Saturday 14 November 2026 · 20:00 Beijing time
Fluctuation-Response Relations and Causal Memory in non-Markovian Systems
Jiming Zheng 郑纪鸣
University of North Carolina at Chapel Hill
- Where
- Tencent Meeting 876-345-373
- Host
- Shiling Liang (梁师翎), MPI-PKS
Seminar
SPYC, short for Statistical Physics Youth Communications (统计物理青年通讯), has run an online seminar for early-career researchers in statistical and biological physics since 2023.
Next talk
About
Statistical Physics Youth Communications (SPYC) is a Chinese-language community of early-career researchers in statistical and biological physics. Shiling Liang founded it in 2022, and in 2026 it had more than 400 members. Since 2023 it has run an online seminar series in which researchers present their recent work on topics ranging from active matter and biophysics to quantum systems and machine learning. Most talks are given in Chinese, but talks in English are welcome too.
The talks are online and open to everyone, and most start at 20:00 Beijing time. Each talk is listed with its meeting number, usually on Tencent Meeting, and with a link to its recording when one is available.
Organizer Shiling Liang · Contact sl-liang@outlook.com
统计物理青年通讯（SPYC）是面向统计物理与生物物理方向青年学者的中文交流社区，由梁师翎于2022年发起，2026年成员已达400余人。社区自2023年起举办线上学术报告，邀请研究者介绍最新工作，主题涵盖活性物质、生物物理、量子系统与机器学习等方向。报告以中文为主，也欢迎英文报告。
报告均在线上举行，欢迎所有感兴趣的人参加。多数报告在北京时间晚上8点开始，会议号（通常为腾讯会议）见每场报告的信息。如有录像，也会附上回放链接。
2026
Fluctuation-Response Relations and Causal Memory in non-Markovian Systems
Jiming Zheng 郑纪鸣
University of North Carolina at Chapel Hill
- Where
- Tencent Meeting 876-345-373
- Host
- Shiling Liang (梁师翎), MPI-PKS
How the past shapes the response of a fluctuating system is a basic question in nonequilibrium physics. Fluctuation-response relations connect spontaneous fluctuations to the effects of external perturbations. In systems with memory, however, the same observed state can respond differently depending on how it was reached. The challenge is to identify which features of the past matter and how their influence appears in fluctuations and responses.
In this talk, I will present a general fluctuation-response theory for non-Markovian jump processes, formulated directly in terms of observed histories. The key quantity is an event’s effect on the expected future value of an observable. It determines both the event’s contribution to fluctuations and the response to perturbing its rate. This connection yields exact fluctuation-response relations and kinetic bounds in both the time and frequency domains. It also reveals a fluctuation contribution that is lost when responses are averaged over histories. We call this contribution causal memory. It measures how the past changes the consequences of present events under perturbations. I will illustrate the framework with a solvable model and experimental ion-channel recordings. In these recordings, dwell-time correlations and causal memory rank cell-membrane and mitochondrial channels in opposite orders. These results give a physical criterion for deciding which historical distinctions must be retained to preserve dynamically causal information.
References
- Jiming Zheng and Zhiyue Lu, “The Memory Hidden in Response Fluctuations: Trajectory-Level Fluctuation-Response Theory and Inequalities for Non-Markovian Jump Dynamics.” arXiv:2608.20328 (2026)
Speaker
Jiming Zheng (郑纪鸣) is a PhD student in Physical Chemistry at the University of North Carolina at Chapel Hill, advised by Prof. Zhiyue Lu. His research interests include nonequilibrium statistical physics, stochastic thermodynamics, and fluctuation-response theory for systems far from equilibrium and with memory. As first and corresponding author, he has published a series of works on nonequilibrium response relations in Physical Review E and The Journal of Chemical Physics. He has also co-authored papers in Physical Review Research and The Journal of Physical Chemistry Letters.
Criticality and Universality of the Generalized Kuramoto Model
Zhongpu Qiu 邱仲普
Beijing Normal University & Shanghai AI Lab
- Host
- Shiling Liang (梁师翎), MPI-PKS
- Recording
- Watch the recording
How symmetry and spatial dimensionality jointly determine the universality class of phase transitions is a key question in statistical physics. For equilibrium O(n) models this picture is textbook knowledge, but for nonequilibrium synchronization transitions of coupled oscillators with quenched disorder, a unified understanding has remained elusive.
In this talk, I will present our recent study of synchronization transitions in the generalized Kuramoto model, where oscillators are n-dimensional unit vectors with O(n) symmetry, on both complete graphs and d-dimensional lattices. On the complete graph, analytical expansions of the self-consistency equations with finite-size corrections reveal universal critical exponents β = 1/2 and ν̄ = 5/2 shared by all even n, indicating an unconventional upper critical dimension du = 5 — in contrast to du = 4 for equilibrium O(n) models. For lattices, we generalize the very definition of synchronization from scalar phases to vector oscillators, introducing a synchronization functional network method together with percolation-type and generalized Edwards–Anderson order parameters, which allow us to locate the transition and extract critical exponents. Combining spin-wave theory with large-scale simulations, we uncover a family of universality classes that are insensitive to the order-parameter dimension n but mainly determined by the spatial dimension d. These results show that the equilibrium paradigm “symmetry and dimensionality determine universality” extends to nonequilibrium coupled-oscillator systems with quenched noise, offering a unified statistical-physics perspective on the ubiquitous phenomenon of synchronization.
References
- Zhongpu Qiu, Tianyi Wu, Sheng Fang, Jun Meng, and Jingfang Fan, “Criticality and Universality of the Generalized Kuramoto Model.” Phys. Rev. Lett. 137, 147201 (2026)
Speaker
Zhongpu Qiu is a PhD candidate in Systems Theory at the School of Systems Science, Beijing Normal University. His research interests include fundamental problems in statistical physics, synchronization dynamics, active matter, and AI for Science. His first-author work on the criticality and universality of the generalized Kuramoto model was recently published in Physical Review Letters. He has also co-authored papers in Nature Machine Intelligence and npj Climate and Atmospheric Science.
Fluctuation–Response Theory of Non-Equilibrium Complex Fluids · in English
Ryota Takaki
Max Planck Institute for the Physics of Complex Systems (MPI-PKS)
- Host
- Shiheng Zhao (赵世恒), MPI-PKS
- Recording
- Watch the recording
A central challenge in nonequilibrium statistical physics is to derive macroscopic constitutive laws of fluids from microscopic fluctuations far from equilibrium. Near equilibrium, Onsager regression and Green–Kubo relations connect fluctuations to response, but a general framework for nonequilibrium steady states with memory is lacking.
Here, we develop a fluctuation–response formulation of generalized hydrodynamics at finite wavevectors and frequencies. Using stationary correlation identities, we derive transport coefficients directly from steady-state correlations, extending Onsager regression and Green–Kubo relations beyond equilibrium while retaining broken time-reversal symmetry. Applied to chemically driven active fluids, the theory reveals active viscoelastic memory, in which chemo-mechanical feedback renormalizes the mechanical response and can generate a negative storage modulus at finite frequencies. Our framework provides a statistical-mechanical basis for active and nonequilibrium complex fluids with spatiotemporal memory.
References
- Ryota Takaki and Frank Jülicher, “Fluctuation–Response Theory of Non-Equilibrium Complex Fluids.” arXiv:2510.15559 (2026)
Speaker
Ryota Takaki is a theoretical biological physicist at the Max Planck Institute for the Physics of Complex Systems in Dresden. His research focuses on nonequilibrium statistical physics of soft biological matter, including biomolecular condensates and molecular motors. He received his PhD in physics from the University of Texas at Austin and currently works on developing statistical-mechanical approaches to understand the dynamics, rheology, and organization of living and active systems.
