Core Concepts
The authors propose a novel numerical framework that leverages multipole expansions to accelerate the simulation of crystalline defects by systematically enhancing the accuracy of approximate multipole tensor evaluations and employing continuous Green's functions.
Abstract
The content presents a comprehensive approach to accelerating the convergence of crystalline defect simulations through the use of higher-order far-field boundary conditions. The key highlights and insights are:
The authors review the variational formulation for the equilibration of crystalline defects and the multipole expansion framework that characterizes the discrete elastic far-fields surrounding point defects.
They introduce a numerical scheme that exploits the low-rank structure of the multipole expansions to minimize the domain size effects in defect simulations. The scheme iteratively improves the boundary condition by systematically following the asymptotic expansion of the far field.
To enable efficient implementation, the authors propose the use of continuous multipole expansions instead of discrete ones, leveraging continuous Green's functions to avoid the computational complexities associated with discrete Green's functions and their derivatives.
Rigorous error estimates are provided for the method, and a range of numerical tests are conducted to assess the convergence and robustness of the approach.
The numerical experiments demonstrate that the proposed framework for higher-order boundary conditions effectively achieves accelerated convergence rates with respect to computational cell size for various point defect scenarios, including vacancies, divacancies, interstitials, and micro-cracks.
Stats
The authors provide the following key figures and metrics to support their analysis:
The decay of strains |D¯u_i,R_dom(ℓ)| for different orders of predictors (i = 0, 1, 2) obtained using the proposed algorithm, showing the improved decay rates for higher-order predictors.
The convergence of geometry error ∥D¯u - D¯u_i,R∥_ℓ2 and energy error |E(¯u) - E(¯u_i,R)| with respect to the computational cell size R, demonstrating the accelerated convergence achieved by the higher-order boundary conditions.