Font Size: a A A

New Method For Constructing Multi-Dimensional Chemical Reaction Potential Energy Surface

Posted on:2017-01-18Degree:MasterType:Thesis
Country:ChinaCandidate:L Y ChenFull Text:PDF
GTID:2271330482995351Subject:Analytical Chemistry
Abstract/Summary:PDF Full Text Request
Potential energy surface is the basis of theory of chemical reaction process, and by the potential energy surface of reaction system we can get reactive dynamic information such as reaction rate, reaction cross section, so the construction of high precision of chemical reaction potential energy surface has been the most important key of chemical reaction. In this paper, we propose a method based on the complex reaction system with multiple channels to construct the potential energy surface which is based on the special configuration of the reaction path. In the first two chapters, we introduce the method of ab initio potential energy surface construction and related theory as well as the basic theory of classical and quantum molecular reaction dynamics. And the nine dimensional quantum dynamics model for C2H+H2 reaction is introduced in detail in this paper.In the third chapter, we use the method of polynomial fitting and partition method to construct the full dimensional potential energy surface of H3 system under UCCSD(T)/AVTZ method. We take three nuclear spacing representation system configuration and on the reaction coordinate plane of H exchange reaction channel 7 center were selected and the fitting data set in accordance with the rule of the distance divided into 7 cross domain data, the were polynomial fitting regional potential energy surface, and finally by the weight function integration for full dimensional potential energy surface. Because the reaction energy barrier (0.41 eV) is very small, we can use the partition method to construct the corresponding full dimensional potential energy surface, which is based on the data set in the 30000cm-1 and 50000cm-1. The results showed that the three full dimensional potential energy surface fitting data, parameters and mean root mean square error RMSE values were 127500/1179/11.99cm-1,57296/5537/0.49cm-1,12749/7603/1.12cm-1. Due to the reaction to low barrier, three potential energy surface contour diagrams are consistent, describing the reaction process well. But the partition method can significantly improve the fitting parameters to reduce the fitting error.In the fourth chapter, we use artificial neural network and quasi classical trajectory supplement combination method to construct the OH3 system partition full dimensional potential energy surface under UCCSD(T)/AVTZ method. We used nuclear spacing indicates that the system configuration, because of the replacement of H atoms in the system, we maintain RoH1≤ROH2≤ROH3 and eventually get 23104 data points. As OH3 system exists the exchange and capture of H atoms, we selected 5 center points on the two channels and the 6-50-80-1 neural network model was selected to get regional potential energy surface to integrate a full dimensional potential energy surface. The results show that the asymptotic region data fitting error is smaller and reaction zone data fitting deviation larger, which need more configurations, but the error of fitting all fall within the [-20,15]cm-1. The RMSE is only 3.2cm-1, significantly less than the literatural 12.99 and 11.29cm-1. The potential energy surface contour map and the quantum dynamics calculation are performed. And the results show that the potential energy surface accurately describe the two reaction channels and reaction probability curve are consistent with literatural values, indicating that the partitioning method can improve the accuracy of the potential energy surface.In the fifth chapter, we use the neural network method to fit the full -dimensional potential energy surface of the C2H3 system under the UCCSD(T)-F12a/AVTZ, and the C2H+H2 reaction is calculated by the nine - dimensional quantum dynamical model. System configurations are displayed with the internuclear distance. Considering H atom replacement in H2, maintain RCH1≥RCH2. The potential energy surface is fitted with the 10-70-90-1 neural network model, acquiring the final reactive barrier for 2.06eV and total 68478 data points, RMSE value for 4.28 MeV. These data points were randomly removed 15% to refit the new potential energy surface with fitting RMSE 4.12 meV, and the probability values of its quantum reaction are accordance with ones of the original potential energy surface, indicating that the potential energy surface has been converged. In order to test the accuracy of the potential energy surface, we analyze the potential energy surface contour graph and energy values of data points and energy deviations distribution and quantum dynamics calculation results, indicating that the potential energy surface accurately describes the C2H + H2 reaction process and the variation of the energy and configuration. There are more data points in the reaction zone and the fitting deviation values of 95% of data points are fall in [-10,10] meV, illustrating that the distribution of the data points are more reasonable. The kinetic results show that the excitation of H2 can greatly promote the occurrence of the reaction. In fragment, excitation of the C-C bond inhibits the reaction occurance, which is more obvious in the excitation of H2; The bending vibration of the C2H has the same effect, which is more obvious in the excitation of H2 as well; As the H atom is not involved in the reaction, C-H vibration has no effect for the reaction.
Keywords/Search Tags:H3, OH3, C2H3, ab initio potential energy surface, partition potential energy surface, time-dependent quantum wave packet
PDF Full Text Request
Related items