Phase separation induced by active polymerization makes protocells robust against environmental changes · in English
Xi Chen 陈曦
Leibniz Institute of Polymer Research Dresden
- Host
- Linge Li (李林格), MPI-PKS
- Recording
- Watch the recording
The origin of life remains a scientific mystery, particularly the emergence of protocells. One hypothesis proposes that protocells arose as droplets formed via liquid–liquid phase separation of polymers. The work on this hypothesis leaves open how protocells survived in fluctuating or cyclic environments. We consider a model system incorporating both spontaneous polymerization and droplet-facilitated fuel-driven polymerization. We show that the resulting droplets display a stationary hysteresis with respect to available fuel. Droplets can remain stable even after the fuel-driven polymerization reactions significantly diminish, suggesting a potential mechanism for protocell formation and resilience to environmental fluctuations. This robustness would have enabled protocells to endure early environmental challenges, such as energy shortages in a famine.
References
- Xi Chen, Jens-Uwe Sommer, Tyler S. Harmon, Proc. Natl. Acad. Sci. U.S.A. 123 (16) e2524346123 (2026)
Speaker
Xi Chen has recently finished his PhD at Leibniz Institute of Polymer Research. He is interested in the dynamics of biomolecular condensates, particularly in how nonequilibrium processes and regulatory mechanisms shape their behavior. He is also interested in the spatial organization of active chemical reactions, such as chemically active droplets.
Hamiltonian description of nonreciprocal interactions
Yubo Shi 施宇博
National University of Singapore
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
In many systems, including sedimenting particles and bird flocks, interactions do not derive from a potential and are generally non-reciprocal, meaning that they do not obey the action–reaction principle. As a result, one cannot define a conventional energy function or use analytical and numerical tools that rely on it. Here we address this limitation by constructing a Hamiltonian with auxiliary degrees of freedom that, under a constraint, generates the original non-reciprocal dynamics. We show that Monte Carlo simulations based on the constrained Hamiltonian reproduce both stationary and non-stationary states of the original Langevin dynamics, as we illustrate for dissipative XY spins with vision-cone interactions. The symplectic structure inherent to the construction also lets us apply established ideas from Hamiltonian engineering, which we demonstrate by varying the amplitude of a periodic (Floquet) drive to tune the spin interactions between square- and chain-lattice geometries. Overall, our construction paves the way towards extending statistical mechanics and Hamiltonian dynamics to non-reciprocal systems.
References
- arXiv:2505.05246 (to appear in Nature Physics, 2026)
Speaker
Yubo Shi is a postdoctoral researcher from National University of Singapore. His current interests include nonequilibrium physics and reinforcement learning for quantum systems. He is also curious about human intelligence, exemplified by questions like how to be an SG instead of an SF in a basketball game.
Learning in Artificial and Biological Systems — From In-Context Learning to Animal Behavioral Learning
Wenping Cui 崔文平
Princeton University
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
How do artificial and biological systems learn structures from their environments? I present two projects that explore shared principles of learning across these domains.
In artificial systems, we study transformers, which exhibit in-context learning: the ability to infer task-specific structure from limited input data without updating model parameters. We provide a mechanistic and dynamical characterization of in-context generalization in a transformer trained on discrete stationary Markov chains. Training gives rise to two distinct algorithmic phases, a unigram phase and a bigram phase, with the bigram solution implemented by a statistical induction head. We derive an effective theory for the learning dynamics of this induction head, explain its abrupt emergence, and show that the transition time is controlled by statistical biases in the data.
In biological systems, we investigate how rats learn continuous pose templates through reinforcement learning. Using an analysis pipeline for high-dimensional trajectory data, we show that behavior can be decomposed into reusable motion motifs, most of which appear to be innate. Our results suggest that biological motor learning follows a select-and-refine strategy, in which innate motifs are recruited and refined through reinforcement rather than generated from scratch.
Speaker
Wenping Cui is a postdoctoral researcher at Princeton University. He received his B.S. from the University of Science and Technology of China, M.S. from the University of Bonn, and Ph.D. from Boston College. From 2021 to 2024, he was a postdoctoral researcher at the Kavli Institute for Theoretical Physics, UC Santa Barbara. His research uses statistical physics to study fundamental principles of complex biological and computational systems, including ecology and evolution, cellular sensing, animal behavior, and attention learning in artificial neural networks.
Coercivity Landscape Characterizes Dynamic Hysteresis
Chen Miao 陈苗
Beijing Normal University
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
Hysteresis, with rich dynamical behaviors—especially in interacting systems—has drawn broad research interest. Yet dynamic scalings across time scales lack a unified description, and their transitions remain unclear. Here, we introduce the coercivity landscape that unifies hysteresis dynamics across time scales by characterizing how coercivity evolves with driving rate. The landscape reveals distinct scaling laws across different dynamical regimes and a universal coercivity plateau originating from the noncommuting thermodynamic and quasistatic limits. Furthermore, by studying magnetic hysteresis in the Curie–Weiss model, we demonstrate that universality governs coercivity scaling over most driving regimes, with model-specific behavior emerging only under fast driving. This framework establishes a unifying connection between equilibrium criticality, non-equilibrium dynamics, and experimental observations.
References
- M. Chen*, X. H. Zhao*, and Y. H. Ma†, Phys. Rev. Lett. 136, 117102 (2026), Editor’s Suggestion
- M. Chen, X. H. Zhao†, and Y. H. Ma†, Phys. Rev. E 113, 034124 (2026)
Speaker
Chen Miao is currently a master’s student in theoretical physics at Beijing Normal University under the supervision of Prof. Yu-Han Ma. Her research interests include phase transitions, finite-system thermodynamics, and non-equilibrium statistical physics.
Dual-symmetry-guided assembly of complex lattices
Huang Fang 方煌
Fudan University
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
Complex lattices with mixed rotational symmetries are important in many functional materials, but their assembly often requires anisotropic building blocks, directional interactions, or fully prescribed templates, which can hinder defect relaxation and kinetically accessible ordering. Here we introduce a dual-symmetry-guided (DSG) strategy for assembling complex two-dimensional lattices in colloidal monolayers. By decomposing a target tiling into two mutually dual sublattices and sparsely pinning only one of them with optical traps, we show that the complementary sublattice can self-organize through purely isotropic repulsive interactions to reconstruct the full lattice. Using this approach, we experimentally realize a broad class of Archimedean lattices and extend the method to two-dimensional quasicrystalline structures, in agreement with simulations. We further show that DSG preserves connected free volume for mobile particles, promotes defect annealing, and reduces kinetic barriers compared with full pinning. These results establish partial pinning based on dual symmetry as a simple route to complex-symmetry assembly with isotropic interactions.
References
- Fang, Huang, Xiaotian Li, Wensi Sun, Chengxin Wang, Nuo Chen, Yining Gan, Jiping Huang, Yuqiang Ma, Hajime Tanaka, and Peng Tan. “Dual-symmetry-guided assembly of complex lattices.” Nature (2026): 1-8.
Speaker
Huang Fang is currently a postdoctoral researcher at Fudan University under the supervision of Prof. Peng Tan. He received his Ph.D. in Physics from Brandeis University. His research interests include DNA nanotechnology, soft-matter self-assembly, and equilibrium and nonequilibrium statistical physics.
Edge Currents Shape Condensates in Chiral Active Matter
Boyi Wang 王博艺
Max Planck Institute for the Physics of Complex Systems (MPI-PKS)
- Host
- Yu Tian, CSBD
- Recording
- Watch the recording
Inspired by biological systems that break parity and time-reversal symmetries, we study theoretical models for chiral active matter. In this talk, I will introduce a minimal two-dimensional lattice gas where chiral activity enters merely through stochastic, biased local rotations. At low temperatures, this system coarsens into faceted condensates featuring persistent, unidirectional edge currents. To generalise these findings, we propose a continuum theory by adding an active chiral edge current term to Model B. This continuum theory reveals the essential role of active chiral transport in interfacial dynamics: edge currents with n-fold symmetry naturally produce condensates shaped like regular n-sided polygons. To determine the geometry of condensates analytically, we construct an effective interface potential governing edge currents, providing an intuitive mathematical framework applicable to both the lattice model and the continuum description.
Speaker
Boyi Wang is a postdoctoral researcher at the Max Planck Institute for the Physics of Complex Systems (MPI-PKS) in Dresden, Germany. She received a Ph.D. in Condensed Matter Physics in January 2025 from the Institute of Physics, Chinese Academy of Sciences (IOP-CAS), after a B.Sc. in Biology in June 2018 from University of Chinese Academy of Sciences (UCAS).
