Tunneling and ionization phenomena in GaAs pin diodes

Ionization coefficients in AlxGa1-xAs (x= 0 - 0.60)

“…To remove the possibility of leakage currents affecting the results, the incident light was chopped and the resulting ac signal was detected with a lock-in amplifier as a function of reverse bias. This technique was especially important in the ultra-thin ( 0.1 µm) GaAs P + -I-N + s where there are high tunnelling currents, in agreement with other measurements of the dark current in such devices by Milledge et al (1995) and Liebig et al (1992), for example. The measured photocurrent data were normalized for the magnitude of the injected photocurrent, including a correction for the bias dependence of the collection efficiency, to give M e and M h .…”

“…When the i-region thickness, w, was reduced below 0.5 µm, α and β were reduced, especially at low fields. Several works including Liebig et al (1992) and Di Carlo et al (2000) have also modelled this non-local behaviour of α and β: using a Monte Carlo (MC) model, Liebig et al (1992) reproduced the measured dark current characteristics of various sub-micrometre structures and extracted how α and β change with position. They showed that the dead space is a significant fraction (≈50%) of the depletion region thickness for a 200 Å thick P-+ I-N + .…”

“…In recent years therefore, there has been a general consensus in the literature that non-local effects are very important for the accurate calculation of multiplication in sub-micrometre devices but that such analysis is complex. This is reflected in the fact that several works, including the aforementioned one by Liebig et al (1992), illustrate nonlocal behaviour but make no attempt to quantify it. However, in contrast to these beliefs, we recently showed in Plimmer et al (2000a) that the local models can replicate experimental multiplication characteristics reasonably well even in GaAs P + -I-N + s as thin as 0.3 µm or in P + -N junctions doped to 2.2 × 10 17 cm −3 when a simple dead-space correction of the type used by Bulman et al (1985) was employed with their α and β data.…”

Hierarchy of simulation approaches for hot carrier transport in deep submicron devices

“…Statistical methods to reduce the variance of averaged quantities are useful in eliminating a large part of the noise, thus providing a way to attempt an estimate of rare hot carrier events (e.g. injection into the oxide of a MOSFET) [42][43][44]. It could be argued that a Monte Carlo approach may simulate naturally occurring noise in devices, and therefore a well set up Monte Carlo model could yield more realistic results for a single device than the actual Boltzmann equation, where the results are implied to be an average over a large number of similar devices.…”