Effects of the Core Location on the Structural Stability of Ni-Au Core-Shell Nanoparticles.

Structural changes of Ni-Au core-shell nanoparticles with increasing temperature are studied at atomic resolution. The bimetallic clusters, synthesized in superfluid helium droplets, show a centralized Ni core, which is an intrinsic feature of the growth process inside helium. After deposition on SiN x , the nanoparticles undergo a programmed temperature treatment in vacuum combined with an in situ transmission electron microscopy study of structural changes. We observe not only full alloying far below the actual melting temperature, but also a significantly higher stability of core-shell structures with decentralized Ni cores. Explanations are provided by large-scale molecular dynamics simulations on model structures consisting of up to 3000 metal atoms. Two entirely different diffusion processes can be identified for both types of core-shell structures, strikingly illustrating how localized, atomic features can still dictate the overall behavior of a nanometer-sized particle.

Introduction

Bimetallic systems at the nanoscale level have recently received increased attention as the combination of intermetallic interactions and surface size effects can trigger unexpected physical behavior and new phenomena (see, e.g., ref (1) for a recent review). Potential applications cover a wide range of different fields, including biomedical applications,[2−6] optics,[7−10] heterogeneous catalysis,[11−14] electrochemistry,[15] and electronics.[15,16] Additionally, magnetic core nanoparticles have been suggested for the activation of stem cells,[17,18] as enhancers of supercapacitors[19] or for the optomagnetic fine-tuning of semiconductors.[20]

Bimetallic nanoparticles which combine magnetic and noble metals are particularly interesting for special physicochemical applications.[21] Our research on the Ni–Au system is partly motivated by the fact that optical, catalytic and photocatalytic properties of gold nanoparticles can be tuned via the insertion of a magnetic Ni core, for example, for a magnetic-field-induced synthesis of wire-like structures.[22] An additional handle on positional control could also become relevant for the 3D-printing of nanostructures.[23] Magnetic nanoparticles and external fields have further been used to target biomolecules, and multifunctional plasmonic-shell magnetic-core nanoparticles are being successfully applied for diagnosis, isolation, and photothermal destruction of cancer cells.[5] Sensitive sensing devices in human serum have been built using Ni–Au core–shell nanoparticles.[6] However, the specific activity of a core–shell nanoparticle is related to the actual position of its core, and its structural integrity is compromised at elevated temperatures due to the onset of diffusion/alloying processes. Therefore, it is necessary to understand the correlation between the magnetic core location and the structural stability of the particle upon heating.

For the synthesis of nanoparticles a variety of methods are currently being exploited. Among them, wet chemistry methods are the most common, but also techniques such as laser ablation synthesis[24] are applied in some cases. An interesting alternative to these standard methods is the usage of superfluid helium nanodroplets as containers for a fully inert, controlled synthesis of mixed-metallic nanoparticles. This is achieved via a sequential pickup of metallic atoms by helium droplets created in the process of a supersonic expansion of helium through a cooled nozzle. Because of the extremely low helium temperature of 0.37 K[25] and the practically inert environment provided by the droplets it is possible to synthesize metallic nanoparticles with arbitrary core–shell or onion-type structures. The low temperature of the droplets even enables the synthesis of particles in metastable structures, as metallic systems can be trapped in the desired local minimum during growth. Recently, we took advantage of this feature and produced Ni–Au core–shell particles where the highly reactive Ni core is protected by a few layers of gold.[26] While this previous work had its focus set on the avoidance of Ni-oxidation by a protective layer of gold, the current study investigates the structural behavior of a bimetallic system at the nanoscale in greater detail.

We demonstrate that it is indeed possible to synthesize bimetallic nanoparticles in a local minimum structure, deposit them on a substrate, and observe their relaxation to the global minimum structure in situ. In terms of experimental effort, this objective is rather challenging: any presence of oxygen during heating after deposition must be avoided at all cost as it would lead to an inevitable chemical reaction of diffusing Ni atoms with oxygen on the particle surface.[26] We show that for Ni–Au core–shell particles in the nanometer range a fully centered position of the Ni core within a gold shell, a structure intrinsic to the pickup-mediated growth process inside the helium nanodroplets, experiences a faster transition toward a fully alloyed or mixed state than their decentralized counterparts. We also demonstrate that in contrast to surface studies of island growth at room temperature,[27] the nanoparticles remain intact until the alloying temperature is reached.

