Hierarchy of entropy production and thermodynamic trade-off relations in non-Markovian systems

Ken Funo, Tan Van Vu, Keiji Saito · 2026

Read Hierarchy of entropy production and thermodynamic trade-off relations in non-Markovian systems by Ken Funo, Tan Van Vu, Keiji Saito (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-25245.txt.

Introduction

Funo, Vu and Saito study entropy in non-Markovian systems. A nesting of Langevin motion from a Gaussian bath shows reduced entropy splits into bigger Markovian entropy plus a link part, giving a ladder. The ladder yields trade-offs: entropy bounds, power-use trade-offs, uncertainty rules, speed limits. Bath memory thus constrains speed and engine output. The ladder extends beyond Gaussian baths to non-quantum and quantum baths.

Catalog record

CREATORKen Funo, Tan Van Vu, Keiji Saito
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE96653
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.25245

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Hierarchy of entropy production and thermodynamic trade-off relations in non-Markovian systems Ken Funo, Tan Van Vu, Keiji Saito Abstract Non-Markovian dynamics arise when a system is coupled to a bath with finite correlation time, giving rise to memory effects that allow the bath to temporarily store and return excitations. However, how memory modifies irreversibility and whether it can be exploited to improve thermodynamic performance is not well established. We address this question by employing a Markovian embedding of generalized Langevin dynamics, in which bath memory is encoded in auxiliary modes and irreversible dissipation in a residual Markovian bath. Here we show that the entropy production defined for the original non-Markovian system upper bounds that of the embedded system, thereby establishing a hierarchy of entropy production under Markovian embedding. Leveraging this hierarchy, we derive non-Markovian extensions of the thermodynamic uncertainty relation, speed limit, and power-efficiency trade-off. For underdamed generalized Langevin systems, sufficiently structured baths allow finite heat currents at vanishingly small entropy production, whereas Carnot efficiency at finite power remains unattainable for ordinary spectral densities. In the overdamped regime, memory effects can simultaneously reduce entropy production and current fluctuations, thereby enhancing the precision-to-dissipation ratio. We further discuss the extension of the hierarchy to generic bath models and the quantum regime. These results provide a quantitative framework for exploiting memory as a thermodynamic resource and for designing energy-efficient protocols in structured environments. Introduction Many open systems of experimental and theoretical interest across biology, chemistry, and physics are coupled to structured environments whose response is not instantaneous 55 ; 24 ; 10 ; 31 ; 17 . The resulting non-Markovian dynamics, in which fluctuations and dissipation depend on the system’s past evolution, reflects bath memory and the backflow of information from the environment 6 ; 11 . Within the framework of stochastic thermodynamics 42 ; 36 ; 47 ; 43 ; 19 ; 51 ; 21 , entropy production quantifiying irreversibility is always nonnegative, but memory effects can render its rate transiently negative 48 ; 29 , indicating that the bath can temporarily store and return energy to the system 17 . This suggests that memory is a potential physical resource, as has been explored in the contexts of channel capacities 8 , metrology 58 , information erasure 7 , and heat engines 41 ; 26 . A powerful framework for analyzing such dynamics is the Markovian embedding technique 24 ; 10 ; 22 ; 23 ; 21 ; 15 ; 52 ; 33 ; 49 ; 54 , which represents the non-Markovian bath as a combination of auxiliary modes encoding the memory and a residual Markovian bath. In this picture, the auxiliary modes can temporarily store and return excitations to the system, whereas the residual bath accounts for irreversible dissipation. Beyond reproducing the reduced system dynamics, this embedding reveals two distinct thermodynamic system-bath boundaries, one associated with the original non-Markovian system description and another associated with the enlarged embedded Markovian system (see Fig. 1 ). These two boundaries naturally define two entropy productions: the former captures both memory and dissipative effects, whereas the latter isolates the irreversible, memoryless contribution associated with the residual bath. Understanding their relation is thus central to disentangling memory effects from irreversibility. Entropy production is fundamental to finite-time thermodynamics, where trade-off relations, such as thermodynamic uncertainty