(from the article " The Influence of HF Treatment of Si on Surface Electron Kinetics", S. Grabtchak, M.Cocivera, Progress in Surface Science v.50, 1995)
A sample of p-Si (3-5 ohm cm, e'=15, e"=45, where e is a dielectric constant) was irradiated by a laser pulse (337 nm, 0.6 nsec) which created an extra carrier density up to 10^20 electrons/cm3. According to previous estimations, one would expect a very complex behavior of these transients because neither the real nor the imaginary part of Eq. ( 4, 5 ) (go to the appropriate equations in The AMTMP, theory) dominate. This complication could arise also from opposite signs of these terms. The calculated changes in the real and imaginary parts of dielectric constant before and after HF treatment are shown in Figures 1 and 2, respectively. The passivated sample showed a smaller time constant for the change of the real part of dielectric constant, but no corresponding change occurred for the imaginary part. Because of the complex nature of the signal behavior, no description can be given. On the other hand, the photoresponse measured at the resonance frequency does not contain these peculiarities.
The situation simplifies for a more conductive sample having approximately e"/e' = 7. The kinetics of the change in the real and the imaginary parts before and after HF treatment are shown in Figures 3 and 4. In spite of the sufficiently large ratio, regions at the beginning of these transients had contributions from both terms according Eqn. ( 6, 7 ). The sign of the change in the imaginary part is positive, but the change in the real part is negative (because the cavity quality factor change is positive). This does not correspond to a true decrease in the real part of dielectric constant. Instead, the appropriate equations from the the theoretical section give only the effective value, which must be analyzed further to obtain a true value of dielectric constant. Such effective negative values of dielectric constants were reported earlier bu some authors for highly conductive silicon spheres. The "cross-talk" condition can only be observed at the tail of the transient (Fig.5). The same time dependence for the change of the real and imaginary parts is consisted with a model involving fast thermal equilibrium between conduction band electrons and electrons in shallow traps. Multiple trapping occurs until the conduction electrons recombine with trapped holes. The negative sign of the shift of the resonance frequency indicates that it is caused by electrons in shallow traps. Unless thermal equilibrium is established, the observed time constants cannot be the same for the cavity quality factor change and the frequency shift. In case of low conductivity materials equal time constants are observed again (see above) despite the inverse proportionality.
HF treatment caused a slight increase in the time constant from 16 to 18.5 microseconds for the real and imaginary parts. These results show that the silicon oxide interface provides subband gap states, which act more like recombination centers rather than trapping centers. When the sample is treated with HF the surface of the Si wafer is passivated. This passivation greatly reduces the surface recombination velocity apparently because the surface is covered by Si-H bonds leaving virtually no surface dangling bonds to act as recombination centers. Once the contributions from free and trapped electrons are separated by the AMTMP method, a couple of approaches can be used to separate bulk and surface components within the transient decays.
(This section is still inder construction, figures are coming soon!)