Transcript
Page 1: Transport in the metallic regime of Mn-doped III-V semiconductors

Transport in the metallic regime of Mn-doped III-V semiconductors

Louis-François Arsenault,* B. Movaghar,† P. Desjardins, and A. YelonDépartement de Génie Physique and Regroupement Québécois sur les Matériaux de Pointe (RQMP), École Polytechnique de Montréal,

Case Postale 6079, Succursale “Centre-Ville,” Montréal, Québec, Canada, H3C 3A7Received 7 August 2007; revised manuscript received 8 January 2008; published 26 March 2008

The standard model of Mn doping in GaAs is subjected to a coherent potential approximation CPAtreatment. Transport coefficients are evaluated within the linear response Kubo formalism. Both normal andanomalous contributions to the Hall effect are examined. We use a simple model density of states to describethe undoped valence band. The CPA band structure evolves into a spin split band caused by the p-d exchangescattering with Mn dopants. This gives rise to a strong magnetoresistance, which decreases sharply withtemperature. The temperature T dependence of the resistance is due to spin-disorder scattering increasingwith T, CPA band-structure renormalization, and charged impurity scattering decreasing with T. The calcu-lated transport coefficients are discussed in relation to the experiment, with a view of assessing the overalltrends and deciding whether the model describes the right physics. This does indeed appear to be the case,bearing in mind that the hopping limit needs to be treated separately, as it cannot be described within the bandCPA.

DOI: 10.1103/PhysRevB.77.115211 PACS numbers: 75.50.Pp, 72.10.d, 72.25.Dc, 75.47.m

I. INTRODUCTION

Magnets, which can be made using semiconductors bydoping with magnetic impurities, are potentially very impor-tant new materials because they are expected to keep some ofthe useful properties of the host system. One important ques-tion is how much of the “doped semiconductor” band struc-ture does the system still have? Is it a completely new alloyor does the system still behave like a doped semiconductor,which can, for example, be described using Kane theory, astaught in standard textbooks.1 Many of these questions havebeen seriously investigated in extensive recent reviews.2–5

The objective of this paper is to present a simple and mostlyanalytically tractable description of the magnetism and trans-port properties of GaMnAs. We need a theory which capturesthe essential physics, can be parametrized, and can be ap-plied to magnetic nanostructures without having to carry outab initio calculations each time some parameter has changed.Previous theories of GaMnAs and related materials havebeen either computational or semianalytical. However, thetreatment has often been split into three parts: magnetism,resistivity, and magnetotransport have been treated sepa-rately. Here, we show that the magnetism and transport canbe treated on the same footing using a one-band model. Theone-band model is of course not enough to explain the opti-cal properties. However, we show that magnetism, spin, andcharged impurity scattering can be formulated in the sameframework, using the coherent potential approximationCPA and a generalization thereof. In this paper, we focus,however, only on the magnetism and spin scattering aspectsand leave the computation of the charged impurity scatteringand the full quantitative comparison to a later paper. One ofthe most difficult and controversial aspects for all conductingmagnets is the origin of the anomalous Hall effect AHE.This is why we have devoted a section to remind the readerof the basic definitions and results. The full history of theAHE is given in Ref. 3.

We focus on GaMnAs as a much-studied prototype. It isnow generally accepted that the observed ferromagnetic or-

der of the localized spins in Mn-doped GaAs is due to Mnimpurities acting as acceptor sites, which generate holes.These are antiferromagnetically coupled to the local 5 /2 Mnspins, lowering their energy when they move in a sea ofaligned moments. It is the configuration of lowest energybecause spin-down band electrons move in attractive poten-tials of spin-up Mn and vice versa. Given that the spins arerandomly distributed in space, the magnetic state is also thestate of maximum “possible order.” Thus, free and localizedspins are intimately coupled. These materials exhibit strongnegative magnetoresistance because aligning the spins re-duces the disorder, and this lowers the resistance. The degreeto which the magnetization of the spins affects the scatteringprocess is dependent on the degree to which the magneticspin scattering is rate determining for resistance. We shall, inthis paper, only include the spin scattering process in order tofirst gain an intuitive understanding of the processes in-volved.

The “hole mediated” magnetism point of view is notshared by everyone. Mahdavian and Zunger6 argue that themobile hole induced magnetism cannot explain the magne-tism in high band gap, strongly bound, hole materials such asMn-doped GaN. Their model, based on first principles super-cell calculations, predicts that the Mn-induced hole takes ona more d-like character as the host band gap increases, mak-ing the magnetism a d-p coupling property.

The material of this paper is structured as follows. Wefirst present a short review of the phenomenology of the Halleffect in magnets. The Hamiltonian which describes theproperties of Mn-doped semiconductors is then introduced.We then discuss, in more detail, how to formulate transportin magnetically doped semiconductors. Following that, werecall how one can calculate the self-consistent CPA self-energy caused by the spin dependent term in the generallyaccepted Hamiltonian, for a one-band system. Once the self-energy is known, we compute the longitudinal and transverseconductivities using the Kubo formulas. When the Fermilevel is just above the mobility edge of the hole band, theKubo formula is still valid and describes the strong-

PHYSICAL REVIEW B 77, 115211 2008

1098-0121/2008/7711/11521113 ©2008 The American Physical Society115211-1

Page 2: Transport in the metallic regime of Mn-doped III-V semiconductors

scattering random phase limit. The localized hopping regimeis treated in another paper.7 The CPA equations only need thedensity of states of the active free hole band as input. We willtherefore consider the semicircular Hubbard band which re-produces the correct free hole band edges and is mathemati-cally convenient for illustrating the effects of band inducedmagnetism.

II. HALL EFFECT

The experimentally measured Hall coefficient RH is oftenwritten as

RH = RN + RA,

RA = axx + bxx2 M/Bz, 1

where a, b are constants and xx is the resistivity. The firstterm, RN, is the normal Hall coefficient and scales with re-sistivity in the usual way and the second, RA, is the anoma-lous term, which, in general, can have two components, onelinear and the other quadratic with resistivity2 and is propor-tional to the magnetization M. The general relation for theHall coefficient is

RH =RexyBz

BzRexxBz2 , 2

where xy and xx denote the transverse and normal conduc-tivities, respectively, and Bz is the magnetic field.

Karplus and Luttinger8 pointed out that the B field in-volves the magnetic moment of the material via the internalmagnetization M,

Bz = 0Hzext + 1 − NM Bz

ext + 01 − NM , 3

where N is the demagnetizing factor. Thus, the magnetizationterm is implicit in the normal contribution as a shift in themagnetic field. However, this form is not normally sufficientto explain the much larger magnetization contribution ob-served in ferromagnets.8 Throughout, we shall, for simplic-ity, suppose a thin film with perpendicular-to-plane magneticfield and thus take N=1, unless otherwise mentioned.

