Introduction to Linear Response Propagators and Green’s Functions

27
Many-Body Theory Application to Electrons in Solids Charles Patterson [email protected] School of Physics, Trinity College Dublin Introduction to Linear Response Propagators and Green’s Functions Green’s Function for Schrödinger Equation Functions of a Complex Variable Contour Integrals in the Complex Plane Schrödinger, Heisenberg, Interaction Pictures Occupation Number Formalism Field Operators Wick’s Theorem Many-Body Green’s Functions Equation of Motion for the Green’s Function Evaluation of the Single Loop Bubble The Polarisation Propagator The GW Approximation The Bethe-Salpeter Equation Numerical Aspects of Many-Body Theory The GW Approximation Hybrid Density functionals

description

Many-Body Theory Application to Electrons in Solids Charles Patterson [email protected] School of Physics, Trinity College Dublin. Introduction to Linear Response Propagators and Green’s Functions Green’s Function for Schr ö dinger Equation Functions of a Complex Variable - PowerPoint PPT Presentation

Transcript of Introduction to Linear Response Propagators and Green’s Functions

Page 1: Introduction to Linear Response Propagators and Green’s Functions

Many-Body TheoryApplication to Electrons in Solids

Charles Patterson [email protected]

School of Physics, Trinity College Dublin

• Introduction to Linear Response• Propagators and Green’s Functions• Green’s Function for Schrödinger Equation• Functions of a Complex Variable• Contour Integrals in the Complex Plane• Schrödinger, Heisenberg, Interaction Pictures• Occupation Number Formalism• Field Operators • Wick’s Theorem• Many-Body Green’s Functions• Equation of Motion for the Green’s Function• Evaluation of the Single Loop Bubble• The Polarisation Propagator• The GW Approximation• The Bethe-Salpeter Equation• Numerical Aspects of Many-Body Theory• The GW Approximation• Hybrid Density functionals

Page 2: Introduction to Linear Response Propagators and Green’s Functions

Recommended Texts

• A Guide to Feynman Diagrams in the Many-Body Problem, 2nd Ed. R. D. Mattuck, Dover (1992).

• Quantum Theory of Many-Particle Systems, A. L. Fetter and J. D. Walecka, Dover (2003).

• Many-Body Theory of Solids: An Introduction, J. C. Inkson, Plenum Press (1984).

• Green’s Functions and Condensed Matter, G. Rickaysen, Academic Press (1991).

• Mathematical Methods for Physicists, 5th (Int’l) Ed. G. B. Arfken and H. J. Weber, Academic Press (2001)

• Elements of Green’s Functions and Propagation,G. Barton, Oxford (1989).

Page 3: Introduction to Linear Response Propagators and Green’s Functions

Overview

• Propagation of single particles or holes• Zero temperature formalism (c.f. finite temperature formalism)• Applicable in first principles, model Hamiltonian, … methods• Scattering of particles and holes from each other and external potentials

– Renormalisation of particle or hole energies– Finite lifetimes for particles or hole excitations (quasiparticles)– Particle-hole bound states (excitons, plasmons, magnons, …)

• New concepts such as – N-body Green’s functions particle, hole, particle-hole, particle-particle, …

propagators – Self energy (energy renormalisation and excitation lifetime)

Page 4: Introduction to Linear Response Propagators and Green’s Functions

Overview

• Approximations to propagators derived from expansion in (Feynman) diagrams or by functional derivative technique– Wick’s Theorem for evaluating time-ordered products of operators

– Lehmann Representation to extract retarded functions

• Integral equations which arise in the new concepts– Dyson’s equation G = Go + GoG

– Bethe-Salpeter Equation = o + o v

– Effective Potential V = v + v o V

• Applications of Concepts and Methods in Many-Particle Theory – Correlation Effects and the Total Energy (No-body propagator)

– Self-Energies and the GW approximation (1-body propagator)

– Collective Excitations and the Bethe-Salpeter Equation (2-body propagator)

Page 5: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Average value of variable A• Probability distribution in phase space

