Projected gradient descent algorithm for $\textit{ab initio}$ crystal structure relaxation under a fixed unit cell volume

Yukuan Hu,Junlei Yin,Xingyu Gao,Xin Liu,Haifeng Song

2024-05-05

Abstract:This paper is concerned with $\textit{ab initio}$ crystal structure relaxation under a fixed unit cell volume, which is a step in calculating the static equations of state and forms the basis of thermodynamic property calculations for materials. The task can be formulated as an energy minimization with a determinant constraint. Widely used line minimization-based methods (e.g., conjugate gradient method) lack both efficiency and convergence guarantees due to the nonconvex nature of the feasible region as well as the significant differences in the curvatures of the potential energy surface with respect to atomic and lattice components. To this end, we propose a projected gradient descent algorithm named PANBB. It is equipped with (i) search direction projections onto the tangent spaces of the nonconvex feasible region for lattice vectors, (ii) distinct curvature-aware initial trial step sizes for atomic and lattice updates, and (iii) a nonrestrictive line minimization criterion as the stopping rule for the inner loop. It can be proved that PANBB favors theoretical convergence to equilibrium states. Across a benchmark set containing 223 structures from various categories, PANBB achieves average speedup factors of approximately 1.41 and 1.45 over the conjugate gradient method and direct inversion in the iterative subspace implemented in off-the-shelf simulation software, respectively. Moreover, it normally converges on all the systems, manifesting its unparalleled robustness. As an application, we calculate the static equations of state for the high-entropy alloy AlCoCrFeNi, which remains elusive owing to 160 atoms representing both chemical and magnetic disorder and the strong local lattice distortion. The results are consistent with the previous calculations and are further validated by experimental thermodynamic data.

Computational Physics,Materials Science,Optimization and Control

What problem does this paper attempt to address?

The paper aims to address the problem of ab initio crystal structure relaxation under fixed cell volume conditions. Specifically, this issue is an important step in calculating the static equation of state (EOS) in materials science and serves as the foundation for thermodynamic property calculations. Mathematically, this task can be formulated as an energy minimization problem with a determinant constraint. However, due to the non-convexity of the feasible region and the significant curvature differences in the potential energy surface between atomic and lattice components, existing line search methods (such as the conjugate gradient method) have deficiencies in efficiency and convergence. To overcome these challenges, the authors propose a projected gradient descent algorithm named PANBB. The main features of PANBB include: 1. **Projection of search direction**: Projecting the search direction of lattice vectors onto the tangent space of the non-convex feasible region to avoid severe oscillations caused by directly using lattice forces. 2. **Curvature-aware initial trial step size**: Designing different initial trial step sizes for atomic and lattice updates to accommodate the curvature differences in the potential energy surface across different components. 3. **Non-monotonic line search criterion**: Employing a non-restrictive non-monotonic line search criterion as the stopping rule for the inner loop, allowing small energy increments to improve the overall performance of the algorithm. With these improvements, PANBB not only has theoretical convergence guarantees but also demonstrates significant acceleration and robustness in practical applications. The paper validates the performance of PANBB through a benchmark test set, showing that PANBB is on average approximately 1.41 times faster than the conjugate gradient method and 1.45 times faster than the direct iterative subspace inversion method in terms of runtime, and it consistently converges across all systems. Additionally, PANBB was applied to the static EOS calculation of the high-entropy alloy AlCoCrFeNi, with results consistent with previous computational and experimental data, further verifying its effectiveness and reliability.

Projected gradient descent algorithm for $\textit{ab initio}$ crystal structure relaxation under a fixed unit cell volume

Efficient numerical methods for computing the stationary states of phase field crystal models

First Order Methods for Geometric Optimization of Crystal Structures

Scalable Crystal Structure Relaxation Using an Iteration-Free Deep Generative Model with Uncertainty Quantification

An Adaptive Block Bregman Proximal Gradient Method for Computing Stationary States of Multicomponent Phase-Field Crystal Model

Accelerating crystal structure search through active learning with neural networks for rapid relaxations

Global optimization in the discrete and variable-dimension conformational space: The case of crystal with the strongest atomic cohesion

First Order Methods for Geometric Optimization of Crystals: Experimental Analysis

Gaussian Statistics of the Cosmic Microwave Background: Correlation of Temperature Extrema In the COBE DMR Two-Year Sky Maps

Exploring parameter dependence of atomic minima with implicit differentiation

Efficient high-resolution refinement in cryo-EM with stochastic gradient descent

A Second Order Accurate Scalar Auxiliary Variable (SAV) Numerical Method for the Square Phase Field Crystal Equation

Gradient Type Optimization Methods for Electronic Structure Calculations

Nonadiabatic Molecular Dynamics with Subsystem Density Functional Theory: Application to Crystalline Pentacene

Crystal structure prediction at finite temperatures

Structure Optimization in a Three‐dimensional Off‐lattice Protein Model

Annealing for prediction of grand canonical crystal structures: Efficient implementation of n-body atomic interactions

Crystal Structure Prediction of Binary Alloys via Deep Potential

Automated optimization and uncertainty quantification of convergence parameters in plane wave density functional theory calculations

Machine-learning correction to density-functional crystal structure optimization

Efficient ab initio schemes for finding thermodynamically stable and metastable atomic structures: benchmark of cascade genetic algorithms