III. MICROSCOPIC DESCRIPTION

A. Hamiltonian

In this section, we formulate the basic model. The Hamil-tonian for this Mn-doped “alloy,” which most workers in thefield accept as the correct description,2 is given by

Htot = Hp + Hloc + Hd + Hpd + Hp−B + Hd−B + Hso. 4

In Eq. 4, the first term describes the free band holes. In thelocal Wannier or tight-binding representation, it is

Hp = m,n,s

tmncms† cns, 5

with tmn denoting the overlap or jump terms from sites m ton, cns

† , cns are creation and annihilation operators for a carrierof spin s at site n, and where the indices include multiband

transfers, if any. In the presence of a magnetic field, theoverlap has a Peierls phase, so that we write it as assumingthin film geometry, no internal field effect for a field perpen-dicular to plane

tmn = tmn0 e−ie/2Bext·RnRm, 6

where ext stands for external, e is the electronic charge, andRn the position vector of site n. The second term in Eq. 4describes the diagonal atomic orbital energy that the va-lence band p hole experiences when it is sitting on an impu-rity site which may or may not be magnetic,

Hloc = m,s

Em,cms† cms, 7

where stands for magnetic M or nonmagnetic NM.The third term is the direct exchange coupling between

the Mn d-localized spins, which we neglect here becausewhen mobile carriers are present, they mediate the exchangecoupling and in the case of a low concentration, the Mn spinsshould be, on average, far away from each other. The fourthterm is the antiferromagnetic spin exchange coupling be-tween the valence p hole and the local d-Mn and ultimatelythe reason for magnetism in these materials,9–11

Hpd =Jpd

2 mNS

s,s

cms† ss · Smcms, 8

where ss is a vector containing Pauli’s matrices, i.e.,x ,y ,z. The indices s and s indicate which terms of the22 matrix we are considering. Finally, Sm is the Mn spinoperator at site m. The fifth and sixth terms are the Zeemanenergies of the holes and Mn spins in an external magneticfield along the z axis, Bz

ext.

Hp−B = −g*

2B

m

s

Bz,mext cms

† ssz cms, 9

Hd−B = − 2B mNs

Bz,mext Sm

z , 10

where g* is the effective g factor. Finally, we have the spin-orbit coupling

Hso = m,n

s,s

s,mhson,scms† cns, 11

where

hso =

4m2c2 · Vr p 12

and where Vr is the total potential acting at a point r andwill include both normal crystal host sites, and impuritysites, and p is the momentum operator. Finally, c is the speedof light. In the tight-binding representation, disorder is nor-mally in the diagonal energies. For the spin-orbit term, dis-order will enter the Hamiltonian through variations in the sitepotential.

ARSENAULT et al. PHYSICAL REVIEW B 77, 115211 2008

115211-2

Page 3: Transport in the metallic regime of Mn-doped III-V semiconductors

B. Spin-orbit coupling

It is generally believed that the spin-orbit coupling is thecause of the anomalous Hall effect. This is one of the mostfascinating and universal observations made in conductingmagnets. In this section, we briefly discuss the basic phe-nomenology of spin-orbit coupling. It is useful to recall thebasic premises.

When the Bloch wave function ansatz is inserted into theHamiltonian, the spin-orbit coupling produces two newterms which enter the Hamiltonian for the periodic part ofthe wave function unk nkr=unkeik·r,

hso =

4m2c2 Vr p · + Vr k · .

13

The first is the usual term and enters the band-structure cal-culation. The second is called the Rashba term12 and in acrystal is, in general, less important than the first because inthe nucleus, momenta are larger than the lattice momenta k.

The “Rashba term” is not normally treated in the compu-tation of unk. However, in a lattice, it may be enhanced. Inthe Kane model, Chazalviel13 and De Andrada e Silva et al.14

have shown how to handle the spin-orbit terms in a latticeand the Rashba term in the presence of a triangular potentialproduced by a gate in an inversion layer. In particular, Ref.14 has shown how to renormalize the coefficient of theRashba term and that this term acts when inversion symme-try is broken in an external field or triangular potential, forexample. The origin of the enhancement of both these spin-orbit effects can perhaps be understood as similar to the ori-gin of the change of the effective mass in the lattice. Thespin-orbit energy can be enhanced by the presence of thelattice, but the enhancement can be very different from caseto case, and needs to be examined in each case separately.For example, naively speaking, one may consider the spin-orbit force acting on a particle moving in a slowly varyingpotential of longer range than the lattice spacing as a spincurrent of an effective mass electron. One can think of thescattered particle as having an effective mass m* and gener-ating a spin-orbit interaction which scales as

4m*c2instead

of

4mc2 . When the particle is scattered from an atomic sizeimpurity in a lattice, it is forced to come back many timesand spends a longer time on the impurity than in free space.Leaving aside these intuitive pictures, one can, in any case,make the analysis quantitative using the Kane Hamiltonian,which takes the lattice modulation into account in a kind ofrenormalized perturbation theory and gives explicit resultsfor the unk part of the wave function.1 The Kane Hamiltoniancan then be used to treat scattering from an impurity. Thus,for example, the matrix elements of the position operatorkrk are very different for plane wave states and for theKane solutions.8,13

In tight binding TB, the spin-orbit matrix elements arecalculated using atomic orbitals, with two terms: the intra-and the interatomic contributions. The magnitude of theintra-atomic term is just what it would be for the correspond-ing orbitals on the atoms in question. The interatomic term is

normally small and not included in tight-binding band-structure calculations. When an electron jumps from site tosite, it experiences a net magnetic field produced via thecross product of its velocity and the electric field gradient ofthe neighboring ions. This field then couples to its spin.Since it is related to the intersite momentum or velocity, itdepends on the transfer rate tmn where the indices m, n in-clude a band index m=m ,. Remember that the velocity xdirection operator is given by

vx =i

m,n,sRm,nxtmncms

† cns, 14

where Rm is the position vector of site m and Rm,n=Rm−Rn. An enhancement of this spin-orbit energy, if any, has tobe calculated by taking the expectation value of the spin-orbit term using the calculated Bloch states as one would forthe dispersion relation ks and for the effective masses. Weshall examine the spin-orbit coupling in more detail below.

IV. KUBO TRANSPORT EQUATIONS

In order to compute the transport properties of the mag-netically doped semiconductors in the linear response re-gime, we need to introduce the Kubo formulas. In this sec-tion, we show how to compute the transport coefficients.

