Thermal Concurrence in Quantum Spin Systems

• Thermal concurrence is a measure of entanglement in thermal quantum states, defined via the Wootters formula and applied to both isolated two-qubit systems and reduced states from many-body ensembles.
• It serves as a diagnostic tool for quantum phase transitions, magnetization plateaus, and crossover behavior in models like the Heisenberg and Ising chains.
• Engineered platforms such as superconducting qubits and semiconductor quantum dots allow tunable control of thermal concurrence through temperature, coupling strength, and detuning.

Thermal concurrence is the concurrence of a thermal quantum state, or of a two-site reduced state obtained from a many-body thermal or pseudo-thermal ensemble. In the standard Gibbs construction,

$$\rho(T)=\frac{e^{-\beta H}}{Z},\qquad Z=\operatorname{Tr}[e^{-\beta H}],\qquad \beta=\frac{1}{k_B T},$$

and thermal concurrence evaluates the Wootters entanglement of $\rho(T)$ itself for a two-qubit system, or of $$\rho_{ij}(T)=\operatorname{Tr}_{\text{rest}}\rho(T)$$ for a pair embedded in a larger lattice. Across the literature, it functions both as a finite-temperature entanglement measure and as a diagnostic for collective phenomena such as quantum phase transitions, magnetization plateaus, and crossover behavior induced by disorder, with its practical usefulness depending strongly on Hamiltonian structure, symmetry, and the choice between full thermal and low-lying pseudo-thermal state constructions (Biswas et al., 2022, Saleem et al., 16 Jun 2026, Ayoub et al., 2022).

1. Definition, reduced states, and standard formulas

For a two-qubit mixed state $\rho$, thermal or otherwise, concurrence is defined by the Wootters formula

$$C(\rho)=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\},$$

where $\lambda_i$ are the square roots of the eigenvalues of $\rho\tilde{\rho}$ in nonincreasing order, and

$$\tilde{\rho}=(\sigma_y\otimes\sigma_y)\rho^*(\sigma_y\otimes\sigma_y).$$

In many-body applications, one first constructs the global thermal state and then reduces it to a two-spin density matrix by tracing out all other degrees of freedom. This is the operative definition in frustrated Heisenberg chains, transverse-field Ising chains, kagome-lattice cluster states, and disordered XX chains (Biswas et al., 2022, Hide, 2011).

Two simplifications recur. First, for X-states,

$$\rho=\begin{pmatrix} \rho_{11}&0&0&\rho_{14}\ 0&\rho_{22}&\rho_{23}&0\ 0&\rho_{32}&\rho_{33}&0\ \rho_{41}&0&0&\rho_{44} \end{pmatrix},$$

the concurrence reduces to

$$C_X=2\max\Big\{0,\ |\rho_{14}|-\sqrt{\rho_{22}\rho_{33}},\ |\rho_{23}|-\sqrt{\rho_{11}\rho_{44}}\Big\}.$$

Second, for Bell-diagonal Gibbs states, concurrence depends only on the largest Bell-state population: $\rho(T)$0 These two cases underlie most exact closed-form treatments of thermal concurrence (Saleem et al., 16 Jun 2026, Ayoub et al., 2022).

The state construction need not be the full canonical ensemble. In the many-body spin systems studied in "Shared purity and concurrence of a mixture of ground and low-lying excited states as indicators of quantum phase transitions" (Biswas et al., 2022), thermal concurrence is also evaluated on pseudo-thermal mixtures: truncated Boltzmann-weighted sums over only the lowest few energy levels, including a ground-plus-first-excited construction and an exponentially weighted low-lying ansatz. This distinction is consequential, because the concurrence of the full thermal state can fail as a phase-transition indicator even when a truncated low-energy mixture succeeds.

2. Exact two-qubit thermal concurrence and temperature response

For the general bilinear two-qubit interaction

$\rho(T)$1

local rotations reduce the Hamiltonian to the canonical anisotropic form

$\rho(T)$2

with $\rho(T)$3. In this representation the Gibbs state is Bell-diagonal, with Bell-basis populations

$\rho(T)$4

and partition function

$\rho(T)$5

The lowest eigenvalue is $\rho(T)$6, so the entangled branch is

$\rho(T)$7

whenever $\rho(T)$8. The threshold inverse temperature $\rho(T)$9 is defined implicitly by

$$\rho_{ij}(T)=\operatorname{Tr}_{\text{rest}}\rho(T)$$0

This yields an exact entanglement-sudden-birth condition for interacting two-qubit Gibbs states (Saleem et al., 16 Jun 2026).

