Skip to content

Correlation Function Definitions

A correlation function is not specified by the symbols ⟨AB⟩\langle A B\rangle alone. The state, operator order, time prescription, Fourier convention, normalization, statistics, and subtraction rule determine which physical object is being calculated.

This page is a convention ledger. It defines the principal correlator families side by side and gives the shortest reliable conversions among them. The linked canonical pages retain the derivations, asymptotic physics, analytic structure, measurement models, and numerical methods.

Before quoting a correlator, state:

  1. the state or ensemble defining ⟨⋅⟩\langle\cdot\rangle;
  2. the generator used for real or imaginary-time evolution;
  3. the operators, their order, and their fermion parity;
  4. whether one-point or other disconnected pieces are subtracted;
  5. lattice, continuum, volume, and internal-index normalizations;
  6. the real-time, imaginary-time, angular-frequency, or energy variable;
  7. the Fourier-transform sign and factors of 2π2\pi and ℏ\hbar;
  8. whether the object is raw, symmetrized, time ordered, retarded, advanced, or Matsubara;
  9. whether a plotted “spectral function” is an ordered spectrum, response density, single-particle spectrum, or measured intensity;
  10. the one-sided prescription at equal time and any coincident-point regulator.

Most disagreements between two correct formulas can be traced to one item in this ledger.

For a normalized density operator ρ\rho,

⟨X⟩ρ≡Tr⁡(ρX).\langle X\rangle_\rho \equiv \operatorname{Tr}(\rho X).

Real-time Heisenberg evolution under HH is

A(t)=eiHt/ℏAe−iHt/ℏ.A(t) = e^{iHt/\hbar} A e^{-iHt/\hbar}.

If

[ρ,H]=0,[\rho,H] = 0,

the state is stationary. A two-time correlator then depends only on the time difference:

⟨A(t)B(t′)⟩ρ=⟨A(t−t′)B(0)⟩ρ.\langle A(t)B(t')\rangle_\rho = \langle A(t-t')B(0)\rangle_\rho.

In a grand-canonical equilibrium state,

ρβ=e−β(H−μN)Tr⁡e−β(H−μN),\rho_\beta = \frac{ e^{-\beta(H-\mu N)} }{ \operatorname{Tr}e^{-\beta(H-\mu N)} },

stationarity under HH still holds when [H,N]=0[H,N]=0. Single-particle Green functions are also often evolved with H−μNH-\mu N so that frequencies are measured relative to μ\mu. That choice must be stated rather than inferred.

For a nonstationary state, use center and relative times,

T=t+t′2,tr=t−t′,T = \frac{t+t'}2, \qquad t_{\mathrm r} = t-t',

because a single-frequency transform in trt_{\mathrm r} does not capture the full TT dependence.

This sheet defines raw ordered products without hidden factors of ii or ℏ\hbar:

CAB>(t)≡⟨A(t)B(0)⟩,CAB<(t)≡⟨B(0)A(t)⟩.\begin{aligned} C_{AB}^{>}(t) &\equiv \langle A(t)B(0)\rangle, \\ C_{AB}^{<}(t) &\equiv \langle B(0)A(t)\rangle. \end{aligned}

For a stationary state,

CAB<(t)=CBA>(−t).C_{AB}^{<}(t) = C_{BA}^{>}(-t).

Complex conjugation gives

[CAB>(t)]∗=CA†B†<(t)=CB†A†>(−t).\begin{aligned} \left[ C_{AB}^{>}(t) \right]^* &= C_{A^\dagger B^\dagger}^{<}(t) \\ &= C_{B^\dagger A^\dagger}^{>}(-t). \end{aligned}

The greater and lesser labels refer to operator order, not numerical magnitude. They remain distinct even when AA and BB are Hermitian.

An equal-time correlator fixes both insertion times to one common value:

CAB(x,y;t)=⟨A(x,t)B(y,t)⟩.C_{AB}(\mathbf x,\mathbf y;t) = \left\langle A(\mathbf x,t) B(\mathbf y,t) \right\rangle.

Equal time does not imply:

  • equal position;
  • commuting operators;
  • connected subtraction;
  • stationarity;
  • normal ordering;
  • a probability distribution.

Common lattice examples are

Γij(1)=⟨di†dj⟩,Cnn(i,j)=⟨ninj⟩,CSSab(i,j)=⟨SiaSjb⟩,CΔ(i,j)=⟨Δi†Δj⟩.\begin{aligned} \Gamma_{ij}^{(1)} &= \langle d_i^\dagger d_j\rangle, \\ C_{nn}(i,j) &= \langle n_i n_j\rangle, \\ C_{SS}^{ab}(i,j) &= \langle S_i^a S_j^b\rangle, \\ C_{\Delta}(i,j) &= \langle\Delta_i^\dagger\Delta_j\rangle. \end{aligned}

They diagnose one-body coherence, density fluctuations, spin order, and pair coherence, respectively. The Equal-Time Correlations page owns their normalizations, contact terms, long-distance limits, and measurement interpretation.

