Bulk Immiscibility at the Edge of the Nanoscale Michael Chatzidakis,† Sagar Prabhudev,† Peyman Saidi,‡ Cory N. Chiang,† Jeffrey J. Hoyt,† and Gianluigi A. Botton*,† †
Department of Materials Science and Engineering, McMaster University, Hamilton, ON L8S 4L8, Canada Department of Mechanical and Materials Engineering, Queen’s University, Kingston, ON K7L 3N6, Canada
‡
S Supporting Information *
ABSTRACT: In the quest to identify more effective catalyst nanoparticles for many industrially important applications, the Au−Pt system has gathered considerable attention. Despite considerable effort the interplay between phase equilibrium behavior and surface segregation in Au−Pt nanoparticles is still poorly understood. Here we investigate the phase equilibrium behavior of 20 nm Au−Pt nanoparticles using a combination of high-resolution scanning transmission electron microscopy and a hybrid Monte Carlo and molecular dynamics atomistic simulation technique. Our approach takes into account the effects of immiscibility, elastic strain, interfacial free energy, and surface segregation. This is used to explain two key phenomena taking place in these nanoparticles. The first is whether the binary system remains immiscible at the nanoscale, and if so what morphology would the secondary phase take. Our findings suggest that even at sizes of 20 nm, thermally equilibrated Au−Pt nanoparticles remain largely immiscible and behave thermodynamically as bulk-like systems. We explain why 20 nm Au−Pt nanoparticles phase separate into hemispheres as opposed to a thick-shelled core−shell structure. These insights are central to further optimization of Au−Pt nanoparticles toward enhanced catalytic activities. The phase-separated Janus particles observed in this study offer enhanced material functionality arising from the nonuniformity of their plasmonic, catalytic, and surface properties. KEYWORDS: phase separation, nanoparticles, Monte Carlo, electron-energy-loss spectroscopy, molecular dynamics, scanning transmission electron microscopy, Janus particles
T
minor depression in the melting temperature occurring at larger sizes around 20 nm. Hoyt showed that the process of spinodal decomposition can be suppressed at small sizes,8 and Yang et al. demonstrated that the critical point from the order−disorder reaction in FePt decreases with decreasing particle radii.9 The Au−Pt system is interesting from a phase equilibria point of view because the bulk alloy exhibits a coherent miscibility gap, whereas it is known that Au tends to surface segregate on Pt at the bulk10−12 and nanoscale13−16 due to the intrinsic lower surface energy of Au atoms over Pt atoms. The phase-separating tendency can produce hemispheric lobe structures,17 and, by contrast, the surface segregation effect would favor a core−shell structure of Pt surrounded by Au. Petkov and co-workers have recently shown experimentally that Au−Pt nanoparticles smaller than 5 nm in diameter exhibit miscibility at some temperatures and compositions traditionally within the miscibility gap of the bulk Au−Pt binary phase diagram;18 however the extent of the depression of the miscibility gap as a function of particle size, composition, and
he self-assembly of multicomponent nanomaterials by thermal equilibration is a scalable device fabrication technique that produces complex nanoscale architectures capable of having unique properties and narrow size distributions. When constrained to a nanoparticle geometry, immiscible metals can phase-separate into hemispheres. Nanoparticles exhibiting this two-lobe structure are termed Janus particles,1−4 and these morphologies allow highly functional materials to be synthesized. The exploitation of Janus particles has already been used in the Au−Fe3O4 heteronanoprobe system, where the difference in surface properties allowed unique surface ligands to be attached to each half of the probe.5 Despite the potential applications of Janus particles, a complete understanding of the phase equilibrium behavior of such systems is not straightforward, and there exist several examples where the high surface-area-to-volume ratios of nanoscale materials change the thermodynamics of these systems to differ drastically from the bulk. Navrotsky et al. has shown that nanoparticles of iron oxide have a different crystal structure than the bulk due to the effect of Laplace pressure.6 In the pioneering study by Buffat and Borel on Au nanoparticles,7 a sharp drop in melting point was observed when particles were smaller than 10 nm in diameter, with only © 2017 American Chemical Society
Received: July 11, 2017 Accepted: October 23, 2017 Published: October 26, 2017 10984
DOI: 10.1021/acsnano.7b04888 ACS Nano 2017, 11, 10984−10991
Article
www.acsnano.org
Cite This: ACS Nano 2017, 11, 10984-10991
Article
ACS Nano
Figure 1. (a) Au nanoparticles used as substrates for the deposition of Pt (scale marker is 10 nm). (b) Central Au nanoparticle with surfacenucleated Pt nanoclusters (scale marker is 10 nm). (c) Representative loading of Au+Pt nanoparticles (highlighted with arrows) on carbon black (scale marker is 200 nm). (d1−d3) Three separate nanoparticles after 1 h of annealing (scale marker is 10 nm). (e1−e3) Three separate nanoparticles after 2.5 h of annealing (scale marker is 10 nm). (f1−f3) Three separate nanoparticles after 18 h of annealing (scale marker is 10 nm). (g) High-resolution HAADF STEM image of a Au−Pt nanoparticle after 18 h of annealing.
