---
title: Third-Order Negativity in Quantum Entanglement
url: https://www.emergentmind.com/topics/third-order-negativity
type: topic
---

# Third-Order Negativity in Quantum Entanglement

Third-order negativity denotes a class of third-moment diagnostics built from a partial transpose and used to characterize mixed-state entanglement. In the recent literature, two closely related constructions are prominent. The first is the third Rényi negativity,
\[
R_3 = -\log \left( \frac{\mathrm{Tr}\left[ (\rho^{T_A})^3 \right]}{\mathrm{Tr} \rho^3} \right),
\]
used as a proxy of mixed-state entanglement in finite-temperature many-body systems. The second is the tripartite invariant
\[
I_5 = \operatorname{Tr}\left[ (\rho_{BC}^{\Gamma})^3 \right],
\]
a permutation-invariant polynomial function of a reduced density matrix that serves as a separability criterion in multipartite settings. Both constructions are based on the third moment of a partially transposed density operator, but they are deployed for different tasks: diagnosing mixed-state entanglement structure in extended systems, resolving disorder-averaged negativity spectra, and providing necessary and sufficient criteria for full separability in tripartite pure and mixed states [1912.03313] [2605.02097].

## 1. Definitions and formal structure

The standard entanglement negativity for a bipartite density matrix $\rho$ on $\mathcal{H}_A \otimes \mathcal{H}_B$ is
\[
E_N = \log \| \rho^{T_A}\|_1,
\]
with $\rho^{T_A}$ the partial transpose and $\|\cdot\|_1$ the trace norm. Rényi negativities generalize this through moments of the partial transpose,
\[
R_n = -\log \left( \frac{\mathrm{Tr}\left[ (\rho^{T_A})^n \right]}{\mathrm{Tr} \rho^n} \right),
\]
and the third Rényi negativity is the $n=3$ case. For pure states, $R_n$ is related to the usual Rényi entanglement entropy; for even $n$, analytic continuation as $n\to 1$ provides the ordinary negativity. Because $n=3$ is odd, $R_3$ is typically used as a computable proxy rather than as the analytic continuation itself [1912.03313].

In multipartite quantum information, the third-order negativity is formulated as
\[
I_5 = \operatorname{Tr}\left[ (\rho_{BC}^{\Gamma})^3 \right],
\]
for a three-qubit or three-qudit pure state $|\Psi\rangle_{ABC}$ with $\rho_{BC} = \operatorname{Tr}_A |\Psi\rangle\langle\Psi|$. Using the replica trick, the same invariant can be written as
\[
I_5 = \langle \Psi^{\otimes 3} | \Omega_A \Omega_B \Omega_C | \Psi^{\otimes 3} \rangle,
\]
where the $\Omega_X$ are permutation operators acting on replicas of subsystem $X$ [2605.02097].

A third line of usage appears in the negativity spectrum of disordered systems, where one studies
\[
M_\alpha^{T_2} = \mathrm{tr} \left(\rho_A^{T_2}\right)^\alpha.
\]
The third negativity moment is then $M_3^{T_2}$, and its scaling encodes information about the full spectrum of the partially transposed reduced density matrix rather than only the logarithmic negativity [1910.09571].

| Quantity | Definition | Primary setting |
|---|---|---|
| Third Rényi negativity | $R_3 = -\log \left( \frac{\mathrm{Tr}\left[ (\rho^{T_A})^3 \right]}{\mathrm{Tr} \rho^3} \right)$ | Bipartite mixed states at finite temperature |
| Third-order negativity | $I_5 = \operatorname{Tr}\left[ (\rho_{BC}^{\Gamma})^3 \right]$ | Tripartite separability of pure and mixed states |
| Third negativity moment | $M_3^{T_2} = \mathrm{tr} \left(\rho_A^{T_2}\right)^3$ | Negativity spectrum in disordered systems |

## 2. Finite-temperature many-body entanglement and the third Rényi negativity

In the two-dimensional transverse field Ising model, the third Rényi negativity was computed with quantum Monte Carlo simulations across the finite-temperature phase transition. For local models, the Rényi negativity obeys an area law,
\[
R_3 = \alpha(T) L - \gamma + \mathcal{O}(1/L),
\]
where $L$ is the boundary length between subsystems and $\alpha(T)$ is the area-law coefficient. The principal numerical result is that the area-law coefficient is singular across the transition, specifically at $T = 3 T_c$ for the third Rényi negativity because of the replica construction [1912.03313].

