Abstract
<jats:p>Accelerating materials discovery requires closing the characterization loop. Over the last two decades, theory, computation, and, more recently, autonomous synthesis have all advanced rapidly. However, the ability to identify the structural, chemical, defect, and functional variables that control materials properties remains comparatively slow. Laboratory photon-based methods are indispensable for validation, but they often probe too much material, too slow, and too late to guide early-stage materials discovery and are limited in exploration of defects, interfaces, disorder, and nonequilibrium states. We argue that electron microscopy is now ideally positioned to provide the missing low-latency feedback, enabling materials to be read locally at the length scales where function emerges and then promoted to classical characterization, device testing, and manufacturing pipelines. This fills the gap between atomistic computation and macroscopically averaged properties which ultimately offers the possibility to create and explore materials as fast as we can compute them, while adding the unique capability for direct atomic manipulation. Realizing this transition requires microscopy development to advance beyond the traditional priorities of spatial and energy resolution, instrumental stability, and beam-damage mitigation. Four additional capabilities are required: quantifying images and spectra as physical variables; multiplexing sample architectures to encode many hypotheses and sample histories in one experiment; intervening through active beam/probe perturbation and atomic fabrication; and orchestrating experiments through ML-driven rewards, digital twins, and federated instrument networks.</jats:p>