Solvent Design Method Based On Reaction Kinetics

The reaction solvent plays an important role in organic synthesis and is able to regulate the reaction rate constant of liquid-liquid homogeneous organic synthesis by solvation effects.The optimization method in process systems engineering has been recently applicated to computer-aided molecular design for reaction solvents,which is able to enhance the reaction kinetic rate and selectivity,improve the yield and purity of products,reduce or eliminate byproducts from the source,and realize clean production for organic synthesis.However,reaction solvent design faces many challenges.On the one hand,there is a lack of reaction kinetic model with the characteristics of strong extension,high accuracy,and low redundancy,which makes it hard to quantitatively control the reaction rate constant considering solvation effects.On the other hand,it is urgent to consider the reaction rate constant and other solvent properties such as the green property simultaneously in order to meet the requirements of clean production for organic synthesis.In this thesis,the solvent design method is developed based on reaction kinetics.The mechanism of solvation effects on the reaction rate constant is investigated,and the reaction kinetic model is successfully established to quantitatively regulate the reaction rate constant.The quantitative structure-property relationships between molecular structures and solvent properties are established to realize the high-throughput optimization-based design for reaction solvents.This thesis also develops a synthesis path design method and integrates it with the reaction solvent design method through a step-by-step integration strategy.The main research contents and results of this thesis are as follows:(1)Due to the lack of reaction kinetic model with the characteristics of strong extension,high accuracy,and low redundancy considering solvation effects,a hybrid modeling method for the prediction of reaction rate constant is proposed to construct the semi-empirical reaction kinetic model.The method consists of three steps,including conventional transition state theory derivation,additional descriptor selection,and model identification,which improve the extension,accuracy,and redundancy of the kinetic model,respectively.Two case studies from literatures are used to test the proposed hybrid modeling method,and the same solvent descriptors are identified in two semi-empirical reaction kinetic models with great regression results,which proves the extension and accuracy of the established kinetic model,and verifies the feasibility and effectiveness of the hybrid modeling method.The semi-empirical reaction kinetic model lays a foundation for the subsequent reaction solvent design.(2)The infinite dilution activity coefficient in the semi-empirical reaction kinetic model is fast calculated with the molecular surface charge density distribution.However,the charge density distribution is calculated by the density functional theory costly,which does not meet the speed requirements for high-throughput reaction solvent design.To solve this problem,a machine learning-based atom contribution method is proposed to greatly improve the calculation speed of charge density distribution by sacrificing minor precision.The weighted atom-centered symmetry functions are used to describe the three-dimensional local atomic environment.Based on this descriptor,a high-dimensional neural network modeling method is used for nonlinear fitting to construct the quantitative structure-property relationship between the weighted atom-centered symmetry functions and the atomic surface charge density distribution,which achieves the goal of high-throughput and accurate predictions for molecular surface charge density distribution.Compared with the density functional theory,the calculation speed of atom contribution method is improved by about three orders of magnitude.Compared with other accelerating methods for the density functional theory(e.g.,group contribution method),the atom contribution method gives a smaller prediction error and can identify isomers.This method provides the technical support for high-throughput predictions of infinite dilution activity coefficients in the semi-empirical reaction kinetic model.(3)In order to solve the problem that the existing quantitative reaction solvent design methods only involve a limited number of solvent properties(e.g.,reaction rate constant)and have not simultaneously consider other solvent properties such as the green property,a generic framework for molecular and mixture product design is proposed to design reaction solvents considering the reaction rate constant and other solvent properties(e.g.,toxicity)simultaneously.The framework consists of preliminary design,CAMD,as well as product evaluation and verification.The optimization-based mixed-integer nonlinear programming model for molecular and mixture design is established,and the decomposition-based algorithm and two-step algorithm are used to solve the solution issue caused by the strong nonlinear equations in the optimization model.One case study for crystallization solvent design illustrates that the proposed framework is universal,while the other one shows that the design for reaction solvent is feasible and effective.This method provides the theoretical guidance and technical support for high-throughput and intelligent reaction solvent design.(4)For the organic synthesis path with multi-step elementary reactions,there are still some limitations in synthesis path though the optimal design for reaction solvents has been considered,such as small reaction equilibrium constants,poor green and safety properties for raw materials and intermediates,the unavailable raw materials,and excessive synthesis steps.In order to solve these problems,an organic reverse synthesis path design method integrating reaction solvents is proposed.First,a multi-objective optimization design framework based on the reaction template and thermodynamics-decision tree model is constructed to high-throughput design the optimal synthesis path within the constraints of reaction thermodynamics.Furthermore,the reaction solvent design method is integrated with the path design method through a step-bystep strategy to optimize and evaluate the apparent reaction rate constants for top candidate paths.This method lays a theoretical foundation for the design of thermodynamically feasible,kinetically efficient,as well as green,safe,and low-consumption organic synthesis paths.