---
title: 'Floquet Damping Matrix: Concepts & Applications'
url: https://www.emergentmind.com/topics/floquet-damping-matrix
type: topic
---

# Floquet Damping Matrix: Concepts & Applications

Searching arXiv for the specified papers and closely related work on Floquet damping matrices and Floquet-Lindblad/open-system generators.
I’ll retrieve arXiv records for the cited works and related papers before composing the article.
arXiv search query: Floquet damping matrix open quantum systems Floquet Lindblad propagator exceptional points non-Markovian open system dynamics
A Floquet damping matrix is a finite-dimensional object that encodes decay, growth, or dissipative mode structure in a periodically driven system over one driving period. In the cited literature, the term does not have a single universal meaning. It can denote a dissipative Floquet generator obtained from the matrix logarithm of a one-period open-system propagator, a Floquet electronic friction tensor governing Langevin or Fokker–Planck dynamics, or the logarithm of a monodromy matrix in Floquet–Lyapunov theory for linear systems with periodic coefficients [2511.08754, 2306.06128, 1206.0716, 2011.02054]. The common mathematical core is the replacement of time-local periodic dynamics by a stroboscopic operator over one period, followed by spectral analysis of that operator or its logarithm.

## 1. Terminological scope and formal setting

The cited works use closely related but non-identical constructions. The shared structure is a \(T\)-periodic evolution law and a one-period propagator whose spectrum determines long-time behavior.

| Setting | Object | Definition or role |
|---|---|---|
| Periodic linear system | \(R\) | \(R=\tfrac{1}{T}\ln \Phi(T)\) |
| Floquet-Lindblad evolution | \(\mathcal L_F\) | Defined implicitly by \(G(T)=e^{T\mathcal L_F}\) |
| Non-Markovian driven open system | \(G\) | \(Q_F=e^{GT}\) |
| Floquet electronic friction | \(\gamma_{\alpha\beta}(\mathbf R)\) | Friction or damping tensor |

In the linear periodic setting, the fundamental solution \(\Phi(t)\) admits a Floquet–Lyapunov factorization
\[
\Phi(t)=P(t)e^{Rt}, \qquad P(t+T)=P(t), \qquad R=\tfrac1T\ln \Phi(T),
\]
with \(M=\Phi(T)\) the monodromy matrix. The eigenvalues of \(R\) separate growth or damping from oscillation by their real and imaginary parts [1206.0716].

In open quantum systems, the analogous object is constructed from a stroboscopic propagator on Liouville space or on an enlarged system–auxiliary space. The generator extracted by a matrix logarithm is the dissipative Floquet analogue of a Floquet Hamiltonian, but its eigenvalues generally have negative real parts rather than being purely imaginary [2511.08754].

A central terminological caution follows immediately. In the Floquet-Lindblad qubit study, the analysis is performed entirely in the time domain through the one-period propagator and an implicitly defined \(\mathcal L_F\); that work does not introduce an explicit infinite-dimensional harmonic-space Floquet matrix or a truncated Fourier-space “damping matrix” \(M_F\) [2011.02054].

## 2. Stroboscopic generators and matrix logarithms

For a periodically driven open system in Lindblad form, vectorization of the density matrix yields
\[
\partial_t|\rho^{v}(t)\rangle=\mathcal L(t)|\rho^{v}(t)\rangle,
\]
where \(\mathcal L(t)\) is an \(N^2\times N^2\) Liouvillian matrix,
\[
\mathcal{L}(t) =
 - i[\,1\!\otimes\!H_S(t)-H_S^{T}(t)\!\otimes\!1\,]
-\sum_k [\gamma_k(t)/2]\,[\,1\!\otimes\!F_k^\dagger F_k +(F_k^\dagger F_k)^T\!\otimes\!1 -2F_k^*\!\otimes\!F_k\,].
\]
If \(\mathcal L(t+T)=\mathcal L(t)\), the one-period propagator is
\[
G(T)=\mathcal T\exp\!\left[\int_0^Tdt'\,\mathcal L(t')\right]\equiv e^{T\mathcal L_F},
\]
and the Floquet problem is formulated through
\[
G(T)|\rho_k^v\rangle=\lambda_k|\rho_k^v\rangle, \qquad |\lambda_k|\le 1.
\]
The steady state satisfies \(\lambda_0=1\), while the remaining \(\lambda_{k>0}\) control decay toward it [2011.02054].