Concerning the theoretical treatment of structural changes in nanoparticles, two diametrically opposed approaches can be distinguished. One is to employ top-down models such as the CALPHAD method,[28] which extends to well-known bulk models by additional surface terms in order to handle the increasingly important surface size effects emerging at small scales. Very recently, a bond-centric model has been developed specifically for the description of monometallic and bimetallic nanoparticles, addressing the energetics in massive nanoalloy structures.[29] The alternative is bottom-up approaches, which aim for a basic understanding at the atomic level via molecular dynamics (MD) simulations.[30−34] In this article, we employ a force field-based bottom-up approach in order to interpret and understand the results of our in situ heating experiments.

Materials and Methods

Nanoparticle Synthesis

Superfluid He nanodroplets serving as “nanolabs” are used to synthesize the desired Ni–Au particles via sequential pickup of different metal atom species. For a more detailed description we refer to the previous publications.[35,36]

High purity He (99.9999%) at a pressure of 20 bar is expanded through a 5 μm nozzle at 7 K in vacuum. This expansion process results in the formation of superfluid He droplets containing an average number of 107 He atoms.[37] These droplets are then sequentially doped when passing through two resistively heated pickup cells, containing Ni and Au. By varying the individual vapor pressures, the amount of doped metal atoms can be adjusted. This way, the droplet first gets doped with Ni, resulting in the formation of a spherical Ni cluster, which is then coated by layers of gold atoms while passing through the second pickup cell. Using this technique, it is possible to synthesize any desired core–shell combination in the core–shell configuration requested. The cluster sizes are log-normally distributed with an average diameter of d = 6.2 nm and a log-normal standard deviation of 1.3 nm. A Ni/Au ratio of 3:7 has been chosen for the experiment in order to obtain perfect conditions for a complete coating in the particle size regime between 2 and 10 nm. More than 90% of the synthesized structures show the desired Ni–Au core–shell structure. The rest is formed by Au clusters, most likely with traces of Ni below our detection efficiency. Note that our method of production in helium does not emphasize the appearance of “magical” numbers, as there is no selective mechanism with regard to stability involved. The release of binding and kinetic energy from the metal atoms throughout the cluster formation process results in the partial evaporation of the helium droplet. This reduction in the droplet size is monitored via a residual gas analyzer (Balzer QMA 200/QME 200). Finally, the beam is terminated on a heatable transmission electron microscopy (TEM) grid (DENSsolutions Nano-Chip XT SiN), where the clusters are deposited by a soft-landing process in which the excess helium is vaporized.[38−41]

Electron Microscopy Characterization

A probe corrected FEI Titan3 G2 60-300 is used to record high angle annular dark-field images of temperature-induced restructuring processes. Elemental analysis was carried out with a four-quadrant energy-dispersive X-ray spectroscopy detector and a Gatan Quantum energy filter for electron energy loss spectroscopy.

Computational Details

All MD simulations have been performed using the large-scale atomic/molecular massively parallel simulation (LAMMPS) package.[42] Newton’s equations of motion have been solved through numeric integration via the velocity Verlet algorithm,[43] with time steps ranging from 1.0 to 3.0 fs.

A potential based on the embedded atom method (EAM) reported by Zhou et al.[44] for Ni–Au alloys has been used to describe all pair interactions. The EAM potential includes contributions for the embedding of each metal atom in the surrounding metal atoms. Therefore, it is expected to be more sensitive to surface nanoparticle effects than other additive pairwise interatomic potentials.

