Search

CN-122024872-A - Construction method of photolytic full-dimensional potential energy surface of first excited state of nitrous acid molecule

CN122024872ACN 122024872 ACN122024872 ACN 122024872ACN-122024872-A

Abstract

The invention discloses a construction method of a photolytic full-dimensional potential energy surface of a first excited state of nitrous acid molecules, and relates to the technical field of computational chemistry and molecular reaction dynamics. The invention adopts a multi-reference configuration interaction method to carry out systematic evaluation and optimization selection on the molecular activity space, and carries out data point sampling and calculation in the molecular full-dimensional global configuration space, thereby completely covering the key reaction channel for generating hydroxyl and nitric oxide products by dissociation of nitrous acid molecules. The method has the advantages that the key reaction channels for generating hydroxyl and nitric oxide products by nitrous acid molecule dissociation are completely covered by adopting a multi-reference configuration interaction method system to evaluate and optimally select the active space, the potential energy description precision in a strong electron correlation area such as O-N bond dissociation is improved by the strategy, the potential energy surface deviation problem possibly caused by the fact that the theoretical precision is insufficient in the existing method is overcome, and a reliable full-dimensional potential energy surface data base is provided for accurately describing the dynamic evolution process of excited state molecules along the main reaction channels.

Inventors

  • CHEN QI
  • Xue Xiucong

Assignees

  • 西北大学

Dates

Publication Date
20260512
Application Date
20260130

Claims (10)

