Descriptor Of Alloy Formation Energy And Molecule Activation Energy On Metal Surfaces

In the field of catalytic chemistry,it has been a great topic for long to design catalysts rationally,that is,to understand mechanisms,predict performance and screen high-quality material,which is of significance both scientifically and industrially.With the booming of scientific computing data,a new paradigm activated by data science has developed,through which researchers can extract trends and laws in a large number of databases.The machine learning techniques,especially the descriptor methods,provide us new knowledge for traditional projects and opportunities to filter catalyst working well.Based on chemistry databases and descriptor methods mentioned above,the alloy formation energy and the activation energy on pure metal surfaces are studied.The main content of the thesis is listed below.(1)The descriptor model of AB2 alloy formation energy is established through mining experimental data,owning lower errors of both fitting and predicting when comparing with Miedema model and its derivated ones.According to sensitivity analysis,the charge difference ΔZ between A and B plays a key role in most of the B-based alloy formation energies.In the meantime,at least one of the atomic volume V of metal A and volume difference ΔV can act notably,while the formation energies between A and B,as well as electronegativity of A contributes weakly to the alloy formation energy.Further discussion reveals the complexity of this problem,and offers prediction of more experimental formation energies of AB2 alloy.(2)The activation energy descriptor model is constructed after identifying traditional linear BEP relation.In the BP-not-applicable dissociation reaction of COOH*(to CO2*and H*),the descriptor can fit well and predict accurately.Likewise,the descriptor fitted by methanol dissociation data varying slightly gives good performance in other testing samples,which is superior to BEP relation.Furthermore,the surface energy and the work function are discussed using correlation.This work provides a descriptor model of activation energy and a routine to predict it,as well as new insights into the topic.