Symmetric Versus Nonsymmetric Structure of the Phosphorus Vacancy on InP(110)

Abstract

The atomic and electronic structure of positively charged P vacancies on InP(110) surfaces is determined by combining scanning tunneling microscopy, photoelectron spectroscopy, and densityfunctional theory calculations. The vacancy exhibits a nonsymmetric rebonded atomic configuration with a charge transfer level 0.75 ± 0.1 eV above the valence band maximum. The scanning tunneling microscopy (STM) images show only a time average of two degenerate geometries, due to a thermal flip motion between the mirror configurations. This leads to an apparently symmetric STM image, although the ground state atomic structure is nonsymmetric. Copyright 2000 by the American Physical Society. 73.20.Hb, 61.16.Ch, 71.15.Mb, Although it is well known that point defects can exert a profound influence on the physical properties of semiconductors, the determination of the atomic scale geometric and electronic structure of point defects has remained a challenging task for theoretical as well as experimental research. One particularly striking example is anion vacancies on (110) surfaces of III-V semiconductors, where no agreement has been reached regarding even basic properties, such as (i) the symmetry of the atomic structure and (ii) the energy of defect levels in the band gap: Although recent density-functional theory (DFT) calculations for positively charged arsenic vacancies on the GaAs (110) surface agreed that the gallium atoms neighboring the vacancy relax into the surface layer (in contrast with an earlier tight-binding calculation [1]), one calculation found a rebonded configuration breaking the mirror symmetry of the surface [2], whereas the other predicted a fully symmetric configuration to have the lowest energy [3]. High resolution scanning tunneling microscopy (STM) images show a density of states preserving the mirror symmetry of the surface at the defect site [1,4]. Although this seems to favor a symmetric atomic structure, the experimental results could not be matched to results of any of the DFT calculations [5][6][7][8]. Furthermore, scanning tunneling spectroscopy (STS) yielded a local downward band bending of 0.1 eV [1], whereas surface photovoltage measurements found a band bending of 0.53 ± 0.3 eV [9] at the site of the positively charged As vacancy on pdoped GaAs(110) surfaces. Concerning the energy levels, the two different DFT calculations predicted the charge transfer levels (+/0) to be 0.32 eV [2] and 0.1 eV [3], and the lowest Kohn-Sham eigenvalues in the band gap to be 0.73 eV [2] and 0.06 eV [3] above the valence band maximum (VBM). In view of this puzzling situation it is obvious that the interpretation of the experimental STM images as well as the structure of the anion vacancies on (110) surfaces of III-V semiconductors are still under debate [5][6][7][8].Discrepancies such as those pointed out above can arise due to limitations of the methods used. On the theoretical side DFT calculations have proven to be powerful to determine the structure of point defects [10]. Yet differences in the size of the s...

Quasiparticle Corrections to the Electronic Properties of Anion Vacancies at GaAs(110) and InP(110)

“…Our LDA results are similar to those reported previously [3,8] and well below the experimentally deduced value of 0:75 0:1 eV [3]. The G 0 W 0 approximation, on the other hand, yields a value for =0 that lies within the experimental error bar.…”

“…(2) and the corresponding formula for 0=ÿ within the LDA. Quantitatively, our calculated value of 0.47 eV for =0 in InP exhibits the same systematic underestimation of the experimentally derived chargetransition level as earlier studies at this level of approximation [3,8].…”

“…Besides, aspects like the self-interaction are treated inappropriately. As a consequence, the band gaps of semiconductors and the energies of localized defect states are not given correctly: for the P vacancy at InP(110) the experimentally determined =0 level of 0:75 0:1 eV [3] contrasts noticeably with the calculated values 0.52 eV [3] and 0.388 eV [8]. The variation between the two theoretical results can be traced to differences in the pseudopotentials and parallels the variation of the corresponding band gaps.…”

“…Experimental investigations are thwarted by the fact that many traditional spectroscopic techniques are not appli-cable due to the low concentration of native point defects, while capacitance methods like deep-level transient spectroscopy [2] are very sensitive but provide no elemental or structural information to identify the type of defect. For surfaces, at least, Ebert et al [3] now demonstrated how the electronic structure of individual defects can be deduced using a combination of scanning tunneling microscopy (STM) and photoelectron spectroscopy, providing the first reliable experimental analysis of a charge-transition level for the P vacancy at InP(110). Theoretical approaches, on the other hand, must include accurate exchange-correlation contributions, the coupling between electronic and lattice degrees of freedom, and, in general, require the treatment of open systems in which the number of particles is not constant [4].…”

Mapping the wave function of transition metal acceptor states in the GaAs surface

“…Larger biases cause the switching to occur more rapidly resulting in noise in the topography around the feature, similar to that observed due to switching behavior in a variety of other systems with STM. 27,28 At large biases the switching occurs quickly enough that only the average is observed, 29 as in the insets of Fig. 2.…”