Bulletin of the American Physical Society
2023 APS March Meeting
Volume 68, Number 3
Las Vegas, Nevada (March 510)
Virtual (March 2022); Time Zone: Pacific Time
Session S02: Stochastic Thermodynamics of Biological and Artificial Information ProcessingFocus

Hide Abstracts 
Sponsoring Units: GSNP Chair: Martin van Hecke, AMOLF Amsterdam & Leiden University; Marc Serra Room: Room 125 
Thursday, March 9, 2023 8:00AM  8:36AM 
S02.00001: Extending computational complexity theory to include thermodynamic resource costs Invited Speaker: David Wolpert The central concern of computational complexity theory is how the minimal "resource costs" needed to perform a given computation on a given type of computational machine depend on the size of the input. One of the most common choices for resource costs is the number of iterations the computation requires to complete. The canonical example is the classes P and NP, which are concerned with how the number of iterations it takes a Turing machine to perform a computation depends on the size of the input. However, in the real world, some of the most important resource costs of performing a computation are thermodynamic, e.g., the amount of free energy required to perform the computation and the amount of heat it produces. In this talk I will summarize recent results on how thermodynamic resource costs depend on the computational machine that is used and on the computation being performed. I will start by reviewing some results concerning the thermodynamic costs of performing a given computation in a (loopfree and branchfree) digital circuit, and on a Turing machine (TM  implemented as the associated `singleiteration' partial function). In particular, I will review results on how considering the minimal entropy production (EP) of computing a desired output on a TM, rather than the minimal size of an input string that causes the TM to produce that output (the output's Kolmogorov complexity), results in a correction term to Kolmogorov complexity. After this review I will summarize a set of new results concerning determinisitc finite automata (DFA). The first of these results are derived in an inclusive Hamiltonian framework, and relate the minimal EP required by a DFA to recognize a language to the state complexity of that DFA. The other results I will review are formulated in a continuoustimeMarkov chain framework. The first of these results concern the unavoidable mismatch cost contribution to EP that arises if the DFA's rate matrix is periodic, undergoing the same trajectory repeatedly for each iteration. The second set of results are stopping time fluctuation theorems giving total EP generated by a DFA by the time it reaches its halt state. I will end by describing the vast new set of research issues at the intersection of stochastic thermodynamics and computer science theory, issues that expand both fields. 
Thursday, March 9, 2023 8:36AM  8:48AM 
S02.00002: Fluctuations Beyond Detailed Balance in VoltageGated Channels Mikhael Semaan, James P Crutchfield To survive, living systems interact with and adapt to their fluctuating environments. In the process they must maintain homeostasis, which supports the thermodynamic conditions for internal biochemical and metabolic processing. We show how to separate this thermodynamic behavior into two pieces. The first is "homeostatic"associated with maintaining nonequilibrium steady states. The second is "adaptive"energy required as a system attempts to relax to its steady states. The theory culminates in a nonequilibrium steadystate trajectory class fluctuation theorem, valid for a broad class of mesoscopic complex systems. We apply this to directly compare the energetic behavior of the sodium and potassium ion channels responsible for generating and propagating signals in mammalian neurons. We explore how one of the channels violates detailed balance while the other does not, driving both of them with a biologicallyplausible action potential spike. This uncovers new quantitative structures that facilitate how nonequilibrium steadystate systems function while necessarily violating detailedbalanced thermodynamics. This mechanism leverages an "extra dimension" of Second Law violations accessible only to nondetailedbalanced systemsdimensions that must be thermodynamically accounted for. Practically, our results point towards necessary procedures for experimentally probing how complex biological systems absorb and dissipate energy. 
Thursday, March 9, 2023 8:48AM  9:00AM 
S02.00003: Stochastic Thermodynamics of Computing in Autonomous Mechanical Systems Zabreen Nissar, Finn T Bohte, Marc Serra Garcia Since the formulation of Landauer's limit for irreversible erasures, stochastic thermodynamics has provided novel insights in the thermodynamics of information processing, including the extension to thermodynamically reversible systems. However, current computers still exceed irreversible bounds by several orders of magnitude, and reversible approaches remain largely untested. Mechanical degrees of freedom (e.g. microresonators, colloidal particles) can be used to process information near thermodynamic limits, as has been shown in recent experiments using timedependent trapping potentials. However, nonautonomous experiments omit energy required to break detailed balance and prescribe a sequential computation. Here, we theoretically investigate nanomechanical systems performing autonomous, thermalnoise limited computations based on realistic nonlinear elastic interactions. A harmonic oscillator, called 'power clock' stores the energy that prescribes an arrow of time and causes the computation to evolve forward. 
