Method for predicting bond dissociation energy

Abstract

The invention discloses a method for predicting bond dissociation energy, which comprises the following steps of: (1) drawing structures after different substituent groups replace H at two different positions of No.1 C and No.2 C on a five-membered heterocyclic compound ring, optimizing and calculating Gibbs free energy to obtain bond dissociation energy of C-H bonds at No.3 C and No.4 C, and forming an output data set; selecting proper data as descriptors; (2) processing the acquired data, and normalizing the acquired data into a range of [0, 1] according to a formula; (3) constructing a random forest model, and selecting proper parameter settings; and (4) predicting through a machine learning model to obtain key dissociation energy. By means of the method, the problems that experimental equipment is expensive, experimental operation is complex, calculation time of software calculation is long, calculation cost is high and the like can be solved, prediction of the key dissociation energy can be achieved by running codes, and the method has the advantages of being short in consumed time, low in cost and the like.

Description

Technical Field

[0001] The present invention relates to the field of bond dissociation energy prediction, and in particular to a method for predicting bond dissociation energy. Background Art

[0002] Bond dissociation energy is a physical quantity that measures the strength of chemical bonds. It indicates the energy required to break a chemical bond under specific conditions. Bond dissociation energy usually refers to the energy required to split a molecule in the ground state into atoms in the ground state at absolute zero. People are usually accustomed to using the bond dissociation energy value at 298.15K. The calculation of bond dissociation energy usually involves structural optimization and vibration analysis of reactants and products, as well as single-point energy calculations. These calculations can help determine the energy required to break a chemical bond under specific conditions.

[0003] There are two main ways to predict bond dissociation energy in the prior art, the first is through experimental methods, and the second is through calculation methods. Bond dissociation energy is measured by traditional experimental methods: for example, ① spectroscopy, including infrared spectroscopy, ultraviolet spectroscopy, nuclear magnetic resonance spectroscopy, etc. The bond dissociation energy is mainly determined by analyzing the energy difference generated by the substance during the absorption or emission spectrum. ② Mass spectrometry. The bond dissociation energy is determined by measuring the mass-to-charge ratio of ions released by the breaking of bonds in compound molecules. ③ Thermochemical methods, including combustion heat determination, reaction heat determination, etc. The bond dissociation energy is determined by observing the energy difference generated by the reaction substance during the heat change process. The bond dissociation energy is calculated by theoretical calculation methods, such as computer simulation and calculation using Gaussian software.

[0004] However, the traditional experimental method for measuring bond dissociation energy requires expensive experimental instruments and complicated experimental operation procedures. Theoretical calculation methods for calculating bond dissociation energy have the problems of long calculation time and slow speed for Windows systems, and high calculation cost and difficulty in obtaining for Linux systems. All of these have hindered the use of bond dissociation energy in high-throughput and large-scale research.

[0005] Therefore, the present invention provides a method that can be obtained neither through experiments nor calculations, but only through predictions of machine learning models. Summary of the invention

[0006] In view of this, the present invention provides a method for predicting bond dissociation energy.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0008] A method for predicting bond dissociation energy, characterized in that it comprises the following steps:

[0009] (1) Obtaining the dataset:

[0010] Draw the structure of H at different positions of C1 and C2 on the five-membered heterocyclic compound after different substituents are substituted, optimize the structure, and calculate the Gibbs free energy to obtain the bond dissociation energy of the C-H bond at C3 and C4 to form the output data set; count the number of different types of atoms in different substituents R, query the electronegativity of the atom connected to C in RC by itself, obtain the bond length and bond angle of RC bond, the dipole moment and electronic energy of the molecule through Gaussian09, and use these data as descriptors;

[0011] (2) Normalization processing:

[0012] The acquired data is processed and normalized to the range of [0, 1] by using the formula;

[0013] (3) Machine Learning Model:

[0014] Build a machine learning model based on Python-Jupyternotebook and select parameters;

[0015] (4) Prediction of bond dissociation energy:

[0016] The bond dissociation energy is predicted by this machine learning model.

[0017] Preferably, in the step (1), the structure after different substituents replace the H on the ring of the five-membered heterocyclic compound is drawn using GaussView5 software, and the different substituents include 28 substituents, namely: -CH3, -OH, -CH2OH, -NH2, -N(CH3)2, -F, -Cl, -Br, -CH2NHCH3, -NOHCH, -PH, -O - , -SH, -NHCH3, -NHCOCH3, -CHCH2, -COCH3, -COCH2CH3, -OCOCH3, -OCOCH2CH3, -CN, -NO2, -OSOCH3, -CHO, -CF3, -CCl3, -COOH, -O2SOH.

