Many-Body Negativity in Mixed-State Quantum Systems

• Many-body negativity is a measure of mixed-state entanglement derived from the partial transpose of reduced density matrices, offering a sharp probe of genuinely quantum correlations.
• It incorporates Rényi, fermionic, and symmetry-resolved extensions to overcome limitations of entanglement entropy in complex, mixed, and finite-temperature states.
• Advanced tensor-network, perturbative, and Monte Carlo methods enable accurate computation of negativity in large many-body systems despite exponential state-space growth.

Searching arXiv for recent and foundational papers on many-body negativity.

Many-body negativity denotes a family of mixed-state entanglement diagnostics built from the partial transpose of reduced density matrices, together with Rényi and fermionic extensions that are tailored to many-body settings in which subsystems are generically mixed rather than pure. In this role, negativity is used to probe genuinely quantum correlations in disjoint regions, open systems, finite-temperature states, localized phases, driven steady states, and Gaussian or fermionic many-body states, often in regimes where entanglement entropy is either inapplicable or does not isolate entanglement from total correlations (Gray, 2018, Gray et al., 2019, Murciano et al., 2022).

1. Core definitions and scope

For a bipartite state $\rho_{AB}$, the basic construction is the partial transpose $\rho_{AB}^{T_B}$. The Peres–Horodecki positive partial transpose criterion states that separability implies $\rho_{AB}^{T_B}$ has only positive eigenvalues. Negativity then quantifies the failure of positivity. In the notation used across the literature surveyed here,

$$\mathcal{N}(\rho_{AB})=\frac{\|\rho_{AB}^{T_B}\|_1-1}{2},$$

equivalently,

$$\mathcal{N}(\rho)\coloneqq \sum_{\lambda<0}\left|\lambda\!\left(\rho^{T_B}\right)\right|.$$

The associated logarithmic negativity is written either as

$E_{\mathcal N}(\rho)=\log_2\|\rho^{T_B}\|_1$

or as

$N_L=\log\left(\left\|\rho^{T_L}\right\|_1\right),$

depending on convention (Cresswell et al., 2018, Wanisch et al., 25 Feb 2025).

In low dimensions, specifically $2\otimes2$ and $2\otimes3$, PPT is necessary and sufficient for separability. In generic many-body settings, however, negativity is primarily used as a computable mixed-state entanglement monotone and as an upper bound on distillable entanglement, rather than as a universal separability criterion (Cresswell et al., 2018, Gray, 2018).

A central reason for its use in many-body physics is that ordinary entanglement entropy is directly tied to pure bipartitions, whereas reduced states of subregions, disjoint intervals, open systems, and finite-temperature states are mixed. For pure states, logarithmic negativity reduces to a Rényi entropy at index $1/2$ (Kudler-Flam et al., 2019).

Quantity	Definition	Role
Negativity	$\rho_{AB}^{T_B}$0	Mixed-state entanglement from partial transpose
Logarithmic negativity	$\rho_{AB}^{T_B}$1	Entanglement monotone; upper bound on distillable entanglement
Third Rényi negativity	$\rho_{AB}^{T_B}$2	Computable proxy in mixed-state dynamics
Fermionic rank-$\rho_{AB}^{T_B}$3 Rényi negativity	$\rho_{AB}^{T_B}$4	Fermionic mixed-state entanglement via fermionic partial transpose

A recurring misconception is that negativity is interchangeable with mutual information. The literature distinguishes them sharply: mutual information measures total correlations, whereas negativity is used to isolate the genuinely quantum part. This distinction is explicit in studies of open XXZ transport, of the many-body localization transition, and of finite-temperature fermionic systems, where mutual information can remain nonzero even when negativity is small or vanishing (Wanisch et al., 25 Feb 2025, Gray et al., 2019, Wang et al., 2023).

2. Rényi, fermionic, and symmetry-resolved extensions

Rényi versions of negativity are introduced because direct evaluation of $\rho_{AB}^{T_B}$5 is often intractable in large many-body systems. In open Wannier–Stark dynamics, the third Rényi negativity

$\rho_{AB}^{T_B}$6

is used as an accessible proxy. For pure states,

$\rho_{AB}^{T_B}$7

while the first and second moments are trivial: $\rho_{AB}^{T_B}$8 The same work emphasizes that $\rho_{AB}^{T_B}$9 is “not an exact entanglement monotone” but “captures the relevant dynamics as it behaves quantitatively similar as the negativity” (Wybo et al., 2021).

