Ab initio investigation of potential energy curves of He2, He2+, and extrapolation by the machine learning method.

Diatomic molecules He2+$$ \mathrm{H}{\mathrm{e}}_2^{+} $$, He2$$ \mathrm{H}{\mathrm{e}}_2 $$ are common compositions in plasmas using the helium as the working gas, and play key roles in quenching O2 and N2. However, the potential energy curve (PEC) of the first state He2a3Σu+$$ \mathrm{H}{\mathrm{e}}_2\left({a}^3{\Sigma}_u^{+}\right) $$ obtained at the CASSCF level of theory is not accuracy and data of higher excited states of He2+$$ \mathrm{H}{\mathrm{e}}_2^{+} $$ are lacking. Here, we calculated PECs of four excited states of He2$$ \mathrm{H}{\mathrm{e}}_2 $$ and nine lowest electronic states of He2+$$ \mathrm{H}{\mathrm{e}}_2^{+} $$ by employing CASSCF+MRCI/Aug‐cc‐pV5Z and CCSD(T)/Aug‐cc‐pV5Z level of theory with sample internuclear distance interval of 0.005 Å. PECs were extrapolated to the complete basis sets. There are double local maxima in PECs of three excited states He2+B2Σu+$$ \mathrm{H}{\mathrm{e}}_2^{+}\left({B}^2{\Sigma}_u^{+}\right) $$, He2+C2Πu$$ \mathrm{H}{\mathrm{e}}_2^{+}\left({C}^2{\Pi}_u\right) $$, He2+D2Πg$$ \mathrm{H}{\mathrm{e}}_2^{+}\left({D}^2{\Pi}_g\right) $$. The machine learning method based on the Gaussian process was used to interpolate and extrapolate the PECs, and then vibrational wave functions are solved numerically. This ab initio investigation provided rotational‐vibrational constants and new data for experimental research in the future. [ABSTRACT FROM AUTHOR]