Research

Active matter refers to a nonequilibrium system whose components self-propel according to their internal degrees of freedom. Unlike traditional nonequilibrium systems, active matter creates self-structured collective formations and movements without the need for an externally applied bias. The scope of active matter is diverse, encompassing numerous biological and synthetic systems, from molecular motors and colloidal particles to bird flocks and bumper cars. Serving as an effective paradigm, active matter offers insights into the physics of various biological phenomena and aids in the creation of metamaterials with novel response characteristics.

Using theoretical and computational tools of statistical physics, our group studies the following aspects of active matter.

• In active matter, an asymmetric obstacle generically creates a flow via an effect called current rectification. Recently, we found that the motion of the obstacle can trigger a similar effect, even when the obstacle itself is entirely isotropic. This prompts several questions: when does this occurrence manifest, what kinds of collective phenomena can it induce, and how can we harness this to regulate the flow of matter and energy? We investigate these issues for both dilute active fluids and dense active granular systems.
  [Ref: K.-W. Kim, Y. Choe, and Y. Baek, "Generic symmetry-breaking motility in active fluids", arXiv:2304.01645, Phys. Rev. E 109, 014614 (2024)]

• Energy dissipation is a signature trait of active matter. However, existing theoretical approaches, which predominantly concentrate on the dynamic aspects of active matter, often fail to provide satisfactory solutions to queries such as the required level of dissipation to sustain a specific structure or to attain a desired power output. To remedy this shortcoming, we explore the models of active matter that incorporate the system's thermodynamic details.
  [Ref: Y. Oh and Y. Baek, "Effects of the self-propulsion parity on the efficiency of a fuel-consuming active heat engine", arXiv:2302.13870, Phys. Rev. E 108, 024602 (2023)]

Recent years have seen machine learning emerge as a powerful instrument for inferring general functional relationships (known as supervised learning) or generating new data (known as unsupervised learning) from provided samples. The questions that arise are:

• How can we apply machine learning efficiently to develop a data-driven approach to the statistical physics of nonequilibrium and complex systems?
  [Ref: E. Kwon and Y. Baek, "α-divergence improves the entropy production estimation via machine learning", arXiv:2303.02901, Phys. Rev. E 109, 014143 (2024)]

• What insights can statistical physics offer us about the inner workings of machine learning? In particular, can we understand under what conditions machine learning exhibits the optimal performance?
  [Ref: G. Kim, H. Lee, J. Jo, and Y. Baek, "Tradeoff of generalization error in unsupervised learning", arXiv:2303.05718, J. Stat. Mech.: Theor. Exp. 2023, 083401 (2023)]

Historically, thermodynamics has been confined to inequalities whose saturation requires macroscopic systems undergoing quasistatic processes. Nevertheless, recent theoretical developments have comprehensively reshaped this field to include equalities and inequalities satisfied even by microscopic systems experiencing regular dynamic processes. Aided by techniques from stochastic thermodynamics, quantum thermodynamics, and information geometry, we delve into the following questions:

• What kinds of universal equalities and inequalities apply to nonequilibrium processes? How do they interrelate? And how can we express them in empirically observable ways?
  [Ref: E. Kwon, J.-M. Park, J. S. Lee, and Y. Baek, "Unified hierarchical relationship between thermodynamic tradeoff relations", arXiv:2311.01098, Phys. Rev. E 110, 044131 (2024)]

• How can we characterize the thermodynamics of quantum systems, considering the impacts of entanglement, coherence, and quantum correlations between the system and its environment? What are the possible consequences for quantum engines, batteries, and computers?