Automatic efficient DFTB repulsive potential fitting method

Abstract

The invention provides an automatic efficient DFTB repulsive potential fitting method. The method comprises the following steps: step (1) optimizing molecule containing atom pair; performing atom pairspacing scanning according to a repulsive potential effect interval; step (2) performing high-precision energy calculation on a scanning result; step (3) energy different calculation power and energysecond derivative; step (4) saving the structure, the energy, the power and the energy second derivative in a database; step (5) obtaining the energy not containing the repulsive potential, the power, and the energy second derivative so as to obtain a repulsive potential target value; step (6) converting the separated repulsive potential target value into a spline curve; step (7) splicing multiple spline curves at an average location between the balance location points to obtain brand new training data; step (8) constructing a training set by using a unnecessary sample and a selection sample;and step (9) decomposing the fitting repulsive potential by using a singular value. The key type imbalance fitting problem is overcome.

Description

technical field [0001] The invention belongs to biological information technology and relates to the design of functional materials, in particular to an automatic and efficient DFTB repulsion potential fitting method. Background technique [0002] DFTB is a high-precision quantum chemical method with fast calculation, which achieves an effective balance between calculation scale and precision, and has a wide range of applications in the fields of biological protein simulation and functional material design. Repulsion potential is an important part of DFTB parameters, and its reliability seriously affects the accuracy of calculation and simulation. [0003] The current DFTB parameterization scheme establishes the training set manually and uses a large number of training sample points for spline curve fitting to determine the repulsion potential curve. Since the same repulsion potential needs to describe multiple bonding forms of two atom pairs, if Using a large number of tra...

AI Technical Summary

Problems solved by technology

[0003] The current DFTB parameterization scheme establishes the training set manually and uses a large number of training sample points for spline curve fitting to determine the repulsion potential curve. Since the same repulsion potential needs to describe multiple bonding forms of two atom pairs, if Using a large number of training sample points for fitting can easily lead to the inability to accurately cover different bonding forms. For example, C...O interaction has two bonding modes of C-O / C=O, and the equilibrium positions are obviously different. Direct spline Fitting is difficult to balance two bond lengths

[0004] The spline fitting method can only fit one atom pair curve at a time, and the fitting sequence has an obvious sequence relationship, which not only requires extremely high complexity of the training set, but also causes compatibility and matching problems of the fitting repulsion potential

For example, if you want to develop O...O interaction, you must prepare H...H and O...H repulsion potential before you can develop O...O, but the developed O...O interaction is not Determine whether it is suitable for a system other than these three functions. It will cause repeated iterations of the development parameter process, the development speed is slow, and the fitting involves the establishment of a large number of training sets and test sets, parameter adjustments, and test corrections, requiring a lot of manual intervention

Images