The Raman spectrum in porous GaAs layers was excited by light with hν>Eg. This approach was used in order to avoid the contribution of the substrate to the measured Raman signal. The micro-Raman spectrum of porous GaAs is presented in Figure 4.32, upper curve. The spectrum consists of a strong LO phonon at 292 cm-1, a comparatively intense TO phonon at 268 cm-1 and a weak structure in the high frequency wing of the TO phonon. According to the spectral deconvolution (Figure 4.33, upper part), the last structure represents a comparatively broad band with the maximum at 275 cm-1. One can note that this band is absent in the Raman spectrum of the as-grown GaAs (lower part in Figure 4.33, where the chain curve represents the calculated LO-phonon-plasmon mode contribution under 514.5 nm line excitation).
From the complete disappearance of the LOPC modes and the appearance of a strong LO phonon, a low carrier concentration in the porous GaAs network could be deduced. The most probable reason for this is the surface space-charge effect which results in a depletion of the skeleton, the free carriers being transferred to the substrate. According to the Schottky model the thickness of the depletion layer was estimated to be about 50 nm, which should be compared with the typical skeleton thickness of about 100 nm.
Another possible explanation of the disappearance of the LOPC modes is connected with their wavevector dependence. The minimum wavevector transferred in the light scattering process is of the order of 2π/d with d being a typical size parameter. According to the Lindhard-Mermin approach, with increasing wavevector the L- band is shifted to the TO phonon and the L+ band to the LO phonon. Finally, for small skeleton sizes, the transferred wavevector is large enough to excite single-particle excitations apart from only weakly disturbed phonons.
Let us discuss the peculiarities of the TO-phonon band in more details. The TO phonon is forbidden due to symmetry arguments in (100) backscattering geometry and, therefore, should be absent in the bulk-GaAs spectrum. The observation of a weak TO peak in the as-grown crystals seems to be caused by a small deviation from the true backscattering geometry. This small deviation, however, cannot explain the observation of an intense TO phonon in the porous layer. It is highly probable that the breakdown of the polarization selection rules may be related, at least partially, to multiple reflection of the light in the porous GaAs network. Since internal reflections occur in all directions, the information about initial scattering geometry is lost. Possible alternative reasons for the breakdown of the selection rules may be phonon confinement and/or lattice disorder . It is well known that broad Raman bands centred at 70, 180 and 250 cm-1 are inherent to heavily damaged and amorphous GaAs (see  and references therein). Since none of these bands were observed in porous GaAs, it may be considered that the porous layer is built up mainly by an undisturbed crystalline structure.
A subject of special interest is the appearance of a Raman peak at 275 cm-1, i.e. in the gap between the bulk optical phonons, in porous GaAs. According to theoretical predictions [114, 108], in the frequency region under consideration surface vibrational modes should occur in microstructures of heteropolar semiconductors. One can note, in this regard, that surface phonon modes were observed previously at 274 and 278 cm-1 in GaAs nanocrystals obtained by sequential ion implantation of As and Ga into SiO2 and &alfa;-Al2O3 matrices with subsequent thermal annealing .
So, anodic etching was found to modify considerably the micro-Raman scattering spectrum of (100)-oriented n-GaAs. The Raman spectrum of as-grown crystals is governed by the LOPC modes, while the porous layers exhibit a strong 'unscreened' LO-phonon band. Evidence for a new Raman peak centred at 275 cm-1 was found in porous GaAs and this seems to correspond to a surface-related vibrational mode.