Probability that system is in infinitesimal region of phase space d

Elements of position p and momenta q

Average value of variable A

Density in phase space

1),(d

),P(d

),P(),(

),P(d

),)P(,A(dA

...dddd

...dddd

ddd

)d,P(

N21

N21

qp

qp

qpqp

qp

qpqp

qqq q

ppp p

qp

qp

p1, q1

p2, q2

Page 6: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Average value of variable A in NVT Canonical Ensemble

Probability depends on total energy of state

Canonical partition function normalises

Helmholtz free energy

Average value of variable A

E)/kT-(F

F/kT

)/kT,E(-

)/kT,E(-

)/kT,E(-

)/kT,E(-

)/kT,E(-

)/kT,E(-

AedA

ez

1

lnz kTF

edz

ed

)e,A(dA

ed

e),(

e),P(

qp

qp

qp

qp

qp

qp

qp

qp

qp

Page 7: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Correspondence with Quantum Mechanics

Density operator

Variables represented as operators

Complete set of states

Expectation value of A

Trace of A

Sum of diagonal elements with probability weights

Pure state

Mixed state

0p 1p

0p 1p

a pa pA

...Tr

ˆATrAˆTrA

aA

iji

iji

nnn

iji

ij

i

statesallnjiiin

nstatesall

n

i

ji

ii

Page 8: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Linear Response in Classical Mechanics

Hamiltonian contains Ho and perturbation B

Average value of A in absence of perturbation B = 1/kT

<A> changes in presence of perturbation B

Defines the linear response

BH-

BH-

H-

H-

H-

H-

o

o

0

o

o

o

o

o

ed

AedA

A

ed

AedAAA

ed

AedA

BHH

Page 9: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Linear Response in Classical Mechanics

Expansion of exponential with scalar exponent

Linear response function

ooo20

2

o

o

o

o

o

o

2

o

o

32

2

BAABV

UV'-VU'lim

BAABV

UV'-VU'

ed

e Bd

ed

eA d

ed

e ABd

V

UV'-VU'

e Bd V'

e ABd U'

...A3!

1A

2!

1A1e

V V'

U U'

V

UV'-VU'

V

U

B-H-

B-H-

B-H-

B-H-

B-H-

B-H-

B-H-

B-H-

A

Page 10: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Example of application of Static Linear Response

Molecule in gas phase with permanent dipole moment

-p.E term in Hamiltonian

<px>o vanishes in absence of field

p is magnitude of permanent moment

P is polarisation

is molecular susceptibility of gas phase

kT3

p

EkT3

pE

V3

NpP

3

pp

Epp

EpppEppEppp

EpB-

pA

2

x

2

x

2

x

2

o

2x

xo

2xx

xoxoxo

2xoxxoxoxxxx

xx

x

Page 11: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Time-dependent linear response

Relaxation after static perturbation in time domain

B(0)A(t)

B(0)A(t)A(t)

ed

e B(0)1A(t)d

ed

e A(t)dA(t)

Bt)(HH

0

o

o

o

H-

H-

H-

H-

• Consider a system where a steady perturbation –B is applied from t→-• The perturbation is switched off abruptly at t = 0• To obtain <A(t)> we must use the perturbed system density at t = 0 (full H)

(-t)

t

1

0

Page 12: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Time-dependent linear response• Onsager regression hypothesis

The way in which spontaneous fluctuations in a system relax back to

equilibrium is the same as the way in which a perturbed system relaxes

back to equilibrium, so long as the perturbation is small

Relaxation after static perturbation in time domain(t)(0)pp(t)

E(t)(0)pp(t)p

(t)(0)p(0)Ep(t)p

B(0)A(t)

xx

xxxx

xxxx

kT

Page 13: Introduction to Linear Response Propagators and Green’s Functions

B(0)A(t))(χ d)(χ dA(t)

0 t 0)f(t'

0 t)f(t'

t τ0 t' τ- t'dt'dτ

t'tτ

))f(t't'-(tχdt'A(t)

)t'-(tχ)t'(t,χ

t' t 0)t'(t,χ

))f(t't'(t,χdt'A(t)

t

AB

t

AB

t

-

AB

ABAB

AB

-

AB

Classical Statistical Mechanics

• Time-dependent linear response

Define the response function AB

Causality requires this for t < t’

Response depends only on time difference

Introduce change of variable

f(t’) switches off at t’ = 0

Linear response in terms of response function

Page 14: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Linear response function

Linear response function is correlation function

0 0)(χ

0 )(AB(0)-)(χ

(t)AB(0)B(0)A(t)dt

d

rule Leibniz' (t)χ)(χ ddt

d

B(0)A(t))(χ d

AB

.