temperature is still debated.15,17−21 For example, according to simulations by Xiao et al. using an angular embedded atom method, 5 nm Au−Pt nanoparticles (∼2000 atoms) are largely still immiscible similar to the bulk.13 The aim of the present study is to understand the morphology of Au−Pt nanoparticles and elucidate the interplay between the surface segregation effect and the depression of the miscibility gap at small particle sizes. The study involves a combination of advanced characterization tools and atomistic simulations. First we utilize scanning transmission electron microscopy and electron energy loss spectroscopy (STEMEELS) to characterize the morphology and chemical composition of individual 20 nm Au−Pt nanoparticles after long annealing times. To understand the thermodynamic equilibrium morphology, we then simulate Au−Pt nanoparticles using a hybrid Monte Carlo (MC) and molecular dynamics (MD) approach, where the interatomic potential describing the Au−Pt system was derived from the bulk properties of the alloy. We present a method for predicting the morphology of large (>100 000 atoms) binary surfacesegregated nanoparticles (single phase alloy and phaseseparated) that accurately predicts the behavior seen in experiments. Such models take into account surface energy, surface segregation, interfacial strain, and immiscibility. By combining the results of high-resolution STEM-EELS chemical maps of the distribution of Pt and Au of individual 20 nm Au− Pt nanoparticles as well as atomistic simulations, we characterize the morphology of single- and dual-phase Au−Pt nanoparticles without the use of bulk-ensemble-averaged characterization techniques. We demonstrate that the Au−Pt phase separates even at particle sizes of 20 nm, and, although a thermodynamic model based on surface segregation and surface energies suggest a core−shell morphology, inclusion of elastic strain effects favors a lobe structure.
ultrasonicated with carbon black (Vulcan Carbon XC-72R) to achieve a monodisperse loading of 1 particle per 100−300 nm2. Such low loading was required to ensure annealing that is free from coarsening interactions such that isolated systems could be studied. Au@Pt/C precursors solvated by ethylene glycol were heated at 190 °C under N2 for solvent evaporation. Dried Au@Pt/C nanoparticles were annealed in a quartz tube furnace for 1, 2.5, and 18 h at 403 ± 5 °C in 1 atm of N2. The temperature was chosen to compare results to two prior experimental studies by Braidy et al.17 and Wanjala et al.19 The full procedure can be found in the Methods section. Observations were carried out using conventional imaging in transmission electron microscopy (TEM) and high-angle annular-dark-field (HAADF) STEM (microscope details and conditions in the Methods section). The evolution of the complex particles toward an equilibrium structure can be observed in Figure 1. After 1 h of annealing (Figure 1(d)), the original 2−3 nm Pt nanoclusters, originally isolated on the surface of Au, sinter together and create rounded 5−6 nm particles on the surface of the Au nanoparticles. Following 2.5 h of annealing treatment, most of the high contact angles between the Pt particles and the Au substrate significantly decrease, indicating that there is wetting of Pt on Au and only a few domains of Pt remain visible on each Au nanoparticle. This preliminary surface wetting, as opposed to the direct absorption of the Pt clusters into the Au, is consistent with prior studies on the increased mobility of surface atoms 24−27 in nanoparticles. Since Au and Pt interdiffuse by substitution, we can expect a high rate of diffusion when Pt diffuses on the surface and a slower rate of diffusion when Pt is in the interior of the Au nanoparticle. STEM-EELS. Elemental maps showing the distribution of Pt and Au nanoparticles annealed for 18 h were obtained with STEM-EELS using the quantification procedure described in the Methods section EELS Quantification. Au and Pt signals (originally overlapping) were separated by a multiple linear least-squares (MLLS) technique using reference spectra derived from the pure components. The full raw spectra and MLLS fits can be found in the Supporting Information (Figures S1 and S2). The compositions of each nanoparticle in Figure 2 were estimated to be 36 ± 10 at. % Pt (left) and 39 ± 10 at. % Pt (right). Both of these phase-separated nanoparticles have
