Pure states for subregions in gravity and their entanglement entropy

Zixia Wei · 2026

Read Pure states for subregions in gravity and their entanglement entropy by Zixia Wei (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-03977.txt.

Introduction

Wei proposes a new way to assign states to spatial parts of quantum gravity. The standard view gives each part a mixed reduced density matrix. Wei instead assigns a pure state to each subregion. A partly frozen path integral prepares that state. The frozen part holds a spacetime zone that contains the spatial part.

In the semiclassical regime, Wei gives a holographic rule for split entropy. The rule uses a frozen-zone version of the homology constraint. It passes strong tests, including strong subadditivity and complementarity. Entanglement wedge nesting also holds under the rule. Known entropy formulas in holography appear as special cases. The build hints at an observer-linked wedge set by the frozen zone.

Catalog record

CREATORZixia Wei
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE27 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.03977

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Pure states for subregions in gravity and their entanglement entropy Zixia Wei Abstract It is proposed that spatial subregions in quantum gravity can be assigned pure states, rather than mixed reduced density matrices. The state is prepared by a partially frozen gravitational path integral, in which a spacetime subregion containing the spatial subregion is fixed while the field configurations and ambient geometry are summed over. In the semiclassical regime, we further propose a holographic prescription for the entanglement entropy of bipartitions of this state, with a frozen-region analogue of the homology constraint. The prescription satisfies nontrivial self-consistency conditions, including strong subadditivity, complementarity, and entanglement wedge nesting, and reproduces several known entropy formulas in holography and gravity as special cases. The construction suggests an observer-dependent entanglement wedge labeled by the frozen subregion. Introduction.— In a closed quantum many-body system, the entire system is described by a pure state, while a proper subsystem is generically described by a mixed reduced density matrix, reflecting the lack of information about the degrees of freedom outside the subsystem. What is the corresponding statement for a subregion of a universe with gravity? At the semiclassical level, one expects little qualitative difference: tracing over exterior degrees of freedom yields a mixed state for the subregion. In the quantum gravity regime, however, the holographic principle 36 ; 30 ; 23 suggests that the information which appears to lie outside the subregion may instead be encoded on its boundary. This raises the possibility that the subregion itself admits a pure-state description. In this Letter, we make this intuition precise by proposing that each spatial subregion in quantum gravity can be associated with a pure state prepared by a gravitational path integral (GPI) with fixed subregion data. To characterize this pure state, we further propose a Ryu–Takayanagi-like prescription 28 ; 27 for the entanglement entropy (EE) associated with its bipartitions, and verify that the prescription obeys nontrivial consistency conditions, including strong subadditivity (SSA), complementarity, and entanglement wedge nesting. For simplicity, we restrict to time-reflection-symmetric cases. Let us begin with the question of what it means to specify a subregion in quantum gravity. Some data must clearly be supplied. A natural choice is the intrinsic geometry of the subregion, although other choices are possible. For instance, one may specify only geometry of its boundary. In any case, these specified data will be held fixed and enter the GPI as input boundary conditions for the GPI. As we will see, this may be viewed as subregion generalization gravitationally prepared states 9 . We now formulate this generalization and show that the resulting object is naturally interpreted as a pure state associated with the subregion. Gravitationally prepared states for subregions.– A gravitationally prepared state on a closed spatial \Sigma slice is proposed in 9 and defined as follows. Consider a D -dimensional ( D -D) Einstein gravity coupled to a QFT described by the Euclidean action I_{E} , and start with a spacetime tube \Sigma\times I on which the gravity is turned off. Let us use \tau\in[-\epsilon,+\epsilon] to parameterize the direction along I and interpret it as the Euclidean time. We then glue a gravitating region {\mathcal{M}}_{G} along \partial\left(\Sigma\times I\right) such that \partial{\mathcal{M}}_{G}=\partial\left(\Sigma\times I\right) . For the gravity sector, we impose the Dirichlet boundary condition (DBC) at the interface between {\mathcal{M}}_{G} and \Sigma\times I , while for the QFT sector, the interface is transparent and it lives across the whole spacetime. Let us use {\mathcal{M}}={\mathcal{M}}_{G}\cup(\Sigma\times I) to denote the entire spacetime manifold and \phi to denote the quantum fields. We consider the Euclidean GPI with the geometry \Sigma\times I serving as the input data, \displaystyle Z_{\rm grav}[\Sigma\times