Opening the book…
The preceding chapters established the statistical mechanics of equilibrium fluctuations. We now develop the dynamical theory: how a system responds to a weak external perturbation, and how that response is entirely encoded in the spontaneous equilibrium fluctuations. Three complementary routes to the same result are presented in sequence. First, we derive the Kubo formula directly from Liouville dynamics-the most fundamental and assumption-free path. Second, we approach the same physics through the Generalized Langevin Equation (GLE), making contact with friction kernels, noise, and causality. Third, we reverse the logic and show how the Kubo relation follows from the Fluctuation-Dissipation Theorem (FDT) together with the Wiener-Khinchin theorem, yielding the Green-Kubo relations for transport coefficients as a corollary. All three routes are classical; the quantum generalization is treated in the final chapter.
7.1 Correlation Functions, Susceptibility, and the FDT: Overview
The "memory" of a system's fluctuations is captured by the equilibrium auto-correlation function of an observable \(A(t)\) :
\[ C(t)=\langle A(t) A(0)\rangle . \tag{7.1} \label{7.1} \]
When a small time-dependent external force \(h(t)\) couples linearly to \(A\), the system's mean response is governed by the causal susceptibility \(\chi(t)\) :
\[ \langle\Delta A(t)\rangle=\int_{-\infty}^{t} \chi\left(t-t^{\prime}\right) h\left(t^{\prime}\right) d t^{\prime} \tag{7.2} \label{7.2} \]
In the frequency domain, \(\chi(\omega)=\chi^{\prime}(\omega)+i \chi^{\prime \prime}(\omega)\); the imaginary part \(\chi^{\prime \prime}(\omega)>0\) represents the energy irreversibly dissipated by the system. In the classical limit \(\hbar \omega \ll k_{B} T\), the FDT asserts:
\[ \chi^{\prime \prime}(\omega)=\frac{\beta \omega}{2} S(\omega), \tag{7.3} \label{7.3} \]
where \(\beta=\left(k_{B} T\right)^{-1}\) and \(S(\omega)=2 \int_{0}^{\infty} C(t) \cos (\omega t) d t\) is the one-sided power spectral density. The following three sections prove this relation by different methods, each illuminating a different facet of the physics.
7.2 Derivation I: Microscopic Route via Liouville Dynamics
7.2.1 Poisson Brackets and the Liouville Equation
We consider a classical system whose state is a point \(\Gamma=(\mathbf{q}, \mathbf{p})\) in phase space. The Poisson bracket of two phase-space functions \(f\) and \(g\) is
\[ \{f, g\}=\sum_{i}\left(\frac{\partial f}{\partial q_{i}} \frac{\partial g}{\partial p_{i}}-\frac{\partial f}{\partial p_{i}} \frac{\partial g}{\partial q_{i}}\right) . \tag{7.4} \label{7.4} \]
The phase-space probability density \(\rho(\Gamma, t)\) evolves according to the Liouville equation:
\[ \frac{\partial \rho}{\partial t}=\{H, \rho\} \equiv-L \rho, \tag{7.5} \label{7.5} \]
where \(L \equiv-\{H, \cdot\}\) is the Liouville operator. We apply a perturbation \(\Delta H=-A(\Gamma) h(t)\), so the full Hamiltonian is \(H=H_{0}-A h(t)\).
