Transcript Document
5. Magnetic properties and disorder; recent developments; questions for the future
Magnetic properties and disorder • Anisotropic RKKY interaction • Effect of
weak
disorder on the magnetization • Disorder and magnetization: Strong disorder approach • Lightly doped, strongly disordered DMS: Percolation picture • Stability of collinear magnetic state, magnetic fluctuations • Spin-charge coupling: possible stripe order • Magnetic order and transport: Resistive anomaly Recent experiments – new puzzles Magnetic semiconductors – what next?
Magnetic properties and disorder
Complicated system of coupled carriers and spins
Experiments:
Potashnik
et al.
(2001) Park
et al.
, PRB
68
, 085210 (2003)
(Ga,Mn)As Mn
-implanted
GaAs:C
more disorder → straight/convex magnetization curves annealing seems to
reduce
disorder (curves more mean-field-like) magnetization for
T
!
0
is significantly less than saturation
Anisotropic RKKY interaction
Reminder: local moment polarizes carrier band, other local moments see oscillating magnetization ( “integrate out the carriers”) RKKY interaction for local carrier-impurity exchange
J
pd and parabolic band (Lecture 3) is isotropic in real and spin space isotropic in real space isotropic in spin space Zaránd & Jankó, PRL
89
, 047201 (2002): local
J
pd →
H
and spherical (4-band) approximation with spin-orbit interaction RKKY isotropic in real space, highly anisotropic in spin space → frustration Brey & Gómez-Santos: PRB
68
, 115206 (2003): non-local
J
pd → with Gaussian form, realistic 6-band weakly anisotropic in real and spin space
k
¢
p
Hamiltonian
Attempt at more realistic description: C.T. & MacDonald, PRB
71
, 155206 (2005) (a) Start from Slater-Koster tight binding theory for
GaAs
with spin-orbit coupling [Chadi (1977)] 16 bands (b) Incorporation of
Mn
d-orbitals hybridization with
GaAs
sp-orbitals: photoemission (Okabayashi) ,
ab-initio
(Sanvito) interactions: Hubbard
U
and Hund‘s 1st rule
J H
to preserve spherical symmetry of d-shell in real and spin space: Parmenter, PRB
8
, 1273 (1973) – 5-orbital Anderson model does
not
(c) Canonical perturbation theory for strong
U
,
J H
for single
Mn
Method: Chao
et al.
, PRB
18
, 3453 (1978) , related to Schrieffer-Wolff transformation in (
+
), out (
–
) unitary transformation does not change the physics expand in choose
T
(hermitian) to make linear order vanish approximation: truncate after 2nd order set
= 1
approximation: project onto ground state with
N
= 5
,
S
= 5/2
E F
projected energies:
Inserting one finds
k
dependent ↔ nonlocal in real space
spin scattering
with
d
5 !
d
4 For small
k
,
k
´
only
=
´ = p
x,y,z
antiferromagnetic correct order of magnitude
d
5 !
d
6
(d) RKKY interaction two
Mn
impurities at
0
and
R
→ canonical transformation integrate out the carriers
S
1
J G G J
S
2 oscillating ferromagnetic at small
R
highly anisotropic in real space (from bands) anisotropic in spin space for larger
R
(spin-orbit)
Can the spin anisotropy lead to non-collinear magnetization reduction of average magnetization at
T
= 0
and (Zaránd & Jankó) ?
For magnetization in
z
direction typical effective field in transverse (
x
) direction is given by With our results for
J ij
→ this is small even at no strong non-collinearity
x
= 0.01
or reduction due to
this
mechanism Complementary
ab-initio
approach: Do LDA (
+
U
) – usually without spin-orbit coupling – for supercell with 2
Mn
→ impurities in various positions, extract
J
(
R
)
from total energy highly anisotropic in real space, isotropic in spin space (no spin-orbit!) Typically overestimates
J
(
R
)
and thus
T c
(while we underestimate it)
Effect of weak disorder on the magnetization
Results for magnetization and
T c
in Lecture 3 did not include disorder Inclusion of disorder: Coherent potential approximation (CPA) Takahashi & Kubo, PRB
66
, 153202 (2002); also Bouzerar, Brey etc.
impurity sites only
local
Coulomb disorder potential (or zero in most other works) (local
J
, simple semicircular DOS assumed)
CPA:
Approximate true scattering by multiple scattering at a single impurity embedded in an effective medium – good for low impurity concentration Cannot describe localization, difficult to treat extended scatterers (Coulomb)
Typical for CPA and DMFT calculations with point-like impurities:
T c
decreases for high hole concentrations (weak compensation)
x
= 0.05
Origin: Impurity band broadens for higher spin polarization (since it is mostly due to large
J
) → spin polarization unfavorable for band filling &
1/2
1 hole per
Mn (Ga,Mn)As
: Takahashi & Kubo Problem: Always requires unphysically large
J
to obtain reasonable
T c
, since no long-range Coulomb potentials → incorrect impurity-band physics
Disorder and magnetization: Strong disorder approach
Zener model with potential disorder: Method: mean-field theory with disorder convex → concave for less disorder
Lightly doped, strongly disordered DMS: Percolation picture
For low concentrations
x
of magnetic impurities in
III-V
DMS Kaminski & Das Sarma, PRL
88
, 247202 (2002); PRB
68
, 235210 (2003), Alvarez
et al.