The dynamic range of solid Au and Ni indicates that a time step within the range of 3–5 fs would be acceptable. However, it is recommended that a smaller time step of about 1 fs is used instead when surface effects and initial instabilities are expected to appear, in particular at the initial steps of the simulation. For comparison, very recent state-of-the-art MD simulations on nickel and gold nanoparticles used a time step of 2 fs.[45,46]

Results and Discussion

In ref (26) we discussed the interplay between oxygen presence and enhanced, temperature-induced, atomistic diffusion of Ni–Au core–shell clusters. We could demonstrate that even small traces of oxygen can completely dominate the outcome of the experiment, resulting in the core–shell inversion in Ni–Au to Au–NiO. As a consequence, studies of the unperturbed Ni–Au phase diagram at the nanoscale require an entirely oxygen-free atmosphere throughout all measurements. We start with a description of the experimental observations under these conditions in our attempt to cover the alloying, spinodal decomposition (SD) and demixing effects at high temperatures. In the second part, we use the LAMMPS suite of programs for MD simulations of the alloying process in a solid phase by looking at mixing rates and as a function of initial composition at elevated temperatures.

Experimental Observations

An overview of all reversible and irreversible processes covered in this article is provided in Figure , together with the corresponding scanning TEM (STEM) images taken at various temperatures. For the sake of completeness, we have also added results from our previous studies on Ni–Au oxidation ref (26) to further illustrate the impact of oxygen.

Figure 1

Temperature-induced structural transitions in Ni–Au core–shell nanoparticles with diameters in a nanometer range in a symbolic representation (left) and as visualized by STEM imaging (right). The temperatures T0, T1 correspond to 25 and 400 °C and t1 and t2 correspond to 10 min and 12 h, respectively. Note that even in clusters with a decentralized arrangement, the Ni core remains protected by a layer of Au.

Starting from centralized core–shell structures (CC) in the upper left corner, synthesized as described in Section , several pathways are possible regarding structural conversion. Within 4 s the temperature of the grid was increased from room temperature to 400 °C. At this temperature, the system forms an alloy (All) within the scanning time of approximately 1 min. This alloyed structure remains alloyed and no traces of oxidation are found over the following observation time of 4 h. To prevent Ni-oxidation, we use SiN grids as an alternative to the widely used amorphous carbon supports in order to reduce the amount of molecular oxygen adsorption and operate with steep heating ramps to quickly reach temperatures where the adsorption of oxygen from the microscope vacuum is decreased. However, if a sufficiently large electron beam dose is applied to the cluster, a spontaneous demixing occurs due to a selective enhancement of Au atomic mobility inside the cluster. The necessary dosage is inversely proportional to the actual cluster temperature; a separate study has been dedicated to analyze this effect.[47] At a temperature of 400°C an electron dose of 2 × 1010 electrons is required to trigger such a phase separation which leads to a decentralized Ni–Au core–shell cluster (DC). In the latter, the Ni core is typically residing in a sub-surface position, still embedded underneath a protective layer of Au atoms. This configuration seems to be a more stable minimum energy structure of the NiAu system, as neither continued heating up to melting temperature nor a cooling down of the particles to room temperature again is affecting the decentralized core–shell structure within the observation times of about 1 h. Note that all of these structural changes appear in the solid phase, as can be derived from the clearly visible lattice structures in the STEM images in Figure .

A different pathway is observed if an alloyed structure (All), obtained by heating of a centralized core–shell particle to 400 °C, is allowed to cool down to room temperature directly. In this case, SD is observed, which leads to the aggregation of small Ni particles inside the Au cluster. This interesting structure also seems to represent a local minimum of the total energy at room temperature, which is analogous to the centralized Ni core structure, and is also unstable upon heating as it turns into a fully alloyed particle at 400 °C.

Transitions between three completely different solid phase structures can be observed for the NiAu clusters under UHV conditions (5 × 10–10 mbar). In order to provide representative statistics, we investigate the phase transitions of more than 100 randomly chosen clusters with either a centralized core–shell arrangement or a spinodally decomposed structure. Long-time observations are performed for 28 decentralized clusters. The clusters are log-normally distributed with an average diameter of d = 6.2 ± 1.3 nm. The following subsections are dedicated to the temperature behavior of each corresponding structural motive.

