Abstract: The invention pertains to the field of organic molecular crystal structure prediction, and particularly related to a double-layer neural network algorithm for high-precision energy calculation of organic molecular crystal structure, including the first round of conventional crystal structure prediction; extract all molecular conformations from existing crystals and calculate their energies; extract all molecular dimers within the Van der Waals radius of the central unit cell and calculate the intermolecular interaction energies; perform molecular conformation analysis to build a convolutional neural network of single-molecule conformational energies; build a molecular dimer energy-corrected convolutional neural network; calculate the total crystal energies.

Abstract: A high-precision energy ranking method used for the crystal structure prediction of organic molecules, including: determining a quantum mechanical radius of a center cell; carrying out an energy calculation in the center cell through a density fragment interaction algorithm; calculating an interaction energy of the molecules outside the center cell acting on the molecules in the center cell within the radius R under quantum mechanical precision; calculating an interaction energy of peripheral extension cells beyond the radius R acting on the molecules in the center cell under molecular mechanical precision; and calculating a total crystal energy.

Abstract: A method for automatically and efficiently fitting repulsive potentials through DFTB includes: optimizing a molecule containing an atomic pair, and performing spacing scan on the atomic pair according to an acting interval of repulsive potentials; performing high-accuracy energy calculation on a scanning result; performing difference calculation on obtained energy to obtain second-order derivatives of force and energy; saving a structure, the energy, the second-order derivatives of force and energy into a database; acquiring energy without a repulsive potential as well as the second-order derivatives of force and energy to obtain target values of the repulsive potentials; transforming the separated target values of the repulsive potentials into splines; splicing multiple splines on a mean position between equilibrium position points to obtain new train data; constructing a train set by means of necessary samples and choosable samples; and fitting the repulsive potentials through singular value decomposition.

Abstract: A method for automatically and efficiently fitting repulsive potentials through DFTB includes: optimizing a molecule containing an atomic pair, and performing spacing scan on the atomic pair according to an acting interval of repulsive potentials; performing high-accuracy energy calculation on a scanning result; performing difference calculation on obtained energy to obtain second-order derivatives of force and energy; saving a structure, the energy, the second-order derivatives of force and energy into a database; acquiring energy without a repulsive potential as well as the second-order derivatives of force and energy to obtain target values of the repulsive potentials; transforming the separated target values of the repulsive potentials into splines; splicing multiple splines on a mean position between equilibrium position points to obtain new train data; constructing a train set by means of necessary samples and choosable samples; and fitting the repulsive potentials through singular value decomposition.