Thursday, March 9, 2023 9:00AM  9:12AM 
S02.00004: Finitetime Landauer principle at strong coupling Alberto Rolandi, Martí Perarnau Llobet Landauer's principle gives a fundamental limit to the thermodynamic cost of erasing information. Its saturation requires a reversible isothermal process, and hence infinite time. We develop a finitetime version of Landauer's principle for a quantum dot strongly coupled to a fermionic bath. By solving the exact nonequilibrium dynamics, we optimize erasure processes (taking both the dot's energy and systembath coupling as control parameters) in the slow driving regime through a geometric approach to thermodynamics. We find analytic expressions for the thermodynamic metric and geodesic equations, which can be solved numerically. Their solution yields optimal finitetime processes that allow us to characterize a fundamental finitetime correction to Landauer's bound, fully taking into account nonmarkovian and strong coupling effects. 
Thursday, March 9, 2023 9:12AM  9:24AM 
S02.00005: A Maxwell demon that can work at macroscopic scales Massimiliano Esposito, Nahuel Freitas Maxwell's demons work by rectifying thermal fluctuations. They are not expected to function at macroscopic scales where fluctuations become negligible and dynamics become deterministic. We propose an electronic implementation of an autonomous Maxwell's demon that indeed stops working in the regular macroscopic limit as the dynamics becomes deterministic. However, we find that if the power supplied to the demon is scaled up appropriately, the deterministic limit is avoided and the demon continues to work. The price to pay is a decreasing thermodynamic efficiency. Our Letter suggests that novel strategies may be found in nonequilibrium settings to bring to the macroscale nontrivial effects so far only observed at microscopic scales. 
Thursday, March 9, 2023 9:24AM  9:36AM 
S02.00006: Symmetries and isometries of the thermodynamic metric Adam G Frim, Michael R DeWeese Geometric approaches to nonequilibrium thermodynamics have developed into a versatile toolkit for finding optimal protocols and fundamental physical limitations in slowlydriven microscopic systems. Typically, one makes a mapping from the space thermodynamic control variables (e.g. volume or temperature) to a smooth, Riemannian manifold in which geometric lengths correspond to physical energy dissipation; geometric methods can then be applied to assist in analyzing problems for a specific model thermal system when, for example, operated as a heat engine or ratchet. In this talk, we will instead focus more directly on the geometric side of this mapping and investigate possible physical interpretations of symmetries directly encoded in thermodynamic metrics. Notably, in certain simple cases, such symmetries have important physical implications, connecting (minimallydissipative) geodesic paths of the underlying geometric space to entropypreserving adiabats in a thermodynamic space. We will attempt to address this question more broadly, studying the deep connections between geometry and the thermodynamics of slowlydriven nonequilibrium systems. 
Thursday, March 9, 2023 9:36AM  9:48AM 
S02.00007: The tightest finitetime Landauer's principle: applications of speed limit Sangyun Lee, Jae Sung Lee, Hyunggyu Park, Hyukjoon Kwon For stochastic processes including Langevin and Markov jump processes, we can define entropy production in terms of trajectory probabilities. Based on this definition, physicists have derived diverse tighter versions of the second law such as thermodynamic uncertainty. Those inequalities bounds entropy production with a nonzero value. Thermodynamic uncertainty relation, a tradeoff relation between entropy production and the precision of an observable, is one example of entropy inequalities. Speed limit that bounds minimum time for transforming probability distribution is another example. Applications of these inequalities draw a lot of interest. Recently we derived the tightest finitetime Landauer's principle as an application of the speed limit. In this talk, I will present our recent research titled "Speed limit for a highly irreversible process and tight finitetime Landauer's bound" [PRL 129, 120603]. The finitetime Landauer's principle states fundamental entropy production when we erase one bit, and implies more heat is dissipated in a highly irreversible computing. 
Thursday, March 9, 2023 9:48AM  10:00AM 
S02.00008: Stochastic thermodynamics of coevolving systems  beyond multipartite processes Farita Tasnim, David H Wolpert A multipartite process (MPP) is a set of variables jointly evolving according to a continuoustime Markov chain for which only one variable can change its state at any time. Recent research in stochastic thermodynamics has produced a rich body of results concerning MPPs, highlighting the important role played by the network specifying which variable's value can directly affect the dynamics of which other variables. However, in many real systems multiple variables can change simultaneously, and so these systems cannot be modeled as MPPs. Indeed, some systems contain variables that cannot change state without a simultaneous change to some other variable's state. Examples of such "composite systems" range from moleculelevel models of chemical reaction networks to electronlevel models of circuits. Here, we analyze the stochastic thermodynamics of composite systems. Specifically, we derive novel decompositions of the entropy production and information flows for composite systems, based on the hypergraph specifying which variables are allowed to change state simultaneously. We also show how the hypergraph of a composite system can be used to strengthen thermodynamic bounds. In particular, we derive a tighter thermodynamic speed limit theorem, which suggests that given a fixed amount of dissipation, network structural constraints tend to cost extra time for evolution. 