The same stroboscopic logic appears in exact non-Markovian open-system dynamics, but there the propagator is built on a combined system–bond space rather than directly from a Lindbladian. One defines a finite-dimensional \(Q_F\) over one period and then sets
\[
Q_F=e^{GT}, \qquad \gamma_i=\frac{1}{T}\ln \lambda_i,
\]
where \(\lambda_i\) are the eigenvalues of \(Q_F\). The real parts of \(\gamma_i\) are damping rates and the imaginary parts are Floquet quasi-frequencies modulo \(2\pi/T\) [2511.08754].

The matrix logarithm is therefore not a formal ornament but the step that converts stroboscopic evolution into a generator-like object. The branch choice of \(\ln \lambda_i\) matters because quasi-frequencies are defined only modulo \(2\pi/T\) [2511.08754].

## 3. Floquet-Lindblad decay, Bloch reduction, and exceptional contours

A concrete Floquet-Lindblad realization is provided by a periodically driven qubit with a single dissipator, treated either as mode-selective spontaneous emission or as pure phase damping. In Bloch-vector form,
\[
\rho=\frac{1+s\cdot \sigma}{2}, \qquad s=(s_x,s_y,s_z)^T, \qquad \frac{ds}{dt}=A(t)s+b(t).
\]
For mode-selective spontaneous emission with \(F=\sigma_-\), \(H_S(t)=-J(t)\sigma_x\), \(b(t)=[0,0,\gamma_-(t)]^T\), and
\[
A(t)=-
\begin{pmatrix}
(\gamma_-/2+2\gamma_z) & 0 & 0\\
0 & (\gamma_-/2+2\gamma_z) & -2J\\
0 & 2J & \gamma_-
\end{pmatrix},
\]
with \(\gamma_z=0\) in pure \(\sigma_-\) dissipation. For pure phase damping with \(F=\sigma_z\), \(H_S(t)=-J(t)\sigma_x\), \(b(t)=0\), and
\[
A(t)=-
\begin{pmatrix}
2\gamma_z & 0 & 0\\
0 & 2\gamma_z & -2J\\
0 & 2J & 0
\end{pmatrix},
\]
with \(\gamma_-=0\) [2011.02054].

Exceptional points are detected from the eigenvectors of the one-period propagator rather than from a Fourier-expanded Floquet matrix. The criterion is based on the maximum inner product among the three trace-zero eigenmatrices,
\[
IP=\max_{m>n\ge 1}\,|\langle \rho_m^v|\rho_n^v\rangle|,
\]
with an exceptional point identified when \(IP\to 1\). By scanning drive or dissipation strength together with \(\Omega\), one obtains contours of exceptional points in parameter space. For sinusoidal \(J(t)\), the contours emerge periodically at subharmonic resonance \(\Omega_n=2J/n\); for modulated \(\gamma(t)\), they appear at \(\Omega_m=4J/(2m+1)\) [2011.02054].

The physical interpretation is stroboscopic critical damping. In the static Lindblad problem, an exceptional point marks the transition between underdamped and overdamped relaxation toward the steady state. In the periodically driven case, the same distinction is read off from the eigenvalues \(\lambda_k\) of the one-period map: real \(\lambda_k\) correspond to overdamped decay, while a complex-conjugate pair corresponds to underdamped oscillatory decay in the stroboscopic dynamics. The reported Floquet exceptional points persist down to vanishingly small dissipation when \(\Omega\) is tuned to a multiphoton resonance, a feature absent in the static Lindblad spectrum [2011.02054].

## 4. Exact non-Markovian construction on system–bond space

In the strong-coupling, non-Markovian setting, the starting point is a small driven system with \(T\)-periodic Hamiltonian \(H_{\rm sys}(t)=H_{\rm sys}(t+T)\), coupled bilinearly to a bosonic environment. Time is discretized as \(T=M\delta t\). The environment is traced out through the Feynman–Vernon influence functional, which is represented as a periodic matrix product operator. For a Gaussian bath, the undriven influence functional is compressed into matrices \(q^{(\mu\mu')}\) of bond dimension \(D\ll \dim(\mathcal H_{\rm env})\), together with boundary vectors \(v_l,v_r\) [2511.08754].

