Correlation effects in transport through quantum wires: a - - PowerPoint PPT Presentation

correlation effects in transport through quantum wires a
SMART_READER_LITE
LIVE PREVIEW

Correlation effects in transport through quantum wires: a - - PowerPoint PPT Presentation

Correlation effects in transport through quantum wires: a functional renormalization group approach Kurt Sch onhammer Institut f ur Theoretische Physik Universit at G ottingen Collaborators from Stuttgart Sabine Andergassen


slide-1
SLIDE 1

Correlation effects in transport through quantum wires: a functional renormalization group approach

Kurt Sch¨

  • nhammer

Institut f¨ ur Theoretische Physik Universit¨ at G¨

  • ttingen
slide-2
SLIDE 2

Collaborators

  • from Stuttgart
  • Sabine Andergassen
  • Tilman Enss
  • Walter Metzner
  • from Aachen
  • Ulrich Schollw¨
  • ck
  • from G¨
  • ttingen
  • Volker Meden
  • Xavier Barnab´

e-Th´ eriault

  • Abdelouahab Sedeki
slide-3
SLIDE 3

Outline

  • Experimental systems — carbon nanotubes and other quantum wires
  • Transport problems to be studied — correlations and impurities
  • The method — functional renormalization group
  • A single impurity — power-laws and one-parameter scaling
  • Resonant tunneling — “universal” and “non-univeral” behavior
  • Transport in more complex geometries — Y-junctions, X-junctions, . . .
  • Summary and outlook
slide-4
SLIDE 4

Experimental systems

single-walled carbon nanotube

(theory ’97: Egger & Gogolin; Kane, Balents, & Fisher) (Dekkers group)

  • ther quantum wires
  • semiconductor heterostructures
  • chains of atoms on surfaces
  • . . .
slide-5
SLIDE 5

Transport problems:

junction

lead impurity contact gate voltage t U

ingredients:

  • 1d quantum wire
  • electron correlations — Tomonaga-Luttinger liquid
  • non-interacting leads
  • contacts
  • impurities, gate voltage
  • more complex geometries
slide-6
SLIDE 6

Scattering on 1d lattices: noninteracting fermions

The Hamiltonian for a single particle on the lattice is given by H = Hleads + VLR + Hs, with Hleads = −

  • n=−∞

(|n − 1n| + H.c.) −

  • n=N+1

(|nn + 1| + H.c.) VLR = tL (|01| + H.c) + tR (|NN + 1| + H.c.) , and Hs is an arbitrary Hamiltonian in the scattering segment from site 1 to N which is connected to ideal leads with nearest neighbor hopping equal t = −1, corresponding to the energy dispersion ǫk = −2 cos k and the bandwidth B = 4. The transmission probability |t(ǫk)|2 can be expressed in terms of the full resolvent as |t(ǫk)|2 = 4t2

Lt2 R sin2 k|N|G(ǫk + i0)|1|2

slide-7
SLIDE 7

Landauer-B¨ uttiker-formula

For spinless fermions on the lattice the Landauer-B¨ uttiker formula for the stationary current reads ˆ j∞ = e h B/2

−B/2

[fL(ǫ) − fR(ǫ)] |t(ǫ)|2dǫ , where the fa(ǫ) =

  • eβa(ǫ−µa) + 1

−1 are the Fermi functions of the leads. For βR = βL the linear conductance G(T) is given by G(T) = e2 h B/2

−B/2

(−d f dǫ)|t(ǫ)|2dǫ

T =0

= e2 h |t(µ)|2

slide-8
SLIDE 8

Example for Hs: quantum dot with ND sites

dot region: 1 < jl ≤ n ≤ jr < N, jr = jl + ND − 1 Hs = ˜ Hs + (tl|jl − 1jl| + tr|jrjr + 1| + H.c.) , where ˜ Hs has no matrix elements connecting the dot to the wire. The sites neighboring the dot are given the indices ˜ ja, i.e. ˜ jl = jl − 1, ˜ jr = jr + 1. With (Λa,∆a real functions) Γa(ǫ + i0) = t2

ja|(ǫ + i0 − Hta=0)−1|˜ ja = Λa(ǫ) − i∆a(ǫ) the transmission probability can then be expressed as |t(ǫ)|2 = 4∆l(ǫ)∆r(ǫ)|Gjr,jl(ǫ + i0)|2

slide-9
SLIDE 9

Example continued: resonant tunneling

For well separated dots (|ta| ≪ 1) ND narrow resonances

  • ccur near the

eigenvalues ǫα of the isolated dot. For ǫ ≈ ǫα one obtains the generalized Breit-Wigner form |t(ε)|2 ≈ 4∆(α)

l

(ε)∆(α)

r (ε)

