arXiv subjects

Felix Thelen

Publications and source records attributed to Felix Thelen.

4 recordsLinked to original sources

arXiv2026

Work Function and High-Coverage Adsorption Energy as Hydrogen-Evolution Descriptors on Ag-Au-Pd-Pt Alloys

Hydrogen-evolution activity is commonly rationalized through hydrogen adsorption energies and the Sabatier principle, yet this descriptor picture becomes ambiguous on multimetallic surfaces, where each composition exposes a distribution of local adsorption environments. Here we investigate whether the bare-surface work function, previously shown to add predictive information for monometallic surfaces, remains an activity descriptor for compositionally complex surfaces. We test this on three combinatorial Ag-Au-Pd-Pt thin-film materials libraries screened for acidic hydrogen evolution by scanning electrochemical cell microscopy. Graph neural networks provide adsorption-energy distributions and work functions for each measured composition. A work-function-only model explained most of the activity variation (mean $R^2_\mathrm{log}$ = 0.903), as did a coverage-corrected adsorption model (mean $R^2_\mathrm{log}$ = 0.955), outperforming dilute adsorption (mean $R^2_\mathrm{log}$ = 0.758). Combining work function and coverage-corrected adsorption yielded the highest fit quality (mean $R^2_\mathrm{log}$ = 0.969), but only a small gain over coverage-corrected adsorption alone. For these four metals the coverage-corrected adsorption energy and work function follow a similar trend, producing similar activity rankings, hence including both adds little beyond either one individually, although both are strong predictors.

cond-mat.mtrl-sci↗

arXiv2026

Autonomous scanning electrochemical cell microscopy enables rapid exploration of large compositionally complex material spaces

Alloying is a central strategy in electrocatalysis, enabling fine-tuning of electronic structure. In particular, compositionally complex solid solutions (CCSS) often called high-entropy alloys are of high interest as they allow active site design. However, the "combinatorial explosion" in the number of possible compositions poses a critical bottleneck for the discovery of active CCSS electrocatalysts. We present an autonomous scanning electrochemical cell microscopy (SECCM) system for ultrahigh-throughput and large-scale CCSS activity screening. The platform rapidly establishes composition-electrocatalytic activity relationships for large compositional spaces across multiple thin-film CCSS materials libraries via active learning and automated library exchange. Embedding analytical expressions of voltammetry in the algorithm enables the learning of whole voltammograms rather than a single selected metric. As a demonstration, we investigated hydrogen evolution reaction (HER) activities of Au-Ir-Rh, where Ir and Rh exhibit strong metal-hydrogen binding and Au exhibits relatively weak binding as derived from the HER volcano plot. The composition-activity trend was accurately predicted after measuring only 15% of all 966 measurement areas. Au30Ir20Rh50 and Au10Ir35Rh55 exhibit highest activities with standard rate constants of about 0.012 cm/s, demonstrating positive synergistic contributions from elemental mixing. The autonomous robotic SECCM platform is broadly applicable to a wide range of electrocatalytic reactions, providing a general pathway for accelerating CCSS electrocatalyst discovery and optimization.

cond-mat.mtrl-sci↗

arXiv2024

A Python-Based Approach to Sputter Deposition Simulations in Combinatorial Materials Science

Magnetron sputtering is an essential technique in combinatorial materials science, enabling the efficient synthesis of thin-film materials libraries with continuous compositional gradients. For exploring multidimensional search spaces, minimizing preliminary experiments is essen-tial, as numerous materials libraries are required to adequately cover the space, making it crucial to fabricate only those libraries that are absolutely necessary. This can be achieved by Monte Carlo particle simulations to model the deposition profile, e.g. by SIMTRA, which is an established package mainly designed for single cathode simulations. A strong enhance-ment of its capabilities is the development of a Python-based wrapper, designed to simulate multi-cathode sputter processes through parallel Monte Carlo simulations. By modeling a sputter chamber and determining the relationship between deposition power and rate for an exemplary quaternary system Ni-Pd-Pt-Ru, we achieve a match between simulated and measured compositions, with a mean Euclidean distance of 3.5%. The object-oriented design of the package allows easy customization and enables the definition of complex sputter sys-tems. Due to parallelization, simulating multiple cathodes results in no additional simulation time. These additions extend the capabilities of SIMTRA making it applicable in combinatorial materials research.

cond-mat.mtrl-sci↗

arXiv2023

Speeding up high-throughput characterization of materials libraries by active learning: autonomous electrical resistance measurements

High-throughput experimentation enables efficient search space exploration for the discovery and optimization of new materials. However, large search spaces of, e.g., compositionally complex materials, require decreasing characterization times significantly. Here, an autonomous measurement algorithm was developed, which leverages active learning based on a Gaussian process model capable of iteratively scanning a materials library based on the highest uncertainty. The algorithm is applied to a four-point probe electrical resistance measurement device, frequently used to obtain indications for regions of interest in materials libraries. Ten materials libraries with different complexities of composition and property trends are analyzed to validate the model. By stopping the process before the entire library is characterized and predicting the remaining measurement areas, the measurement efficiency can be improved drastically. As robustness is essential for autonomous measurements, intrinsic outlier handling is built into the model and a dynamic stopping criterion based on the mean predicted covariance is proposed. A measurement time reduction of about 70-90% was observed while still ensuring an accuracy of above 90%.

cond-mat.mtrl-sci↗