Surprising Effects of Electronic Correlations in Solids Dieter Vollhardt Center for Electronic Correlations and Magnetism University of Augsburg Sólidos 2015, November 10, 2015; La Plata, Argentina
Surprising Effects of Electronic Correlations in Solids
Dieter Vollhardt
Center for Electronic Correlations and MagnetismUniversity of Augsburg
Sólidos 2015, November 10, 2015; La Plata, Argentina
Outline:
• What are electronic correlations?
• Modeling of correlated materials
• Ab initio approach to correlated electron materials
• Applications:
- Lattice stability of Fe
- FeSe: Parent compound for Fe-based superconductors
Correlations
Correlation [lat.]: con + relatio ("with relation")
Correlations in mathematics, natural sciences:
Definition of Correlations (I): Effects beyond factorization approximations (e.g., Hartree-Fock)
AB A B≠
2( ) ( ') ( ) ( ')n n n n n≠ =r r r r
e.g., densities:
Temporal/spatial correlations in everyday life
Correlations?
But: External periodic potential long-range order enforced no genuine correlations
Temporal/spatial correlations in everyday life
Time/space average inappropriate
Temporal/spatial correlations in everyday life
Electronic Correlations in Solids
Narrow d,f-orbitals/bands electronic correlations important
Correlated electron materials have unusual properties
} How to study correlated systemstheoretically?
• sensors, switches, Mottronics• spintronics• thermoelectrics• high-Tc superconductors
With potential for technological applications:
• functional materials:oxide heterostructures …
• huge resistivity changes• gigantic volume anomalies• colossal magnetoresistance• high-Tc superconductivity• metallic behavior at interfaces of insulators
material realistic modelÞmodeling
quantum many-particle problem
material realistic modelÞmodeling
maximal reduction: Hubbard modelÞU
time
†
, ,σ σ
σ↑ ↓= +− ∑ ∑H c c n nt Ui j i i
i j i
Hubbard model
time
Dimension of Hilbert spaceL: # lattice sites
(4 )LO
Computational time for N2 molecule: ca. 1 year with 50.000 compute nodes
n n n n↑ ↓ ↑ ↓≠i i i i
Static (Hartree-Fock-type) mean-field theories generally insufficient
Gutzwiller, 1963Hubbard, 1963Kanamori, 1963
†
, ,σ σ
σ↑ ↓= +− ∑ ∑H c c n nt Ui j i i
i j i
Theoretical challenge of many-fermion problems:Construct reliable, comprehensive
non-perturbative approximation schemes
Hubbard model
time
n n n n↑ ↓ ↑ ↓≠i i i i
Static (Hartree-Fock-type) mean-field theories generally insufficient
Purely numerical approaches (d=2,3): hopeless
Gutzwiller, 1963Hubbard, 1963Kanamori, 1963
How to combine?
Held (2004)
time
Non-perturbative approximation schemes for real materials
Dynamical Mean-Field Theory (DMFT)of Correlated Electrons
Metzner, Vollhardt (1989)
,d Z→∞→
Z=12
time
dynamicalmean-field
Theory of correlated electrons
Hubbard model
Face-centered cubic lattice (d=3)
Georges, Kotliar (1992))
Self-consistent single-impurity Anderson model
†
, ,σ σ
σ↑ ↓= +− ∑ ∑H c c n nt Ui j i i
i j i
Solve with an „impurity solver“, e.g., QMC, NRG, ED,...
Exact time resolved treatment of local electronic interactions
Dynamical mean-field theory (DMFT) of correlated electrons
U
Spec
tral
fun
ctio
n
( , )ωΣ k ( )ωΣ
Metzner, DV (1989)Georges, Kotliar (1992)
Kotliar, DV (2004)
U
Landau quasiparticles, not “electrons“
Exact time resolved treatment of local electronic interactions
Dynamical mean-field theory (DMFT) of correlated electrons
U
Spec
tral
fun
ctio
n
Experimentallydetectable!