  • ε − ǫα − Λ(α)

l

(ε) − Λ(α)

r (ε)

2 +

  • ∆(α)

l

(ε) + ∆(α)

r (ε)

2 , where ∆(α)

a

= |ja|ǫα|2∆a and correspondingly for the Λ(α)

a

. Questions: What replaces the LB-formula when the interaction is turned on? What happens to the resonances when the interaction is turned on?

slide-10
SLIDE 10

Meir-Wingreen-formula

Using the Keldysh nonequilibrium Green function technique Meir and Wingreen (PRL 68,2512 (1992)) derived a formula for the stationary current for the transport through an interacting region connected to noninteracting leads. For

  • ur geometry the (nonsymmetrized) stationary current (on the link 0 → 1) is

given by ˆ j∞ = e h B/2

−B/2

2i∆L(ǫ)

  • fL(ǫ) (Gr

11(ǫ) − Ga 11(ǫ)) + G< 11(ǫ)

  • dǫ,

which can easily shown to reduce to the LB-formula in the noninteracting case. The simplification under the special assumption discussed by MW cannot be used for our geometry (in the interacting model). The correction to the LB-formula as presented vanishes if Σr − Σa = 0 and Σ< = 0 holds (approximately).

slide-11
SLIDE 11

Luttinger liquids with impurities

much is known from bosonization and exactly solved models

  • density response of homogeneous system:

(Luther & Peschel ’74; Mattis ’74)

χ(q ≈ 2kF) ∝ |q − 2kF|2K−2

  • spinless fermions, nearest neighbor interaction, half-filling:

(Haldane ’80)

K−1 = 2 π arccos

  • −U

2

  • local sine-Gordon model :

(e.g. Apel & Rice ’82)

  • perturbative RG for K < 1:

(Kane & Fisher ’92)

VkF ,−kF relevant; hopping between open ends irrelevant

  • numerics: no intermediate fixed point

(Moon et al. ’93; Egger & Grabert ’95)

  • Bethe ansatz: no intermediate fixed point

(Saleur et al. ’95)

alternative fermionic approach for weak two body interaction:

  • leading-log resummation for transmission

(Glazman et al. ’93)

slide-12
SLIDE 12

Functional renormalization group

(Polchinski ’84, Wetterich ’93, Morris ’94, Salmhofer ’98, . . . )

idea

  • generating functional Γ of grand can. pot., self-energy, m-particle interaction
  • cutoff Λ in free propagator G0,Λ, cuts out infrared divergencies (if necessary)
  • take ∂Λ, expand in sources ⇒ exact hierachy of differential equations

formalism

  • “interacting” part of action:

Sint

  • { ¯

ψ}, {ψ}

  • =
  • k′,k

Vk′,k ¯ ψk′ψk + 1 4

  • k′

1,k′ 2,k1,k2

¯ uk′

1,k′ 2,k1,k2 ¯

ψk′

1 ¯

ψk′

2ψk2ψk1

  • generating functional of connected Green functions:

Wc,Λ ({¯ η}, {η}) = ln 1 ZΛ

  • D ¯

ψψ e

  • ¯

ψ,[G0,Λ]

−1ψ

  • −Sint({ ¯

ψ},{ψ})−( ¯ ψ,η)−(¯ η,ψ)

  • Legendre transformation:

ΓΛ {¯ φ}, {φ}

  • = −Wc,Λ

{¯ ηΛ}, {ηΛ}

¯ φ, ηΛ −

  • ¯

ηΛ, φ

  • +
  • ¯

φ,

  • G0,Λ−1 φ
slide-13
SLIDE 13
  • differentiate:

