18: Quantum Mori-Zwanzig Formulation of the Magalinski Model

Fluctuations Near Equilibrium · LibreTexts

0%

Contents

    Text

    Size
    100%
    Theme
    Typeface
    Line length
    Spacing

    Read aloud

    How

    Speed: 1×
    Light up
    Highlight

    Symbols: Mulberry Symbols, © 2018–2026 Steve Lee, CC BY-SA 4.0 (licence, the set), the site's own copies.

    From LibreTexts, under CC BY-NC-SA 4.0, as the source states it; its equations are written in TeX, which the PDF shows typeset. This page at LibreTexts.

    18: Quantum Mori-Zwanzig Formulation of the Magalinski Model

    Chapter 14 applied the Mori-Zwanzig (M-Z) projection technique to the Magalinski model purely classically: the Liouvillian was the Poisson-bracket generator, the projector \(\mathcal{P}\) was built from the classical equilibrium average, and the equipartition theorem fixed the correlation matrix \(\mathbb{C}\). Every one of those ingredients has a quantum-mechanical counterpart, and the purpose of this chapter is to carry the Magalinski-model calculation through operator by operator: the Liouvillian becomes the commutator generator of Heisenberg motion, the projector is built from Mori's canonical (Kubo-transform) inner product rather than the naive operator trace, and the equipartition theorem is replaced by its exact quantum generalization. We shall find that, once this dictionary is in place, the frequency matrix \(\Omega\) and the memory kernel \(K(t)\) obtained in Chapter 14 carry over unchanged, operator for operator, to all orders in the system-bath coupling. The genuinely new quantum content appears only when one asks what the noise operator \(\hat{F}(t)\) actually is as a physical, non-commuting object: its Mori-transformed autocorrelation reproduces the classical fluctuation-dissipation relation exactly, while its symmetrized (measurable) autocorrelation carries the familiar thermal factor \(\hbar \omega \operatorname{coth}\left(\hbar \omega / 2 k_{\mathrm{B}} T\right)\) derived independently, via the Kubo-Martin-Schwinger condition, in Appendix A.

    A second, more subtle point concerns the Hamiltonian itself. Chapter 14 worked with the bare bilinear coupling \(H_{\text {int }}=\sum_{j} c_{j} q_{0} q_{j}\) and implicitly set the system-bath position correlation \(\left\langle q_{0} q_{j}\right\rangle_{0}\) to zero when forming \(\mathbb{C}\) and \(\Omega\). This is exact only to leading order in the coupling; at the operator level, making the Mori inner product well defined non-perturbatively requires the Ford-Kac-Mazur/Caldeira-Leggett renormalization counterterm discussed at length, for the single-oscillator system-plus-bath problem, in Appendix A ("Details of potential renormalization") of Ghosh et al. [6]. We include this counterterm from the outset below; doing so is what allows every statement in this chapter to hold exactly, not merely to leading order in \(c_{j}\). This same counterterm construction - the Caldeira-Leggett Hamiltonian of Eq. (H-CL) - is already adopted, for a completely general central potential \(V(X)\), in Appendix C ("Generalized Langevin Dynamics for an Arbitrary Potential"), which simply invokes "the quantum-mechanical FDT" at the point where it is needed; the present appendix supplies the missing first-principles derivation of that FDT, and of the underlying quantum generalized Langevin equation, for the exactly solvable harmonic case from which Appendix C's anharmonic analysis departs.

    C. 1 The Quantum Hamiltonian and the Renormalization Counterterm

    We promote the coordinates and momenta of Chapter 14 to Hermitian operators, \(q_{0} \rightarrow \hat{q}_{0}\), \(p_{0} \rightarrow \hat{p}_{0}, q_{j} \rightarrow \hat{q}_{j}, p_{j} \rightarrow \hat{p}_{j}\), satisfying the canonical commutation relations

    \[ \left[\hat{q}_{0}, \hat{p}_{0}\right]=i \hbar, \quad\left[\hat{q}_{j}, \hat{p}_{k}\right]=i \hbar \delta_{j k}, \quad j, k=1, \ldots, N, \tag{C.1} \label{C.1} \]

    with all other commutators among \(\left\{\hat{q}_{0}, \hat{p}_{0}, \hat{q}_{j}, \hat{p}_{j}\right\}\) vanishing. Rather than the bare Hamiltonian of Equation \ref{14.1} of Chapter 14, we take, following Ford, Kac and Mazur and the independent-oscillator model reviewed in [6],

    \[ \hat{H}=\frac{\hat{p}_{0}^{2}}{2 m_{0}}+\frac{m_{0} \omega_{0}^{2}}{2} \hat{q}_{0}^{2}+\sum_{j=1}^{N}\left[\frac{\hat{p}_{j}^{2}}{2 m_{j}}+\frac{m_{j} \omega_{j}^{2}}{2}\left(\hat{q}_{j}+\frac{c_{j}}{m_{j} \omega_{j}^{2}} \hat{q}_{0}\right)^{2}\right] . \tag{C.2} \label{C.2} \]

    Expanding the square reproduces exactly the bilinear coupling of Chapter 14 together with one extra, \(j\)-summed, term quadratic in \(\hat{q}_{0}\) :

    \[ \hat{H}=\underbrace{\frac{\hat{p}_{0}^{2}}{2 m_{0}}+\frac{m_{0} \omega_{0}^{2}}{2} \hat{q}_{0}^{2}}_{\hat{H}_{S}}+\underbrace{\sum_{j}\left[\frac{\hat{p}_{j}^{2}}{2 m_{j}}+\frac{m_{j} \omega_{j}^{2}}{2} \hat{q}_{j}^{2}\right]}_{\hat{H}_{B}}+\underbrace{\sum_{j} c_{j} \hat{q}_{0} \hat{q}_{j}}_{\hat{H}_{\mathrm{int}}}+\underbrace{\frac{1}{2} \sum_{j} \frac{c_{j}^{2}}{m_{j} \omega_{j}^{2}} \hat{q}_{0}^{2}}_{V_{r}\left(\hat{q}_{0}\right)} . \tag{C.3} \label{C.3} \]

    The last piece, \(V_{r}\left(\hat{q}_{0}\right)\), is the renormalization counterterm. It is not optional: without it, the bath-induced frequency shift \((\delta \omega)^{2}=\left(1 / m_{0}\right) \sum_{j} c_{j}^{2} /\left(m_{j} \omega_{j}^{2}\right)\) pulls the true, measurable oscillation frequency of the system away from the parameter \(\omega_{0}\) appearing in \(\hat{H}_{S}\) - indeed, for a strongly or continuously coupled bath \((\delta \omega)^{2}\) can exceed \(\omega_{0}^{2}\) and the potential in Equation \ref{14.1} ceases to be bounded from below, exactly as explained in Appendix A of [6]. With \(V_{r}\) included, as we now show, \(\omega_{0}\) in Eq. (C.2) is the physical frequency, to all orders in \(\left\{c_{j}\right\}\) and for any \(N\), including \(N \rightarrow \infty\).