, PRL
89
, 277202 (2002), etc.
at
T
= 0
: ferromagnetism if aligned clusters percolate at
T
> 0
: two clusters align if the weak coupling
J
´ → between them is & express by
k B T T
-dependent BMP radius For lower
T
the aligned region growths and eventually percolates at
T c
weak link
J
´ (Kaminski & Das Sarma) Exponentially small for small hole concentration, but not zero (since no quantum fluctuations) – compare VCA/MF:
T c
/
n h
1/3
n i
Global magnetization is carried by sparse cluster at
T
.
T c
: small Percolation theory: Kaminski & Das Sarma (2002) Monte Carlo simulations: Mayr
et al.
, PRB
65
, 241202(R) (2002) Highly convex magnetization curves (upwards curvature) MC: magnetization at
T
!
0
strongly reduced compared to saturation For
clustered
→ defects: large ordered clusters exist for
T
&
T c
, percolate at
T c
more rapid (Brillouin-function-like) increase of bulk magnetization
Stability of collinear magnetic state, magnetic fluctuations
Question: stable Is the collinear state found by approximate (mean-field) methods against magnetic fluctuations? In particular with disorder?
Expand energy around mean-field solution → density of states of magnetic excitations Schliemann & MacDonald, PRL
88
, 137302 (2002) etc.: spin disorder, no Coulomb disorder DOS at negative energies: state not stable …collective spin excitations involving many spins (high participation ratio) Schliemann, PRB
67
, 045202 (2003): 6-band
k
¢
p
model solution not even stationary Goldstone mode
C.T., J. Phys.: Cond. Mat.
15
, R1865 (2003): Spin
and
Coulomb disorder, parabolic band DOS at negative energies: state not stable clustered defects: DOS shifts away from zero → stiffer magnetic order Collinear state unstable due to anisotropic magnetic interactions – but argument from RKKY interaction suggests that the deviation is small
Spin-charge coupling: possible stripe order
C.T., cond-mat/0509653 Carrier-mediated ferromagnetism → strong dependence of magnetism on hole concentration in
(In,Mn)As
,
(Ga,Mn)As
Landau theory for magnetization
m
carrier concentration
n
: and excess Ohno
et al.
(2000) magnetization charge with Introduce electrostatic potential : for p-type DMS
Spin- and charge-density waves (stripes) lowest energy periodic, anharmonic magnetization and carrier density for typical solutions
Magnetic order and transport: Resistive anomaly in dirty itinerant ferromagnets
Zumsteg & Parks, PRL
24
, 520 (1970)
Ni
Potashnik
et al.
, APL
79
, 1495 (2001)
(Ga,Mn)As
R
(
T
)
ρ
(
T
)
T c dR
/
dT T c
Theories for paramagnetic regime,
T
>
T c
:
(1)
de Gennes & Friedel, J. Phys. Chem. Solids
4
, 71 (1958) scattering from magnetic fluctuations close to
T c
: critical slowing down → static, elastic Approach equivalent to: perturbation theory, similar to inverse quasiparticle lifetime, but transport rate involves factor anomaly from small momentum transfers
q
Ornstein/Zernicke (sharp maximum)
(2)
Fisher & Langer, PRL
20
, 665 (1968) disorder damping for large length scales ↔ small
q
: electronic Green function decays exponentially on scale
l
(mean free path) no de Gennes-Friedel singularity from small
q
weak singularity from large
q
¼
2
k F
, have to go beyond Ornstein/Zernicke/Landau theory:
α
:
small
anomalous specific-heat exponent Equivalent to Boltzmann equation approach (Lecture 4) , disorder and magnetic scattering treated on equal footing
Problem:
fails for magnetic correlation length
(
T
)
À
l
(mean free path), magnetization variations are explored by diffusive carriers
Beyond the Boltzmann approach C.T., Raikh & von Oppen, PRL
94
, 036602 (2005) (a) Description of transport on large length scales (phase coherence length) 3D resistor network magnetization ~ constant in cells conductivity of network: spatial average h
…
i large system: equivalent to average over • quenched disorder • magnetization (thermal)
(b) Two spin subbands: ↑ , ↓ UCF UCF Correlation function spin ↑ , ↓ carriers have different Fermi energies but see same disorder universal conductance fluctuations (UCF)
is a scaling function of
x
= eff. Zeeman energy
£ (Stone 1985, Altshuler 1985, Lee and Stone 1985)
diffusion time
Correlations decrease with increasing Zeeman energy (c) With spin-orbit coupling: realistic case is a scaling function
H
(
y
)
of
y
= eff. Zeeman energy
£
spin-orbit time
, increases by factor of 2 in strong effective Zeeman field (d) Typical magnetization assuming Gaussian fluctuations: long-wavelength modes
with scaling function For Gaussian fluctuations: maximum at
T c
Beyond Gaussian fluctuations:
Stronger singularity
than in Fisher/Langer and de Gennes/Friedel theories
Condition:
Transport disorder-dominated at
T c
(low
T c
, strong disorder) –
(In,Mn)Sb
?