∂ΛΓΛ = Tr

  • ∂Λ
  • G0,Λ−1G0,Λ

− Tr

  • GΛ∂Λ
  • G0,Λ−1V

δ2ΓΛ δφδφ, GΛ

  • expand in sources:

ΓΛ {¯ φ}, {φ}

  • =

  • m=0

(−1)m (m!)2

  • k′

1,...,km

γΛ

m (k′ 1, . . . , k′ m; k1, . . . , km) ¯

φk′

1 . . . ¯

φk′

mφkm . . . φk1

  • exact hierachy of flow equations:
  • with GΛ =
  • G0,Λ−1 + γΛ

1

−1 , SΛ = GΛ∂Λ

  • G0,Λ−1GΛ
  • and G0,Λ0 = 0 , G0,Λ=0 = G0 , ΓΛ0

{¯ φ}, {φ}

  • = Sint

φ}, {φ}

  • approximation: γΛ

mc+1 = γΛ0 mc+1 = 0 for mc ≥ 2; γΛ=0 m

correct to order mc

slide-14
SLIDE 14

Transport in a quantum wire – fRG approach (linear response)

t U

set up:

  • cutoff in Matsubara frequency G0,Λ(iω) = Θ (|ω| − Λ) G0(iω), for T = 0
  • approximations:
  • γΛ

3 = γΛ=∞ 3

= 0

  • no self-energy corrections in ∂ΛγΛ

2

  • γΛ

2 frequency independent ⇒ ΣΛ freq. indep. ⇒ no bulk power-laws

  • interaction remains nearest neighbor

equations: ∂ΛΣΛ

j,j = − 1

2πU Λ GΛ

j+1,j+1(iΛ) + GΛ j−1,j−1(iΛ) + (iΛ → −iΛ)

  • ∂ΛΣΛ

j,j±1 = 1

2πU Λ GΛ

j,j±1(iΛ) + GΛ j,j±1(−iΛ)

  • GΛ(iω) =
  • G0(iω)

−1 − ΣΛ − Σleads(iω) −1 , U Λ = f(U, Λ) Σleads(z) =

  • t2

l |1 1| + t2 r |N N|

  • G0,leads

boundary(z)

initial condition: ΣΛ=∞

j,j(±1) = Vj,j(±1)

conductance:

G(N) = e2 h |t(0, N)|2 , |t(ε, N)|2 ∼ |G1,N(ε)|2

slide-15
SLIDE 15

Two examples for the effective scattering potential

400 450 500 550 600 j 1 2 V

j,j

400 450 500 550 600 j 1 2 Σ Σ j,j+1

j,j

400 450 500 550 600 j 0.2 0.4 0.6 0.8 1 V j,j+1

400 450 500 550 600 j 0.3 0.32 0.34 Σ j,j+1

long range oscillatory potentials → power law behaviour in |t(ǫ)|2

slide-16
SLIDE 16

Comparison with DMRG for “small” N

  • G, no impurity, abrupt contacts:

1 2 3 4 5

U

0.2 0.4 0.6 0.8 1

G/(e

2/h)

DMRG fRG

N=12

  • no impurity, smooth contacts: G = e2

h

  • Friedel oscillations, open boundaries:

50 100

j

0.4 0.5 0.6 0.7

<nj>

DMRG fRG

N=128, U=1

  • nj requires its own flow-equation

(composite operator)

slide-17
SLIDE 17

Limiting cases

  • infinitesimal imp., smooth contacts:

e2 h − G ∝ N 2(1−KRG)

  • 2
  • 1

1 2 U

  • 1
  • 0.5

0.5 1 1-K

exact leading order fRG

  • weak link from jimp to jimp + 1,

smooth contacts: ρj≈jimp ∝ N −αRG

B

⇒ G ∝ N −2αRG

B

  • 2
  • 1

1 2 U

  • 1
  • 0.5

0.5 1 αB

exact leading order fRG

  • KRG and αRG

B

“almost” fulfill the Tomonaga-Luttinger relation αB = 1/K−1

  • we can distinguish between 1 − K and 1/K − 1
slide-18
SLIDE 18

Single impurity, smooth contacts — one parameter scaling

G = e2

h ˜

GK(x) , x = [T/T0(U, V )]K−1 or x = [N0(U, V )/N]K−1

10

  • 2

10

  • 1

10 10

1

x=[T/T0(U,V)]