    It is convenient to introduce the shifted bath coordinate

    \[ \hat{Q}_{j} \equiv \hat{q}_{j}+\frac{c_{j}}{m_{j} \omega_{j}^{2}} \hat{q}_{0}, \tag{C.4} \label{C.4} \]

    which is simply the combination that is perfect-squared in Eq. (C.2). Because \(\left[\hat{Q}_{j}, \hat{p}_{j}\right]=\) \(\left[\hat{q}_{j}, \hat{p}_{j}\right]=i \hbar\) and \(\left[\hat{Q}_{j}, \hat{p}_{k}\right]=0\) for \(k \neq j\), the pair \(\left(\hat{Q}_{j}, \hat{p}_{j}\right)\) remains canonically conjugate.

    C. 2 The Quantum Liouvillian and Heisenberg Equations of Motion

    Following the convention used throughout this book for the quantum case (and matching the "Key Foundations in Quantum Mechanics" used elsewhere), we define the quantum Liouvillian superoperator through the commutator,

    \[ i \mathcal{L} \hat{X} \equiv \frac{i}{\hbar}[\hat{H}, \hat{X}], \tag{C.5} \label{C.5} \]

    so that the Heisenberg equation of motion for any operator \(\hat{X}(t)=e^{i \hat{H} t / \hbar} \hat{X} e^{-i \hat{H} t / \hbar}\) reads \(d \hat{X} / d t=i \mathcal{L} \hat{X}\), in exact formal parallel with the classical \(\dot{A}=\mathcal{L} A\) of Chapter 14 and of the general formalism of Chapter 15.

    Using the canonical commutation relations (C.1) and the Hamiltonian (C.3), the Heisenberg equations of motion for the relevant vector \(\hat{\mathbf{A}}(t)=\left(\hat{q}_{0}(t), \hat{p}_{0}(t)\right)^{T}\) are obtained exactly as in

    elementary quantum mechanics, \(i \mathcal{L} \hat{q}_{0}=(i / \hbar)\left[\hat{H}, \hat{q}_{0}\right]=\partial \hat{H} / \partial \hat{p}_{0}\) and \(i \mathcal{L} \hat{p}_{0}=-\partial \hat{H} / \partial \hat{q}_{0}\) (the operator ordering is unambiguous here because \(\hat{H}\) is at most quadratic):

    \[ i \mathcal{L} \hat{q}_{0}=\frac{\hat{p}_{0}}{m_{0}}, \quad i \mathcal{L} \hat{p}_{0}=-m_{0} \omega_{0}^{2} \hat{q}_{0}-\sum_{j} c_{j} \hat{q}_{j}-\hat{q}_{0} \sum_{j} \frac{c_{j}^{2}}{m_{j} \omega_{j}^{2}} . \tag{C.6} \label{C.6} \]

    Substituting \(\hat{q}_{j}=\hat{Q}_{j}-c_{j} \hat{q}_{0} /\left(m_{j} \omega_{j}^{2}\right)\) from Eq. (C.4), the last two terms in \(i \mathcal{L} \hat{p}_{0}\) combine as

    \[ -\sum_{j} c_{j} \hat{q}_{j}-\hat{q}_{0} \sum_{j} \frac{c_{j}^{2}}{m_{j} \omega_{j}^{2}}=-\sum_{j} c_{j} \hat{Q}_{j}+\hat{q}_{0} \sum_{j} \frac{c_{j}^{2}}{m_{j} \omega_{j}^{2}}-\hat{q}_{0} \sum_{j} \frac{c_{j}^{2}}{m_{j} \omega_{j}^{2}}=-\sum_{j} c_{j} \hat{Q}_{j}, \nonumber \]

    and the renormalizing pieces cancel exactly. We are left with

    \[ i \mathcal{L} \hat{q}_{0}=\frac{\hat{p}_{0}}{m_{0}}, \quad i \mathcal{L} \hat{p}_{0}=-m_{0} \omega_{0}^{2} \hat{q}_{0}-\sum_{j} c_{j} \hat{Q}_{j} . \tag{C.7} \label{C.7} \]

    These are operator-identical in form to Equation \ref{14.5} of Chapter 14, with the single replacement \(q_{j} \rightarrow Q_{j}\) : the renormalization counterterm's entire effect is to replace the bare bath coordinate by its recentred counterpart in the force law on the oscillator. This is the quantum-operator statement of the physics behind Appendix A of [6], and it is the technical fact that makes everything that follows exact rather than perturbative.

