Equilibrium Cu-Ag nanoalloy structure formation revealed by in situ scanning transmission electron microscopy heating experiments

New physical and chemical phenomena related to nanoscale effects are currently the focus of intense research.^1–4 Nanoparticles (NPs) are characterized by small volumes and a large number of surface atoms, which inherently possess a significantly larger mobility than bulk atoms. As a result, novel structures, physical properties, and processes may occur with NPs that are not observed for their bulk counterparts.^1–4 For instance, the melting point of a metallic NP is typically lower than that of the bulk metal.^5–7 For the metallic copper (Cu) and silver (Ag) system, it is well known that the solid solution of the binary Ag-Cu alloy is unstable, and upon heating an initially amorphous or nanocrystalline Ag-Cu alloy in bulk form will separate into Ag and Cu phases due to the large miscibility gap and positive enthalpy of mixing for this eutectic system.^8–11 It was of interest to determine if the same phenomena would also take place when the Ag-Cu alloy is at the nanoscale.

Molecular Dynamics (MD) simulations have shown the formation of a Ag-rich phase segregated at the surface during thermally induced phase separation processes in nanoscale Ag-Cu alloy rods and wires.² Recent experiments also revealed formation of core-shell Cu-Ag NPs in a self-assembled, free-standing Cu-Ag NP alloy synthesized by magnetron sputtering;¹² however, it is not clear if the NPs formed under these conditions were at equilibrium and, in addition, the process of the core-shell formation was not reported. There are also reports that show the alloying temperature decreases more than several hundred degrees for Ag-Cu samples in a surface geometry, indicative of the reduction of the miscibility gap for the low-dimensional system.¹³ Nevertheless, it remains unclear what the equilibrium Cu-Ag NP alloy structure is, how it forms, and particularly how the new NP alloy structure is related to the equilibrium phase diagram of the Cu-Ag bulk system. The present study is designed to better understand these questions by thermally reacting Cu (∼12 nm) and Ag NPs (∼6 nm) and observing the reaction processes and products using in situ scanning transmission electron microscopy (STEM). The formation of a 3-dimensional (3D), mainly cube-on-cube epitaxial Cu-core and Ag-shell structure was observed during thermal interaction of Cu and Ag NPs at temperatures as low as 150 °C. The structure was formed by the Ag NPs initially wetting the Cu NP along its {111} surfaces at multiple locations, forming epitaxial Ag/Cu (111) interfaces, and subsequently Ag atoms diffusing along the Cu surface until the shell was complete. Results were compared to Monte Carlo simulations to understand the stability of the structure and provide insight into the driving force(s) for the formation process.

Generally, in situ electron microscopy is performed in transmission electron microscopy (TEM) mode, which offers the ability to observe the process in real-time. However, during this study, it was determined that electron beam irradiation of the sample during TEM/STEM imaging severely impacted the Cu-Ag interaction process, preventing continuous, real-time observations from being made. In particular, under normal TEM or STEM imaging conditions, exposure to the beam for as little as 60 s rendered the NPs inactive (see Figure S1 of the supplementary material¹⁴). This change in reactivity is likely due to the formation of a carbon shell around the NPs, similar to that found on Au NPs under electron irradiation by Sutter et al.¹⁵ As a result, the in situ heating experiments had to be performed with the electron beam off or with the electron beam on only for the period necessary to obtain the STEM images (typically, less than 30 s). For this study, the in situ experiment was performed in STEM mode to facilitate direct identification of Cu and Ag NPs based on high-angle annular dark-field (HAADF) Z-contrast imaging;¹⁶ in this imaging mode, the contrast of Ag is brighter relative to Cu due to its higher atomic number.

In situ STEM heating experiments were carried out using the Protochips Aduro^TM in situ heating stage, which is capable of extremely fast heating and cooling rates (1000 °C/s).¹⁷ A FEI Titan^TM G2 80-200 STEM with a Cs probe corrector and ChemiSTEM^TM technology (X-FEG^TM and SuperX^TM energy-dispersive X-ray spectroscopy (EDS) with four windowless silicon drift detectors), operated at 200 kV, was used in this study. The Cu and Ag NPs were synthesized according to the modified literature preparation routes described previously,¹⁸ and had an average diameter of ∼12 and 6 nm, respectively. The TEM specimen was made by pre-mixing the NPs in toluene and placing a drop of the mixed solution on a thin carbon film (<5 nm) supported by the TEM grid. By adjusting the solution concentration, Cu and Ag NPs could be assembled on the carbon film such that a monolayer of NPs was formed (see Figure S2 of the supplementary material¹⁴). Here, the larger Cu and smaller Ag NPs are well mixed, with the smaller Ag NPs present in between and in contact with Cu NPs. This intimate contact is critical for initiating the Cu-Ag reaction, since there is a competitive, low temperature Ag-Ag reaction that also takes place during heating.