Physics-uninformed machine learning
Yuanzhao Zhang 章元肇
Santa Fe Institute
- Host
- Yu Meng, Fudan University
- Recording
- Watch the recording
How much can we learn about a complex system when governing equations are unknown and data are scarce? This talk explores how modern machine-learning models can extrapolate beyond limited training data without physics priors. First, I will show that simple recurrent neural networks with no built-in physics can unexpectedly reconstruct basins of attraction in multistable systems, even for basins not seen during training. This unexpected generalization capability raises questions about what hidden inductive biases are implicitly regularizing the model. Second, I will examine time-series foundation models and their ability to forecast entirely new dynamical systems from a short context trajectory. These models often achieve strong zero-shot performance in forecasting chaotic systems by exploiting a strategy we term context parroting. Analyzing the parroting strategy provides insights into the capabilities of current foundation models and explains observed neural scaling laws. Exploring strategies beyond parroting may further reveal how both artificial and natural intelligence extract information from limited data.
Speaker
Yuanzhao is an Omidyar Fellow at the Santa Fe Institute, where he studies collective dynamics on networks and data-driven modeling of nonlinear systems. Previously, he was a Schmidt Science Fellow at Cornell’s Center for Applied Mathematics. He received his Ph.D. in Physics from Northwestern.
Dynamical Theories and Methods for Complex Biological Systems
Bo-Wei Qin 秦伯韡
Fudan University
- Host
- Yu Meng, Fudan University
- Recording
- Watch the recording
Complex biological systems usually exhibit novel dynamical phenomena and functions owing to interactions among individuals. These emergent behaviors are significantly influenced by key properties such as interaction network topology and interaction types. Utilizing dynamical models and methods is essential for elucidating the fundamental principles governing complex biological systems. In this talk, we will introduce recent theoretical and methodological advancements in this direction, with a particular focus on investigating crucial physical characteristics including, for instance, order parameters, pattern complexity [1], frequency and amplitude of oscillations [2,3] when the state of a system is close to its critical state. Furthermore, we will discuss the insights and implications these studies offer for understanding those systems from microscale to macroscale.
References
- Xuzhe Qian, Bo-Wei Qin*, Huaiping Zhu, Wei Lin*. Physical Review E(Letter) 113, L012301 (2026).
- Zhaoyue Zhong, Wei Lin*, Bo-Wei Qin*. Physical Review Letters 131, 138401 (2023).
- Bo-Wei Qin*, Lei Zhao, Wei Lin*. Nature Communications 12, 5894 (2021).
Speaker
Bo-Wei Qin (秦伯韡) is currently an associate professor at Research Institute of Intelligent Complex Systems in Fudan University and serves as the secretary-general of the Shanghai Society for Nonlinear Science. In 2017, he received the Ph.D. from Department of Mathematics in City University of Hong Kong. His research focuses on applied mathematics and complexity science, particularly applying bifurcation theory to analyze gene regulatory networks, neuronal networks, ecological systems and human behaviors. For more information please visit: https://iics.fudan.edu.cn/c8/92/c43565a641170/page.htm
Preparing steady states in dissipative many-body systems
Yu-Min Hu 胡渝民
Max Planck Institute for the Physics of Complex Systems (MPI-PKS)
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
Dissipative couplings in open quantum systems not only enable the preparation and control of novel quantum states but also lead to nonequilibrium phases without counterparts in closed systems. In this talk, I will present our two recent developments in this direction. In the first part, I will introduce the notion of asymptotic exceptional steady states in dissipative dynamics [1]. Although exceptional points are forbidden for exactly degenerate steady states of Liouvillians, a large class of open many-body systems can asymptotically approach an exceptional steady state as the system size increases. Such an asymptotic structure has an intriguing connection to the computational complexity of dissipative state preparation. In the second part, I will discuss a dissipative quantum link model that naturally exhibits the many-body non-Hermitian skin effect—an anomalous boundary localization phenomenon [2]. Despite the complexity arising from strong interactions, local gauge symmetry enables the exact construction of steady states. These results provide an analytical method for constructing exact many-body steady states and demonstrate how dissipation can be used in quantum many-body experiments to engineer unexpected steady states.
References
- Y.-M. Hu and J. C. Budich, Phys. Rev. Lett. 135, 250402 (2025).
- Y.-M. Hu, Z. Wang, B. Lian, and Z. Wang. Phys. Rev. Lett. 135, 260401 (2025).
Speaker
Yu-Min Hu (胡渝民) is a postdoctoral researcher in Prof. Jan Carl Budich’s group at the Max Planck Institute for the Physics of Complex Systems. His research interests span nonequilibrium many-body physics and open quantum systems. He received his Ph.D. in 2024 from the Institute for Advanced Study, Tsinghua University, under the supervision of Prof. Zhong Wang. From February to June 2023, he visited Princeton University, where he worked with Prof. Biao Lian.
2025
Statistical mechanics of chunking – from natural language to human memory
Weishun Zhong 钟纬舜
Institute for Advanced Study (IAS)
- Host
- Wenping Cui, Princeton University
- Recording
- Watch the recording
Natural language is organized hierarchically in chunks, from paragraphs and sentences down to phrases and words. These chunks serve as the fundamental building blocks of language at multiple scales, forming a backbone that supports the expression and communication of meaning at different levels of abstraction. Using Large Language Models (LLMs), we automatically extract these tree-like chunking structures across vast corpora of text. Remarkably, despite substantial variation in content and style across human writings, the ensemble statistics of these chunking trees can be captured by an exactly solvable model. The entropy derived from our model closely matches cross-entropy measurements by LLMs on real-world texts. Furthermore, grounded in the hierarchical organization of language, our model predicts human memory performance in natural language tasks. It quantitatively explains data from large-scale narrative recall experiments and reveals that cognitive phenomena like summarization and abstraction naturally emerge from the underlying chunking structure in language. Our results bridge the gap between computational models of language and memory, offering a unified framework for understanding chunking in both natural language processing and human cognition.
References
- Semantic Chunking and the Entropy of Natural Language, Weishun Zhong, Tankut Can, Mikhail Katkov, Misha Tsodyks, in preparation.
- Random Tree Model of Meaningful Memory, Weishun Zhong, Tankut Can, Atonis Georgiou, Ilya Shnayderman, Mikhail Katkov, Misha Tsodyks, Physical Review Letters (2025).
Speaker
Weishun Zhong (钟纬舜) is currently an Eric and Wendy Schmidt Postdoctoral Member at the Simons Center for Systems Biology, School of Natural Sciences, Institute for Advanced Study (IAS) in Princeton. His research area lies in the intersection of statistical physics, neuroscience, and AI. He earned his Ph.D. in Physics from MIT in June 2023 under the supervision of Haim Sompolinsky (Harvard) and Mehran Kardar (MIT).
Inferring collective dynamics in biological systems: from social mice to microbial communities
Xiaowen Chen 陈笑闻
CNRS/LPENS
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
From social animals to ecological systems, collective behavior is ubiquitous in living systems. How are these behaviors encoded in interactions, and how do they drive biological functions? Recent insights from statistical physics applied to biological data have offered exciting new perspectives. However, previous research has mostly focused on the statics, i.e. the steady-state distributions of the collective behavior, without taking into consideration of time. In this talk, I will present two recent progresses tapping into the temporal domain. First, I will present a study of collective behavior in social mice from their co-localization patterns [1]. To capture both static and dynamic features of the data, we developed a novel inference method termed the generalized Glauber dynamics (GGD) that can tune the dynamics while keeping the steady state distribution fixed. The inferred interactions characterize sociability for different mice strains. In the second example, we studied the resource competition network in microbial communities [2]. We show that while simple correlations are often misleading, spectral methods that account for time-delayed effects are better at inferring the structure of resource competition in communities. Finally, if there is time, I will talk about a third example in the fine-tuning problem of neuronal networks [3].
References
- X. Chen, M. Winiarski, A. Puścian, E. Knapska, A. M. Walczak, and T. Mora, “Generalized Glauber dynamics for inference in biology,” Physical Review X, 13, 041053 (2023).
- X. Chen, K. Crocker, S. Kuehn, A. M. Walczak, and T. Mora, “Inferring resource competition in microbial communities from time series,” PRX Life, 3, 023019 (2025).
