Nondestructive characterization for crystalline wafers

A method of analyzing crystalline defects comprising:

• providing a crystalline wafer that comprises a first face and a second face that opposes the first face;
• accessing nondestructive data captured from at least a portion of the first face;
• and
• inferring one or more predicted etch features that would be present in destructive data captured from the second face, wherein the one or more predicted etch features correspond to defects in the crystalline wafer.

A method of analyzing crystalline defects comprising:

• providing a silicon carbide (SiC) wafer that comprises a silicon face and a carbon face;
• accessing photoluminescence data captured from at least a portion of the silicon face;
• detecting one or more defects from the photoluminescence data;
• and
• identifying the presence or the absence of one or more threading edge dislocations in the one or more defects based on destructive data captured from other SiC wafers, wherein the destructive data are obtained using destructive characterization steps.

A method of analyzing silicon carbide (SiC) wafers comprising:

• capturing nondestructive data from at least a portion of a SiC wafer;
• etching one or more surfaces of the SiC wafer to delineate etch defects;
• correlating the delineated etch defects with the nondestructive data to form a defect map identifying one or more defects in the nondestructive data according to one or more defect categories based on the delineated etch defects;
• coupling the defect map to train a deep neural network;
• and
• accessing nondestructive data of other SiC wafers with the deep neural network to detect and identify one or more defects in the nondestructive data of the other SiC wafers according to the one or more defect categories.

A method of analyzing silicon carbide (SiC) wafers comprising:

• capturing nondestructive data from at least a portion of a SiC wafer;
• forming one or more epitaxial layers on the SiC wafer;
• determining one or more characteristics of the one or more epitaxial layers;
• correlating the one or more characteristics with the nondestructive data to form a characteristic map;
• coupling the characteristic map to train a deep neural network;
• and
• accessing nondestructive data of other SiC wafers with the deep neural network to infer one or more characteristics of one or more epitaxial layers that would be formed on the other SiC wafers.