---
title: Perturbative Separability Criterion
url: https://www.emergentmind.com/topics/perturbative-separability-criterion
type: topic
---

# Perturbative Separability Criterion

Searching arXiv for the primary paper and closely related separability-criterion work.
{"query":"1705.05557 Separability criterion for three-qubit states with a four dimensional norm", "max_results": 5}
The perturbative separability criterion, in the three-qubit setting developed in "Separability criterion for three-qubit states with a four dimensional norm" [1705.05557], is a complete characterization of separability for three-qubit \({\sf X}\)-states in terms of a diagonal margin and a dual norm of the anti-diagonal sector. An \({\sf X}\)-state is an \(8\times 8\) self-adjoint matrix whose only potentially nonzero entries are on the diagonal and anti-diagonal, written as \(\varrho=X(a,b,c)\) with \(a,b\in\mathbb R^4\) and \(c\in\mathbb C^4\). In this framework, separability is reduced to an explicit inequality of the form \(\Delta_\varrho\ge \|c\|^\prime\), which is especially useful for near-diagonal or “perturbed diagonal” states because it isolates the anti-diagonal perturbation as the only nontrivial obstruction to separability.

## 1. State class and structural reduction

For three-qubit \({\sf X}\)-states, the matrix is parameterized as
\[
\varrho=X(a,b,c),
\]
with
\[
a=(a_1,a_2,a_3,a_4),\qquad b=(b_1,b_2,b_3,b_4)\in\mathbb R^4,\qquad c=(c_1,c_2,c_3,c_4)\in\mathbb C^4.
\]
The anti-diagonal part \(c\) is the key object, while the diagonal data enter only through a single scalar margin. The paper’s main theorem states that separability is equivalent to comparing a scalar built from the diagonal entries with the dual norm of a norm determined by the anti-diagonal entries [1705.05557].

The diagonal quantity is
\[
\Delta_\varrho := \min\Bigl\{\sqrt{a_i b_i}\ (i=1,2,3,4),\ \sqrt[4]{a_1 b_2 b_3 a_4},\ \sqrt[4]{b_1 a_2 a_3 b_4}\Bigr\}.
\]
This reduction is significant because it transforms the separability problem from a decomposition problem over product states into a comparison between explicitly computable quantities. In the \({\sf X}\)-state sector, that comparison is exact rather than merely necessary or merely sufficient.

## 2. Full criterion and witness formulation

The full separability criterion is
\[
\varrho=X(a,b,c)\ \text{is separable}
\quad\Longleftrightarrow\quad
\Delta_\varrho \ge \|c\|^\prime.
\]
Here \(\|c\|^\prime\) is the dual norm of a specific norm on \(\mathbb C^4\) induced by the anti-diagonal structure [1705.05557].

The paper also gives an equivalent witness formulation. The state is separable if and only if, for every \(z\in\mathbb C^4\),
\[
C(z)\,\Delta_\varrho \ge \mathcal L(\varrho,z),
\]
where
\[
\mathcal L(\varrho,z)=\operatorname{Re}\bigl(z_1c_1+z_2c_2+z_3c_3+z_4\bar c_4\bigr).
\]
The witness coefficient \(C(z)\) is then simplified to a norm \(B(z)\) through
\[
C(z)=B(z_1,z_2,z_3,\bar z_4),
\]
with
\[
B(z)=\max_\sigma\left(|z_1e^{i\sigma}+\bar z_4|+|z_2e^{i\sigma}+\bar z_3|\right).
\]

This reformulation is the operational core of the criterion. It shows that the anti-diagonal sector is tested against a dual norm determined by the same geometry that appears in the witness bound. For \({\sf X}\)-states, the resulting inequality is necessary and sufficient, so no additional PPT-, range-, or decomposition-based check is required within that family.