Define fluctuation operators

δA≡A−⟨A⟩,δB≡B−⟨B⟩.\delta A \equiv A-\langle A\rangle, \qquad \delta B \equiv B-\langle B\rangle.

The connected two-point function in the displayed order is

CABc≡⟨δA δB⟩=⟨AB⟩−⟨A⟩⟨B⟩.\begin{aligned} C_{AB}^{c} &\equiv \langle\delta A\,\delta B\rangle \\ &= \langle AB\rangle - \langle A\rangle \langle B\rangle. \end{aligned}

At unequal times in a stationary state,

CABc(t)=⟨A(t)B(0)⟩−⟨A⟩⟨B⟩.C_{AB}^{c}(t) = \langle A(t)B(0)\rangle - \langle A\rangle \langle B\rangle.

Connected is not the same as symmetrized, retarded, irreducible in a diagrammatic channel, or entangled.

For a fixed three-operator order,

⟨ABC⟩c=⟨ABC⟩−⟨A⟩⟨BC⟩−⟨B⟩⟨AC⟩−⟨C⟩⟨AB⟩+2⟨A⟩⟨B⟩⟨C⟩.\begin{aligned} \langle ABC\rangle_c ={}& \langle ABC\rangle - \langle A\rangle\langle BC\rangle \\ &- \langle B\rangle\langle AC\rangle - \langle C\rangle\langle AB\rangle \\ &+ 2\langle A\rangle \langle B\rangle \langle C\rangle. \end{aligned}

Higher connected functions subtract every set partition into lower moments. In noncommuting problems, the sequence and source-order convention must remain fixed. The canonical treatment is Connected Correlation Functions.

For fluctuation operators,

CABsym(t)≡12⟨{δA(t),δB(0)}⟩.C_{AB}^{\mathrm{sym}}(t) \equiv \frac12 \left\langle \left\{ \delta A(t),\delta B(0) \right\} \right\rangle.

The commutator correlator is

CABcom(t)≡⟨[A(t),B(0)]⟩.C_{AB}^{\mathrm{com}}(t) \equiv \left\langle [A(t),B(0)] \right\rangle.

Connected subtraction does not change a commutator because scalar one-point functions commute. The symmetrized correlator measures fluctuations with the two orders averaged; the commutator controls causal response. They are related at thermal equilibrium by the fluctuation–dissipation theorem, not by an identity valid in every state.

For a stationary function F(t)F(t), this page uses angular frequency:

F(ω)=∫−∞∞dt eiωtF(t),F(t)=∫−∞∞dω2π e−iωtF(ω).\begin{aligned} F(\omega) &= \int_{-\infty}^{\infty} dt\, e^{i\omega t}F(t), \\ F(t) &= \int_{-\infty}^{\infty} \frac{d\omega}{2\pi}\, e^{-i\omega t}F(\omega). \end{aligned}

Energy transfer is

E=ℏω.E = \hbar\omega.

Changing from a spectrum per unit angular frequency to one per unit energy introduces a Jacobian:

SE(E)=1ℏSω(E/ℏ).S_E(E) = \frac1\hbar S_\omega(E/\hbar).

Never change the horizontal axis from ω\omega to EE without changing the units of the vertical axis.

Adopt the source convention

Hpert(t)=−f(t)B.H_{\mathrm{pert}}(t) = -f(t)B.

The response of AA is

δ⟨A(t)⟩=∫−∞∞dt′ χABR(t−t′)f(t′).\delta\langle A(t)\rangle = \int_{-\infty}^{\infty} dt'\, \chi_{AB}^{\mathrm R}(t-t') f(t').

The retarded susceptibility is

χABR(t)=iℏθ(t)⟨[A(t),B(0)]⟩.\chi_{AB}^{\mathrm R}(t) = \frac{i}{\hbar} \theta(t) \left\langle [A(t),B(0)] \right\rangle.

The sign follows from the displayed minus sign in HpertH_{\mathrm{pert}}. If a source is coupled as +fB+fB, the response kernel changes sign.

Retarded means

χABR(t)=0(t<0).\chi_{AB}^{\mathrm R}(t) = 0 \qquad (t<0).