Definition ofCorrelations (II):
Transfer of spectral weight
( , )ωΣ k ( )ωΣ
Metzner, DV (1989)Georges, Kotliar (1992)
Kotliar, DV (2004)
U
Material specific electronic structure (Density functional theory: LDA, GGA, ...) or GW
Computational scheme for correlated electron materials:
+Local electronic correlations
(Many-body theory: DMFT)
Anisimov et al. (1997)Lichtenstein, Katsnelson (1998)Held et al. (2003)Kotliar et al. (2006)
LDA+DMFT
=
Material specific electronic structure (Density functional theory: LDA, GGA, ...) or GW
Computational scheme for correlated electron materials:
+Local electronic correlations
(Many-body theory: DMFT)
=X+DMFT X=LDA, GGA; GW, …
1) Calculate LDA band structure: ' 'ˆ( )lml m LDAk Hε →
ˆLDAH
,
ˆd
d
i mi i
dm
n σσ
ε==
− ∆∑∑
double counting correction''mmU σσ
local Coulomb interaction
basis' ' e.g. LMTO
ˆ( )lml m LDAk Hε →
LDA+DMFT: Simplest version
Held, Nekrasov, Keller, Eyert, Blümer, McMahan, Scalettar, Pruschke, Anisimov, DV (Psi-k, 2003)
Goal: Dynamical mean-field approach with predictive power for strongly correlated materials
FOR 1346 Research Unit
Fe
Application of LDA+DMFT
• Most abundant element by mass on Earth
• Ferromagnetism: Longest known quantum many-body phenomenon
• Still most widely used metal in modern day industry (“iron age”)
Narrow d,f-orbitals/bands electronic correlations important
DMFT: Ferromagnetism in the one-band Hubbard modelUlmke (1998)
Generalized fcc lattice ( )Z →∞
ferromagnetic metal
Ferromagnetic order of itinerant local moments
Lichtenstein, Katsnelson, Kotliar (2001)
LDA+DMFT
DMFT: Ferromagnetism in the one-band Hubbard modelUlmke (1998)
Generalized fcc lattice ( )Z →∞
ferromagnetic metal
Ferromagnetic order of itinerant local moments
Lichtenstein, Katsnelson, Kotliar (2001)
LDA+DMFT
Exceptional:• Abundance of allotropes: α, γ, δ, ε, … phases• bcc-phase stable for P,T0• Very high Curie temperature (TC = 1043 K)
Fe
ferrite
austenite
hexaferrum
Fe
ferrite
austenite
hexaferrum
©Lawrence Livermore National Laboratory
Fe
ferrite
austenite
hexaferrum
©Lawrence Livermore National Laboratory
Mikhaylushkin, Simak, Dubrovinsky, Dubrovinskaia, Johansson, Abrikosov (2007)
Until recently: LDA+DMFT investigations of correlated materials for
given lattice structure
• How do electrons + ions influence each other ? • Which lattice structure is stabilized ?
Investigation of the structural stability of Fe
Ivan Leonov (Augsburg)Vladimir Anisimov (Ekaterinburg)Alexander Poteryaev (Ekaterinburg)
Collaborators:
Leonov, Poteryaev, Anisimov, DV; Phys. Rev. Lett. 106, 106405 (2011)
Fe
ferrite
austenite
hexaferrum
DFT(GGA): finds paramagnetic bcc phase to be unstable
- What stabilizes paramagnetic ferrite?- What causes the bcc-fcc structural phase transition?
Tstruct
Bain and Dunkirk (1924)
• continuous transformation path from bcc-phase to fcc-phase• volume per atom fixed at exp. value of α-Fe
Total energies calculated along bcc-fcc Bain transformation path:
bcc-fcc structural transition in paramagnetic Fe
bcc (c/a=1) fcc (c/a = )2
Z = 8 Z = 12 c/a
21
c c
aa
Pressure, GPa
Coulomb interaction between Fe 3d electrons: U=1.8 eV, J=0.9 eV
bcc-fcc structural transition in paramagnetic Fe
Construct Wannier functions for partially filled Fe sd orbitals
First-principles multi-band Hamiltonianincluding local interactions
Goal: Understand the structural stability ofparamagnetic bcc phase
GGA+DMFT:
bcc-fcc structural transition at Tstruct ≈ 1.3 TC
Pressure, GPa
GGA+DMFT total energy Etot - Ebcc
fccbcc
bcc-fcc structural transition in paramagnetic Fe
Construct Wannier functions for partially filled Fe sd orbitals
What determines the strong temperature dependence of the total energy?