RESULTS AND DISCUSSION Synthesis of Phase-Separated Nanoparticles. We synthesized Au−Pt nanoparticles with a two-step approach. First ∼15 nm Au nanoparticle precursors were synthesized by tetrachloroauric acid reduction with citric acid using the method developed by Frens.22 Next, small domains ( 0. Being able to accept some unfavorable switches is essential for an accurate simulation. By only accepting downhill switches, it is possible to end in a local minima and not at a true energy-minimized equilibrium. Being able to accept some unfavorable switches allows the simulation to traverse over small energy barriers. The outlined acceptance probability will adjust concentration until some equilibrium concentration corresponds to the fixed initial conditions {N, V, T, Δμ}. However, it should be noted that since Δμ must be unique for the corresponding equilibrium concentration, the SGC ensemble is not suitable for multiphase simulations. We use the SGC-MC method in the present study for three main purposes. First SGC-MC was used to construct the bulk phase diagram by creating chemical potential curves. This method is also used to benchmark the accuracy of the potentials. Using the same technique as the bulk phase diagram generation, a chemical potential curve of free 20 nm nanoparticles was also created. These single-phase particles before and after the miscibility gap are shown in Figure 3. The chemical potential of the discontinuity in the two-phase regions of 20 nm nanoparticles was also resolved in simulating the single-phase particles. This value is required for the later simulations of multiphase Au−Pt nanoparticles in the varianceconstrained SGC-MC (VC-SGC-MC) ensemble. A full description explaining the VC-SGC-MC ensemble can be found in the Supporting Information. Accuracy of the EAM Potential. The EAM potential was validated by constructing the bulk phase diagram using an SGC-MC approach similar to Hoyt et al.35 The bulk phase diagram was compared to the experimental Au−Pt phase diagram after refining the interatomic potential. Recently, Xu et al. reassessed experimentally the Au−Pt phase diagram using the diffusion couple method. They found the critical point of the miscibility gap to be 1473 K at 56 at. % Pt.21 The critical 10987
DOI: 10.1021/acsnano.7b04888 ACS Nano 2017, 11, 10984−10991
Article
ACS Nano
Figure 4. Quarter slices of nanoparticles to show the interior of the nanoparticles. The direction vectors shown are from the simulation initialization. Each principal stress is shown and color coded to show tension and compressive effects. (a) Quarter slice of a Au0.56Pt0.44 nanoparticle equilibrated by VC-SGC-MC at 673 K. This energy-minimized structure exhibits a hemispheric “lobe” morphology. (b) Quarter slice of a Au0.56Pt0.44 nanoparticle by manually creating a spherical secondary phase and annealing for 0.6 ns at 673 K using MD for equilibration. The color scaling is identical to the “lobe” separated nanoparticle.
which the balance between stress and interface energies is minimized. Although the results presented here focus on the 20 nm particle size, we have also investigated the change in the miscibility gap and the phase morphology for two additional sizes, 5 and 8 nm. The 5 nm truncated cuboctohedral Au−Pt particle was simulated, and its miscibility gap was characterized at 673 K (Figure S3). This particle is still shown to have a miscibility gap despite its small size. Although no VC-SGC-MC simulations were performed on faceted particles to show the influence on morphology, we have provided experimental verification that lobe separation happens on faceted 20 nm particles. Additionally, there are other studies with more clear HRTEM images,17 which show clearly the influence of faceting on the morphology of secondary phases in 20 nm Au−Pt particles. To further probe at the extent of miscibility gap suppression as a function of particle size, the miscibility gap of a 5 nm faceted truncated cuboctohedral Au−Pt particle was characterized using an SGC-MC approach. A chemical potential curve was created by finding the equilibrium concentration corresponding to a fixed imposed chemical potential. The bulk (periodic), spherical 20 nm, and truncated cuboctohedral 5 nm Au−Pt chemical potential curves at 673 K are shown in Figure S3 in the Supporting Information. The miscibility gap at 5 nm at 673 K is reduced when compared to the bulk by a Pt concentration of ∼35 at. %. The miscibility gap is not entirely suppressed outright, even at 5 nm size scales. An 8 nm spherical nanoparticle was also simulated using VC-SGC-MC to determine the morphology of the secondary phase, and it was also found to produce a lobe morphology (Figure S4). In an attempt to understand the morphology of these subsurface lobe Pt-rich phase structures, a comparison was made to a core−shell structure consisting of a spherical Pt-rich phase manually created in LAMMPS and heated at 673 K for
The Au surface atoms of the nanoparticle are not uniform in energy despite having a predominantly Au shell and differ depending on the underlying Au-rich or Pt-rich phase. The Au surface-segregated layer over top of the Pt-rich phase is destabilized by the Pt-rich phase underneath it. Our calculations demonstrate the interfacial free energy of Au segregation on a Pt-rich phase to be 40% greater than a pure Au/vacuum interface. Since the surface energy of pure Au is less than that of a surface-segregated Au−Pt alloy, it is not immediately obvious why a hemispherical lobe structure of the Pt-rich phase is thermodynamically favorable over a so-called “core−shell” structure characterized by a spherical Pt phase surrounded by a thick Au shell. We propose that the observed morphology is driven by elastic strain energy effects. In addition to this, the extent of surface segregation is