[0018] Preferably, in step (1), the five-membered heterocyclic compound is:

[0019]

[0020] Preferably, in step (1), the method for structural optimization is to import Gaussian09, and the calculation of Gibbs free energy is performed at the B3LYP / 6-31G (D, P) level.

[0021] Preferably, in step (1), the method for acquiring the descriptor data is:

[0022] The number of different types of atoms in a substituent (R) can be directly counted;

[0023] Look up the electronegativity of the atom connected to C in RC on the web [Electronegativity table URL: https: / / wikipredia.net / zh / Electronegativities_of_the_elements_(data_page)];

[0024] Draw the molecule in GaussView5, import it into Gaussian09, input "OPT FREQ" for structure optimization, and then output the OUT file. Open the OUT file in GaussView5 to query the information of bond length and bond angle.

[0025] Draw the molecule in GaussView5, import it into Gaussian09, input "OPT FREQ" to optimize the structure, and then output the OUT file. Open the OUT file in GaussView5 and click results-summary to view the dipole moment.

[0026] Draw the molecule in GaussView5, import it into Gaussian09, input "OPT FREQ" for structure optimization, and then output the OUT file. Open the OUT file in GaussView5 and click results-summary to view the electron energy.

[0027] Preferably, in step (1), the bond dissociation energy is:

[0028] AH→A · +H ·

[0029] BDE(AH)=G(A · )+G(H · )-G(AH)

[0030] Wherein, H· represents hydrogen free radical, A· represents the corresponding free radical after hydrogen dissociation, G(A·) represents the Gibbs free energy of A·, G(H·) represents the Gibbs free energy of H·, and G(AH) represents the Gibbs free energy of AH molecule.

[0031] Preferably, in step (2), the acquired data is normalized using a Python-Jupyter notebook script, and the normalization formula is:

[0032]

[0033] Among them, x i is the input data, x minis the minimum value of the input data, x max is the maximum value of the input data.

[0034] Preferably, in step (3), the model refers to a random forest model.

[0035] Preferably, in step (3), the descriptor is selected from the number of different types of atoms in different substituents R, the electronegativity of the atom connected to C in RC, the bond length and bond angle of the RC bond, the dipole moment and electronic energy of the molecule.

[0036] Preferably, in step (4), the parameters are set as n_estimators=100, random_state=42, 80% of the data is randomly divided into a training set and 20% is divided into a test set, the bond dissociation energy data is used as output, and the other data is used as input features / descriptors.

[0037] Compared with the prior art, the present invention has achieved the following technical effects:

[0038] The present invention avoids the problems of expensive experimental equipment, complicated experimental operation, long calculation time and high calculation cost of software calculation, and can be realized by just running the code, which has the advantages of short time consumption and low cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 The present invention is a flow chart of a method for predicting bond dissociation energy. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] Embodiment 1:

[0042] A method for predicting bond dissociation energy comprises the following steps:

[0043] (1) Obtaining the dataset:

[0044] The structures of five-membered heterocyclic compounds with different substituents on the ring H were drawn using GaussView5 software, imported into Gaussian09 for structural optimization, and Gibbs free energy was calculated at the B3LYP / 6-31G (D, P) level, and the bond dissociation energy was obtained to form the output data set;

[0045] Among them, the five-membered heterocyclic compound is:

[0046]

[0047] The bond dissociation energy is:

[0048] AH→A·+H ·

[0049] BDE(AH)=G(A · )+G(H · )-G(AH)

[0050] Wherein, H· represents hydrogen free radical, A· represents the free radical corresponding to hydrogen dissociation, G(A·) represents the Gibbs free energy of A·, G(H·) represents the Gibbs free energy of H·, and G(AH) represents the Gibbs free energy of AH molecule;

[0051] Count the number of different types of atoms in different substituents (R), query the electronegativity of the atoms connected to C in RC by yourself, obtain the bond length and angle of RC bonds, the dipole moment and electronic energy of the molecule through Gaussian09, and use these descriptors as input data;

[0052] (2) Normalization processing:

[0053] The acquired data is normalized using the Python-Jupyternotebook script. The normalization formula is:

[0054]

[0055] Among them, x i is the input data, x min is the minimum value of the input data, x max is the maximum value of the input data.

[0056] (3) Machine Learning Model:

[0057] A random forest machine learning model was built based on Python-Jupyter notebook, with parameters set to n_estimators = 100, random_state = 42, and the data was randomly divided into 80% training set and 20% test set, with bond dissociation energy data as output and other data as input features / descriptors;

[0058] (4) Prediction of bond dissociation energy:

[0059] The bond dissociation energy is predicted by this machine learning model.

