A new mathematical analysis suggests that entropy inequalities beyond strong subadditivity, a restriction on how entropies of overlapping regions fit together, can constrain how entanglement changes along some holographic renormalization-group flows. It does not, however, produce a new universal measure that steadily tracks the flow. The clearest result is narrower: a finite nested radial shell obeys an exact bound, and on a special Markov condition that bound becomes a one-sided limit on the growth rate of conditional mutual information, a measure of the correlation remaining between two regions after conditioning on a third.
The work is a theoretical modeling study rather than an experiment. It examines weighted graph models, mathematical networks in which boundary parties are terminal vertices and entropies are minimum-cut values, as well as common-light-cone regions, RT/HRT extremal-surface configurations and an asymptotically anti-de Sitter domain-wall model. No empirical participant sample or biological data was used.
Why the broader claim remains out of reach
The study tests whether inequalities involving many entangled regions can reveal response information that ordinary pairwise constraints miss. One obstruction appears in a local probe built from disjoint outward deformations, or “bumps,” of the regions. For every nonzero centered and balanced inequality valid for the weighted graph models, the second-order response reduces to a nonnegative sum of pair susceptibilities already controlled by strong subadditivity, abbreviated SSA. In that quadratic test, the higher-party structure adds no new sign information.
A six-party functional called G6 shows why that does not settle the matter. Its coefficient pattern is not a nonnegative combination of elemental Shannon-type entropy constraints, so it contains finite multipartite information that is absent from those pairwise building blocks. The construction can realize all twelve terms on a common light cone, and positivity is direct for the graph models. Applying the result to covariant HRT extremal surfaces requires the additional assumption that the relevant surfaces can be placed on a common slice, or on a common regulated slice.
Other finite constructions also fail to supply a graph-cone building block beyond SSA. Exact deletion of one or two parties reduces the coefficient pattern to nonnegative sums of conditional mutual informations, which SSA controls. Three independent deletions force the entire coefficient array to vanish. A separate finite-scale test reaches the same negative conclusion for a proposed radial construction: with fixed self-similar atomization, finite scale support and exact coefficient identities, no nonzero combination of the allowed inequalities can represent a finite difference between scales.
The result that survives
The successful construction uses five parties arranged so that a new nested radial shell absorbs an already-grown petal into the conditioning system. The exact finite-shell theorem says that the associated increase in conditional correlation is nonnegative and cannot exceed the initial conditional mutual information between two regions, together with two conditioned increases. It is a finite statement about a shell, not an unrestricted claim about every scale in a flow.
On the Markov face, the shell inequality becomes a one-sided conditional speed limit for the right logarithmic growth rate of conditional mutual information. If the Markov equality persists as the scale changes, the construction can define a conditional monotone. The inequality is not generated by strong subadditivity or by the monogamy-of-mutual-information structure, and refining the graph models can approach the bound sharply.
That qualification matters because the construction changes the central atomization as the scale changes while keeping spectator regions fixed. It therefore mixes scale and shape response instead of applying a self-similar dilation to the whole configuration. The analysis does not establish that the required Markov chamber, or the sharply approaching graph configuration, occurs in a smooth homogeneous holographic vacuum with a single boundary.
A separate geometric check
The authors also examine a special conformal transformation in an Einstein gravity null-energy-condition domain-wall model. The pair response is nonnegative in the stated model and is strictly positive unless the connected bulk component being probed is exact anti-de Sitter space. Exact anti-de Sitter space saturates the response. The result is therefore a positivity statement for that particular deformation and model.
What the preprint establishes
Taken together, the results support a qualified answer to the paper’s question: beyond-SSA inequalities can constrain entanglement response in selected holographic models through a trajectory-dependent Markov construction, but the resulting object is a conditional monotone rather than a second universal radial function analogous to the F-function.
The claims are mathematical and conditional. The graph results hold within their stated domain, while covariant HRT applications require common-slice or common-regulated-slice assumptions. The analysis is limited to three boundary spacetime dimensions and leading classical holographic order. Its finite-scale no-go result also depends on exact coefficient identities, finite scale support and fixed self-similar regions.
Paper data and sources
Original title: Beyond Strong Subadditivity: Holographic Entropy Inequalities Along Renormalization Group Flows
Authors: Ning Bao, Christian Ferko
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text