Figure 4.

Quantification of point defect species and point defect classification in VA 2H-MoTe₂ and LI 2H-MoTe₂. (a) (Top)–(Bottom) Variations in the dI/dV line spectra by STS near the point defects such as (i) Te_ad1 or V_Te1 in VA 2H-MoTe₂ and (ii) V_Mo, V_Te1+1O in LI-2H-MoTe₂, respectively. A shift in the LDOS toward the occupied (unoccupied) states as shown by purple (green) lines depict n-type (p-type) character. The legends denote the distance of the tip for the STS measurement from the defect site. The 200 °C vacuum annealing transpires a Te vacancy (Te adatom) inducing a strong n-type character. Conversely, the 532 nm-laser-illumination transpires oxygen adsorbed/chemisorbed at the Te vacancy site (V_Te) inducing strong p-type character. (b) (Top, bottom) Estimated point defect classification results by deep learning for VA 2H-MoTe₂ and LI 2H-MoTe₂ MLs, respectively. The Perfect types are not indicated, though they were classified by deep learning. Color codes are all the same as in Figures 2 and 3 for defect species. Note that V_Mo was undetected in VA 2H-MoTe₂ and Te_ad1 (Te_ad2) was undetected in LI 2H-MoTe₂. (c) (Top, Bottom) Statistical point defect classification by deep learning in VA 2H-MoTe₂ and LI 2H-MoTe₂ MLs respectively to correlate point defect-electrical property modulations. V_Te1 (red), V_Te2 (light-magenta), Te_ad1 (light-purple), and Te_ad2 (light-pink) for VA 2H-MoTe₂; V_Te1+1O (blue), V_Te2+2O (light-blue), and V_Mo (sky-blue) for LI 2H-MoTe₂.