Dissipation-Coherence Trade-Off in Open Systems

• Dissipation-Coherence Trade-Off is a framework that defines how irreversible processes constrain or enhance coherent behavior across stochastic, quantum, and information-theoretic systems.
• It quantifies coherence using observables like correlation time, phase regularity, and metrological gain, employing precision-dissipation bounds and thermodynamic uncertainty relations.
• Structured dissipation can act as a resource by engineering loss mechanisms that stabilize coherent steady states, enhance transport, and improve metrological sensitivity.

The dissipation–coherence trade-off denotes a class of relations in which sustained order, oscillatory regularity, localization, metrological sensitivity, or preserved superposition is constrained—or, in some open-system settings, actively shaped—by irreversible processes. In stochastic thermodynamics, the central question is how much entropy production is required to maintain coherent noisy oscillations. In open quantum dynamics, the same phrase can refer either to the destructive action of local noise on coherent resources or to regimes in which structured dissipation stabilizes coherence-bearing steady states, synchronized motion, or enhanced transport. In information-theoretic formulations, it appears as a complementarity between coherence and irreversible disturbance rather than as a literal heat or entropy-production law (Santolin et al., 30 Jan 2025, Roy et al., 18 Jun 2025, Sharma et al., 2017).

1. Terminological scope and principal observables

In the cited literature, the expression appears in several technically distinct forms. One family concerns stochastic oscillators, where coherence is quantified by a correlation time or by inter-spike-interval regularity, and dissipation is quantified by entropy production per period or per mean spike. A second family concerns open quantum systems, where coherence may mean off-diagonal structure in an energy or site basis, long-range first-order coherence, synchronized phase locking, or reduced-state purity, while dissipation is implemented through Lindblad jump operators or effective non-Hermitian loss terms. A third family is information-theoretic: coherence is measured by the relative entropy of coherence, and the competing quantity is disturbance under a CPTP map rather than thermodynamic entropy production (Santolin et al., 30 Jan 2025, Roy et al., 18 Jun 2025, Sharma et al., 2017).

Setting	Coherence quantity	Dissipation or competing quantity
Weak-noise limit cycles	$\tau_c$, $\mathcal N$	$\Sigma$ per period
Excitable phase oscillators	$F$, $CV$	$\sigma$ per mean ISI
Open quantum steady states	eigenbasis coherence, purity, PR	structured Lindblad loss
Quantum channels	$C_r(\rho)$	disturbance $D(\rho,\mathcal E)$
Quantum sensing	IMG $\mathcal I$	local dephasing or emission rate $\gamma$

This distribution of meanings is not merely terminological. It indicates that “coherence” is model-dependent: in one-dimensional excitable systems it is spike-time regularity, in noisy limit cycles it is phase memory, in Lindblad lattice models it is off-diagonal steady-state structure, and in metrology it is the time-integrated persistence of QFI-generated sensitivity. This suggests that the topic is best understood as a family of trade-offs rather than a single universal law.

2. Entropy-production bounds for noisy autonomous oscillations

For overdamped stochastic limit-cycle oscillators, the strongest thermodynamic formulation is a precision–dissipation bound on phase coherence. In the Langevin setting

$\mathcal N$0

with a stable deterministic periodic orbit $\mathcal N$1, the noisy oscillation has frequency $\mathcal N$2 and long-time decorrelation

$\mathcal N$3

Using

$\mathcal N$4

the main theorem is

$\mathcal N$5

for weak noise and uncorrelated noise components $\mathcal N$6. Near a supercritical Hopf bifurcation in macroscopic limits of Markov jump processes, the extension

$\mathcal N$7

shows that microscopic master-equation dissipation upper-bounds the coarse-grained Langevin cost. The same work emphasizes that the theorem is asymptotic and regime-specific: overdamped first-order Langevin dynamics, weak noise, smooth drift, a stable deterministic limit cycle, and, in its strongest form, isotropic white noise. It explicitly does not establish the inequality for underdamped dynamics, strong noise, colored noise, dynamics without a stable limit cycle, or general correlated-noise systems far from Hopf onset. Its examples further show that the bound can remain parametrically tight far from equilibrium, whereas in the Brusselator the microscopic entropy production can exceed coarse-grained dissipation by orders of magnitude, so coherence is not a reliable quantitative proxy for total biochemical free-energy cost (Santolin et al., 30 Jan 2025).

