Figure 4
From: Accelerated spin dynamics using deep learning corrections
Conservation of energy and magnetization on \({8 \times 8 \times 8}\) cubic lattice. Predictions of \(\mu _{|{\tilde{e}}(t)|}\), std\(({\tilde{e}}(t))\), \(\mu _{|{\tilde{m}}(t)|}\), and std\(({\tilde{m}}(t))\) as a function of time at \(\mathbf{a }, k_{B}T/J=0.4\), \(\mathbf{b }, k_{B}T_{c}/J\approx 1.44\), and \(\mathbf{c }, k_{B}T/J=2.4\). Black line, yellow line, and red line represent data obtained from spin dynamics simulations with \(\uptau =10^{-1}\), \(\uptau =10^{-2}\), and \(\uptau =10^{-3}\), respectively, while blue line represents data from Deep Learning (DL) correction. These figures show that the effect of averaging over disordered spins for \(L=8\) is stronger than for \(L=4\) above the critical temperature \({k_B}T_{c}/J\).