    C. 3 The Mori (Kubo-Transform) Inner Product

    For classical phase-space functions, Chapter 14 used the equilibrium average \((A, B)=\langle A B\rangle_{0}\) to build the projector \(\mathcal{P}\). For non-commuting operators the naive choice \(\left\langle\hat{X}^{\dagger} \hat{Y}\right\rangle_{0}\) is not symmetric and does not make \(\mathcal{L}\) Hermitian with respect to it, so \(\mathcal{P}\) built from it fails to be a genuine orthogonal projector. Mori's resolution [2] is to use instead the canonical (Kubo-transform) inner product,

    \[ (\hat{X}, \hat{Y}) \equiv \frac{1}{\beta} \int_{0}^{\beta} d \lambda\left\langle e^{\lambda \hat{H}} \hat{X}^{\dagger} e^{-\lambda \hat{H}} \hat{Y}\right\rangle_{0}, \quad \beta=\frac{1}{k_{\mathrm{B}} T} \tag{C.8} \label{C.8} \]

    where \(\langle\cdot\rangle_{0}=Z^{-1} \operatorname{Tr}\left[e^{-\beta \hat{H}}(\cdot)\right]\) is the canonical equilibrium average with respect to the full Hamiltonian (C.2). One verifies that \((i \mathcal{L} \hat{X}, \hat{Y})=(\hat{X}, i \mathcal{L} \hat{Y})\), so that \(\mathcal{L}\) is Hermitian with respect to \((\cdot, \cdot)\) and the Mori projector

    \[ \mathcal{P} \hat{B} \equiv(\hat{B}, \hat{\mathbf{A}})(\hat{\mathbf{A}}, \hat{\mathbf{A}})^{-1} \hat{\mathbf{A}}, \quad \mathcal{Q} \equiv 1-\mathcal{P} \tag{C.9} \label{C.9} \]

    is a well-defined orthogonal projector onto the two-dimensional operator subspace spanned by \(\hat{q}_{0}, \hat{p}_{0}\). Note that \((\hat{X}, \hat{Y})\) is bilinear in \(\hat{X}, \hat{Y}\) (for Hermitian operators) and reduces to \(\langle\hat{X} \hat{Y}\rangle_{0}\) whenever \(\hat{X}\) and \(\hat{Y}\) commute - in particular, it reduces to the classical equilibrium average in the limit \(\hbar \rightarrow 0\).

    Theorem C. 1 (Quantum / Kubo-canonical equipartition theorem). Let \(\hat{x}_{a}\) denote any one of the canonical phase-space operators of a Hamiltonian system, \(\left\{\hat{x}_{a}\right\}=\left\{\hat{q}_{0}, \hat{p}_{0}, \hat{q}_{1}, \hat{p}_{1}, \ldots, \hat{q}_{N}, \hat{p}_{N}\right\}\). Then

    \[ \left(\hat{x}_{a}, \frac{\partial \hat{H}}{\partial \hat{x}_{b}}\right)=k_{\mathrm{B}} T \delta_{a b} \tag{C.10} \label{C.10} \]

    for all \(a, b\), exactly as in the classical equipartition theorem invoked in Chapter 14 (there citing Pathria and Beale [4]). This is a standard property of the Kubo-canonical correlation; see Kubo, Toda and Hashitsume [5].

    We shall use Theorem C. 1 exactly as Chapter 14 used its classical counterpart - as a black box that fixes all equal-time correlations - without reproducing its (standard, and not especially illuminating) proof here.

    C.3.1 The correlation matrix \(\mathbb{C}\)

    Applying Theorem C. 1 with \(a=b=q_{0}\) to the full Hamiltonian (C.3), and separately with \(a=q_{0}, b=q_{j}\) (distinct canonical coordinates, so the right-hand side of Eq. (C.10) vanishes), one finds, after eliminating \(\left(\hat{q}_{0}, \hat{q}_{j}\right)\) between the two resulting linear relations exactly as the classical renormalization identity of Appendix C of [6] does for the static susceptibility,

    \[ \left(\hat{q}_{0}, \hat{q}_{0}\right)=\frac{k_{\mathrm{B}} T}{m_{0} \omega_{0}^{2}}, \quad\left(\hat{q}_{0}, \hat{q}_{j}\right)=-\frac{c_{j}}{m_{j} \omega_{j}^{2}}\left(\hat{q}_{0}, \hat{q}_{0}\right)=-\frac{c_{j} k_{\mathrm{B}} T}{m_{0} \omega_{0}^{2} m_{j} \omega_{j}^{2}} . \tag{C.11} \label{C.11} \]

    By bilinearity, the shifted bath coordinate is exactly uncorrelated with the system coordinate:

    \[ \left(\hat{q}_{0}, \hat{Q}_{j}\right)=\left(\hat{q}_{0}, \hat{q}_{j}\right)+\frac{c_{j}}{m_{j} \omega_{j}^{2}}\left(\hat{q}_{0}, \hat{q}_{0}\right)=0 . \tag{C.12} \label{C.12} \]

    Applying Theorem C. 1 with \(a=q_{0}, b=p_{0}\) (distinct variables, \(\partial \hat{H} / \partial \hat{p}_{0}=\hat{p}_{0} / m_{0}\) ) gives \(\left(\hat{q}_{0}, \hat{p}_{0}\right)=0\); with \(a=b=p_{0}\) it gives \(\left(\hat{p}_{0}, \hat{p}_{0}\right)=m_{0} k_{\mathrm{B}} T\); and with \(a=p_{0}, b=q_{j}\) it gives \(\left(\hat{p}_{0}, \hat{q}_{j}\right)=0\) and hence, by the same bilinearity argument, \(\left(\hat{p}_{0}, \hat{Q}_{j}\right)=0\). Collecting these,

    \[ \mathbb{C} \equiv(\hat{\mathbf{A}}, \hat{\mathbf{A}})=\left(\begin{array}{ll} \left(\hat{q}_{0}, \hat{q}_{0}\right) & \left(\hat{q}_{0}, \hat{p}_{0}\right) \\[4pt] \left(\hat{p}_{0}, \hat{q}_{0}\right) & \left(\hat{p}_{0}, \hat{p}_{0}\right) \end{array}\right)=\left(\begin{array}{cc} \frac{k_{\mathrm{B}} T}{m_{0} \omega_{0}^{2}} & 0 \\[4pt] 0 & m_{0} k_{\mathrm{B}} T \end{array}\right), \tag{C.13} \label{C.13} \]

    identical to Equation \ref{14.8} of Chapter 14 - the quantum Mori-canonical correlation matrix coincides exactly with the classical equipartition matrix, with no \(\hbar\)-dependence whatsoever and no restriction to weak coupling. This is the single most important simplifying fact about Mori's inner product: it is constructed precisely so that static, equal-time quantum thermal averages reproduce their classical values.

