---
title: Disentangling Theorem & Entanglement Structure
url: https://www.emergentmind.com/topics/disentangling-theorem
type: topic
---

# Disentangling Theorem & Entanglement Structure

Searching arXiv for the primary paper and closely related work on entanglement negativity and monogamy.
Searching arXiv for papers on entanglement negativity, monogamy, and tripartite structure.
The disentangling theorem is a structural characterization of tripartite pure states in terms of entanglement negativity. In its standard form, for a pure state \(|\Psi_{ABC}\rangle\) on \(\mathcal H_{ABC}\cong \mathcal H_A\otimes\mathcal H_B\otimes\mathcal H_C\), it compares the negativity across the bipartition \(A|BC\) with the negativity of the reduced mixed state \(\rho_{AB}=\operatorname{tr}_C(|\Psi_{ABC}\rangle\langle\Psi_{ABC}|)\) across \(A|B\). The theorem states that equality of these two negativities is equivalent to a rigid factorization of the global state into independent \(AB_1\) and \(B_2C\) components. In this sense, negativity is not merely a mixed-state entanglement monotone, but a witness of an exact tensor-product structure in the physically relevant support of subsystem \(B\) [1401.5843].

## 1. Formal statement

For a bipartite state \(\rho_{AB}\), the entanglement negativity is defined by
\[
\mathcal N^{A|B}=\frac{\|\rho_{AB}^{T_A}\|_1-1}{2},
\]
where \(T_A\) denotes partial transpose on subsystem \(A\), and \(\|\cdot\|_1\) is the trace norm. In the tripartite setting, the theorem compares \(\mathcal N^{A|BC}\), computed on the pure state \(|\Psi_{ABC}\rangle\) across \(A|BC\), with \(\mathcal N^{A|B}\), computed on the mixed state \(\rho_{AB}\) across \(A|B\) [1401.5843].

The precise claim is
\[
\mathcal N^{A|BC}=\mathcal N^{A|B}
\quad\Longleftrightarrow\quad
\exists\, \mathcal H_{B_1},\mathcal H_{B_2},\mathcal H_B^\perp
\text{ such that }
\mathcal H_B\cong (\mathcal H_{B_1}\otimes\mathcal H_{B_2})\oplus \mathcal H_B^\perp
\]
and, with respect to that decomposition,
\[
|\Psi_{ABC}\rangle = |\Psi_{AB_1}\rangle\otimes |\Psi_{B_2C}\rangle.
\]

This is stronger than the statement that \(B\) “splits” abstractly. The relevant content is that the support of the reduced state on \(B\) admits a tensor-product decomposition, while the additional summand \(\mathcal H_B^\perp\) is orthogonal to that support and therefore irrelevant to the state. Equality of the two negativities is thus equivalent to a product decomposition across the cut
\[
(AB_1)\,|\,(B_2C).
\]

A useful equivalent formulation is obtained by choosing an orthonormal basis \(\{|\psi_B^{ij}\rangle\}\) for the support of \(\rho_B\) and identifying it with a product basis
\[
|\phi_{B_1}^i\rangle\otimes|\phi_{B_2}^j\rangle \equiv |\psi_B^{ij}\rangle.
\]
Then the state takes the form
\[
|\Psi_{ABC}\rangle
=
\sum_{ij}\sqrt{p_i q_j}\,
|\phi_A^i\rangle\,|\phi_{B_1}^i\rangle\,|\phi_{B_2}^j\rangle\,|\phi_C^j\rangle
=
|\Psi_{AB_1}\rangle\otimes|\Psi_{B_2C}\rangle,
\]
with
\[
|\Psi_{AB_1}\rangle=\sum_i \sqrt{p_i}\,|\phi_A^i\phi_{B_1}^i\rangle,
\qquad
|\Psi_{B_2C}\rangle=\sum_j \sqrt{q_j}\,|\phi_{B_2}^j\phi_C^j\rangle.
\]

## 2. Structural meaning

The theorem provides a very sharp answer to the question of when tracing out \(C\) leaves the negativity unchanged. If \(\mathcal N^{A|B}\) does not drop under \(\operatorname{tr}_C\), that invariance is not a numerical accident; it means that the full wavefunction was already factorized so that all entanglement between \(A\) and \(BC\) was concentrated entirely in \(AB_1\), with the remaining degrees of freedom of \(B\) correlating only with \(C\) [1401.5843].

