Skip to content

Finite-Temperature and Excited-State Entanglement

Finite-temperature and excited-state entropies mix quantum entanglement with thermal and classical correlations. Clean statements compare matched states or use relative and excess quantities; thermal entropy density, entanglement entropy, and eigenstate entanglement coincide only in specific limits.

Required background. Use regulated subregion entropy. Helpful background. Mutual Information and Regulator-Independent Correlations and Entanglement Negativity in QFT separate total and quantum correlations.

For a state ρ\rho relative to a reference σ\sigma, define a matched excess entropy

ΔSA=S(ρA)S(σA).\Delta S_A=S(\rho_A)-S(\sigma_A).

Ultraviolet terms cancel when the states share the same local short-distance structure and regulator. Relative entropy sharpens the comparison:

D(ρAσA)=ΔKσ,AΔSA0.D(\rho_A\Vert\sigma_A) =\Delta\langle K_{\sigma,A}\rangle-\Delta S_A\geq0.

For a small perturbation, the first law δSA=δKσ,A\delta S_A=\delta\langle K_{\sigma,A}\rangle is the linear term, while relative entropy begins at quadratic order.

The structure figure separates state-dependent entropy from mixed-state and correlation measures.

Thermal and excited-state entanglement begin with a fixed region, algebra, state, and regulator, after which mutual information or negativity can separate different correlation content.

A thermal reduced entropy contains both boundary entanglement and extensive thermal entropy. Mutual information and negativity probe different surviving correlations, so the measure must be chosen before interpreting a temperature crossover. Schematic.

In a two-dimensional CFT at inverse temperature β\beta,

SA(β)=c3log ⁣[βπϵsinh ⁣(πβ)]+c1.S_A(\beta)=\frac{c}{3} \log\!\left[ \frac{\beta}{\pi\epsilon} \sinh\!\left(\frac{\pi\ell}{\beta}\right) \right]+c_1.

For β\ell\ll\beta, this approaches vacuum entanglement plus a small thermal correction. For β\ell\gg\beta, it contains the thermal entropy density times \ell. The latter is not purely bipartite quantum entanglement.

The conformal-map derivation and both limits are given by Calabrese and Cardy 2004, § 3.

For a one-particle excitation in a free field, compare SAS_A, ΔSA\Delta S_A, relative entropy, and mutual information at fixed volume and cutoff. Finite volume matters because a delocalized particle has a subsystem occupation probability that depends on A/VA/V. Degeneracies and coherent superpositions can change the answer.

A high-energy eigenstate may have subsystem entropies close to a thermal ensemble under eigenstate-thermalization assumptions, but this is not a theorem for every QFT or every subsystem fraction. Integrable theories, conserved charges, scars, and finite-size effects require different ensembles or can violate the approximation.

Negativity can retain quantum entanglement after thermal entropy makes SAS_A extensive, but adjacent-region ultraviolet terms and finite-temperature replica continuations must still be controlled.

The validity map lists the choices that must not drift between vacuum, thermal, and excited calculations.

Thermal and excited-state comparisons require the same algebra, geometry, cutoff, volume, Hamiltonian, and ensemble; unmatched subtraction or assuming thermalization can misidentify entropy.

Excess and relative quantities cancel ultraviolet structure only for matched prescriptions. Changing volume, ensemble, energy window, boundary conditions, or regulator can leave a spurious contribution; ETH is an additional state-class assumption. Schematic.

Report temperature or energy, ensemble, total volume, subsystem fraction, boundary conditions, cutoff, and reference state. Separate the short-region, crossover, and volume-law regimes.

  • Calabrese, Pasquale, and John Cardy. “Entanglement Entropy and Quantum Field Theory.” Journal of Statistical Mechanics (2004): P06002. DOI.
  • Alcaraz, Francisco C., Miguel Ibáñez Berganza, and Germán Sierra. “Entanglement of Low-Energy Excitations in Conformal Field Theory.” Physical Review Letters 106 (2011): 201601. DOI. Open preprint.