ICMCTF 2025 Session CM3-1-ThM: Accelerated Thin Film Development: High-throughput Synthesis, Automated Characterization and Data Analysis I

Increasing demand for data and the surge of AI technologies is escalating the needs of telecommunication devices. RF filters are a limiting factor in this regard, where improvements in bandwidth and selectivity are needed. Advanced RF filters rely on piezo thin films like AlN, valued for its linear response and strong electromechanical coupling, with Sc doping enhancing its d₃₃ and overall performance. While AlScN is considered state-of-the-art, many proposed dopants (e.g. B, Y, La) could further improve its properties. High-throughput experiments are instrumental in exploring quaternary or multinary materials, but few studies examine combinatorial screening of piezoelectric materials. This is because piezoelectric materials are difficult to screen effectively, as their device properties strongly depend on the c-axis texture of the film. During combinatorial gradient deposition the shallow deposition angles and static substrate can cause significant grain tilt towards the dominant source, convoluting composition and texture gradients. In high-power impulse magnetron sputtering (HiPIMS), additional energy can be introduced to arriving species by synchronizing a substrate bias to arriving metal ions. This improves the adatom mobility of species, removing grain tilt, and resulting in highly textured films without substrate rotation. In this work, we use AlScYN as an example material to demonstrate how HiPIMS enables effective device screening of piezoelectric properties.

Before the use of HiPIMS, DCMS screening is used to assess solubility limits in the quaternary phase space. This is typically done using X-ray diffraction mapping while applying the disappearing phase method. However, in combinatorial screening of materials this method loses effectiveness, as precipitates can be tilted out of the diffraction plane. Consequently, we use the peak shift of the (0001) plane to track discontinuities from Vegard's law and identify precipitation. Despite the larger ionic radius of Y, the combined solubility of Y and Sc increases, reaching a maximum of ~50%. For context, the solubility limit of AlScN is ~40%. Materials libraries were then made using metal-ion synchronized HiPIMS. The libraries are highly textured, with rocking curve FWHMs of <2°. Following this, the coupling (k) and clamped d₃₃ coefficients of these libraries were mapped, showing their dependence on the combined Sc and Y contents. Importantly, d₃₃ coefficients were mapped using a double beam laser interferometer (DBLI) for improved accuracy.

The discovery of a new structural class of semiconductor is a rare occurrence. For example, in the case of solar absorption, nearly all relevant semiconductors can broadly be described as materials derived from the tetrahedrally-coordinated diamond structure (e.g., Si, III-Vs, II-VIs, chalcopyrites, kesterites). This is part of the reason that new high-performing materials, such as perovskites, which are made up of octahedral bonding motifs, garner so much interest and help generate new materials design concepts. Using high-throughput computational workflows, we recently discovered that several AM₂P₂ (A = Ca, Ba, Sr and M = Cd, Zn) compounds possess the requisite intrinsic materials properties for high optoelectronic performance, including solar-spectrum matched band gaps, strong optical absorption, and benign intrinsic defects, leading to long photoexcited carrier lifetimes. This family of compounds, which exhibits a mixed octahedral + tetrahedral bonding motif, has been known for several decades but the optoelectronic properties had been almost entirely unexplored.

Using a combinatorial synthesis approach based on a hybrid PVD/CVD method, we recently prepared the first thin films of Zintl phosphides CaZn₂P₂ and SrZn₂P₂. By combinatorial sputtering from metallic targets in the presence of PH₃ at low temperature, we prepare films across the ternary composition space. Growths at higher temperatures result in much narrower compositional spreads pinned around the AM₂P₂ composition, indicating an adsorption-controlled growth regime can be realized. Mapping measurements including x-ray fluorescence, x-ray diffraction, UV-vis spectroscopy, photoluminescence, and Raman spectroscopy are used to probe the properties of the zintl phosphide films. To establish the photoactivity and semiconducting nature of the AM₂P₂materials, minority carrier lifetimes and electronic properties were measured via time resolved microwave conductivity, transient absorption, and van der Pauw/Hall effect. To summarize the characterization of CaZn₂P₂ as an example, we observe high optical absorption of ~10⁴ cm^-1 at the ~1.95 eV direct transition, near band edge optical emission, and a photoexcited carrier lifetime of up to 30 ns at a fluence of 2 10¹³ cm^-2. Films are intrinsic, but p-dopable with +1 elements such as Na (A site dopant) or Cu (M site dopant). Such performance metrics are usually not observed in inorganic solar absorber materials until well into their development, highlighting the value of coupling high-throughput theory, high-throughput experiments, and targeted experiments toward new functional materials.

Combinatorial and high-throughput (CHT) methods offer an accelerated pathway for the discovery and development of novel materials with wide ranging biological, electronic, and structural applications. One common approach to accelerate synthesis is the deposition of large compositionally graded thin film arrays with hundreds of distinct samples, often referred to as thin film material libraries. High-throughput characterization techniques are then employed to quickly assess processing-structure-property relationships, which generates large datasets for machine learning models and screens for promising next-generation materials. For assessing mechanical behavior in these libraries, nanoindentation is particularly suitable due to the ease of automation, minimal sample preparation requirements, and compatibility with thin films. However, despite the widespread use of this technique in CHT research, many inconsistencies between reported methodologies in literature can be identified. This work presents a CuNi alloy library to identify how to improve data reliability while minimizing experimental times and costs. Emphasis is given to optimizing the number of indents per sample and the distribution of samples tested. By improving method standardization, both efficiency and reproducibility of combinatorial studies can be enhanced, thus expanding the value of material libraries to the scientific community.

