Extreme-Scale Computational Science Discovery in Fluid Dynamics and Related Disciplines

Digital twins are a powerful complement to experimental science, providing access to data that is unattainable through physical experiments alone, and ultimately leading to theories that can aid in the development and design of experiments. We develop cryoflo, a 3D digital twin of cryo-plunging—a critical step in sample preparation for cryogenic electron microscopy (cryo-EM)—to simulate the fluid dynamics and heat transfer of sample vitrification. Cryo-plunging is the standard protocol for rapidly freezing thin biological samples. However, cryo-EM practitioners report success rates below 25% on initial attempts. Moreover, cooling rates, a known impediment to plunging thicker samples, cannot be directly measured inside typical samples only 10 micrometers thick. Cryoflo enables the spatiotemporal study of cooling rates in individual biological samples. It systematically models the cooling process and is built on AMReX, a parallel supercomputing framework with adaptive mesh refinement capabilities, enabling efficient large-scale simulations. Our digital twin has been benchmarked with experimental thermocouple data, showing an average error of around 5%. In addition, cryoflo can load actual sample geometries from z-stack images, prescribe plunging protocols, and visualize heat transfer and fluid motion. Our computational approach has proven instrumental in guiding the design of new plunging protocols, and improving sample vitrification by engineering plunging speed and plunger geometry.