For fermions, the ordinary bosonic partial transpose is not the natural construction. A major development is the use of partial time-reversal or fermionic partial transpose. In fermionic Gaussian systems, partial time-reversal preserves Gaussianity and can be computed efficiently, whereas the conventional fermionic partial transpose becomes a linear combination of two Gaussian operators that generally do not commute and fails additivity under tensor products (Shapourian et al., 2016). This distinction is physically consequential: in the Kitaev chain, the conventional definition fails to capture the formation of the edge Majorana fermions, while partial time-reversal computes the quantum dimension of the Majorana fermions (Shapourian et al., 2016).

A further refinement distinguishes untwisted and twisted fermionic partial transposes. The untwisted PTDM is generally pseudo-Hermitian, while the twisted PTDM is Hermitian. Both yield the same logarithmic negativity after the appropriate analytic continuation, but they generally give different Rényi negativities at finite integer rank. In interacting fermionic simulations, the twisted Rényi negativity ratio is reported to obey the area law and to decrease monotonically with temperature, in contrast to the untwisted ratio (Wang et al., 10 Mar 2025).

In free fermions with conserved charge, negativity can also be organized by connected charge correlators. The charge-correlator expansion expresses Rényi negativity and, under additional conditions, logarithmic negativity through cumulants, paralleling full counting statistics expansions of entanglement entropy. The replica-limit identification of logarithmic negativity is found numerically to work only for translationally invariant local Hamiltonians (Tang, 2024).

Symmetry can resolve negativity in a different way. For globally charge-conserving many-body states, partial transposition converts charge conservation into imbalance conservation: $\rho_{AB}^{T_B}$0 Hence $\rho_{AB}^{T_B}$1 is block diagonal in the eigenbasis of the imbalance operator $\rho_{AB}^{T_B}$2, and the negativity decomposes into imbalance sectors rather than total-charge sectors (Cornfeld et al., 2018). This shows that negativity is not merely charge-resolved but imbalance-resolved.

3. Computational frameworks

The principal computational bottleneck is the trace norm of the partially transposed density matrix. Exact evaluation generally requires diagonalizing an exponentially large operator. Several complementary frameworks address this difficulty.

A tensor-network stochastic Lanczos quadrature method treats $\rho_{AB}^{T_B}$3 as a lazily evaluated tensor-network operator and estimates

$\rho_{AB}^{T_B}$4

using Hutchinson trace estimation, Lanczos iteration, and Gauss quadrature. This makes many-body logarithmic negativity feasible without explicitly constructing the full partially transposed density matrix. The method is reported to yield useful accuracy at roughly the $\rho_{AB}^{T_B}$5 level, to handle exact full-state representations up to about $\rho_{AB}^{T_B}$6 qubits, and to reproduce conformal-field-theory behavior in a Heisenberg ground state of total length $\rho_{AB}^{T_B}$7. The same paper notes that naively forming the partially transposed density matrix for $\rho_{AB}^{T_B}$8 qubits split into three $\rho_{AB}^{T_B}$9-qubit parts could require on the order of $$\mathcal{N}(\rho_{AB})=\frac{\|\rho_{AB}^{T_B}\|_1-1}{2},$$0 terabytes of storage (Gray, 2018).

A more recent tensor-network advance targets the trace norm directly. The key idea is to combine Zolotarev’s rational approximation to the sign function with a variational formulation solved using a DMRG-like algorithm. For Hermitian $$\mathcal{N}(\rho_{AB})=\frac{\|\rho_{AB}^{T_B}\|_1-1}{2},$$1,

$$\mathcal{N}(\rho_{AB})=\frac{\|\rho_{AB}^{T_B}\|_1-1}{2},$$2

and the estimator

$$\mathcal{N}(\rho_{AB})=\frac{\|\rho_{AB}^{T_B}\|_1-1}{2},$$3

comes with the bound

$$\mathcal{N}(\rho_{AB})=\frac{\|\rho_{AB}^{T_B}\|_1-1}{2},$$4

This makes trace-norm-based quantities, including negativity, practical tensor-network observables beyond exact diagonalization (Lee et al., 10 Jun 2026).

