Reaction-Diffusion Driven Patterns in Immiscible Alloy Thin Films

Controlling the microstructure of thin films is of critical importance for various applications. We demonstrate a methodology for tuning the local microstructure through film-substrate interactions using Ag-Cu as a model system. Metastable single-phase Ag-Cu thin films are deposited on Si substrates pre-patterned by FIB milling. During post-deposition annealing, localized film-substrate reaction around the milled patterns produces a distinct microstructure termed as the ‘halo’. It consists of copper silicide and almost pure Ag, while the far-field film forms a random mixture of Cu and Ag-rich domains through phase separation. We show that the extent of the halo can be controlled by varying the temperature and duration of annealing. We present a semi-analytical kinetic model of product and halo growth that incorporates species balance, diffusional transport and a modified Stefan condition. Predictions from the model reveal two distinct growth regimes of the product with power law indices of 1/2 and 2/7 and experimental data fall into the latter regime. These regimes originate from the dimensionality of growth (2d or 3d) compared to that of solute transport (2d), which in turn depend on film thickness and species diffusivity. Using an inverse optimization procedure, we also estimate the diffusivity, which suggests grain boundary diffusion to be the dominant transport mechanism. This study provides an avenue and framework for microstructural engineering of alloy thin films through interfacial reaction.

💡 Research Summary

In this work the authors present a novel route to locally engineer the microstructure of immiscible Ag‑Cu alloy thin films by exploiting a controlled reaction with a silicon substrate. The experimental procedure consists of three main steps. First, Si(100) wafers are coated with an amorphous silicon‑nitride (SiNₓ) passivation layer and then patterned with focused ion beam (FIB) milling to create circular apertures of 400–1000 nm diameter that expose bare Si. Second, a metastable, single‑phase Ag‑Cu alloy film (≈50 nm thick, nominal composition Ag₀.₅Cu₀.₅) is deposited by co‑sputtering onto the pre‑patterned substrate. As‑deposited films are confirmed by X‑ray diffraction and transmission electron microscopy to be a homogeneous FCC solid solution with nanometer‑scale grains. Third, the samples are annealed under high vacuum (≤10⁻⁵ mbar) at temperatures ranging from 150 °C to 400 °C for times between 10 min and 2 h.

During annealing, the regions where the film directly contacts Si undergo a chemical reaction: Cu diffuses into Si and forms copper silicide, predominantly Cu₃Si, as verified by HAADF‑STEM, EDS mapping, and electron diffraction. The silicide appears as a bright central particle in back‑scattered electron (BSE) images. Surrounding this particle a distinct “halo” region forms, composed of an almost pure Ag matrix with a gradual compositional gradient that returns to the bulk Ag‑Cu phase farther away. In contrast, areas covered by SiNₓ do not react with the substrate; instead the alloy phase‑separates into Cu‑rich and Ag‑rich domains, creating a random two‑dimensional microstructure.

The authors quantify the halo width by measuring the radial distance from the silicide edge to the point where the BSE intensity reaches the bulk level. Systematic variation of annealing temperature and time shows that the halo expands non‑linearly: higher temperatures and longer times produce wider halos. To rationalize this behavior, a semi‑analytical reaction‑diffusion model is developed. Assuming axial symmetry, the model couples mass conservation at the moving reaction front with Fickian diffusion of Cu (the rate‑limiting species) in the film. A modified Stefan condition relates the front velocity to the concentration gradient and the concentration jump across the interface. Solving the governing equations yields two distinct growth regimes. When the film thickness is large enough for the reaction front to advance in three dimensions, the halo radius follows a power law R ∝ t¹ᐟ². When the film is thin and diffusion is effectively two‑dimensional, the model predicts R ∝ t²⁄⁷. Experimental data fall squarely in the latter regime, indicating that the growth is limited by lateral diffusion within the thin film.

An inverse optimization procedure is employed to extract the effective diffusion coefficient of Cu in the Ag‑Cu matrix from the measured halo evolution. The fitted values (≈10⁻¹⁴–10⁻¹³ cm² s⁻¹ at 400 °C) are orders of magnitude larger than bulk lattice diffusion, pointing to grain‑boundary diffusion as the dominant transport mechanism. Arrhenius analysis of the temperature dependence yields an activation energy of ~0.8 eV, consistent with literature values for Cu diffusion along grain boundaries in Ag‑Cu alloys.

The study demonstrates three key contributions: (1) a practical method to create spatially localized reaction zones by FIB patterning, enabling “on‑demand” microstructural features such as silicide islands and Ag halos; (2) a clear theoretical framework that links the dimensionality of the reaction front to observable growth exponents, providing predictive capability for designing annealing schedules; and (3) a quantitative inverse‑modeling approach that turns microstructural measurements into estimates of otherwise inaccessible material parameters like diffusivity. These insights open pathways for engineering functional thin‑film devices—such as interconnects, thermoelectrics, or catalytic layers—where precise placement of distinct phases is required. The methodology is readily extensible to other immiscible alloy systems and to more complex substrate chemistries, suggesting a broad impact on microstructural engineering of nanoscale materials.