typically just a few atomic layers,37,38 and as such, there is minimal driving force (via interfacial free energy) for a thick-shelled structure (10+ layers) to form. One important point in studying the morphology of complex systems is the cross effect of energy components (in this case elastic energy and interface energies) on thermodynamics stability. The stress in the system can be considered approximately as the sum of three parts: (i) vibrational displacement, (ii) structural contribution (the stress due to structural distortions caused mainly by the Gibbs−Thomson effect and curvature), and (iii) chemical contribution (due to bonding to dissimilar neighbors).39 Energy minimization in simulations filters the vibrational displacement; thus, the combination of structural and chemical stresses contributes to the free energy of the system. The other energy components are the interface energies, where the formation of facets might result in the alteration of interface energies. However, for systems where the contributions of both elastic and interfacial energies are significant, the final morphology is the one for 10988
DOI: 10.1021/acsnano.7b04888 ACS Nano 2017, 11, 10984−10991
Article
ACS Nano
integral part of the stability of a phase-separated lobe morphology. Such a lobe morphology was also identified experimentally in 20 nm Au−Pt nanoparticles after extracting Au and Pt EELS signals using an MLLS technique. The same EELS chemical maps also suggest Au surface segregation on top of the Pt-rich phase, which agrees with the results of the simulation and the literature.
0.6 ns of simulation time in MD. Both configurations are Au0.56Pt0.44. Both configurations were conjugate gradient minimized at the final time-step, while LAMMPS was outputting the stresses40 for mapping. The key difference between these two morphologies is the amount of surface area of each of the interfaces (γ(Au/Vac), γ(Au‑Segr(Pt)/Vac), and γ(Au/Pt)) as well as the curvature of each interface, which affects the strain acting on each phase. The color-coded scaling on the stress maps for both morphologies of the Pt-rich phase in Figure 4 are identical. It is also clear from Figure 4 that the lobe morphology (top) is a stress-minimized structure considering a large volume fraction of the nanoparticle is relaxed. This configuration is even favorable despite having an unfavorable surface-segregated interface (when compared with the pure Au interface). A comparison of the total internal energy of both configurations can be seen in Figure S4 in the Supporting Information, which does indeed show the hemispherical lobe structures are lower in internal energy. It should be noted that while Au does completely surround the Pt-rich phase in the lobe configuration, as it does in the spherical morphology, there are energetic differences between a Au-rich phase and Au segregation. The stress of the surface atoms shown in Figure 4, however, should be interpreted with caution. The stress mapping40 approach used is applied to a conjugate gradient minimized structure (forces atoms into original lattice positions) that undoes surface relaxation. Therefore, accurate values of surface stresses cannot be calculated using this method, and surface atoms were excluded for our strain energy calculation. Au monolayers on top of a Pt-rich phase should indeed be compressed due to the smaller lattice parameter of Pt compared with Au. To calculate the atomic strain energies, it was necessary to find the per atom stress-tensor. This was calculated by finding the force and distance between neighbors in the deformed solid. The global strain energy of the lobe nanoparticle was found to be ∼8 % less than the thick-shelled core−shell configuration. The results shown in Figure 4(a) represent the theoretical maximum allowable internal stresses (a defect-free coherent interface). Such defects are present in the MD-equilibrated simulation in Figure 4b, which reduce internal stress. However, the total strain energy of the lobe structure (top) is less than the core−shell structure containing interfacial dislocations (bottom). Since MD simulations are of a limited time range, we cannot definitively conclude whether interfacial defects should be present in a fully equilibrated lobe-structure nanoparticle. Nevertheless, the presence of interfacial defects will reduce the total elastic strain energy, and we can conclude that the lobe structure is energetically favorable over the spherical core−shell morphology.
METHODS Synthesis of Au@Pt Precursors. A two-step reduction method was used to synthesize the Au@Pt nanoparticles where gold nanoparticles were first synthesized and then used as seeds for the heterogeneous nucleation of platinum. Gold seed nanoparticles were synthesized using the citrate reduction technique developed by Frens22 wherein chloroauric acid (10−2 wt %) was reduced using trisodium citrate (1 wt %) solution. The wine red colloidal Au dispersion (average size ∼17 nm) was maintained at 273 K by boiling for a reaction time of 40 min. These Au nanoparticles were coated with atomic clusters/nanoparticles (