Thursday, March 9, 2023 10:00AM  10:12AM 
S02.00009: Importance of nonequilibrium activity in information storage and processing in perceptronbased systems and systems with promiscuous interactions Agnish Kumar Behera, Madan Rao, Srikanth Sastry, Suriyanarayanan Vaikuntanathan Motivated by advances in the field of active matter where nonequilibrium forcing has been shown to activate new assembly pathways, here we study how nonequilibrium driving in prototypical memory formation models can affect their information processing capabilities. Building on previous work done with the simple spherical Hopfield model where activity was shown to improve the memory retrieval properties, we will discuss a model with promiscuous interactions among its constituents where dynamics with nonequilibrium noise sources can lead to a higher memory capacity than a zerotemperature equilibrium version that is not subject to any noise. We also demonstrate better signal retrieval properties using perceptron layers with nonequilibrium dynamics when compared to the equilibrium counterpart. Our results demonstrate the generality of the enhancement of memory capacity arising from nonequilibrium, active dynamics. 
Thursday, March 9, 2023 10:12AM  10:24AM 
S02.00010: Resource Theory of Thermodynamic Computation Jinghao Lyu, Alexander B Boyd, James P Crutchfield Landauer’s Principle is a wellknown demonstration of the thermodynamic cost of computation. It establishes the minimum work needed to reset one bit at a given temperature. Recently, thermodynamics was reformulated as a resource theory, where the work reservoir is integrated into the system of interest, realizing dynamical behavior rather than operating as an external idealized energy source. We show how to implement any computation with a thermal operation using a twolevel work reservoir. We compute the minimum work cost, finding that, unlike Landauer's bound, it is robust to unexpected changes in the input distribution. We study the entropy production of those information processing tasks and show that realizing those generalized Landauer’s bounds requires infinitesize thermal baths. 
Thursday, March 9, 2023 10:24AM  10:36AM 
S02.00011: Szilard engines and informationbased work extraction for active systems Holger F Stark, Paolo Malgaretti The out of equilibrium nature of active systems can be exploited for the design of informationbased engines. We design two types of an active Szilard engine that use a Maxwell demon to extract work from an active bath composed of noninteracting Active Brownian Particles (ABPs). The two engines exploit either the quasistatic active pressure of ABPs or the long correlation time of their velocities. For both engines the active bath allows to overcome the Landauer principle and to extract larger work compared to conventional Szilard engines operating in quasi thermal equilibrium. For both of our engines,we identify the optimal regimes at which the work extracted and the efficiency are maximized. Finally, we discuss them in the context of synthetic and biological active systems. 
Thursday, March 9, 2023 10:36AM  10:48AM 
S02.00012: Phase Transition in a noisy information engine is avoided by using Bayesian inference John Bechhoefer, Tushar K Saha, Joseph Neil E Lucero, Jannik Ehrich, David A Sivak Information engines use observations of thermal fluctuations as "fuel" to extract work or produce directed motion. Inspired by the classic thought experiment of Maxwell, they have clarified our understanding of the interplay between information and the second law of thermodynamics. We investigated the role of uncertainties in the measurements used to power the information engine. Surprisingly, there is a critical level of measurement noise, above which a pure information engine — one where no external work is needed — is no longer possible. The reason can be traced back to a bias in the naive use of measurements by the information engine. The bias has two sources: time delays and a subtle error arising whenever decisions are based on rare events and depend on noisy measurements. We then used a more sophisticated Bayesian way of incorporating the entire history of past measurements (predictive Kalman filter). By removing both biases, the Bayesian approach allows the information engine to function at all levels of measurement noise. In the regime where thermal fluctuations are comparable to the measurement noise, the performance gain over a naive approach is significant, pointing to practical information engines that operate with minimal informationprocessing costs. 
Follow Us 
Engage
Become an APS Member 
My APS
Renew Membership 
Information for 
About APSThe American Physical Society (APS) is a nonprofit membership organization working to advance the knowledge of physics. 
© 2024 American Physical Society
 All rights reserved  Terms of Use
 Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 207403844
(301) 2093200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 5914000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 200452001
(202) 6628700