Facile Optimization of Combinatorial Sputtering Processes with Arbitrary Numbers of Components for Targeted Compositions

Combinatorial sputtering is a physical vapor deposition method that enables the high-throughput synthesis of compositionally varied thin films. Using this technique, the effects of stoichiometry on specific properties of alloy thin films with analog composition gradients can be mapped using high-throughput characterization. To obtain specific stoichiometries, such as those desired for an equiatomic, intermetallic, or doped compounds, the sputter power of each target must be simultaneously tuned to optimize the deposition rate of each component. This optimization problem increases in complexity with the number of components, which commonly leads to iterative guess-and-check processing and can limit the intrinsic high-throughput advantages of this synthesis method. To circumvent this challenge, this work introduces a composition optimization procedure that enables the facile synthesis of sputtered combinatorial films with targeted compositions. This procedure leverages the expeditious mapping of composition using wavelength dispersive x-ray fluorescence and is capable of optimizing processing for an arbitrary number of components. As a demonstration, this method is leveraged to sputter a combinatorial Cr${v}$Fe${w}$Mo${x}$Nb${y}$Ta$_{z}$ film with an equiatomic composition near the wafer center.

💡 Research Summary

The paper addresses a long‑standing bottleneck in combinatorial magnetron sputtering: the difficulty of simultaneously tuning the sputter powers of multiple targets to achieve a desired overall composition, especially when the number of components exceeds two or three. Traditional approaches rely on iterative trial‑and‑error adjustments, which erode the intrinsic high‑throughput advantage of combinatorial methods. The authors propose a systematic, data‑driven optimization workflow that leverages rapid, non‑destructive wavelength‑dispersive X‑ray fluorescence (WDXRF) mapping to quantify the areal mass density of each element across a full‑size wafer.

First, they demonstrate that the spatial distribution of a single‑element film deposited by direct‑current sputtering can be accurately described by a two‑dimensional Gaussian function. By fitting WDXRF‑derived mass‑density maps to this model, they extract five parameters: amplitude (A), standard deviations (σx, σy), and the Gaussian centre coordinates (x0, y0). Repeating the single‑element depositions at several power densities (2.6–6.6 W cm⁻²) reveals that A scales linearly with power while σx, σy, x0, and y0 remain essentially constant for a given target geometry and chamber configuration. This linear relationship provides a simple “mass‑per‑energy” calibration factor for each element (e.g., 5.39 × 10⁻⁶ g cm² J⁻¹ for Mo, 3.11 × 10⁻⁶ g cm² J⁻¹ for Cr).

Next, the authors test whether the single‑element calibrations remain valid when two elements are co‑deposited. A Mo–Cr co‑deposition at identical power density (3.9 W cm⁻²) produces mass‑density maps whose Gaussian parameters differ by less than 3 % from the single‑element fits, and the amplitudes agree within ±0.5 %. Consequently, the co‑deposition can be accurately predicted by simply superimposing the calibrated single‑element Gaussian profiles, confirming that inter‑target interactions are negligible under the studied conditions.

With these calibrations in hand, the workflow proceeds as follows: (1) perform a minimal set of single‑element depositions to determine the Gaussian shape parameters and linear mass‑per‑power coefficients; (2) solve a set of linear equations that relate the desired atomic fractions to the required power densities for each target; (3) apply the calculated powers in a simultaneous multi‑target sputter run; (4) map the resulting composition with WDXRF; and (5) optionally iterate a single correction step if the measured composition deviates beyond tolerance. All data processing is automated using the Python lmfit library, enabling rapid turnaround (a full‑wafer WDXRF map takes ~2 h).

The authors validate the method by fabricating a five‑component Cr‑Fe‑Mo‑Nb‑Ta alloy on a 6‑inch Si wafer, targeting an equiatomic composition (≈20 % each). The WDXRF map shows that the central region of the wafer attains the target composition within a few percent, and energy‑dispersive X‑ray spectroscopy (EDS) measurements on selected spots confirm agreement to within 2 % of the intended stoichiometry.

Key insights include: (i) the Gaussian description of sputter deposition profiles is robust across a wide power range; (ii) the shape parameters are largely independent of power, allowing reuse for any target combination; (iii) linear mass‑per‑power scaling eliminates the need for complex, non‑linear models; (iv) WDXRF provides a fast, quantitative compositional map suitable for feedback control; and (v) the workflow scales trivially to arbitrary numbers of components, making it a universal tool for high‑throughput thin‑film libraries.

In summary, the paper delivers a practical, mathematically grounded, and experimentally verified protocol that transforms compositional optimization in combinatorial sputtering from a labor‑intensive guessing game into a predictable, automated process. This advancement promises to accelerate the discovery of multi‑principal‑element alloys, high‑entropy materials, and compositionally complex functional thin films across catalysis, energy conversion, and electronic applications.