The same analysis provides exact derivatives of thermal concurrence with respect to inverse temperature. In the entangled branch,

$$\rho_{ij}(T)=\operatorname{Tr}_{\text{rest}}\rho(T)$$1

and

$$\rho_{ij}(T)=\operatorname{Tr}_{\text{rest}}\rho(T)$$2

The thermal quantum Fisher information is

$$\rho_{ij}(T)=\operatorname{Tr}_{\text{rest}}\rho(T)$$3

which gives the bounds

$$\rho_{ij}(T)=\operatorname{Tr}_{\text{rest}}\rho(T)$$4

and

$$\rho_{ij}(T)=\operatorname{Tr}_{\text{rest}}\rho(T)$$5

Temperature uncertainty likewise induces an entanglement loss bounded by thermal quantum Fisher information: $$\rho_{ij}(T)=\operatorname{Tr}_{\text{rest}}\rho(T)$$6 or, equivalently,

$$\rho_{ij}(T)=\operatorname{Tr}_{\text{rest}}\rho(T)$$7

In this exact setting, thermal concurrence is not only a static measure but a temperature-response observable constrained by equilibrium energy fluctuations (Saleem et al., 16 Jun 2026).

A standard example is the antiferromagnetic Heisenberg XXX model with $$\rho_{ij}(T)=\operatorname{Tr}_{\text{rest}}\rho(T)$$8. Its Gibbs spectrum consists of one ground Bell state with energy $$\rho_{ij}(T)=\operatorname{Tr}_{\text{rest}}\rho(T)$$9 and three degenerate excited states with energy $\rho$0, so

$\rho$1

The concurrence is then

$\rho$2

and vanishes for $\rho$3. The critical temperature is

$\rho$4

3. Thermal concurrence as a probe of quantum criticality

In extended spin systems, thermal concurrence is typically evaluated on nearest-neighbor reduced states extracted from exact-diagonalization spectra. Its behavior is strongly model dependent. In the one-dimensional $\rho$5–$\rho$6 Heisenberg chain,

$\rho$7

with frustration parameter $\rho$8, the ground-state transition occurs at $\rho$9. For the full thermal state at low temperature $$C(\rho)=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\},$$0, nearest-neighbor concurrence fails to detect this quantum phase transition; no nonanalyticity or sharp feature identifies $$C(\rho)=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\},$$1. A pseudo-thermal truncation including levels up to the third excited state also fails. By contrast, the ground-plus-first-excited pseudo-thermal mixture produces clear discontinuities in both concurrence and shared purity, with finite-size critical points converging to $$C(\rho)=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\},$$2 (Biswas et al., 2022).

The finite-size scaling in that model is unusual. The deviations obey

$$C(\rho)=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\},$$3

with distinct exponents for near and far discontinuities and for odd and even system sizes: $$C(\rho)=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\},$$4 The same odd-even dichotomy persists for exponentially weighted low-lying mixtures. The paper treats a Möbius strip-like boundary identification for odd $$C(\rho)=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\},$$5 as a plausible explanation, whereas even $$C(\rho)=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\},$$6 is compatible with the usual periodic boundary condition (Biswas et al., 2022).

In the one-dimensional transverse-field Ising model,

$$C(\rho)=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\},$$7

with critical point $$C(\rho)=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\},$$8, concurrence performs better. For the ground-plus-first-excited pseudo-thermal mixture, extrema of $$C(\rho)=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\},$$9 identify finite-size transition points

$\lambda_i$0

with scaling

$\lambda_i$1

For the exponentially weighted mixture, the corresponding estimates are

$\lambda_i$2

with exponent $\lambda_i$3. No odd-even dichotomy appears, consistent with the absence of next-nearest-neighbor frustration (Biswas et al., 2022).

The two-dimensional $\lambda_i$4–$\lambda_i$5 Heisenberg model on a $\lambda_i$6 square lattice shows a more stringent limitation. Two phase transitions are expected, near $\lambda_i$7 and $\lambda_i$8. In both the ground-plus-first pseudo-thermal mixture and the exponential mixture, concurrence detects the first transition via a sharp drop, with $\lambda_i$9 for the exponential mixture. Beyond $\rho\tilde{\rho}$0, however, nearest-neighbor concurrence vanishes and therefore fails to detect the second transition. Shared purity remains nonzero, changes curvature from convex to concave, and yields $\rho\tilde{\rho}$1 or $\rho\tilde{\rho}$2, depending on the low-lying mixture used. A common misconception is that concurrence is generically an equally effective phase-transition indicator as other bipartite quantum-correlation measures; in this model, it is not (Biswas et al., 2022).