The reaction between the Cu and Ag NPs was found to take place at temperatures as low as 150 °C and up to about 300 °C. Over 300 °C, the volatility of the Ag NPs becomes an issue, with most of the Ag NPs evaporating. In the 150–300 °C temperature range, the Cu and Ag NPs react to form a Cu-core and Ag-shell structure. Figure 1(a) shows a high-resolution HAADF image of the core-shell structure formed at 150 °C with a Cu-core of ∼9 nm in diameter (marked by the circle) and Ag-shell thickness of ∼3 nm. The core-shell structure can be clearly discerned due to use of Z-contrast in the HAADF imaging in Figure 1(a). The structure was further confirmed by an EDS element map shown in Figure 1(b) and an EDS line-profile across the particles shown in Figure 1(c). It should be noted that the EDS map in Figure 1(b) reveals a particle with a somewhat different shape than that observed in Figure 1(a). This is due to the continuously changing Ag-shell that evolves when exposed to inherent electron beam irradiation during EDS acquisition.

The composition of the Ag shell was determined by EDS quantification from an area marked by a small circle in Figure 1(b) where there is no overlapping with the neighboring Cu particles. Within the EDS detection limit of about 1%, the shell was determined to be pure Ag (see Figure S3 of the supplementary material¹⁴). The FFT image from the particle (shown in inset in Figure 1(a)) reveals that the Ag shell forms predominately in a cube-on-cube orientation relationship with the Cu core, or variants of the orientation relationship related by (111) twinning. Stacking faults and twin-boundaries are present in parts of the Ag shell, as marked by arrows in Figure 1(a). The moiré patterns, visible most clearly in the right side of the core-shell particle in Figure 1(a), show overlap between Ag and Cu atoms in the electron beam direction, indicating that the Ag is present on the top and/or bottom surface(s) of the Cu core.

A series of HAADF images showing the process of the core-shell particle formation during heating can be found in Figure 2. Figure 2(a) shows a particle with a core-shell structure that has partially formed. It can be observed that an additional Ag NP (marked by a dashed circle near the top of the image) has started to become part of the Ag shell but has not completely merged. Continuous heating at 150 °C for ∼3 min leads to a particle with a more fully formed core-shell structure, as shown in Figure 2(b). Further heating for an additional 3 min gives rise to the final core-shell structure (Figure 2(c)). The moiré patterns at the Cu core become better developed over time as can be clearly observed from Figure 2(a) to Figure 2(c), indicating that the Ag has gradually enveloped the Cu-core. The Cu-Ag core-shell structure (Figure 2(c)) appears to be stable as long as it stays isolated and no new reactions are initiated with neighboring particles. Figure 2(d) shows the particle after an additional 7 min of in situ heating at 150 °C. As can be observed, with the continued heating the Ag shell began new interactions with the Cu particles to its right and in the lower-right corner.

The initial Cu-Ag NP reaction involves Ag wetting on Cu surfaces, forming predominately Ag{111}/Cu{111} interfaces. Figure 3 shows the initial structure of the reaction between Cu and Ag NPs at 200 °C. The Ag, which has a brighter contrast in the HAADF image, has wet several Cu NPs in multiple locations as marked by arrows in the image. The inset image in Figure 3 shows that a sharp Ag{111}/Cu{111} epitaxial interface is formed between the Ag surface layers (as thin as 2–3 {111} monolayers) and the Cu. Although it is difficult to determine the exact chemical composition of the thin surface Ag layer directly, the lattice spacing measured from the brighter Ag layer is about 0.236 nm, consistent with pure Ag (see Figure S4 of the supplementary material¹⁴). This observation indicates that the miscibility gap was likely still present for the nanoalloy system, as even a few monolayers of Ag remain segregated at the surface. Note several Ag NPs can wet single Cu NP at multiple locations (Figure 3). Since the epitaxial Ag/Cu (111) interfaces have several orientation variants, the twins are often formed between different parts of Ag shell (Figure 1(a)) when several Ag NPs involving with interaction with a single Cu NP merge into a single Ag shell.

In effort to understand the stability and driving force for forming the Cu-Ag core-shell structure, the thermal interaction between the Cu and Ag NPs was modeled using Monte Carlo (MC) simulations. These calculations use the embedded atom method (EAM) with the Cu-Ag alloy potential developed by Williams et al.,¹⁹ which gives a reasonably accurate description of the alloy phase diagram, as well as lattice constants, thermal expansion, and surface energies. Simulations were performed using a home-grown code²⁰ which performs Metropolis MC, allowing swaps between atomic types with a small random translation added.

