Reflected Entropy and Proposed Entanglement-Wedge Cross-Section Duals

Reflected entropy is the entropy of a specified canonical purification of $\rho_{AB}$. In semiclassical holographic states with a connected entanglement wedge, its leading $O(G_N^{-1})$ term is proposed—and derived through replica saddles under their assumptions—to equal twice the minimal entanglement-wedge cross section. The factor of two comes from the doubled canonical purification. Disconnection sets the leading geometric term to zero but does not make the exact boundary reflected entropy vanish; bulk correlations and finite-$N$ terms remain. We therefore treat the cross-section relation as an order- and saddle-dependent holographic statement, not as the definition of the measure.

Required background. Reflected entropy and canonical purification defines the measure, and entanglement-wedge nesting supplies the bulk region whose cross section is used.

Helpful background. Rényi analytic continuation supplies the two-index replica caveat, and holographic entropy cones distinguishes von Neumann inequalities from mixed-state proposals.

Canonical purification and the geometric proposal

Let

$$\rho_{AB}=\sum_i\lambda_i|i\rangle\langle i|.$$

The canonical purification in $ABA^*B^*$ is

$$|\sqrt{\rho_{AB}}\rangle =\sum_i\sqrt{\lambda_i} |i\rangle_{AB}|i\rangle_{A^*B^*},$$

with the second factor understood through the canonical vectorization. Reflected entropy is

$$S_R(A:B)=S(AA^*)_{|\sqrt\rho\rangle} =S(BB^*)_{|\sqrt\rho\rangle}.$$

If the classical entanglement wedge $\mathcal E_W[AB]$ is connected, let $\Sigma_{A:B}$ be its least-area codimension-two surface that partitions the wedge into portions adjacent to $A$ and $B$. Define

$$E_W(A:B)=\frac{\operatorname{Area}(\Sigma_{A:B})}{4G_N}.$$

The cross section was first advanced as a holographic candidate for entanglement of purification (Takayanagi and Umemoto 2018); reflected entropy supplies a different boundary measure and replica construction.

The leading holographic relation is

$$S_R(A:B)=2E_W(A:B)+O(G_N^0).$$

Dutta and Faulkner obtained this relation by constructing a bulk dual of the canonical purification and its replica path integral (Dutta and Faulkner 2021, §§2–4). The hypotheses include a dominant replica-symmetric semiclassical saddle and a specified continuation of the reflected-entropy replicas.

Two intervals with a connected wedge

Take the vacuum of a holographic CFT$_2$ on the line and intervals

$$A=[x_1,x_2], \qquad B=[x_3,x_4], \qquad x_1<x_2<x_3<x_4.$$

The conformal cross ratio is

$$\eta=\frac{x_{12}x_{34}}{x_{13}x_{24}}, \qquad 0<\eta<1,$$

where $x_{ij}=|x_i-x_j|$. In the leading vacuum-block regime, the connected RT pairing dominates for $\eta>1/2$. The minimal hyperbolic distance between the two bounding geodesics is

$$\operatorname{Length}(\Sigma_{A:B}) =L\log\!\left(\frac{1+\sqrt\eta}{1-\sqrt\eta}\right).$$

One can verify it in the symmetric configuration $A=[-b,-a]$, $B=[a,b]$: then $\sqrt\eta=(b-a)/(b+a)$, so the logarithm becomes $\log(b/a)$, the radial distance between the concentric geodesics.

Using $c=3L/(2G_3)$,

$$E_W(A:B)=\frac{c}{6} \log\!\left(\frac{1+\sqrt\eta}{1-\sqrt\eta}\right),$$

and the proposed leading reflected entropy is

$$S_R(A:B)=\frac{c}{3} \log\!\left(\frac{1+\sqrt\eta}{1-\sqrt\eta}\right) +O(c^0).$$

This is the first application. It compares a boundary replica observable with a geometrically independent cross-section calculation in the connected phase. The expression is cutoff independent for separated intervals because reflected entropy measures correlations rather than the individual area-law divergences in this configuration.

Disconnection is a phase test, not a redefinition

For $\eta<1/2$, the disconnected RT pairing dominates at order $c$, so $\mathcal E_W[AB]$ has two components and no cross section joining $A$ to $B$. The leading geometric prescription gives

$$E_W^{(0)}=0, \qquad S_R^{(0)}=0.$$

The exact density matrix need not factorize. Exchange of bulk fields and finite-$c$ CFT operators produces mutual information and reflected entropy of order $c^0$ or smaller. Declaring the exact reflected entropy to be zero would redefine the boundary measure to match the classical geometry. The correct conclusion is only that its $O(c)$ term vanishes in the disconnected saddle.

At the classical connectivity transition, the dominant cross section jumps. Close enough to the crossing, $O(c^0)$ terms can shift the transition and smooth finite-$c$ observables. Every comparison must keep the same saddle order on both sides.

Quantum corrections and a second control

In the doubled bulk associated with the canonical purification, the semiclassical entropy receives both a geometric term and the reflected entropy of bulk quantum fields across the wedge partition. Schematically,

$$S_R(A:B) =\frac{2\operatorname{Area}(\Sigma_{A:B})}{4G_N^{\rm ren}} +S_{R,\rm bulk}(a:b) +S_{\rm local}^{(1)}+O(G_N).$$

The local terms renormalize the geometric coupling, and the relevant cross section can become a quantum extremum in the doubled geometry. Computing only $2E_W$ and comparing it to a boundary result that includes $O(c^0)$ effects creates an order mismatch. Conversely, tuning the cross section by hand until it equals the exact measure is not evidence for the proposal.

The adversarial procedure is therefore: move through the connected/disconnected transition; compute the exact or controlled boundary reflected entropy at the same order; include bulk and counterterm corrections; and record any residual mismatch. A mismatch can diagnose replica-symmetry breaking, an incorrect continuation, a subdominant saddle, or limits of the cross-section proposal.

Status and handoff

The leading relation is well supported for semiclassical holographic CFT states and admits replica derivations in important regimes. It is not a theorem for arbitrary quantum states, arbitrary gravity actions, or finite $N$. Higher derivatives change the geometric functional, and gauge/algebra choices can add edge terms.

Negativity and purification proposals compares this result with inequivalent mixed-state measures whose replicas and defining properties differ.

The chapter overview contains the structure diagram and validity and failure diagram. They are embedded there once so that their shared chapter-level context is not repeated on every article.

For the chapter-wide comparison of assumptions, counterevidence, falsifiers, and claim ceilings, see the claim-domain table.

References

• Dutta, S., and Faulkner, T. (2021). “A canonical purification for the entanglement wedge cross-section.” Journal of High Energy Physics 2021(3), 178. DOI.
• Takayanagi, T., and Umemoto, K. (2018). “Entanglement of purification through holographic duality.” Nature Physics 14, 573–577. DOI.