Centralized Core–Shell Clusters

Centralized Ni cores are the standard configuration in which the clusters are synthesized by the sequential pickup technique, where exposure to Ni vapor is followed by exposure to Au vapor. The synthesis conditions (droplet size, and vapor pressure of metals) are set to single-center growth; see Section for further details. The Ni core is fully encapsulated in a shell of Au, which would prevent it from oxidation even if exposed to ambient air. The structural rearrangements of the deposited clusters at elevated temperature are illustrated in Figure . Upon increasing the temperature, the mobility also increases, resulting in the intermixing of the two metal species at a temperature of around 400°C. We note that in this temperature range the danger of undesired oxidation is largest due to the thermally enhanced mobility of the Ni atoms and the inevitable traces of oxygen produced during electron microscopy scans. In order to minimize the risk of Ni-oxidation, we use SiN chips as supports, which reduce the hydrocarbon background drastically in comparison to substrates based on amorphous carbon. Additionally, we employ a very steep heating ramp (see the Supporting Information) from room temperature to 400°C in 4 s in order to minimize the exposure time of the clusters to any oxygen present in the microscope vacuum (10–8 mbar) or adsorbed at the grid. After alloy formation, the cluster remains in this mixed, solid structure until it undergoes a first order phase transition into the liquid state.

Figure 2

Thermal evolution of an initially centralized core–shell cluster. (a) room temperature, (b) 400°C, and (c) 700°C.

Spinodally Decomposed Clusters

SD becomes visible when performing an in-chamber heating to temperatures above 300°C, keeping this temperature constant for about 30 min, followed by a final cooling of the sample back to room temperature. Reduced mobility in this final step typically results in the formation of several smaller Ni agglomerates embedded in a Au matrix (see Figure ). It can be assumed that the system gets trapped in local minimum structures with energies above that of the decentralized Ni core. However, the experiment does not answer whether the multi-centered structure is lower or higher in energy than a centralized Ni core. Upon increasing the temperature, spinodally decomposed NiAu structures undergo the same pathway as the initially centralized core–shell clusters by forming an alloy at about 400°C.

Figure 3

Evolution of spinodally decomposed clusters at elevated temperatures. Similar to centralized clusters, they form an alloy at 400°C.

Decentralized Core–Shell Clusters

Our experiments indicate that a decentralized position of the Ni core corresponds to a more thermally stable structure for nanometer-sized NiAu clusters in the solid phase. A closer investigation of the typical Ni core positions and shapes in this “final” configuration suggests that a single agglomeration of Ni directly under a facet of the Au shell is the most feasible structural motive. While a centralized core structure becomes mixed within a minute at 400 °C, a decentralized cluster is observed to be stable for several hours, exceeding the standard measurement time. In order to determine whether entropy or inner energy dominates the final configuration, we synthesized a total number of 28 decentralized clusters “on the fly” by electron beam exposure at 700°C, where the intrinsic electron dosage applied by STEM is sufficient to induce the desired structural changes almost immediately, and keep them at 450°C for 12 h without further manipulation. After that, a final STEM examination is performed. This final investigation confirms that also decentralized clusters are metastable as they end up in a mixed state with entropy at its maximum. From this outcome we conclude that, independent of the initial configuration, alloying takes place at temperatures above 300°C. However, the time required for this transition to take place is strongly dependent on the initial configuration of the bimetallic cluster. As shown in Figure , a random group of three initially decentralized core–shell clusters is kept at 450°C for 12 h.

Figure 4

Example transformations from decentralized core–shell structures to mixed clusters, visualized by a comparison of STEM images taken before and after 12 h at 450°C.

MD Simulations

To reproduce realistic experimental conditions we have performed large-scale atomistic simulations relying on the EAM[44] to describe inter-atomic interactions. As a reasonable representative of the experimental cluster size distribution (2–10 nm) and metal donation ratio (3:7), we build a core–shell model system from 900 Ni and 2100 Au atoms, which yields a nanoparticle with a diameter of about 5 nm. An even larger nanoparticle composed of 6266 atoms has been simulated as well in order to test for size dependencies.