    For completeness we record, by the identical argument applied to \(a=b=p_{j}\) and \(a=p_{j}, b=\) \(p_{k}(j \neq k)\), that

    \[ \left(\hat{p}_{j}, \hat{p}_{j}\right)=m_{j} k_{\mathrm{B}} T, \quad\left(\hat{p}_{j}, \hat{p}_{k}\right)=0(j \neq k), \quad\left(\hat{Q}_{j}, \hat{Q}_{j}\right)=\frac{k_{\mathrm{B}} T}{m_{j} \omega_{j}^{2}}, \quad\left(\hat{Q}_{j}, \hat{p}_{j}\right)=0 \tag{C.14} \label{C.14} \]

    the last two following from \(\partial \hat{H} / \partial \hat{Q}_{j}=m_{j} \omega_{j}^{2} \hat{Q}_{j}\) and \(\partial \hat{H} / \partial \hat{p}_{j}=\hat{p}_{j} / m_{j}\) evaluated at fixed \(\hat{q}_{0}\). That is: in every Mori-canonical moment that matters below, the shifted operator pair \(\left(\hat{Q}_{j}, \hat{p}_{j}\right)\) behaves exactly like an isolated quantum harmonic oscillator of mass \(m_{j}\) and frequency \(\omega_{j}\) in thermal equilibrium at temperature \(T\), decoupled from \(\hat{q}_{0}, \hat{p}_{0}\) and from every other bath mode.

    C. 4 The Frequency Matrix

    With \(\mathbb{C}\) in hand, the frequency matrix is defined, exactly as in Equation \ref{14.9} of Chapter 14,

    \[ \Omega \equiv(i \mathcal{L} \hat{\mathbf{A}}, \hat{\mathbf{A}}) \mathbb{C}^{-1} . \tag{C.15} \label{C.15} \]

    Using Eq. (C.7) and the vanishing of \(\left(\hat{Q}_{j}, \hat{q}_{0}\right),\left(\hat{Q}_{j}, \hat{p}_{0}\right)\) established in Eqs. (C.12)-(C.14):

    \[ \left(i \mathcal{L} \hat{q}_{0}, \hat{q}_{0}\right)=\frac{1}{m_{0}}\left(\hat{p}_{0}, \hat{q}_{0}\right)=0, \quad\left(i \mathcal{L} \hat{p}_{0}, \hat{q}_{0}\right)=-m_{0} \omega_{0}^{2}\left(\hat{q}_{0}, \hat{q}_{0}\right)-\sum_{j} c_{j}\left(\hat{Q}_{j}, \hat{q}_{0}\right)=-k_{\mathrm{B}} T, \nonumber \]

    \[ \left(i \mathcal{L} \hat{q}_{0}, \hat{p}_{0}\right)=\frac{1}{m_{0}}\left(\hat{p}_{0}, \hat{p}_{0}\right)=k_{\mathrm{B}} T, \quad\left(i \mathcal{L} \hat{p}_{0}, \hat{p}_{0}\right)=-m_{0} \omega_{0}^{2}\left(\hat{q}_{0}, \hat{p}_{0}\right)-\sum_{j} c_{j}\left(\hat{Q}_{j}, \hat{p}_{0}\right)=0 \nonumber \]

    so that, with \(\mathbb{C}^{-1}=\operatorname{diag}\left(m_{0} \omega_{0}^{2} / k_{\mathrm{B}} T, 1 /\left(m_{0} k_{\mathrm{B}} T\right)\right)\),

    \[ \Omega=\left(\begin{array}{cc} 0 & 1 / m_{0} \\[4pt] -m_{0} \omega_{0}^{2} & 0 \end{array}\right) . \tag{C.16} \label{C.16} \]

    This is identical to the boxed result, Equation \ref{14.12}, of Chapter 14, and - unlike the classical derivation there, which quietly assumed the system-bath position correlation to vanish - it is here exact to all orders in \(\left\{c_{j}\right\}\), a consequence of including the renormalization counterterm of Sec. C.1.

    C. 5 Orthogonal Dynamics and the Quantum Random-Force Operator

    Just as in Chapter 14, we must compute \(i \mathcal{Q} \mathcal{L} \hat{p}_{0}=i \mathcal{L} \hat{p}_{0}-\mathcal{P}\left(i \mathcal{L} \hat{p}_{0}\right)\). From Eq. (C.7), \(i \mathcal{L} \hat{p}_{0}=\) \(-m_{0} \omega_{0}^{2} \hat{q}_{0}-\sum_{j} c_{j} \hat{Q}_{j}\); since \(\left(\hat{Q}_{j}, \hat{q}_{0}\right)=\left(\hat{Q}_{j}, \hat{p}_{0}\right)=0\) for every \(j\) (Sec. C.3), the \(\hat{Q}_{j}\) piece has no overlap at all with the relevant subspace, and projection removes only the first term:

    \[ \mathcal{P}\left(i \mathcal{L} \hat{p}_{0}\right)=-m_{0} \omega_{0}^{2} \hat{q}_{0}, \quad \hat{f} \equiv i \mathcal{Q} \mathcal{L} \hat{p}_{0}=-\sum_{j} c_{j} \hat{Q}_{j} . \tag{C.17} \label{C.17} \]

    This matches Equation \ref{14.13} of Chapter 14 with \(q_{j} \rightarrow Q_{j}\).

