Kps.or.kr

South Korea's government and one of Asia's leading physics institutions awarded the 2026 Lee Hwi-so Prize to University of Maryland physicist Christopher Jarzynski on Monday, recognizing the equation he derived in 1997 that brought the Second Law of Thermodynamics into the world of individual molecules — and that now underpins a growing class of generative AI training algorithms. The prize was jointly announced by South Korea's Ministry of Science and ICT and the Asia-Pacific Center for Theoretical Physics.

What the Jarzynski Equality Actually Does

Classical thermodynamics offers a clean guarantee: if you transform a system from one state to another infinitely slowly — a so-called quasistatic or reversible process — the work you perform equals the difference in free energy between the two states. Free energy is the thermodynamic quantity that tells scientists how much usable work a system can produce, and knowing its change between states is essential to understanding whether a chemical reaction will proceed spontaneously, how tightly a drug binds to a protein, or how a protein folds.

The trouble is that real experiments are never infinitely slow. Drive a system too fast, and it falls out of thermal equilibrium. Classical thermodynamics still offers a statement in that case — the irreversible work done will always exceed the free energy difference — but that statement is an inequality, not an equation. It tells you the free energy is somewhere below the work you measured. It doesn't tell you where.

Jarzynski's equality changes this. It says that by performing the same irreversible transformation many times and computing the average of the exponential of the negative work — written formally as ⟨exp(−W/kT)⟩ — a scientist obtains a quantity that equals exactly the exponential of the negative free energy difference, regardless of how fast or violently the process occurs. The key operation is the exponential average, which gives disproportionate weight to the rare, low-dissipation realizations of the experiment and extracts from them the equilibrium information that classical thermodynamics said was only accessible through reversible processes. The 1997 paper announcing this result appeared in Physical Review Letters under the title "Nonequilibrium Equality for Free Energy Differences."

As a mathematical bonus, the Second Law inequality — the statement that work always exceeds the free energy difference — drops out of the Jarzynski equality for free. Because the exponential function is convex, Jensen's inequality from mathematics guarantees that the average of exp(−βW) is at least as large as exp(−β⟨W⟩), which immediately recovers the classical result ⟨W⟩ ≥ ΔF. The Second Law is not contradicted by the Jarzynski equality; it is contained within it as a limiting case.

From Nobel Stages to Drug Discovery to Generative AI

The original paper appeared on April 7, 1997, in Physical Review Letters under the title "Nonequilibrium Equality for Free Energy Differences." According to Jarzynski's University of Maryland faculty profile, the work has accumulated thousands of citations in the scientific literature.

Experimental confirmation came from multiple directions. In 2002, a team led by Carlos Bustamante at the University of California, Berkeley, demonstrated the equality using optical tweezers to repeatedly unfold a single RNA molecule, comparing the work distribution against the molecule's known equilibrium free energy. A 2014 experiment verified a quantum-mechanical analog of the equality in a system of trapped ions, published in Nature Physics.

The Nobel Committee for Physics took note: when the 2018 Nobel Prize in Physics honored the invention of optical tweezers — the laser-based tools capable of gripping and manipulating individual biological molecules — the committee's scientific background document explicitly cited experimental tests of the Jarzynski equality as among the landmark achievements made possible by the technology. That citation placed Jarzynski's equation directly within the Nobel record for physics, even though the prize went to Arthur Ashkin, Gérard Mourou, and Donna Strickland for the tools themselves.

Today, the equality's footprint extends well beyond biophysics. Computational chemists use Jarzynski-based free-energy methods routinely in drug discovery, computing binding affinities for candidate molecules with simulations that deliberately drive the system out of equilibrium and apply the equality to recover the equilibrium answer. A growing body of work published since 2023 has connected the equality to generative AI: energy-based generative models — which represent probability distributions as Boltzmann-type exponentials, exactly the mathematical language of statistical mechanics — can be trained more efficiently by deploying the Jarzynski equality to importance-sample across the nonequilibrium dynamics of model training, as demonstrated in a 2023 NeurIPS paper on energy-based models by Carbone, Hua, Coste, and Vanden-Eijnden. The APCTP noted Jarzynski's own recent POSTECH lecture on this connection, titled "Nonequilibrium Statistical Physics, Generative AI, and Free Energy Estimation," delivered Monday in Pohang.

Why Measuring Free Energy the Hard Way Still Matters

A reader might reasonably ask why measuring free energy from irreversible experiments is so useful when equilibrium experiments exist. The answer is practical. At the scale of individual molecules, quasistatic processes are often experimentally impossible: a protein does not wait for the experimenter to move infinitely slowly. Force spectroscopy, the technique of applying mechanical forces to single molecules with optical or magnetic tweezers, inevitably drives the molecule out of equilibrium on experimental timescales.

The Jarzynski equality makes the irreversibility an asset rather than a contamination. Instead of needing to perform the process slowly enough to avoid dissipation — which may require days of computer simulation per molecule — a researcher can drive the transformation fast, collect the work values from many trajectories, apply the exponential-average formula, and recover the equilibrium free energy. The practical price is statistical: the rare, low-work trajectories carry the most weight, so convergence requires enough repetitions to sample them. Researchers have developed extensions — the Crooks fluctuation theorem, bidirectional free-energy methods, escorted Jarzynski protocols — that improve convergence and reduce the number of required trajectories.

At the University of Maryland, Jarzynski holds joint appointments in the Department of Chemistry and Biochemistry, the Department of Physics, and the Institute for Physical Science and Technology, as detailed on his Maryland faculty profile page. His current research interests include the thermodynamics of information processing, shortcuts to adiabaticity in quantum systems, and the arrow of time — questions that connect the same nonequilibrium framework to fundamental puzzles about why time runs in one direction.