The singularity is visible in the temperature derivative of the boundary-density contribution. The quantity $d(R_3/|\partial A|)/dT$ displays a cusp at $T = 3 T_c$ and scales as $\log L$ at criticality, mirroring the scaling of the specific heat in the two-dimensional Ising universality class with $\alpha = 0$ and $\nu = 1$. This identifies a sharply localized entanglement response at the entangling boundary, even though the thermal transition itself has a divergent correlation length [1912.03313].

The subleading constant $\gamma$ was isolated using a Levin-Wen-type subtraction scheme,
\[
\gamma = -\frac{1}{2} [ R_3(S_2) - R_3(S_1) - R_3(S_3) + R_3(S_4) ].
\]
Quantum Monte Carlo results found $\gamma$ to be zero within statistical error for all studied system sizes and across the phase transition temperature. Each individual $R_3(S_i)$ is singular at the transition, but the subtraction yields no net long-range or non-local quantum entanglement contribution. The corresponding interpretation in the paper is explicit: entanglement at the finite-temperature critical point is short-ranged, despite the divergent classical correlation length [1912.03313].

## 3. Solvable models, universality, and the absence of long-range entanglement

The same qualitative picture appears in several exactly solvable models. In the quantum spherical model and the two-dimensional Gaussian (free boson) model, the temperature derivative of the Rényi negativity is singular at $T = n T_c$, in agreement with the finite-temperature Monte Carlo results for the non-integrable two-dimensional transverse field Ising model. In both solvable and non-integrable cases, the long-range part $\gamma$ remains strictly zero in the thermodynamic limit, which the paper identifies as a universal feature of finite-temperature phase transitions in these systems [1912.03313].

In the solvable Gaussian and mean-field models, the subleading term vanishes exponentially with system size even at the critical point,
\[
\gamma \sim e^{-L/\xi_Q},
\]
where $\xi_Q$ is a finite quantum correlation length unrelated to the diverging classical correlation length. This supports a separation between classical criticality and the range of quantum entanglement in Gibbs states. A plausible implication, stated in the paper’s interpretive discussion, is that the Gibbs state near or at the transition can be decomposed into minimally entangled typical thermal states (METTS), that is, into a mixture of area-law pure states [1912.03313].

This framework also addresses a common misconception: long-range correlations at a thermal critical point do not, by themselves, imply long-range quantum entanglement. The singularity in the area-law coefficient is boundary-local, whereas the non-local term vanishes. The paper further argues that earlier linked-cluster claims regarding the absence of area-law singularity are likely artifacts of small system sizes, since the larger-scale quantum Monte Carlo simulations observe the singularity unambiguously [1912.03313].

## 4. Third negativity moments in the random singlet phase

In the random singlet phase, the third moment of the partially transposed reduced density matrix is part of the negativity spectrum. For a single disorder realization, the moments depend on the numbers of singlets crossing subsystem boundaries, and for $\alpha=3$ one obtains
\[
M_3^{T_2} = 2^{-2(n_{A:B} + n_{A_1: A_2})},
\]
with $n_{A:B}$ the number of singlets shared between $A$ and $B$, and $n_{A_1:A_2}$ the singlets between $A_1$ and $A_2$ [1910.09571].

The disordered problem distinguishes two inequivalent averages:
\[
\hat{\mathcal{E}}_\alpha = \overline{\log M_\alpha^{T_2}}, \qquad \mathcal{E}_\alpha = \log \overline{M_\alpha^{T_2}}.
\]
For adjacent intervals of size $\ell$, the average of the log for the third moment obeys
\[
\hat{\mathcal{E}}_3^o = -\log 2 \cdot \log \ell,
\]
whereas the log of the average obeys
\[
\mathcal{E}_3^o = 3 \frac{\sqrt{6}-3}{4}\log \ell \approx -0.413 \log \ell.
\]
The two coefficients are therefore distinct. This is one of the paper’s central results: negativity and logarithmic negativity are not trivially related after the average over the disorder, and higher negativity moments reveal finer information about the spectrum than the standard negativity alone [1910.09571].

The scaling coefficients are universal in the sense used by the random-singlet literature: they do not depend on microscopic disorder details once the system flows to the infinite-randomness fixed point. Analytic predictions from strong disorder renormalization group were checked against SDRG numerics and exact computations for the random XX chain, with excellent agreement for both $\hat{\mathcal{E}}_3^o$ and $\mathcal{E}_3^o$. The paper characterizes the inequivalence between different disorder averages as genuine multifractal, or “multiscaling,” behavior [1910.09571].

