Differentiable simulations return a value and a gradient with respect to some input parameters. Analagously, functional simulations return a function valid over some region of parameter space. I will show how functional atomic simulations of a prototypical tasks such as geometry minimization[1] and free energy calculation[2] enable rapid forward propagation of parameter uncertainties[3,4] and new approaches for inverse design strategies. Motivated by applications in physical metallurgy, I will discuss how these same methods can apply to atomic simulations in electrochemistry.
[1] Maliyov, Grigorev, TDS NPJ 2025
[2] TDS, Lapointe, Marinica Nat Comm 2025 (to appear)
[3] TDS and Perez, ML:S&T 2025
[4] Perez, Subramanyam, Maliyov, TDS NPJ 2025
Copyright. All Rights Reserved.