Abstract
Mathematical modeling of magnets is of great importance for both fundamental science and applications. In this work, we present the numerical results of simulating statistical properties of the classical Heisenberg magnet using the atomistic approach implemented in the StatASD code. For the calculations the modified fourth-order Runge–Kutta scheme with temperature noise compensation is used. Parallelization of calculations is performed using the OpenMP library. The data is accessed via the three-dimensional Morton Z-curve. The crystal lattice structure is specified during compilation based on auto-generated C++ code segments. As a result, a parallelization efficiency of 40 \(\%\) is achieved, and performance is improved by an order of magnitude compared to the widely known UppASD code. For managing calculations, the Results and Algorithms Control System (RACS) is used. RACS is non-relational DBMS designed for massive HPC applications. RACS integrates effortlessly into Python scripts and provides a high-level command-line interface for launching calculations. This tool enables large-scale calculations on a cluster with automatic load balancing of compute nodes. Once the calculations are completed, RACS enables comprehensive analysis and post-processing of the results. The described approach made it possible to calculate the integral coefficients of the equations of correlation magnetodynamics, which describes the magnet in the continuum approximation. Based on the analysis of the StatASD results, a new algorithm for calculating the magnet’s entropy was developed, specifically, the entropy can be calculated as the sum of entropies of one- and two-particle distribution functions for various coordination spheres.