Synthesis of Model Nanoparticles

In order to obtain nanoparticles with geometries as close as possible to the outcome of the assumed growth process inside the helium matrix,[48,49] we proceed in a two-step fashion: first, a Ni core of 900 atoms is allowed to fully relax. Then, a shell of Au atoms is added to the system and allowed to structurally relax while the Ni structure is kept frozen. This procedure takes into account that the actual synthesis inside the He droplet is considered to take place “atom by atom”. In other words, at the time when the droplet arrives the second pickup and starts to collect Au atoms, a Ni core in a minimum energy configuration can be assumed inside the He droplet.

In our “virtual” synthesis, 900 Ni atoms are randomly located inside a sphere with a radius of 14.0 Å and cooled down using a Berendsen thermostat at 0.001 K, allowing a maximum atomic displacement of 0.01 Å. The value of 14 Å is consistent with an estimate of the Ni core diameter d following the relationship d [nm] = 31.72–282.85 N–1/3 as suggested in ref (45), resulting in a value of about 24 Å for a structure containing 900 Ni atoms. Next, an annealing procedure is applied, where the Ni nanoparticle is heated up to 900 K and kept at this temperature for 5 ns. Then, the nanoparticle was cooled down again to 0.4 K, resulting in its crystallization. The addition of the 2100 Au atoms follows a similar approach. First, they are randomly localized in a spherical shell with an internal (external) radius of 14 Å (24 Å) around the Ni core. Second, the same heating and annealing procedure described is performed as above, allowing the Au2100 shell to crystallize as well, while the geometry of the Ni core is kept frozen.

Dynamics at Elevated Temperatures

Both the centralized and the decentralized Ni900Au2100 nanoparticles were heated to 300 K in 0.5 ns (i.e., applying a heating rate ΔT/Δt of 6 × 1010 K s–1). Using a Langevin thermostat, the temperature of 300 K was kept for 5 ns. Next, the nanoparticles were heated by increasing the temperature in steps of 50 K within 0.5 ns (i.e., with a heating rate of 5 × 109 K s–1). Each value of the temperature was then kept fixed for 5 ns before further heating, until a final value of 750 K was reached. The first signal of the diffusion of Ni atoms was observed at 550 K.

Next, the nanoparticle was heated up to 900 K in 0.5 ns, keeping this value of the temperature during 5 ns. Then, the time step of the simulation was set to 2 fs for 245.5 ns. Finally, the time step was increased to 3 fs, and the temperature of the nanoparticle was kept at 900 K for 186.5 ns. This procedure allowed us to test if the time step of 3 fs was still small enough to avoid the appearance of instabilities or abrupt changes in the behavior of the nanoparticle. After confirming that a time step of 3 fs can be applied, this value was used in the last steps of the simulation, allowing us to study long-time diffusion processes on timescales of several hundred ns.

Comparison between CC and DC Dynamics

At 900 K, the diffusion process is fast enough to be studied via MD simulations. To analyze this behavior, we found the average Ni atom radial distance from the center-of-mass to be the most suitable parameter for modelling. The simulations were carried out for both the model particles, consisting of 3000 and 6266 atoms, respectively. Both the cluster models show the same interesting behavior: while a decentralized Ni cluster has a higher overall diffusion, a centralized cluster exhibits a larger radial diffusion toward a fully intermixed state. As can be seen in Figure (see also Figure S10 of the Supporting Information), the diffusion of the decentralized Ni cores is mainly limited to subsurface positions, which leaves the remaining Au cluster intact, resulting in a slower convergence toward a fully intermixed bimetallic cluster. In contrast, for the centralized Ni cores, the nickel atoms diffuse radially through the gold shell toward the subsurface region in the centralized nanoparticle.

Figure 5

Simulation snapshots of a nanoparticle consisting of 6266 atoms, showing the different diffusion processes of Ni atoms through the Au shell observed in centralized (a) and decentralized (b) nanoparticles at 900 K, taken after 216 ns (see also Figure S10 of the Supporting Information).