- X. Chen and W. Bialek, “Searching for long time scales without fine tuning,” Physical Review E, 110, 034407 (2024).
Speaker
Chen Xiaowen is a theoretical physicist and a researcher at CNRS / Laboratory of Physics at ENS. Prior to this, she received a PhD in Physics at Princeton University under the advising of William Bialek in 2020, and conducted postdoctoral research at LPENS under the supervision of Aleksandra Walczak and Thierry Mora. She is interested in understanding the collective behavior in living systems using a combination of analytical, numerical, and data analysis methods.
Matrix-weighted networks for modeling multidimensional dynamics
Yu Tian 田雨
CSBD & MPI-PKS/CBG
- Host
- Boyi Wang, MPI-PKS
- Recording
- Watch the recording
Networks are powerful tools for modeling interactions in complex systems. While traditional networks use scalar edge weights, many real-world systems involve multidimensional interactions. For example, in social networks, individuals often have multiple interconnected opinions that can affect different opinions of other individuals, which can be better characterized by matrices. We propose a general framework for modeling such multidimensional interacting dynamics: matrix-weighted networks (MWNs). We present the mathematical foundations of MWNs and examine consensus dynamics and random walks within this context. Our results reveal that the coherence of MWNs gives rise to non-trivial steady states that generalize the notions of communities and structural balance in traditional networks.
References
- Tian, Y., Kojaku, S., Sayama, H., & Lambiotte, R. (2025). Matrix-Weighted Networks for Modeling Multidimensional Dynamics: Theoretical Foundations and Applications to Network Coherence. Physical Review Letters, 134(23), 237401.
Speaker
Yu Tian (田雨) is currently an ELBE postdoctoral fellow at the Center for Systems Biology Dresden (CSBD), Max Planck Institute for Physics of Complex Systems (MPI PKS), and Max Planck Institute for Cell Biology and Genetics (MPI CBG). Before this, she was a postdoctoral fellow at NORDITA, Stockholm. She earned her PhD in Mathematics from the University of Oxford in October 2022. Her research focuses on the interaction between structure and dynamics in network systems, specifically when the relationship between entities is necessarily characterized by a negative or complex value, or a matrix. She also works on the downstream applications in biological and quantum systems, and machine learning.
Not-so-glass-like Caging and Fluctuations of an Active Matter Model · in English
Mingyuan Zheng 郑铭媛
HKUST (Guangzhou)
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
Simple active models of matter recapitulate complex biological phenomena. The out-of-equilibrium nature of these models, however, often makes them beyond the reach of first-principle descriptions. This limitation is particularly perplexing when attempting to distinguish different glass-forming mechanisms. We here consider a minimal active system in various spatial dimensions to identify the processes underlying their sluggish dynamics [1]. Simulations of an active Brownian particle random Lorentz gas (ABP-RLG) reveal that activity leads to the exploration of lower-dimensional caging features (see Figure 1). This mechanism significantly enhances cage escape processes, shifting the mean-field-like dynamical transition to higher densities and saturating at the percolation threshold, thereby sidestepping instantonic (or activated) processes typical of standard glass formers. Near the dynamical arrest, activity also introduces a distinct fluctuation signature, as evinced by the growth of a short-time peak in the non-Gaussian parameter. This finite-dimensional echo of mean-field criticality is readily accessible to both simulations and experiments, and provides a robust basis for contrasting standard and active glass formers. Our findings bridge finite-dimensional simulations and mean-field theory, thus offering insight into the critical dynamics of active matter and paving the way toward a first-principle description of glassy systems.
References
- Zheng, Mingyuan, Dmytro Khomenko, and Patrick Charbonneau. “Not-so-glass-like Caging and Fluctuations of an Active Matter Model.” Physical Review Letters 134.22 (2025): 228301.
Speaker
Mingyuan Zheng (郑铭媛) earned her bachelor’s degree in Chemistry from Fudan University and a doctoral degree in Theoretical and Computational Chemistry from Duke University. She worked with Prof. Patrick Charbonneau on statistical physics during her Ph.D period, focusing on self-assembly systems including equilibrium microphase formers and non-equilibrium active matter models. She is currently a postdoctoral researcher in Jiaxing Yuan’s group at HKUST(GZ), exploring machine learning-based simulation methods to address challenges in soft and liquid matter systems. Her research interests include microphase formation, cluster-based Monte Carlo methods, and active matter models related to glass transition and other soft matter phenomena.
Thermomajorization Mpemba effect · in English
Tan Van Vu
Kyoto University
- Host
- Hongchao Li, University of Tokyo
- Recording
- Watch the recording
The Mpemba effect is a counterintuitive physical phenomenon where a hot system cools faster than a warm one. In recent years, theoretical analyses of the Mpemba effect have been developed for microscopic systems and experimentally verified. However, the conventional theory relies on a specific choice of distance measure to quantify relaxation speed, leading to several theoretical ambiguities. In this Letter, we derive a rigorous quantification of the Mpemba effect based on thermomajorization theory, referred to as the thermomajorization Mpemba effect. This approach resolves all existing ambiguities and provides a unification of the conventional Mpemba effect across all monotone measures. Furthermore, we demonstrate the generality of the thermomajorization Mpemba effect for Markovian dynamics, rigorously proving that it can occur in any temperature regime with fixed energy levels.
References
- Van Vu, T. and Hayakawa, H., 2025. Thermomajorization Mpemba Effect. Physical Review Letters, 134(10), p.107101.
Universal phenomena, irreversible processes, and thermodynamics in deep representation learning
深度表征学习中的普适现象，不可逆过程，和热力学
Liu Ziyin 刘子寅
MIT
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
With the rapid discovery of emergent phenomena in deep learning and large language models, explaining and understanding their cause has become an urgent need. In this seminar, I present a rigorous entropic-force theory for understanding the learning dynamics of neural networks trained with stochastic gradient descent (SGD) and its variants. Building on the theory of parameter symmetries and an entropic loss landscape, I show that representation learning is crucially governed by emergent entropic forces arising from stochasticity and discrete-time updates. These forces systematically break continuous parameter symmetries and preserve discrete ones, leading to a series of gradient balance phenomena that resemble the equipartition property of thermal systems. These phenomena, in turn, (a) explain the universal alignment of neural representations between AI models and lead to a proof of the Platonic Representation Hypothesis, and (b) reconcile the seemingly contradictory observations of sharpness- and flatness-seeking behavior of deep learning optimization. I present theory and experiments to demonstrate that a combination of entropic forces and symmetry breaking is key to understanding emergent phenomena in deep learning.
Speaker
I (Liu Ziyin) am a Postdoctoral Fellow at MIT and NTT Research. At MIT, I work with Prof. Isaac Chuang. I also collaborate with Prof. Tomaso Poggio in the BCS department. My research focus is on the theoretical foundation of deep learning and, recently, computational neuroscience. Prior to coming to MIT, I received my PhD in physics at the University of Tokyo under the supervision of Prof. Masahito Ueda. I received a Bachelor’s degree in physics and mathematics at Carnegie Mellon University.
Driving AI with Science
Ziming Liu 刘子鸣
MIT
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
We are entering an era where a handful of mainstream models dominate the AI landscape. While these models have achieved remarkable successes, it is unwise to place all our bets on a single paradigm. To ensure both robustness and progress, we must explore alternative AI models that are not only scientifically grounded but also address the limitations of today’s dominant approaches.
In this talk, I will explore how scientific principles can inspire new directions in AI — how mathematical insights led to the development of the Kolmogorov-Arnold Network, an efficient and interpretable architecture, and how physical intuition gave rise to Poisson Flow Generative Models, a new class of generative models. I will also briefly discuss my broader research agenda, including using science to illuminate the inner workings of foundation models (Science of AI) and accelerating scientific discovery through AI (AI for Science).
