Abstract
The effects of diluted nitrogen impurities on the valence- and conduction-band states of have been predicted and measured experimentally. The calculation uses state-of-the-art atomistic modeling: we use large supercells with screened pseudopotentials and consider several random realizations of the nitrogen configurations. These calculations agree with photoluminescence excitation (PLE) measurements performed for nitrogen concentrations up to 0.035 and photon energies up to above the GaP optical-absorption edge, as well as with published ellipsometry data. In particular, a predicted nitrogen-induced buildup of the character near the valence- and conduction-band edges accounts for the surprising broad-absorption plateau observed in PLE between the and the critical points of GaP. Moreover, theory accounts quantitatively for the downward bowing of the indirect conduction-band edge and for the upward bowing of the direct transition with increasing nitrogen concentration. We review some of the controversies in the literature regarding the shifts in the conduction band with composition, and conclude that measured results at ultralow N concentration cannot be used to judge behavior at a higher concentration. In particular, we find that at the high concentrations of nitrogen studied here the conduction-band edge (CBE) is a hybridized state made from the original GaP band-edge state plus all cluster states. In this limit, the CBE plunges down in energy as the N concentration increases, in quantitative agreement with the measurements reported here. However, at ultralow nitrogen concentrations , the CBE is the nearly unperturbed host , which does not sense the nitrogen cluster levels. Thus, this state does not move energetically as nitrogen is added and stays pinned in energy, in agreement with experimental results.
- Received 28 June 2006
DOI:https://doi.org/10.1103/PhysRevB.74.155303
©2006 American Physical Society