A nonequilibrium equality for free energy differences

C. JarzynskiarXiv 1996 · DOI · arXiv

TL;DR

This seminal paper derives the Jarzynski Equality, a fundamental identity in statistical mechanics that relates equilibrium free energy differences to the work performed during non-equilibrium, finite-time processes. It proves that by averaging the exponential of work over many irreversible trials, one can recover exact equilibrium information, effectively turning the Second Law's inequality into an exact equality.

What Problem Does This Paper Address?

In classical thermodynamics, calculating the free energy difference ($\Delta F$) between two states usually requires a "quasi-static" process—one performed infinitely slowly so the system remains in equilibrium at every step. For any real-world, finite-time process, the Second Law of Thermodynamics dictates that the average work performed ($\overline{W}$) will always be greater than or equal to the free energy difference ($\overline{W} \geq \Delta F$). The "extra" work is dissipated as heat due to irreversibility.

The challenge was that while we had inequalities for non-equilibrium processes, we lacked a general equality that could extract precise equilibrium values from experiments or simulations that happen away from equilibrium. Jarzynski sought to bridge this gap, providing a way to calculate $\Delta F$ from an ensemble of fast, irreversible measurements.

Methods

The paper provides a theoretical derivation rooted in Hamiltonian mechanics and statistical ensembles. The author employs three main logical steps:

  1. Isolated Systems: First, the equality is derived for a deterministic, isolated Hamiltonian system using Liouville’s Theorem (which states that phase-space density is conserved along trajectories). By defining a "work accumulated" function, the author shows the equality holds when starting from a canonical distribution.
  2. Coupling to a Reservoir: The proof is extended to a system in contact with a heat bath. By treating the system and reservoir together as one large isolated Hamiltonian system and assuming weak coupling, the author demonstrates that the equality remains valid.
  3. Numerical Thermostats: To ensure the result is useful for computational science, the author proves the equality also holds for systems evolved using the Nosé-Hoover thermostat and Metropolis Monte Carlo algorithms, which are standard tools for "mocking up" heat reservoirs in molecular dynamics simulations.

Datasets

Not discussed in the paper. This is a theoretical derivation; no empirical datasets or specific simulation results are presented. The author references existing literature to show how his result reduces to well-known equilibrium identities.

Key Insights and Ideas

  • Independence of Path and Rate: The most striking insight is that the equality $\overline{\exp(-\beta W)} = \exp(-\beta \Delta F)$ holds regardless of the path taken between states or how fast the parameters are switched. Whether you move a piston in a microsecond or a millennium, the identity remains true.
  • Rare Events Matter: The average is dominated by trajectories where the work performed is significantly less than the average work—specifically, those rare events that seem to "defy" the Second Law locally.
  • Unifying Equilibrium Identities: The paper shows that two famous identities—Kirkwood’s thermodynamic integration (the slow limit) and Zwanzig’s free energy perturbation (the instantaneous limit)—are actually just special cases of this new, more general non-equilibrium result.
  • Microscopic Basis for the Second Law: By applying Jensen’s inequality to his result, Jarzynski derives the macroscopic Second Law ($\overline{W} \geq \Delta F$) directly from microscopic Hamiltonian dynamics, rather than simply assuming an increase in entropy.

Key Results

  • The Jarzynski Equality: The central result is Equation 2a: $$\overline{\exp(-\beta W)} = \exp(-\beta \Delta F)$$ where $\beta = 1/k_B T$.

  • Fluctuation-Dissipation Relation: For processes where the work distribution is Gaussian (typical for slow switching), the paper derives a specific relation: $\Delta F = \overline{W} - \beta \sigma^2 / 2$. This relates the dissipated work directly to the variance (fluctuations) of the work performed.

  • Computational Utility: The proof that the equality holds for Nosé-Hoover dynamics provided a theoretical green light for researchers to use non-equilibrium simulations to calculate equilibrium free energies, potentially saving massive amounts of computation time.

Limitations and Future Work

  • The Sampling Problem: The author honestly notes a major practical hurdle: if the fluctuations in work ($\sigma$) are much larger than the thermal energy ($k_B T$), the average will be dominated by extremely rare trajectories. In such cases, an "unreasonably large number of measurements" would be needed to get an accurate result.
  • Scale Limits: Because of this sampling issue, the equality is most applicable to nanoscale systems (like single-molecule experiments or small clusters) where fluctuations are significant relative to the total work. It is less practical for macroscopic systems where fluctuations are negligible.
  • Quantum Effects: The derivation is strictly classical. The author notes that the result should be valid "provided... quantal effects may be ignored," leaving the door open for quantum generalizations.

Summary updated

Abstract

An expression is derived for the classical free energy difference between two configurations of a system, in terms of an ensemble of finite-time measurements of the work performed in parametrically switching from one configuration to the other. Two well-known equilibrium identities emerge as limiting cases of this result.
BibTeX
@misc{jarzynski1996nonequilibrium,
  author = {C. Jarzynski},
  title = {{A nonequilibrium equality for free energy differences}},
  year = {1996},
  eprint = {cond-mat/9610209},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/cond-mat/9610209},
}
Figure