GGA:Only paramagnetic fcc structure stable
Goal: Understand the structural stability ofparamagnetic bcc phase
GGA+DMFT total energy Etot - Ebcc
fccbcc
3.6 TC
• kinetic energy favors fcc structure• correlation energy indifferent as in GGA fcc structure stable}
bcc-fcc structural transition in paramagnetic Fe
Contributions to the GGA+DMFT total energy:
Etot=Ekin+Eint
1.8 TC
bcc
• kinetic energy favors fcc structure• correlation energy increases fcc structure still stable}
GGA+DMFT total energy Etot - Ebcc
fcc
bcc-fcc structural transition in paramagnetic Fe
Contributions to the GGA+DMFT total energy:
Etot=Ekin+Eint
1.2 TC
bcc
• kinetic energy favors fcc structure• correlation energy increases bcc structure becomes stable}
GGA+DMFT total energy Etot - Ebcc
fcc
bcc-fcc structural transition in paramagnetic Fe
Contributions to the GGA+DMFT total energy:
Etot=Ekin+Eint
GGA+DMFT:
bcc-fcc structural transitionat Tstruct ≈ 1.3 TC
0.9 TC
bcc
• kinetic energy favors fcc structure• correlation energy increases bcc structure remains stable}
GGA+DMFT total energy Etot - Ebcc
fcc
bcc-fcc structural transition in paramagnetic Fe
GGA+DMFT:
bcc-fcc structural transitionat Tstruct ≈ 1.3 TC
Contributions to the GGA+DMFT total energy:
Etot=Ekin+Eint
Electronic correlations responsible for Tstruct > TC
Pressure, GPa
Fe
Tstruct
bcc-fcc structural transition in paramagnetic Fe
Conclusion:
Vexp ~ 165 / 158 au3
• V ~ 161.5/158.5 au3 in bcc/fcc phaseincreased by electronic repulsion
Non-magnetic GGA:
• V ~ 141/138 au3 in bcc/fcc phasetoo small density too high
GGA+DMFT:
Agrees well with exp. data:
GGA+DMFT
bcc Fe
fcc Fe∆V ~ -2%
Equi
libri
um v
olum
e (a
u3) Pressure, GPa
bcc-fcc structural transition in paramagnetic Fe
Equilibrium volume V
Leonov, Poteryaev, Anisimov, DV; Phys. Rev. B 85, 020401(R) (2012)
Lattice dynamics and phonon spectra of Fe
Pressure, GPa
Lattice dynamics of paramagnetic bcc iron
Non-magnetic GGA phonon dispersion
Exp.: Neuhaus, Petry, Krimmel (1997)
1. Brillouin zone
Dynamically unstable +elastically unstable (C11, C‘ < 0)
Pressure, GPa
GGA+DMFT phonon dispersion at 1.2 TC
• phonon frequencies calculated with frozen-phonon method
• harmonic approximation
Stokes, Hatch, Campbell (2007)
Calculated:• equilibrium lattice constant
a~2.883 Å (aexp~2.897 Å)• Debye temperature Θ~458 K
Lattice dynamics of paramagnetic bcc iron
Exp.: Neuhaus, Petry, Krimmel (1997)
Lattice dynamics of paramagnetic fcc iron
Non-magnetic GGA phonon dispersion 1. Brillouin zone
Elastic constants much too large
Pressure, GPa
Exp.: Zarestky, Stassis (1987)
GGA+DMFT phonon dispersion at 1.4 TC
Exp.: Zarestky, Stassis (1987)
Calculated:• equilibrium lattice constant
a~3.605 Å (aexp~3.662 Å)• Debye temperature Θ~349 K
Pressure, GPa
Lattice dynamics of paramagnetic fcc iron
Strong anharmonic effectsT1
Dynamically and elastically unstable
Pressure, GPa
Instability of bcc phase due to soft transverse T1 acoustic mode near N-point •
Leonov, Poteryaev, Gornostyrev, Lichtenstein, Katsnelson, Anisimov, DV;Scientific Reports 4, 5585 (2014)
N [110] displacement at 1.4 TC:
What causes the bcc-fcc structural phase transition ?