To try to understand this remarkable behavior, we have calculated the average displacement distance Δd(t),as well as the radial displacement Δr(t)with r⃗Ni(t) denoting the position of Ni atom j at time t. It is found that the radial diffusion (see the right graph in Figure ) for a centralized cluster behaves exponentially, as expected. On average, the Ni atoms changed their radial position by 2 Å after 250 ns. For a decentralized cluster, it can be seen that the change in the radial distance after 250 ns hardly happens at all.

Figure 6

Average overall diffusion distance Δd (left) and radial diffusion distance Δr (right) of Ni atoms in a decentralized (red) and centralized (blue) cluster, plotted as a function of time.

In addition, we found that the diffusion of Ni atoms in decentralized clusters is mostly limited to the positions below the surface. This phenomenon is caused by the reduced lattice mismatch underneath the surface. The average change in absolute distance Δd after 250 ns for the decentralized cluster is 10 Å in comparison to 5.5 Å for a centralized core (see the left graph in Figure ). Furthermore, we observed that the energetic “cost” of placing the Ni atoms at the surface is rather high in comparison with placing a Au atom at the surface. The energetic “gain” of exchanging one Ni atom from the surface to a subsurface position is about 0.2 eV higher than the energetic “gain” when transferring a Au atom from the surface to a subsurface position. This is analyzed in greater detail in the Supporting Information. A similar interplay between bulk, surface, and strain effects and their impact on the morphology of a bimetallic nanoparticle has recently been discussed for the closely related CuAg system.[30] In this theoretical study of Rahm and Erhart, the segregation of the two metals is strongly influenced by local lattice strain. Agglomerations of Cu are preferred at regions of high lattice strain in the Ag matrix. Depending on the shape of the nanoparticle, this can be either in the center of the metal cluster or at a subsurface position. Our findings regarding Ni diffusion in a Au matrix fit nicely to these predicted lattice strain tendencies.

Another scenario may be the SD of the alloyed system. The earliest findings in this direction have been revealed by Nelli and Ferrando in a very recent work.[50] However, note that our systems are about 10 times larger than those treated in ref (50). Due to the computational effort, which increases with the 2nd power of the system size at least, combined with the necessity of even smaller timesteps due to the lower temperature, a simulation of spinodal demixing is currently beyond our reach.

Conclusions

Irrespective of the future usage of bimetallic core–shell nanoparticles, for example, for medical, optical or chemical purposes, it is clear that the structural integrity of the layers, also at higher temperatures, is a knockout criterion for any planned industrial application.

Taking advantage of the He-mediated synthesis of mixed-metallic structures in combination with meticulously controlled electron beam dosage after particle deposition, we were able to prepare NiAu core–shell samples in different geometries for a follow-up study of their structural stability at higher temperatures. Comparing centralized Ni-cores, decentralized Ni-cores and spinodally decomposed Ni-cores, we found that these differences in the initial geometries have a tremendous impact on intermetallic diffusion behavior and therefore, affect the times for an inevitable alloying at higher temperatures. We were able to confirm that, starting at a temperature of 300°C, all the initially separated structures undergo a transition toward a fully alloyed state. The lower alloying temperature for Ni–Au nanoparticles in comparison to the bulk materials confirms the predictions made via the CALPHAD approach (see the Supporting Information).[28] However, an analysis of structural changes via TEM image reveals that clusters featuring a decentralized core possess a pronounced structural integrity, which we link to a different diffusion behavior. From computationally costly large-scale MD simulations over several hundred nanoseconds, we conclude that a decentralized core affects the overall structure of the particle in a way which promotes diffusion processes along the intermetallic interface but quenches atomic intermixing along the radial coordinate. In the concrete case, this subtle change in the diffusion mechanism leads to an extension of the alloying times by 2 orders of magnitude. Depending on the actual position of the Ni core, lattice strain appears to be distributed differently; while a central position seems to increase Ni mobility throughout the cluster, a decentralized core position, where the Ni core is located at a subsurface position, shows less pronounced diffusion tendencies which also remain mostly within the subsurface region of the cluster. This indicates that the inclusion of Ni atoms near the surface (but not at the surface) allows for an effective compensation of lattice strain which stems from the finiteness of the nanoparticle.