It does not mean “any correlator with a small positive imaginary part,” and it is not the same object as a positive fluctuation spectrum.

The matching advanced function is

χABA(t)=−iℏθ(−t)⟨[A(t),B(0)]⟩.\chi_{AB}^{\mathrm A}(t) = -\frac{i}{\hbar} \theta(-t) \left\langle [A(t),B(0)] \right\rangle.

It has support only for t<0t<0. The retarded–advanced difference removes the step functions:

χABR(t)−χABA(t)=iℏ⟨[A(t),B(0)]⟩.\chi_{AB}^{\mathrm R}(t) - \chi_{AB}^{\mathrm A}(t) = \frac{i}{\hbar} \left\langle [A(t),B(0)] \right\rangle.

The Retarded and Advanced Response page owns adjoint identities, half-plane analyticity, dispersion relations, stability, and source-response derivations.

Let pA,pB∈{0,1}p_A,p_B\in\{0,1\} be the fermion parities of parity-homogeneous operators. Real-time graded ordering is

TA(t)B(t′)=θ(t−t′)A(t)B(t′)+(−1)pApBθ(t′−t)×B(t′)A(t).\begin{aligned} \mathcal T A(t)B(t') &={} \\[-0.2em] &\quad \theta(t-t')A(t)B(t') \\ &\quad+ (-1)^{p_Ap_B} \theta(t'-t) \\[-0.2em] &\qquad{}\times B(t')A(t). \end{aligned}

The generic time-ordered expectation is

CABT(t−t′)≡⟨TA(t)B(t′)⟩.C_{AB}^{\mathrm T}(t-t') \equiv \left\langle \mathcal T A(t)B(t') \right\rangle.

For t′=0t'=0,

CABT(t)=θ(t)CAB>(t)+(−1)pApBθ(−t)CAB<(t).\begin{aligned} C_{AB}^{\mathrm T}(t) ={}& \theta(t)C_{AB}^{>}(t) \\ &+ (-1)^{p_Ap_B} \theta(-t)C_{AB}^{<}(t). \end{aligned}

At t=0t=0, the value of θ(0)\theta(0) does not replace the need for one-sided limits. Canonical (anti)commutators can produce a jump or contact term.

Many-body Green functions often multiply CTC^{\mathrm T} by −i/ℏ-i/\hbar; relativistic QFT and some condensed-matter texts use different prefactors. Time ordering itself is the operator-ordering rule, not that conventional prefactor.

For canonical annihilation operators dad_a, define

η={+1,bosons,−1,fermions,\eta = \begin{cases} +1, & \text{bosons},\\ -1, & \text{fermions}, \end{cases}

and

[X,Y]η≡XY−ηYX.[X,Y]_\eta \equiv XY-\eta YX.

A common normal convention is

Gab>(t)=−iℏ⟨da(t)db†(0)⟩,Gab<(t)=−iηℏ⟨db†(0)da(t)⟩.\begin{aligned} G_{ab}^{>}(t) &= -\frac{i}{\hbar} \left\langle d_a(t)d_b^\dagger(0) \right\rangle, \\ G_{ab}^{<}(t) &= -\frac{i\eta}{\hbar} \left\langle d_b^\dagger(0)d_a(t) \right\rangle. \end{aligned}

Then

GabR(t)=θ(t)[Gab>(t)−Gab<(t)]=−iℏθ(t)⟨[da(t),db†]η⟩,\begin{aligned} G_{ab}^{\mathrm R}(t) &= \theta(t) \left[ G_{ab}^{>}(t)-G_{ab}^{<}(t) \right] \\ &= -\frac{i}{\hbar} \theta(t) \left\langle [d_a(t),d_b^\dagger]_\eta \right\rangle, \end{aligned}

and

GabA(t)=−θ(−t)[Gab>(t)−Gab<(t)]=iℏθ(−t)⟨[da(t),db†]η⟩.\begin{aligned} G_{ab}^{\mathrm A}(t) &= -\theta(-t) \left[ G_{ab}^{>}(t)-G_{ab}^{<}(t) \right] \\ &= \frac{i}{\hbar} \theta(-t) \left\langle [d_a(t),d_b^\dagger]_\eta \right\rangle. \end{aligned}

These single-particle retarded functions use a statistics-dependent bracket and a conventional −i/ℏ-i/\hbar prefactor. They should not be substituted into the observable-response formula without checking the channel and source convention. Green Functions in Many-Body QM owns the full family.

Let

K=H−μN,Lτ=βℏ.\mathcal K = H-\mu N, \qquad L_\tau = \beta\hbar.

