Data for performing wavepacket propagation using Fourier Neural Operators

<p>Datasets for training Fourier neural operator (FNO) [Li, Z., et al, 2021] and the trained models in&nbsp;<strong>Accelerating wavepacket propagation with Machine Learning&nbsp;</strong>by Kanishka Singh, Ka Hei Lee, Daniel Pel&aacute;ez and Annika Bande.</p> <ul> <li>data-gaussian-pulse-10000.pickle: Training dataset for 2D FNO model. The dataset contains 10000 wavefunction propagations under 1D double well which is generated by the split-operator method. Each propagation is influenced by a different laser pulse.</li> <li>data-gaussian-pulse-10000.pt: The trained 2D FNO model for wavefunction propagation under 1D double well influenced by laser pulse.</li> <li>anharmonic_FNO_dens.pt: Training dataset for 3D FNO model. The dataset contains 5100 wavefunction propagations under 2D anharmonic potential which is generated by the multi-configurational time-dependent Hartree (MCTDH) method [Beck, M., 2000]. The dataset is structured and a dictionary with input 'x_conv' and output 'y'.</li> <li>alllambda-anharmonic.pt: The trained 3D FNO model for wavefunction propagation under 2D anharmonic potential.</li> <li>anharmonic_FNO_dens-potential.pt: The 2D anharmonic potential of the 2D wavefunction propagations.</li> <li>anharmonic_FNO_dens-true.pt: The ground truth of the 2D wavefunction propagations by the MCTDH method.</li> <li>anharmonic_FNO_dens-pred.pt: The prediction of the 2D wavefunction propagations by the 3D FNO model.</li> </ul>