Constructing free energies from partition functions
Partition functions are normalization constants for equilibrium ensembles, but their deeper role is that their logarithms produce the thermodynamic potentials governing macroscopic behavior (see statistical free energy and thermodynamic equilibrium ).
Canonical free energy from the canonical partition function
In the canonical ensemble (fixed ), the partition function is constructed by the canonical partition-function construction .
The Helmholtz free energy (as a function of ) is defined by
with the inverse temperature . Equivalently, using explicitly and the Boltzmann constant ,
This agrees (in the thermodynamic limit and under standard regularity assumptions) with the thermodynamic Helmholtz free energy .
Grand potential from the grand partition function
In the grand canonical ensemble (fixed ), the grand partition function is constructed by the grand-canonical partition-function construction .
The corresponding thermodynamic potential is the grand potential
matching the thermodynamic grand potential .
Why the logarithm?
The use of (and ) reflects extensivity and additivity:
- For (approximately) independent subsystems, partition functions multiply, while free energies add: if , then .
- In the thermodynamic limit , one expects to scale like volume, so scales like an extensive quantity.
Connection to Legendre structure
Switching which variables are held fixed corresponds to Legendre transforms between potentials. For example, the grand potential is the Legendre transform of with respect to particle number, as explained in the construction from $F$ to $\Omega$ . This mirrors the conjugacy between and the chemical potential .
Physical interpretation
- measures the “useful work” available at fixed once entropic effects are included; minimizing at fixed characterizes equilibrium.
- is the potential adapted to particle exchange; minimizing at fixed characterizes equilibrium in the presence of a particle reservoir.
- Derivatives of and generate observable averages and fluctuations; see constructing observables from log partition functions .