A. General considerations

The conductivity in linear response to an electric field isusually written as15–17

=ie2

lim →0

,

vv − + + i

f − f −

,

15

where the v are the velocity operators in the respectivedirection . For a given Hamiltonian, they are derived fromthe Heisenberg relation iv= x ,H with x the positionoperator. and are the exact wave functions and energylevels. The wave functions include any disorder and mag-netic and spin-orbit couplings. f is the Fermi function and is the frequency of the applied electric field; e is the elec-tronic charge. This formula can now be written in the par-ticular representation selected, e.g., TB or Bloch states.Within the band-structure approach for semiconductors, onewould substitute the eight-band k ·p wave functions and en-ergies into Eq. 15, include spin splitting through a Weissfield and spin-orbit coupling, and treat disorder as a lifetimecontribution in the energy levels. In this picture, the dopedsemiconductor retains the pure band-structure features, apartfrom a complex lifetime shift. We shall use the one-bandCPA for the coupling of the holes to the magnetic impurities.For the calculation of the ordinary transport coefficients, wemay neglect the spin-orbit coupling. We obtain, for the lon-gitudinal conductivity of the one-band TB model for =0and per spin s,

TRANSPORT IN THE METALLIC REGIME OF Mn-DOPED… PHYSICAL REVIEW B 77, 115211 2008

115211-3

Page 4: Transport in the metallic regime of Mn-doped III-V semiconductors

xxs =

e2

dE −

fEE

dXImGsE,2,

16

where

GsE, =1

E − − s, 17

s = sR − isI, 18

and

X =1

N

kvx

2 − k . 19

sE is the CPA self-energy for spin state s, to be calculatedself-consistently using the CPA condition and where the ve-locity and effective masses of the free electron band aregiven by

v =1

k

k

, 20

M−1 =

1

2

2k

kk

, 21

where M is the effective mass tensor. The CPA self-energycontains the information on the net spin splitting producedby the magnetic impurities and the scattering time. It definesthe effective medium which is to be discussed below.

The Hall conductivity, given by the antisymmetric part ofthe transverse conductivity,18,19 is somewhat more compli-cated because the magnetic field has to be dealt with first. Tofirst order in magnetic field and with one TB band, we findfor the antisymmetric part referred to by the index a for=0,15,17,18

xya =

2e32B

3

s dE −

fEE

dY

ImGsE,3, 22

where

Y 1

N

k vx

2

Myy+

vy2

Mxx−

2vxvy

Mxy − k . 23

In Eq. 22, the dispersion relation k can be any one-bandstructure any choice of tmn

0 where k is defined throughtmn0 = 1

Nkeik·Rmnk. As was shown by Matsubara and

Kaneyoshi,18 for a weak magnetic field, the Peierls phase canbe factored out of the Green function and this is why we onlyneed the Green function Eq. 17 that does not include theorbital effect of the field. The effects of the different phasesfrom the tmn and from the Green functions, after an expan-sion to first order in the field, are reflected in the topologydependent term Y. Furthermore, Eq. 22 does not containthe spin-orbit scattering contribution. It can be included as anextrinsic effect. In the spirit of the skew scattering Bornapproximation19–21 and in the weak scattering limit, we can,

at the end of the calculation, introduce an extra internal mag-netic field so that the total field entering Eq. 22 is

B = Bz +m*

e 1

sz , 24

where Bz includes the internal magnetization. However, fromEq. 3, for a field perpendicular to plane, N=1. This result,Eq. 24, is written in the notation of Ballentine.19 The effec-tive spin-orbit field depends on the thermally averaged spinpolarization z, with its sign to be discussed in Sec. IV B,and 1

s is the so-called skew scattering rate. The second

term of Eq. 24 involves, in addition to the sign of thecarrier, the sign of the spin orientation.

Engel et al.21 claimed that the impurity spin-orbit cou-pling in the lattice environment can be “6 orders of magni-tude” larger than in vacuum. The large enhancement canmean that, in some cases, the skew scattering dominateswhere there should exist a substantial intrinsic contributionas well.21,22 This is why developing a way to calculate theextrinsic contributions to the anomalous Hall effect, as we dohere, is important and useful.

B. Anomalous Hall effect

Let us now show how one can derive such an anomalousHall effect, in our formalism TB approximation, from theexistence of a spin-orbit coupling and how one can arrive atthe concept of an effective magnetic field, as given inEq. 24.

In tight binding, the spin-orbit coupling is given by Eq.11 and Vr is the total potential experienced by the chargeat point r, with p the momentum operator. The spin-orbitcoupling can be separated into an “intra-” and an “inter-atomic” contribution by splitting the momentum operator asp=pa+pinter. For spherically symmetric potentials, the twoterms are given by

Vso = i

irli · + n

r − Rnnr pinter · ,

25

nr =

4m2c2Vnr − Rnr − Rn

, 26

where nr is related to the spin-orbit coupling strength.Here, pinter is the usual zero-field interatomic momentum op-erator and li is the atomic orbital operator and is oftenquenched so that its expectation value is zero, but nondiago-nal same site interorbital matrix elements can be veryimportant.23–25

The second contribution in Eq. 25 is due to the inter-atomic motion and must be evaluated using pinter=mv, wherev is given by Eq. 14 and involves the intersite transferenergy t and position operator

r = m,s

Rmcms† cms. 27

Here, we have a spin-orbit coupling only because the particlecan jump to another orbital. When substituting the site rep-

ARSENAULT et al. PHYSICAL REVIEW B 77, 115211 2008

115211-4

Page 5: Transport in the metallic regime of Mn-doped III-V semiconductors

resentation in the second term of Eq. 25 and assumingspherical symmetric potentials, either impurity with effectivelocal charge eZn=eZimp or host eZh, we have

iVso2j =

4mc2in,l eZn

40r − Rn3r − Rnl i

Rljtlj · , 28

where i is the imaginary number, while i is the orbital atsite i. The analysis of this term is not trivial and depends onthe lattice and potential distribution in question. If one hasdecided that the extrinsic contributions are dominant, thenone only needs to sum around the impurities. In general, thesimplest approach is to take only the largest terms in thesum, i.e., the diagonal terms l= i. In this case, we get

iVso2j = iijtij ,

iij =

4mc2in eZn

40r − Rn3r − Rni i

Rij · . 29

It is convenient to treat this effect as a spin dependent phasein the transfer, in analogy to the Peierls phase, to first order.Indeed, to first order ex1+x, so that we can write iij=eiij −1 and Hp+Hso as

Hp + Hso = i,j

,

tijeiijci† cj. 30

Then, we also note that the Rn=Ri and Rn=R j terms in Eq.29 are zero after the vector product, leaving the gradientfield contributions due to the nearest neighbors as the largestof the remaining terms in the n sum. A similar result has beenderived in the context of the Rashba coupling by Damkeret al.26

With Eq. 29, we can rewrite the spin-orbit matrix ele-ment as

hijso = i

e

2tij

n,ni,jR j Rin · Bn,so, 31

where Rin=Ri−Rn and where we have introduced a spin-orbit magnetic field defined by

Bn,soi → j =

2mc2i Zn

40Ri − Rn3i . 32

In summary, let us write in this approximation a new totalphase for Eq. 6,

ij =e

2 − Bz · R j Ri + n,ni,j

R j Rin · Bn,so .

33

We thus see what the spin-orbit coupling does to the TBHamiltonian. It is, in terms of energy, a small effect. Itseffect on the band structure and magnetism will be neglected.

