Using a hard dead space impact ionization model, the dependence of breakdown probabilities on overbias ratio in single photon avalanche diodes is investigated theoretically in a variety of semiconductor materials for the simple case of constant electric field, that is, in a p þ -i-n þ diode structure. By using avalanche widths of 2 mm, the effects of dead space are minimized so that the breakdown probability results are determined primarily by the enabled ionization coefficients of the materials. The results illustrate how the slope of breakdown probability with overbias ratio is affected by the enabled ionization coefficients ratio and by the field dependences of ionization coefficients, which should be taken into account when choosing semiconductor materials for single photon avalanche diodes.
IntroductionSingle photon avalanche diodes (SPADs), also known as Geiger-mode avalanche photodiodes, operate at a bias above the avalanche breakdown voltage V b , so that the electron-hole pair generated by a photon absorbed in the SPAD will initiate a self-sustaining avalanche breakdown, with a finite breakdown probability, P b (V ), where V is the reverse bias voltage. When the resultant large electrical current is detected, the arrival of the photon is registered. Increasing the bias voltage and hence the electric field F, increases P b because the carriers are more likely to impact ionize and cause avalanche breakdown. The dark current of the SPAD, which increases with bias, can also trigger avalanche breakdown (dark counts), giving rise to noise and reducing the sensitivity of the SPAD. It is therefore desirable to have a large dP b /dV in order to compete with the undesirable rise in the noise with bias.Using a local impact ionization model which assumes negligible dead space d, the experimental and theoretical works of Oldham et al.[1] and McIntyre [2] showed that dP b /dV depends on the field dependence of the ionization coefficients, on their ratio and on the carrier injection conditions. The theory of breakdown probability in SPAD was later improved by McIntyre [3] to include the effects of dead space and was used by Hayat et al. [4] and Wang et al. [5] in their studies, which surprisingly did not investigate the effects of dead space. Other than the effects of ionization coefficient ratio and dead space, the slope of P b (V ) is also affected by the field dependences of ionization coefficients, which are unique to each semiconductor material. Although there were limited comparisons between