Defect & Dopant Predictions for Thermoelectric Materials
Experimental training sets concerning dopability have
been assembled and serve as the basis for machine learning models
to predict the maximum dopability range in materials (left Figure). We
have achieved predictive dopability, with carrier concentration
predicted within one order of magnitude.
Rather than relying on experimental literature and our
on-going experimental efforts to build a training set, a similar
database has been constructed using DFT calculations of defects in
model systems. We have validated the defect calculations against
Looking forward, we will combine these two training sets to
provide greater predictive accuracy.