    To obtain the random-force operator \(\hat{F}(t)=e^{i \mathcal{Q} \mathcal{L} t} \hat{f}\) we must know how \(\hat{f}\) evolves under the orthogonal, projected dynamics. Compute first the unrestricted derivative,

    \[ i \mathcal{L} \hat{f}=-\sum_{j} c_{j} i \mathcal{L} \hat{Q}_{j}=-\sum_{j} c_{j}\left(\frac{\hat{p}_{j}}{m_{j}}-\frac{c_{j}}{m_{j} \omega_{j}^{2}} i \mathcal{L} \hat{q}_{0}\right)=-\sum_{j} \frac{c_{j}}{m_{j}} \hat{p}_{j}+\frac{\hat{p}_{0}}{m_{0}} \sum_{j} \frac{c_{j}^{2}}{m_{j} \omega_{j}^{2}}, \nonumber \]

    using \(i \mathcal{L} \hat{q}_{0}=\hat{p}_{0} / m_{0}\) from Eq. (C.7). At first sight this leaks into the relevant subspace through the \(\hat{p}_{0}\) term: \(\hat{Q}_{j}\) itself "feels" \(\hat{q}_{0}\) because its very definition, Eq. (C.4), involves \(\hat{q}_{0}\). However, this leakage is projected straight back out again. Using \(\left(\hat{p}_{j}, \hat{q}_{0}\right)=\left(\hat{p}_{j}, \hat{p}_{0}\right)=0\) (Theorem C. 1 applied to \(a=p_{j}, b=q_{0}\) or \(p_{0}\) ) and \(\left(\hat{p}_{0}, \hat{p}_{0}\right)=m_{0} k_{\mathrm{B}} T\),

    \[ \left(i \mathcal{L} \hat{f}, \hat{q}_{0}\right)=0, \quad\left(i \mathcal{L} \hat{f}, \hat{p}_{0}\right)=\frac{k_{\mathrm{B}} T}{m_{0}} \sum_{j} \frac{c_{j}^{2}}{m_{j} \omega_{j}^{2}} \cdot m_{0}=k_{\mathrm{B}} T \sum_{j} \frac{c_{j}^{2}}{m_{j} \omega_{j}^{2}}, \nonumber \]

    so that \(\mathcal{P}(i \mathcal{L} \hat{f})=\left[\sum_{j} c_{j}^{2} /\left(m_{j} \omega_{j}^{2}\right)\right] \hat{p}_{0} / m_{0}\) - precisely the term that must be subtracted to cancel the leaked piece of \(i \mathcal{L} \hat{f}\) itself:

    \[ i \mathcal{Q} \mathcal{L} \hat{f}=i \mathcal{L} \hat{f}-\mathcal{P}(i \mathcal{L} \hat{f})=-\sum_{j} \frac{c_{j}}{m_{j}} \hat{p}_{j}=-\left.\sum_{j} c_{j} i \mathcal{L} \hat{Q}_{j}\right|_{\hat{q}_{0} \rightarrow 0} . \tag{C.18} \label{C.18} \]

    In other words, on the orthogonal subspace the generator \(i \mathcal{Q} \mathcal{L}\) acting on \(\hat{f}\) coincides exactly with the free-bath generator \(i \mathcal{L}_{B}\), under which \(\hat{Q}_{j}, \hat{p}_{j}\) evolve as an isolated oscillator, \(i \mathcal{L}_{B} \hat{Q}_{j}=\hat{p}_{j} / m_{j}\), \(i \mathcal{L}_{B} \hat{p}_{j}=-m_{j} \omega_{j}^{2} \hat{Q}_{j}\). This is the operator version of the "Rosetta Stone" identity of Chapter 14, \(\mathcal{Q} \mathcal{L} f(0)=\mathcal{L}_{B} f(0)\) : the bilinear-coupling closure property survives quantization unchanged, provided one tracks the system-dependence hidden inside the shifted bath coordinate \(\hat{Q}_{j}\), as we have just done explicitly. Exponentiating,

    \[ \hat{F}(t)=e^{i \mathcal{Q} \mathcal{L} t} \hat{f}=-\sum_{j=1}^{N} c_{j} \hat{Q}_{j}(t), \quad \hat{Q}_{j}(t)=\hat{Q}_{j}(0) \cos \left(\omega_{j} t\right)+\frac{\hat{p}_{j}(0)}{m_{j} \omega_{j}} \sin \left(\omega_{j} t\right), \tag{C.19} \label{C.19} \]

    exactly reproducing Equation \ref{14.14} of Chapter 14 with \(q_{j} \rightarrow Q_{j}\).

    C. 6 Quantum Memory Kernel and the Mori-Form Fluctuation-Dissipation Theorem

    We need the two-time Mori-canonical correlation of a single shifted bath mode, \(\left(\hat{Q}_{j}(0), \hat{Q}_{j}(t)\right)\). Since \(\left(\hat{Q}_{j}, \hat{p}_{j}\right)\) is, as established in Sec. C.3, in every relevant respect an isolated harmonic oscillator of mass \(m_{j}\) and frequency \(\omega_{j}\) at temperature \(T\), we may compute this directly from Eq. (C.8) using ladder operators \(\hat{a}_{j}, \hat{a}_{j}^{\dagger}\) for that mode, \(\hat{Q}_{j}=\sqrt{\hbar / 2 m_{j} \omega_{j}}\left(\hat{a}_{j}+\hat{a}_{j}^{\dagger}\right)\), with \(\hat{Q}_{j}(t)\) as in Eq. (C.19) and thermal occupation \(n_{j} \equiv\left\langle\hat{a}_{j}^{\dagger} \hat{a}_{j}\right\rangle_{0}=\left(e^{\beta \hbar \omega_{j}}-1\right)^{-1}\).

    Lemma C. 2 (Two-time Mori correlation of a free quantum oscillator). For a free quantum harmonic oscillator of mass \(m\), frequency \(\omega\), in thermal equilibrium at temperature \(T\),

    \[ (\hat{q}(0), \hat{q}(t))=\frac{k_{\mathrm{B}} T}{m \omega^{2}} \cos (\omega t) . \tag{C.20} \label{C.20} \]

    Proof. With \(\hat{q}(-i \hbar \lambda)=\sqrt{\hbar / 2 m \omega}\left(\hat{a} e^{-\omega \hbar \lambda}+\hat{a}^{\dagger} e^{\omega \hbar \lambda}\right)\) and \(\hat{q}(t)=\sqrt{\hbar / 2 m \omega}\left(\hat{a} e^{-i \omega t}+\hat{a}^{\dagger} e^{i \omega t}\right)\),

    \[ \langle\hat{q}(-i \hbar \lambda) \hat{q}(t)\rangle_{0}=\frac{\hbar}{2 m \omega}\left[(n+1) e^{-\omega \hbar \lambda} e^{i \omega t}+n e^{\omega \hbar \lambda} e^{-i \omega t}\right] . \nonumber \]