relations (TURs) 44 ; 18 ; 3 ; 16 , speed limits 1 ; 45 ; 56 , entropic bounds 13 , and power-efficiency bounds 46 ; 39 , constrain current precision and operational speeds through entropy production, and are well established in Markovian systems. In the non-Markovian regime, however, entropy production acquires an additional memory-induced contribution, and it remains unclear how this modifies finite-time trade-off relations. A precise relation between the entropy production of the original non-Markovian description and that of the embedded Markovian description is therefore essential. Figure 1: Hierarchy of entropy production under Markovian embedding. a Original non-Markovian description ( 1 ) with the entropy production \Sigma and heat Q_{\rm sys} based on system observables. b Equivalent Markovian embedded description ( 3 ), in which the bath is given by auxiliary modes A and a residual Markovian bath. The inner system-bath boundary recovers (\Sigma,Q_{\rm sys}) , whereas the outer embedding boundary gives (\Sigma_{\rm emb},Q_{\rm emb}) . c The difference between \Sigma and \Sigma_{\rm emb} is the non-negative memory contribution \Sigma_{\rm mem} between the system and auxiliary modes, yielding the hierarchy \Sigma\geq\Sigma_{\rm emb} ( 5 ). We plot \Sigma and \Sigma_{\rm emb} by varying a parameter \epsilon that controls \Sigma_{\rm mem} (see Methods for details). Here we establish a hierarchy between the entropy productions of the original non-Markovian system and the embedded system, and apply it to derive finite-time thermodynamic trade-off relations in the non-Markovian regime. We further obtain a decomposition of the entropy production into a memoryless contribution associated with dissipation to the residual bath and a memory contribution with the auxiliary modes (see Fig. 1 ). Leveraging this hierarchy, we derive entropic bounds on currents and power-efficiency trade-off relations for underdamped generalized Langevin systems, and thermodynamic speed limits and TURs in the overdamped regime. We discuss in detail how the memory effect quantitatively modifies each bound. In the underdamped case, we identify spectral conditions under which the memory-controlled prefactor of the entropic bound diverges, allowing finite heat currents at vanishingly small entropy production, and conditions under which it remains finite, prohibiting Carnot efficiency at finite power. In the overdamped case, we extend the hierarchy further and show that memory forces give a negative correction to the apparent Markovian entropy production. We show that the precision-to-dissipation ratio can exceed its Markovian value, demonstrating that bath memory is an exploitable thermodynamic resource. Results Gaussian bath model.— We start by assuming a Gaussian bath model or the Zwanzig model 59 in which the bath is given by a collection of harmonic oscillators that are linearly coupled to the system. In the following, we assume that the system Hamiltonian is given by H_{S}^{\lambda_{t}}=\frac{P^{2}}{2M}+V^{\lambda_{t}}_{S}(X) , where \lambda_{t} describes the time-dependence of the potential. Starting from the system-bath Hamiltonian dynamics, the reduced dynamics of the system is described by the generalized Langevin equation 59 ; 10 ; 24 : \displaystyle\frac{dX}{dt} \displaystyle=\frac{P(t)}{M} \displaystyle\frac{dP}{dt} \displaystyle=-\partial_{X}V^{\lambda_{t}}_{S}-\int^{t}_{0}dsK(t-s)\frac{P(s)}{M}+\eta_{t}, (1) where K(t):=\frac{2}{\pi}\int_{0}^{\infty}d\omega\frac{J(\omega)}{\omega}\cos\omega t is the memory kernel, J(\omega) is the bath spectral density that reflects the property of the bath, and \eta_{t} is the colored Gaussian noise satisfying \mathbb{E}[\eta_{t}]=0 and the fluctuation-dissipation relation \mathbb{E}[\eta_{t}\eta_{s}]=K(t-s)/\beta . Here, \beta is the inverse temperature of the bath. The entropy production \Sigma from t=0 to t=\tau for the non-Markovian system described by Eq. ( 1 ) is defined as 36 ; 47 ; 43 ; 19 ; 51 \displaystyle\Sigma:=\Delta S_{S}-\beta Q_{\rm sys}\geq 0, (2) where \Delta S_{S} is the change of the Shannon entropy of the system and Q_{\rm sys}:=\int^{\tau}_{0}dt\dot{Q}_{\rm sys} . Here, the heat current is defined based on the first law of thermodynamics as \dot{Q}_{\rm sys}:=\frac{d}{dt}\langle H_{S}^{\lambda_{t}}\rangle-\dot{W} , where \dot{W}:=\dot{\lambda}_{t}\langle\partial_{\lambda_{t}}H_{S}^{\lambda_{t}}\rangle is the work rate. The entropy production quantifies the total entropy generated over the entire system and serves as a measure of thermodynamic irreversibility. Crucially, the entropy production ( 2 ) depends only on the probability distribution f^{S}_{t} and the observables of the system. Markovian embedding.— Let us now focus on the

Read the complete text