Thermally Averaged Magnetic Anisotropy Tensors via Machine Learning Based on Gaussian Moments

<ul> <li><strong>D-Tensor Data</strong></li> </ul> <p>The reference data for 3500 configurations of&nbsp;[Co(N₂S₂O₄C₈H₁₀)₂]^2&minus; (CoSar),&nbsp;[Fe(tpa)^Ph]^&minus; (FeTPAPh), and&nbsp;[Ni(HIM₂&minus;py)₂NO₃]⁺ (NiComplex) is generated employing Molpro package [1]. For more details see the original publication [2]. The data is stored in python compressed array format (.npz) with the D-Tensor in cm^-1. The data set contains four <span class="math-tex">\(np.ndarray\)</span></p> <pre><code class="language-python">import numpy as np data = np.load('CoSar.npz') R = data['R'] # Cartesian coordinates of nuclei in Ang. D = data['MAT'] # D-Tensor values in cm-1, D = (D11, D12, D13, D22, D23, D33) N = data['N'] # Number of atoms in each structure Z = data['Z'] # Nuclear charges</code></pre> <ul> <li><strong>AIMD Data</strong></li> </ul> <p>To propagate the periodic cell containing four CoSar molecules, for which D-Tensor was computed above, a data set containing total energies as well as atomic forces of 3500 structures was generated employing VASP package [3-6]. For more details see the original publication [2]. The data is stored in python compressed array format (.npz) with the total energy in kcal/mol and atomic forces in kcal/mol/Ang. The data set contains six <span class="math-tex">\(np.ndarray\)</span></p> <pre><code class="language-python">import numpy as np data = np.load('CoSar_bulk.npz') R = data['R'] # Cartesian coordinates of nuclei in Ang. C = data['C'] # Cell vectors in Ang. E = data['E'] # Total energy in kcal/mol F = data['F'] # Atomic forces in kcal/mol/Ang. N = data['N'] # Number of atoms in each structure Z = data['Z'] # Nuclear charges</code></pre> <ul> <li><strong>References</strong></li> </ul> <p>[1] H.-J. Werner, P. J. Knowles, G. Knizia, F. R. Manby, M. Sch&uuml;tz,et al.,&ldquo;Molpro, version 2020.0, a package of ab initio programs,&rdquo; (2020), see https://www.molpro.net.</p> <p>[2] V. Zaverkin, J. Netz, F. Zills, A. K&ouml;hn, and J. K&auml;stner, &ldquo;Thermally Averaged Magnetic Anisotropy Tensors via Machine Learning Based on Gaussian Moments,&rdquo;<strong> submitted</strong> (2021).</p> <p>[3]&nbsp;P. E. Bl&ouml;chl, &ldquo;Projector augmented-wave method,&rdquo; Phys. Rev. B 50, 17953 (1994).</p> <p>[4] G. Kresse and J. Hafner, &ldquo;Ab initio molecular dynamics for liquid metals,&rdquo; Phys. Rev. B 47, 558 (1993).</p> <p>[5] G. Kresse and J. Furthm&uuml;ller, &ldquo;Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set,&rdquo; Comput. Mater. Sci. 6, 15 &ndash; 50 (1996).</p> <p>[6] G. Kresse and J. Furthm&uuml;ller, &ldquo;Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set,&rdquo; Phys. Rev. B 54, 11169 (1996).</p> <p>&nbsp;</p>