The correlation study of the electronic and spin properties in the buckled III–V monolayers using the transfer learning network approach is performed. Band engineering based on such approach has been applied to the target parameters of bandgap without spin–orbit coupling and with spin–orbit coupling, spin splitting at K-point in either of valence and conduction bands, and band splitting at Γ-point in the valence band. It is also found that some special nonlinear combination of Group-III and V atomic numbers together with their corresponding atomic mass can accurately describe the mentioned band engineering parameters of III–V monolayers. In this regard, a nonlinear descriptor relation as a function of atomic number and atomic mass is reported for every parameter. The maximum decoupling of Eg, WOS from other compounds is observed for InN, BSb and GaN. For this parameter, heavier compounds tend to have more correlation with the other ones. A similar pattern is observed for EgWS, although the correlation for heavier compounds is more evident. In addition, more compounds demonstrate discriminatory behavior for EgWS. Furthermore, the maximum correlating parameter among III–V compounds is valence band splitting. For heavier compounds, valence and conduction spin-splitting demonstrate sensible discrimination between the band engineering predictions and DFT calculations. Interestingly, for AlSb, a clear independent behavior for conduction band spin-splitting is also reported.
Read full abstract