For open systems, a tree tensor operator representation writes the density operator in manifestly positive form,

$$\mathcal{N}(\rho_{AB})=\frac{\|\rho_{AB}^{T_B}\|_1-1}{2},$$5

so positivity is guaranteed by construction. In the appropriate unitary gauge, bipartite entanglement across the two halves is concentrated in the root tensor, which gives direct access to entanglement monotones such as logarithmic negativity and entanglement of formation in boundary-driven Lindblad dynamics (Wanisch et al., 25 Feb 2025).

A different computational direction is perturbative rather than variational. Patterned matrix calculus yields an all-orders expansion of negativity for arbitrary variations of the density matrix, together with explicit first and second derivatives in terms of the eigendecomposition of $$\mathcal{N}(\rho_{AB})=\frac{\|\rho_{AB}^{T_B}\|_1-1}{2},$$6. This addresses the non-analyticity of negativity as a function of the density matrix and is proposed as a practical framework for entanglement growth, decay, and sudden-death analyses in many-body and open settings (Cresswell et al., 2018).

Quantum Monte Carlo provides a complementary route for interacting fermions. Determinantal QMC represents the interacting density matrix as a weighted sum of Gaussian states,

$$\mathcal{N}(\rho_{AB})=\frac{\|\rho_{AB}^{T_B}\|_1-1}{2},$$7

and the partially transposed density matrix as the corresponding sum of partially transposed Gaussian components. This enables rank-$$\mathcal{N}(\rho_{AB})=\frac{\|\rho_{AB}^{T_B}\|_1-1}{2},$$8 Rényi negativity in the half-filled Hubbard model and the spinless $$\mathcal{N}(\rho_{AB})=\frac{\|\rho_{AB}^{T_B}\|_1-1}{2},$$9-$$\mathcal{N}(\rho)\coloneqq \sum_{\lambda<0}\left|\lambda\!\left(\rho^{T_B}\right)\right|.$$0 model, and, in later work, stable computations of both untwisted and twisted high-rank Rényi negativities through stabilized Green’s-function formulas and incremental algorithms (Wang et al., 2023, Wang et al., 10 Mar 2025).

4. Dynamics, decoherence, and nonequilibrium transport

Many-body negativity has become a central diagnostic for mixed-state dynamics. In a tilted-field XXZ chain, the third Rényi negativity $$\mathcal{N}(\rho)\coloneqq \sum_{\lambda<0}\left|\lambda\!\left(\rho^{T_B}\right)\right|.$$1 is used to track the entanglement dynamics of a Wannier–Stark many-body localized system coupled to a Markovian dephasing bath. For $$\mathcal{N}(\rho)\coloneqq \sum_{\lambda<0}\left|\lambda\!\left(\rho^{T_B}\right)\right|.$$2, $$\mathcal{N}(\rho)\coloneqq \sum_{\lambda<0}\left|\lambda\!\left(\rho^{T_B}\right)\right|.$$3 captures the characteristic logarithmic growth of interacting localized phases in the intermediate-time regime, while for $$\mathcal{N}(\rho)\coloneqq \sum_{\lambda<0}\left|\lambda\!\left(\rho^{T_B}\right)\right|.$$4 it does not show such growth. Under dephasing, the relevant window is

$$\mathcal{N}(\rho)\coloneqq \sum_{\lambda<0}\left|\lambda\!\left(\rho^{T_B}\right)\right|.$$5

which in practice means

$$\mathcal{N}(\rho)\coloneqq \sum_{\lambda<0}\left|\lambda\!\left(\rho^{T_B}\right)\right|.$$6

In that regime, $$\mathcal{N}(\rho)\coloneqq \sum_{\lambda<0}\left|\lambda\!\left(\rho^{T_B}\right)\right|.$$7 still reproduces the onset of logarithmic entanglement growth even though the state is mixed (Wybo et al., 2021).

