THE CERTIFIED EFFICIENCY OF MOST COMMERCIALLY PRODUCED CRYSTALLINE SI (C SI) SOLAR CELLS UNDER STANDARD TEST CONDITIONS DOES NOT REFLECT THE EFFICIENCY IN THE FIELD OVER THE LIFETIME OF THE PANELS DUE TO PERFORMANCE DEGRADATION FROM LIGHT-INDUCED DEFECTS. TO ELIMINATE LIGHT-INDUCED DEGRADATION (LID) IN P-TYPE PASSIVATED EMITTER REAR CONTACT (P-PERC) SOLAR CELLS, OVER THE LAST THREE YEARS, THE C-SI PHOTOVOLTAICS (PV) INDUSTRY RAPIDLY TRANSITIONED FROM B- TO GA-DOPED CZ SI. HOWEVER, GA-DOPED P-PERC CELLS STILL UNDERGO LIGHT- AND ELEVATED-TEMPERATURE-INDUCED DEGRADATION (LETID). UNLIKE LID, REGENERATION IN LETID PROCEEDS VERY SLOWLY, AND FIELD REGENERATION CAN TAKE UP TO 25 YEARS—ESSENTIALLY THE LIFE OF THE MODULE. THEREFORE, THE HIGH EFFICIENCY AND RELIABILITY OF P-PERC MODULES CAN ONLY BE FULLY REALIZED IF LETID CAN BE MITIGATED DURING MANUFACTURING PRIOR TO DEPLOYMENT. IN THIS PROJECT, WE WILL STUDY LETID USING GA-DOPED CZ SI WAFERS OBTAINED FROM AN INDUSTRY PARTNER. WE WILL PERFORM A SYSTEMATIC EXPERIMENTAL AND COMPUTATIONAL EXPLORATION OF HOW THE EXTENT OF LETID IS INFLUENCED BY THE INITIAL VACANCY CONCENTRATIONS IN THE PRESENCE OF H. WE WILL EMPLOY THERMAL ANNEALING EXPERIMENTS WITH CONTROLLED WAFER BOUNDARY CONDITIONS, PERFORM MEASUREMENTS OF CARRIER LIFETIME, AND USE ELECTRON PARAMAGNETIC RESONANCE (EPR) TO MEASURE DEFECT CONCENTRATIONS, AND THEN GENERATE A DATABASE FOR INFORMING A PREDICTIVE PHYSICS-INFORMED NEURAL NETWORKS (PINNS) MODEL. THE END-OF-PROJECT GOAL IS TO USE THIS MODEL TO PREDICT THE MOST PROMISING INDUSTRIALLY-VIABLE TECHNOLOGY FOR LETID SUPPRESSION IN GA-DOPED CZ SI, AND VERIFY THAT IN 90% OF THE EXPERIMENTS, LETID IS REDUCED BY AT LEAST 50%.