Speaker
Ziming Liu is a final-year PhD student at the Massachusetts Institute of Technology (MIT), and an incoming assistant professor of CollegeAI at Tsinghua University. With a strong foundation in both artificial intelligence and physics, he is uniquely positioned to study the foundations of AI through a scientific lens. His research bridges Science for AI—leveraging scientific principles to design and understand AI models—and AI for Science—applying AI to accelerate scientific discovery. Ziming’s work has attracted wide attention in both the AI and physics communities, including a notable contribution to the development of Kolmogorov-Arnold Networks. His research has been cited over 4,000 times, published in leading AI conferences and physics journals, and featured by mainstream media outlets such as Scientific American, Quanta Magazine, and MIT News.
Impossible ecologies: interaction networks and stability of coexistence in ecological communities
Yu Meng 孟雨
Universität Oldenburg
- Host
- Shiheng Zhao, MPI-PKS
- Recording
- Watch the recording
Does an ecological community allow stable species coexistence? Approaches based on random matrix theory can describe the statistical effect of the proportions of competitive, mutualistic, and predator-prey interactions on stability. Still, identifying such general principles remains a central problem of systems ecology because such approaches cannot account for the full network arrangement of these interactions. Here, we therefore develop a different approach: We analyse all interaction networks of 𝑁 ⩽ 5 species with Lotka–Volterra dynamics by combining exact results for 𝑁 ⩽ 3 species and numerical exploration. Surprisingly, we find that a very small subset of these networks are “impossible ecologies”, in which stable coexistence is non-trivially impossible. We prove that the possibility of stable coexistence in general ecologies is determined by similarly rare “irreducible ecologies”. By statistical sampling, we show that the probability of stable coexistence varies over orders of magnitude even in ecologies that differ only in the network arrangement of identical interactions and analyse assembly of these ecological networks and evolutionary or environmental perturbations of them. Strikingly, this reveals that these network arrangements can influence stability of coexistence more than the proportions of interaction types. In this way, our approach of exhaustive analysis of small ecologies reveals the importance of the full structure of the network of interactions for stability of coexistence in ecological communities.
References
- Meng, Y., Horvát, S., Modes, C.D. and Haas, P.A., 2023. Impossible ecologies: Interaction networks and stability of coexistence in ecological communities, Cell Systems (in press) arXiv:2309.16261.
Speaker
Dr Yu Meng is currently a research associate at University of Oldenburg, working with Prof Ulrike Feudel. She received her PhD at University of Aberdeen, under the supervision of Prof Celso Grebogi and co-supervision of Prof Ying-Cheng Lai (from Arizona State University). Before working at University of Oldenburg, she worked as a postdoctoral researcher at the Max Planck Institute for the Physics of Complex Systems and at the Max Planck Institute of Molecular Cell Biology and Genetics. Her main research interests are tipping points in complex networks and nonlinear dynamics including prediction, control and recovery of tipping points, transient , theoretical ecology, population dynamics, effects of network topology on the stability of nonlinear systems, and ecological spatial networks. *yu.meng@uni-oldenburg.de
Geometrically Frustrated Rose Petals
Yafei Zhang 张亚飞
The Hebrew University of Jerusalem
- Host
- Wencheng Ji, Weizmann Institute
- Recording
- Watch the recording
The way nature shapes its creations is amazing, inspiring and often beyond human imagination. Understanding the principles of shape-morphing in nature has long been a cornerstone of biology, material science, physics, applied mathematics and art. Roses, lauded widely for their aesthetic appeal and symbolism, have long served as the muse for numerous artistic and literary creations. Surprisingly, the physical mechanism, which leads to their unique form, remains unknown–until now!
Growth and form are deeply interconnected, in a manner often mediated by mechanical instabilities arising from geometric incompatibilities. Although Gauss incompatibility has long been recognized as the source of morphing in naturally growing slender organs, here [1] we show that the growth profile of rose petals remains Gauss compatible. Their distinctive shape emerges from a different type of geometric incompatibility, the Mainardi-Codazzi-Peterson (MCP) incompatibility, which leads to the formation of localized cusps along the petal margins. We validated this mechanism in model disc petals theoretically, computationally, and experimentally. Our study reveals distinct morphological regimes, ranging from smooth edges to cusp-forming configurations, and demonstrates how stress focusing at cusps influences subsequent petal growth. These findings position MCP incompatibility as a generic mechanism for cusp formation in both natural and manmade self-morphing sheets.
References
- Yafei Zhang, Omri Y. Cohen, Michael Moshe, and Eran Sharon. “Geometrically frustrated rose petals.” Science 388, 6746 (2025): 520-524.
Speaker
Dr. Yafei Zhang is currently an IASH-CHE postdoctoral researcher at the Hebrew University of Jerusalem, working with Prof. Eran Sharon & Prof. Michael Moshe on nonlinear physics. He received his Ph.D. in Solid Mechanics at Tsinghua University, under the supervision of Prof. Changqing Chen and Prof. Quanshui Zheng. His research mainly focuses on the development and application of geometric formalism to model and solve problems in soft-matter systems. Of specific interest are the geometry, topology and mechanics of living matter, growing sheets, and soft metamaterials
“Deriving” Statistical Mechanics from Thermodynamics
从热力学“推导”统计力学
Ruoxun Zhai 翟若迅
GSCAEP
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
传统上，热现象常被解释为微观粒子统计行为的结果，意味着可以从统计物理出发“推导”出热力学定律。然而，这一做法依赖的基础假设——如等概率原理和各态历经假设等——难以通过直接实验验证。相比之下，热力学建立在热机和热循环等实验的坚实基础之上，因此从热力学基本原理反推平衡态统计分布具有重要意义。我们从量子力学出发，通过量子态上待定的统计分布定义内能，再利用功的力学定义，自洽地定义了传热。进而将逆温度 𝜷 作为热的积分因子，我们引入了“热力学可积性”的概念，这把 Carathéodory于1909年提出的可积性思想推广至量子热力学领域，建立了“量子热力学可积性”这一基本框架。
在热力学流形上，该可积性体现为熵的路径无关性。由此构造的熵可积方程决定了统计分布的形式。进一步引入“细致平衡条件”后，该方程的唯一对称解即为玻尔兹曼分布。当细致平衡条件不满足时，熵可积方程仍有非正则解，刻画了有限系统尺寸下的非平衡行为，如有限尺度下的标度现象以及小黑洞辐射中的量子关联。这些非正则解为解决黑洞信息丢失问题提供了新的物理视角。
References
- R. X. Zhai and C. P. Sun, “Quantum Thermodynamic Integrability for Canonical and Noncanonical Statistics”. Phys. Rev. Lett. 134, 160404
Speaker
翟若迅是中国工程物理研究院研究生院(GSCAEP)孙昌璞教授课题组的博士研究生。目前的工作集中于有限时间热力学和量子热力学可积性。感谢各位老师和同学的关注、交流与讨论。
Design Principles of Transcription Factors with Intrinsically Disordered Regions
Wencheng Ji 吉文成
Weizmann Institute
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
Transcription factors (TFs) are proteins that play a crucial role in regulating gene expression by binding to regulatory DNA regions associated with genes. However, in eukaryotic cells, the process by which TFs rapidly find their specific binding targets within large genomes remains a long-standing puzzle. It is found that many eukaryotic TFs have one or more intrinsically disordered regions (IDRs). The functionality of IDRs has garnered a growing interest in the research community in recent years. Experimental findings indicate that the IDR of certain TFs takes a key part in the search process. However, from a theoretical perspective, a qualitative model that explains the search mechanism in eukaryotes is still lacking.
In our paper [1], we use polymer physics to conduct a theoretical investigation into the influence of IDRs on binding probability and search time in TF search. We reveal the optimality principles for binding probability and search time, respectively, and observe significant enhancements in both compared to TFs lacking IDRs. These results highlight the crucial role of IDRs in TF search.
Specifically, we find an optimal TF search process involving a single round of 3D-1D search. Initially, the TF undergoes 3-dimensional diffusion, followed by binding to the DNA antenna region and subsequent 1-dimensional diffusion (which we term “octopusing”, reminiscent of the seminal works by De Gennes on polymer reptation) until it locates its target. Our theoretical predictions and numerical results on search time demonstrate good agreement. This mechanism differs from the well-known facilitated diffusion mechanism observed in bacteria and provides distinct advantages in eukaryotes. Moreover, our model aligns well with existing experimental observations and suggests new experiments to further test its predictions.