7.2.2 A Key Thermal Identity
The unperturbed equilibrium distribution is the canonical ensemble \(\rho_{0}=e^{-\beta H_{0}} / Z_{\text {part }}\), where \(Z_{\text {part }}=\int e^{-\beta H_{0}(\Gamma)} d \Gamma\). Using the chain rule for Poisson brackets, \(\{A, f(g)\}=(d f / d g)\{A, g\}\), which follows from \(\partial f(g) / \partial p_{i}=(d f / d g)\left(\partial g / \partial p_{i}\right)\) and likewise for \(q_{i}\), we compute
\[ \left\{A, \rho_{0}\right\}=\left\{A, \frac{e^{-\beta H_{0}}}{Z_{\text {part }}}\right\}=\frac{1}{Z_{\text {part }}} \frac{d\left(e^{-\beta H_{0}}\right)}{d H_{0}}\left\{A, H_{0}\right\}=-\beta \rho_{0}\left\{A, H_{0}\right\} . \tag{7.6} \label{7.6} \]
Since Hamilton's equations give \(\dot{A}=\left\{A, H_{0}\right\}\), this yields the fundamental thermal identity:
\[ \left\{A, \rho_{0}\right\}=-\beta \rho_{0} \dot{A} . \tag{7.7} \label{7.7} \]
7.2.3 First-Order Perturbation Theory
Writing \(\rho(t)=\rho_{0}+\Delta \rho(t)\) and linearising the Liouville equation in \(h(t)\) :
\[ \frac{\partial \Delta \rho}{\partial t}=\left\{H_{0}, \Delta \rho\right\}+\left\{-A h(t), \rho_{0}\right\} . \tag{7.8} \label{7.8} \]
The second term becomes \(h(t) \beta \rho_{0} \dot{A}\) via Equation \ref{7.7}. Solving with the propagator \(e^{L_{0}\left(t-t^{\prime}\right)}\) :
\[ \Delta \rho(t)=\beta \rho_{0} \int_{-\infty}^{t} h\left(t^{\prime}\right) \dot{A}\left(t^{\prime}-t\right) d t^{\prime} \tag{7.9} \label{7.9} \]
The mean deviation \(\langle\Delta A(t)\rangle=\int A(\Gamma) \Delta \rho(t) d \Gamma\). Shifting time evolution to the observable (passing to the Heisenberg picture) and using the stationarity of the equilibrium average:
\[ \langle\Delta A(t)\rangle=\int_{-\infty}^{t} h\left(t^{\prime}\right) \beta\left\langle\dot{A}(0) A\left(t-t^{\prime}\right)\right\rangle_{0} d t^{\prime} \tag{7.10} \label{7.10} \]
By time-translation invariance, \(\frac{d}{d t}\langle A(0) A(t)\rangle=\langle A(0) \dot{A}(t)\rangle=-\langle\dot{A}(0) A(t)\rangle\), so we identify the susceptibility as
\[ \chi\left(t-t^{\prime}\right)=-\beta \frac{d}{d t} C\left(t-t^{\prime}\right) \theta\left(t-t^{\prime}\right) . \tag{7.11} \label{7.11} \]
This is the Kubo formula in the time domain: the linear response function is the negative \(\beta\)-scaled time derivative of the equilibrium correlation function, multiplied by the Heaviside step \(\theta\) to enforce causality.
7.2.4 Connection to the Spectral Density
Taking the Fourier transform of Equation \ref{7.11}:
\[ \chi(\omega)=-\beta \int_{0}^{\infty} \dot{C}(t) e^{i \omega t} d t \tag{7.12} \label{7.12} \]
Integration by parts, with \(C(\infty) \rightarrow 0\), gives \(\chi(\omega)=\beta C(0)+i \omega \beta \int_{0}^{\infty} C(t) e^{i \omega t} d t\). The imaginary part is:
\[ \chi^{\prime \prime}(\omega)=\omega \beta \int_{0}^{\infty} C(t) \cos (\omega t) d t=\frac{\beta \omega}{2} S(\omega) \tag{7.13} \label{7.13} \]
recovering the classical FDT stated in Section 7.1. The quantum generalization, which replaces the factor \(\beta \omega / 2\) with \(\left(1-e^{-\beta \hbar \omega}\right) /(2 \hbar)\), is treated in Kubo [4] and in the final chapter of this book.
7.3 Derivation II: The Generalized Langevin Equation Route
The Liouville derivation is exact but abstract. A complementary, more physical route proceeds through the stochastic equation of motion for a distinguished degree of freedom embedded in a thermal bath.
7.3.1 The Generalized Langevin Equation
Consider a particle of mass \(m\) in a harmonic potential with intrinsic frequency \(\omega_{0}\), subject to non-Markovian friction described by a memory kernel \(\gamma(t)\) [3, 5, 3]. The equation of motion is:
\[ m \ddot{x}(t)+m \omega_{0}^{2} x(t)+m \int_{-\infty}^{t} \gamma\left(t-t^{\prime}\right) \dot{x}\left(t^{\prime}\right) d t^{\prime}=F_{\mathrm{ext}}(t)+f(t), \tag{7.14} \label{7.14} \]
where \(F_{\text {ext }}(t)\) is a deterministic external force and \(f(t)\) is the random (fluctuating) force exerted by the bath. The memory kernel encodes the retarded friction: the drag at time \(t\) depends on the particle's velocity at all earlier times \(t^{\prime}<t\).