4. Frustrated kagome magnetism and thermal concurrence plateaus

In the kagome-lattice model for the third layer of fluid $\rho\tilde{\rho}$3He on graphite, thermal concurrence is studied within a variational mean-field-like treatment based on the Gibbs–Bogoliubov inequality. The exchange Hamiltonian contains two-site and three-site permutations,

$\rho\tilde{\rho}$4

together with a Zeeman term

$\rho\tilde{\rho}$5

Using the kagome geometry, the model is regrouped over triangles and replaced by independent triangle clusters in self-consistent fields. The trial Hamiltonian on each triangle is

$\rho\tilde{\rho}$6

with optimized parameters

$\rho\tilde{\rho}$7

The corresponding three-spin thermal state is diagonalized exactly at the cluster level (Ananikian et al., 2011).

Tracing out one spin yields a nearest-neighbor two-spin X-state,

$\rho\tilde{\rho}$8

for which the concurrence simplifies to

$\rho\tilde{\rho}$9

The effective field entering the triangle spectrum is related to the lattice magnetization through

$$\tilde{\rho}=(\sigma_y\otimes\sigma_y)\rho^*(\sigma_y\otimes\sigma_y).$$0

This makes thermal concurrence a self-consistent observable: it depends on temperature, field, and exchange couplings both through Boltzmann weights and through the mean-field magnetization (Ananikian et al., 2011).

The principal finding is that in the antiferromagnetic region the behavior of concurrence coincides with the behavior of magnetization. For $$\tilde{\rho}=(\sigma_y\otimes\sigma_y)\rho^*(\sigma_y\otimes\sigma_y).$$1, $$\tilde{\rho}=(\sigma_y\otimes\sigma_y)\rho^*(\sigma_y\otimes\sigma_y).$$2, and $$\tilde{\rho}=(\sigma_y\otimes\sigma_y)\rho^*(\sigma_y\otimes\sigma_y).$$3, the model exhibits magnetization plateaus at $$\tilde{\rho}=(\sigma_y\otimes\sigma_y)\rho^*(\sigma_y\otimes\sigma_y).$$4, attributed to stabilized trimer configurations, and the concurrence displays corresponding plateaus. At $$\tilde{\rho}=(\sigma_y\otimes\sigma_y)\rho^*(\sigma_y\otimes\sigma_y).$$5, the recursive mean-field equations have a single stable solution $$\tilde{\rho}=(\sigma_y\otimes\sigma_y)\rho^*(\sigma_y\otimes\sigma_y).$$6 and no plateau structure. At $$\tilde{\rho}=(\sigma_y\otimes\sigma_y)\rho^*(\sigma_y\otimes\sigma_y).$$7, both $$\tilde{\rho}=(\sigma_y\otimes\sigma_y)\rho^*(\sigma_y\otimes\sigma_y).$$8 and concurrence decrease with temperature and vanish at the same characteristic temperature $$\tilde{\rho}=(\sigma_y\otimes\sigma_y)\rho^*(\sigma_y\otimes\sigma_y).$$9. In the strong-field limit, the pair density matrix approaches $$\rho=\begin{pmatrix} \rho_{11}&0&0&\rho_{14}\ 0&\rho_{22}&\rho_{23}&0\ 0&\rho_{32}&\rho_{33}&0\ \rho_{41}&0&0&\rho_{44} \end{pmatrix},$$0 and concurrence vanishes (Ananikian et al., 2011).

These results are model-specific and method-specific. The cluster Hamiltonian is exact only for the trial triangles, while inter-triangle quantum correlations are neglected beyond mean-field products, and six-site exchange interactions are omitted. Accordingly, the coincidence between magnetization plateaus and concurrence plateaus should be interpreted within the variational cluster framework rather than as a general property of kagome antiferromagnets (Ananikian et al., 2011).

