Entanglement and Spacetime

Entanglement and spacetime research asks when selected quantum correlations admit a geometric description and which geometric properties that description determines. The useful question is not whether every entangled state “is space,” but which state families, observables, and limits support a controlled dictionary. Reviewed 11 October 2026, this page uses Van Raamsdonk’s entanglement argument and Harlow’s bulk-emergence lectures as theoretical entry points.

The selected technical baseline spans the 2006 entropy proposal, 2013 monogamy result, 2016 random-network construction, and 2018 overview. It supports the conditional tests below; it is not a comprehensive assessment of all developments through the search date.

Helpful background. Entanglement entropy and mutual information supply the quantum quantities. Holography and quantum information addresses encoding and reconstruction.

An entropy pattern is not a unique geometry

In an appropriate semiclassical holographic regime, boundary entropies are related to extremal-surface areas plus quantum corrections. This is a statement inside a particular duality and approximation, not a definition valid for arbitrary many-qubit states.

Even before gravity, bipartite entropies cannot specify a general quantum state. States can share many entropy values while differing in phases, multipartite correlations, or dynamics. Consequently, a reconstruction from entropies needs a uniqueness statement or a description of its unresolved equivalence class.

For disjoint subsystems,

$$I(A:B)=S(A)+S(B)-S(AB)\ge0.$$

A small value bounds some correlations but is not itself a universal distance. Converting it into a geometric length requires additional assumptions about operators, states, and the proposed bulk model.

A finite calculation can reject an unjustified geometric dictionary

Established under the classical Ryu–Takayanagi area prescription: disjoint boundary regions satisfy monogamy of mutual information, equivalently

$$I_3(A:B:C)={}S(A)+S(B)+S(C) -S(AB)-S(AC)-S(BC)+S(ABC)\le0.$$

Hayden, Headrick, and Maloney derive this restriction from the minimal-surface construction. It is not an inequality obeyed by all quantum states. For the four-qubit state $(|0000\rangle+|1111\rangle)/\sqrt2$, every nonempty proper subset has entropy $\log2$, so $I_3=\log2>0$. This state cannot have its entire entropy vector represented by that leading classical area prescription. The conclusion concerns a specified dictionary; bulk quantum corrections can alter the entropy inequalities, and this calculation does not forbid every gravitational interpretation of a GHZ state.

A tensor network also supplies a bound rather than automatic equality. Cutting bonds $e$ that separate $A$ from its complement bounds the Schmidt rank by $\prod_eD_e$, hence

$$S(A)\le\min_{\text{cuts}}\sum_e\log D_e.$$

The tensors determine whether the bound is saturated. A four-leg GHZ tensor with bond dimension two has entropy $\log2$ across a two-versus-two split, although cutting two boundary legs gives $2\log2$. Random-network results establish appropriate large-bond regimes; finite bond dimension, competing cuts, and nonrandom tensors must be checked rather than assumed away.

These tests separate two failures: data incompatible with a proposed area rule, and data compatible with more than one reconstruction. Passing an entropy inequality is necessary for that model class, not sufficient to recover a unique geometry or its dynamics.

Productive research questions

Reconstruction stability. If entropies or correlators have finite error, how uncertain is the reconstructed geometry? A formal inverse map may be unusable if ill conditioned.

Missing information. Which nongeometric states reproduce the same selected entropies? Explicit counterexamples help determine which additional observables are necessary.

Quantum corrections. At finite system size or beyond leading semiclassical order, an area-based description becomes approximate. Quantifying its failure is more informative than treating a tensor-network picture as exact gravity.

Dynamics. A static graph reproducing an entropy pattern need not reproduce propagation, constraints, or gravitational equations. A geometry claim must state which of these are part of the target.

Van Raamsdonk’s work motivates the connection between entanglement and connected geometry. Harlow’s lectures explain the more precise holographic framework and its limitations. They support constrained research programs, not a model-independent identification of entanglement with spacetime.

An entry project: test uniqueness before reconstruction

Baseline. Reproduce the GHZ monogamy and cut-bound examples, then compare a small random-tensor network with its minimal-cut estimates. Use several random seeds and bond dimensions. Fix a graph class and a set of measured boundary regions before reconstructing anything.

Resources. A laptop, tensor contraction, singular-value decomposition, and elementary graph algorithms. Exact contraction of a small network is preferable to uncontrolled large-network approximations.

Deliverable. Find two networks or states with matching selected entropies but different correlators. A first check is that the two Bell states $(|00\rangle\pm|11\rangle)/\sqrt2$ have identical bipartite entropies and opposite $\langle X\otimes X\rangle$; this exposes local-basis ambiguity, not two inequivalent bulk geometries. The graph project must additionally identify transformations regarded as physically equivalent. Fit only a subset of region entropies, predict withheld regions, and add controlled input errors to quantify the remaining graph ambiguity.

Failure criterion. If the reconstruction succeeds only because the generating graph was supplied as an input constraint, do not claim model-independent recovery. A minimal-cut rule that fails for the chosen tensors must not be enforced by definition.

Research extension. Identify an additional observable set that makes a constrained reconstruction stable, or derive bounds on the ambiguity. State the graph and state families, error norm, and any minimality claim’s quantifiers. A first project can enumerate all small candidate graphs on a laptop; scaling to an unrestricted graph inverse problem is a separate task. Novelty lies in a new identifiability or stability result, not rediscovering that entropy discards phases.

What would count as progress?

A useful contribution gives a controlled dictionary, an error bound, or a sharp obstruction. A visually convincing network is insufficient. New evidence should change this assessment when it broadens a proven regime, makes reconstruction operationally robust, or distinguishes candidate geometries through independent observables.

This is a good direction for students combining quantum information, geometry, and computation, provided they are willing to learn the gravitational assumptions before interpreting toy models as spacetime.

References

• D. Harlow, “TASI Lectures on the Emergence of the Bulk in AdS/CFT,” Proceedings of Science, TASI2017, 002 (2018). arXiv:1802.01040.
• P. Hayden, M. Headrick, and A. Maloney, “Holographic Mutual Information is Monogamous,” Physical Review D 87, 046003 (2013). DOI — the conditional entropy inequality used above.
• P. Hayden, S. Nezami, X.-L. Qi, N. Thomas, M. Walter, and Z. Yang, “Holographic duality from random tensor networks,” Journal of High Energy Physics 2016(11), 009 (2016). arXiv:1601.01694 — controlled network assumptions and large-bond behavior.
• S. Ryu and T. Takayanagi, “Holographic Derivation of Entanglement Entropy from AdS/CFT,” Physical Review Letters 96, 181602 (2006). DOI — leading geometric entropy prescription.
• M. Van Raamsdonk, “Building up spacetime with quantum entanglement,” General Relativity and Gravitation 42, 2323–2329 (2010). arXiv:1005.3035.

Entry points

Reproduce one entropy/minimal-cut comparison, including a counterexample, before reading geometric reconstructions. Track which information is supplied by the model class and which is inferred from the data.