Automated Materials Discovery Platform Realized: Scanning Probe Microscopy of Combinatorial Libraries (2412.18067v1)

Abstract: Combinatorial libraries are a powerful approach for exploring the evolution of physical properties across binary and ternary cross-sections in multicomponent phase diagrams. Although the synthesis of these libraries has been developed since the 1960s and expedited with advanced laboratory automation, the broader application of combinatorial libraries relies on fast, reliable measurements of concentration-dependent structures and functionalities. Scanning Probe Microscopies (SPM), including piezoresponse force microscopy (PFM), offer significant potential for quantitative, functionally relevant combi-library readouts. Here we demonstrate the implementation of fully automated SPM to explore the evolution of ferroelectric properties in combinatorial libraries, focusing on Sm-doped BiFeO3 and ZnxMg1-xO systems. We also present and compare Gaussian Process-based Bayesian Optimization models for fully automated exploration, emphasizing local reproducibility (effective noise) as an essential factor in optimal experiment workflows. Automated SPM, when coupled with upstream synthesis controls, plays a pivotal role in bridging materials synthesis and characterization.

• The paper introduces an innovative automated SPM system that leverages Bayesian optimization to navigate combinatorial materials libraries.
• It validates the approach by revealing non-trivial hysteresis patterns in Sm-doped BiFeO3 and characterizing key metrics in ZnxMg1-xO systems.
• The study demonstrates reduced manual intervention and enhanced reproducibility, paving the way for high-throughput materials characterization.

Automated Materials Discovery Platform Realized: Scanning Probe Microscopy of Combinatorial Libraries

The presented paper introduces an automated materials discovery platform employing scanning probe microscopy (SPM) to investigate combinatorial libraries. The focus is primarily on the use of piezoresponse force microscopy (PFM) within this framework, showcasing the evolution of ferroelectric properties in Sm-doped BiFeO3 (SmBFO) and ZnxMg1-xO (ZMO) systems. The paper emphasizes the importance of automation in connecting the synthesis and characterization paths in materials science.

Key Contributions

The research addresses significant challenges associated with the characterization of combinatorial libraries. The core innovation lies in the development and deployment of an automated SPM system, which facilitates the exploration of combinatorial material spaces. This system integrates two key concepts: 1) the use of Gaussian Process-based Bayesian Optimization models for automated explorative tasks, and 2) the validation of this framework across multicomponent systems with different material classes.

The paper presents several distinctive results:

1. Sm-BFO Library Exploration: The paper revealed a non-trivial trend in hysteresis loop areas across the Sm-doped BiFeO3 library, with a double peak observed at the morphotropic phase boundary. Such findings highlight the capacity of this automated setup to uncover intricate details in material behavior that might be overlooked with conventional grid-based approaches.
2. Optimization Framework: A comparative evaluation of vanilla Bayesian Optimization (vBO) and measured noise Bayesian Optimization (nBO) algorithms illustrated their respective strengths. Notably, nBO incorporates noise predictions from experimental variations, leading to improved decision-making in certain contexts, albeit with greater susceptibility to outlier influences.
3. ZMO Library Experiments: Automated exploration was expanded to ZnxMg1-xO, where both loop height and turn-on voltage were characterized. The framework confirmed consistency across the material library and effectively identified key functional metrics with minimal manual intervention.

Theoretical and Practical Implications

This research signifies an advancement towards fully automating the materials discovery process. Practically, the use of automated SPM in combinatorial studies expedites characterization, reduces human error, and offers reproducibility in analyzing complex material systems. The multifaceted analysis enabled by SPM functions, such as electromechanical property assessments via PFM, underscores the approach's adaptability.

Theoretically, the work challenges existing paradigms in the interpretation of noise and variance in experimental measurements. It suggests the incorporation of heteroscedastic noise models could yield more precise insights into materials' intrinsic properties, especially in systems characterized by spatial inhomogeneities.

Future Directions

The approach outlined broadens the prospects for efficiently exploring large compositional spaces, potentially catalyzing the development of new functional materials. Future research might explore other machine learning models and optimization algorithms tailored to cope with diverse noise characteristics and target functions.

Moreover, integration with other SPM techniques, such as conductive or magnetic probes, can enhance the platform’s applicability across multiple domains. The resulting high-throughout characterization capabilities will be critical for accelerating discoveries in energy materials, electronic devices, and other technologically relevant fields.

In summary, the paper provides a compelling demonstration of automated material exploration via advanced microscopy techniques, setting a foundation for subsequent innovations in materials science automation.