7.3.2 Mean Response and the Deterministic Equation
Taking the ensemble average of Eq. (B.10) and using \(\langle f(t)\rangle=0\), the mean displacement \(\bar{x}(t)=\) \(\langle x(t)\rangle\) satisfies:
\[ m \ddot{\bar{x}}(t)+m \omega_{0}^{2} \bar{x}(t)+m \int_{-\infty}^{t} \gamma\left(t-t^{\prime}\right) \dot{\bar{x}}\left(t^{\prime}\right) d t^{\prime}=F_{\mathrm{ext}}(t) . \tag{7.15} \label{7.15} \]
The random force drops out entirely at this level: it governs fluctuations, not the mean response. This separation is the point of departure for the two Fluctuation-Dissipation relations below.
7.3.3 Susceptibility and the Green's Function
The causal Green's function \(G(t)\) (also called the response function) is defined as the mean displacement produced by an impulsive force \(F_{\text {ext }}(t)=\delta(t)\), with \(G(t)=0\) for \(t<0\). The susceptibility is its one-sided Fourier transform:
\[ \chi(\omega)=\int_{0}^{\infty} G(t) e^{i \omega t} d t \tag{7.16} \label{7.16} \]
Fourier-transforming Equation \ref{7.15} and denoting the transform of \(\gamma(t)\) by \(\tilde{\gamma}(\omega)\), one finds
\[ \chi(\omega)=\frac{1}{m\left(\omega_{0}^{2}-\omega^{2}\right)-i \omega m \tilde{\gamma}(\omega)} . \tag{7.17} \label{7.17} \]
The complex mobility \(\mu(\omega)=-i \omega \chi(\omega)\) relates the force to the mean velocity in the frequency domain.
7.3.4 The Two Fluctuation-Dissipation Relations
The noise \(f(t)\), which vanishes on average, carries statistical information that is tightly constrained by thermal equilibrium.
Second FDT (noise-friction relation). The power spectral density of the random force, \(S_{f f}(\omega)=\int_{-\infty}^{\infty}\langle f(t) f(0)\rangle e^{i \omega t} d t\), is related to the real part of the frequency-dependent friction by:
\[ S_{f f}(\omega)=2 k_{B} T m \operatorname{Re}[\tilde{\gamma}(\omega)] . \tag{7.18} \label{7.18} \]
This is the Second Fluctuation-Dissipation Theorem: the spectral density of the bath noise is proportional to the dissipative part of the friction kernel. For Markovian (white-noise) friction \(\gamma(t)=2 \gamma_{0} \delta(t), \operatorname{Re}[\tilde{\gamma}]=\gamma_{0}\) and one recovers the Langevin relation \(\langle f(t) f(0)\rangle=2 \gamma_{0} k_{B} T \delta(t)\).
First FDT (Kubo formula for mobility). The complementary relation connects the response (mobility) to the equilibrium velocity autocorrelation:
\[ \mu(\omega)=\frac{1}{k_{B} T} \int_{0}^{\infty} e^{i \omega t}\langle v(t) v(0)\rangle_{\text {eq }} d t \tag{7.19} \label{7.19} \]
This is the First Fluctuation-Dissipation Theorem (equivalently, the Kubo formula for the mobility): the admittance of the system equals the \(\beta\)-weighted one-sided transform of the velocity autocorrelation. Taking \(\omega \rightarrow 0\) yields the Einstein-Green-Kubo formula for the DC mobility, connecting it to the integral of the velocity autocorrelation.