K-1

10

  • 3

10

  • 2

10

  • 1

10 G/(e

2/h)

U=0.5,n=1/2

1-x

2

x

  • 2/K

10

  • 1

10 10

1

10

2

x=[N0(U,V)/N]

  • 1/2

10

  • 8

10

  • 6

10

  • 4

10

  • 2

10

G/(e

2/h)

U=2.23, n=1/2

0.0 0.2 0.4 0.6 0.8 1.0 |R| 10

  • 6

10 10

6

10

12

N0(U,V) U=0.5 U=1 U=2.23

(V.M. et al. ’03)

  • leading-log approach always gives

K = 1 scaling function: ˜ GK=1(x) = 1 1 + x2

slide-19
SLIDE 19

Transport at finite T — smooth contacts

10

  • 4

10

  • 3

10

  • 2

10

  • 1

10 10

1

T 10

  • 3

10

  • 2

G/(e

2/h)

U=0.5 U=0

V=10, N=10

4

T

2αB

10

  • 4

10

  • 3

10

  • 2

10

  • 1

10 10

1

T 10

  • 3

10

  • 2

10

  • 1

10 G/(e

2/h)

V=10 V=1 V=0.1

U=0.5, N=10

4

T

2αB

  • V. Meden, T. Enss, S. Andergassen, W. Metzner, and K.S., PRB 71, 041302(R)

(2005)

slide-20
SLIDE 20

“Universality”, but . . . 10

  • 2

10

  • 1

10 10

1

x=[T/T0(U,n,V)]

K-1

10

  • 3

10

  • 2

10

  • 1

10 G/(e

2/h)

K=0.85

length band

slide-21
SLIDE 21

Resonant tunneling I — experiment versus theory

(Dekkers group ’01)

theory

  • no exact solution for model with two local sine-Gordon terms
  • Gp(T)meas not consistent with UST exponent αB − 1 (Furusaki & Nagaosa ’93)
  • new suggestion: CST with exponent 2αB − 1

(Thorwart et al. ’02)

  • not found in “leading-log” method (Nazarov & Glazman ’03; Polyakov & Gornyi ’03)
  • not found in exactly solvable model

(Komnik & Gogolin ’03)

  • QMC data interpreted to be consistent with CST

(H¨ ugle & Egger ’03)

slide-22
SLIDE 22

Resonant tunneling – Conductance as function of the gate voltage

gate voltage smooth

  • 2

2 Vg 0.00 0.01 0.02 0.03 G/(e

2/h)

T=0.3 T=0.1 T=0.03 T=0

ND=6, Vl/r=10, U=0.5, N=10

4

  • at T = 0, width ∼ N K−1
slide-23
SLIDE 23

Simplest case: ND = 1

V t’ t’

  • With g = µ − V − 2t′2Gaux({Σ}, ǫ = µ + i0) the transmission probability is

|t(0)|2 = (Im g)2/|g|2

0.5 1 1.5

Re g

0.5 1 1.5

Im g U=0

0.5 1 1.5

Re g

0.5 1 1.5

Im g U=0

  • “perfect chain” FP: stable for U < 0
  • “decoupled chain” FP: stable for U > 0

(Kane & Fisher ’92, Enss et al. ’05)

slide-24
SLIDE 24

Resonant tunneling — “universal” scaling of Gp(T)

1 2 3 4 5

p

ln G (T) lnT

  • regime 1: e2

h − Gp(T) ∼ T 2 for T < ∆W

  • regime 2: e2

h − Gp(T) ∼ T 2K for ∆W < T < T ∗ (small ND, small Vl/r)