## 3. The four-dimensional norm and phase dependence

The relevant norm on \(\mathbb C^4\) is
\[
\|z\|_= \max_{\sigma}\left(|z_1 e^{i\sigma}+\bar z_4|+|z_2 e^{i\sigma}+\bar z_3|\right), \qquad z\in\mathbb C^4,
\]
and its dual norm is
\[
\|c\|^\prime=\max_{z\ne 0}\frac{\operatorname{Re}(c,z)}{\|z\|_= \max_{\sigma}\left(|z_1 e^{i\sigma}+\bar z_4|+|z_2 e^{i\sigma}+\bar z_3|\right)}.
\]
With this notation, the separability criterion becomes simply
\[
\Delta_\varrho\ge \|c\|^\prime.
\]

A central structural fact is that this norm depends only on the magnitudes \(|c_i|\) and the phase difference
\[
\phi_c=(\theta_1+\theta_4)-(\theta_2+\theta_3)\pmod{2\pi},
\]
if \(c_i=r_ie^{i\theta_i}\). The dependence on phases is therefore highly compressed. In particular, if only one anti-diagonal entry vanishes, the remaining phases become irrelevant to separability [1705.05557].

This phase reduction is one of the reasons the criterion is suitable for “routine computations.” The anti-diagonal contribution is not controlled by four independent phases; it is controlled by the four magnitudes together with a single invariant phase combination. A common misconception is that all anti-diagonal phases contribute independently. In this setting they do not.

## 4. Explicitly computable cases

The paper computes \(\|c\|^\prime\) in several cases that are practically important because they replace the abstract dual-norm optimization by closed formulas or direct geometric tests [1705.05557].

For real anti-diagonals \(c\in\mathbb R^4\), the dual norm is determined by the sign pattern of \(c\). Using
\[
\lambda_5=2(c_1+c_2+c_3+c_4),\quad
\lambda_6=2(-c_1-c_2+c_3+c_4),
\]
\[
\lambda_7=2(-c_1+c_2-c_3+c_4),\quad
\lambda_8=2(-c_1+c_2+c_3-c_4),
\]
together with quantities \(t_1,t_2,t_3,t_4\), the analysis splits into three cases. In cases (A) or (B),
\[
\|c\|^\prime=\|c\|_\infty.
\]
In case (C),
\[
\|c\|^\prime=\frac18\,\Lambda(\lambda_5,\lambda_6,\lambda_7,\lambda_8),
\]
where
\[
\Lambda(a_1,a_2,a_3,a_4)
:=
\sqrt{\frac{(a_1a_2+a_3a_4)(a_1a_3+a_2a_4)(a_1a_4+a_2a_3)}{a_1a_2a_3a_4}}.
\]
This gives a complete closed formula for real anti-diagonals.

If one of the four entries of \(c\) is zero, the norm reduces to a simple geometric expression. If exactly three entries are nonzero and their magnitudes do not form an acute triangle, then
\[
\|c\|^\prime=\|c\|_\infty.
\]
If they form an acute triangle, then
\[
\|c\|^\prime=
\frac{2|c_ic_jc_k|}{
\sqrt{(|c_i|+|c_j|+|c_k|)(-|c_i|+|c_j|+|c_k|)(|c_i|-|c_j|+|c_k|)(|c_i|+|c_j|-|c_k|)}
}.
\]
If at least two entries of \(c\) are zero, then
\[
\|c\|^\prime=\|c\|_\infty.
\]

If the entries can be partitioned into two pairs with equal magnitudes,
\[
|c_{i_1}|=|c_{i_2}|=r,\qquad |c_{i_3}|=|c_{i_4}|=s,
\]
with \(r,s\neq 0\), then the phase difference \(\phi_c\) governs the answer. If \(\phi_c=0\),
\[
\|c\|^\prime=\max\{r,s\}.
\]
The especially important symmetric case
\[
|c_1|=|c_2|=|c_3|=|c_4|=r
\]
gives
\[
\|c\|^\prime=r\sqrt{1+|\sin(\phi_c/2)|}.
\]

These formulas explain why the criterion is operationally sharp. In the real, zero-entry, and symmetric two-pair regimes, the abstract dual norm is replaced by explicit algebraic or geometric expressions, so separability can be decided by direct computation rather than by a search over separable decompositions.

## 5. Perturbed-diagonal and semi-perturbative analysis

The perturbative aspect of the criterion is most explicit for states close to diagonal. The paper emphasizes that the \({\sf X}\)-part of any separable three-qubit state is again separable, so the criterion gives a necessary test for arbitrary states by looking only at their diagonal and anti-diagonal entries [1705.05557].

