Thermal States and Gibbs Distributions

Position in Knowledge Graph

System: Quantum Foundations Structural Domain: Quantum Thermodynamics Node: 9.1

This is the foundation node of the quantum thermodynamics module. The Gibbs state is the unique density matrix that maximizes von Neumann entropy subject to a fixed expected energy. It is the equilibrium state in quantum statistical mechanics, and it is the structural object on which all subsequent nodes — work, heat, fluctuation theorems, Landauer's principle — are built.

Formal Definition

For a quantum system with Hamiltonian

H

at inverse temperature

\beta = 1/(k_B T)

, the thermal state (or Gibbs state) is

\rho_\beta = \frac{e^{-\beta H}}{Z(\beta)}, \qquad Z(\beta) = \text{tr}(e^{-\beta H}).

The normalization

Z(\beta)

is the partition function. In the energy eigenbasis

\{|n\rangle\}

with eigenvalues

E_n

, the Gibbs state is diagonal with populations

p_n = \frac{e^{-\beta E_n}}{Z(\beta)}.

The thermodynamic free energy is

F(\beta) = -\frac{1}{\beta} \log Z(\beta),

and entropy, internal energy, and other thermodynamic quantities follow by standard derivatives.

Variational Characterization

The Gibbs state is the unique density matrix that maximizes the von Neumann entropy subject to a constraint on the expected energy:

\rho_\beta = \arg\max_{\rho : \text{tr}(\rho H) = U} S(\rho).

The Lagrange multiplier of the energy constraint is exactly

\beta

. Equivalently, it minimizes the free energy functional

F(\rho) = \text{tr}(\rho H) - T \, S(\rho)

. Both characterizations are quantum analogues of the Boltzmann maximum-entropy and Helmholtz minimum-free-energy principles.

Intuition

The Gibbs state is "what you get when a quantum system equilibrates with a thermal bath at temperature

T

." It has two structural properties that distinguish it from other density matrices:

Diagonal in the energy eigenbasis. Energy is conserved, so coherences between eigenstates of $H$ are unstable under thermal fluctuations and decay.
Boltzmann-weighted populations. Energetically lower states are exponentially more likely. The exponential structure is the structural fingerprint of thermal equilibrium.

The Gibbs state is also the fixed point of any Lindblad dynamics that satisfies detailed balance with respect to

H

at temperature

T

. This connects quantum thermodynamics to the open-system formalism of Module 8.

Structural Role

The Gibbs state organizes the entire module:

Quantum work and heat (9.2) are defined as energy changes during processes that take a system from one Gibbs state to another (or that drive it out of equilibrium).
Fluctuation theorems (9.3) constrain the statistics of work in non-equilibrium processes.
Landauer's principle (9.4) bounds the thermodynamic cost of erasing information in terms of $k_B T \log 2$ .
Quantum heat engines (9.5) are cyclic processes between Gibbs states at different temperatures.
Resource theories (9.6) take Gibbs states as the "free" resource and quantify deviations from equilibrium as the costly resource.

Mathematical Development

High and low temperature limits

$T \to 0$ ( $\beta \to \infty$ ): the Gibbs state collapses onto the ground state (or ground subspace, if degenerate). Thermal fluctuations are suppressed.
$T \to \infty$ ( $\beta \to 0$ ): the Gibbs state becomes maximally mixed, $\rho_\beta \to I/d$ . All energy levels are equally populated.

In between, the populations interpolate exponentially.

Connection to entropy

Substituting the Gibbs state into the von Neumann entropy formula:

S(\rho_\beta) = -\sum_n p_n \log p_n = \beta U + \log Z = \beta(U - F).

This is exactly the thermodynamic relation

S = (U - F)/T

, now derived from the underlying quantum-mechanical formalism rather than postulated.

Quantum partition function

For a single qubit with

H = \frac{1}{2}\omega Z

Z(\beta) = e^{-\beta\omega/2} + e^{\beta\omega/2} = 2\cosh(\beta\omega/2),

\langle E \rangle = -\frac{\partial \log Z}{\partial \beta} = -\frac{\omega}{2}\tanh(\beta\omega/2).

In the high-

T

limit

\langle E \rangle \to 0

(qubit unbiased); in the low-

T

limit

\langle E \rangle \to -\omega/2

(qubit in ground state).

Worked Example

Two-level atom in a thermal bath

Consider a two-level atom with energies

0

and

\hbar\omega

. At temperature

T

, the populations are

p_0 = \frac{1}{1 + e^{-\beta\hbar\omega}}, \qquad p_1 = \frac{e^{-\beta\hbar\omega}}{1 + e^{-\beta\hbar\omega}}.

This is the standard Boltzmann distribution. As

T

increases,

p_1

approaches

1/2

; as

T \to 0

p_1 \to 0

. The von Neumann entropy is the binary entropy

H_2(p_1)

, which is exactly the classical Shannon entropy of the population distribution. This is no coincidence: when a state is diagonal in a fixed basis, von Neumann entropy reduces to Shannon entropy.

Cross-Links

Module 5.1: Density matrices — Gibbs states are a particularly important family.
Module 7.4: Von Neumann entropy is the Lagrange-dual quantity that the Gibbs state maximizes.
Module 8: Open-system dynamics with detailed balance converge to the Gibbs state.
Statistics: Maximum entropy principle in classical statistics; the Gibbs state is the quantum analogue.
VQA: Variational quantum eigensolvers minimize $\langle H \rangle$ , which is the $T \to 0$ limit of free-energy minimization.

Structural Compression

Invariant: The Gibbs state is the maximum-entropy state at fixed expected energy.

Mechanism:

\rho_\beta = e^{-\beta H}/Z

— the unique density matrix with this variational property.

Cross-System Echo: Boltzmann distribution in classical statistical mechanics; maximum-entropy distributions in classical information theory. Same structural identity, non-commutative generalization.

Exercises

Derive the Gibbs state by maximizing $S(\rho)$ subject to $\text{tr}(\rho) = 1$ and $\text{tr}(\rho H) = U$ , using Lagrange multipliers.
Compute the partition function and average energy for a harmonic oscillator with $H = \hbar\omega(a^\dagger a + 1/2)$ .
Show that the Gibbs state is the unique fixed point of any Lindblad evolution that satisfies detailed balance with respect to $H$ at temperature $T$ .
Verify that for a diagonal density matrix, von Neumann entropy reduces to Shannon entropy of the population distribution.
Show that the free energy $F = U - TS$ is minimized over all density matrices by the Gibbs state, and that the minimum value equals $-k_B T \log Z$ .