DMN seminar - Mohammed Benaissa (IPR, Rennes)

Optimization Algorithms in Computational Materials Science:
Methods, Applications, and Case Studies
illustration of an optimization procedure

 

Computational materials science faces intricate challenges that necessitate efficient optimization methods for navigating complex material configurations and properties. This abstract delves into the pivotal role of optimization algorithms in addressing these challenges and discovering optimal material configurations. The study explores two optimization techniques: genetic algorithms (GAs) and Bayesian optimization, applied to three case studies involving density functional theory (DFT) simulations. The first case study examines the optimal placement of hydrogen impurities in ZnSb₂O₆, a transparent conducting oxide [1], revealing similarities in behavior to ZnO and suggesting potential improvements in electrical conductivity [2]. The second case study focuses on Cu2S, a solar cell absorber, and unveils insights into hydrogen impurity preferences, akin to Cu2O [3]. These case studies employ genetic algorithms, generating optimal solutions by evaluating energy levels for potential hydrogen impurity positions. Additionally, we highlight the substantial reduction in computational time and energy consumption achieved through optimization algorithms in DFT simulations. A Bayesian optimization is conducted on a simulation of graphene structure. This demonstrates a remarkable 25% reduction in calculation time without compromising accuracy by optimizing charge mixing parameters.

In conclusion, optimization algorithms offer invaluable tools for computational materials science, facilitating efficient exploration of material configurations and properties. The presented case studies exemplify their utility in uncovering optimal solutions and fundamental insights, while also highlighting practical benefits, such as significant reductions in computational time and energy consumption.

 

References:

  1. A.J. Jackson et al., "Computational prediction and experimental realization of earth-abundant transparent conducting oxide Ga-doped ZnSb2O6," Chemistry, 2022.
  2. C.G. Van de Walle, "Hydrogen as a Cause of Doping in Zinc Oxide," Phys. Rev. Lett. 85 (2000) 1012–1015.
  3. D.O. Scanlon and G.W. Watson, "Uncovering the Complex Behavior of Hydrogen in Cu2O," Physical Review Letters, 106 (2011) 186403.
  4. P. Schneider, Bayesian optimization (part 2), Bayesian Optimization (Part 2). (n.d.). -bayesian-optimization (accessed September 28, 2023).