Lattice-gas model for electron-hole coupling in disordered media V. I. Yudson Center for Chemical Physics, Department of Physics and Astronomy, University of Western Ontario, London, Ontario, Canada N6A 3K7 and Institute of Spectroscopy, Russian Academy of Sciences, Troitzk, Moscow region 142092, Russia M. R. Singh Center for Chemical Physics, Department of Physics and Astronomy, University of Western Ontario, London, Ontario, Canada N6A 3K7 Received 23 February 1998 We study an effective lattice-gas model for electron-hole coupling in disordered semiconductor structures. Despite its simplicity, the model turns out to be quite rich. It possesses several crossover regimes between phases of bound electron-hole pairs excitonsand unbound electrons and holes. It has been shown that a sufficiently strong disorder promotes dissociation of bound electron-hole pairs and may decrease considerably the range of existence of exciton gas. S0163-18299805948-7 I. INTRODUCTION There is considerable research interest in systems of ex- citons in semiconductor quantum wells QW’s. This interest is connected with the important role excitons play in light- matter interaction processes in light-emitting devices and semiconductor microcavities. Another reason for the re- search interest is related to a possibility of Bose-Einstein condensation of excitons in QW’s. 1 In this respect, special attention is attracted to spatially indirect excitons formed by spatially separated electrons eand holes ( h ). 2 Due to a strongly enhanced annihilation time of spatially indirect ex- citons, these systems are especially attractive with respect to the search for collective phenomena. There is a rapidly in- creasing amount of publications devoted to both theoretical 3–6 and experimental 7,8 investigations of indirect e -h coupling; see also references 4–6 and 8. Some experi- mental evidence has been reported 8 for a stable excitonic ground state in a strong magnetic field, which favors the stability of the excitonic phase. 9 One of the intrinsic physical problems one meets when dealing with excitons in semiconductor nanostructures is the presence of a disorder that hinders manifestation of collec- tive or coherent effects of exciton-exciton or exciton-light interaction. The disorder corresponds to structure imperfec- tions unavoidable in the course of fabrication. For thin QW’s and, especially, in the case of indirect excitons, the disorder is mainly determined by an interface roughness and a thick- ness variation of QW’s. The presence of a moderate disorder results in scattering of excitons and is responsible for rela- tively small values of the exciton diffusion coefficient. With an increase of the disorder excitons may become localized. With respect to two-dimensional 2Dsystems a more cor- rect statement is that the exciton localization length is be- coming small as compared to the system size.However, even in this ‘‘strong-localization’’ regime, typical variations V 0 of the random potential V ( r) may be still small with respect to the exciton binding energy E 0 , so that the internal structure of excitons is weakly affected by the disorder. The question addressed in this paper relates to the oppo- site case when the random potential variation V 0 is greater than the exciton binding energy E 0 . This situation may oc- cur in sufficiently thin QW’s. For instance, as small as 0.1 nm thickness variation L in a GaAs QW of an average thickness L =3 nm this corresponds to the experiment 8 would result in the variation of the electron confinement en- ergy 2 2 L /( m e L 3 ) 30 meV, which is considerably greater than the exciton binding energy. In general, electrons and holes in semiconductors are subject to different varia- tions of their potential energies, V e ( r) and V h ( r), respec- tively. Moreover, the correlation between V e ( r) and V h ( r) may be rather weak. The difference in the electron and hole random potentials influences the internal degrees of freedom of an e -h pair. Our task is to investigate under which cir- cumstances this may lead to the exciton dissociation. An exciton created at an arbitrary place is not in the most favorable—lowest potential energy—state of the e -h pair. At large as compared to the exciton binding energyand non- correlated random-potential variations there is a good chance for the electron and the hole to find lower energy positions that may be quite far from each other, so that at first glance, the e -h pair should dissociate see Fig. 1. On the other hand, for finite and finite-size correlated potentials, there is always a possibility for the e -h pair to find a position that corre- sponds to the minimum of both V e ( r) and V h ( r). Electron- hole coupling in the vicinity of this point and the formation of the exciton state leads to an additional lowering of the energy. Thus, speaking about thermodynamically equilib- rium states one might come to the opposite conclusion that the disorder may have only a little influence on the existence of excitons. However, the latter reasoning relates to a single exciton. But a single exciton cannot survive even in an or- dered macroscopic sample at thermal equilibrium with a nonzero temperature T, it will unavoidably dissociate into an electron and a hole. At T 0, only a finite-density gas of bound e -h pairs coexisting with a gas of free electrons and holes may be in the ‘‘ionization equilibrium.’’ 10 The subject of the present study is the influence of a random potential on the ionization equilibrium condition for e -h pairs. We shall restrict our consideration to the case of low electron and hole densities when we may neglect screening effects. The quantum-mechanical problem of a particle subjected PHYSICAL REVIEW B 15 DECEMBER 1998-II VOLUME 58, NUMBER 24 PRB 58 0163-1829/98/5824/162027/$15.00 16 202 ©1998 The American Physical Society