Investigation of the Statistical Spread of the Time-Dependent Dielectric Breakdown in Polymeric Dielectrics for Galvanic Isolation

We present a study of the statistical spread of the Time-Dependent Dielectric Breakdown (TDDB) in thick (>10 µm) polymeric dielectrics for galvanic isolation devices. By performing Monte Carlo simulations based on a thermochemical percolative model, we demonstrate, first of all, that in the case of a homogeneous dielectric the intrinsic TDDB spread arising from the statistics of local degradation of the material and percolative conduction across its entire thickness is negligible with respect to what typically observed experimentally. The experimental TDDB spread is, then, reproduced in the modeling framework by introducing inhomogeneities in the local material properties, giving rise to additional variability in the degradation dynamics leading to device breakdown. This approach is, finally, shown to be capable to explain the dependence of the TDDB spread on the magnitude of the electric field stressing the device.