Imaginary-time evolution is

A(τ)=eτK/ℏAe−τK/ℏ.A(\tau) = e^{\tau\mathcal K/\hbar} A e^{-\tau\mathcal K/\hbar}.

Graded imaginary-time ordering is

TτA(τ)B(τ′)=θ(τ−τ′)A(τ)B(τ′)+(−1)pApBθ(τ′−τ)×B(τ′)A(τ).\begin{aligned} \mathcal T_\tau A(\tau)B(\tau') &={} \\[-0.2em] &\quad \theta(\tau-\tau')A(\tau)B(\tau') \\ &\quad+ (-1)^{p_Ap_B} \theta(\tau'-\tau) \\[-0.2em] &\qquad{}\times B(\tau')A(\tau). \end{aligned}

A channel-dependent convention is

CAB(τ−τ′)≡sAB⟨TτA(τ)B(τ′)⟩β.\mathcal C_{AB}(\tau-\tau') \equiv s_{AB} \left\langle \mathcal T_\tau A(\tau)B(\tau') \right\rangle_\beta.

The fixed sign sABs_{AB} is a definition. The exchange sign (−1)pApB(-1)^{p_Ap_B} follows from fermion parity. For example:

  • a normal fermion propagator often uses sAB=−1s_{AB}=-1;
  • a normal boson propagator also often uses sAB=−1s_{AB}=-1;
  • a density or coordinate correlator often uses sAB=+1s_{AB}=+1.

For a two-point function whose transported insertion has fermion parity pp,

Cp(τ+Lτ)=(−1)pCp(τ).\mathcal C_p(\tau+L_\tau) = (-1)^p \mathcal C_p(\tau).

Thus even channels are periodic and odd single-particle channels are antiperiodic around the thermal circle.

Using Matsubara energies ζn\zeta_n,

C(iζn)=1ℏ∫0βℏdτ eiζnτ/ℏC(τ),C(τ)=1β∑ne−iζnτ/ℏC(iζn).\begin{aligned} \mathcal C(i\zeta_n) &= \frac1\hbar \int_0^{\beta\hbar} d\tau\, e^{i\zeta_n\tau/\hbar} \mathcal C(\tau), \\ \mathcal C(\tau) &= \frac1\beta \sum_n e^{-i\zeta_n\tau/\hbar} \mathcal C(i\zeta_n). \end{aligned}

The allowed grids are

ζnB=2πnβ,ζnF=(2n+1)πβ.\begin{aligned} \zeta_n^{\mathrm B} &= \frac{2\pi n}{\beta}, \\ \zeta_n^{\mathrm F} &= \frac{(2n+1)\pi}{\beta}. \end{aligned}

The corresponding angular frequencies are ωn=ζn/ℏ\omega_n=\zeta_n/\hbar. The canonical Bosonic and Fermionic Matsubara Frequencies page owns indexing, cutoff, and summation details.

Analytic continuation is not justified by replacing iζni\zeta_n with E+i0E+i0 in an arbitrary fitted formula. A spectral representation and control of static bosonic terms are required.

The word spectral is overloaded:

ObjectDefining ingredientTypical sign property
ordered transition spectrumFourier transform of ⟨A(t)B(0)⟩\langle A(t)B(0)\ranglenonnegative for B=A†B=A^\dagger
response spectral densityFourier transform of a commutatorgenerally signed
single-particle spectral functionretarded–advanced discontinuity of GGpositive semidefinite in standard fermionic channels
dynamic structure factormomentum-resolved ordered spectrumnonnegative in a conjugate channel
density of statestrace or momentum sum of a single-particle spectrumnonnegative in standard use
measured intensityprobe and detector forward modelnonnegative counts, but not a universal correlator

The formulas below use compatible conventions, but their units still differ.

Define

SAB(ω)≡12π∫−∞∞dt eiωtCAB>(t).\mathcal S_{AB}(\omega) \equiv \frac1{2\pi} \int_{-\infty}^{\infty} dt\, e^{i\omega t} C_{AB}^{>}(t).

Let

ρ=∑npn∣n⟩⟨n∣,H∣n⟩=En∣n⟩.\rho = \sum_n p_n \lvert n\rangle\langle n\rvert, \qquad H\lvert n\rangle = E_n\lvert n\rangle.