A later TUR-based derivation reformulated the same theme using a different normalization of coherence, $\mathcal N$8, and proved

$\mathcal N$9

The same construction yields a thermodynamic speed limit,

$\Sigma$0

with the two bounds arising from dual observables associated with the tangent Floquet mode and its dual covector. This makes the dissipation–coherence relation and the thermodynamic speed limit dual consequences of a short-time TUR in the weak-noise limit (Nagayama et al., 8 Sep 2025).

For finite Markov jump processes with autonomous noisy oscillations, the OBS conjecture,

$\Sigma$1

was refined to the rigorous bound

$\Sigma$2

where

$\Sigma$3

is a mode-uniformity factor of the dominant oscillatory eigenmode. This shows that an eigenvalue-only prefactor can fail when the oscillatory mode is localized. Translation-invariant rings satisfy $\Sigma$4, and the drift–diffusion limit on a circle saturates the bound (Gu, 3 Jun 2026).

3. Excitable oscillators: thermodynamic and dynamical constraints

Excitable phase oscillators introduce a distinct notion of coherence based on first-passage-time variability rather than phase diffusion. In the one-dimensional model

$\Sigma$5

one spike is the first-passage time $\Sigma$6 for one full turn, and coherence is measured by

$\Sigma$7

The entropy production during the mean ISI is exactly

$\Sigma$8

and the first-passage-time TUR gives

$\Sigma$9

For the active rotator, however, thermodynamics is not the only restriction. The critical $F$0 SNIC curve is a dynamical bound: it is a lower bound in the subthreshold excitable regime and an upper bound in the superthreshold oscillatory regime. At criticality,

$F$1

The same paper shows that coherence resonance is also bounded by the TUR, so the most coherent response occurs at an optimal noise strength rather than at maximal dissipation. In strongly coupled ensembles, synchronization reduces effective noise and yields the collective bound

$F$2

so the same coherence can be achieved at lower per-unit thermodynamic cost (Zheng, 2024).

This excitable case is important because it rules out a purely thermodynamic reading of the trade-off. In ordinary limit cycles, the main issue is phase diffusion around an already rotating orbit. In excitable systems, threshold geometry and refractory excursion create an additional dynamical source of timing variability, so more dissipation is necessary but not sufficient for arbitrarily high coherence.

4. Structured dissipation as a generator of coherence

Several open-system quantum models reverse the naive expectation that dissipation must suppress coherence. In a clean one-dimensional tight-binding chain with spatially phase-modulated two-site Lindblad operators,

$F$3

aperiodic dissipation alone can stabilize a localized steady state although the Hamiltonian has no disorder and no quasiperiodic potential. In the slow incommensurate regime with $F$4, $F$5, and $F$6, the steady state has

$F$7

whereas for fast modulation $F$8 these become

$F$9

The mechanism is coherence between Hamiltonian eigenstates generated by slowly varying aperiodic dissipation; rapid variation instead acts effectively like dephasing (Roy et al., 18 Jun 2025).

In frustrated flat-band lattices, local Markovian loss can likewise generate mobility and long-range first-order coherence. For the sawtooth lattice at the flat-band point, coherent hopping between Wannier states vanishes in the projected Hamiltonian, but local site loss becomes nonlocal in the Wannier basis after projection. The resulting dissipator couples otherwise decoupled localized modes, yielding finite remote occupation and long-range

$CV$0

The effect requires loss asymmetry across sublattices; for uniform loss, the nonlocal terms cancel, and interactions reduce but do not destroy the induced coherence and mobility (Owen et al., 2016).

In a spin-2 $CV$1 Bose-Einstein condensate, spin-dependent particle loss produces phase synchronization among the five Zeeman components and drives the gas toward a nearly fully magnetized transverse state. The observed transverse magnetization magnitude,

$CV$2

is close to the fully magnetized spin-2 value $CV$3, even though the conservative spin interaction is not ferromagnetic. The mechanism is selective loss: states with larger $CV$4 are less lossy because the imaginary part of the coefficient of $CV$5 is non-negative (Eto et al., 2018).