This yields an immediate corollary:
\[
\mathcal N^{A|BC}=\mathcal N^{A|B}
\quad\Longrightarrow\quad
\mathcal N^{A|C}=0.
\]
Indeed, from the factorized form one obtains
\[
\rho_{AC}=\rho_A\otimes \rho_C,
\]
so \(A\) and \(C\) are unentangled. The theorem therefore isolates an extremal regime in which all entanglement between \(A\) and the rest of the system is localized in a proper part of \(B\), and none can be shared with \(C\).

The result can also be read as a support-level factorization theorem. What is forced is not a decomposition of the entire ambient Hilbert space \(\mathcal H_B\) in an arbitrary sense, but a decomposition of the physically relevant support \(\mathcal H_B^\parallel\subseteq \mathcal H_B\). A plausible implication is that the theorem is best understood as a statement about the operational content of the reduced state rather than about arbitrary unused degrees of freedom in \(B\).

## 3. Proof architecture

The proof proceeds in three stages. First, one takes a Schmidt decomposition of the pure state across \(A|BC\):
\[
|\Psi_{ABC}\rangle=\sum_i \sqrt{p_i}\,|\phi_A^i\phi_{BC}^i\rangle,
\]
with orthonormal families \(\{|\phi_A^i\rangle\}\) and \(\{|\phi_{BC}^i\rangle\}\). Each \(|\phi_{BC}^i\rangle\) is then expanded in product bases of \(B\) and \(C\):
\[
|\phi_{BC}^i\rangle=\sum_{jk} T^i_{jk}\,|\phi_B^j\phi_C^k\rangle,
\]
where the coefficients satisfy
\[
\sum_{ab} T^{i*}_{ab}T^j_{ab}=\delta_{ij}.
\]

Second, the partially transposed operator \(\rho^{T_A}_{ABC}\) is explicitly diagonalized. Its eigenvalues are \(p_i\), \(+\sqrt{p_ip_j}\), and \(-\sqrt{p_ip_j}\) for \(i<j\), with eigenvectors denoted by \(|0_i\rangle\), \(|+_{ij}\rangle\), and \(|-_{ij}\rangle\). Hence
\[
\rho^{T_A}_{ABC} = (1+n)\rho^+_{ABC}-n\rho^-_{ABC},
\qquad
n=\mathcal N^{A|BC},
\]
with \(\rho^\pm_{ABC}\ge 0\) and
\[
\operatorname{tr}(\rho^+_{ABC}\rho^-_{ABC})=0.
\]

Third, one traces out \(C\):
\[
\rho^{T_A}_{AB}
=
\operatorname{tr}_C(\rho^{T_A}_{ABC})
=
(1+n)\rho^+_{AB}-n\rho^-_{AB}.
\]
This decomposition need not remain optimal after tracing out \(C\). The equality \(\mathcal N^{A|B}=n\) holds if and only if the traced positive and negative parts remain orthogonal,
\[
\operatorname{tr}(\rho^+_{AB}\rho^-_{AB})=0.
\]
That orthogonality condition is then translated into constraints on the coefficients \(T^i_{jk}\), and those constraints force the factorization of the support of \(B\). The proof therefore converts preservation of a trace-norm quantity into a rigid tensor-product structure [1401.5843].

## 4. Optimal decomposition and orthogonality criterion

A key technical ingredient is Lemma 2 from Vidal and Werner, used to analyze optimal decompositions of Hermitian operators. Any Hermitian matrix \(A\) can be written as
\[
A=a_+\rho^+-a_-\rho^-,
\]
with \(\rho^\pm\ge 0\) and \(a_\pm\ge 0\). Such a decomposition is optimal if and only if several equivalent conditions hold, including
\[
\|A\|_1=a_++a_-,
\qquad
\mathcal N=a_-,
\qquad
\operatorname{tr}(\rho^+\rho^-)=0
\]
[1401.5843].

In the disentangling theorem, the crucial condition is the orthogonality of the positive and negative parts. For \(\rho^{T_A}_{ABC}\), this orthogonality is automatic in the spectral decomposition. After tracing out \(C\), however, the positive and negative components may cease to be orthogonal. Equality of negativities before and after tracing out \(C\) is equivalent to the claim that this orthogonality survives the partial trace.