With Anm=⟨n∣A∣m⟩A_{nm}=\langle n\lvert A\rvert m\rangle and Bmn=⟨m∣B∣n⟩B_{mn}=\langle m\lvert B\rvert n\rangle,

SAB(ω)=∑n,mpnAnmBmn×δ ⁣(ω−Em−Enℏ).\begin{aligned} \mathcal S_{AB}(\omega) ={}& \sum_{n,m} p_n A_{nm}B_{mn} \\ &\times \delta\!\left( \omega-\frac{E_m-E_n}{\hbar} \right). \end{aligned}

If B=A†B=A^\dagger,

SAA†(ω)≥0\mathcal S_{AA^\dagger}(\omega) \ge 0

as a spectral measure. Its zeroth moment is the equal-time ordered correlator:

∫−∞∞dω SAB(ω)=⟨AB⟩.\int_{-\infty}^{\infty} d\omega\, \mathcal S_{AB}(\omega) = \langle AB\rangle.

At thermal equilibrium, detailed balance is

SBA(−ω)=e−βℏωSAB(ω).\mathcal S_{BA}(-\omega) = e^{-\beta\hbar\omega} \mathcal S_{AB}(\omega).

This relation compares reversed operator order and reversed frequency. It is not a statement that one arbitrary cross-spectrum is even or positive.

For the observable response convention above, define

ρAB(ω)≡12πℏ∫−∞∞dt eiωt×⟨[A(t),B(0)]⟩.\begin{aligned} \rho_{AB}(\omega) \equiv{}& \frac{1}{2\pi\hbar} \int_{-\infty}^{\infty} dt\, e^{i\omega t} \\ &\times \left\langle [A(t),B(0)] \right\rangle. \end{aligned}

It is the difference of ordered spectra:

ρAB(ω)=1ℏ[SAB(ω)−SBA(−ω)].\rho_{AB}(\omega) = \frac1\hbar \left[ \mathcal S_{AB}(\omega) - \mathcal S_{BA}(-\omega) \right].

The retarded–advanced discontinuity is

χABR(ω)−χABA(ω)=2πi ρAB(ω).\chi_{AB}^{\mathrm R}(\omega) - \chi_{AB}^{\mathrm A}(\omega) = 2\pi i\, \rho_{AB}(\omega).

For a conjugate Hermitian equilibrium channel in this sign convention,

Im⁡χAAR(ω)=πρAA(ω).\operatorname{Im} \chi_{AA}^{\mathrm R}(\omega) = \pi\rho_{AA}(\omega).

Detailed balance gives the equilibrium preview

ρAB(ω)=1−e−βℏωℏSAB(ω).\rho_{AB}(\omega) = \frac{ 1-e^{-\beta\hbar\omega} }{\hbar} \mathcal S_{AB}(\omega).

The Fluctuation–Dissipation Theorem owns the full conversion to symmetrized spectra and the zero-frequency limit.

For a normal fermionic Green function written per unit energy,

Aab(E)≡i2π[GabR(E)−GabA(E)].A_{ab}(E) \equiv \frac{i}{2\pi} \left[ G_{ab}^{\mathrm R}(E) - G_{ab}^{\mathrm A}(E) \right].

For a diagonal channel,

Aaa(E)=−1πIm⁡GaaR(E).A_{aa}(E) = -\frac1\pi \operatorname{Im} G_{aa}^{\mathrm R}(E).

Canonical anticommutation gives

∫−∞∞dE Aaa(E)=1\int_{-\infty}^{\infty} dE\, A_{aa}(E) = 1

for a complete normalized orbital. This positive single-particle spectrum is not the signed response density ρAB\rho_{AB}, even though both are retarded–advanced discontinuities.

For a Hermitian lattice operator OjO_j, define

Oq≡∑j=1Le−iq⋅rjOj.O_{\mathbf q} \equiv \sum_{j=1}^{L} e^{-i\mathbf q\cdot\mathbf r_j} O_j.

With

δOq≡Oq−⟨Oq⟩,\delta O_{\mathbf q} \equiv O_{\mathbf q} - \langle O_{\mathbf q}\rangle,

one connected static convention is

SOc(q)≡1L⟨δOqδO−q⟩.S_O^c(\mathbf q) \equiv \frac1L \left\langle \delta O_{\mathbf q} \delta O_{-\mathbf q} \right\rangle.

For O−q=Oq†O_{-\mathbf q}=O_{\mathbf q}^\dagger,

SOc(q)≥0.S_O^c(\mathbf q) \ge 0.

At q=0\mathbf q=0, a density structure factor measures total-number fluctuations under this normalization. It vanishes in an exact fixed-number state but need not vanish in a grand-canonical ensemble.