7.3.5 Causality and the Kramers-Kronig Relations
Causality requires \(G(t)=0\) for \(t<0\). This condition is equivalent to \(\chi(\omega)\) being analytic in the closed upper half of the complex frequency plane. By Cauchy's theorem applied to the contour integral of \(\chi\left(\omega^{\prime}\right) /\left(\omega^{\prime}-\omega\right)\), one obtains the Kramers-Kronig relations:
\[ \begin{align*} & \operatorname{Re}[\chi(\omega)]=\frac{2}{\pi} \mathcal{P} \int_{0}^{\infty} \frac{\omega^{\prime} \operatorname{Im}\left[\chi\left(\omega^{\prime}\right)\right]}{\omega^{\prime 2}-\omega^{2}} d \omega^{\prime}, \tag{7.20} \label{7.20}\\[4pt] & \operatorname{Im}[\chi(\omega)]=-\frac{2 \omega}{\pi} \mathcal{P} \int_{0}^{\infty} \frac{\operatorname{Re}\left[\chi\left(\omega^{\prime}\right)\right]}{\omega^{\prime 2}-\omega^{2}} d \omega^{\prime}, \tag{7.21} \end{align*} \]
where \(\mathcal{P}\) denotes the Cauchy principal value. These relations are model-independent consequences of causality: a measurement of \(\chi^{\prime \prime}(\omega)\) over all frequencies uniquely determines \(\chi^{\prime}(\omega)\), and vice versa. Combined with the FDT, they imply that the full equilibrium noise spectrum determines both the reactive and dissipative parts of the linear response.
7.4 Derivation III: From the FDT and Wiener-Khinchin to the Kubo Formula
The two preceding sections derived the FDT from microscopic dynamics. It is illuminating to run the argument in the opposite direction: assuming the classical FDT as a postulate (or as experimentally established), one can derive the Kubo relation and the Green-Kubo transport formulas. This route is the one most directly useful for practical calculations.
7.4.1 Starting Point: The Classical FDT
In the classical limit \(k_{B} T \gg \hbar \omega\), the FDT relates the power spectral density \(S_{A A}(\omega)\) of an observable \(A(t)\) to the dissipative part of the susceptibility:
\[ S_{A A}(\omega)=\frac{2 k_{B} T}{\omega} \operatorname{Im}[\chi(\omega)] . \tag{7.22} \label{7.22} \]
7.4.2 Spectral Density via the Wiener-Khinchin Theorem
The Wiener-Khinchin theorem [18, 19] states that the power spectral density is the Fourier transform of the equilibrium autocorrelation function \(C_{A A}(t)=\langle A(t) A(0)\rangle\) :
\[ S_{A A}(\omega)=\int_{-\infty}^{\infty}\langle A(t) A(0)\rangle e^{i \omega t} d t \tag{7.23} \label{7.23} \]
For a classical equilibrium system the correlation function is stationary and even, \(C(t)=C(-t)\), so:
\[ S_{A A}(\omega)=2 \operatorname{Re} \int_{0}^{\infty}\langle A(t) A(0)\rangle e^{i \omega t} d t=2 \int_{0}^{\infty}\langle A(t) A(0)\rangle \cos (\omega t) d t \tag{7.24} \label{7.24} \]
7.4.3 Linear Response and Susceptibility
The mean response to an external force \(f(t)\) linearly coupled to \(A\) is
\[ \delta\langle A(t)\rangle=\int_{-\infty}^{t} \phi\left(t-t^{\prime}\right) f\left(t^{\prime}\right) d t^{\prime} \tag{7.25} \label{7.25} \]
where the response function \(\phi(t)\) (causal, so \(\phi(t)=0\) for \(t<0\) ) is related to the susceptibility by
\[ \chi(\omega)=\int_{0}^{\infty} \phi(t) e^{i \omega t} d t=\int_{0}^{\infty} \phi(t)[\cos (\omega t)+i \sin (\omega t)] d t \tag{7.26} \label{7.26} \]
The dissipative part is therefore
\[ \operatorname{Im}[\chi(\omega)]=\int_{0}^{\infty} \phi(t) \sin (\omega t) d t \tag{7.27} \label{7.27} \]
7.4.4 The Kubo Relation in the Time Domain
Substituting Eqs. Equation \ref{7.24} and Equation \ref{7.27} into the FDT (A.29):
\[ \int_{-\infty}^{\infty}\langle A(t) A(0)\rangle e^{i \omega t} d t=\frac{2 k_{B} T}{\omega} \int_{0}^{\infty} \phi(t) \sin (\omega t) d t \tag{7.28} \label{7.28} \]
Using the symmetry of \(C(t)\) and inverting the Fourier relation (see, e.g., Zwanzig [3]), the time-domain identity for \(t>0\) is the Kubo relation:
\[ \phi(t)=-\frac{1}{k_{B} T} \frac{d}{d t}\langle A(t) A(0)\rangle . \tag{7.29} \label{7.29} \]
This is the same result as Equation \ref{7.11} derived from Liouville dynamics, now obtained by a purely spectral argument.
