Photocapacitance effects of deep traps in epitaxial GaAs

Abstract

A photocapacitance technique which allows rapid characterization of deep traps in a semiconductor is described. Continuous illumination with light of photon energy slightly below the band gap provides for the occupation of a trap level by both holes and electrons, and at the same time reduces the time constants associated with population changes to typically 0.1 sec. The first feature enables both hole‐emission and electron‐emission processes to be detected in a single spectrum and the second eliminates the effects of long time constants which could be masked by slow drifts. Sharp features due to individual traps are displayed by electronic differentiation with respect to energy of the photocapacitance signal. Thus we refer to the technique as double source differentiated photocapacitance (DSDP). Our data for deep levels in GaAs show that the variation of cross section with respect to energy is much more rapid than described by the frequently applied Lucovsky theory. This effect can be understood in terms of band structure throughout the reduced zone, as is more appropriately considered for very deep levels. Characteristic differences of the trap distribution for GaAs grown under different conditions are described. A series of traps less than 0.6 eV above the valence band in vapor‐phase epitaxial material are not observed in liquid‐phase epitaxial material. A strong feature representing a trap ∼0.65 eV below the conduction band mainly in vapor epitaxial material is associated with chromium. It clearly demonstrated that the transitions detected in this optical cross‐section DSDP experiment relate not precisely to the band edge, but to points higher in the bands, influenced by the detailed density of states. We find the complementary energies, of peak hole and electron optical‐emission processes in the derivative spectra of a given very deep level, total ∼1.9 eV, considerably greater than the minimum (direct) band gap of GaAs and expected value from the Lucovsky model.