References
- W. Ji, O. Hachmo, N. Barkai, A. Amir, “Design principles of transcription factors with intrinsically disordered regions”, eLife, (2025).
Speaker
Wencheng Ji (吉文成) studied condensed matter theory during his bachelor’s and master’s studies at Peking University under the supervision of Prof. Junren Shi. He completed his PhD at EPFL, working with Prof. Matthieu Wyart on statistical physics, specifically focusing on local excitations in amorphous solids. He is currently a postdoctoral researcher at the Weizmann Institute, working with Prof. Ariel Amir on biological physics, following an earlier affiliation with Harvard University. His research interests broadly include the dynamics of complex systems in biology, glasses, and artificial neural networks, particularly through the exploration of their energy landscapes.
Active Bath-Induced Negative Friction
活性热库诱导的负摩擦力
Jihui Pei 裴继辉
Peking University and Leuven University
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
将不同尺度的物理描述关联起来是统计物理的一个重要目标。我们知道，介观尺度的胶体粒子在水中的运动时会感受到来自水分子的摩擦力和热噪声，从而表现为布朗运动。这样的胶体粒子处于一个平衡态的环境当中，而当下越来越多的研究着眼于带有活性的非平衡环境。
我们考察一个球状探测粒子在活性物质组成的热库中的运动。这些活性物质在理论上被描述为1维跑动-翻转（run-and-tumble）粒子或2维活性布朗（active Brownian）粒子。我们发现，对于远离平衡的活性热库以及特定的相互作用形式，线性摩擦系数可能会是负数，此时探测粒子的运动与布朗运动截然不同，探测粒子将会自动向某个方向加速，直到达到某个饱和速度。
更进一步，通过将活性热库的自由度积掉，我们从理论上得到了对探测粒子随机动力学的完整描述。此动力学由非线性摩擦力以及速度依赖的热噪声来刻画。我们据此对粒子的运动模式进行了分类。在非线性摩擦力存在负数区域的情况下，探测粒子的稳态速度分布是多峰的，从而也表现出自推进活性运动的特性。相关结果已由数值模拟所定量验证。此项研究揭示了活性运动可以借助负摩擦力效应由更小尺度的热库粒子向更大尺度的探测粒子传递。
References
- Ji-Hui Pei, and Christian Maes, Phys. Rev. E 111, L032101 (2025).
- Ji-Hui Pei, and Christian Maes, arXiv:2504.03210
Speaker
裴继辉，北京大学物理学院与鲁汶大学物理与天文系联合培养博士生。研究兴趣为涨落定理与活性热库理论。
Response kinetic uncertainty relations for classical and quantum Markovian processes
Kangqiao Liu
Xihua University
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
The thermodynamics and kinetics of a nonequilibrium classical system fundamentally constrain the precision of an observable regarding the celebrated thermodynamic uncertainty relation (TUR) and the kinetic uncertainty relation (KUR). They have also been extended to open quantum systems obeying the Lindblad master equation where unique quantum effects are identified. Recently, a new set of principles that further incorporate the response of a classical system to an external perturbation has been discovered, named the response TUR [1]. In this talk, we present two response KURs (R-KUR). In the first part, we introduce a fundamental trade-off between the precision of response for generic observables and dynamical activity in classical Markovian systems [2]. The practical applicability and tightness of the derived bound are demonstrated through illustrative examples. Our results apply to a broad spectrum of Markov jump processes, ranging from currents to non-current variables, from steady states to time-dependent driving, from continuous to discrete time, and including Maxwell’s demon or absolute irreversibility. In the second part, we generalize the derived classical R-KUR to the steady state of the Lindblad master equation for open quantum systems [3]. The precision and the sensitivity of a measured observable are bounded from above by the conventional quantum dynamical activity and a perturbation-induced inter-subspace transition term. We provide sufficient conditions such that either of the two contributions vanishes, implying their equal importance. We verify our result and illustrate the conditions in an exactly solvable two-level atom.
References
- Krzysztof Ptaszyński, Timur Aslyamov and Massimiliano Esposito, Dissipation Bounds Precision of Current Response to Kinetic Perturbations, Physical Review Letters 133, 227101 (2024).
- Kangqiao Liu and Jie Gu, Dynamical activity universally bounds precision of response in Markovian nonequilibrium systems, Communications Physics 8, 62 (2025).
- Kangqiao Liu and Jie Gu, Response Kinetic Uncertainty Relation for Markovian Open Quantum System, arXiv:2501.04895.
Particle exchange statistics beyond fermions and bosons · in English
Zhiyuan Wang
Max Planck Institute of Quantum Optics
- Host
- Xiao Wang, Shanghai Jiaotong University
- Recording
- Watch the recording
It is commonly believed that there are only two types of particle exchange statistics in quantum mechanics, fermions and bosons, with the exception of anyons in two dimension [1-3]. In principle, a second exception known as parastatistics, which extends outside of two dimensions, has been considered [4] but was believed to be physically equivalent to fermions and bosons [5,6]. In this talk I present a recent work of mine [7] which shows that nontrivial parastatistics inequivalent to either fermions or bosons can exist in physical systems. I first formulate a second quantization theory of paraparticles that is significantly different from previous theories, which turns out to be the key to get new physics. I then present a family of exactly solvable quantum spin models in one and two dimensions where free paraparticles emerge as quasiparticle excitations. Next, I demonstrate a distinctive physical consequence of parastatistics by proposing a challenge game [8] that can only be succeeded using physical systems (in 2D or 3D) hosting paraparticles, as passing the challenge requires the two participating players to secretly communicate information by exploiting the nontrivial exchange statistics of the emergent paraparticles. Finally, I describe the universal properties of emergent paraparticles in gapped topological phases in the framework of tensor category theory, and show that in 2D and 3D, parastatistics correspond to an exotic type of symmetric fusion categories. This work demonstrates the possibility of a new type of quasiparticle in condensed matter systems, and, more speculatively, the potential for previously unconsidered types of elementary particles.
References
- J. M. Leinaas and J. Myrheim, Nuovo Cim. B 37, 1 (1977).
- F. Wilczek, Phys. Rev. Lett. 48, 1144 (1982); Phys. Rev. Lett. 49, 957 (1982).
- C. Nayak, S. H. Simon, A. Stern, M. Freedman, and S. Das Sarma, Rev. Mod. Phys. 80, 1083 (2008).
- H. S. Green, Phys. Rev. 90, 270 (1953).
- S. Doplicher, R. Haag, and J. E. Roberts, Commun. Math. Phys. 23, 199 (1971); 35, 49 (1974).
- S. Doplicher and J. E. Roberts, Commun. Math. Phys. 28, 331 (1972).
- Z. Wang and K. R. A. Hazzard, Nature 637, 314 (2025).
- Z. Wang, arXiv:2412.13360 (2024).
2024
Law of Balance, Stationary Distribution and Noise Equilibrium of Stochastic Gradient Descent (SGD)
Hongchao Li
University of Tokyo
- Host
- Ruicheng Bao, University of Tokyo
- Recording
- Watch the recording
How the stochastic gradient descent (SGD) navigates the loss landscape of a neural network remains poorly understood. In this seminar, I am going to introduce our recent work towards understanding the stationary solution and noise equilibrium in the dynamics of SGD. We directly show that the minibatch noise of SGD regularizes the solution towards a noise-balanced solution whenever the loss function contains a rescaling symmetry. We prove that when the rescaling symmetry exists, the SGD dynamics is limited to only a low-dimensional subspace and prefers a special set of solutions in an infinitely large degenerate manifold, which offers a partial explanation of the effectiveness of SGD in training neural networks. We then apply this result to derive the stationary distribution of stochastic gradient flow for a diagonal linear network with arbitrary depth and width, which is the first analytical expression of the stationary distribution of SGD in a high-dimensional non-quadratic potential. The stationary distribution exhibits complicated nonlinear phenomena such as phase transitions, loss of ergodicity, memory effects, and fluctuation inversion. For a general exponential symmetry, we show the existence and uniqueness of the fixed point in the SGD and apply the result to directly calculate the noise equilibrium of a fully-connected deep neural network.
