Temperature dependence of electronic defect in semiconductors: How ab initio calculations can complement PAC experiments.

L.A. ERRICO, M. RENTERÍA, A. G. BIBILONI, G.N. DARRIBA.

La Habana

Simposio; Simposio Latinoamericano de Física del Estado Sólido (SLAFES 2004); 2004

SLAFES

The electronic charge density (r) in a solid and its temperature dependence can be studied by measuring a physical magnitude - the electric-field gradient tensor (EFG) – that is very sensitive to small changes in (r). The Time-Differential Perturbed-Angular-Correlation technique (PAC) is specially suited to study the temperature dependence of the EFG because its sensitivity is temperature independent. In metallic systems, the temperature dependence of the EFG is relatively well understood in terms of lattice vibrations. On the other hand, in the case of semiconductors, several behaviours have been observed, which have been explained (with more or less success) in terms of host properties or processes induced by the presence of the hyperfine probes (generally impurities in the host). In particular, the wide gap semiconductors that crystallize in the cubic structure of the mineral bixbyite Mn2O3 have been subject of systematic studies with 111 Cd tracers [1]. In these compounds, the presence of dynamic hyperfine interactions, originated in the electron-capture decay after-effects, and a strong positive linear temperature dependence of the EFG appear selectively, depending on the fact that if the host cations present closed electronic shells (such as Sc, Y, and In) or if it has incomplete electronic shells (as is the case of the 4f-orbitals of the rare-earths) [1].Lutetium is the only rare-earth that presents a closed-shell electronic structure in its 3+ oxidation state and its sesquioxide does not undergo a phase transition like others in the group, converting this oxide into an interesting “laboratory” to check the models proposed to explain the phenomena mention above. The experimentally observed anomalous “step-like” EFG temperature dependence at 111 Cd sites in Lu2O3 was explained in the frame of a “two-state” model that considers an extremely fast fluctuation between two static EFG configurations [2], which enabled the experimental determination of an acceptor energy level introduced by the Cd impurity in the band-gap of the semiconductor and the estimation of the oxygen vacancy density in the sample.For a long time, point-charge model in combination with Sternheimer factors has been used to interpret experimental EFG results in solids through the assignment of some effective charges to the ions of the crystal. These calculations, however, depend on empirical parameters and often deviate significantly from the experimental values. To interpret the experimental EFG and extract the electronic structure a system, ab initio calculations which do not need external parameters (like the Sternheimer ones), are indispensable. About 15 years ago, Blaha et al. have developed such first-principles calculations using the Full-Potential Linearized Augmented Plane Wave (FLAPW) method, which is one of the most accurate methods for band-structurecalculations of solids in the framework of the Density Functional Theory. Recently, increasing computer power and progress in method developing have made it possible to applied the FLAPW method to fairly complicated systems like doped semiconductors [3] and, in particular, to oxide semiconductors with diluted impurities [4], enabling the description of structural andelectronic properties on these complex. More recently, we have showed that an ab initio calculation can be used to explain the temperature dependence of the EFG at Cd impurities in TiO2 [5]. In the present communication we have applied the FLAPW method to the case of Lu2O3 doped with Cd. Our results show that we can reproduce not only the experimental results at RT but also the temperature dependence of the EFG. In this new insight, the EFG thermal dependence arises from the ionization of an impurity acceptor level introduced in the band-gap of the semiconductor, in good agreement with the previously proposed two-statemodel.[1] J. Shitu et al., Mod.. Phys. Lett. B 12, 281 (1998).[2] L. Errico et al., Hyp. Interact. 120/121, 457 (1999).[3] S. Lany et al., PRB 62, R2259 (2000).[4] L. A. Errico et al., PRL 89, 55503 (2002).[5] L. A. Errico, Hyp. Interact., in press (2004).This work was partially supported by CONICET, Fundación Antorchas, and ANPCyT (PICT98 03-03727), Argentina,and TWAS, Italy. M.R. is member of CONICET. L.A.E and G.N.D are fellows of CONICET.