The same study shows that the open-system dynamics of $$\mathcal{N}(\rho)\coloneqq \sum_{\lambda<0}\left|\lambda\!\left(\rho^{T_B}\right)\right|.$$8 can be fit in an intermediate-time regime by stretched exponentials of the form

$$\mathcal{N}(\rho)\coloneqq \sum_{\lambda<0}\left|\lambda\!\left(\rho^{T_B}\right)\right|.$$9

while for small finite systems the late-time regime crosses over to simple exponential decay. It also identifies a crossover time in the isolated tilted system associated with the quadratic correction in the potential, a feature stated to be absent in disorder-induced MBL (Wybo et al., 2021).

In a boundary-driven open XXZ chain, logarithmic negativity reveals entanglement transitions controlled both by the anisotropy $E_{\mathcal N}(\rho)=\log_2\|\rho^{T_B}\|_1$0 and the boundary coupling $E_{\mathcal N}(\rho)=\log_2\|\rho^{T_B}\|_1$1. Starting from the fully polarized product state and quenching by turning on the environment coupling, the arrival of spin current at the center of the chain coincides with the onset of bipartite entanglement between the two halves. The transport regimes map onto distinct entanglement classes: ballistic transport corresponds to volume-law entanglement, subdiffusive transport to area-law entanglement, and insulating transport to a separable state with vanishing entanglement. The same work reports that spin current and entanglement rise and fall together as $E_{\mathcal N}(\rho)=\log_2\|\rho^{T_B}\|_1$2 is varied (Wanisch et al., 25 Feb 2025).

A different usage of negativity appears in quasiprobability dynamics. The first-time negativity of the Margenau–Hill quasiprobability,

$E_{\mathcal N}(\rho)=\log_2\|\rho^{T_B}\|_1$3

is introduced as a dynamical marker of when local measurement sequences in an interacting many-body system begin to exhibit genuinely nonclassical behavior. In the Ising chain, FTN discriminates between interaction-dominated and field-dominated regimes, is reshaped by temperature, and reveals a finite-time light-cone-like buildup of operator incompatibility when measurements are performed on different sites (Shukla et al., 1 Jan 2026). This is a distinct notion from entanglement negativity, but it extends the broader many-body use of negativity as a witness of nonclassicality.

5. Localization, criticality, and spatial organization

Negativity is a sharp diagnostic of localization phenomena because it tracks entanglement propagation rather than merely local memory. In disordered spin-$E_{\mathcal N}(\rho)=\log_2\|\rho^{T_B}\|_1$4 and spin-1 Heisenberg chains, nearest-neighbor negativity increases across the many-body localization transition, while the geometric entanglement decreases. This opposite behavior is interpreted through entanglement monogamy: localization concentrates entanglement into local pockets, whereas thermal states distribute entanglement broadly across the system. In the localized phase, both concurrence and negativity between sites separated by distance $E_{\mathcal N}(\rho)=\log_2\|\rho^{T_B}\|_1$5 decay exponentially with distance (West et al., 2018).

At the many-body localization transition itself, logarithmic negativity between two disjoint blocks provides a two-length-scale probe of criticality. For equal block size $E_{\mathcal N}(\rho)=\log_2\|\rho^{T_B}\|_1$6 and separation $E_{\mathcal N}(\rho)=\log_2\|\rho^{T_B}\|_1$7, the normalized bond negativity is found to collapse as a function of $E_{\mathcal N}(\rho)=\log_2\|\rho^{T_B}\|_1$8, and at the transition it obeys a scale-invariant exponential decay,

$E_{\mathcal N}(\rho)=\log_2\|\rho^{T_B}\|_1$9

By contrast, the mutual information also collapses but decays with a power law. This difference is used to argue that negativity reveals the genuinely quantum part of the critical correlations and gives direct microscopic evidence for scale invariance at the MBL transition (Gray et al., 2019).

