First Principles Electronic Structure of Mn doped GaAs, GaP, and GaN semiconductors

Abstract

We present first-principles electronic structure calculations of Mn doped III-V semiconductors based on the local spin-density approximation (LSDA) as well as the self-interaction corrected local spin density method (SIC-LSD). We find that it is crucial to use a self-interaction free approach to properly describe the electronic ground state. The SIC-LSD calculations predict the proper electronic ground state configuration for Mn in GaAs, GaP, and GaN. Excellent quantitative agreement with experiment is found for magnetic moment and p-d exchange in (GaMn)As. These results allow us to validate commonly used models for magnetic semiconductors. Furthermore, we discuss the delicate problem of extracting binding energies of localized levels from density functional theory calculations. We propose three approaches to take into account final state effects to estimate the binding energies of the Mn-d levels in GaAs. We find good agreement between computed values and estimates from photoemisison experiments.

Full text

Page 1

arXiv:cond-mat/0610378v1 [cond-mat.mtrl-sci] 13 Oct 2006
First Principles Electronic Structure of Mn doped GaAs, GaP,
and GaN semiconductors
T. C. Schulthess1, W. M. Temmerman2, Z. Szotek2, A. Svane3, L. Petit1
1Computer Science and Mathematics Division and Center for Nanophase Materials Sciences,
Oak Ridge National Laboratory, Oak Ridge, Tennesee 37831-6164, USA
2Daresbury Laboratory, Daresbury, Warrington WA4 4AD, UK
3Department of Physics and Astronomy,
University of Aarhus, DK-8000 Aarhus C, Denmark
(Dated: May 13, 2010)
1

Page 2

Abstract
We present first-principles electronic structure calculations of Mn doped III-V semiconductors
based on the local spin-density approximation (LSDA) as well as the self-interaction corrected local
spin density method (SIC-LSD). We find that it is crucial to use a self-interaction free approach to
properly describe the electronic ground state. The SIC-LSD calculations predict the proper elec-
tronic ground state configuration for Mn in GaAs, GaP, and GaN. Excellent quantitative agreement
with experiment is found for magnetic moment and p-d exchange in (GaMn)As. These results allow
us to validate commonly used models for magnetic semiconductors. Furthermore, we discuss the
delicate problem of extracting binding energies of localized levels from density functional theory
calculations. We propose three approaches to take into account final state effects to estimate the
binding energies of the Mn-d levels in GaAs. We find good agreement between computed values
and estimates from photoemisison experiments.
PACS numbers: 75.30.Et, 75.30.Hx, 75.50.Pp
2

Page 3

I. INTRODUCTION
Semiconductor spintronics aims to exploit both the spin and charge of electrons in new
generations of fast, low dissipation, non-volatile integrated information storage and pro-
cessing devices. The Mn doped Ga-V semiconductors are amongst the most interesting
materials for applications in such new devices. In particular, (GaMn)As has been estab-
lished as a well-behaved mean field ferromagnet with the Curie temperature Tc linearly
dependent on the concentration of the substitutional Mn,1a magnetic moment per Mn close
to its free ion value,2and a possible low concentration of carriers further promoting Tc.3
Recent advances in thin-film growth of III-V semiconductors doped with Mn have led to
synthesis of (GaMn)As dilute magnetic semiconductor (DMS) with Curie temperatures of
the order of 173 K4. Also, recent experiments by Edmonds et al.5indicate a carrier-induced
nature of the ferromagnetic exchange, but a small, finite, density of unoccupied Mn d states
is found close to the Fermi level, reflecting hybridization with the host valence bands. Burch
et al.,6on the other hand, claim observing impurity band conduction in Ga1−xMnxAs, with
large effective masses of the carriers, so far not confirmed by other experiments.
The most interesting characteristic of DMS is the carrier induced nature of the magnetic
coupling, which has two important practical implications. First, the carriers are polarized
and DMS can serve as efficient sources for spin injection, owing to the fact that they are
structurally compatible with semiconductors used in devices, which alleviates the problems
of interfacial disorder that prohibits efficient spin-injection from traditional ferromagnets
into semiconductors. Second, because the Curie temperature is correlated with the car-
rier concentration, the magnetic order can be manipulated with voltage7. Therefore, efforts
aimed at increasing the Curie temperature of magnetic semiconductors have to be concerned
3