The Prize and the Korean Physicist It Honors

The Lee Hwi-so Prize — formally called the Benjamin W. Lee Professorship — was established in 2012 by APCTP and the Korean Physical Society to honor the memory of Benjamin Whisoh Lee (이휘소; January 1, 1935 – June 16, 1977), Korea's most celebrated theoretical physicist. The award is presented annually at the KPS autumn academic conference, where the recipient delivers a keynote lecture. Jarzynski is scheduled to address the KPS autumn conference keynote on October 21 (Korean Standard Time) at the Jeju International Convention Center.

Benjamin Whisoh Lee was born in Seoul during the period of Japanese colonial rule, arrived in the United States in 1955, and rose to become head of the Theoretical Physics Department at Fermi National Accelerator Laboratory near Chicago. His contributions to the renormalization of non-Abelian gauge theories — the mathematical procedure that gives the Standard Model of particle physics its predictive accuracy — were foundational to a series of discoveries later honored with Nobel Prizes in Physics. He predicted the approximate mass of the charm quark and helped establish the theoretical basis for what became the modern understanding of weak interactions, as documented in a 2025 Physics Today retrospective on his contributions.

Lee died on June 16, 1977, in an automobile accident near Kewanee, Illinois, at the age of 42, cutting short a career that Nobel laureate C.N. Yang, in his memorial address at Fermilab, described as having peaked at the moment of Lee's death.

In March 2026, South Korea's Overseas Koreans Agency honored Lee as its "Overseas Korean of the Month," describing him as "the Korean Oppenheimer" and noting that more than 60 of his 140-plus papers have each accumulated more than 10,000 citations in the scientific literature.

Is This Equation the Reason You Should Know Nonequilibrium Thermodynamics?

For a reader without a physics background, the relevant takeaway is this: the Second Law of Thermodynamics has long been taught as a rule about averages — systems dissipate energy, processes lose efficiency, entropy only increases. What Jarzynski's equality revealed is that hidden within the randomness of dissipation, at the level of individual trajectories, there is an exact mathematical identity. That identity connects the messy, fluctuating work done in a real finite-time experiment to the pristine equilibrium free energy. It is the kind of result that, once seen, makes the field appear to have been waiting for it.

The award presented in Seoul today places Jarzynski's work in the company of previous Lee Hwi-so Prize recipients — international figures in theoretical physics — and gives formal institutional recognition to a result that has already reshaped biophysics, computational chemistry, and, most recently, the mathematical foundations of machine learning.

APCTP director Misao Sasaki noted at the announcement that the center intends to continue strengthening networks between world-class scholars and Korean researchers. South Korea's Ministry of Science and ICT official Yoon Kyeong-suk said the event "will inspire young scientists' intellectual curiosity and present a new vision for research built on the enormous legacy of Dr. Lee Hwi-so."

Frequently Asked Questions

What exactly is the Jarzynski equality, and why is it considered a breakthrough?

The Jarzynski equality is a mathematical identity in statistical mechanics that lets scientists extract exact equilibrium thermodynamic information — specifically, the free energy difference between two states of a system — from experiments conducted far from equilibrium. Before Jarzynski's result, classical thermodynamics said this was only possible when a process was carried out infinitely slowly. The breakthrough is that the equality holds regardless of how fast or violently the transformation occurs, provided the experimenter averages a specific exponential function of the work over many repetitions. The Second Law of Thermodynamics — the statement that real processes always dissipate energy — follows as a mathematical consequence of the equality rather than an independent postulate.

How is a thermodynamics equation from 1997 now relevant to artificial intelligence?

The mathematics of thermodynamics and the mathematics of machine learning share a deep structural connection. A class of generative AI models called energy-based models represents probability distributions using the same Boltzmann-type exponential functions that appear in statistical mechanics. Training these models efficiently requires computing quantities analogous to thermodynamic free energies. Researchers have found that the Jarzynski equality — which was designed to estimate free energies from nonequilibrium processes — can be applied directly to improve how energy-based generative models are trained. Papers published from 2023 onward explicitly invoke the equality in the context of diffusion models and other generative architectures, including a NeurIPS 2023 Jarzynski training paper by Carbone, Hua, Coste, and Vanden-Eijnden demonstrating improved training efficiency.

Who was Benjamin Whisoh Lee, and why does a South Korean physics prize bear his name?

Benjamin Whisoh Lee (1935–1977) was a Korean-born American theoretical physicist who became one of the architects of the Standard Model of particle physics. He headed the Theoretical Physics Department at Fermilab, contributed foundational work on the renormalization of gauge theories, and helped establish the theoretical basis for the charm quark. He died in an automobile accident at 42, at what colleagues described as the peak of his career. The prize bearing his name was established in 2012 by the Asia-Pacific Center for Theoretical Physics and the Korean Physical Society to honor his legacy and to recognize international scholars who advance theoretical physics at the same level of ambition Lee himself embodied.

What are the practical applications of the Jarzynski equality beyond theoretical physics?

Three major application domains have emerged. In biophysics, the equality enables single-molecule force-spectroscopy experiments — using optical or magnetic tweezers — to recover the thermodynamic stability of proteins, RNA molecules, and DNA structures from irreversible pulling experiments, without requiring the infinitely slow processes that classical thermodynamics would demand. In drug discovery, computational chemistry pipelines use Jarzynski-based free-energy perturbation methods to estimate how tightly candidate drug molecules bind to biological targets, accelerating simulations that would otherwise require far longer run times. In AI research, energy-based generative models can be trained using importance-sampling methods that draw directly on the equality's mathematical framework, making the training of certain generative architectures more statistically efficient.