A compatible angular-frequency convention is

SOc(q,ω)≡12πL∫−∞∞dt eiωt×⟨δOq(t)δO−q(0)⟩.\begin{aligned} S_O^c(\mathbf q,\omega) \equiv{}& \frac1{2\pi L} \int_{-\infty}^{\infty} dt\, e^{i\omega t} \\ &\times \left\langle \delta O_{\mathbf q}(t) \delta O_{-\mathbf q}(0) \right\rangle. \end{aligned}

Its zeroth-moment relation is

∫−∞∞dω SOc(q,ω)=SOc(q).\int_{-\infty}^{\infty} d\omega\, S_O^c(\mathbf q,\omega) = S_O^c(\mathbf q).

The factor 1/L1/L, the sign in OqO_{\mathbf q}, and the factor 1/(2π)1/(2\pi) vary across disciplines. Scattering cross sections also include probe form factors, polarization tensors, kinematic prefactors, resolution, and background. The canonical home is Structure Factors.

With internal indices, a spectrum is a matrix:

S(ω)=[Sab(ω)].\mathbf S(\omega) = \left[ S_{ab}(\omega) \right].

Positivity means

v†S(ω)v≥0\mathbf v^\dagger \mathbf S(\omega) \mathbf v \ge 0

for every vector v\mathbf v in a conjugate ordered channel. It does not require each off-diagonal entry to be real or positive.

Hermiticity, reciprocity, time reversal, inversion, and point-group symmetry impose different index and momentum relations. Apply only the symmetries actually possessed by the state, Hamiltonian, operators, and boundary conditions.

For canonical modes,

⟨aiaj†⟩=δij+⟨aj†ai⟩,⟨cicj†⟩=δij−⟨cj†ci⟩.\begin{aligned} \langle a_i a_j^\dagger\rangle &= \delta_{ij} + \langle a_j^\dagger a_i\rangle, \\ \langle c_i c_j^\dagger\rangle &= \delta_{ij} - \langle c_j^\dagger c_i\rangle. \end{aligned}

These relations fix jumps in one-sided single-particle Green functions. In a continuum,

δij⟶δ(d)(x−y),\delta_{ij} \longrightarrow \delta^{(d)}(\mathbf x-\mathbf y),

so coincident-point expressions can be distributional or ultraviolet divergent. Normal ordering removes specified contractions; it is not a universal instruction to discard every contact term.

ChannelOperatorsTypical object
densityδnq\delta n_{\mathbf q}, δn−q\delta n_{-\mathbf q}structure factor and density response
spinSqaS_{\mathbf q}^a, S−qbS_{-\mathbf q}^bmagnetic structure factor and susceptibility
currentJaJ^a, JbJ^bconductivity kernel with possible contact term
single particledkd_{\mathbf k}, dk†d_{\mathbf k}^\daggerGreen function and addition/removal spectrum
pairΔq\Delta_{\mathbf q}, Δq†\Delta_{\mathbf q}^\daggerpair susceptibility and coherence
order parameterδOq\delta O_{\mathbf q}, δO−q\delta O_{-\mathbf q}static scaling and collective spectrum

The same operator pair can generate a raw correlator, response, Matsubara function, or structure factor. The time prescription and normalization select the physical question.