When that condition is unpacked, the paper derives two families of constraints:
\[
\sum_a T^{i*}_{am}T^j_{an}=0
\qquad (i\neq j),
\]
and
\[
\sum_a T^{i*}_{am}T^i_{an}
=
\sum_a T^{j*}_{am}T^j_{an}
\qquad \forall i\neq j.
\]
Together they imply the compact relation
\[
\sum_a T^{i*}_{am}T^j_{an}=\delta_{ij}C_{mn},
\]
for some matrix \(C_{mn}\). This is the structural constraint that forces factorization. Once the \(C\)-basis is chosen so that \(\rho_C\) is diagonal, \(C_{jk}=\delta_{jk}q_j\), one defines
\[
|\tilde\psi_B^{ik}\rangle=\sum_j T^i_{jk}|\phi_B^j\rangle,
\]
and obtains
\[
\langle \tilde\psi_B^{mn}|\tilde\psi_B^{ij}\rangle
=
\delta_{mi}\delta_{nj}q_j.
\]
After normalization, these vectors form an orthonormal basis for the support of \(\rho_B\), which can then be relabeled as a product basis of \(B_1\otimes B_2\). This is the precise mechanism by which orthogonality of the traced decomposition becomes a support factorization.

## 5. Relation to monogamy of negativity

The theorem is closely tied to monogamy. The paper places it in the context of the conjectured inequality
\[
(\mathcal N^{A|BC})^2 \ge (\mathcal N^{A|B})^2 + (\mathcal N^{A|C})^2.
\]
In the regime singled out by the disentangling theorem, one has \(\mathcal N^{A|C}=0\), so the inequality is saturated in a trivial way. The theorem therefore identifies a special extremal case of monogamy: if tracing out \(C\) does not reduce the negativity, then \(A\) must already be completely disentangled from \(C\) [1401.5843].

The paper further reports numerical evidence supporting the squared form of the monogamy relation, especially in three-qubit and higher-dimensional examples. By contrast, plain negativity does not seem to satisfy the unsquared inequality
\[
\mathcal N^{A|BC}\ge \mathcal N^{A|B}+\mathcal N^{A|C}.
\]

This places the disentangling theorem in a broader conceptual role. It is not itself a generic monogamy inequality; rather, it characterizes the boundary case in which one subsystem can be discarded without any loss of negativity. The significance of that boundary case is that it admits an exact structural interpretation, not merely an inequality.

## 6. Scope, significance, and broader usage of the term

Within quantum information theory, the theorem gives negativity a dual status. It remains a measure of mixed-state entanglement, increasingly used to investigate and characterize quantum many-body phenomena including quantum criticality and topological order, but it also functions as a structural witness for disentangling in tripartite pure states [1401.5843].

A common misconception is to treat equality of \(\mathcal N^{A|BC}\) and \(\mathcal N^{A|B}\) as a weak statement about robustness under partial trace. The theorem shows that the condition is far stronger: it is equivalent to an exact product decomposition of the global state on the support of \(B\). Another misconception is to read the conclusion as merely saying that \(A\) and \(C\) are unentangled. That corollary is true, but the theorem asserts substantially more, namely the existence of a tensor-product decomposition of the support of \(B\) under which the global wavefunction separates across \((AB_1)|(B_2C)\).

The expression “disentangling theorem” has also acquired broader meanings in later literature. In condensed-matter and lattice-symmetry settings, it can denote the statement that any finite, internal, anomaly-free symmetry in a \(1+1\)d quantum spin chain can be converted into an on-site symmetry after adding ancillas and applying a finite-depth quantum circuit [2503.09717]. In quantum complexity, “disentangler” refers to a channel whose image approximates separable states, and recent work proves exponential lower bounds for the input dimension of strong disentanglers [2402.08981]. These usages are conceptually related only at a high level. The theorem of primary importance in entanglement negativity remains the tripartite factorization criterion of [1401.5843].

## 7. Conceptual legacy

The lasting importance of the disentangling theorem lies in the rigidity of its conclusion. Entanglement measures often supply inequalities, monotonicity statements, or order-theoretic comparisons. Here, by contrast, equality of two negativities yields an if-and-only-if structural theorem about the wavefunction itself [1401.5843].

That rigidity makes the theorem unusual. It turns a condition formulated entirely in terms of partial transpose and trace norm into a precise decomposition of the support of one subsystem and a product form of the global state. The result therefore forms a bridge between entanglement quantification and wavefunction structure. A plausible implication is that it is best viewed not only as a theorem about negativity, but as an exact criterion for when the entanglement carried by \(A\) is already fully localized inside a proper subfactor of \(B\), leaving \(C\) completely outside that entanglement structure.

Source: https://www.emergentmind.com/topics/disentangling-theorem