arXiv subjects

Takahiro Sagawa

Publications and source records attributed to Takahiro Sagawa.

At least 19 recordsLinked to original sources

arXiv2026

Optimizing optimal transport: Role of final distributions in finite-time thermodynamics

Performing thermodynamic tasks within finite time while minimizing thermodynamic costs is a central challenge in stochastic thermodynamics. Here, we develop a unified framework for optimizing the thermodynamic cost of performing various tasks in finite time for overdamped Langevin systems. Conventional optimization of thermodynamic cost based on optimal transport theory leaves room for varying the final distributions according to the intended task, enabling further optimization. Taking advantage of this freedom, we use Lagrange multipliers to derive the optimal final distribution that minimizes the thermodynamic cost. Our framework applies to a wide range of thermodynamic tasks, including particle transport, thermal squeezing, and information processing such as information erasure, measurement, and feedback. Our results are expected to provide design principles for information-processing devices and thermodynamic machines that operate at high speed with low energetic costs.

cond-mat.stat-mech↗

arXiv2026

Stochastic Thermodynamics for Autoregressive Generative Models: A Non-Markovian Perspective

Autoregressive generative models -- including Transformers, recurrent neural networks, classical Kalman filters, state space models, and Mamba -- all generate sequences by sampling each output from a deterministic summary of the past, producing genuinely non-Markovian observed processes. We develop a general theoretical framework based on stochastic thermodynamics for this class of architectures and introduce the entropy production, which can be efficiently estimated from sampled trajectories without exponential overhead, despite the non-Markovian nature of the observed dynamics. As a proof-of-concept experiment with a large language model (LLM), we evaluate the entropy production for a pre-trained Transformer-based model, GPT-2. We find that the token-level entropy production is dominated by a syntactic artifact, while the sentence-level entropy production tends to be larger for causally ordered than for non-causal text sets. This observation is supported by a re-evaluation with a substantially larger model, Qwen3-4B-Base. We also demonstrate the framework in the linear Gaussian case, where the model reduces to the Kalman innovation representation and the entropy production admits an analytical expression. We also show that the entropy production decomposes exactly into non-negative per-step contributions in terms of retrospective inference, and each of those terms further splits into information-theoretically meaningful terms: a compression loss and a model mismatch. Our results establish a bridge between stochastic thermodynamics and modern generative models, and provide a starting point for using irreversibility as a quantitative probe of the highly non-Markovian processes generated by models such as LLMs.

cond-mat.stat-mech↗

arXiv2026

Quantum enhancement of information-mediated energy transfer

Thermodynamics of information identifies information flow as a thermodynamic resource, but whether quantum coherence and collective coupling can enhance it at low entropy-production cost remains unresolved. We address this question for interacting open quantum systems by deriving a thermodynamic uncertainty relation that bounds information flow in terms of entropy production in nonequilibrium steady states. We construct a quantum engine in which N-fold degenerate ground and excited states are collectively coupled to heat baths. Collective jumps enhance heat currents and information flow linearly with N, while entropy production remains independent of N, realizing a high-power, low-dissipation autonomous quantum Maxwell's demon that leverages collectively enhanced information flow to pump heat against a temperature gradient. Beyond steady states, collective interactions amplify the unitary component of quantum information flow, yielding a quadratic enhancement of the free-energy charging power of quantum batteries. Our results reveal scalable advantages of quantum coherence and collective effects in quantum engines.

cond-mat.stat-mech↗

arXiv2026

Optimal Work Extraction from Finite-Time Closed Quantum Dynamics

Extracting useful work from quantum systems is a fundamental problem in quantum thermodynamics. In scenarios where rapid protocols are desired -- whether due to practical constraints or deliberate design choices -- a fundamental trade-off between power and efficiency is yet to be established. Here, we investigate the problem of finite-time optimal work extraction from closed quantum systems, subject to a constraint on the magnitude of the control Hamiltonian. We first reveal the trade-off relation between power and work under a general setup, showing that these fundamental performance metrics cannot be maximized simultaneously. We then identify a solvable class of finite-time optimal work-extraction problems. This class includes nontrivial many-body models such as the Heisenberg model and the SU(n)-Hubbard model. The key assumption is that the control Hamiltonian is optimized over a Lie algebra preserved by the uncontrolled dynamics. Within this class, the optimal work-extraction problem admits an exact reduction to a nonlinear self-consistent equation, circumventing extensive search over time-dependent control paths. The resulting optimal protocol turns out to be particularly simple: it suffices to use a time-independent control Hamiltonian in the interaction picture, determined by that equation. By exploiting the Lie-algebraic structure of the controllable terms, our approach is applicable to quantum many-body systems through efficient numerical computation. Our results highlight the necessity of rapid protocols to achieve the maximum power and provide an exact route to finite-time optimal work extraction in many-body quantum systems.