  1. 1. The construction method of the photolytic full-dimensional potential energy surface of the first excited state of the nitrous acid molecule is characterized by comprising the following steps: step a, evaluating and optimizing the selection of an active space by adopting a multi-reference configuration interaction method system, and carrying out sampling calculation of data points in a full-dimensional configuration space based on the active space, so that a key reaction channel for generating hydroxyl and nitric oxide products by dissociation of nitrous acid molecules is completely covered; b, in a full-dimensional configuration space covering the dissociation channel, combining sampling in a normal mode space with a dense sampling strategy aiming at key dissociation coordinates, and performing large-scale de-novo energy point sampling; step c, screening initial data points obtained by sampling based on geometric differences among configurations to construct a training data set with high quality and low redundancy; Step d, adopting a neural network to fit a potential energy surface, and using a multi-layer feedforward neural network architecture comprising an input layer, two hidden layers and an output layer, wherein each hidden layer comprises 50 neurons, and the neural network input layer is constructed by adopting a low-order displacement invariant polynomial (PIPs) to realize the displacement invariance of equivalent atoms; Setting segmentation weights in neural network training, setting segmentation weights for samples in different energy intervals, gradually decreasing the weight values along with the increase of energy, and adopting the same weights when fitting potential energy surfaces of S 0 and S 1 ; Step f, iteratively updating parameters of the neural network by minimizing a loss function, determining weights and bias items of the neural network by minimizing a residual error square sum, and performing iterative optimization on the weights and bias items of the neural network by a Levenberg-Marquardt algorithm; Step g, calculating molecular characteristics by utilizing a trained neural network potential energy surface, and performing structural optimization and frequency calculation to obtain geometric parameters and vibration frequency of trans-HONO, cis-HONO and transition states; And h, comparing and verifying the root mean square error of the ground state and the excited state of the potential energy surface obtained after the neural network method is fitted, and comparing the obtained geometric parameters and vibration frequency with the existing theoretical and experimental data to verify the accuracy and reliability of the potential energy surface.
  2. 2. The method for constructing the photolytic full-dimensional potential energy surface of the first excited state of the nitrous molecule according to claim 1, wherein the method comprises the following steps: The step a specifically comprises adopting a multi-reference configuration interaction (MRCI) method, performing state average complete active space self-consistent field (SA-CASSCF) calculation by using an aug-cc-pVTZ base group in an active space containing the lowest five single states, further adopting a Davidson correction multi-reference configuration interaction method (MRCI +Q), testing a plurality of active spaces containing different electron numbers and track numbers, stretching bond lengths of O-N bonds at the MRCI +Q level based on the active spaces, fixing other coordinates in a ground state balance configuration, and finally selecting an active space with a smooth potential energy curve in a dissociation region for subsequent calculation, wherein the finally selected active space contains 14 electrons and 9 tracks, namely (14 e, 9O).
  3. 3. The method for constructing the photolytic full-dimensional potential energy surface of the first excited state of the nitrous molecule according to claim 1, wherein the method comprises the following steps: the step b specifically comprises the steps of firstly, at Sampling initial points in normal mode space of states trans-HONO, cis-HONO and Transition State (TS), then densely sampling in the range of O-N atomic distance (R O-N ) being [1.0,10.0] A, randomly distributing the rest five coordinates in the following interval ：R H-O ∈[0.7,3.0]Å、R N=O ∈[0.8,2.5]Å、θ ONO ∈[0.0,180.0]°、θ HON ∈[0.0,180.0]°、φ HONO ∈[0.0,180.0]°,, screening out the number of energy points for potential energy surface fitting, wherein the number of energy points is about 63,000, and the energy is relative to that of the potential energy surface fitting The trans-HONO state is below 10.0eV.
  4. 4. The method for constructing the photolytic full-dimensional potential energy surface of the first excited state of the nitrous molecule according to claim 1, wherein the method comprises the following steps: The step c specifically includes introducing Euclidean distance (Euclidean distance) as a screening standard, when Euclidean distance between a new molecular configuration and an existing molecular configuration is smaller than 0.1A, considering that the two configurations are highly close, and the sampling result is not included in subsequent data set construction, wherein the Euclidean distance has a calculation formula as follows:
  5. 5. The method for constructing the photolytic full-dimensional potential energy surface of the first excited state of the nitrous molecule according to claim 1, wherein the method comprises the following steps: The symmetric polynomial expression of the low-order Permutation Invariant Polynomials (PIPs) in the step d is: Wherein l ij is the order of a single item, Representing a symmetrization operator, N representing the number of nuclei, Is a morse-like variable (parameter α=2/3 a -1 ）,r ij represents the atomic spacing).
  6. 6. The method for constructing the photolytic full-dimensional potential energy surface of the first excited state of the nitrous molecule according to claim 1, wherein the method comprises the following steps: The mathematical expression of the kth neuron of the (i+1) th layer of the neural network in the step d is as follows: where N i represents the number of neurons in the ith layer, To connect the weight of the ith layer jth neuron with the (i+1) layer kth neuron Is the bias of the (i + 1) th layer of the kth neuron, Representing the transfer function of layer (i+1).
  7. 7. The method for constructing the photolytic full-dimensional potential energy surface of the first excited state of the nitrous molecule according to claim 1, wherein the method comprises the following steps: The specific value of the segmentation weight in the step e is that the weight is 1.0,6-6.5eV, 0.8,6.5-7eV, 0.6,7-7.5eV, 0.4,7.5-8eV, 0.1,8-10eV and 0.001 when the energy is <6.0 eV.
  8. 8. The method for constructing the photolytic full-dimensional potential energy surface of the first excited state of the nitrous molecule according to claim 1, wherein the method comprises the following steps: The minimized loss function expression in the step f is: Wherein N dat represents the total number of energy points, w is an energy-dependent weight factor, E represents the adiabatic energy in the S 0 or S 1 state, where The term is a regular term, p represents all parameters to be optimized (weight and deviation) in the neural network function, and t is a tiny constant (10 -5 is taken in calculation).
  9. 9. The method for constructing the photolytic full-dimensional potential energy surface of the first excited state of the nitrous molecule according to claim 1, wherein the method comprises the following steps: the geometrical parameters in steps g and H include bond length R O-H 、R N=O 、R O-N and bond angle θ HON 、θ ONO and dihedral angle Φ HONO , and the vibration frequencies include simple harmonic frequencies corresponding to H-O stretching (v 1 ), n=o stretching (v 2 ), HON bending (v 3 ), O-N stretching (v 4 ), ONO bending (v 5 ) and torsion (v 6 ) modes.
  10. 10. The method for constructing the photolytic full-dimensional potential energy surface of the first excited state of the nitrous molecule according to claim 1, wherein the method comprises the following steps: The reliability of the potential energy surface in the step h is verified by ① calculating the Root Mean Square Error (RMSE) of the ground state (S 0 ) and the excited state (S 1 ), ② comparing the optimized geometric parameters (key length, key angle and dihedral angle) and vibration frequency with reported experimental data and high-precision theoretical results, and confirming the coincidence degree.

