Jarzynski equality (theorem)

Non-equilibrium work identity: the exponential work average equals the equilibrium free-energy difference for a system driven by an arbitrary protocol from an initial Gibbs state.
Jarzynski equality (theorem)

Theorem (work identity)

Consider a system with Hamiltonian H(x;λ)H(x;\lambda) depending on an external control parameter λ\lambda. Assume the system starts in equilibrium at inverse temperature β\beta with parameter value λ0\lambda_0, i.e. with initial distribution given by the .

Drive the system with an arbitrary protocol λt\lambda_t from t=0t=0 to t=τt=\tau, producing a random work value WW along each realization (trajectory). Then the Jarzynski equality states

eβW=eβΔF, \left\langle e^{-\beta W}\right\rangle = e^{-\beta \Delta F},

where ΔF=F(λτ)F(λ0)\Delta F = F(\lambda_\tau)-F(\lambda_0) is the equilibrium Helmholtz free-energy difference (see and ).

This statement is the theorem-level form of .

Work along a protocol (common classical convention)

For Hamiltonian dynamics, the (inclusive) work performed on the system is often written as

W=0τλ˙tλH(xt;λt)dt, W = \int_0^\tau \dot{\lambda}_t\,\partial_\lambda H(x_t;\lambda_t)\,dt,

where xtx_t is the phase-space trajectory generated by the driven dynamics.

The distribution of WW is the .

Immediate corollary (second-law inequality)

By Jensen’s inequality applied to eβW\langle e^{-\beta W}\rangle,

WΔF, \langle W\rangle \ge \Delta F,

consistent with the in isothermal processes (temperature as in ).

Conditions (typical sufficient assumptions)

Different derivations use different dynamics, but common sufficient conditions include:

  • Equilibrium start: initial state is Gibbs/canonical at λ0\lambda_0.
  • Microreversibility: underlying dynamics satisfy time-reversal invariance (Hamiltonian case) or, for stochastic dynamics, are consistent with with respect to the instantaneous equilibrium family.
  • Well-defined work functional: depends on a clear convention for what counts as work vs. heat.

Connections

  • The Jarzynski identity is closely related to and implies it under standard assumptions.
  • It provides a practical route to compute from nonequilibrium experiments/simulations via exponential averaging.