In order to obtain Eq. 22 with B defined as Eq. 24, onehas to decouple the trace over the spin polarization from theremaining expression and this automatically makes z theoverall mobile spin polarization. In reality, the skew scatter-ing is due to carriers near the Fermi level and therefore zshould be the average weighted at the Fermi level. We willreturn to this point when we attempt to fit Eqs. 22 and 24to the experiment in Sec. VI B.

C. Transport coefficients for a simple cubic band

To perform a calculation, one needs to specify a latticetopology tmn

0 . The simplest one is a simple cubic latticewith nearest neighbor hopping. The dispersion relation isgiven, for a d-dimensional system, by k=−2t=1

d coska,where a is the lattice constant of the simple cubic. With thisparticular form, we can show27 that the longitudinal and Hallconductivities at zero frequency can be written, using Eqs.16 and 22, with no approximations, as

xx =e2

dad−2s −

fEE

ImGsE,2

− zD0zdzddE , 34

xya =

4e3B

3dd − 1ad−42s −

fEE

ImGsE,3

− zD0zdzddE ,

35

where GsE , is given by Eq. 17, d is the number ofdimensions, a stands once again for antisymmetric, andD0z is the density of states of the pure crystal. In three-dimensions, D0z can be found knowing that, for the abovek, for a three-dimensional cubic lattice, the Green functionis28

G0E =1

22t

0

dxKx , 36

where Kx is the complete elliptic integral of the first kind,with x= 4t

E+i 2t cos .Now, we can return to the calculation of the new bands

generated by the hole-spin coupling, and for this, we use theCPA. Later, we use Eq. 24 to estimate the magnitude of theskew scattering Hall effect.

V. COHERENT POTENTIAL APPROXIMATIONEQUATIONS FOR A ONE-BAND MODEL WITH

COUPLING TO LOCAL SPINS

In this section, we show how to compute the new energybands in the presence of a high concentration of dopants,magnetic and nonmagnetic. The magnetic coupling betweenthe Mn spin and charge and the holes in the valence bandcreates a new valence band structure. This new Mn-induced

TRANSPORT IN THE METALLIC REGIME OF Mn-DOPED… PHYSICAL REVIEW B 77, 115211 2008

115211-5

Page 6: Transport in the metallic regime of Mn-doped III-V semiconductors

band will be magnetic at low enough temperatures. Since theMn are randomly distributed, we cannot use Bloch’s theo-rem. One way forward is to use the powerful self-consistentsingle site approximation known as the CPA. The CPAself-energy29 E is determined by the condition that if at asingle site, the effective medium is replaced by the true me-dium, then the configurationally averaged t matrix producedby scattering from the difference between the true mediumand effective medium potentials must vanish.29

In the present case, we have localized Mn spins of mag-nitude S=5 /2 so that there are six possible states Sz=−5 /2, . . . ,5 /2 and thus seven possible sites in the systemincluding the nonmagnetic ones. We have for a “spin-up”carrier the normal nonmagnetic sites with concentration 1−x with local site energy ENM and magnetic field interaction− g*

2BBz,m

ext and the magnetic sites with concentration x with

local site energy EM and magnetic couplingJpd

2 Smz

− g*

2BBz,m

ext . However, there is also the possibility of a spinflip of both the carrier and the impurity with interactionJpd

2 Sm+ . For a “spin-down” carrier, we have ENM at a nonmag-

netic site and magnetic field interaction g*

2BBz,m

ext . For a mag-netic site, we still have EM as the local site energy but themagnetic coupling is now −

Jpd

2 Smz + g*

2BBz,m

ext and the spin-flip

interaction is given byJpd

2 Sm− . The concentration of spin car-

rying impurities is defined by x and isNS

NL, where NS is the

number of sites with an impurity, while NL is the total num-ber of sites in the system. The CPA conditions are givenby27,30

1 − xtm↑↑NM th + xtm↑↑

M th = 0,

1 − xtm↓↓NM th + xtm↓↓

M th = 0, 37

and the thermal average of an operator is

OmSzth =TrOmSze−HS

Tre−HS

=

Sz

OSze−hmSzSz

Sz

e−hmSzSz

Sz

OSzPSz . 38

In order to find the spin dependent t-matrix operator for theCPA, it is useful to rewrite the Hamiltonian as a 22 matrixin spin space. The t↑↑ and t↓↓ operators are the diagonal ele-ment of the t matrix. The expressions for the t matrix areessentially the same as the ones obtained previously by Ta-kahashi and Mitsui.30 The minor difference is that in thepresent case, we include the magnetic field, so that the Zee-man term enters the t operator. The inclusion of the Zeemanterm is straightforward: one has to add, when N=1, g*

2BBz,m

ext

with the appropriate sign in the expressions of Ref. 30. Tocalculate the CPA t matrices, we need the effective mediumGreen’s function which is no longer dependent on the site m.

GsE GmmsE =1

Nk

1

E − ks − sE. 39

In the previous equation, GijsE refers to the Green’s func-tion of a carrier of spin s between sites i and j for oneparticular configuration of disorder. The average is over allpossible realization of disorder.

The local density of states is

DsE = −1

ImGmmsE , 40

and the hole-spin concentration with spin s at site m can bewritten as

pms = dEfEDsE . 41

Assuming mean field theory for the energy entering the Bolt-zmann factor, the probability that the local spin has a valueSz is

PSz =e−hmSz

Sz

e−hmSz , 42

where

hm =Jpd

2pm↑ − pm↓ − 2BBz,m

ext . 43

To compute the CPA self-energy sE, we need to specifythe lattice dispersion of the hole band. In effect, it suffices tospecify the bare pure lattice density of states D0E sincethe real and imaginary parts are simply related. The Fermilevel is determined by the condition that the total hole con-centration p is known and fixed relative to the total impurityconcentration with x as a maximum value. In general, therewill be fewer holes than dopants, but the number is notknown and remains a fit parameter, so we have

p = dEfED↑E + D↓E . 44

This now allows us to compute the CPA self-energy self-consistently and then to determine the new density of statesvia Eq. 40 and the local spin polarization and mobile spinconcentration via Eqs. 38 and 41. Thus, we can determinethe magnetization and the transport coefficients, the resistiv-ity and the conductivity as a function of B via Eq. 34, andthe magnetoresistance as the relative change with B. Finally,the normal and anomalous Hall RH can be obtained viaEq. 35.

VI. APPLICATIONS OF THE COHERENT POTENTIALAPPROXIMATION

For numerical calculations, to simplify the analysis andfor proof of principle, instead of the D0E found with Eq.36, we use the Hubbard function, given by

ARSENAULT et al. PHYSICAL REVIEW B 77, 115211 2008

115211-6

Page 7: Transport in the metallic regime of Mn-doped III-V semiconductors

D03DE =

2W − EW2

W2 − E2, 45

where W is half the bandwidth and is the Heaviside stepfunction. Equation 45 has the same band edge behavior asthe D0E calculated with Eq. 36. With the simple form ofEq. 45, the first integration in Eqs. 34 and 35 is analyti-cally tractable.