[0060] The above description is only a preferred embodiment of the present invention and does not limit the technical scope of the present invention. Therefore, any slight modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for predicting bond dissociation energy, characterized in that The following steps are involved: (1) Obtaining the dataset: Draw the structure of H at different positions of C1 and C2 on the five-membered heterocyclic compound after different substituents are substituted, optimize the structure, and calculate the Gibbs free energy to obtain the bond dissociation energy of the C-H bond at C3 and C4 to form the output data set; count the number of different types of atoms in different substituents R, query the electronegativity of the atom connected to C in RC by itself, obtain the bond length and bond angle of RC bond, the dipole moment and electronic energy of the molecule through Gaussian09, and use these data as descriptors; (2) Normalization processing: The acquired data is processed and normalized to the range of [0, 1] by using the formula; (3) Machine Learning Model: Build a machine learning model based on Python-Jupyternotebook and select parameters; (4) Prediction of bond dissociation energy: The bond dissociation energy is predicted by this machine learning model.

2. A method for predicting bond dissociation energy according to claim 1, characterized in that: In the step (1), the structure after different substituents replace the H on the ring of the five-membered heterocyclic compound is drawn using GaussView5 software, and the different substituents include 28 substituents, namely: -CH3, -OH, -CH2OH, -NH2, -N(CH3)2, -F, -Cl, -Br, -CH2NHCH3, -NOHCH, -PH, -O － , -SH, -NHCH3, -NHCOCH3, -CHCH2, -COCH3, -COCH2CH3, -OCOCH3, -OCOCH2CH3, -CN, -NO2, -OSOCH3, -CHO, -CF3, -CCl3, -COOH, -O2SOH.

3. A method for predicting bond dissociation energy according to claim 1, characterized in that: In the step (1), the five-membered heterocyclic compound is:

4. A method for predicting bond dissociation energy according to claim 1, characterized in that: In the step (1), the method for structural optimization is to import Gaussian09, and the calculation of Gibbs free energy is performed at the B3LYP / 6-31G (D, P) level.

5. A method for predicting bond dissociation energy according to claim 1, characterized in that: In the step (1), the method for obtaining the descriptor data is: The number of different types of atoms in a substituent (R) can be directly counted; Look up the electronegativity of the atom connected to C in RC on the web [Electronegativity table URL: https: / / wikipredia.net / zh / Electronegativities_of_the_elements_(data_page)]; Draw the molecule in GaussView5, import it into Gaussian09, input "OPT FREQ" to optimize the structure, and then output the OUT file. Open the OUT file in GaussView5 to query the information of bond length and bond angle. Draw the molecule in GaussView5, import it into Gaussian09, input "OPT FREQ" to optimize the structure, and then output the OUT file. Open the OUT file in GaussView5 and click results-summary to view the dipole moment. Draw the molecule in GaussView5, import it into Gaussian09, input "OPT FREQ" to optimize the structure, and then output the OUT file. Open the OUT file in GaussView5 and click results-summary to view the electron energy.

6. A method for predicting bond dissociation energy according to claim 1, characterized in that: In the step (1), the bond dissociation energy is: A-H→A · +H · BDE(A-H)==G(A · )+G(H · )-G(A-H) Wherein, H· represents hydrogen free radical, A· represents the corresponding free radical after hydrogen dissociation, G(A·) represents the Gibbs free energy of A·, G(H·) represents the Gibbs free energy of H·, and G(AH) represents the Gibbs free energy of AH molecule.

7. A method for predicting bond dissociation energy according to claim 1, characterized in that: In step (2), the acquired data is normalized using the Python-Jupyter notebook script, and the normalization formula is: Among them, x i is the input data, x min is the minimum value of the input data, x max is the maximum value of the input data.

8. A method for predicting bond dissociation energy according to claim 1, characterized in that: In the step (3), the model refers to a random forest model.

9. A method for predicting bond dissociation energy according to claim 1, characterized in that: In step (3), the descriptor selects the number of different types of atoms in different substituents R, the electronegativity of the atom connected to C in RC, the bond length and bond angle of RC bonds, the dipole moment and electronic energy of the molecule.

10. A method for predicting bond dissociation energy according to claim 1, characterized in that: In the step (4), the parameters are set as n_estimators=100, random_state=42, 80% of the data is randomly divided into a training set and 20% is divided into a test set, the bond dissociation energy data is used as output, and the other data is used as input features / descriptors.

Images