Searching Stable Chemical Space through Computational Methods using Multi-Objectives Genetic Algorithm

HUSEN, AMIR; Munoz, Jorge

doi:10.5281/zenodo.17268398

Searching Stable Chemical Space through Computational Methods using Multi-Objectives Genetic Algorithm

HUSEN, AMIR; Munoz, Jorge

Publisher: Zenodo
DOI: 10.5281/zenodo.17268398
Repository: zenodo.org
Language: en

Abstract

Abstract: Understanding the stability of crystalline materials is central to designing new functional compounds. In this work, we integrate Phonopy, a first-principles phonon calculation package, with PyGAD, a Python-based library for multi-objectives genetic algorithms, to systematically search stable regions of chemical space and structural stability. Phonopy enables the calculation of phonon dispersion relations, where the presence or absence of imaginary vibrational modes determines dynamical stability. These results are coupled with PyGAD’s multi-objective genetic algorithm (MOGA), where Born–von Kármán force constants evolve across generations to efficiently search parameter space and identify Pareto-optimal solutions. The workflow produces a “stability map” for body-centered cubic (BCC) structures such as Fe and V, balancing stability criteria with agreement to experimental phonon spectra. Importantly, all simulations are executed on the National Energy Research Scientific Computing Center (NERSC) JupyterHub platform, which provides scalable high-performance computing resources and a reproducible environment for running optimization-driven phonon calculations. By combining phonon-based lattice dynamics with evolutionary algorithms on HPC infrastructure, this framework demonstrates a transferable and efficient strategy for navigating chemical space, accelerating materials discovery, and enabling high-throughput stability screening.

Full text

Abstract: Crystal stability plays a key role in the design of new materials. This work combines Phonopy [1, 2] for phonon with PyGAD [3], a multi-objective genetic algorithm (MOGA) to explore stable regions of chemical space. By evolving Born–von Kármán force constants [4] and assessing phonon spectra, the workflow generates stability maps for BCC crystal structures. Implemented on NERSC HPC, this approach enables scalable, highthroughput stability screening and accelerates materials discovery. Atomic-level insight: Enhances understanding of BCC crystal behavior to explain thermal and mechanical properties. Computational efficiency: Evolutionary algorithms reduce analysis time and resource use compared to experiments. Foundation for discovery: Provides a scalable framework to explore stability and performance across metallic systems. Searching Stable Chemical Space through Computational Methods using MultiObjectives Genetic Algorithm Authors: Amir Husen1, Jorge A. Munoz2 1. Amir Husen, Research Associate & PhD candidate in Computational Science, The University of Texas at El Paso; Email: [email protected] 2. Jorge A. Munoz, Associate Professor in Physics & Computational Science, The University of Texas at El Paso; Email: [email protected] Objectives: Computational Workflow: Setup BCC crystal system & BvK model Generate initial force constants (up to 2nd coordination shell) Calculate phonon using Phonopy Optimize force constants with MOGA - Stability constraints for fitness Evaluate phonon spectra & convergence Generate stability map & identify stable parameter space Abstract Results and Discussion: Fig-1: Phonon dispersion of Fe Fig-2: Phonon dispersion of V Fig-3: BCC crystal stability map in parameter space Fig-1 shows excellent agreement between simulated and experimental phonon spectra for BCC Fe [5], validating the model parameters. For BCC V (Fig. 2), noticeable deviations appear compared to experimental data [6], likely due to differences in lattice dynamics or parameter-fitting limitations. Fig-3 illustrates stable regions in the mass–lattice constant parameter space. These regions indicate parameter sets that yield dynamically stable BCC structures. The computational framework integrates phonon analysis with genetic algorithms to efficiently map stability regions in BCC crystals. Results validate the method for Fe and reveal parameter sensitivities for V, supporting highthroughput stability exploration. Conclusion: References: Scan here!