A. dc conduction and magnetoresistivity

Here, we calculate the dc conductivity xx from Eq. 34and the magnetoresistance. The calculations have been doneassuming N=1 thin film. The calculations for the magne-toresistance have been performed without taking into ac-count the spin-orbit interactions.

Figures 1 and 2 are plots of the CPA density of states fortwo values of p and Jpd assuming that the scattering is solelydue to the spin potential, that is, when EM =ENM =0. Thediagrams shows the evolution of the spin dependent densityof states as a function of temperature for four different tem-peratures in each figure. The temperature is measured inunits of the bandwidth W. In Figs. 1d and 2d, one can nolonger see the spin splitting. The position of the Fermi levelis also shown. Note that all energy parameters are normal-ized by W E=E /W. From Figs. 1 and 2, one can see thatwhen the magnetic coupling Jpd is very small, the split banddisappears see Ref. 27 for details at low Jpd, whereas in theopposite limit, the spin bands split off and a pseudogap ap-pears.

Figure 3 shows the resistivity in zero B field as a functionof temperature with p as a parameter. Results for two valuesof Jpd are presented. The results for lower Jpd than 0.35W aresimilar to Fig. 3a.27 When the impurity bands completelysplit from the valence band for example, for Jpd=0.5W, thequalitative behavior changes, as there are now two well de-fined bands, above and below the Fermi level impurity andvalence separated by a gap see Fig. 2 above and Ref. 27.At high hole concentration p0.8x, the system is metallicand resistance increases with spin disorder, at first rapidly,and then decreases again above Tc. This is intuitively to beexpected because at first, as temperature increases, the disor-der increases, and then in the paramagnetic phase, thermalbroadening overcomes the potential scattering disorder andthe lifetime averages out. We should remember that in thesimplest Boltzmann approach the conductivity is given by

xx = e2s DsE −

fEE

vs2EsEdE 46

and at low temperatures, depending on the product of thescattering time and the density of states at the Fermi level. Inan alloy, either quantity can change with B and T and deter-mine the conductivity. Above Tc, the spin splitting disappearsand the quantities involved are, in the absence of chargedimpurity scattering, only weak functions of temperature. Atvery low hole concentration, the Fermi level is in a region ofsmall density of states where we expect localization. How-ever, CPA is a mean field method and does not produce lo-calization. Even though the resistance increases with de-

−1.5 −1 −0.5 0 0.5 1 1.50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

E’

D(E

’)

−1.5 −1 −0.5 0 0.5 1 1.50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

E’

D(E

’)

−1.5 −1 −0.5 0 0.5 1 1.50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

E’

D(E

’)

−1.5 −1 −0.5 0 0.5 1 1.50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

E’

D(E

’)

(b)

(a)

(c)

(d)

FIG. 1. Density of states DOS for different temperatures, akBT=0, b kBT=510−3W, c kBT=7.9910−3W, and d kBT=8.110−3W, when x=0.053, ENM =EM =0, BB=0, Jpd=0.4W,and p=0.8x. The full line curve is the DOS for the spin-down car-rier, the dashed one is the DOS for the spin up, and the vertical lineshows the Fermi level. One can see that the density of states is,relative to a nonmagnetic system, a strong function of temperature.

TRANSPORT IN THE METALLIC REGIME OF Mn-DOPED… PHYSICAL REVIEW B 77, 115211 2008

115211-7

Page 8: Transport in the metallic regime of Mn-doped III-V semiconductors

creasing density of states at the Fermi level, the conductionprocess continues to be a band conduction albeit with a shortrelaxation time. Even if the mean free path reaches the latticespacing random phase limit, the conductivity is still fargreater than hopping conductivity between localized levels.31

The effect of temperature on resistivity above Tc is not toosignificant in the nonlocalized intermediate cases. The tem-perature dependence of the resistance is, in general, a com-plex interplay of density of states, velocity, and relaxationtime imaginary part of the self-energy. The temperatureinduced lowering of the charged impurity screening lengthsee Eqs. 49–51 is not included here. In this range ofexchange coupling, Jpd does not seem to strongly influencethe structure of the resistivity versus T curves but doeschange their magnitude. In Fig. 3, for p=0.1x, a low carrierconcentration, one may see that the resistance increases withJpd. In contrast, in the high hole concentration limit p=x,there is only a relatively weak variation of resistance behav-ior with Jpd. The complex but regular behavior of resistancewith temperature is a manifestation of Eq. 46, reflecting thedifferent elements which determine the value of resistancefor a given set of parameters.

−1.5 −1 −0.5 0 0.5 1 1.50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

E’

D(E

’)

−1.5 −1 −0.5 0 0.5 1 1.50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

E’

D(E

’)

−1.5 −1 −0.5 0 0.5 1 1.50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

E’

D(E

’)

−1.5 −1 −0.5 0 0.5 1 1.50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

E’

D(E

’)

(b)

(a)

(c)

(d)

FIG. 2. Density of states DOS for different temperatures, akBT=0, b kBT=2.510−3W, c kBT=4.510−3W, and d kBT=5.410−3W, when x=0.053, ENM =EM =0, BB=0, Jpd=0.5W,and p=0.3x. The solid line curve is the DOS for the spin-downcarrier, the dashed one is the DOS for the spin up, and the verticalline shows the Fermi level. One can see that the density of states is,relative to a nonmagnetic system, a strong function of temperature.

0 0.005 0.01 0.0150

2

4

6

8

10

12

14

16

kB

T/W

ρ(m

Ωcm

)

p = 0.1x

p = 0.3x

p = 0.5xp = 0.6xp = 0.7xp = 0.8xp = 0.9xp = x

0 0.005 0.01 0.0150

5

10

15

20

25

kB

T/W

ρ(m

Ωcm

)

p = 0.1x

p = 0.3xp = 0.5x

p = 0.8x

p = 0.9xp = x

(b)

(a)

FIG. 3. Resistivity as a function of temperature for two Jpd, aJpd=0.35W and b Jpd=0.4W for an impurity concentration x=0.053, ENM =EM =0, and BB=0, for two values of Jpd, with p asa parameter.

ARSENAULT et al. PHYSICAL REVIEW B 77, 115211 2008

115211-8

Page 9: Transport in the metallic regime of Mn-doped III-V semiconductors

Figure 4 shows the resistivity at one temperature as afunction of impurity concentration for varying degrees ofhole doping. The results of Figs. 3 and 4 are based on thespin scattering model with no additional potential scatteringterms and no electron-phonon interaction. A fit to experimen-tal data must take these other mechanisms into account aswell. Thus, the complete CPA self-energy should includeother sources of random potential scattering nonzero valuesof ENM, EM, in particular, the charged impurity scatteringterms discussed below. We should also add, when necessary,the electron-phonon self-energy epE. The imaginary partof the charged impurity self-energy will contribute anothersource of temperature dependent lifetime broadening, andthe real part will enter the density of states.