Thin films, or micro- and nano-scale materials with unique functional properties, enable many clean energy industries, such as lithium-ion batteries, solar cells, and carbon conversion. Minimal-waste manufacturing of high-performance thin films will be vital for achieving society’s net-zero carbon emission goals. Because existing thin film sensors cannot operate within most sections of manufacturing lines, manufacturers often complete hundreds of processing steps before they can test (a small portion of) their products and detect problems. Current thin-film manufacturing thus often yields many low-quality products, or off-spec products that must be discarded.

SirenOpt has developed a real-time low-temperature plasma-based thin film metrology sensor. The sensor is compact, can be integrated within existing manufacturing lines, and can collect multiple thin film property measurements (such as thickness with a precision of 1/10th of a nanometer, density, resistivity, chemical composition, and contaminate identification) in parallel and in real-time, thus allowing for the collection of critical manufacturing data that is otherwise unobtainable. SirenOpt is currently applying the sensor to industrial thin film coating processes such as Lithium-ion battery electrodes, where this flexible sensor can be used to accelerate R&D and process optimization, improve quality control, and enable real-time process control.

In combinatorial materials science, magnetron sputtering plays a key role for the exploration of future high-performance materials due to its capability to produce well-defined, continuous compositional gradients in form of thin-film libraries. Its scalability from laboratory settings to industrial applications, relatively high deposition rates, and compatibility with a wide range of materials make it an effective choice for combinatorial synthesis [1]. However, achieving precise control over the deposition profile and compositional distribution often requires multiple preliminary experiments to optimize process parameters - an approach that can be time- and resource-intensive.

Aiming to predict those properties, several analytical and numerical simulations were reported in literature over the past decade [2-4]. However, only the magnetron sputter deposition model SIMTRA [4] was made publicly available. Based on the Monte Carlo approach, it allows to simulate the deposition profile of a single magnetron source, while taking into account the dimensions of the components through a graphical user interface.

In order to make this tool more suitable for the application in combinatorial materials science, the command line version of the SIMTRA application was wrapped in a Python environment, enabling the definition of sputter chambers through code and executing the time-consuming Monte Carlo calculations through user defined scripts. This approach also enables parallel simulation of multiple magnetrons by using multi-threading, decreasing simulation times significantly, especially when simulating co-sputtering systems with 5-8 cathodes. The accuracy of SIMTRA and the capabilities of the Python wrapper are demonstrated by comparing the compositions predicted by simulation and measured by energy-dispersive X-ray spectroscopy of seven materials libraries in the system Ni-Pd-Pt-Ru.

References:

[1] Gregoire, J. M., Zhou, L., and Haber, J. A. (2023). ‘Combinatorial synthesis for AI-driven materials discovery’. Nature Synthesis, vol. 2, no. 6.

[2] Ekpe, S. D., Bezuidenhout, L. W. and Dew, S. K. (2004) ‘Deposition rate model for magnetron sputtered particles’, Thin Solid Films, vol. 474, no. 1.

[3] Bunn, J. K., Metting, C. J. and Hattrick-Simpers, J. (2014) ‘A semi-empirical model for titled-gun planar magnetron sputtering accounting for chimney shadowing’, The Journal of The Minerals, Metals & Materials Society, vol. 67, no. 1.

Transition metal nitride semiconductors are rapidly emerging as a promising class of materials for advanced optoelectronic and energy conversion applications. Compared to oxides, nitrides offer narrower bandgaps, stronger bond covalency, and improved carrier transport properties that make them well suited for harvesting sunlight in photovoltaic and photoelectrochemical systems. Despite this considerable promise, far fewer nitrides than oxides have been experimentally investigated due to their synthetic complexity and a broad range of new compounds remain to be explored. Furthermore, synthesis challenges have led to poorly controlled defect and impurity properties within this class of materials. In this work, we overcome these limitations using reactive co-sputtering to synthesize thin film nitride semiconductors with controlled compositions, exploring both dopants and new compounds in the Ti-Ta-N, Zr-Ta-N, and Hf-Ta-N composition spaces. Starting with orthorhombic Ta₃N₅, which stands as the best performing photoanode material within this class, we investigate the critical roles of native and impurity defects on carrier transport and recombination, showing that substitutional Ti and Zr doping with rationally optimized concentrations can be used to improve photoconversion efficiencies. While high Ti contents in Ta₃N₅ lead to the precipitation of a secondary TiN phase, different behavior is observed for the case of Hf and Zr. In particular, solid solutions with broadly tunable compositions across the Hf-Ta-N-(O) and Zr-Ta-N-(O) composition spaces are investigated, leading to bandgap-tunable compounds that exhibit remarkably large refractive indices suitable for photonics applications. Moreover, deposition of a stoichiometric 1:1 Zr:Ta ratio leads to formation of a new ternary nitride compound, bixbyite-type ZrTaN₃, that it is a strong visible light absorber, functioning as an active photoanode material. Complementary DFT calculations indicate a direct bandgap that is tunable based on cation site occupancy. Thus, this material offers exciting prospects not only for solar energy conversion but also for optoelectronics applications. Overall, these results highlight the promise of both established and new transition metal nitride semiconductors for solar energy harvesting, as well as the importance of precise composition engineering to tune optoelectronic and charge transport characteristics. Considering the compositional complexities of these compounds, exploration and optimization can be dramatically accelerated through use of gradient sputtering and rapid characterization approaches.