Research Roadmap

Year 1: 2020–2021


Develop *version zero* of the full system simulation: prototype connectivity software and a dummy decision engine in Julia to communicate inputs and boundary conditions between all relevant software components of the simulation.

We are also developing interface code between Julia and LAMMPS, as well as between Julia and the Atomic Simulation Environment (ASE). Moreover, we have begun using Julia-native density functional theory (DFT) and molecular dynamics (MD) codes as alternatives to QuantumESPRESSO and LAMMPS. These options are proving useful for workflow design and testing, and may become key elements even of our production-scale simulations.


Spin up center resources for software development, e.g., continuous integration and deployment. Compile an initial selection of simulation components using Tapir/LLVM and integrate Tapir/LLVM into TACO and Tiramisu.


Compute aerothermal loads due to hypersonic flows using Exasim, and an ideal gas and equilibrium chemistry models with LES on canonical flow geometries (e.g., double-ramp and cylindrical cowl leading edge). Initial development and validation of fixed-charged and reactive potential models. Initial validation of range-separated-hybrid density functionals with higher-level wavefunction theory methods.


Develop a comprehensive inventory of uncertainty sources in multiscale simulation approach and conduct (non-intrusive) forward UQ studies on each component. Set up and test high-temperature heating system combined with transient electrical heating; carry out initial oxidation experiments on HfB$_2$. \textit {[Progress: We have developed a UQ roadmap and a comprehensive inventory of uncertainty sources, and are currently building new UQ software in Julia; this is on track for the end of Year 1. The experimental effort has been split into two: a high-temperature thermogravimetric analyzer (TGA), as described in this milestone, and high-temperature plasma arc jet experiment that is newly conceived. TGA experiments on Hf oxidation are set to begin in June 2021.]}

Year 2: 2021–2022


**Full system simulation:** calculation of the oxidation profile in HfB2 using reactive potentials on canonical flow geometry. Simulation will incorporate one-way coupling with flow to simulation from Year 1 using and chemically reacting equilibrium gas models. This will demonstrate our version zero of the software to connect simulation components.


Incorporate vibrational excitation (two temperature model) in hypersonic flow calculation. Develop robust UQ procedures for updating classical potentials “on the fly” from AIMD information. Force matching of fixed-charge and reactive potential models on AIMD results. First validation of fixed-charge and reactive potential models using Year 1 experimental data. Introduce multi-fidelity approaches to MD simulation.


Use Tapir/LLVM to compile the whole simulation codebase and integrate Tapir/LLVM into Julia to enable Tapir/LLVM to optimize parallel scientific software written in Julia. Verification and performance analysis of individual simulation components. Add support for generating code for heterogeneous systems (e.g., multicore CPUs and/or GPUs). Develop Julia’s differentiable-programming capabilities and a CSI tool for legacy codes. Prototype connectivity software using Julia and CSI to communicate UQ information between software components of the simulation.


Develop Bayesian formulations of inter-model calibration; apply using gradient-free sampling algorithms. Provide information on oxidation profiles of HfB$_2$, via total weight change measurements, for a range of starting particle sizes and atmospheres.

Year 3: 2020–2021