Although unable to provide reasonable depictions of intermediate states, MC simulations are well-suited for describing a final equilibrium structure. For these simulations, a single NP, spatially divided as 1:1 Cu:Ag, was used. This initial structure was selected since the random swaps utilized by the MC algorithm can easily maintain this spherical structure, whereas the evolution from two separate spherical NPs into one single spherical core-shell structure would take significantly longer times. In Figure 4(a), the initial structure for the MC simulations is shown, while in Figures 4(b) and 4(c) the structure arrived at after 400 million MC steps, fully rendered and in cross-section, respectively, are presented. While this is not the final structure, it is clear that the preferred evolution is to form a Ag shell around a Cu core. This structure is still only partially developed because of the way in which MC simulations progress: as the Ag shell becomes more developed (i.e., fewer Ag atoms in the interior), it is less likely that a randomly selected Ag atom (for potential swapping with a random Cu atom) will come from the core region. It is therefore expected that the Ag shell will form asymptotically.

Additional MD simulations (the full results of which are reported in a separate paper²¹) were performed to better understand the formation process, and showed that Ag atoms preferentially diffuse along the surface of the Cu NP (generally along step edges) and do not penetrate the core of the Cu NP. This is in agreement with the observed envelopment noted in the STEM experiments (Figures 2). In addition, the MD simulation showed that a monolayer of Ag atoms remains on the surface of Cu, consistent with experimental observations (Figure 3), indicating preservation of the miscibility gap for the nanoalloy system.

The data from both the in situ STEM experiments and the atomic-scale modeling clearly demonstrate that the Cu-Ag core-shell structure is an energetically preferred structure for nanosized Cu-Ag alloys, and the process of forming a Cu-Ag core-shell proceeds by Ag NPs wetting the Cu NP along the Cu {111} surfaces at multiple locations followed by diffusion of Ag atoms along the Cu surface. Although it is difficult to determine precisely the chemical composition of the Ag-shell due to the EDS detection limit (∼1%), our experimental data also seem to indicate that the miscibility gap for the Cu-Ag nanoalloy system is still present.

The core-shell structure is believed to be energetically preferred for two primary reasons. First, the surface energy of the Cu is significantly higher than that of the Ag. Upon being coated by the Ag, the surface energy of the Cu is lowered, leading to a significant reduction in the system's energy due to the large surface-to-volume ratio of the NPs. In contrast, the increase in energy due to the formation of the Ag/Cu interface is relatively small.^22,23 Experimentally, the Ag shell was found to adopt a cube-on-cube orientation relationship (or its variants by (111) twins) with the Cu core, and formed predominately Ag{111}/Cu{111} interfaces, which were found to have the lowest interfacial energy.²³ Second, the Cu atoms are smaller, which leads to their preferential location inside the core as a mechanism for relieving strain.^3,24–26 As the size of the NP decreases, both of these effects become increasingly important since the surface energy and the strain energy related to the particle size and geometry become increasingly dominant.

These reasons may also explain why the miscibility gap in the nanosized Cu-Ag alloys is still likely preserved. It should be noted that both the surface and strain energies present in the NP core-shell alloy should make the Cu-Ag segregation more favorable in the nanosized alloy than that in the comparable Cu-Ag bulk. This would result in an increase in the miscibility gap or the alloying temperature in the nanosized Cu-Ag core-shell alloy. Nevertheless, the reduction of miscibility gap for Ag-Cu samples in a surface system has been reported previously.¹³ The observation, however, was made on a 2-dimensional Cu-Ag film, which was epitaxially grown on a Ru (0001) substrate. The Cu epitaxy on Ru (0001) substrate had strained the Cu films substantially and made the Cu-Cu distance 5.5% larger than in the bulk Cu. In bulk, the Ag atoms are about 13% larger than Cu atoms. The expansion, therefore, facilitated accommodation of the Ag atoms and likely resulted in the observation of immiscibility reduction in the system.¹³ Finally, considering that both the Ag surface and Ag/Cu interface energies are significantly lower than that of the Cu, that the Ag atoms do not like to be mixed with Cu atoms in the core, and that the surface diffusivity is much higher than the bulk diffusivity, the process of forming a Cu-Ag core-shell by initial Ag wetting of the Cu NPs and surface diffusion along the Cu surface is to be expected.

In summary, for the first time in situ STEM analysis that carefully avoids electron beam irradiation effects revealed the thermal interactions of Cu and Ag NPs that ultimately lead to the formation of a thermodynamically stable Cu-Ag core-shell structure. The initial step of the interaction between distinct Cu and Ag NPs at elevated temperatures (150–300 °C) was found to involve Ag NPs wetting a Cu NP along the {111} surfaces at one or multiple locations, forming Ag{111}//Cu{111} interfaces. The sharp interfaces were found consistently between the epitaxial Ag-shell and the Cu-core. The formation of this core-shell structure was attributed to a nanoscale effect related to the large surface-to-volume ratio of the particles. Finally, although the observations of this letter are limited to the Cu-Ag system, a similar formation process and equilibrium core-shell structure would be expected to extend to other binary systems where NPs with different atomic sizes and differing surface energies are thermally interacted.

Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the US Department of Energy's National Nuclear Security Administration under Contract No. DE-AC04-94AL85000.