## 5. Tripartite separability and the invariant \(I_5\)

For normalized tripartite pure states, the principal separability statement is exact:
\[
I_5 = 1 \quad \Longleftrightarrow \quad |\Psi\rangle \text{ is fully separable (product state)}.
\]
This is the main proposition in the 2026 work on multipartite measures. If $I_5=1$, the state is a product across all three parties $A|B|C$; if the state is fully separable, then direct computation gives $I_5=1$. The proof uses properties of the partial transpose and the spectrum of $\rho_{BC}^{\Gamma}$, showing that only product states saturate the maximal value [2605.02097].

For mixed states, the paper introduces convex-roof extensions. One measure is based on
\[
E_{\phi_{ABC}}(\rho) = \inf_{\{p_\ell, |\psi_\ell\rangle\} \in \mathbb{D}(\rho)} \sum_\ell p_\ell\, \phi_{ABC}(\psi_\ell),
\]
with
\[
\phi_{ABC} = 12(1 - I_5) + 27(\tau_{A|B} + \tau_{A|C} + \tau_{B|C}) + \frac{81}{2}\tau_{ABC}.
\]
A second measure is
\[
E_{1 - I_5}(\rho) = \inf_{\{p_\ell, |\psi_\ell\rangle\} \in \mathbb{D}(\rho)} \sum_\ell p_\ell\, [1 - I_5(\psi_\ell)]^{2/3},
\]
where the $2/3$ exponent ensures monotonicity as an entanglement measure. The vanishing criteria are again exact:
\[
E_{\phi_{ABC}}(\rho) = 0 \quad \Longleftrightarrow \quad \rho \text{ is fully separable},
\]
and
\[
E_{1 - I_5}(\rho) = 0 \quad \Longleftrightarrow \quad \rho \text{ is fully separable}.
\]
Within the paper’s framing, third-order negativity thus extends the role played by bipartite negativity and the PPT criterion into a compact, permutation-invariant criterion for full separability in three-party systems [2605.02097].

This separates two notions that are often conflated. Bipartite negativity is a criterion on a chosen cut; third-order negativity in the $I_5$ sense is a multipartite criterion for full product structure. The latter is designed to bypass the combinatorial complexity of testing all bipartitions individually in multipartite systems [2605.02097].

## 6. Multipartite generalization, four-qubit structure, and conformal field theory

The replica-based formulation of $I_5$ generalizes to a hierarchy of multipartite invariants. For an $N$-partite system, the paper constructs
\[
\mathcal{Z}_{\Omega_1, ..., \Omega_q}^{(q)}(\Psi) = \langle \Psi^{\otimes N} | \Omega_1 \cdots \Omega_q | \Psi^{\otimes N} \rangle,
\]
where the $\Omega_i$ are distinct permutation operators on the replicas associated with each subsystem. The general proposition is
\[
|\mathcal{Z}_{\Omega_1,...,\Omega_q}^{(q)}(\Psi)| = 1 \quad \Longleftrightarrow \quad \Psi \text{ is fully product}.
\]
The paper states that this yields a permutation-symmetric necessary and sufficient criterion for full separability in arbitrary multipartite systems, including arbitrary qudit dimensions, and that a convex-roof extension preserves its role as an entanglement monotone and separability criterion for mixed states [2605.02097].

A concrete organizational example is the four-qubit pure state, which has 18 independent parameters and a corresponding set of 18 entanglement measures: 6 two-tangles, 8 tripartite measures, and 4 quadripartite measures. Among the tripartite measures are four $\phi$-type quantities derived from $I_5$, one for each three-party reduction. Their vanishing is equivalent to full separability of the corresponding three-party mixed state. If all four vanish, only the four-tangle may remain nonzero, corresponding to generalized GHZ-type states [2605.02097].

The same paper also places third-order negativity in quantum field theory. In CFT$_2$, the third-order negativity corresponds to a three-point function of twist operators, and in the AdS/CFT correspondence this three-point function can be evaluated using geometric data and OPE coefficients. The paper further states that the universal value for adjacent intervals in CFT$_2$, at large central charge, can be determined. This links third-order negativity to field-theoretic observables rather than treating it solely as an abstract invariant [2605.02097].

Source: https://www.emergentmind.com/topics/third-order-negativity