Including the local system drive by a Trotter breakup yields micro-step dissipative propagators \(Q_n\) acting on system-\(\otimes\)-bond space:
\[
(Q_n)_{r,r'}^{\mu,\mu'}
=\sum_{\alpha,\alpha'}
\bigl[\mathcal U_{\rm sys}^{\mu,\alpha}(t_n,t_n-\tfrac{\delta t}{2})\bigr]\,
q_{r,r'}^{(\alpha,\alpha')}\,
\bigl[\mathcal U_{\rm sys}^{\alpha',\mu'}(t_n-\tfrac{\delta t}{2},t_{n-1})\bigr].
\]
Periodic driving gives \(Q_{n+M}=Q_n\), so the one-period stroboscopic propagator is
\[
Q_F\coloneqq Q_MQ_{M-1}\cdots Q_1.
\]
By construction, \(Q_F\) is a finite-dimensional matrix of size \(d^2D\times d^2D\), where \(d=\dim(H_{\rm sys})\), and it gives the exact stroboscopic open-system evolution including non-Markovian memory [2511.08754].

The Floquet damping matrix is then defined as
\[
G=\frac1T\ln Q_F,
\]
or equivalently through the spectral decomposition
\[
Q_F=\sum_i \lambda_i |v_i\rangle\langle w_i|,\qquad
G=\frac1T\sum_i \ln\lambda_i\,|v_i\rangle\langle w_i|,
\]
with \(\langle w_i|v_j\rangle=\delta_{ij}\). Typically \(|\lambda_i|<1\) except for one steady-state mode with \(\lambda_1=1\), so the corresponding exponents \(\gamma_i=(1/T)\ln\lambda_i\) have negative real parts. Their imaginary parts, taken modulo \(2\pi/T\), are the Floquet quasi-frequencies of damped modes [2511.08754].

This construction reduces continuously to the closed-system Floquet problem. In the limit \(g_\lambda\to 0\), the tensors \(q^{(\mu\mu')}\) become \(\delta_{\mu\mu'}I_D\), the \(Q_n\) become pure-unitary superoperators, and \(Q_F\to \mathcal U_F=\mathcal U_{\rm sys}(T,0)\) with
\[
Q_F\to e^{-i\mathcal H_FT}.
\]
Accordingly,
\[
G\to -i\mathcal H_F,
\]
all eigenvalues lie on the unit circle, and the damping rates vanish [2511.08754].

## 5. Spectral information, heating, and transient entanglement

The utility of a Floquet damping matrix lies in what its spectrum organizes. In the driven spin–boson model,
\[
H_{\rm sys}(t)=\frac12\Omega\sigma_x+\epsilon_d\cos(\omega_dt)\sigma_z,\qquad
S=\sigma_z,\qquad
J(\omega)=\frac{\alpha}{2}\omega e^{-\omega/\omega_c},
\]
the steady-state eigenvector \(v_1\) of \(Q_F\) with \(\lambda_1=1\) gives the Floquet stationary density matrix \(\rho_{ss}\). The period-averaged heat current density
\[
\bar j(\omega)=\frac{1}{T}\int_0^Tdt\,j(t,\omega)
\]
is obtained from the End–Matter formula for \(j(t,\omega)\), with two-point correlations \(\langle S(t)S(s)\rangle\) reconstructed by interleaving the micro-motion operators \(Q_n\) with the system-state vector in bond space. The spectral decomposition of \(Q_F\) also yields the long-time decay rates of correlations, which enter the frequency dependence of the heat current [2511.08754].

For two driven qubits with
\[
S=\frac{\sigma_z^A+\sigma_z^B}{2},\qquad
H_{\rm sys}(t)=\frac12\Omega(\sigma_x^A+\sigma_x^B)+\frac12\epsilon_d\cos(\omega_dt)(\sigma_x^A+\sigma_x^B),
\]
the spectrum of \(G=(1/T)\ln Q_F\) contains one steady mode \(\gamma_1=0\) and a ladder of damped modes with nonzero \(\operatorname{Im}\gamma_i\) and negative \(\operatorname{Re}\gamma_i\). In the undriven limit, the auxiliary propagator \(q\) alone has eigenmodes governing quench dynamics,
\[
\rho(t)=\sum_i e^{\gamma_it}\rho_i.
\]
One transient mode \(\gamma_*\) has an associated \(\rho_*\) with high concurrence. Choosing \(\omega_d\approx \operatorname{Im}\gamma_*\) resonantly stabilizes this transient entangled mode in the Floquet steady state, and numerically the asymptotic concurrence from the leading eigenvector of \(Q_F\) reaches a maximum \(\simeq 0.5\) near \(\omega_d\approx 2\Omega\) [2511.08754].

In the Floquet-Lindblad qubit problem, the spectrum of the one-period propagator plays a different but related role. There the relevant structure is not entanglement stabilization or heat transport but the boundary between underdamped and overdamped approach to the Lindblad steady state. The exceptional contours of the stroboscopic map therefore act as critical-damping manifolds in parameter space [2011.02054].