quant-ph↗

arXiv2026

Stochastic Thermodynamics on Time-Evolving Curved Spaces

We construct stochastic thermodynamics of overdamped Langevin systems on nonrelaticvistic curved spaces with time-dependent metrics. The time dependence of the metric contributes to the energy balance by performing work on the kinetic energy, which is instantaneously dissipated as heat in the overdamped regime. This contribution makes our framework thermodynamically consistent so that entropy production satisfies the second law of thermodynamics. As a special case, when the metric evolves according to backward Ricci flow, the entropy balance exhibits a structure similar to Perelman's entropy functional. Our framework provides a way to quantify thermodynamic costs in dynamics on time-evolving spaces such as diffusion on membranes.

cond-mat.stat-mech↗

arXiv2026

Symmetry and Topology of Successive Quantum Feedback Control

We establish a symmetry classification for a general class of quantum feedback control. For successive feedback control with a non-adaptive sequence of bare measurements (i.e., with positive Kraus operators), we prove that the symmetry classification collapses to the ten-fold AZ$^\dagger$ classes, specifying the allowed topology of CPTP maps associated with feedback control. We demonstrate that a chiral Maxwell's demon with Gaussian measurement errors exhibits quantized winding numbers. Moreover, for general (non-bare) measurements, we explicitly construct a protocol that falls outside the ten-fold classification. These results broaden and clarify the principles in engineering topological aspects of quantum control robust against disorder and imperfections.

cond-mat.mes-hall↗

arXiv2026

Thermodynamic uncertainty relation under continuous measurement and feedback with quantum-classical-transfer entropy

We derive a thermodynamic uncertainty relation (TUR) under quantum continuous measurement and feedback control. By incorporating the quantum-classical-transfer entropy, which quantifies the information gained by continuous measurement, we show that the precision of currents is constrained by information-thermodynamic costs such as the entropy production and information gain. Our result shows that information gain has the potential to enhance the precision of currents beyond the bounds set by the conventional TUR. We illustrate the bound with a driven two-level system under continuous measurement and feedback, demonstrating that feedback achieves higher precision of currents while suppressing the entropy production.

cond-mat.stat-mech↗

arXiv2026

Testing the equivalence to thermal states via extractable work under LOCC

Understanding the thermal behavior of quantum many-body pure states is one of the most fundamental issues in quantum thermodynamics. It is widely known that typical pure states yield vanishing work, just as thermal states do, when one restricts to local operations that cannot access correlations among subsystems. However, it remains unclear whether this equivalence to thermal states persists under LOCC (local operations and classical communication), where classically accessible correlations can be exploited for work extraction. In this work, we establish criteria for determining whether many-body pure states remain equivalent to thermal states even under LOCC, and show that this thermal equivalence is governed by their multipartite quantum correlation structure. We show that states with asymptotically maximal multipartite entanglement, such as Haar-random states, cannot yield extensive work under LOCC, whereas some states with limited multipartite entanglement, such as constant-degree graph states, allow extensive work extraction despite being locally indistinguishable from thermal states. Thus, our work provides a refined operational notion of thermal equivalence beyond the traditional local regime, which is becoming increasingly important due to the recent expansion of experimentally accessible operations.

quant-ph↗

arXiv2025

Thermodynamic uncertainty relation for feedback cooling

Feedback cooling enables a system to achieve low temperatures through measurement-based control. Determining the thermodynamic cost required to achieve the ideal cooling efficiency within a finite time remains an important problem. In this work, we establish a thermodynamic uncertainty relation (TUR) for feedback cooling in classical underdamped Langevin systems, thereby deriving a trade-off between the cooling efficiency and the entropy reduction rate. The obtained TUR implies that simultaneous achievement of the ideal cooling efficiency and finite entropy reduction rate is asymptotically possible by letting the fluctuation of the reversible local mean velocity diverge. This is shown to be feasible by using a feedback control based on the Kalman filter. Our results clarify the thermodynamic costs of achieving the fundamental cooling limit of feedback control from the perspective of the TUR.