A two-species Bose–Josephson junction furnishes a dynamical version of the same theme. With incoherent hopping rate $CV$6, weak interactions $CV$7 support synchronized phase-locked oscillations. Stronger interactions destabilize this regime and produce transient chaos, but dissipation later suppresses scrambling and restores coherence by driving the dynamics toward a stable self-trapped attractor. A controlled tilt destabilizes that attractor and converts transient chaos into persistent steady-state chaos, eliminating long-time coherence recovery (Mondal et al., 18 Feb 2026).

Taken together, these results suggest that the decisive variable is not dissipation strength alone but dissipation structure: symmetry, phase modulation, dark-state geometry, or attractor formation can turn loss into a coherence-preserving or even coherence-generating resource.

5. Quantum thermodynamic and metrological trade-offs

In open quantum thermodynamics, coherence can either worsen or relax current–dissipation trade-offs depending on its spectral location. For Davies-type Lindblad dynamics, coherence between different energy eigenspaces is never beneficial: $CV$8 For fully dephased states,

$CV$9

whereas block-diagonal states with coherence inside degenerate subspaces satisfy

$\sigma$0

In the collective $\sigma$1-state model studied there, $\sigma$2 while $\sigma$3, enabling an $\sigma$4 heat current with $\sigma$5 entropy production in the large-$\sigma$6 limit (Tajima et al., 2020).

A complementary result arises in pure dephasing. For a central qubit coupled to a finite Ising-like spin environment, the system energy does not change,

$\sigma$7

but the mean heat deposited into the environment equals the coherent-energy contribution in a reformulated first law: $\sigma$8 In the model’s numerical dynamics, each peak of $\sigma$9 coincides with a minimum of the qubit $C_r(\rho)$0-coherence $C_r(\rho)$1, while coherence revivals coincide with reduced heat; the revivals are simultaneously positive BLP information backflow (Lenzarini et al., 28 Mar 2026).

In quantum sensing, the trade-off becomes operationally metrological. For magnetic-field estimation with $C_r(\rho)$2 spins under local dephasing, the integrated metrological gain

$C_r(\rho)$3

is

$C_r(\rho)$4

for GHZ probes and

$C_r(\rho)$5

for separable product probes. Under local emission,

$C_r(\rho)$6

The short-time GHZ gain is Heisenberg-enhanced, but the same entanglement accelerates dissipative decay, so long-time integrated performance scales inversely with the local dissipation rate and becomes comparable to that of unentangled probes (Takahashi et al., 25 Dec 2025).

6. Information-theoretic formulations and limits of generality

Not every “dissipation–coherence” relation is thermodynamic in the narrow sense. For a finite-dimensional quantum system under a CPTP map $C_r(\rho)$7, the relative entropy of coherence

$C_r(\rho)$8

and Maccone disturbance obey

$C_r(\rho)$9

For measurement channels, the tighter relation

$D(\rho,\mathcal E)$0

holds. Here disturbance is defined via coherent information and quantifies irreversibility of state change, not thermodynamic entropy production, heat, or work (Sharma et al., 2017).

A neighboring but still indirect framework is the dissipation–relaxation trade-off for detailed-balance relaxation. In the quantum extension based on a quantum logarithmic-Sobolev inequality,

$D(\rho,\mathcal E)$1

with a corresponding inverse speed limit

$D(\rho,\mathcal E)$2

This constrains decay of total nonequilibrium structure, including but not isolating coherence (Bao et al., 2023).

The scope conditions in the direct theorems are therefore decisive. The strongest oscillator bound is weak-noise and overdamped (Santolin et al., 30 Jan 2025). The aperiodic-dissipation localization mechanism is a one-dimensional single-particle Lindblad problem with numerical rather than full analytic Liouvillian control (Roy et al., 18 Jun 2025). The pure-dephasing heat–coherence identity assumes a finite environment and a TPM-based energetic decomposition (Lenzarini et al., 28 Mar 2026). The metrological trade-off is established for local dephasing and local emission, not for arbitrary noise models (Takahashi et al., 25 Dec 2025). Taken together, these results suggest that there is no single scalar law of the form “more dissipation always destroys coherence” or “more coherence always requires the same thermodynamic cost” across all regimes. What is robust is a narrower statement: the relation between irreversibility and coherence is structural, and its precise form depends on what counts as coherence, how irreversibility is measured, and whether dissipation acts as unbiased noise or as a specifically engineered dynamical resource.