The Design Of Germanene-based Materials For Photovoltaic And Photocatalytic Applications Based On Machine Learning

High-throughput computing of two-dimensional（2D）materials proved helpful in searching for potential candidates from the mass of new materials,which assists the research of advanced materials and meet the performance requirements of semiconductor devices from stability and photoelectric characteristics.Only relying on the traditional simulation,the high throughput calculation often needs a lot of computing resources and an extended research cycle.In exploring the photoelectric properties for two-dimensional Ge₈H_nX_8-n（n=0-8,X=F,Cl,Br,I）materials,a material design method for photovoltaic and photocatalysis applications based on machine learning is studied in this paper.The machine learning（ML）method efficiently completes the high-precision prediction for stability and photoelectric properties of Ge₈H_nX_8-n（n=0-8,X=F,Cl,Br,I）with fewer experiment costs.Finally,13 kinds of Ge₈H_nX_8-n（n=0-8,X=F,Cl,Br,I）materials and germanene-based heterojunctions with high photoelectric conversion efficiency（PCE）have been successfully designed.Through the experimental research in this paper,the main conclusions are as follows:On the one hand,a compound model combining random forest classification（RTC）,ridge regression（RR）,support vector regression（SVR）,and fitting method can effectively utilize 33 kinds of simple features（such as concentration,atomic radius,etc.）,which achieve the high precision prediction of stability（binding energy）and electrical properties（band-type,band gap,minimum conduction bottom）for 2D Ge₈H_nX_8-n（n=0-8,X=F,Cl,Br,I）materials.In the composite model,the accuracy coefficient（AUC）of RTC algorithm,the regression correlation coefficient（R²）of the RR algorithm and SVR algorithm reached 0.96 and 0.97,respectively.Meanwhile,three photovoltaic Ge₈H_nX_8-n（n=0-8,X=F,Cl,Br,I）materials,14 Ge₈H_nX_8-n（n=0-8,X=F,Cl,Br,I）materials with potential photocatalytic applications and three high PCE germanene-based heterojunctions were successfully obtained according to predicted results of the composite model.On the other hand,17 Ge₈H_nX_8-n（n=0-8,X=F,Cl,Br,I）materials with photovoltaic or photocatalytic application and three high PCE germanene-based heterojunctions show that the predicted results of the composite model can be a high agreement with the physical properties obtained by density functional theory（DFT）.DFT verification proves the high accuracy of the ML composite model.According to the DFT results,Ge₈H₆F₂（1000＿1000 and 2000＿0000）and Ge₈H₃FCl₂ have high stability,a suitable band gap,a higher light absorption coefficient,and optical anisotropy compared with the prototype material（germanane,Ge H）.The three 2D Ge₈H_nX_8-n（n=0-8,X=F,Cl,Br,I）materials are particularly suitable for photovoltaic applications.For the selected new materials of photocatalytic applications,the band properties of Ge₈H₇F,Ge₈H₇Cl,Ge₈H₇Br,Ge₈H₇I,Ge₈H₆Br₂,Ge₈H₆Br I,and Ge₈H₆I₂ are consistent with the redox reaction conditions of water decomposition,which have a prospect of photocatalytic applications.At the same time,three potential germanene-based heterojunctions by DFT simulation were found to have high PCE.Ge₈H₆Cl Br/Ge₈H₅Br₂ has the highest PCE（23.24%）among the three heterojunctions and a strong light absorption coefficient compared with the monolayer materials.Finally,molecular dynamics simulation and phonon spectrum calculation also both proved that the selected Ge₈H_nX_8-n（n=0-8,X=F,Cl,Br,I）with the potential application has good thermodynamic and dynamics stability.