Figure 5 shows the resistivity as a function of temperaturefor various magnetic fields, for one value of x, p, and Jpd,and again with EM =ENM =0. Apart from very low and veryhigh temperatures, increasing the magnetic field decreasesthe resistivity, by favoring the alignment of the impurityspins. The magnetic field also eliminates the sharp metal-insulator transition around Tc, replacing it with a smoothtransition.

Figure 6 shows the relative magnetoresistivity defined by

MR xxBext − xxBext = 0

xxBext = 0, 47

as a function of B, for various value of T. The overall trendin this exclusively spin scattering model is the strong nega-tive magnetoresistivity, as one would expect, since increas-ing the Mn spin alignment reduces disorder and thus reducesthe scattering lifetime, and no other sources of scattering areconsidered. However, lifetime is not the only quantity enter-ing the conductivity. As shown in Eq. 46, the density ofstates DEF at EF also plays an important role. There arealso regimes of positive magnetoresistivity. From Fig. 6a,we see that it is also possible for the resistance to increasewith B at low T when the hole density is very low. This isprobably because in this limit, the density of states at theFermi level decreases with magnetic field and this effect isstronger than the concomitant increase of the carrier lifetimedue to the suppression of spin disorder with B. However, wealso know that for low concentrations, when the Fermi levelis at the band edge or in the region of localized states, otherchanges arise which are not due to spin-disorder scatteringand which require another approach which is based on local-ization. In the hopping regime, not describable by CPA, the

0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1

100

101

102

x

ρ(m

Ωcm

) p = 0.1x

p = 0.3x

p = 0.5x

p = 0.8xp = x

FIG. 4. Resistivity as a function of impurity concentration x, forvarying degrees of hole doping. The other parameters are Jpd

=0.35W, ENM =EM =0, kBT=3.510−3W, and BB=0.

0 0.005 0.01 0.0150123456789

1011

kB

T/W

ρ(m

Ωcm

)

FIG. 5. Resistivity as a function of temperature for various mag-netic fields. Other parameters are x=0.053, ENM =EM =0, Jpd

=0.4W, and p=0.8x. The solid line is the zero-field result, thedashed line is for BB=110−5W, the dotted line is for BB=110−4W, and the dashed-dotted line is for BB=310−4W.

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5−16

−14

−12

−10

−8

−6

−4

−2

0

2

µB

B/W ( 10 −4 )

MR

(%)

kB

T = 3.64x10 −4W

kB

T = 1.31x10 −3W

kB

T = 1.80x10 −3W

kB

T = 0.015W

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5−50−45−40−35−30−25−20−15−10−505

µB

B/W ( 10 −4 )

MR

(%)

kB

T = 3.64x10 −4W

kB

T = 4.03x10 −3W

kB

T = 5.78x10 −3W

kB

T = 0.015W

(a)

(b)

FIG. 6. Relative magnetoresistivity for two carrier concentra-tions, a p=0.1x and b p=0.8x, for impurity concentration x=0.053, ENM =EM =0, and Jpd=0.35W, with T as a parameter.

TRANSPORT IN THE METALLIC REGIME OF Mn-DOPED… PHYSICAL REVIEW B 77, 115211 2008

115211-9

Page 10: Transport in the metallic regime of Mn-doped III-V semiconductors

magnetic field can, for example, squeeze the localized wavefunctions and increase the resistance by reducing overlap.However, it also shifts the energy and the mobility edge suchas to reduce resistance.32 When the calculated resistance ishigh and when the Fermi level is in a region of small densityof states 1019 /cm3 eV, we should therefore not trust theCPA, which gives a good description only of the metallicregime.

B. Hall effect

By definition, the Hall resistance is given by

yx =− Reyx

Rexx2 − ReyxRexy. 48

Equation 48, being a general definition, the normal and theanomalous spin-orbit generated Hall resistances are included.

In the CPA, the sign of the normal Hall effect is not de-termined by a simple relationship but depends on the disper-sion of the lattice and the behavior of the real part of theGreen’s function at the Fermi level.15,27 A full analyticalanalysis of the CPA Hall sign is beyond the scope of thispaper but we can state that there is no simple rule. The clos-est to one is that RH is electronlike when the density of statesincreases with energy and holelike when it decreases. In theapproximation of skew scattering, the sign of the anomalouseffect follows the sign of the normal effect as long as thecarriers at the Fermi level are polarized in the same directionas the total magnetization. This is true here, as can be seenfrom Fig. 1. The majority spin band at the Fermi level hasthe same sign as the overall magnetization, which is domi-nated by the localized spins. Thus, in the phase addition ap-proximation of Eq. 33, in order to get the right sign, it isessential to keep the correlation between the spin and energy.The decoupling of the spin magnetization out of the Kuboformula can give rise to the wrong sign. Here, the overallpolarization, found using Eq. 41, is opposite to the direc-tion of the field as it should be. We correct for this effect byassuming that z is in the same direction as the overallmagnetization.

Figure 7 shows, for a selected class of parameters, theoverall behavior of the Hall resistivity in the skew spin scat-tering model. In order to explain the experimental data, weneed

s

12t =0.2, which means that the skew scattering energy

has to be 0.4 times the tight-binding overlap. This is a verytoo strongly enhanced skew scattering rate. It suggests thatthe correct interpretation of the AHE in GaMnAs is mostlikely the intrinsic mechanism proposed by Sinova et al.3 Inthis work, the AHE is due to the intrinsic spin-orbit field andrelies on the multiple band nature of this class of semicon-ductors. The usual simple nearest neighbor tight-bindingmodel only gives a skew scattering contribution. The univer-sality and order of magnitude of the AHE in ferromagnetssuggest that the intrinsic process dominates in most cases.

At high temperature, the magnetism disappears. The nor-mal Hall conduction, linear in B field, is recovered see thekBT=0.015W results Figs. 7a and 7b. Note that for sim-plicity of notation, we refer to the applied field in the figureas B, even if it was defined otherwise previously.

VII. EXPERIMENTAL RELEVANCE

Figure 8 shows results of the CPA calculations for thethree transport coefficients: resistance, magnetoresistance,and Hall effect in a range of parameters for which a behaviorclose to the ones observed experimentally by Ruzmetov etal.33 and Ohno et al.34 is observed. We have added a constantcontribution to the resistivity so that the relative change inresistivity, such as discussed in connection with Fig. 3, be-tween Tc and T=0 is of the same order of magnitude, asobserved in the experiment. The constant term is chosen tobe of the same order of magnitude as the spin scattering rate.The qualitative temperature structure is satisfactory in the“metallic regime” of the material, but because we have ne-glected the charged impurity scattering, we are underestimat-ing the temperature drop of the resistivity with temperature,at higher temperatures. The latter is due to a reduction inscreening length when the density of states at the Fermi levelincreases see Eq. 51. Figure 8 also illustrates the strongtemperature dependence of the magnetoresistance, whichreaches 40%–50% at kBT0.0057W this is 60 K for W=1 eV, and how it is intimately connected to the magneticorder, as observed experimentally. For an even clearer pic-

−5 −4 −3 −2 −1 0 1 2 3 4 5−2.5

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

2.5

µBB/W

ρ H/W

(Ωm

/J)(

x1014

)

kBT = 3.64x10−4 W

kBT = 1.31x10−3W

kBT = 1.80x10−3 W

kBT = 0.015W

( 10-4 )

−5 −4 −3 −2 −1 0 1 2 3 4 5−1.5

−1

−0.5

0

0.5

1

1.5

µB

B/W ( 10 −4 )

ρ H/W

m/J

)(x1

014

)

kB

T = 3.64x10 −4W

kB

T = 4.03x10 −3W

kB

T = 5.78x10 −3W

kB

T = 0.015W

(a)

(b)

FIG. 7. Hall resistivity as a function of the applied magneticfield for two carrier concentrations, a p=0.1x and b p=0.8x, foran impurity fraction x=0.053, ENM =EM =0, and Jpd=0.35W, withT as a parameter.

