Description
Chemical vapor deposition (CVD) has emerged as key technology for bringing next-generation materials to market, yet process discovery still largely follows an Edisonian trial-and-error approach, blind and constraint by conventional reactors. This is not only wasteful and frustratingly slow, but hinders scientific breakthroughs in crystal growth and innovation in new deposition technology. This talk will introduce a versatile micro-CVD platform that enables direct in-line process screening by electron microscopy as well as interaction in real time - synchronously or on-the-fly - with ongoing atomic-scale reactions. We show how this can transform process exploration, focussing on the crystal growth of atomically thin transition metal dichalcogenide layers.[1] We use large datasets on spatio-temporal basal plane nucleation kinetics and propagation of tens of thousands of individual 1D reaction facets to self-consistently guide atomistic reaction exploration via machine-learned interatomic potentials based on the MACE architecture.[2]
[1] Yang et al., Chem. Mat. 37, 989 (2025)
[2] Gsanyi et al., arXiv:2401.00096, 2023