This leads directly to a near-diagonal workflow. If one starts from a diagonal or semi-diagonal state and adds small anti-diagonal terms, separability can be tested by checking whether the diagonal margin \(\Delta_\varrho\) dominates the dual norm of those terms. The basic comparison remains
\[
\Delta_\varrho\ge \|c\|^\prime.
\]
The paper also derives lower and upper bounds
\[
\|c\|_\infty \le \|c\|^\prime \le \|c\|_1,
\]
together with sharper bounds depending on phase difference and pairwise magnitudes. One estimate is
\[
\|c\|^\prime\ge
\max_\phi \frac{\|(c_1,c_2,c_3,c_4e^{i\phi})\|_}
{2\sqrt2\sqrt{1+|\cos(\phi/2)|}},
\]
which provides computable sufficient conditions for separability of the \({\sf X}\)-part of general states.

Two points delimit the scope of the result. First, for matrices whose only nonzero entries are diagonal and anti-diagonal, the criterion is complete:
\[
X(a,b,c)\ \text{is separable}
\quad\Longleftrightarrow\quad
\Delta_{X(a,b,c)}\ge \|c\|^\prime.
\]
Second, for a general three-qubit state, the criterion applies to its \({\sf X}\)-part. Failure of the inequality rules out separability, because the \({\sf X}\)-part of a separable state must itself be separable. Satisfaction of the inequality certifies separability only in the \({\sf X}\)-sector.

In that precise sense, the criterion is “perturbative” and “semi-perturbative.” It is perturbative because it is especially effective for perturbed diagonal states, and semi-perturbative because it yields a sharp necessary test for arbitrary states after projection to the diagonal-plus-anti-diagonal sector.

## 6. Related near-threshold and perturbative separability frameworks

The broader separability literature contains several adjacent formulations in which separability is controlled by a restricted perturbation, a spectral deviation, or a neighborhood of a distinguished state. A geometric near-identity criterion treats density matrices as points in Euclidean space with Hilbert–Schmidt distance and derives exact thresholds for Werner families, including
\[
\rho_W \text{ is separable } \iff p\le \frac{1}{N+1}
\]
for bipartite Werner states and
\[
\rho_W \text{ is separable } \iff p\le \frac{1}{d^{\,n-1}+1}
\]
for \(n\)-qudit maximally entangled Werner states, together with an explicit separability ball around the normalized identity [1608.06145]. This is a different criterion from the \({\sf X}\)-state norm inequality, but it shares the same near-threshold logic: a state remains separable if the entangled component is not too large.

A spectral perturbation-style criterion appears in the channel-state-duality approach, where separability of a bipartite state is equivalent to the existence of rank-one Kraus operators for the associated CP map. In that setting, a refined criterion uses the singular-value perturbation inequality
\[
\sum_{i=1}^q \left(\sigma_i(A)-\sigma_i(B)\right)^2 \le \|A-B\|_2^2,
\]
which sharpens a basic spectral test and detects some entangled states missed by the first criterion [1909.13309]. The relation to the perturbative \({\sf X}\)-state criterion is methodological rather than formal: both reduce separability to explicit inequalities after isolating a structured part of the state.

A further variant is the separability-gap viewpoint for Hamiltonians. There the comparison is between the true ground-state energy \(E_0(H)\) and the minimal expectation value over product states,
\[
\lambda_{\rm min}^{\otimes}(H)=\min_{\ket{\psi_{\rm sep}}}\langle \psi_{\rm sep}|H|\psi_{\rm sep}\rangle,
\]
with gap
\[
\Delta_{\rm sep}(H)=\lambda_{\rm min}^{\otimes}(H)-E_0(H).
\]
If an expectation value falls below the product-state bound, entanglement is witnessed [1812.09251]. This does not reproduce the norm inequality for \({\sf X}\)-states, but it expresses the same general principle: separability is benchmarked by a restricted optimization, and sufficiently large deviation from that benchmark certifies entanglement.

Taken together, these results locate the perturbative separability criterion within a wider program of replacing decomposition-based separability questions by explicit inequalities. In the three-qubit \({\sf X}\)-state problem, that replacement is exact and particularly sharp:
\[
\boxed{\ \varrho=X(a,b,c)\ \text{separable} \iff \Delta_\varrho\ge \|c\|^\prime\ }.
\]
Its distinguishing feature is that the anti-diagonal perturbation is encoded by a four-dimensional norm whose dual is computable in several nontrivial regimes, while the diagonal sector enters only through the single scalar margin \(\Delta_\varrho\).

Source: https://www.emergentmind.com/topics/perturbative-separability-criterion