Description

Construction method of photolytic full-dimensional potential energy surface of first excited state of nitrous acid molecule Technical Field The invention relates to the technical field of computational chemistry and molecular reaction dynamics, in particular to a construction method of a photolytic full-dimensional potential energy surface of a first excited state of nitrous acid molecules. Background The first excited state full-dimensional potential energy surface of nitrous acid (HONO) molecules is obtained by a high-precision first calculation electronic structure calculation method under the similar framework of Bern-Oppenheimer, and the system description system is a (3N-6) dimensional potential energy hypersurface with the change of electron energy along with nuclear coordinates when in a first electronic excited state, and the potential energy can represent the integral change characteristic of the excited state potential energy in the process of evolving the system from a Franck-Condon region to an OH+NO dissociation outlet channel along a main reaction coordinate, and is a key theoretical basis for developing HONO photodissociation quantum dynamics simulation and predicting product distribution. Theoretical research on the photodissociation process of nitrous acid molecules in the prior art mainly depends on construction of ground state and excited state potential energy surfaces. In recent years, a machine learning potential energy surface construction method based on large-scale ab initio data points is gradually applied to the system, the coverage of the potential energy surface is expanded to a certain extent, but the reliability of the potential energy surface in a dissociation region still needs to be further verified. The existing method still has obvious limitations, namely, the existing method is limited by the adopted quantum chemical calculation method, the conventional CASSCF method is difficult to achieve quantitative accuracy in calculating O-N bond breakage and other areas, the existing data sampling strategy is concentrated on a molecular equilibrium configuration area, the configuration space coverage of a dissociation outlet channel is insufficient, so that a potential energy surface has a larger error on a photodissociation key path, in addition, in the potential energy surface fitting process, the influence of different energy intervals on the potential energy surface accuracy is not fully considered, the kinetic applicability of the potential energy surface is further limited, and the problems restrict the reliable theoretical description of HONO photodissociation rate and product state distribution, so that the deep understanding and prediction of the photochemical behavior of nitrous acid atmosphere are restricted. Therefore, we provide a method for constructing a photolytic full-dimensional potential energy surface of a first excited state of nitrous acid molecules, which is used for solving the above problems. Disclosure of Invention The invention aims to provide a construction method of a photodecomposition full-dimensional potential energy surface of a first excited state of nitrous acid molecules, which solves the problems of insufficient precision and sparse sampling of high-energy areas such as a dissociation channel in the prior art by adopting a strategy of combining multi-reference configuration interaction calculation and neural network fitting. In order to solve the technical problems, the invention is realized by the following technical scheme: The invention discloses a construction method of a photolytic full-dimensional potential energy surface of a first excited state of nitrous acid molecules, which comprises the following steps: step a, evaluating and optimizing the selection of an active space by adopting a multi-reference configuration interaction method system, and carrying out sampling calculation of data points in a full-dimensional configuration space based on the active space, so that a key reaction channel for generating hydroxyl and nitric oxide products by dissociation of nitrous acid molecules is completely covered; b, in a full-dimensional configuration space covering the dissociation channel, combining sampling in a normal mode space with a dense sampling strategy aiming at key dissociation coordinates, and performing large-scale de-novo energy point sampling; step c, screening initial data points obtained by sampling based on geometric differences among configurations to construct a training data set with high quality and low redundancy; Step d, adopting a neural network to fit a potential energy surface, and using a multi-layer feedforward neural network architecture comprising an input layer, two hidden layers and an output layer, wherein each hidden layer comprises 50 neurons, and the neural network input layer is constructed by adopting a low-order displacement invariant polynomial (PIPs) to realize the displacement invariance of equivalen