A fluctuation-sensitivity-timescale trade-off in feedback-controlled dynamics
Ka Kit Kong
Peking University
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
Signal transduction is ubiquitous in biological systems and vital for most biological functions, which is, however, inevitably impeded by noise due to the stochastic nature of biochemical reactions. While noise attenuation mechanisms, such as feedback control, have been studied for decades, noise is not the sole factor determining the performance of signal transduction. Effective signal transduction generally necessitates small noise, high response sensitivity, and short response timescale, along with other factors. However, how noise attenuation mechanisms affect these factors simultaneously remains unclear.
Feedback control is a renowned mechanism for buffering intrinsic fluctuation in regulatory networks. Here, we study a general feedback-controlled network in which the feedback is achieved by a complex interactive module, and analytically derive a fundamental trade-off between fluctuation, sensitivity, and timescale altered by the feedback. We show that feedback control cannot infinitely suppress fluctuation without the cost of reducing sensitivity or response speed. Furthermore, the lower bound for this trade-off can be reduced up to half in non-gradient dynamical systems compared to gradient systems. We validate this trade-off as a tight bound for high-dimensional systems in nonlinear regime through numerical simulations. These results elucidate the fundamental limitation of feedback control in enhancing the information transmission capacity of regulatory networks.
References
- Ka Kit Kong, and Feng Liu. “A fluctuation-sensitivity-timescale trade-off in feedback-controlled dynamics.” arXiv preprint arXiv:2410.00559 (2024).
- Ka Kit Kong, Chunxiong Luo, and Feng Liu. “A phase diagram structure determines the optimal sensitivity-precision trade-off in signaling systems.” Communications Physics 7.1 (2024): 80.
Robust Edge Flows in Swarming Bacteria Colonies
He Li
IGBMC
- Host
- Shiling Liang, MPI-PKS
- Recording
- Watch the recording
Understanding if and how the chirality of biomolecules is transferred across scales into larger components and active processes remains elusive. For instance, flagellated bacteria swim in helical trajectories but chirality seems absent from the active turbulence dynamics they display in dense suspensions. We address this question by examining multi-scale dynamics in Paenibacillus vortex colonies. We find active turbulence without manifest chirality in the bulk, but wide, clockwise (viewed from the air side) circulation all along the tortuous centimeter-scale external boundary, while similar but counter-clockwise flows follow internal boundaries. We trace back the origin of these robust edge flows to an unexpected asymmetry at the individual level that is amplified by local interactions. We rationalize our findings with a model of noisy self-propelled particles immersed in a Stokes fluid that accounts faithfully for our observations. Our modeling and experimental efforts reveal that local nematic alignment and hydrodynamic interactions amplify the weak chiral bias in individual motion, promoting the formation of strong edge flows. The likely topological protection of these flows provides robust transport mechanism over large scales. Such robust boundary phenomena in weakly chiral active fluids may inspire new control strategies for active and biological matter.
References
- He Li, Hugues Chaté, Masaki Sano, Xia-qing Shi, and H. P. Zhang, Robust edge flows in swarming bacterial colonies, Physical Review X (accepted) (2024)
- C. J. Ingham and E. Jacob, Swarming and complex pattern formation in Paenibacillus vortex studied by imaging and tracking cells, BMC Microbiology 8, 36 (2008).
- K. Dasbiswas, K. K. Mandadapu, and S. Vaikuntanathan, Topological localization in out-of-equilibrium dissipative systems, Proceedings of the National Academy of Sciences 115, E9031 (2018).
- V. Soni, E. S. Bililign, S. Magkiriadou, S. Sacanna, D. Bartolo, M. J. Shelley, and W. T. M. Irvine, The odd free surface flows of a colloidal chiral fluid, Nature Physics 15, 1188 (2019).
- E. Lauga, W. R. DiLuzio, G. M. Whitesides, and H. A. Stone, Swimming in circles: Motion of bacteria near solid boundaries, Biophysical Journal 90, 400 (2006).
Designing self-adaptive network with thermodynamics
Mingyang Bai
Beihang University
- Host
- Shiling Liang, EPFL
To realize survival goals in unknown environments, animals and plants learn the environment property and adjust their behavior accordingly. Such adaptability is widely believed to be the essence of general intelligence. Compared to fixed strategies, adaptive strategy seems more reasonable for designing complex systems in uncertain environments. However, many adaptive control methods are challenged by numerous degrees of freedom and limited information for complex systems. Inspired by thermodynamics, we propose a self-adaptation strategy enabling complex systems to adaptively realize the desired landscape in unknown environments with only macroscopic information.
We choose the macroscopic landscape as our target, which corresponds to free energy landscape in thermodynamics. In our strategy, when environmental disturbance drives system from one microstate to another, systems only use macroscopic information to decide whether to accept or reject. We derive the suitable acceptance probability from target landscape and environment entropy under detailed balance condition. To estimate environment entropy in unknown environments, we propose a simple method enabling system to update estimation on environment entropy by extending the Wang-Landau method. Thus system could adaptively realize desired target landscape in unknown environments. Then we employ our self-adaptation strategy to design a self-adaptive network programmed with desired target landscape represented by topological measures, which realizes target under several scenarios. In addition, the adaptation process is described by unique power law distinguishing our adaptive systems from non-adaptive systems. Our self-adaptive strategy can be applied to designing future intelligent complex systems and help understand the principle behind complex system adaptability and system intelligence.
References
- Bai, Mingyang, and Daqing Li. “Designing self-adaptive network with thermodynamics.” arXiv preprint arXiv:2407.04930 (2024).
Geometry of optimal control in chemical reaction networks
Yikuan Zhang
Peking University
- Host
- Shiling Liang, EPFL
Although optimal control (OC) has been studied in stochastic thermodynamics for systems with continuous state variables, less is known in systems with discrete state variables, such as Chemical Reaction Networks (CRNs). Here, we develop a general theoretical framework to study OC of CRNs for changing the system from an initial distribution of states to a final distribution with minimum dissipation. We derive a “Kirchhoff’s law” for the probability current in the adiabatic limit, from which the optimal kinetic rates are determined analytically for any given probability trajectory. By using the optimal rates, we show that the total dissipation is determined by a L2-distance measure in the probability space and derive an analytical expression for the metric tensor that depends on the probability distribution, network topology, and capacity of each link. Minimizing the total dissipation leads to the geodesic trajectory in the probability space and the corresponding OC protocol is determined by the Kirchhoff’s law. To demonstrate our general approach, we use it to find a lower bound for the minimum dissipation that is tighter than existing bounds obtained with only global constraints. We also apply it to simple networks, e.g., fully connected 3-state CRNs with different local constraints and show that indirect pathway and non-functional transient state can play a crucial role in switching between different probability distributions efficiently. Future directions in studying OC in CRNs by using our general framework are discussed.
References
- Zhang, Y., Ouyang, Q. and Tu, Y., 2024. Geometry of optimal control in chemical reaction networks. arXiv preprint arXiv:2407.05620.
Emergence of Memory in Equilibrium versus Nonequilibrium Systems
Xizhu Zhao
Max Planck Institute for Multidisciplinary Sciences
- Host
- Shiling Liang, EPFL
Experiments often probe observables that correspond to low-dimensional projections of high-dimensional dynamics. In such situations distinct microscopic configurations become lumped into the same observable state. It is well known that correlations between the observable and the hidden degrees of freedom give rise to memory effects. However, how and under which conditions these correlations emerge remain poorly understood. Here we shed light on two fundamentally different scenarios of the emergence of memory in minimal stationary systems, where observed and hidden degrees of freedom either evolve cooperatively or are coupled by a hidden nonequilibrium current. In the reversible setting the strongest memory manifests when the timescales of hidden and observed dynamics overlap, whereas, strikingly, in the driven setting maximal memory emerges under a clear timescale separation. Our results hint at the possibility of fundamental differences in the way memory emerges in equilibrium versus driven systems that may be utilized as a “diagnostic” of the underlying hidden transport mechanism.
References
- Zhao, X., Hartich, D., & Godec, A. (2024). Emergence of Memory in Equilibrium versus Nonequilibrium Systems. Physical Review Letters, 132(14), 147101.