AB

.

AB

t

AB

t

AB

Page 15: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Example: Mobility of particle in a fluid

Phenomenological relation

Force acting on charged particle in uniform field

Linear response approach

Need to know velocity auto-correlation function

Mobility

0

xx

..

0

xxx

0

x

.

x

0

x

.

xx

0

xvx

t

-

xvxx

xx

xx

)((0)vv d

τ)(tBA(t)τ)(t)B(tA )((0)vv dF

0τ)A(t)B(tdt

d )((0)vx dF

)(vx(0) dF (t)v

)(χ dF)t'-(tχdt'F(t)v

kxρ(x)φ(x)HqEF

F (t)v

xx

Page 16: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Example: Electric conductivity

Hamiltonian for charged particle in vector potential, A

Omit nonlinear A2 term

Define current density

Hamiltonian as Ho + perturbing part

t),(.t,dc

1HH

)'-((t)et),(c

1 U

2m

pH

AO.mc

eHH

UA2mc

e.

mc

e

2m

pH

Uc

e

2m

1H

o

.

pot.

2

o

2o

pot.2

2

22

pot.

2

rj)r A(r

rrvrj

AE

pA

pA

Ap

-

Page 17: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Example: Electric conductivity of harmonic oscillator e.g. phonon

Equation of motion

Set driving force to zero

Equation of motion in absence of driving force

Solution in absence of driving force

A, depend on initial conditions

Required for EoM to be satisfied – defines 1

t/2

t/2

eδ)tsin(2

-δ)tcos(A(t)x

4-

δ)tsin(Aex(t)

0xxx

0 F(t)

F(t)x mxmxm

111

.

22o

21

1

2o

...

2o

...

Page 18: Introduction to Linear Response Propagators and Green’s Functions

)t'-(t )t'-G(tdt

d

dt

d 2o2

2

Classical Statistical Mechanics

• Example: Electric conductivity of harmonic oscillator e.g. phonon

• Impulse Response Function

We can view the continuous force applied to a mass on a spring as

a sequence of delta function impulses. If we know the response of

the system to a single impulse, provided the system is linear, we

can immediately write down the solution in terms of the impulse

response function.

G(t-t’) is the Green’s functionDirac delta function is a unit impulse function

Velocity of oscillator when given an unit impulse at t = 0

t/2et)sin(2

-t)cos(m

1(t)x 111

1

.

Page 19: Introduction to Linear Response Propagators and Green’s Functions

)F(t'

m

e )t'(tsindt'x(t)

e )t(tsin Am

1x(t)

e )t'(tsin2

)t'(tcosm

1(t)

e )t'(tsinm

1x(t)

m

1

m

)F(t'dt'

m

)F(t'lim(t)dtxlim)t'(x)t'(x

t

0 1

1

ii1

ii

1

1111

11

t'

t'0

t'

t'

..

0

..

)t'-(t2Γ-

)t-(t2Γ-

)t'-(t2Γ-.

x

)t'-(t

2Γ-

Classical Statistical Mechanics

• Example: Electric conductivity of harmonic oscillator e.g. phonon

• Impulse Response Function

Unit impulse F(t’) = 1

t t+ t t+ t'

Position following unit impulseVelocity following unit impulsePosition following impulse seriesPosition following continuous force

Page 20: Introduction to Linear Response Propagators and Green’s Functions

)F(t')t't(m

e)t'(tsin

2-)t'(tcos dt'

)F(t')t'G(tdt'dt

d(t)x

)F(t')t'G(tdt'x(t)

)t'-(t )t'-G(tdt

d

dt

d

)t't(m

e )t'(tsin )t'-G(t

1111

t

t.

