Interface-mediated Kirkendall effect and nanoscale void migration in bimetallic nanoparticles during interdiffusion

Abstract

At elevated temperatures, bimetallic nanomaterials change their morphologies because of the interdiffusion of atomic species, which also alters their properties. The Kirkendall effect (KE) is a well-known phenomenon associated with such interdiffusion. Here, we show how KE can manifest in bimetallic nanoparticles (NPs) by following core-shell NPs of Au and Pd during heat treatment with in situ transmission electron microscopy. Unlike monometallic NPs, these core-shell NPs did not evolve into hollow core NPs. Instead, nanoscale voids formed at the bimetallic interface and then, migrated to the NP surface. Our results show that: (1) the direction of vacancy flow during interdiffusion reverses due to the higher vacancy formation energy of Pd compared to Au, and (2) nanoscale voids migrate during heating, contrary to conventional assumptions of immobile voids and void shrinkage through vacancy emission. Our results illustrate how void behavior in bimetallic NPs can differ from an idealized picture based on atomic fluxes and have important implications for the design of these materials for high-temperature applications.

Fig. 1

Heating-induced changes in the morphology of Au core–Pd shell (Au–Pd) nanoparticles (NPs). a In situ scanning transmission electron microscopy (STEM) images of the same NP after heating at 500 °C for 5 and 25 min, and their corresponding energy-dispersive X-ray (EDX) elemental maps. In the STEM images, voids show up as round areas of dark contrast. Dashed boxes highlight some of the voids that appeared in the NP during the heating. b (Left) Diffusion profiles extracted from the EDX measurements presented in the panel (a) and their respective error-function fits (black curves). (Right) Diffusion coefficients of Au diffusing in Pd obtained from extended studies (Supplementary Fig. 2), where we followed the diffusion of Au in these NPs as a function of temperature and time. The error bars indicate the standard deviation in the diffusion coefficients obtained at each temperature. An Arrhenius fit (black line) to the averaged diffusion coefficients gives the activation energy of E_a = 1.6 ± 0.2 eV. c In situ transmission electron microscopy (TEM) image sequence of a NP during heating to 500 and 600 °C. In the TEM images, the voids (highlighted with dashed boxes) show up as dark rings with a light core. d Images recorded after increasing the temperature to 650 °C (Supplementary Movie 1). The time-stamps are pegged to time in the movie. The position of the void denoted as 1 is tracked over 400 s in e. At t = 440 s, the void annihilated at the surface on the right of the NP

Fig. 2

Heating-induced changes in the morphology of Pd core–Au shell (Au–Pd) NPs. a In situ STEM images of the same NP after heating at 500 °C for 3 and 20 min, and their corresponding EDX maps. The dashed box highlights a void in the NP. b (Left) Diffusion profiles extracted from the EDX measurements presented in the panel (a) and their respective error-function fits (black curves). (Right) Diffusion coefficients of Pd diffusing in Au obtained from extended studies (Supplementary Fig. 8), where we followed the diffusion of Pd in these NPs as a function of temperature and time. The error bars indicate the standard deviation in the diffusion coefficients obtained at each temperature. An Arrhenius fit (black line) to the averaged diffusion coefficients gives the activation energy of E_a = 1.0 ± 0.6 eV. c In situ TEM image sequence of a NP during heating to 500 and 600 °C

Fig. 3

Formation energy of Pd vacancies and a schematic illustrating interface-mediated Kirkendall effect (KE). a Density functional theory (DFT) calculations of the formation energy of a Pd vacancy at different locations of the Au–Pd layered structure (Supplementary Fig. 11). Migration of the Pd vacancy across the Au–Pd interface is accompanied by diffusion of a Au atom into a Pd lattice site. b A schematic comparison of the conventional KE and the observations of interfacial void formation and void migration in this study. The conventional view for KE is that interdiffusion leads to a flux of vacancies towards the faster-diffusing side. When the vacancy concentration in the diffusion zone reaches supersaturation, the vacancies can condense at defect sinks on the faster-diffusing side to form voids that grow with time^,. In our study, we see void formation at the bimetallic interface due to the clustering of vacancies at the interface leading to void formation and subsequent void migration to the surface. Green and red spheres represent Au and Pd atoms, respectively

Fig. 4

Void formation and migration in Au–Pd–Au sandwich NPs at elevated temperatures. a In situ STEM (upper) and EDX (lower) images of the as-synthesized NP morphology and after heating to 500 °C. The images indicate that the voids formed in these NPs are larger than the bi-layer NPs. Voids are denoted by dashed squares. b–d In situ TEM image sequences of void migration in a Au–Pd–Au NP after heating to 500 °C, 600 °C (Supplementary Movie 4), and further to 650 °C (Supplementary Movie 5). The time-stamps in c and d are pegged to time in the supplementary movies. Note that void 1 increased in size between t = 0 s and t = 180 s at 600 °C. It can be seen from Supplementary Movie 4 that this increase was due to two void coalescence events between t = 60–80 s. Void 2 got annihilated at the surface between t = 180 s and t = 360 s at 600 °C (Supplementary Movie 4), resulting in a small hole on the NP surface. The sequence at 650 °C shows the extended migration of void 1 (Supplementary Movie 5). Movie 5 is recorded 70 s after the end of Movie 4