Page 4

with the nature of the magnetic exchange coupling in order not to loose these main advan-
tages of carrier induced magnetism. However, in order that these materials be relevant for
practical spin- and magneto-electronics8applications, their Curie temperatures have to be
raised above room temperature.
The successful description of magnetic properties in Ga1−xMnxAs motivated Dietl et
al.9to use the Zener model10,11description to predict Curie temperatures of various Mn
doped group IV, III-V, and II-VI semiconductors. In particular, their prediction of a high
Curie temperature of Mn doped GaN inspired many groups to synthesize this system, and
several reports of Curie temperatures well in excess of room temperature now exist in the
literature12. However, despite important advances made for the Ga1−xMnxAs system, the
microscopic nature of the electronic structure and magnetic exchange of Ga-V systems, and
especially (GaMn)N, is far from understood. In particular, it still needs to be established
whether the magnetism in these materials complies with the description provided by the
Zener model. More fundamentally, it is not clear to what extent the Kondo-like Hamiltonian,
describing Mn spins of S=5/2 interacting with free carriers, is a justified starting point
for describing systems other than Mn doped GaAs. For example, electron spin resonance
measurements, which indeed support the picture of divalent Mn in GaAs13, clearly favour
the trivalent d4configuration for Mn impurities in GaP14. For GaN it has also been reported
that Mn is in a divalent state when electrons are doped15, but in a trivalent state when holes
are doped to the system.16Hwang et al.,17using photoemission and soft X-ray absorption
spectroscopy, confirm that in the n-type doped GaN the Mn state is divalent, while for the
non-doped one it is trivalent.
Most model descriptions of Mn doped GaAs assume the Mn impurity to have a local-
ized moment of S = 5/2, formed by five occupied atomic-like d-orbitals (d5) that interact
4

Page 5

weakly with the host valence band through level repulsion, leading to the simple (d5+ h)
picture where the Mn spin couples antiferromagnetically to a polarized hole, h. The result-
ing Kondo-like Hamiltonian appears well justified with parameters that can be established
experimentally. However, its solution is still under debate. Dietl et al.9,18, as well as Mac
Donald and co-workers19use the Zener10,11description, in which it is assumed that the
ferromagnetic interaction between Mn ions is mediated by the induced spin-density that is
anti-aligned with the Mn moments, and the resulting prediction of the Curie temperature,
Tc, appears in agreement with experiment for Ga1−xMnxAs, with x up to 0.094.
The key question that needs to be addressed from an electronic structure point of view
is whether at concentrations of several atomic percent of substitutional Mn, the acceptor
level forms an impurity band broad enough to merge with the valence band. The electronic
structure of Mn in Ga-V semiconductor hosts is best studied with first principles elec-
tronic structure calculations. Since the Local (Spin-)Density Approximation20(L{S}DA)
to Density Functional Theory21,22(DFT) has proven very successful in predicting ground
state properties in semiconductors23as well as transition metals and their alloys24, several
groups25,26,27,28,29,30have performed LSDA calculations of the electronic structure and mag-
netism of various transition metal doped semiconductor systems soon after the discovery of
ferromagnetism in Mn doped GaAs. These studies all agree in one important point28: the
Mn d-orbitals are not atomic-like but hybridize rather strongly with the host valence band
(As-p in the case of Ga1−xMnxAs). In this, the LSDA picture differs substantially from the
one of model calculations. We have recently shown31, that first principles calculations based
on the self-interaction corrected local spin density (SIC-LSD) method, lead to a picture that
is more consistent with what is expected for a system with strongly correlated electrons. The
purpose of the present paper is to make a more in-depth comparison between the electronic
5