626
A. Jariwala et al.
light followed by generation of surface photovoltage, scattering and reflection of
illumination light do not interfere with the spectra.
The bandgap determination in SPS relies on the large increase in absorption coefficient near the band-edge, which is a typical characteristic of most semiconductors.
This increase in the absorption of light near band-edge causes a drastic change of the
SPV that is reflected in SPS by means of a sharp variation in slope of the SPV spectrum. An example is shown in Fig. 10, where the SPV signal, measured at our lab on
SI-GaAs wafer using the setup describer earlier, is plotted as a function of illumination photon energy. As evident from the figure, the change in the slope corresponding
to the bandgaps of GaAs is identified clearly. The bandgaps of a very large number
of semiconductors have been analyzed using this approach, either by pinpointing the
corresponding knee in the SPV spectrum [93, 94] or by looking for the maximum of
its derivative [15, 95]. In a very interesting application of this technique, Lagowski
et al. have studied the SPV response of silicon-on-sapphire films [96]. They discovered that apart from the fundamental bandedge of Silicon at 1.1 eV, an another,
distinct, knee in the SPV spectrum is observed at 1.4 eV. It was established that this
value corresponds to the onset of significant absorption in the amorphous silicon
films and thus, they could deduce a partial amorphization of the films from above
results.
It must be noted here that the obtained value of Eg is only approximate. The onset
of SPV is generally broad. Hence the exact position of Eg within it is difficult to
determine. Usually, the bandgap is underestimated by such an analysis, and the real
bandgap could, at instance, be higher by as 100 meV [24]. The disagreement between
the actual bandgap and the observed one is mainly attributed to the sub-bandgap
transitions in the SPS, which blurs the real onset of the band to band transitions.
This holds true even for the high-quality surfaces with negligible surface states, e.g.
GaAs and InP single crystals, mainly due to Franz–Keldysh effect followed by photoassisted charge transfer between one of the bands and the shallow tail states [96].
Since the SPS is very sensitive to sub-bandgap effects in the surface and buried SCR,
the tail states contribution is quite prominent and an accurate value of Eg can rarely
be obtained from the SPS curve.
The experimental comparisons between SPS and the absorption spectra show that
the two are mostly similar but never the same [97], It has been established that the
SPS mimics the absorption spectrum in cases where the injection level of excess
carriers is low, the absorption length is much larger than the diffusion length, and the
surface recombination velocity is independent of the injection level. As discussed
above, the surface states may trigger an underestimation of the bandgap in SPS.
However, it must be noted that the response of surface states is much slower than
that of the bulk in most cases. Hence, use of higher chopping frequency can easily
eliminate the contribution from surface states [98], which can improve the accuracy
of the bandgap determination.
A. Jariwala et al.
light followed by generation of surface photovoltage, scattering and reflection of
illumination light do not interfere with the spectra.
The bandgap determination in SPS relies on the large increase in absorption coefficient near the band-edge, which is a typical characteristic of most semiconductors.
This increase in the absorption of light near band-edge causes a drastic change of the
SPV that is reflected in SPS by means of a sharp variation in slope of the SPV spectrum. An example is shown in Fig. 10, where the SPV signal, measured at our lab on
SI-GaAs wafer using the setup describer earlier, is plotted as a function of illumination photon energy. As evident from the figure, the change in the slope corresponding
to the bandgaps of GaAs is identified clearly. The bandgaps of a very large number
of semiconductors have been analyzed using this approach, either by pinpointing the
corresponding knee in the SPV spectrum [93, 94] or by looking for the maximum of
its derivative [15, 95]. In a very interesting application of this technique, Lagowski
et al. have studied the SPV response of silicon-on-sapphire films [96]. They discovered that apart from the fundamental bandedge of Silicon at 1.1 eV, an another,
distinct, knee in the SPV spectrum is observed at 1.4 eV. It was established that this
value corresponds to the onset of significant absorption in the amorphous silicon
films and thus, they could deduce a partial amorphization of the films from above
results.
It must be noted here that the obtained value of Eg is only approximate. The onset
of SPV is generally broad. Hence the exact position of Eg within it is difficult to
determine. Usually, the bandgap is underestimated by such an analysis, and the real
bandgap could, at instance, be higher by as 100 meV [24]. The disagreement between
the actual bandgap and the observed one is mainly attributed to the sub-bandgap
transitions in the SPS, which blurs the real onset of the band to band transitions.
This holds true even for the high-quality surfaces with negligible surface states, e.g.
GaAs and InP single crystals, mainly due to Franz–Keldysh effect followed by photoassisted charge transfer between one of the bands and the shallow tail states [96].
Since the SPS is very sensitive to sub-bandgap effects in the surface and buried SCR,
the tail states contribution is quite prominent and an accurate value of Eg can rarely
be obtained from the SPS curve.
The experimental comparisons between SPS and the absorption spectra show that
the two are mostly similar but never the same [97], It has been established that the
SPS mimics the absorption spectrum in cases where the injection level of excess
carriers is low, the absorption length is much larger than the diffusion length, and the
surface recombination velocity is independent of the injection level. As discussed
above, the surface states may trigger an underestimation of the bandgap in SPS.
However, it must be noted that the response of surface states is much slower than
that of the bulk in most cases. Hence, use of higher chopping frequency can easily
eliminate the contribution from surface states [98], which can improve the accuracy
of the bandgap determination.