t

2o2

2

1

1

)t'-(t2Γ-

)t'-(t2Γ-

Classical Statistical Mechanics

• Example: Electric conductivity of harmonic oscillator e.g. phonon

• Impulse Response Function

Green’s function

Defining relation for G

Position, from G for a given F

Velocity, from G for a given F

Page 21: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Example: Electric conductivity of harmonic oscillator e.g. phonon

• Impulse Response Function

Change of variable

Example: impulse at t’=0

Recover original solution from impulse response function

)t(m

et)sin(

2-t)cos((t)x

)-(t)(m

e)sin(

2-)cos( d(t)x

)-(t)(t')F(t'

)-F(t)(m

e)sin(

2-)cos(ω d(t)x

0t t' - t'dt'd t't

)F(t')t't(mω

e)t'(tωsin

2-)t'(tωcosω dt'(t)x

1111

.

1111

0

.

1111

0

.

1111

t.

2Γt-

2Γ-

2Γ-

)t'-(t2Γ-

Page 22: Introduction to Linear Response Propagators and Green’s Functions

ii

i ii

i

ii

22o

22o

.

1111

0

.

tt

-t2Γ-

-tt'

e

dt

dRe

m

1e Re

m

1x

e )(m

e)sin(

2-)cos( dRe(t)x

eRe e Re)F(t'

Classical Statistical Mechanics

Example: Periodic forcing

• Example: Electric conductivity of harmonic oscillator e.g. phonon

• Impulse Response Function

If we set the lower limit in the integral dt’ to 0 instead of – we obtain additional transient terms which depend on initial conditions

Page 23: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

2Γt-

2Γt-

2Γ-

2)Γ(t-

x

x2Γt-

et)sin(2

-t)cos(m

kT(t)x(0)x

et)sin(2

-t)cos(1

m

kT(t)x

m

kT(0)x

2

(0)xm

2

kTE

e)sin(2

-)cos(2

e)(tsin2

-)(tcos

et)sin(2

-t)cos(dt)(t(t)vvdt

1111

..

1111

...

1111

111

111

00

xx

• Example: Electric conductivity of harmonic oscillator e.g. phonon

• Evaluate velocity auto-correlation function

Page 24: Introduction to Linear Response Propagators and Green’s Functions

T

t')t'(T

'

e f *f dT

ee )(t'f )t'-(Tf dt'dt)()g(g

dT dt t'-T tt'tT

e )(t'f dt'e (t)fdt )()g(g

e (t)fdt )(g

f*f)'(tf)t'-(tf dt'f*f

21

2121

2121

11

122121

i

ii

titi

ti

Classical Statistical Mechanics

• Example: Electric conductivity of harmonic oscillator e.g. phonon

• Parseval’s Theorem

Convolution of f1 and f2

Fourier transform of f1

Product of Fourier transforms

Fourier transform of convolution

Page 25: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Example: Electric conductivity of harmonic oscillator e.g. phonon

)().()( )(mV

e

e)(e

)sin(2

-)cos(dmV

e)G(

)-(tE)(e)sin(2

-)cos(m

kT

V

e)-(t).()((0)

V

e

)()((0).c

)-(td

V

e(t)

)((0))G( )'((t)et),(

))F(G()X( )-F(t)G( d)-F(t)G( d)F(t')t'G(tdt'x(t)

22o

2

1111

2

1111

22

0

.2

.

0

t

2Γ-

2Γ-

EJ

Evv

vvA

j

ABrrvrj

i

i

i

Page 26: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Example: Electric conductivity of harmonic oscillator e.g. phonon

Same result as Drude when o tends to zero

22

2

22

2

2

2

22

Drude

22

2

oSHO

1

1

m

e1

1

m

e

1

1

m

e

1

1

m

e)(

m

e)0;(

ii

i

i

i

Page 27: Introduction to Linear Response Propagators and Green’s Functions

Classical Statistical Mechanics

• Conclusions

• Linear response functions, e.g. transport coefficients, are derived

from correlation functions

• The correlation function is independent of the external stimulus

(Onsager)

• The reponse function contains the step function () to satisfy

causality