2026 IEEE 38TH INTERNATIONAL CONFERENCE ON MICROELECTRONIC TEST STRUCTURES, ICMTS
Abstract
Capacitance–Voltage (C–V) measurement is an important but slow device testing technique. We introduce an active learning methodology to reduce the required number of measurement needed by using Gaussian Process Regression to predict unmeasured values. The proposed method results in a 4x reduction in the number of needed measurements while generating almost perfectly identical curves with a 0.24% capacitance equivalent thickness error, and 0.9977 R2 score.