    Using \(\frac{1}{\beta} \int_{0}^{\beta} d \lambda e^{\mp \omega \hbar \lambda}=\frac{1-e^{\mp \beta \hbar \omega}}{ \pm \beta \hbar \omega}\) together with the Bose identities \((n+1)\left(1-e^{-\beta \hbar \omega}\right)=1\) and \(n\left(e^{\beta \hbar \omega}-1\right)=1\), both integrals collapse to \(1 /(\beta \hbar \omega)\), giving

    \[ (\hat{q}(0), \hat{q}(t))=\frac{\hbar}{2 m \omega} \cdot \frac{1}{\beta \hbar \omega}\left(e^{i \omega t}+e^{-i \omega t}\right)=\frac{1}{\beta m \omega^{2}} \cos (\omega t)=\frac{k_{\mathrm{B}} T}{m \omega^{2}} \cos (\omega t) . \nonumber \]\(\square\)

    Lemma C. 2 is the dynamical counterpart of Theorem C. 1 all explicit \(\hbar\) - and temperature-dependence cancels, and the Mori-canonical two-time correlation of a harmonic coordinate is exactly the classical bath correlation \(\left\langle q_{i}(0) q_{i}(t)\right\rangle_{e q}=k_{\mathrm{B}} T \cos \left(\omega_{i} t\right) /\left(m_{i} \omega_{i}^{2}\right)\) used in Chapter 14 immediately before its boxed result, Equation \ref{14.17}, at every temperature. Applying it to each shifted bath mode in Eq. (C.19),

    \[ (\hat{F}(0), \hat{F}(t))=\sum_{j} c_{j}^{2}\left(\hat{Q}_{j}(0), \hat{Q}_{j}(t)\right)=k_{\mathrm{B}} T \sum_{j} \frac{c_{j}^{2}}{m_{j} \omega_{j}^{2}} \cos \left(\omega_{j} t\right), \tag{C.21} \label{C.21} \]

    and dividing by \(\mathbb{C}_{p p}=m_{0} k_{\mathrm{B}} T\) as in Chapter 14 (and Eq. (C.14) confirms \(\mathcal{Q}\) does not mix the \(\hat{Q}_{j}\)-sector back into \(\hat{p}_{0}\), so the \(p p\)-component is the only non-zero one), we obtain the quantum memory kernel

    \[ K(t)=\frac{(\hat{F}(0), \hat{F}(t))}{m_{0} k_{\mathrm{B}} T}=\frac{1}{m_{0}} \sum_{j=1}^{N} \frac{c_{j}^{2}}{m_{j} \omega_{j}^{2}} \cos \left(\omega_{j} t\right), \tag{C.22} \label{C.22} \]

    identical to the boxed result, Equation \ref{14.17}, of Chapter 14. The complete quantum generalized Langevin equation for the Heisenberg operators \(\hat{q}_{0}(t), \hat{p}_{0}(t)\) is therefore

    \[ \frac{d \hat{q}_{0}}{d t}=\frac{\hat{p}_{0}}{m_{0}}, \quad \frac{d \hat{p}_{0}}{d t}=-m_{0} \omega_{0}^{2} \hat{q}_{0}-\int_{0}^{t} K(t-s) \frac{\hat{p}_{0}(s)}{m_{0}} d s+\hat{F}(t), \tag{C.23} \label{C.23} \]

    operator-identical to Equation \ref{14.18} of Chapter 14, with the exact Mori-form ("second") fluctuation-dissipation theorem

    \[ (\hat{F}(t), \hat{F}(s))=k_{\mathrm{B}} T m_{0} K(|t-s|) \tag{C.24} \label{C.24} \]

    holding by construction, in exact parallel with Equation \ref{15.25} of the general formalism of Chapter 15.

    C. 7 From Mori Correlations to Physical Noise: the Thermal Factor and Operator Non-Commutativity

    Equation (C.24) is an exact algebraic identity of the Mori formalism, but \((\hat{F}(t), \hat{F}(s))\) is not what an experiment measures: a measurement of the force operator at two times yields expectation values of the physical, symmetrized correlator \(\frac{1}{2}\langle\{\hat{F}(t), \hat{F}(s)\}\rangle_{0}\), together with the (purely quantum, operator-ordering) commutator \(\langle[\hat{F}(t), \hat{F}(s)]\rangle_{0}\). Both are obtained from the same mode expansion used in the proof of Lemma C.2.

    Lemma C. 3 (Symmetrized correlator and commutator of a free quantum oscillator). For the same oscillator as in Lemma C.2,

    \[ \begin{align*} \frac{1}{2}\langle\{\hat{q}(0), \hat{q}(t)\}\rangle_{0} & =\frac{\hbar}{2 m \omega} \operatorname{coth}\left(\frac{\hbar \omega}{2 k_{\mathrm{B}} T}\right) \cos (\omega t), \tag{C.25} \label{C.25}\\[4pt] {[\hat{q}(0), \hat{q}(t)] } & =\frac{i \hbar}{m \omega} \sin (\omega t), \tag{C.26} \end{align*} \]

    with Eq. (C.26) a c-number, independent of temperature.

    Proof. From \(\hat{q}(0)=\sqrt{\hbar / 2 m \omega}\left(\hat{a}+\hat{a}^{\dagger}\right)\) and \(\hat{q}(t)=\sqrt{\hbar / 2 m \omega}\left(\hat{a} e^{-i \omega t}+\hat{a}^{\dagger} e^{i \omega t}\right)\),

    \[ \langle\hat{q}(0) \hat{q}(t)\rangle_{0}=\frac{\hbar}{2 m \omega}\left[(n+1) e^{i \omega t}+n e^{-i \omega t}\right], \quad\langle\hat{q}(t) \hat{q}(0)\rangle_{0}=\frac{\hbar}{2 m \omega}\left[(n+1) e^{-i \omega t}+n e^{i \omega t}\right] . \nonumber \]