ARSENAULT et al. PHYSICAL REVIEW B 77, 115211 2008

115211-10

Page 11: Transport in the metallic regime of Mn-doped III-V semiconductors

ture of the connection, see Fig. 5. The reader should note thatVan Esch et al.35 have observed magnetoresistances of500% at T=4 K and 50% at T=20 K in their more resis-tive samples with 1 /cm. In this sample, the magneti-zation dropped by 100% between 4 and 20 K.

Hwang and Das Sarma36 and Lopez-Sancho and Brey37

quite rightly pointed out that when charged impurities arepresent, they will normally dominate the scattering rate andthe temperature dependence of resistance in a wide range ofparameters. They then decided to use two different conduc-tivity models for the transport in Mn-doped GaAs, one toexplain the temperature dependence of the resistance whichemphasizes charged impurity scattering and one for the mag-netoresistance which emphasizes spin scattering. We believethat this is not necessary. It is enough to add the configura-tionally averaged charged impurity self-energy to the basisband dispersion ks→ks+ks, and then to neglect, in thesimplest limit, the real part of the impurity scattering self-energy ks. Thus, we replace the CPA self-energy s, whichappears in Eq. 16, by

ImsE + ImksE = ImsE

+ dk 1

23

Vk−k2

1 − cos k−k E − ks ,

49

where = 2 Ni, with Ni the impurity concentration, and

Vk−k =4Zimpe

2

0k − k21

1 + qsc/k − k2 . 50

In Eq. 50, Zimp is the impurity charge number, and

qsc2 = 4e2

s dEDsE −

fEE

51

is the inverse screening length. Since the results of Sec. VIshow that DsE at the Fermi level change with temperatureand field, the screening length will change as well. Note thatusually, in Boltzmann transport, one does not take into ac-count the effect of the real part of the self-energy the imagi-nary part gives the scattering on the actual band structurebecause the concentration of dopants is low. Here, the con-centration is not low. The real part should, in principle,modify the band gap and effective masses and can be in-cluded in our formalism.

We agree with Hwang and Das Sarma36 that the completeproblem in parameter space is enormously complex. In ef-fect, we should want the impurity scattering to only representan additional self-consistent, density of states dependent,lifetime process. A fully self-consistent Coulomb potential spin, multiband CPA, is far too complex and of little value.In addition, the CPA band theory cannot account for the hop-ping regime, as pointed out in the previous section.

VIII. PROBLEM OF THE INTRINSIC AND EXTRINSICHALL EFFECT

In the Kane-Luttinger theory of Jungwirth et al.,2 the Mndopants only cause a lifetime broadening and no change tothe Kane-Bloch band structure. In this formalism, the intrin-sic spin-orbit interaction is Bloch invariant and only causesband mixing. All disorder is treated only as a lifetime effect.

0 0.005 0.01 0.0153456789

1011121314

kB

T/W

ρ 0+

ρ(T)

(mΩ

cm)

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5−50−45−40−35−30−25−20−15−10−505

µB

B/W ( 10 −4 )

MR

(%)

kB

T = 3.64x10 −4W

kB

T = 4.03x10 −3W

kB

T = 5.78x10 −3W

kB

T = 0.015W

−5 −4 −3 −2 −1 0 1 2 3 4 5−1.5

−1

−0.5

0

0.5

1

1.5

µB

B/W ( 10 −4 )

ρ H/W

m/J

)(x1

014

)

kB

T = 3.64x10 −4W

kB

T = 4.03x10 −3W

kB

T = 5.78x10 −3W

kB

T = 0.015W

(b)

(a)

(c)

FIG. 8. Transport coefficients which can be directly compared tothe experiment, a =0+T, b MRB, and c HB, forENM =EM =0, Jpd=0.35W, and p=0.8x. Two possible constant con-tributions are shown in a, the solid line curve has 0

=10 m cm, and the dashed line curve has 0=3.6 m cm.

TRANSPORT IN THE METALLIC REGIME OF Mn-DOPED… PHYSICAL REVIEW B 77, 115211 2008

115211-11

Page 12: Transport in the metallic regime of Mn-doped III-V semiconductors

This is not so in the CPA where the dopant scattering causesa substantial renormalization of the band structure via thereal part of the self-energy and indeed gives an explanationof the magnetism. The spin-orbit coupling produces newterms in the Hamiltonian which allow the electrons jumpingor tunneling from one site to the next site to experience theelectric field of the neighboring atoms. The three-site pro-cesses we invoke are not normally included in the tight-binding modeling of semiconductors. They represent onlysmall modifications of the band structure, much smaller thanthe on-site spin-orbit admixture. The anomalous Hall mecha-nism in one-band tight binding is basically one of “skewscattering,” except that the jump is always from orbit to or-bit, and the effective magnetic field can only come fromanother third site. We have seen that the one-band TB de-scription is inadequate to describe the AHE in diluted mag-netic semiconductors. On the other hand, the multiple bandKane and/or Kohn-Luttinger38 approach seems to work well.It gives the right order of magnitude with an intrinsic AHE.This suggests that a TB modeling of the AHE must includethe multiple band aspect and at least a second nearest neigh-bor overlap. Since, for example, a 5% GaMnAs alloy cannothave a true Bloch band structure, in the orbital description, itis the short range many orbital aspect which must give rise tothe intrinsic AHE.

IX. CONCLUSION

We have presented a CPA theory which could explain themagnetism in Mn-doped semiconductors in the metallic re-gime and allows one to calculate the transport parameters.The numerical evaluation of resistance, magnetoresistance,and Hall coefficient, normal and anomalous, show that theoverall trends observed experimentally are reproduced by theCPA. The power of the method lies in its simplicity. All theinformation is contained in a spin and energy dependent self-energy sE. In good approximation, we may add the effectof the charged impurity and electron-phonon lifetime correc-

tions to the calculated CPA self-energy. Of the two, the firstis the most important addition because it explains why theexperimental resistance decreases at high temperatures ingeneral, more strongly than shown in Fig. 8a, when theferromagnetism has reduced to paramagnetism, and spin-disorder scattering should have reached its highest value.The “small” resistance drop at high temperature seen in Fig.8a is due to the CPA band-structure renormalization. TheCPA and t-matrix methods can be extended to treat magneticclusters and evaluate effective localized spin-spin couplingmediated by the band.