Recent experiments – new puzzles
Ferromagnetism in superdilute magnetic semiconductors
Dhar
et al.
, PRL
94
, 037205 (2005); Sagepa
et al.
, cond-mat/0509198 (
GaN:Gd
x
with
= 8
£
10
-8 to
Gd
concentrations from
2
£
10
-4 )
7
£
10
15 to
2
£
10
19
cm
-3 wurtzite structure formal valence
Gd
3+ : isovalent, configuration
4d
7 , local spin
S
= 7/2
high concentration of native donors (
N
vacancies) expected Observations: room-temperature ferromagnetism (
T c
»
360K
for
x
= 8
£
10
-8 ) highly insulating field-cooled sweep zero-field-cooled
giant magnetic moment per
Gd
for low
x
effective field acting on VB is reversed
m
absolute moment per Gd Magnetization must be carried by “something else” – native defects?
How does very little
Gd
induce magnetic ordering?
“
d
0 ferromagnetism” in oxides
Coey
et al.
, Nature Mat.
4
, 173 (2005) ferromagnetism in
HfO
2 (no partially filled shells?) magnetization extraplotates to nonzero value for strongly diluted DMS
Sn
1-
x X x
O
2
Two species of substitutional
Mn
Kronast
et al.
in
(Ga,Mn)As
: XAS, XMCD results
(BESSY II Collaboration), submitted
Mn
2+ with
~ 3d
5 configuration, large moment, orders magnetically
Mn
3+ with
~ 3d
4 configuration, strong valence fluctuations, large moment, does not order Questions: What mechanism lifts the
d
5 Why does high-spin
Mn
3+ !
d
4 transition by several eV?
not participate in the ordering?
Magnetic semiconductors – what next?
Goals for DMS experiments:
control of growth dependence, reproducability unconventional DMS (oxides etc.), concentrated magnetic semiconductors other magnetic probes: NMR/NQR/ SR and neutron scattering dynamics: optical pump-probe and noise crossover to antiferromagnetism, superconductivity, QHE…
…and theory:
study of crossover between weak doping (BMP‘s) and band picture unconventional DMS (oxides etc.): different mechanisms?
better
ab-initio
methods to get hydrogenic impurity level of
Mn
in
GaAs
detailed simulation of DMS growth to find defect distribution selfconsistent theory of scattering and carrier-mediated magnetism
DMS/nonmagnetic semiconductor heterostructures
Transport, disorder & magnetism 1.
delta-doped layers, single layer
vs.
superlattice metallic or insulating?
magnetic properties?
2.
FNF structure: RKKY coupling between layers, control by gate voltage – unlike metal structures 3.
DMS quantum dots • many local moments • few local moments 4.
DMS/nonmagnetic interfaces, spin injection MC simulation of interdiffusion
cf.
experiment: Kawakami
et al.
, APL
77
, 2379 (2000)
Electronic correlations & quantum critical points
electronic correlations:
(Ga,Gd)N
?
at least two quantum critical points: • ferromagnetic end point • metal-insulator transition …with overlapping critical regions Griffiths-McCoy singularities: rare regions relevant for poperties Galitski
et al.
, PRL
92
, 177203 (2004)
Vision:
DMS are ideal materials to study the interplay of disorder and electronic correlations: both are important and can be tuned Possible parallels to cuprates: indications that dopand-induced disorder is important in cuprates, McElroy
et al.
, Science
309
, 1048 (2005)
Diluted Magnetic Semiconductors
Prof. Bernhard Heß-Vorlesung 2005
I am grateful for discussions and collaborations with G. Alvarez, W.A. Atkinson, M. Berciu, L. Borda, G. Bouzerar, L. Brey, H. Buhmann, K.S. Burch, S. Dhar, T. Dietl, H. Dürr, S.C. Erwin, G.A. Fiete, E.M. Hankiewicz, F. Höfling, P.J. Jensen, T. Jungwirth, P. Kacman, J. König, J. Kudrnovský, A.H. MacDonald, L.W. Molenkamp, W. Nolting, H. Ohno, F. von Oppen, C. Paproth, M.E. Raikh, F. Schäfer, J. Schliemann, M.B. Silva Neto, J. Sinova, C. Strunk, G. Zaránd and others