The effects of disorder and of final-state electron-hole interactions on acceptor-to-band absorption and photoluminescence spectra of semiconductors are treated theoretically within the effective-mass approximation. The results of the calculations are used to analyze photoluminescence spectra from lightly doped GaAs: Cd measured by Williams and Bebb. It is shown that although a proper one-electron calculation of optical absorption ${\ensuremath{\epsilon}}_{2}(\ensuremath{\omega})$ should include the effects of electron-hole spatial correlations, the excitonic Coulomb final-state interaction exactly cancels the spatial correlations in the case of infinite hole mass. Therefore the usual one-electron approximation, which neglects both spatial and Coulombic correlations, is valid for small electron-hole mass ratios (${m}_{e}\ensuremath{\ll}{m}_{h}$); for GaAs: Cd, the finite-mass-ratio corrections are nearly independent of energy and can be neglected. The major portion of the discrepancies between the Williams-Bebb data and Eagles's one-electron theory cannot be attributed to final-state interactions, but can be attributed to small internal microfields of strength ${10}^{3}$ to ${10}^{4}$ V/cm generated by disorder. The sources of these microfields are most likely to be piezoelectric phonons, but ionized impurities or excitons could possibly generate such fields. It is shown that photoluminescence lines are particularly sensitive to disorder and that the band-to-acceptor luminescence edge $L(\ensuremath{\omega})$ should not be a precisely exponential function of photon energy $\ensuremath{\hbar}\ensuremath{\omega}$. Instead the microfield model predicts that the edge shape should be approximately given by ln $L(\ensuremath{\omega})\ensuremath{\propto}{(\ensuremath{\hbar}{\ensuremath{\omega}}_{0}\ensuremath{-}\ensuremath{\hbar}\ensuremath{\omega})}^{\ensuremath{\alpha}}$, with $1<\ensuremath{\alpha}<\frac{3}{2}$. The existing data are compatible with $\ensuremath{\alpha}=\frac{3}{2}$, but the possibility of $\ensuremath{\alpha}=1$ cannot be ruled out without additional photoluminescence measurements further down the edge. If such measurements indicate that $\ensuremath{\alpha}$ is precisely equal to unity, the electric-microfield theory of exponential absorption edges may require revision.
Read full abstract