Metal-oxide-semiconductor structures incorporating thermally grown silicon dioxide films were implanted with arsenic ions and then annealed at high temperatures. The subsequent trapping sites produced are amphoteric. Coulombic-attractive traps (for electrons) were produced with the avalanche injection of holes from the silicon substrate and the subsequent capture of some of these holes on the arsenic-related sites. During internal photoemission of electrons from a thin aluminum gate, the voltage shifts due to hole annihilation by electrons were recorded and the effective capture cross section was determined. This capture cross section was found to vary from 10-12 to 3×10-15 cm2 for average electric fields ranging from 2×105 to 3×106 V/cm. An average field threshold (1.2×106 V/cm) was found, below which the capture-cross-sectionaverage-field dependence follows a power law with an exponent of approximately -1.5. Above the average field threshold, the power-law exponent was found to be approximately -3.0. Also, when the amphoteric arsenic-related sites are empty, they can form neutral trapping sites for electrons. For these trapping centers, it is found that the neutral capture cross section is relatively independent of the average electric field. For average fields ranging from 5×105 to 6×106 V/cm, the neutral cross section is found to be approximately constant at (12)×10-15 cm2. For the Coulombic electron traps, classical and quantum-mechanical Monte Carlo simulations agree qualitatively with the experimental results. These simulations suggest that the heating of the electron-energy distribution and tunnel detrapping are the primary cause of the decrease in the effective capture cross section in the high-field regime. For the neutral traps in the low-field regime, the classical Monte Carlo simulation also agrees with the experimental results. However, for fields above the electron-heating threshold, the simulation predicts an increase in the capture cross section not found in the experimental data. We suggest that this discrepancy arises since the classical simulation does not account for tunnel detrapping, which would lower the effective cross section. © 1991 The American Physical Society.