Differentiable Characterization of Materials is a software-based approach for analyzing materials microscopy data by converting discrete measurement data into continuous, queryable representations of a material. Instead of limiting analysis to fixed measurement points, this approach allows researchers to study material features across space and extract useful property information from microscopy signals. The technology may help researchers better understand composition, crystal orientation, stress, defects, boundaries, and other material features with greater flexibility than traditional grid-based analysis. It can support higher-resolution review of existing datasets, reduce storage needs by representing large datasets more compactly, and connect characterization data more directly with property prediction and physics-based materials models. This could make materials analysis faster, more informative, and easier to integrate into existing research and commercial microscopy workflows. Background: Materials researchers rely on microscopy and microanalysis tools to understand how a material’s structure affects its properties and performance. Current methods often collect data at fixed points or on fixed grids, which can limit how closely researchers can inspect areas between measurement points. These methods can also introduce noise when estimating gradients, boundaries, or other spatial changes in a material. Existing workflows often require separate steps for data collection, data processing, image segmentation, property extraction, and model-based analysis. This can make the process time-consuming and can cause information to be lost between steps. This technology creates a continuous representation of material data, allowing users to query, analyze, compress, and integrate microscopy data with material property models within a more unified software workflow. Applications:
Advantages: