Quantitative Structure-Property Relationship For Partial Properties Of Ionic Liquids

Ionic Liquid(IL) is a kind of special solvents with ionic properties.Comparied toother traditional solvents, Ionic Liquids have so much advantages,especially “design-ability”.It is that within certain limits,ILs can change their partial physicochemicalproperties by changing the structure of anion,cation and ligand,so people pay moreattention to it.But ILs have much different kinds,then the number of them is very large,hence it is very different for people to try to study them.Based on the rule of the closerelationship between properties and structure,people do their bests to try to build aQuantitative Structure-Property Relationship(QSPR) model to predict the propertiesof ILs,which are able to reduce the blindness for the preparation and study of ILs.Thisdegree paper mainly build three kinds of property QSPR model of meltingpoint,viscosity and toxicity for ILs,and the melting point and viscosity mainly includeimidazolium ILs,then the toxicity mainly conclude imidazolium,pyrid-inium andquaternary ammonium ILs.Part One:One hundred and eighty-six descripters for melting point of forty-onekinds of imidazolium ionic liquids;one hundred and seventy-two descripters forviscosity of twenty kinds of imidazolium ionic liquids;one hundred and seventy-onedescripters for toxicity of twenty-one kinds of imidazolium,pyridinium andquaternary ammonium ILs,which are all respectively calculated by using ChemDrawUltra8.0、Chem3D Ultra8.0、HyperChem7.0and Dragon software.After screeningthe variables,we have a conclusions that melting point is mainly influenced by xt、DipoleLength、xy and Total Connectivity four variables,especially, the variable-xy isequivalent to linear combination of partial3D-Morse variables and the othervariable-xt is equivalent to linear combination of partial information index;Viscosityis mainly influenced by Mor04v and Mor17v two3D-Morse variables and toxicity is mainly influenced by Mor13e，Mor07m，Mor26m and Mor15p four3D-Morsevariables.Part Two:The model of the structure-activity relationship between meltingpoint,viscosity and toxicity and respectively corresponding influencedvariables,quantitative structure and property relationship(QSPR) model,are built byusing Multiple Stepwise Regressin, Back Propagation Artificial Neural Network andSupport Vector Machines three different kinds of QSPR modeling ways,there are nine3D-Morse models.The models above are well tested by internal detection concludingR value,F-test value,T-test value,variance inflation factor(VIF) and so on and theexternal detection concluding residual error plot,independent detection set andleave-one-out(LOO) detection.The results show that the model built is comparativelyreliable.Part Three:The QSPR models for melting point,viscosity and toxicity threeproperties of ILs are built well by using Multiple Stepwise Regressin, BackPropagation Artificial Neural Network and Support Vector Machines three differentways,and the vital factors of influencing melting point,viscosity and toxicity threeproperties of partial ILs are obtained.So the establishment of these models iscomparatively significant in theory and valuable in application for the preparationand study of properties of ionic liquids.And the Artificial Neural Network is the bestfor the accuracy of three models predicting partial properties,then it also havepractical significance for the intelligentized research of QuantitativeStructure-Property Relationship of ionic liquids.