I]=\int_{\partial{\mathcal{M}}_{G}=\partial(\Sigma\times I)}\mathcal{D}g_{\mu\nu}\int_{{\mathcal{M}}}\mathcal{D}\phi\exp\left(-I_{E}\right)\,, (1) where we sum over the QFT configurations on the entire spacetime {\mathcal{M}} and the geometries of {\mathcal{M}}_{G} . This GPI is considered to be computing the ensemble average 1 1 1 Note that although the analyses performed in 9 do not require the ensemble average, the authors are open to this possibility, and it may in general be necessary to avoid the factorization puzzle 22 ; 29 , so we take this more general choice. of the trace of a family of unnormalized QFT density matrices \rho_{\Sigma}\in{\mathcal{B}}({\mathcal{H}}_{\Sigma}^{\rm QFT}) defined on \tau=0 , \displaystyle Z_{\rm grav}[\Sigma\times I]=\overline{{{\text{Tr}}\left(\rho_{\Sigma}\right)}}\,, (2) which is called a gravitationally prepared state. Here the overline denotes the ensemble average. Such a GPI can be dominated by a “bra-ket wormhole”, in which QFT on \Sigma looks entangled with another closed universe and hence mixed at semiclassical level, but one can demonstrate via the replica trick that \rho_{S} is pure 9 . As a consequence, one may alternatively write ( 2 ) as \displaystyle Z_{\rm grav}[\Sigma\times I]=\overline{\braket{\Psi(\Sigma\times I/\mathbb{Z}_{2})|\Psi(\Sigma\times I/\mathbb{Z}_{2})}_{\Sigma}}\,, (3) where the pure state \ket{\Psi(\Sigma\times I/\mathbb{Z}_{2})}_{\Sigma} is labeled by half of the original input data \Sigma\times I , separated by the spatial slice at \tau=0 . More generally, the inner product, trace, and other index-contractions associated to the quantum states are accordingly defined by cutting and pasting in the non-gravitational region following the same rules as that in QFT. Now let us generalize this setup to a broader class of GPI, in which a general spacetime subregion is frozen. In particular, while the frozen subregion \Sigma\times I discussed above includes the full Cauchy slice \Sigma , now we would like to consider a general {\mathcal{M}}_{F} which only covers a spatial subregion. More precisely, let {\mathcal{M}}_{F} be a D -D manifold and consider the GPI \displaystyle Z_{\rm grav}[{\mathcal{M}}_{F}]=\int_{\partial{\mathcal{M}}_{G}=\partial{\mathcal{M}}_{F}}\mathcal{D}g_{\mu\nu}\int_{{\mathcal{M}}}\mathcal{D}\phi\exp\left(-I_{E}\right)\,. (4) Let F be a (D-1) -D spatial slice of {\mathcal{M}}_{F} , we assume the partially frozen GPI computes the trace of an unnormalized reduced transition matrix \mathcal{T}_{F^{\circ}\cup\partial F} on \mathcal{H}_{F^{\circ}}\otimes\mathcal{H}_{\partial F} , averaged, \displaystyle Z_{\rm grav}[{\mathcal{M}}_{F}]=\overline{{{\text{Tr}}\left(\mathcal{T}_{F^{\circ}\cup\partial F}\right)}}\,, (5) considering its similarity to the reduced transition matrix in QFT 24 . Here, \displaystyle F^{\circ}\equiv F\backslash\partial F\,, (6) denotes the interior of F . We also assume that when F is a time-reflection symmetric slice of {\mathcal{M}}_{F} , \mathcal{T}_{F^{\circ}\cup\partial F} becomes a density matrix \rho_{F^{\circ}\cup\partial F} . The reason why we carefully separate F^{\circ} and \partial F will become clear soon later. The partially frozen GPI in ( 4 ) provides a subregion generalization of the gravitationally prepared states defined in 9 . Let us then consider if such a state is pure or mixed via the replica trick. Pure nature of the subregion state.— Consider the case where {\mathcal{M}}_{F} is time-reflection symmetric with respect to F . Prepare two copies of {\mathcal{M}}_{F} , and call them {\mathcal{M}}_{F}^{(1)} and {\mathcal{M}}_{F}^{(2)} respectively. Cut each of {\mathcal{M}}_{F} along the spatial slice F , and denote the two parts separated by it {\mathcal{M}}_{F}^{+} and {\mathcal{M}}_{F}^{-} . Now, with four pieces {\mathcal{M}}_{F}^{(1)\pm} and {\mathcal{M}}_{F}^{(2)\pm} , let us combine them in the following two ways. First, gluing {\mathcal{M}}_{F}^{(1)+} ( {\mathcal{M}}_{F}^{(2)+} ) and {\mathcal{M}}_{F}^{(1)-} ( {\mathcal{M}}_{F}^{(2)-} ) back to {\mathcal{M}}_{F}^{(1)} ( {\mathcal{M}}_{F}^{(2)} ), the GPI with {\mathcal{M}}_{F}^{(1)}\sqcup{\mathcal{M}}_{F}^{(2)} frozen computes \displaystyle Z_{\rm grav}[{\mathcal{M}}_{F}^{(1)}\sqcup{\mathcal{M}}_{F}^{(2)}]=\overline{\left({{\text{Tr}}\left(\rho_{F^{\circ}\cup\partial F}\right)}\right)^{2}}\,. (7) On the other hand, gluing {\mathcal{M}}_{F}^{(1)+} ( {\mathcal{M}}_{F}^{(2)+} ) and {\mathcal{M}}_{F}^{(2)-} ( {\mathcal{M}}_{F}^{(1)+} ) along F to obtain {\mathcal{M}}_{F}^{(1)+(2)-} ( {\mathcal{M}}_{F}^{(2)+(1)-} ), the GPI with {\mathcal{M}}_{F}^{(1)+(2)-}\sqcup{\mathcal{M}}_{F}^{(2)+(1)-} frozen computes \displaystyle Z_{\rm grav}[{\mathcal{M}}_{F}^{(1)+(2)-}\sqcup{\mathcal{M}}_{F}^{(2)+(1)-}]=\overline{{{\text{Tr}}\left(\mathcal{\rho}_{F^{\circ}\cup\partial F}^{2}\right)}}\,. (8) Since {\mathcal{M}}_{F}^{(1)}\sqcup{\mathcal{M}}_{F}^{(2)}={\mathcal{M}}_{F}^{(1)+(2)-}\sqcup{\mathcal{M}}_{F}^{(2)+(1)-} , we have \displaystyle\overline{\left({{\text{Tr}}\left(\rho_{F^{\circ}\cup\partial F}\right)}\right)^{2}}=\overline{{{\text{Tr}}\left(\rho_{F^{\circ}\cup\partial F}^{2}\right)}}\,. (9) If there is no ensemble average, this is already sufficient for claiming {\rm rank}\left(\rho_{F^{\circ}\cup\partial F}\right)=1 . In

Read the complete text