cond-mat.stat-mech↗

arXiv2025

Adversarial Thermodynamics

In thermodynamics, an agent's ability to extract work is fundamentally constrained by their environment. Traditional frameworks struggle to capture how strategic decision-making under uncertainty, particularly an agent's tolerance for risk, determines the trade-off between extractable work and probability of success in finite-scale experiments. Here, we develop a framework for nonequilibrium thermodynamics based on adversarial resource theories, in which work extraction is modeled as an adversarial game for an agent extracting work. Within this perspective, we consider a Szilard-type engine as a game isomorphic to Kelly gambling, an information-theoretic model of optimal betting under uncertainty -- but with a thermodynamic utility function. Extending the framework to finite-size regimes, we apply a risk-reward trade-off to find an interpretation of the Renyi divergences in terms of extractable work for a given failure probability. By incorporating risk sensitivity via utility functions, we show that the guaranteed amount of work a rational agent would accept instead of undertaking a risky protocol is given by a Renyi divergence. This provides a unified picture of thermodynamics and gambling, and highlights how generalized free energies emerge from an adversarial setup.

quant-ph↗

arXiv2025

Thermodynamic approach to quantum cooling limit of continuous Gaussian feedback

Feedback cooling plays a critical role in stabilizing quantum systems and achieving low temperatures, where a key question is to determine the fundamental thermodynamic limits on cooling performance. We establish a fundamental bound on quantum feedback cooling in Gaussian systems, by deriving a generalized second law of thermodynamics involving the kinetic temperatures of the system and a measure of quantum information flow obtained by continuous measurement. In contrast to previously known bounds, the obtained bound can be saturated by experimentally feasible situations using the quantum Kalman filter with a large feedback gain, where the cooling efficiency approaches its maximum. Our theoretical result is numerically demonstrated using parameters from an experiment of levitated nanoparticles. Our theory provides a general framework for understanding the thermodynamic constraints on quantum feedback cooling.

quant-ph↗

arXiv2025

Experimentally probing entropy reduction via iterative quantum information transfer

Thermodynamic principles governing energy and information are important tools for a deeper understanding and better control of quantum systems. In this work, we experimentally investigate the interplay of the thermodynamic costs and information flow in a quantum system undergoing iterative quantum measurement and feedback. Our study employs a state stabilization protocol involving repeated measurement and feedback on an electronic spin qubit associated with a Silicon-Vacancy center in diamond, which is strongly coupled to a diamond nanocavity. This setup allows us to verify the fundamental laws of nonequilibrium quantum thermodynamics, including the second law and the fluctuation theorem, both of which incorporate measures of quantum information flow induced by iterative measurement and feedback. We further assess the reducible entropy based on the feedback's causal structure and quantitatively demonstrate the thermodynamic advantages of non-Markovian feedback over Markovian feedback. For that purpose, we extend the theoretical framework of quantum thermodynamics to include the causal structure of the applied feedback protocol. Our work lays the foundation for investigating the entropic and energetic costs of real-time quantum control in various quantum systems.

quant-ph↗

arXiv2025

Transition from topological to chaos in the nonlinear Su-Schrieffer-Heeger model

Recent studies on topological materials are expanding into the nonlinear regime, while the central principle, namely the bulk-edge correspondence, is yet to be elucidated in the strongly nonlinear regime. Here, we reveal that nonlinear topological edge modes can exhibit the transition to spatial chaos by increasing nonlinearity, which can be a universal mechanism of the breakdown of the bulk-edge correspondence. Specifically, we unveil the underlying dynamical system describing the spatial distribution of zero modes and show the emergence of chaos. We also propose the correspondence between the absolute value of the topological invariant and the dimension of the stable manifold under sufficiently weak nonlinearity. Our results provide a general guiding principle to investigate the nonlinear bulk-edge correspondence that can potentially be extended to arbitrary dimensions.

cond-mat.mes-hall↗

arXiv2025

Thermodynamic speed limit for non-adiabatic work and its classical-quantum decomposition

Understanding the fundamental constraint on work far beyond the adiabatic regime is crucial to investigating fast and efficient energy extraction or consumption processes. In this study, we derive thermodynamic speed limits for non-adiabatic work and quantify the fundamental costs of non-adiabatic work extraction or consumption processes in open quantum systems, where the costs are quantified by geometric and thermodynamic quantities. We further decompose the non-adiabatic work into classical and quantum contributions and derive their thermodynamic speed limits, clarifying the classical and quantum nature of the fundamental costs. The obtained results are numerically demonstrated by driven two-level systems.