  • Static limit: lim⁡ω→0χ(q,ω)\lim_{\omega\to0}\chi(\mathbf q,\omega) can depend on whether q→0\mathbf q\to0 is taken first.
  • Equal-time limit: t→0+t\to0^+ and t→0−t\to0^- can differ by a canonical contact term.
  • Analytic continuation: finite Matsubara data do not determine a stable real-frequency spectrum without additional information.
  • Thermodynamic limit: finite systems have discrete lines and recurrences; continua and irreversible widths emerge only after appropriate limits or coupling to an environment.
  • Connected subtraction: removing ⟨A⟩⟨B⟩\langle A\rangle\langle B\rangle eliminates a zero-frequency elastic piece only when the state is stationary and the normalization is matched.
  • Broadening: replacing δ\delta functions by Lorentzians or Gaussians is a visualization or physical-resolution model, not an exact identity.
  • Energy origin: number-changing spectra shift when evolution changes between HH and H−μNH-\mu N.
  1. Set t=0±t=0^\pm and recover the canonical (anti)commutator jump.
  2. Integrate a dynamic spectrum and recover its equal-time correlator.
  3. Check that a conjugate ordered spectrum and structure factor are nonnegative.
  4. Verify detailed balance only in a thermal stationary state.
  5. Confirm that the retarded kernel vanishes for negative time.
  6. Confirm that the advanced kernel vanishes for positive time.
  7. Check the dimensions after changing from ω\omega to E=ℏωE=\hbar\omega.
  8. Separate fixed convention signs from fermionic exchange signs.
  9. Reconstruct a known free mode or two-level benchmark before analyzing interacting data.
  10. Report finite-time windows, artificial broadening, and resolution functions.
  • Calling every ⟨A(t)B(0)⟩\langle A(t)B(0)\rangle a Green function.
  • Calling a time-ordered function retarded because both contain step functions.
  • Using an anticommutator in an observable Kubo response merely because the microscopic fields are fermionic.
  • Omitting the source sign when defining a susceptibility.
  • Treating connected and symmetrized correlators as synonyms.
  • Assuming a response spectral density is nonnegative at all frequencies.
  • Comparing spectra per unit energy and per unit angular frequency without the factor of ℏ\hbar.
  • Forgetting the fermionic sign in real or imaginary-time ordering.
  • Setting t=0t=0 without declaring a one-sided prescription.
  • Applying thermal detailed balance to a quenched or driven state.
  • Inferring every Hamiltonian eigenvalue from one operator-resolved spectrum.
  • Interpreting broadened finite-size lines as intrinsic lifetimes without a scaling or resolution analysis.
  • G. D. Mahan, Many-Particle Physics, 3rd ed., Springer (2000) — real-time Green functions, response, spectral functions, and many-body conventions.
  • A. L. Fetter and J. D. Walecka, Quantum Theory of Many-Particle Systems, Dover (2003) — Lehmann representations, propagators, response, and sum rules.
  • J. W. Negele and H. Orland, Quantum Many-Particle Systems, Westview Press (1998) — imaginary-time functions, Matsubara transforms, and functional methods.
  • P. C. Martin and J. Schwinger, “Theory of Many-Particle Systems. I,” Physical Review 115, 1342–1373 (1959) — equilibrium Green-function hierarchy and thermal boundary relations.
  • R. Kubo, “Statistical-Mechanical Theory of Irreversible Processes. I,” Journal of the Physical Society of Japan 12, 570–586 (1957) — linear response and fluctuation relations.
  • L. Van Hove, “Correlations in Space and Time and Born Approximation Scattering in Systems of Interacting Particles,” Physical Review 95, 249–262 (1954) — dynamic correlations and scattering.
  • S. W. Lovesey, Theory of Neutron Scattering from Condensed Matter, Vol. 1, Clarendon Press (1984) — structure factors, response channels, and experimental normalization.

1. Show shift invariance of a connected correlator

Section titled “1. Show shift invariance of a connected correlator”

Let A′=A+aIA'=A+aI and B′=B+bIB'=B+bI. Show that

⟨A′B′⟩c=⟨AB⟩c.\langle A'B'\rangle_c = \langle AB\rangle_c.
Solution

The shifted fluctuation operators are

δA′=A+aI−⟨A+aI⟩=A−⟨A⟩=δA,δB′=δB.\begin{aligned} \delta A' &= A+aI-\langle A+aI\rangle \\ &= A-\langle A\rangle = \delta A, \\ \delta B' &= \delta B. \end{aligned}

Therefore

⟨A′B′⟩c=⟨δA′δB′⟩=⟨δA δB⟩=⟨AB⟩c.\begin{aligned} \langle A'B'\rangle_c &= \langle\delta A'\delta B'\rangle \\ &= \langle\delta A\,\delta B\rangle \\ &= \langle AB\rangle_c. \end{aligned}

Connected two-point functions are insensitive to additive scalar offsets.

In a stationary state, prove

CAB<(t)=CBA>(−t).C_{AB}^{<}(t) = C_{BA}^{>}(-t).

Then write the time-ordered correlator for two odd fermionic operators.

Solution

By definition,

CBA>(−t)=⟨B(−t)A(0)⟩.C_{BA}^{>}(-t) = \langle B(-t)A(0)\rangle.

Stationarity allows both times to be shifted by tt:

⟨B(−t)A(0)⟩=⟨B(0)A(t)⟩=CAB<(t).\begin{aligned} \langle B(-t)A(0)\rangle &= \langle B(0)A(t)\rangle \\ &= C_{AB}^{<}(t). \end{aligned}

If pA=pB=1p_A=p_B=1, exchanging the operators under time ordering contributes a minus sign:

CABT(t)=θ(t)CAB>(t)−θ(−t)CAB<(t).\begin{aligned} C_{AB}^{\mathrm T}(t) ={}& \theta(t)C_{AB}^{>}(t) \\ &- \theta(-t)C_{AB}^{<}(t). \end{aligned}

3. Derive the retarded–advanced discontinuity

Section titled “3. Derive the retarded–advanced discontinuity”

Starting from the time-domain definitions, show that

χABR(ω)−χABA(ω)=2πi ρAB(ω).\chi_{AB}^{\mathrm R}(\omega) - \chi_{AB}^{\mathrm A}(\omega) = 2\pi i\,\rho_{AB}(\omega).
Solution