    Adding and halving, and using \(2 n+1=\operatorname{coth}(\beta \hbar \omega / 2)\), gives Eq. (C.25); subtracting gives \((\hbar / 2 m \omega) \cdot(2 i \sin \omega t)=i \hbar \sin (\omega t) /(m \omega)\), which is Eq. (C.26) (and, being a \(c\)-number for this linear system, equals its own equilibrium average, consistent with evaluating it directly from the explicit solution \(\hat{q}(t)=\hat{q}(0) \cos \omega t+\hat{p}(0) \sin \omega t /(m \omega)\) and \([\hat{q}(0), \hat{p}(0)]=i \hbar)\). \(\square\)

    Applying Lemma C. 3 to each mode of \(\hat{F}(t)=-\sum_{j} c_{j} \hat{Q}_{j}(t)\),

    \[ \begin{align*} \frac{1}{2}\langle\{\hat{F}(0), \hat{F}(t)\}\rangle_{0} & =\sum_{j} c_{j}^{2} \frac{\hbar}{2 m_{j} \omega_{j}} \operatorname{coth}\left(\frac{\hbar \omega_{j}}{2 k_{\mathrm{B}} T}\right) \cos \left(\omega_{j} t\right), \tag{C.27} \label{C.27}\\[4pt] {[\hat{F}(0), \hat{F}(t)] } & =i \hbar \sum_{j} c_{j}^{2} \frac{1}{m_{j} \omega_{j}} \sin \left(\omega_{j} t\right) . \tag{C.28} \end{align*} \]

    Equation (C.27) is, in the continuum limit \(\sum_{j} \rightarrow \int d \omega J(\omega) /(\ldots)\), exactly the Callen-Welton/KMS form of the fluctuation-dissipation theorem derived from scratch in Appendix A, Sec. "The KMS Condition and Detailed Balance," and, in particular, the boxed detailed-balance relation \(\chi_{A B}^{\prime \prime}(\omega)=\frac{1}{2 \hbar}\left(1-e^{-\beta \hbar \omega}\right) C_{A B}(\omega)\), Eq. (A.26), of Sec. "The Fluctuation-Dissipation Theorem" there. Combining Eq. (A.26) with the symmetrized spectral density \(S_{A B}(\omega)\) of Eq. (A.25) and the detailed-balance relation \(C_{A B}(\omega)=e^{\beta \hbar \omega} C_{B A}(\omega)\), Eq. (A.21), yields precisely the familiar \(\operatorname{coth}\) form \(S_{A B}(\omega)=\hbar \operatorname{coth}(\beta \hbar \omega / 2) \chi_{A B}^{\prime \prime}(\omega)\) of Eq. (C.27) above. This is, equivalently, the quantum noise correlator of the independent-oscillator model, Eq. (43) of Ghosh et al. [6]: the thermal weight \(\hbar \omega \operatorname{coth}\left(\hbar \omega / 2 k_{\mathrm{B}} T\right)\) replaces the classical \(2 k_{\mathrm{B}} T\) exactly as expected from the KMS condition. Equation (C.28) has no classical analogue at all: classically \(F(t)\) is a \(c\)-number trajectory and \(\left[F\left(t_{1}\right), F\left(t_{2}\right)\right] \equiv 0\) trivially, whereas quantum mechanically the noise operator fails to commute with itself at unequal times - a direct consequence of the operator non-commutativity flagged as a defining feature of the quantum Mori-Zwanzig formalism. Notably, Eq. (C.28) is temperature-independent: it is a statement about the kinematic (causal) structure of the operator algebra, not about the thermal state, and it survives unchanged as \(T \rightarrow 0\).

    Remark C. 1 (Zero-point persistence). Setting \(T \rightarrow 0\) in Eq. (C.27) gives \(n_{j} \rightarrow 0\) but \(\operatorname{coth}\left(\hbar \omega_{j} / 2 k_{\mathrm{B}} T\right) \rightarrow\) 1, so \(\frac{1}{2}\langle\{\hat{F}(0), \hat{F}(t)\}\rangle_{0} \rightarrow \sum_{j} c_{j}^{2} \hbar \cos \left(\omega_{j} t\right) /\left(2 m_{j} \omega_{j}\right) \neq 0\) : the physical noise driving the oscillator, and hence the dissipative dynamics itself, does not switch off at absolute zero. This is the zero-point-fluctuation phenomenon noted for the general quantum Mori-Zwanzig formalism, and it has no counterpart in Chapter 14, where the classical force vanishes identically as \(T \rightarrow 0\).

    C. 8 The Classical Limit

    As \(\hbar \rightarrow 0\) at fixed \(T, \operatorname{coth}\left(\hbar \omega / 2 k_{\mathrm{B}} T\right) \rightarrow 2 k_{\mathrm{B}} T /(\hbar \omega)\), and Eq. (C.25) collapses to

    \[ \frac{1}{2}\langle\{\hat{q}(0), \hat{q}(t)\}\rangle_{0} \longrightarrow \frac{k_{\mathrm{B}} T}{m \omega^{2}} \cos (\omega t), \nonumber \]

    which coincides with the Mori-canonical result of Lemma C.2: in the classical limit, the physical (symmetrized) and the Mori-canonical correlators merge into a single, \(\hbar\)-independent object, which is exactly the classical equal-ensemble correlation function used throughout Chapter 14. Simultaneously, the commutator (C.26), which carries an explicit overall factor of \(\hbar\), vanishes. The entire apparatus of this chapter - the Hamiltonian (C.2), the equations of motion (C.7), the frequency matrix (C.16), the memory kernel (C.22), and the generalized Langevin equation (C.23) - therefore reduces smoothly, term by term, to the corresponding classical objects of Chapter 14 as \(\hbar \rightarrow 0\), while the form of every one of those objects is already exact at any finite \(\hbar\).

