Observations are presented on breakdown events during the anodic growth of tantalum pentoxide in characteristics which plot the rate of breakdown as function of oxide thickness at constant field, F . The characteristics shift to lower thicknesses for an increase in field, or a decrease in temperature. The breakdown process is interpreted by the stochastic succession of avalanche breakdown model. An approximate relation was derived for the rate of breakdown as function of insulator thickness. This relation could be well fitted to experimental data at various fields. The fitting procedure permits the evaluation of the coefficient of impact ionization α, finding for a set of anodizations α=4.4×106expfalse(−21.5/Ffalse)/normalcm with F in MV/cm, in the range of fields 6.1–6.7 MV/cm. There is an uncertainty of roughly ±10% in α, due to shallow minima in fitting procedures, to fluctuations in the rate of breakdown, and to some uncertainty in the magnitude of F .

Breakdown characteristics found during the anodic growth of aluminum oxide start with a range of rapid increase in the rate of breakdown Rnormalr . This is followed by a range of saturation, when observations are not disturbed by propagating breakdowns. The mean value of Rnormalr can be subject to some variations in the saturation range. The breakdown characteristics shift to lower voltages with increasing field of anodization. The observations agree with predictions of the succession of avalanche breakdown theory. The reason for variations of Rnormalr in the saturation range is not clear. The coefficient of impact ionization at a field of 8.7 MV/cm was found to be roughly 2×105 normalmcm . Compared with tantalum pentoxide the breakdown characteristics of aluminum oxide show a relatively poor reproducibility.

Light emission during the anodic growth of tantalum pentoxide was found to grow quasi-exponentially with electric field and oxide thickness. Data on the spectrum were obtained with six narrow bandpass filters. Largest light output was found for 4.8 eV photons. This light component was strongly absorbed in the oxide, since its bandgap is 4.6 eV. The light emission was interpreted as an electroluminescent (EL) effect due to electron avalanching and to recombination of hot electrons with trapped holes for most of the light components. The nonabsorbed light components could in part be due also to first-order kinetics EL processes. Relations derived for the EL output fitted experimental results well, giving ~ : 1.12 • 105/cm for the coefficient of impact ionization at a field of 6.26 MV/cm. Earlier breakdown investigations resulted in a = 1.4 • 105 _ 10%/cm. The difference in ~ values can be due to approximations made in deriving relations for the two phenomena. Since both breakdown and EL theories were developed with the same avalanche model, independent evidence was obtained on impact ionization in tantalum pentoxide from two phenomena.