We have also demonstrated how to include charged impu-rity scattering within the same formalism. The central aim ofthis paper was to achieve an understanding of the importantmechanism which determines the conductivity behavior. Forthis purpose, we have used a one-band approach which isadequate to understand the conductivity. For optical proper-ties and the intrinsic AHE, the many band aspects are essen-tial.

The AHE has been modeled as an extrinsic effect in theframework of skew scattering. We did this using the tight-binding language. The intrinsic AHE, as invoked by Jung-wirth et al.,2 cannot be derived using a nearest neighbor TBformalism, and it is not clear how to recover it in the TBformalism. The problem of the intrinsic AHE and the TBmodel is an interesting one. It should be reexamined in detailbecause so far, the nearest neighbor tight-binding methodshave proved useful as band-structure descriptions of semi-conductors.

ACKNOWLEDGMENTS

The authors gratefully acknowledge financial supportfrom the Natural Sciences and Engineering Research Councilof Canada NSERC during this research. P.D. also acknowl-edges support from the Canada Research Chair Program.L.-F.A. gratefully acknowledges Martine Laprise for helpwith the figures and Alain Rochefort for the use of his com-puters.

*Present address: Département de Physique and RQMP, Universitéde Sherbrooke, Sherbrooke QC, Canada;[email protected]

†Present address: Department of Electrical and Computer Engineer-ing, Northwestern University, Evanston, IL, USA1 S. L. Chuang, Physics of Optoelectronic Devices, Wiley Series in

Pure and Applied Optics Wiley, New York, 1995.2 T. Jungwirth, J. Sinova, J. Masek, J. Kucera, and A. H. Mac-

donald, Rev. Mod. Phys. 78, 809 2006.3 J. Sinova, T. Jungwirth, and J. Cerne, Int. J. Mod. Phys. B 18,

1083 2004.4 H. Ohno, H. Munekata, S. Von Molnar, and L. L. Chang, J. Appl.

Phys. 69, 6103 1991.5 H. Ohno, J. Magn. Magn. Mater. 200, 110 1999.6 P. Mahdavian and A. Zunger, Appl. Phys. Lett. 85, 2860 2004.

7 L.-F. Arsenault, B. Movaghar, P. Desjardins, and A. Yelon un-published.

8 R. Karplus and J. M. Luttinger, Phys. Rev. 95, 1154 1954.9 T. Dietl, H. Ohno, and F. Matsukura, Phys. Rev. B 63, 195205

2001.10 J. Okabayashi, A. Kimura, O. Rader, T. Mizokawa, A. Fujimori,

T. Hayashi, and M. Tanaka, Phys. Rev. B 58, R4211 1998.11 J. Szczytko, W. Mac, A. Twardowski, F. Matsukura, and H.

Ohno, Phys. Rev. B 59, 12935 1999.12 Yu Bychkov and E. I. Rashba, J. Phys. C 17, 6039 1984.13 J.-N. Chazalviel, Phys. Rev. B 11, 3918 1975.14 E. A. de Andrada e Silva, G. C. La Rocca, and F. Bassani, Phys.

Rev. B 50, 8523 1994.15 B. Movaghar and R. W. Cochrane, Phys. Status Solidi B 166,

311 1991.16 B. Movaghar and R. W. Cochrane, Z. Phys. B: Condens. Matter

ARSENAULT et al. PHYSICAL REVIEW B 77, 115211 2008

115211-12

Page 13: Transport in the metallic regime of Mn-doped III-V semiconductors

85, 217 1991.17 L. M. Roth, in Liquid Metals 1976, edited by R. Evans and D. A.

Greenwood, IOP Conf. Proc. No. 30 Institute of Physics, Lon-don, 1977, Chap. 1, Pt. 2, pp. 200-206.

18 T. Matsubara and T. Kaneyoshi, Prog. Theor. Phys. 40, 12571968.

19 L. E. Ballentine, in Liquid Metals 1976, edited by R. Evans andD. A. Greenwood, IOP Conf. Proc. NO. 30 Institute of Physics,London, 1977, Chap. 1, Pt. 2, pp. 188-199.

20 J. Smit, Physica Amsterdam 24, 39 1958.21 H. A. Engel, B. I. Halperin, and E. I. Rashba, Phys. Rev. Lett.

95, 166605 2005.22 Y. K. Kato, R. C. Myers, A. C. Gossard, and D. D. Awschalom,

Phys. Rev. Lett. 93, 176601 2004.23 A. Baldareshi and N. O. Lipari, Phys. Rev. B 8, 2697 1973.24 G. A. Fiete, G. Zarand, and K. Damle, Phys. Rev. Lett. 91,

097202 2003.25 T. P. Pareek and P. Bruno, Pramana, J. Phys. 58, 293 2002.26 T. Damker, H. Bottger, and V. V. Bryksin, Phys. Rev. B 69,

205327 2004.27 L.-F. Arsenault, M. Sc.A. thesis, École Polytechnique de Mon-

tréal, 2006.

28 E. N. Economu, Green’s Functions in Quantum PhysicsSpringer-Verlag, Berlin, 1983.

29 B. Velicky, S. Kirkpatrick, and H. Ehrenreich, Phys. Rev. 175,747 1968.

30 M. Takahashi and K. Mitsui, Phys. Rev. B 54, 11298 1996.31 W. Allen, E. G. Gwinn, T. C. Kreutz, and A. C. Gossard, Phys.

Rev. B 70, 125320 2004.32 B. Movaghar and S. Roth, Synth. Met. 63, 163 1994.33 D. Ruzmetov, J. Scherschligt, D. V. Baxter, T. Wojtowicz, X.

Liu, Y. Sasaki, J. K. Furdyna, K. M. Yu, and W. Walukiewicz,Phys. Rev. B 69, 155207 2004.

34 H. Ohno, H. Munekata, T. Penney, S. von Molnar, and L. L.Chang, Phys. Rev. Lett. 68, 2664 1992.

35 A. Van Esch, L. Van Bockstal, J. De Boeck, G. Verbanck, A. S.van Steenbergen, P. J. Wellmann, B. Grietens, R. Bogaerts, F.Herlach, and G. Borghs, Phys. Rev. B 56, 13103 1997.

36 E. H. Hwang and S. Das Sarma, Phys. Rev. B 72, 0352102005.

37 M. P. Lopez-Sancho and L. Brey, Phys. Rev. B 68, 1132012003.

38 W. Kohn and J. M. Luttinger, Phys. Rev. 108, 590 1957.

TRANSPORT IN THE METALLIC REGIME OF Mn-DOPED… PHYSICAL REVIEW B 77, 115211 2008

115211-13