    C. 9 Ohmic Bath and the Markovian Quantum Langevin Equation

    In the continuum limit, with spectral density \(J(\omega)=\frac{\pi}{2} \sum_{j} c_{j}^{2} /\left(m_{j} \omega_{j}^{2}\right) \delta\left(\omega-\omega_{j}\right)\) as in Chapter 14, Eq. (C.22) becomes \(K(t)=\left(2 / m_{0} \pi\right) \int_{0}^{\infty} d \omega J(\omega) \cos (\omega t)\), exactly as before; for an Ohmic spectrum, \(J(\omega)=m_{0} \gamma \omega\), one again finds \(K(t) \approx 2 \gamma \delta(t)\) in the Markov limit, and Eq. (C.23) reduces to the quantum damped oscillator \(\ddot{\hat{q}}_{0}+2 \gamma \dot{\hat{q}}_{0}+\omega_{0}^{2} \hat{q}_{0}=\hat{F}(t) / m_{0}\). The genuinely quantum content is entirely in the noise strength: whereas the classical Markovian limit of Chapter 14 gives white noise of strength \(2 m_{0} \gamma k_{\mathrm{B}} T \delta(t-s)\), Eq. (C.27) gives, upon restricting to the resonant bath oscillators near \(\omega_{0}\) (the weak-coupling argument reviewed in Sec. 2.3 of [6]),

    \[ \frac{1}{2}\langle\{\hat{F}(t), \hat{F}(s)\}\rangle_{0} \approx 2 m_{0} \gamma \hbar \omega_{0} \operatorname{coth}\left(\frac{\hbar \omega_{0}}{2 k_{\mathrm{B}} T}\right) \delta(t-s), \tag{C.29} \label{C.29} \]

    matching Eq. (47) of Ghosh et al. [6]. The classical white-noise strength \(2 m_{0} \gamma k_{\mathrm{B}} T\) is recovered from Eq. (C.29) only in the high-temperature regime \(k_{\mathrm{B}} T \gg \hbar \omega_{0}\); at low temperature, \(\operatorname{coth}\left(\hbar \omega_{0} / 2 k_{\mathrm{B}} T\right) \rightarrow 1\) and the noise strength saturates at the zero-point value \(m_{0} \gamma \hbar \omega_{0}\), consistent with Remark C. 1 above.

    C. 10 Summary: Structural Parallels and Genuine Quantum Differences

    Table C. 1 collects the dictionary established in this chapter.

    The overall moral is a clean one. Everything in the Mori-Zwanzig construction that is built from the canonical (Kubo-transform) inner product - the correlation matrix \(\mathbb{C}\), the frequency matrix \(\Omega\), the memory kernel \(K(t)\), and the Mori-form fluctuation-dissipation relation - is exactly classical in form, for any \(\hbar\) and any temperature, once the renormalization counterterm of Sec. C. 1 is included; this is precisely the sense in which Mori's inner product was designed.

    Table \(\PageIndex{1}\): Classical versus quantum Mori-Zwanzig treatment of the Magalinski model.
    Ingredient Classical (Ch. 14) Quantum (this chapter)
    Generator \(\mathcal{L} A=\{H, A\}\) \(i \mathcal{L} \hat{X}=(i / \hbar)[\hat{H}, \hat{X}]\)
    Inner product \((A, B)=\langle A B\rangle_{0}\) Kubo transform, Eq. (C.8)
    Equipartition \(\left\langle q_{0}^{2}\right\rangle_{0}=k_{\mathrm{B}} T /\left(m_{0} \omega_{0}^{2}\right)\) \(\left(\hat{q}_{0}, \hat{q}_{0}\right) \quad=\quad k_{\mathrm{B}} T /\left(m_{0} \omega_{0}^{2}\right)\), Thm. C. 1
    Frequency matrix \(\Omega\) Equation \ref{14.12} identical, Eq. (C.16)
    Random force \(F(t)=-\sum_{j} c_{j} q_{j}(t)\) \(\hat{F}(t)=-\sum_{j} c_{j} \hat{Q}_{j}(t)\) (shifted)
    Memory kernel \(K(t)\) Equation \ref{14.17} identical, Eq. (C.22)
    "Second FDT" \(\langle F(t) F(s)\rangle_{0}=k_{\mathrm{B}} T m_{0} K(\mid t- s \mid\) ) \((\hat{F}(t), \hat{F}(s))=k_{\mathrm{B}} T m_{0} K(\mid t- s \mid\) )
    Physical noise strength \(2 k_{\mathrm{B}} T\) \(\hbar \omega \operatorname{coth}\left(\hbar \omega / 2 k_{\mathrm{B}} T\right)\)
    Noise at unequal times commutes with itself \([\hat{F}(t), \hat{F}(s)] \quad \neq \quad 0, \quad T\) -independent
    Noise at \(T=0\) vanishes persists (zero-point)

    Genuine quantum mechanics enters only when the formalism is asked to produce something measurable: the physical, symmetrized noise correlator, which carries the Bose thermal factor \(\hbar \omega \operatorname{coth}\left(\hbar \omega / 2 k_{\mathrm{B}} T\right)\) derived independently in Appendix A from the Kubo-Martin-Schwinger condition, and the noise commutator, which has no classical counterpart and encodes the irreducibly operator-valued character of quantum noise.

    Bibliography

    [1] V. I. Magalinski, JETP 36, 1942 (1959).

    [2] H. Mori, Transport, Collective Motion, and Brownian Motion, Prog. Theor. Phys. 33, 423 (1965).

    [3] R. Zwanzig, Nonequilibrium Statistical Mechanics (Oxford University Press, 2001).

    [4] R. K. Pathria and P. D. Beale, Statistical Mechanics (Elsevier, 2011).

    [5] R. Kubo, M. Toda, and N. Hashitsume, Statistical Physics II: Nonequilibrium Statistical Mechanics, 2nd ed. (Springer, 1991).

    [6] A. Ghosh, M. Bandyopadhyay, S. Dattagupta, and S. Gupta, Independent-oscillator model and the quantum Langevin equation for an oscillator: a review, J. Stat. Mech. (2024) 074002.

    [7] G. W. Ford, M. Kac, and P. Mazur, Statistical mechanics of assemblies of coupled oscillators, J. Math. Phys. 6, 504 (1965).

    [8] A. O. Caldeira and A. J. Leggett, Path integral approach to quantum Brownian motion, Physica A 121, 587 (1983).


    This page titled 18: Quantum Mori-Zwanzig Formulation of the Magalinski Model was last modified on Mon, 02 Nov 2026 06:09:27 GMT and is shared under a CC BY-NC-SA 4.0 license and was authored, remixed, and/or curated by Pabitra N. Sen.