In time,

χABR(t)−χABA(t)=iℏ[θ(t)+θ(−t)]×⟨[A(t),B]⟩.\begin{aligned} \chi_{AB}^{\mathrm R}(t) - \chi_{AB}^{\mathrm A}(t) &= \frac{i}{\hbar} \left[ \theta(t)+\theta(-t) \right] \\ &\quad\times \langle[A(t),B]\rangle. \end{aligned}

Away from the convention-dependent point t=0t=0, the bracket of step functions is one. Hence

χABR(t)−χABA(t)=iℏ⟨[A(t),B]⟩.\chi_{AB}^{\mathrm R}(t) - \chi_{AB}^{\mathrm A}(t) = \frac{i}{\hbar} \langle[A(t),B]\rangle.

Fourier transformation and the definition

ρAB(ω)=12πℏ∫dt eiωt⟨[A(t),B]⟩\rho_{AB}(\omega) = \frac{1}{2\pi\hbar} \int dt\, e^{i\omega t} \langle[A(t),B]\rangle

give the result.

Use the Lehmann representation to show that SAA†(ω)\mathcal S_{AA^\dagger}(\omega) is nonnegative and that its integral equals ⟨AA†⟩\langle AA^\dagger\rangle.

Solution

Set B=A†B=A^\dagger in the ordered spectrum:

SAA†(ω)=∑n,mpn∣Anm∣2×δ ⁣(ω−Em−Enℏ).\begin{aligned} \mathcal S_{AA^\dagger}(\omega) ={}& \sum_{n,m} p_n |A_{nm}|^2 \\ &\times \delta\!\left( \omega-\frac{E_m-E_n}{\hbar} \right). \end{aligned}

Every pnp_n and ∣Anm∣2|A_{nm}|^2 is nonnegative, so the spectral measure is nonnegative.

Integrating over frequency removes the delta function:

∫dω SAA†(ω)=∑n,mpn∣Anm∣2=∑npn⟨n∣AA†∣n⟩=⟨AA†⟩.\begin{aligned} \int d\omega\, \mathcal S_{AA^\dagger}(\omega) &= \sum_{n,m} p_n|A_{nm}|^2 \\ &= \sum_n p_n \langle n|AA^\dagger|n\rangle \\ &= \langle AA^\dagger\rangle. \end{aligned}

Show that the dynamic definition on this page satisfies

∫dω SOc(q,ω)=SOc(q).\int d\omega\, S_O^c(\mathbf q,\omega) = S_O^c(\mathbf q).
Solution

Insert the dynamic definition and use

∫−∞∞dω2πeiωt=δ(t).\int_{-\infty}^{\infty} \frac{d\omega}{2\pi} e^{i\omega t} = \delta(t).

Then

∫dω SOc(q,ω)=1L∫dt δ(t)×⟨δOq(t)δO−q(0)⟩=1L⟨δOqδO−q⟩=SOc(q).\begin{aligned} \int d\omega\, S_O^c(\mathbf q,\omega) &= \frac1L \int dt\, \delta(t) \\ &\quad\times \left\langle \delta O_{\mathbf q}(t) \delta O_{-\mathbf q}(0) \right\rangle \\ &= \frac1L \left\langle \delta O_{\mathbf q} \delta O_{-\mathbf q} \right\rangle \\ &= S_O^c(\mathbf q). \end{aligned}

Let

C(τ+βℏ)=ξC(τ),ξ=±1.\mathcal C(\tau+\beta\hbar) = \xi\mathcal C(\tau), \qquad \xi=\pm1.

Find the allowed Matsubara energies in e−iζnτ/ℏe^{-i\zeta_n\tau/\hbar}.

Solution

The basis function must obey the same boundary condition:

e−iζn(τ+βℏ)/ℏ=ξe−iζnτ/ℏ.e^{-i\zeta_n(\tau+\beta\hbar)/\hbar} = \xi e^{-i\zeta_n\tau/\hbar}.

After canceling the common factor,

e−iβζn=ξ.e^{-i\beta\zeta_n} = \xi.

For ξ=+1\xi=+1,

βζn=2πn,\beta\zeta_n = 2\pi n,

so

ζnB=2πnβ.\zeta_n^{\mathrm B} = \frac{2\pi n}{\beta}.

For ξ=−1\xi=-1,

βζn=(2n+1)π,\beta\zeta_n = (2n+1)\pi,

so

ζnF=(2n+1)πβ.\zeta_n^{\mathrm F} = \frac{(2n+1)\pi}{\beta}.