The optical properties of deep centers and their dependence on general materials parameters are predicted from an analytical eight-band k⋅p model of deep-center states. A wide variety of deep centers in a wide variety of direct-gap semiconductors can be modeled this way. Scanning-tunneling-microscopy images and measured optical dipoles are in excellent agreement with our model. Our model of deep-center optical properties is the most detailed, multiband k⋅p model which remains fully analytical. Our model of deep centers goes beyond previous work in being able to simultaneously explain, within an analytical framework, both the size and spectral shape of the experimentally measured cross sections for optical transitions from deep levels to (i) the valence band, and (ii) the conduction band; as well as, (iii) observed optical selection rules, and (iv) scanning-tunneling-microscopy images of deep-level bound states. Very good agreement is observed between our model and experiment for deep levels in a variety of (large and small band-gap) semiconductors: the arsenic antisite in both GaAs and In0.53Ga0.47As; the chromium substitutional impurity in both GaAs and InP; and the indium vacancy in InSb. Good agreement is achieved between our model and experiment because both the size and spectral shape of the cross sections for direct optical transitions from the deep level (to the conduction or valence-band edge) is found to be determined by the small-wave-vector component of the deep-center wave function. It is precisely the small-wave-vector component of the deep-center wave function which is described well by our eight-band k⋅p model. Significantly, this agreement between our model and experiment is a vindication of the general materials parameters (Kane dipole, nonparabolic effective masses, band-gap energy, spin-orbit splitting) characterizing our eight-band model of deep centers, rather than a result of careful use of adjustable parameters. Our model shows that the spatial extent of the deep-center bound state is proportional to the Kane dipole, and is thus larger (more delocalized) in a smaller band-gap semiconductor. Moreover, our model shows that, in order to successfully predict optical properties, a linear combination of atomic orbitals describing deep centers must extend over many lattice sites: more than just the neighbors and next-nearest neighbors of the deep center.
Read full abstract