7.4.5 The Frequency-Domain Kubo Formula
Inserting Equation \ref{7.29} into the definition of \(\chi(\omega)\) and integrating by parts:
\[ \chi(\omega)=\int_{0}^{\infty}\left(-\frac{1}{k_{B} T} \frac{d}{d t}\langle A(t) A(0)\rangle\right) e^{i \omega t} d t=\frac{1}{k_{B} T}\left[\left\langle A(0)^{2}\right\rangle+i \omega \int_{0}^{\infty}\langle A(t) A(0)\rangle e^{i \omega t} d t\right] . \tag{7.30} \label{7.30} \]
Equation \ref{7.30} is the standard frequency-domain Kubo formula: the susceptibility is the \(\beta\) -weighted Laplace transform of the correlation function.
7.4.6 Green-Kubo Relations for Transport Coefficients
For transport coefficients, one identifies the observable \(A\) with a current and takes the DC \((\omega \rightarrow 0)\) limit of Equation \ref{7.30}. For electrical conductivity, with \(A=J\) (current density):
\[ \sigma(\omega)=\frac{1}{V k_{B} T} \int_{0}^{\infty}\langle J(t) J(0)\rangle e^{i \omega t} d t \tag{7.31} \label{7.31} \]
which at \(\omega=0\) reduces to the DC Green-Kubo formula \(\sigma_{\mathrm{DC}}=\left(V k_{B} T\right)^{-1} \int_{0}^{\infty}\langle J(t) J(0)\rangle d t\). Analogous formulas hold for viscosity, thermal conductivity, and diffusion coefficients, all expressible as time integrals of appropriate equilibrium current-current correlations.
7.5 Defining and Contrasting the Two Types of Fluctuation-Dissipation Theorems
The Fluctuation-Dissipation Theorem (FDT), pioneered by Ryogo Kubo, bridges the gap between microscopic fluctuations and macroscopic dissipative responses. In the context of the Generalized Langevin Equation (GLE), two distinct forms of the FDT are identified: the First Kind (1FDT) and the Second Kind (2FDT) [12].
7.5.1 The First Fluctuation-Dissipation Theorem (1FDT)
The 1FDT is rooted in Linear Response Theory (LRT). It establishes a direct relationship between the linear response of a system to a small external perturbation and the correlation of its spontaneous fluctuations in thermal equilibrium.
- Core Principle: It relates the generalized susceptibility (or mobility) of a system to the auto-correlation function of a physical observable (e.g., velocity).
- Physical Interpretation: This theorem validates Onsager's regression hypothesis-the idea that the relaxation of a system perturbed from equilibrium follows the same laws as the decay of a spontaneous, internal fluctuation.
- Mathematical Relation: For a classical system, the response function \(\chi(t)\) is related to the equilibrium correlation function \(C(t)\) by:
\[ \chi(t)=-\beta \frac{d}{d t} C(t) \theta(t) \tag{7.32} \label{7.32} \]
where \(\beta=\left(k_{B} T\right)^{-1}\) and \(\theta(t)\) is the Heaviside step function.
7.5.2 The Second Fluctuation-Dissipation Theorem (2FDT)
The 2FDT arises within the framework of the Generalized Langevin Equation (GLE) and is often referred to as the Generalized Nyquist Theorem.
- Core Principle: It relates the stochastic random force acting on a particle to the frictional memory kernel that characterizes the energy dissipation to the surrounding heat bath.
- Physical Interpretation: It ensures a balance between the random impacts from the environment (fluctuations) and the systematic drag (dissipation). This balance ensures the system reaches thermal equilibrium rather than heating up indefinitely.
