Fig. 1: CrystalNet System Overview.
From: Towards end-to-end structure determination from x-ray diffraction data using deep learning
As input, CrystalNet takes in 1D powder x-ray diffraction (PXRD) patterns that may be obtained by azimuthally integrating a 2D diffraction pattern such as shown as the black square. It also takes in chemical composition ratios, and the atomic coordinates of known structures. In the Solver it processes each input item with a specialized branch (pink, purple and black). It then fuses them into one unified latent vector of length 512, which is passed through the charge density regressor to produce a voxelized 3D charge density map at arbitrary resolution. See “Methods” for more details.