5. Disorder-averaged thermal concurrence

In disordered quantum spin systems, thermal concurrence becomes a disorder-dependent quantity. For the XX chain with weak, uncorrelated Gaussian bond disorder,

$$\rho=\begin{pmatrix} \rho_{11}&0&0&\rho_{14}\ 0&\rho_{22}&\rho_{23}&0\ 0&\rho_{32}&\rho_{33}&0\ \rho_{41}&0&0&\rho_{44} \end{pmatrix},$$1

$$\rho=\begin{pmatrix} \rho_{11}&0&0&\rho_{14}\ 0&\rho_{22}&\rho_{23}&0\ 0&\rho_{32}&\rho_{33}&0\ \rho_{41}&0&0&\rho_{44} \end{pmatrix},$$2

$$\rho=\begin{pmatrix} \rho_{11}&0&0&\rho_{14}\ 0&\rho_{22}&\rho_{23}&0\ 0&\rho_{32}&\rho_{33}&0\ \rho_{41}&0&0&\rho_{44} \end{pmatrix},$$3

with $$\rho=\begin{pmatrix} \rho_{11}&0&0&\rho_{14}\ 0&\rho_{22}&\rho_{23}&0\ 0&\rho_{32}&\rho_{33}&0\ \rho_{41}&0&0&\rho_{44} \end{pmatrix},$$4 Gaussian of mean zero and variance $$\rho=\begin{pmatrix} \rho_{11}&0&0&\rho_{14}\ 0&\rho_{22}&\rho_{23}&0\ 0&\rho_{32}&\rho_{33}&0\ \rho_{41}&0&0&\rho_{44} \end{pmatrix},$$5, the reduced two-site thermal density matrix is again of X-type after disorder averaging. The disorder-averaged concurrence can therefore be expressed in terms of disorder-averaged thermal correlators (Hide, 2011).

The paper develops a replica-based thermal many-body perturbation theory to first order in $$\rho=\begin{pmatrix} \rho_{11}&0&0&\rho_{14}\ 0&\rho_{22}&\rho_{23}&0\ 0&\rho_{32}&\rho_{33}&0\ \rho_{41}&0&0&\rho_{44} \end{pmatrix},$$6. Rather than averaging concurrence directly, it computes disorder-averaged correlators and then constructs concurrence from them. For the XX chain, the needed inputs are encoded in the fermionic quantity

$$\rho=\begin{pmatrix} \rho_{11}&0&0&\rho_{14}\ 0&\rho_{22}&\rho_{23}&0\ 0&\rho_{32}&\rho_{33}&0\ \rho_{41}&0&0&\rho_{44} \end{pmatrix},$$7

from which one obtains, for example,

$$\rho=\begin{pmatrix} \rho_{11}&0&0&\rho_{14}\ 0&\rho_{22}&\rho_{23}&0\ 0&\rho_{32}&\rho_{33}&0\ \rho_{41}&0&0&\rho_{44} \end{pmatrix},$$8

and

$$\rho=\begin{pmatrix} \rho_{11}&0&0&\rho_{14}\ 0&\rho_{22}&\rho_{23}&0\ 0&\rho_{32}&\rho_{33}&0\ \rho_{41}&0&0&\rho_{44} \end{pmatrix},$$9

The clean limit is recovered from

$$C_X=2\max\Big\{0,\ |\rho_{14}|-\sqrt{\rho_{22}\rho_{33}},\ |\rho_{23}|-\sqrt{\rho_{11}\rho_{44}}\Big\}.$$0

High temperature drives $$C_X=2\max\Big\{0,\ |\rho_{14}|-\sqrt{\rho_{22}\rho_{33}},\ |\rho_{23}|-\sqrt{\rho_{11}\rho_{44}}\Big\}.$$1 and hence $$C_X=2\max\Big\{0,\ |\rho_{14}|-\sqrt{\rho_{22}\rho_{33}},\ |\rho_{23}|-\sqrt{\rho_{11}\rho_{44}}\Big\}.$$2 (Hide, 2011).

The main physical result is that disorder can enhance concurrence in some parameter regimes. In the quenched average, first-order weak disorder decreases nearest-neighbor concurrence below the quantum critical point at $$C_X=2\max\Big\{0,\ |\rho_{14}|-\sqrt{\rho_{22}\rho_{33}},\ |\rho_{23}|-\sqrt{\rho_{11}\rho_{44}}\Big\}.$$3 when $$C_X=2\max\Big\{0,\ |\rho_{14}|-\sqrt{\rho_{22}\rho_{33}},\ |\rho_{23}|-\sqrt{\rho_{11}\rho_{44}}\Big\}.$$4, and does not create entanglement above the critical point at that order. At finite temperature, however, there is a crossover field $$C_X=2\max\Big\{0,\ |\rho_{14}|-\sqrt{\rho_{22}\rho_{33}},\ |\rho_{23}|-\sqrt{\rho_{11}\rho_{44}}\Big\}.$$5: for $$C_X=2\max\Big\{0,\ |\rho_{14}|-\sqrt{\rho_{22}\rho_{33}},\ |\rho_{23}|-\sqrt{\rho_{11}\rho_{44}}\Big\}.$$6, disorder suppresses concurrence, while for $$C_X=2\max\Big\{0,\ |\rho_{14}|-\sqrt{\rho_{22}\rho_{33}},\ |\rho_{23}|-\sqrt{\rho_{11}\rho_{44}}\Big\}.$$7 but still near $$C_X=2\max\Big\{0,\ |\rho_{14}|-\sqrt{\rho_{22}\rho_{33}},\ |\rho_{23}|-\sqrt{\rho_{11}\rho_{44}}\Big\}.$$8, disorder enhances concurrence relative to the clean chain. The annealed average is qualitatively similar, but with a significantly lower crossover field. Next-nearest-neighbor concurrence is smaller and exists over a narrower parameter window, though its disorder trends are analogous. The paper also reports that random-field disorder produces the same qualitative crossover, with generally stronger enhancement near the quantum critical point than random-bond disorder at the same variance (Hide, 2011).