GGA+DMFT phonon dispersion for bcc structure
T1 mode becomes strongly anharmonic high lattice entropy lowers free energy F = E – TS
What stabilizes the high temperature bcc (δ) phase ?
Pressure, GPa
bcc (δ) phase of iron stabilized byelectronic correlations + lattice entropy
Quantitative estimate:
Leonov, Poteryaev, Anisimov, DV; Phys. Rev. B 85, 020401(R) (2012)
Leonov, Skornyakov, Anisimov, DV; Phys. Rev. Lett. 115, 106402 (2015)
FeSe
Fe-based superconductors
‘11’ FeSe ‘111’ LiFeAs ‘1111’ LaFeAsO ‘122’ BaFe2As2
• new class of superconducting materials (Tc ~ 55 K in SmFeAsO1-xFx)• parent compounds are metals (typically do not superconduct)• superconductivity induced by doping or pressure
Y. Kamihara et al. (2008)
similarity to high-Tc cuprates
FeSe: Crystal structure and properties
• How does the electronic and magnetic structure of FeSe change with lattice expansion ?
• Why does Tc increase ?
FeSe FeSe1-xTex , x<0.7 FeTe
no static magnetic order,short-range fluctuations with Qm=(π, π)
TN ~ 70 KQm=(π, 0)
Tc ~ 8 K Tc ~ 14 K -
• Simplest structure of/parent compound forFe-based superconductors
• tetragonal crystal structure• Tc ~ 8 K (~ 37 K under hydrostatic pressure)• isovalent substitution Se → Te:
- lattice expansion- Tc increases up to ~ 14 K
GGA+DMFT total energy U = 3.5 eV, J = 0.85 eV
FeSe: Phase stability and local magnetism
• isostructural transformation upon ~ 10%expansion of the lattice (pc ~ -6.4 GPa)
• strong increase of the fluctuating magnetic moment
Observed in FeTe:- tetragonal to collapsed-tetragonalphase transformation under pressure
- suppression of magnetism > 4-6 GPaExp: Zhang et al. (2009)
FeSe: Spectral function
Low-volume phase
High-volume phase
Van Hove singularityshifts to EF due to correlations
Exp
ansi
on
Exp.: Yokoya et al. (2012)
Spectral weight suppressed with Tesubstitution ( = lattice expansion)
T = 290 K
non-magnetic GGA
Exp
ansi
on
GGA+DMFT
Lifshitz transition
Low-volume phase:
• Orbital-selective shift and renormalization of quasiparticle bands
• Van Hove singularity at M point pushed towards EF
High-volume phase:
• Complete reconstruction of the electronic structure
• Correlation-induced shift of Van Hove singularity above EF
Lifshitz transition
FeSe: Electronic structure
Leonov et al., Phys. Rev. Lett. 115, 106402 (2015)
Shift of Van Hove singularity to EF goes along with increase of Tc in FeSe1-xTex
Conjecture:Superconductivity in FeSe1-xTex depends sensitively on Van Hove singularity
Correlation-induced change of magnetic structure upon expansion of lattice
In accord with experiments on FeSe1-xTex Xia et al. (2009), Tamai et al. (2010), Maletz et al. (2014), Nakayama et al. (2014)
non-magnetic GGA
Low-volume phase:
• Correlations do not change FS• In-plane nesting with Qm=(π,π)
High-volume phase:
• Correlations induce topological change of FS Lifshitz transition
• Electron pocket at M point vanishes• In-plane nesting with Qm=(π,0)
GGA+DMFT
Exp
ansi
on
FeSe: Fermi surface
Leonov et al., Phys. Rev. Lett. 115, 106402 (2015)
Conclusion:
The structural stability of solidsdepends strongly on
Coulomb correlations between electrons