Spatially resolved versions of negativity sharpen this picture. The negativity contour $N_L=\log\left(\left\|\rho^{T_L}\right\|_1\right),$0 is introduced as a quasi-local measure that decomposes logarithmic negativity across real space,

$N_L=\log\left(\left\|\rho^{T_L}\right\|_1\right),$1

For Gaussian fermions, an explicit contour follows from the fermionic partial transpose and is exactly normalized to the total negativity. Although the generic derivative-based contour is not guaranteed to be positive because logarithmic negativity does not satisfy strong subadditivity, positivity is stated to hold rigorously for holographic states and for states obeying the quasi-particle picture (Kudler-Flam et al., 2019).

The contour formalism clarifies the difference between entanglement entropy and negativity at finite temperature and after quenches. In a $N_L=\log\left(\left\|\rho^{T_L}\right\|_1\right),$2-dimensional CFT, the universal part of the negativity contour subtracts the thermal contribution through the characteristic $N_L=\log\left(\left\|\rho^{T_L}\right\|_1\right),$3 term in the coth expression, so it decays away from the entangling cuts rather than supporting a thermal background. In nonequilibrium free-fermion quenches, the entanglement contour approaches a finite thermal profile while the negativity contour relaxes to zero away from the entangling cuts, making the quantum-to-classical crossover spatially explicit (Kudler-Flam et al., 2019).

6. Multipartite structure, operator formulations, and operational meaning

Negativity has structural content beyond bipartite quantification. A disentangling theorem states that, for a tripartite pure state, equality of $N_L=\log\left(\left\|\rho^{T_L}\right\|_1\right),$4 and $N_L=\log\left(\left\|\rho^{T_L}\right\|_1\right),$5 is equivalent to a specific factorization of the wavefunction,

$N_L=\log\left(\left\|\rho^{T_L}\right\|_1\right),$6

with a corresponding decomposition of the Hilbert space of $N_L=\log\left(\left\|\rho^{T_L}\right\|_1\right),$7. As a corollary, if $N_L=\log\left(\left\|\rho^{T_L}\right\|_1\right),$8, then $N_L=\log\left(\left\|\rho^{T_L}\right\|_1\right),$9. The same work reports that negativity itself is not generally monogamous, but that the square of the negativity appears to satisfy a monogamy inequality, supported analytically for three qubits and numerically in higher local dimensions (He et al., 2014).

An operator-level formulation is provided by the negativity Hamiltonian, defined formally through

$2\otimes2$0

This is the mixed-state analogue of the entanglement Hamiltonian, but attached to the partial transpose rather than to the reduced density matrix itself. In fermionic $2\otimes2$1-dimensional conformal field theories and free-fermion chains, the negativity Hamiltonian is found to assume a quasi-local form consisting of a local energy-density term plus a controlled mirror-coupling term. Its non-Hermiticity is not treated as a defect; rather, it encodes the negative or complex spectrum of the partial transpose that underlies logarithmic negativity (Murciano et al., 2022).

For a broad class of symmetric, $2\otimes2$2-uncorrelated many-body Gaussian states, including the free scalar field vacuum, logarithmic negativity is elevated from a convenient witness to a necessary and sufficient entanglement measure. In that setting, vanishing negativity is equivalent to separability, and the same partial-transpose structure determines exact detector profiles that maximize the accessible two-mode negativity. The paper further argues that, in the continuum limit, spacelike vacuum entanglement remains detectable at all distances through sufficiently fine detector resolution and optimized local collective modes (Gao et al., 2024).

The relation between negativity and operational access is also visible in experimental proposals. Charge-imbalance-resolved negativities are formulated through multi-copy protocols using Fourier transforms in copy space and swap-like observables in cold atoms (Cornfeld et al., 2018). Randomized-measurement ideas motivate the use of Rényi negativities as experimentally accessible proxies in open many-body dynamics (Wybo et al., 2021). In this sense, many-body negativity has evolved from a static criterion based on partial transpose into a broader toolkit for diagnosing, decomposing, simulating, and in some settings directly accessing mixed-state quantum correlations.

A distinct use of the phrase also exists in work on negative kinetic energy systems, where “many-body negativity” refers to negative kinetic energy, negative temperature, and reversed variational principles rather than partial-transpose entanglement (Wang, 2023). Within entanglement theory proper, however, the term denotes the family of negativity-based observables that quantify or proxy mixed-state entanglement in many-body quantum matter.