Universal Fluctuation-Response Relations of Nonequilibrium Dynamics: A Trajectory Information Geometry Framework
Jiming Zheng
UNC Chapel Hill
- Host
- Shiling Liang, EPFL
Unraveling the universal principles governing the response of complex systems to environmental changes is crucial for predicting and controlling their behavior. While fluctuation-dissipation relations have been established for systems near equilibrium, a general framework for understanding the responsiveness of systems far from steady states remains elusive. Here, we introduce a novel approach based on the information geometry of stochastic trajectories to derive a set of universal thermodynamic bounds on the response of any Markov system, regardless of its proximity to steady states. This theory establishes a new paradigm in non-equilibrium statistical mechanics, providing a unified perspective on the behavior of non-stationary systems, from biological processes to engineered devices, and paving the way for designing complex responsiveness in far-from-equilibrium systems.
References
- Zheng, Jiming, and Zhiyue Lu. “Information Geometry and Universal Bounds on Non-stationary Responsiveness of Markov Dynamics.” arXiv preprint arXiv:2403.10952 (2024).
2023
Restart Expedites Quantum Walk Hitting Times
Ruoyu Yin
Bar-Ilan University
- Host
- Shiling Liang, EPFL
Classical first-passage times under restart are used in a wide variety of models, yet the quantum version of the problem still misses key concepts. We study the quantum hitting time with restart using a monitored quantum walk. The restart strategy eliminates the problem of dark states, i.e., cases where the particle evades detection, while maintaining the ballistic propagation which is important for a fast search. We find profound effects of quantum oscillations on the restart problem, namely, a type of instability of the mean detection time, and optimal restart times that form staircases, with sudden drops as the rate of sampling is modified. In the absence of restart and in the Zeno limit, the detection of the walker is not possible, and we examine how restart overcomes this well-known problem, showing that the optimal restart time becomes insensitive to the sampling period.
References
- Yin, R. and Barkai, E., 2023. Restart expedites quantum walk hitting times. Physical Review Letters, 130(5), p.050802.
- Yin, R. and Barkai, E., 2023. Instability in the quantum restart problem. arXiv preprint arXiv:2301.06100.
Quantum Reliability
Lianxiang Cui
Beijing Computational Science Research Center
- Host
- Shiling Liang, EPFL
In this talk, we are going to establish a bridge between quantum mechanics and reliability in system engineering, introducing a novel framework for reliability analysis specifically designed for quantum devices with quantum coherence, known as Quantum Reliability. For engineering, the reliability characterizes equipment’s ability to function flawlessly. The concept of reliability gained its precise meaning during the twenties of the last century, coinciding with the birth of quantum mechanics. Today, with the fast advancement of quantum technology, the promotion of quantum devices enhanced by quantum coherence has become a subject of intense fascination. The preservation of a quantum device’s long-term functionality has emerged as a great concern, inevitably leading to the convergence of quantum physics and reliability engineering. The quantum reliability is proposed in a general framework, that the significant influence of quantum coherence on device structure and degradation is clearly characterized. In the future applications to quantum engineering, quantum reliability will serve as a vital measure, enabling the evaluation and optimization of quantum devices’ design. As this captivating journey unfolds, the alliance between quantum reliability and cutting-edge quantum technologies will shape a future of unparalleled reliability and innovation.
References
- L. X. Cui, Y-M. Du and C. P. Sun, Phys.Rev.Lett.131.160203(2023)
Cas9 Target Search and Recognition from a Physical Perspective
Qiao Lu
OIST
- Host
- Shiling Liang, EPFL
Originating from the immune system of prokaryotes, Cas9 protein can recognize any target on DNA under the guidance of RNA. We first use a physical model to analyze and reconcile seemingly contradictory experimental results, confirming that Cas9 can perform facilitated diffusion, and draw an analogy to Anderson localization in condensed matter theory, which can be extended to other protein-DNA interactions. Then we theoretically study the search efficiency of Cas9, finding an “overkill” effect and calculating its optimal solution, which is unique to Cas9 but similar to the “stability-speed paradox” in previously studied protein-DNA interactions.
References
- Q. Lu, D. Bhat, D. Stepanenko, S. Pigolotti, “Search and localization dynamics of the CRISPR/Cas9 system”, Phys. Rev. Lett. 127, 208102, arXiv/2103.10667, recommended by APS online journal Physics, (2021)
Surface Tensions in Active Fluids
Yongfeng Zhao
Soochow University
- Host
- Shiling Liang, EPFL
Surface tensions are state variables that control various interfacial phenomena in equilibrium systems, including Laplace pressure jump, wetting, and interface fluctuations. Phase boundaries are also commonly found in active materials, due to the wealth of phase transitions they display, such as the motility-induced phase separation. However, in the absence of free-energy-abiding dynamics, what controls the properties of interfaces in active media remains elusive. The situation is particularly confusing for active particles interacting with purely repulsive forces, for which a liquid-vapor surface tension has been predicted to be negative [1]. This questions whether a proper surface tension can actually be defined and the properties it may control.
In this talk, we consider fluid-solid and fluid-fluid interfaces. In both cases, we show that tangential forces can be defined and measured, which allows us to provide a mechanical definition of surface tension [2]. The situation is shown to be much richer than for equilibrium systems due to the emergence of non-equilibrium currents and drags, which alter the mechanical balance at interfaces. This is shown to play an important role in wetting phenomena, where we show how the Young-Dupré’s law, which controls the contact angle of a droplet partially wetting a solid surface, has to be generalized for active fluids.
References
- Julian Bialké, Jonathan T. Siebert, Hartmut Löwen, and Thomas Speck, Phys. Rev. Lett. 115, 098301 (2015)
- R. Zakine, Y. Zhao, M. Knežević, A. Daerr, Y. Kafri, J. Tailleur, and F. van Wijland, Phys. Rev. Lett. 124, 248003 (2020)
Expanding Sensory Horizons: Multiplexing and Temporal Pattern Recognition in Stochastic Sensors
Zhiyue Lu
University of North Carolina at Chapel Hill
- Host
- Shiling Liang, EPFL
Macromolecular sensory receptors, despite their microscopic scale, serve as essential information transduction units in biological cells. Contrary to the prevailing assumption that thermal noise impedes information transduction, we find that it can facilitate multiplexing, enabling a single sensor to scramble and convey multiple information channels. We demonstrate that even a simple binary-state ligand-receptor sensor can simultaneously encode multiple independent environmental variables. We also propose a general theory of stochastic sensory multiplexing, suggesting two theoretical upper bounds on a sensor’s sensory ability. Building on this foundation, we focus on the sensor’s state dynamics and explore its capacity for discerning temporal patterns within a limited sensing period. We present a theoretical upper bound for a sensor’s ability to distinguish temporal patterns of different ligand concentration protocols. This upper bound is expressed as a weighted expectation of the cumulative number of binding events within the observation time. This insight illuminates the path for designing stochastic sensors, opening the door to developing more efficient artificial sensors and enhancing our understanding of sensory processes in biology.
The nonequilibrium thermodynamics of flocking of active spins · in English
Qiwei Yu
Princeton University
- Host
- Shiling Liang, EPFL
The collective coherent motion known as flocking is a nonequilibrium phenomenon and is sustained by a continuous input of free energy. By studying the energy dissipation of the active Ising model (AIM), we show that the energy cost can be decomposed into two parts, namely the cost for self-propelled motion and an additional energy dissipation required to align spins in order to maintain the flocking order. We find that this additional alignment dissipation reaches its maximum at the flocking transition point in the form of a cusp with a discontinuous first derivative with respect to the interaction strength. To understand this singular behavior, we analytically solve the two-site and three-site AIM models and obtain the exact dependence of the alignment dissipation on the flocking order parameter and control parameter, which explains the cusped dissipation maximum at the flocking transition. Our results reveal a trade-off between the energy cost of the system and its performance measured by the flocking speed and sensitivity to external perturbations. This trade-off relationship provides a new perspective for understanding the dynamics of natural flocks and designing optimal artificial flocking systems.
References
- Yu, Q. and Tu, Y., 2022. Energy Cost for Flocking of Active Spins: The Cusped Dissipation Maximum at the Flocking Transition. Physical Review Letters, 129(27), p.278001.