Design Of Novel Two Dimensional MXenes Hydrogen Evolution Catalyst By Descriptors Based On High Throughput Computing And Machine Learning

Hydrogen evolution reaction(HER)electrocatalyst plays a key role in electrochemistry and energy conversion technology.Development of suitable HER electrocatalyst is the first step in the industrialization of HER electrocatalysis process.Catalytic descriptors can connect catalysts structures with properties,quickly screen new materials and optimize the performance of existing catalysts.As a new type of HER catalyst,two-dimensional MXenes have attracted much attention in both experiment and theoretical calculation.However,the design of catalytic performance and construction of catalytic descriptors for two-dimensional MXenes materials with single atom doping and ordered alloys are relatively lacking.In this work,we focused on HER catalytic properties of complex two-dimensional MXenes and explored potential excellent HER catalysts by integrating DFT high-throughput computing and machine learning methods.The descriptors were used to predict the catalytic activity and reveal the origin mechanism of catalytic activity at the electronic structure level.The specific research contents are as follows:(1)14 kinds of two-dimensional nonmetal single atom doped(NM)-Ti2CO2 systems were systematically selected by DFT calculation.Firstly,the stability and catalytic activity of NM-Ti2CO2 were studied.Secondly,the doped NM roles and types of catalytic regulation were analyzed.Finally,the intrinsic relationship between the electronic structure and catalytic activity of two-dimensional NM-Ti2CO2 was discussed,and the valence electron number and charge transfer coupling descriptors were constructed,which can accurately predict HER catalytic activity.(2)Taking two-dimensional titanium carbide with oxygen terminal as an example,the performance of single atom catalyst was studied by first principle calculation.Firstly,the catalytic activity(?GH),conductivity and stability of 3d,4d and 5d single transition metal atom(STM)doped with Ti2CO2 were screened.Several Ti2CO2-STM materials with good catalytic activity for HER were obtained theoretically.Then,the effect of STM doping on HER catalytic activity at the electronic structure level was studied.Finally,the descriptors containing Fermi level and local structure information were constructed by machine learning method.This descriptor had clear physical meaning and can accurately reveal the difference of HER catalytic activity at different active sites.(3)By combining DFT calculation with machine learning,we screened out potential two-dimensional MXene ordered binary alloys(OBAs)and revealed the key factors affecting the catalytic performance of two-dimensional MXene-OBAs.Based on Sabatier principle,we performed high-throughput DFT calculations and selected 188 ideal HER catalysts from 420 two-dimensional MXene-OBAs databases with stable O-terminal.Machine learning revealed the relationship between the oxygen and surface metal atoms bond length(dM-O),the distance between nearest neighbor oxygen atoms(dO-O),the ionization energy difference(IEmd)and the average affinity energy(EAmm)of alloying elements,and the valence electron(XV)of X=C or N with the catalytic activity of two-dimensional MXene-OBAs.Finally,the proposed machine learning descriptor was analyzed by electronic hierarchy process.The above work not only provides data reference for the experimental synthesis of various complex two-dimensional MXenes materials,but also provides theoretical guidance for the design of more two-dimensional MXens HER catalysts by using descriptors with clear physical meaning.The integration of high-throughput computing with machine learning and descriptors is expected to be an effective way to efficiently design HER catalysts in the future.