- Mathematical Relation: The auto-correlation of the random force \(F_{\text {ext }}(t)\) is proportional to the memory kernel \(\gamma(t)\) :
\[ \left\langle F_{e x t}(t) F_{e x t}(0)\right\rangle=k_{B} T \gamma(t) \tag{7.33} \label{7.33} \]
7.5.3 Contrast and Comparison
| Feature | First Kind (1FDT) | Second Kind (2FDT) | |
|---|---|---|---|
| Primary work | Frame- | Linear Response Theory (LRT) | Generalized Langevin Equation (GLE) |
| Related ties | Quanti- | Response function vs. Correla- tion function | Random force vs. Memory kernel |
| Level of tion | Descrip- | Macro/Mesoscopic (Observable response) | Microscopic/Stochastic (Dynam- ics) |
| Equilibrium tus | Sta- | Relates non-equilibrium response to equilibrium fluctuations | Relates noise and friction within the heat bath |
| Robustness | Can break down far from equilib- rium | Often valid in non-equilibrium steady states |
7.6 Summary and Logical Connections
The table below collects the key results of this chapter and their derivation routes.
7.7 Summary and Logical Connections
The table below collects the key results of this chapter and their derivation routes.
| Result | Key equation | Origin |
|---|---|---|
| Thermal identity | \(\left\{A, \rho_{0}\right\}=-\beta \rho_{0} \dot{A}\) | Chain rule for Poisson brackets |
| Kubo formula (time domain) | \(\chi(t)=-\beta \dot{C}(t) \theta(t)\) | Liouville + first-order perturbation |
| Classical FDT | \(\chi^{\prime \prime}(\omega)=\frac{\beta \omega}{2} S(\omega)\) | Kubo formula + spectral decomposition |
| GLE susceptibility | Equation \ref{7.17} | Fourier transform of mean GLE |
| Second FDT | \(S_{f f}(\omega)=2 k_{B} T m \operatorname{Re}[\tilde{\gamma}(\omega)]\) | Equilibrium constraint on noise |
| First FDT (mobility) | \(\mu(\omega)=\beta \int_{0}^{\infty}\langle v v\rangle e^{i \omega t} d t\) | Kubo formula for velocity |
| Kramers-Kronig | Eqs. (A.17)-(A.18) | Causality (analyticity in upper half-plane) |
| Kubo relation (from FDT) | \(\phi(t)=-\left(k_{B} T\right)^{-1} \dot{C}(t)\) | FDT + Wiener-Khinchin (reverse route) |
| Green-Kubo (conductivity) | Equation \ref{7.31} | DC limit of frequency-domain Kubo formula |
The central message is that all dissipative and transport properties of a system at thermal equilibrium are encoded in equilibrium time-correlation functions. Whether one starts from Liouville dynamics, the stochastic GLE, or the FDT itself, the same family of results emerges: susceptibility, mobility, conductivity, and noise are not independent quantities-they are different faces of the same underlying fluctuation statistics.
Bibliography
[1] N. Wiener, Acta Math. 55, 117 (1930).
[2] A. Khinchin, Math. Ann. 109, 604 (1934).
[3] J. B. Johnson, Phys. Rev. 32, 97 (1928).
[4] H. Nyquist, Phys. Rev. 32, 110 (1928).
[5] L. E. Reichl, A Modern Course in Statistical Physics, University of Texas Press, Austin TX (1980).
[6] B. I. Bleaney and B. Bleaney, Electricity and Magnetism, Clarendon Press, Oxford (1965).
[7] W. C. van Etten, Introduction to Random Signals and Noise, John Wiley & Sons, England (2005).
[8] R. Kubo, Rep. Prog. Phys. 29, 255 (1966).
[9] R. Kubo, Statistical-Mechanical Theory of Irreversible Processes. I, J. Phys. Soc. Japan 12, 570 (1957).
[10] R. Kubo, The Fluctuation-Dissipation Theorem, Rep. Prog. Phys. 29, 255 (1966).
[11] R. Zwanzig, Nonequilibrium Statistical Mechanics, Oxford University Press (2001).
[12] Balakrishnan, V. (1979). Fluctuation-dissipation theorems from the generalised Langevin equation. Pramana, 12(4), 301-315.
This page titled 7: Classical Linear Response Theory and the Kubo Formula was last modified on Mon, 02 Nov 2026 06:11:06 GMT and is shared under a CC BY-NC-SA 4.0 license and was authored, remixed, and/or curated by .
The book could not be loaded just now.