  • regime 3: Gp(T) ∼ T αB−1 and w(T) ∝ T for T ∗ < T < ∆D
  • regime 4: Gp(T) ∼ T 2αB for ∆D < T < B (large ND, large Vl/r)
  • regime 5: Gp(T) ∼ T −1 for B < T
  • T ∗ ∝ (t2

bar/ND)1/(1−αB)

slide-25
SLIDE 25

Resonant tunneling — details for the scaling of Gp(T)

10

  • 4

10

  • 3

10

  • 2

10

  • 1

10 10

1

T

  • 1
  • 0.5

exponent 10

  • 3

10

  • 2

10

  • 1

10 Gp(T)/(e

2/h)

ND=2 ND=100

Vl/r=10, U=0.5, N=10

4

2αB αB-1 2αB-1

10

  • 4

10

  • 3

10

  • 2

10

  • 1

10 10

1

T

  • 1
  • 0.5

0.5 1 exponent 10

  • 1

10 Gp(T)/(e

2/h)

ND=10 ND=50

Vl/r=0.8, U=1.5, N=10

4

2αB αB-1 2αB-1

  • no indications of 2αB − 1 (“correlated sequential tunneling”)
slide-26
SLIDE 26

Resonant tunneling — off resonance scaling

  • Vl/r = 10, U = 0.5, ND = 100, N = 104:

10

  • 4

10

  • 3

10

  • 2

10

  • 1

10 10

1

T

  • 1

1 2 exponent 10

  • 5

10

  • 4

10

  • 3

10

  • 2

G(T)/(e

2/h)

∆Vg=0.04 ∆Vg=0.001

  • two different temperature regimes with exponent 2αB
slide-27
SLIDE 27

Resonant tunneling — asymmetric barriers

  • Vl = 5, Vr = 50, U = 0.5, ND = 100, N = 104:

10

  • 3

10

  • 2

10

  • 1

10 10

1

T

  • 1
  • 0.5

0.5 exponent 10

  • 4

10

  • 3

10

  • 2

Gp(T)/(e

2/h)

ND=2 ND=20

slide-28
SLIDE 28

X,Y,. . . -junctions

  • dot junction of Nleg ≥ 2 legs, made of N sites, smoothly coupled to leads
  • perfect and open chain fixed points of wire → more fixed points
  • example: symmetric case is a fixed point with GFP = e2

h

  • 2

Nleg

2

  • scaling of spectral weight: ρdot ∼ N αB(1−2/Nleg)
  • more general junction (e.g. “ring” with flux)

(Nayak et al. ’99, Chen et al. ’02, Lal et al. ’02, Egger et al. ’03, Chamon et al. ’03, . . . )

slide-29
SLIDE 29

Y-junction with flux – set up

1 2 3

1 2 3

V t t∆

Y

φ

  • single-particle scattering theory:

g = [−V − t2

Y Gaux ({Σ})]/|t△|

G1,2 = e2 h 4 (Im g)2 e−iφ − g

  • 2

|g3 − 3g + 2 cos φ|2 G2,1 = e2 h 4 (Im g)2 e+iφ − g

  • 2

|g3 − 3g + 2 cos φ|2

  • φ = 0.4 π

0.2 0.4 0.6 0.8 1 Re g Im g

  • 3
  • 2
  • 1

1 2 3 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0.2 0.4 0.6 0.8 1 Re g Im g

  • 3
  • 2
  • 1

1 2 3 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

slide-30
SLIDE 30

Y-junction with flux – flow for U > 0

−3 −2 −1 1 2 3 Re g 0.5 1 1.5 2 Im g −3 −2 −1 1 2 3 0.5 1 1.5 2

φ=0.4π

−3 −2 −1 1 2 3 Re g 0.5 1 1.5 2 Im g −3 −2 −1 1 2 3 0.5 1 1.5 2

φ=0.1π

G1,2 = G2,1 = 0 G1,2 = G2,1 = 4

9 e2 h

G1,2 = e2

h , G2,1 = 0

  • on light blue curve: flow to blue FPs – new scaling dimension
  • general: flow to red line of FPs, but

|G1,2−G2,1| G1,2+G2,1 → 0

⇒ restoring time reversal symmetry by interaction

(Barnab´ e-Th´ eriault et al. PRL 94, 136405 (2005))

slide-31
SLIDE 31

Summary . . .

