥2 nm). At size 1.5 nm, geometry changes from BCC at low X(Ni) to icosahedral at high X(Ni). FCC is stabilized over icosahedral geometry by increasing number of atoms from 561 to 923. In large FCC particles, there is enrichment of Fe atoms from core to shell beneath surface, while surface and core are enriched by Ni atoms. Catalytic enhancement effect in CO methanation was found to be due to Ni incorporating in the active sites which brings adsorption energy of oxygen closer to the optimum. The predicted phase diagrams and implications on catalysis are expected to help rationalization of experimental results and provide guidance for design of Fe/Ni-based nanomaterials.
■
INTRODUCTION
Catalytic properties of Fe/Ni NPs are strongly dependent on geometry, size, and configuration of atoms. For example, Ni0.27Fe0.73 NPs have been shown to produce almost chirally pure SWCNTs while other atomic compositions give raise mixture of nanotubes.5,8 In the CO methanation reaction, Fe0.5−0.25Ni0.5−0.75 alloys are more active and selective than monometallic Fe and Ni.7,9,10 The CO methanation reaction takes place at undercoordinated sites (steps, kinks,corners, or vertices).11,12 Even at geometrically similar steps, the exact locations of Fe and Ni atoms around the reaction site have been demonstrated to change activation energies of different reaction pathways by 0.1−0.5 eV making the Fe/Ni a structurally sensitive methanation catalyst.13 Combination of the size and geometry dependent reactivity, and the synergetic benefits, offers a rich chemistry on the bimetallic iron−nickel particles. In order to fully utilize the potential of bimetallic catalysts, geometry, segregations, and alloy formation must be understood. However, both experimental and computational structure determination is challenging. Two seemingly contrary alternatives have been suggested for atomic arrangement in Fe/ Ni NPs. Adsorption studies of Parks et al.14 showed that Fe locates beneath the surface in Ni-rich small icosahedral NPs. On the contrary, X-ray absorption and Mössbauer studies of
Finding better materials is an ever-growing goal of chemistry and materials science. Bimetallic nanoparticles (NPs) are under extensive study and interest due to their unique properties, often different from their bulk counterparts or monometallic nanoparticles. From a chemical perspective, the most intriguing possibility are synergetic effects in catalysis, where a combination of two metals shows markedly enhanced catalytic properties compared to the pure metals.1,2 For example, higher activity, prolonged stability, better selectivity, and increased conversion have been obtained via synergetic NP catalysis.1 Factors causing the observed improvements are elusive but their origin lies in differences of geometric and electronic factors of bimetallic particles compared to the single metal NPs.2,3 In addition to chemical improvements, use of bimetallic NPs can also have a beneficial economical factor if expensive noble metals are used together with, or even replaced by, cheap abundant elements. In this Article, we focus on the bimetallic NPs of Fe and Nithe most abundant and affordable metals. Besides their availability, Fe/Ni NPs are found to exhibit synergetic catalysis. The Fe/Ni NPs are used as a catalyst in, for example, selective hydrogen generation from hydrous hydrazine,4 synthesis of single-walled carbon nanotubes (SWCNT),5,6 and for CO methanation.7 Furthermore, bulk Fe/Ni alloys are widely used as a soft magnetic material. © XXXX American Chemical Society
Received: October 31, 2016 Revised: December 21, 2016 Published: December 30, 2016 A
DOI: 10.1021/acs.jpcc.6b10926 J. Phys. Chem. C XXXX, XXX, XXX−XXX
Article
The Journal of Physical Chemistry C Margeat et al.15 showed progressive enrichment in Fe atoms from the core to the surface in NPs with diameters of 2.8 ± 0.3 nm. According to computational studies by the density functional theory (DFT), small Fe/Ni NPs favor icosahedral geometries, and Ni enriches on the surface.16−21 Apart from the smallest NPs, pure ab initio methods are computationally too demanding. Faster atomistic models, such as Gupta potential22 or Sutton−Chen potential23 have been used to study nanoalloys but, in general, the accuracy of such empirical 2-body models is uncertain. The fundamental problem is that higher order contributions to total energy are often non-negligible, and, in the case of metallic interactions, the n-body decomposition converges slowly.24 Often empirical models are constructed to reproduce known properties of a material. However, fitting of “correct” empirical parameters can be impossible in cases where correct answers are not known. Fe/Ni NPs are such case. Indeed, only an ab initio-based model can explain all complexity observed in the nanoscale alloys. In bulk alloys, the cluster expansion (CE) method25 has been used with excellent accuracy.26−28 Success of the CE method is due to fact that it coarse-grains an ab initio model to n-body model with sufficient amount of complexity for higher order contributions. No empirical parameters are needed. In fact, combination of DFT, CE, and automated interaction selection scheme allows fully automated modeling for any system for which DFT calculations can be done.28 This is not the case with force field approaches. Instead, a lot of skill and testing is included when force field is selected or constructed to a new system. Furthermore, CE calculations are very fast and easily parallelizable. After DFT calculations are done, CE calculations can be done even by a normal PC. However, CE does not include relaxation of atomic positions which restricts it to rather large systems. Recently, the CE method has been applied to NPs with fixed size and geometry.29−31 However, NPs are usually not completely monodisperse but exhibit tunable distribution of sizes and geometries.32 In order to have a realistic view on nanoscale phenomena, also geometry- and size-dependencies must be taken into account. Very recently, Müller et al.33,34 used the CE method to simultaneously optimize atomic order, size, and shape of a nanoparticle. The main goal of the current work is to understand the structures of bimetallic Fe/Ni NPs. To achieve this, we construct a multiscale CE method to study configurations of different geometries and their relative stabilities as a function of size. Phase diagrams are provided to rationalize and predict physical and chemical properties of the Fe/Ni NPs. Obtained structures are used to predict CO methanation activity on 1.5 nm Fe/Ni NPs. It is shown that 1:2 or 1:3 mixture of Fe and Ni at the active site will provide improved catalysts compared to pure NPs and crystal surfaces. Selection of the Model Systems. Elemental Fe and Ni NPs have received considerable interest mainly due to their rich geometric35 and magnetic36 phase diagrams. When dispersed as NPs, the magnetic and structural phase diagrams become even more complicated, as these are greatly influenced by the particle size. Here, we focus only on the gas phase particles because things get even more complicated when anchored on a surface. Based on both experimental16−18,37−39 and computation20,21 data, small NPs (0.5−1.4 nm diameter) seem to favor al icosahedral or Mackay transformed icosahedral structures, whereas larger NPs adopt a body-centered cubic (BCC) configuration for Fe and face-centered cubic (FCC) for Ni. Size
range for the geometry transitions is unclear. To compromise between the smallest clusters used in experiments and the largest NPs achievable by accurate DFT calculations, we have chosen to do direct DFT calculations for Fe/Ni particles at the size of 1.5 nm with 145−147 atoms. Based on the discussion above, FCC, BCC, and icosahedral geometries were selected for this study. In real systems, NPs are often not perfectly symmetric and can exhibit amorphous phases.32 This is particularly true for NPs with diameter of