A common misconception is that “disorder-enhanced entanglement” is universal. In this treatment it is neither universal nor nonperturbative: the calculation is first order in weak disorder, relies on the free-fermion structure of the XX model, and equates disorder-averaged concurrence with concurrence built from disorder-averaged correlators only to the perturbative order retained (Hide, 2011).

6. Engineered solid-state realizations and correlation beyond entanglement

Thermal concurrence is also a directly tunable quantity in engineered two-qubit platforms. For two superconducting qubits coupled through a tunable coupler qubit, the effective dispersive Hamiltonian is

$$C_X=2\max\Big\{0,\ |\rho_{14}|-\sqrt{\rho_{22}\rho_{33}},\ |\rho_{23}|-\sqrt{\rho_{11}\rho_{44}}\Big\}.$$9

with

$\rho(T)$00

Its thermal state has X-form in the computational basis, with partition function

$\rho(T)$01

where $\rho(T)$02, $\rho(T)$03, and $\rho(T)$04. The resulting concurrence is

$\rho(T)$05

and the critical temperature satisfies

$\rho(T)$06

At resonance, $\rho(T)$07, one has

$\rho(T)$08

The paper emphasizes three trends: concurrence decreases monotonically with temperature, increases with coupling strength, and is suppressed by frequency detuning. In particular, maximally entangled thermal states are favored by equal qubit frequencies, strong effective coupling, and low temperature; the tunable coupler provides control through capacitances, inductances via $\rho(T)$09, and detuning $\rho(T)$10 (Ayoub et al., 2022).

A distinct solid-state realization is a single electron in a semiconductor double quantum dot, where the two qubits are charge $\rho(T)$11 and spin $\rho(T)$12. The Hamiltonian is

$\rho(T)$13

with interdot detuning $\rho(T)$14, tunnel coupling $\rho(T)$15, longitudinal Zeeman splitting $\rho(T)$16, and transverse magnetic field gradient $\rho(T)$17. The four eigenvalues are

$\rho(T)$18

where

$\rho(T)$19

The thermal state

$\rho(T)$20

is generically not an X-state, so concurrence must be obtained from the full Wootters spectrum of

$\rho(T)$21

At $\rho(T)$22, the concurrence approaches the pure-ground-state value $\rho(T)$23; at $\rho(T)$24, $\rho(T)$25 and concurrence vanishes. The paper reports entanglement sudden death at a finite critical temperature $\rho(T)$26, maximal concurrence near resonance $\rho(T)$27, enhancement with increasing $\rho(T)$28, reduction with increasing $\rho(T)$29 and $\rho(T)$30, and a non-monotonic dependence on $\rho(T)$31: small $\rho(T)$32 increases entanglement through spin-charge hybridization, whereas sufficiently large $\rho(T)$33 suppresses it (Leitão et al., 2024).

The double-quantum-dot analysis also sharpens the distinction between thermal concurrence and broader quantum-correlation diagnostics. The paper computes correlated coherence,

$\rho(T)$34

in the local basis that diagonalizes $\rho(T)$35 and $\rho(T)$36. It finds that at low temperature correlated coherence equals concurrence, but as temperature increases concurrence decays faster and vanishes at a finite $\rho(T)$37, while correlated coherence persists over a larger temperature window. This reinforces a broader lesson already visible in the frustrated-lattice study of shared purity: thermal concurrence is a stringent and often useful entanglement diagnostic, but it is not exhaustive as a probe of quantum correlations (Leitão et al., 2024, Biswas et al., 2022).