  • fRG is reliable and flexible tool to study microscopic models of inhomogeneous

1d correlated electrons

  • provides simple physical picture for e.g. transport
  • allows to describe the interplay of contacts, impurities, and correlations
  • the method covers all energy scales
  • example: resonant tunneling — “universal” and “non-universal” behavior

. . . and outlook

  • consider more realistic models for the contacts and leads
  • consider more realistic models for the quantum wire
  • in the long run: combine “ab initio” methods with fRG
  • extend to non-linear transport
slide-32
SLIDE 32

The role of more realistic contacts

  • interaction turned on abruptly at contacts
  • additional reduced hopping in and out of quantum wire

10

2

10

4

10

6

N

10

  • 9

10

  • 6

10

  • 3

10

G/(e

2/h)

no imp., tl/r=1 timp=0.5, tl/r=1 timp=0.5, tl/r=0.7

U=1.5 N

  • 2αB

N

  • 4αB
  • what happens as a function of T?
slide-33
SLIDE 33

Transport at finite T — abrupt contacts

10

  • 4

10

  • 3

10

  • 2

10

  • 1

10 10

1

T 10

  • 2

10

  • 1

10 G/(e

2/h)

tl/r=0.3 tl/r=0.8

U=1, N=10

4

T

αB

10

  • 4

10

  • 3

10

  • 2

10

  • 1

10 10

1

T 10

  • 5

10

  • 4

10

  • 3

10

  • 2

10

  • 1

G/(e

2/h)

jimp=N/2 jimp=N/4 jimp=N/2-20 timp=0.1, tl/r=0.8, U=1, N=10

4

T

αB

T

2αB

10

  • 4

10

  • 3

10

  • 2

10

  • 1

10 10

1

T 10

  • 2

10

  • 1

G/(e

2/h)

jimp=N/2 jimp=N/4 timp=0.5, tl/r=0.8, U=1, N=10

4

T

αB

T

2αB

10

  • 4

10

  • 3

10

  • 2

10

  • 1

10 10

1

T 10

  • 5

10

  • 4

10

  • 3

10

  • 2

10

  • 1

G/(e

2/h)

timp=0.1, tl/r=0.3, U=1, N=10

4

T

αB

T

2αB

slide-34
SLIDE 34

Resonant tunneling V — details for ND = 1 (U = 1)

10

  • 4

10

  • 3

10

  • 2

10

  • 1

10 10

1

T

10

  • 3

10

  • 2

10

  • 1

10

Gp(T)/(e

2/h)

tl/r=0.05 tl/r=0.1 tl/r=0.2 tl/r=0.5 tl/r=0.7

10

  • 4

10

  • 3

10

  • 2

10

  • 1

10 10

1

T

  • 1.0
  • 0.5

0.0 exponent of Gp(T)/(e

2/h)

αB-1

Γl(ε, T ) = t2

l Im ˜

G0

jl−1,jl−1(ε, T )

Ωl(ε, T ) = t2

l Re ˜

G0

jl−1,jl−1(ε, T )

|t(ε, T )|2 =

10

  • 3

10

  • 2

10

  • 1

10 10

1

T 1 2 exponent of 1-Gp(T)/(e

2/h)

2K

4Γl(ε, T )Γr(ε, T ) [ε − Vg − Ωl(ε, T ) − Ωr(ε, T )]2 + [Γl(ε, T ) + Γr(ε, T )]2

slide-35
SLIDE 35

Resonant tunneling — |t(ε, T, N)|2

gate voltage smooth

  • 2
  • 1

1 2 ε 0.0 0.5 1.0 |t(ε,T)|

2

T=10 T=1 T=0.1 T=0

ND=10, Vl/r=10, U=0.5, N=10

4, Vg=0.2404...

  • important T-dependence
  • 2

2 Vg 0.00 0.01 0.02 0.03 G/(e

2/h)

T=0.3 T=0.1 T=0.03 T=0

ND=6, Vl/r=10, U=0.5, N=10

4

  • at T = 0, w ∼ N K−1