Figures & text for Atomate2 paper

[hrushikesh-s]

• [ ] Phonon DOS
• [ ] Phonon Bandstructure
• [ ] F vs T
• [ ] Renormalized Phonon Bandstructure as a function of Temp
• [ ] Grüneisen parameter
• [ ] CTE
• [ ] LTC
• [ ] NAC + LRC correction
• [ ] Workflow diagram -- using Google diagram

[hrushikesh-s]

[hrushikesh-s]

This workflow outlines the process for conducting anharmonic phonon calculations using the hiphive package.

Methodology: The workflow begins with the required input of a primitive structure and adjustable parameters such as supercell matrix, force field calculator (VASP, M3GNet, CHGNet etc), hiphive fitting options, and temperatures to be used for renormalization and lattice thermal conductivity (κ_lat ) calculations. The process involves steps, including structure relaxation, displacement generation, supercell creation & perturbation, static calculations, and force constants fitting using hiphive. These steps are crucial for preparing the system and obtaining the necessary perturbed forces and displacements for subsequent harmonic & anharmonic phonon calculations. Then, harmonic phonon properties are calculated using phonopy, anharmonic phonon properties using phono3py, renormalization using Thermodynamic Integration, and κ_lat calculation using FourPhonon & phono3py.

The workflow incorporates dynamic features of jobflow such as “Detour” and “Replace”, which allow for the creation of additional jobs based on conditional logic. For example, if the fitting RMSE exceeds a certain threshold, the workflow can detour to increase configurations or displacements, ensuring the accuracy and reliability of the calculations.

Outputs and Metadata: Outputs from each job are stored in MongoDB, a NoSQL database, which provides a flexible schema to accommodate the diverse data generated during the workflow. Some of the important outputs that get stored are job directories, pymatgen structure, hiphive fitting rmse for different cutoffs, harmonic & anharmonic thermal props, phonon frequencies, phonon DOS, phonon bandstructure and κ_lat. Metadata for every job is also stored, including configuration details, displacement matrices, and temperature ranges used during renormalization and κ_lat calculations.

Key differences of the current “atomate2 workflow” from “atomate workflow”:

• can make use of MLFF calculators to run structure relaxation and static jobs
• can dynamically add more configs and displacements based on conditional logic
• ability to use different fireworks configurations for running individual jobs of the flow, thus improving the speed of flow completion

Use Cases and Papers: The atomate counterpart of the same workflow has been utilized for calculating lattice dynamical properties from first principles, as detailed in [1]. This paper demonstrates the application of the workflow in calculating interatomic force constants, κ_lat, coefficient of thermal expansion (CTE), and vibrational free energy (Fvib) to name a few. The workflow also performs phonon renormalization for dynamically unstable compounds, highlighting its capability to compute high-temperature phase diagrams in the future. The workflow's adaptability make it a valuable asset for researchers aiming to explore the thermal and vibrational properties of materials.

References [1] Z. Zhu et al., “A High-Throughput Framework for Lattice Dynamics.” ChemRxiv, Mar. 15, 2024. doi: 10.26434/chemrxiv-2024-c82v8.

[Zhuoying]

The figures look nice and we will finish the writing next week.

[hrushikesh-s]

[hrushikesh-s]