## 6. Floquet electronic friction, multidimensional damping tensors, and conceptual distinctions

A distinct usage of “Floquet damping matrix” appears in nonadiabatic dynamics near metal surfaces, where periodic driving of an Anderson–Holstein model leads first to a Floquet classical master equation and, in the fast-drive and fast-electron limits, to a Floquet Fokker–Planck equation. For the one-dimensional nuclear problem, the resulting effective equation is
\[
\frac{\partial A}{\partial t}
=-\frac{P}{M}\partial_RA
+\frac{\partial U(R,t)}{\partial R}\partial_PA
+\gamma_e(R)\,\partial_P[PA]
+D(R)\,\partial_P^2A,
\]
with mean force
\[
\frac{\partial U}{\partial R}
=\partial_RV_0(R)+\partial_RE(R)\,\tilde f(E(R),t),
\]
electronic friction coefficient
\[
\gamma_e(R)
=-\frac{1}{M\Gamma}\,\partial_RE(R)\,\frac{d\bar{\tilde f}(E)}{dR},
\]
and diffusion strength
\[
D(R)=\gamma'_e(R)MkT,\qquad
\gamma'_e(R)=\frac{\beta}{M\Gamma}\,[\partial_RE(R)]^2\,\bar{\tilde f}\,(1-\bar{\tilde f}).
\]
The equivalent Langevin dynamics is
\[
\dot R=\frac{P}{M},\qquad
\dot P=-\frac{\partial U(R,t)}{\partial R}-\gamma_e(R)P+\xi(t),
\]
with \(\langle \xi(t)\xi(t')\rangle=2D(R)\delta(t-t')\) [2306.06128].

In multiple nuclear dimensions, the scalar friction becomes a tensor \(\gamma_{\alpha\beta}(\mathbf R)\). The explicit Floquet expression is
\[
\gamma_{\alpha\beta}(\mathbf R)
=\hbar\sum_{m,n=-\infty}^{\infty}\int_{-\infty}^{\infty}\frac{d\epsilon}{2\pi}\,
J_m(z)J_n(z)\,
\frac{f(\epsilon)-f\!\bigl(\epsilon+(n-m)\hbar\Omega\bigr)}{(n-m)\hbar\Omega}
\Bigl[F_\alpha A(\epsilon)F_\beta A\bigl(\epsilon+(n-m)\hbar\Omega\bigr)\Bigr],
\]
where \(A(\epsilon)=i[G^R(\epsilon)-G^A(\epsilon)]\), \(J_k(z)\) are Bessel weights with \(z=A/(\hbar\Omega)\), and the trace over electronic indices is implicit in the matrix product [2306.06128].

The physical distinction between this tensor and the stroboscopic logarithm \(G=(1/T)\ln Q_F\) is substantial. The former is a friction kernel entering an effective nuclear Langevin description; the latter is a generator of one-period dissipative evolution for an open quantum system. Both encode damping, but they do so in different state spaces and at different levels of coarse graining.

Several common conceptual confusions are resolved by the cited literature. First, a Floquet damping matrix is not necessarily a harmonic-space block-Toeplitz construction. The Floquet-Lindblad qubit analysis extracts all spectral information directly from the finite one-period propagator and does not construct an infinite-dimensional harmonic-space matrix or any truncated \(M_F\) [2011.02054]. Second, a Floquet damping matrix is not necessarily Markovian: the system–bond-space construction produces an exact non-Markovian \(Q_F\) and generator \(G\) [2511.08754]. Third, periodic driving can invalidate equilibrium fluctuation–dissipation structure. In the Floquet electronic friction formulation, \(\gamma'_e\neq\gamma_e\) in general, so the second fluctuation–dissipation theorem is violated, and the driven nuclear degree of freedom heats up to an effective temperature \(T_{\rm eff}>T\) in the long-time limit [2306.06128].

From a computational standpoint, the required numerics depend on which object is meant. In Floquet electronic friction, one truncates sidebands \(m,n\in[-N_{\max},N_{\max}]\), builds the Floquet Hamiltonian in extended space, computes \(G^{R/A}(\epsilon)\) and \(A(\epsilon)\), evaluates the double sum and energy integral, and assembles \(\gamma_{\alpha\beta}(\mathbf R)\) at each geometry. In the exact non-Markovian quantum construction, one compresses the influence functional with an MPO method such as uniTEMPO, builds the periodic micro-step tensors \(Q_n\), forms \(Q_F\), and diagonalizes it or takes its logarithm to obtain the damping spectrum [2306.06128, 2511.08754].

Source: https://www.emergentmind.com/topics/floquet-damping-matrix