quant-ph↗

arXiv2025

Thermodynamic optimization of finite-time feedback protocols for Markov jump systems

In recent advances in finite-time thermodynamics, optimization of entropy production required for finite-time information processing is an important issue. In this work, we consider finite-time feedback processes in classical discrete systems described by Markov jump processes, and derive achievable bounds on entropy production for feedback processes controlled by Maxwell's demons. The key ingredients of our approach is optimal transport theory and an achievable Fano's inequality, by which we optimize the Wasserstein distance over final distributions under fixed consumed information. Our study reveals the minimum entropy production for consuming a certain amount of information, and moreover, the optimal feedback protocol to achieve it. These results are expected to lead to design principles for information processing in various stochastic systems with discrete states.

cond-mat.stat-mech↗

arXiv2025

Experimentally achieving minimal dissipation via thermodynamically optimal transport

Optimal transport theory, originally developed in the 18th century for civil engineering, has since become a powerful optimization framework across disciplines, from generative AI to cell biology. In physics, it has recently been shown to set fundamental bounds on thermodynamic dissipation in finite-time processes. This extends beyond the conventional second law, which guarantees zero dissipation only in the quasi-static limit and cannot characterize the inevitable dissipation in finite-time processes. Here, we experimentally realize thermodynamically optimal transport using optically trapped microparticles, achieving minimal dissipation within a finite time. As an application to information processing, we implement the optimal finite-time protocol for information erasure, confirming that the excess dissipation beyond the Landauer bound is exactly determined by the Wasserstein distance - a fundamental geometric quantity in optimal transport theory. Furthermore, our experiment achieves the bound governing the trade-off between speed, dissipation, and accuracy in information erasure. To enable precise control of microparticles, we develop scanning optical tweezers capable of generating arbitrary potential profiles. Our work establishes an experimental approach for optimizing stochastic thermodynamic processes. Since minimizing dissipation directly reduces energy consumption, these results provide guiding principles for designing high-speed, low-energy information processing.

cond-mat.stat-mech↗

arXiv2024

Finite-time thermodynamic bounds and tradeoff relations for information processing

In thermal environments, information processing requires thermodynamic costs determined by the second law of thermodynamics. Information processing within finite time is particularly important, since fast information processing has practical significance but is inevitably accompanied by additional dissipation. In this paper, we reveal the fundamental thermodynamic costs and the tradeoff relations between incompatible information processing such as measurement and feedback in the finite-time regime. To this end, we introduce a general framework based on the concept of the Pareto front for thermodynamic costs, revealing the existence of fundamental tradeoff relations between them. Focusing on discrete Markov jump processes, we consider the tradeoff relation between thermodynamic activities, which in turn determines the tradeoff relation between entropy productions. To identify the Pareto fronts, we introduce a new Wasserstein distance that captures the thermodynamic costs of subsystems, providing a geometrical perspective on their structure. Our framework enables us to find the optimal entropy production of subsystems and the optimal time evolution to realize it. In an illustrative example, we find that even in situations where naive optimization of total dissipation cannot realize the function of Maxwell's demon, reduction of the dissipation in the feedback system according to the tradeoff relation enables the realization of the demon. We also show that an optimal Maxwell's demon can be implemented by using double quantum dots. Furthermore, our framework is applicable to larger scale systems with multiple states, as demonstrated by a model of chemotaxis. Our results would serve as a designing principle of efficient thermodynamic machines performing information processing, from single electron devices to biochemical signal transduction.

cond-mat.stat-mech↗

arXiv2024

Optimal Finite-time Maxwell's Demons in Langevin Systems

We identify the optimal protocols to achieve the minimal entropy production in finite-time information exchange processes in Langevin systems, on the basis of optimal transport theory. Our general results hold even for non-Gaussian cases, while we derive a concise expression of the minimal entropy production for Gaussian processes. In particular, we apply our results to Maxwell's demons that perform measurement and feedback, and demonstrate Gaussian and non-Gaussian models of optimal demons operating in finite time. Our results provide a general strategy for controlling Langevin systems, including colloidal particles and biomolecules, in a thermodynamically optimal manner beyond the quasi-static limit.

cond-mat.stat-mech↗