Factorization, Ensembles, and Gravitational Path Integrals

Disconnected boundaries create a sharp diagnostic for a quantum theory of gravity: a fixed boundary Hamiltonian gives a product partition function, whereas a Euclidean wormhole can give a connected contribution. At the 2026-08-20 evidence cutoff, ensemble averaging and conditional baby-universe sectors explain this tension in controlled models, especially JT gravity. They do not yet supply a universally accepted nonperturbative interpretation for every gravitational path integral or every fixed holographic theory.

Helpful background. Factorization, Ensembles, and Gravitational Path Integrals develops the topology expansion and the distinction between fixed-theory products and averaged correlators.

The factorization test

For one specified Hamiltonian $H$,

$$Z_H(\beta_1)Z_H(\beta_2)$$

is exactly the partition function of two noninteracting copies. There is no connected covariance to explain. If instead one averages over Hamiltonians,

$$\left\langle Z(\beta_1)Z(\beta_2)\right\rangle -\left\langle Z(\beta_1)\right\rangle \left\langle Z(\beta_2)\right\rangle$$

can be nonzero. A connected bulk saddle is therefore naturally compatible with an ensemble-valued object, but its mere existence does not prove that the microscopic boundary theory is fundamentally random.

Assessment at 2026-08-20	Meaning
Question status	Open
Evidence status	Contested overall; established in JT/ensemble and conditional-sector models
Confidence	Moderate overall; high for the fixed-theory/average distinction
Domain	Specified topology sum, contour, source category, boundary theory, and state or alpha sector
What could change the conclusion	Validation of the 2026 theorem in a UV-complete gravity example

A starting project: calculate one connected amplitude

This topic suits readers who can evaluate Gaussian integrals, distinguish connected from disconnected correlators, and track what an ensemble average means. Start with the linked factorization background, then read Saad, Shenker, and Stanford 2019, §3.4.1, the double-trumpet calculation. Work backward to the trumpet partition function and gluing measure, rather than treating the two-boundary answer as an unexplained wormhole rule.

Evaluate the gluing integral independently for positive boundary inverse temperatures. Check symmetry under interchanging the boundaries, normalization conventions, and agreement between analytic integration and numerical quadrature. Then construct a small explicitly specified random-matrix ensemble and compare an averaged product of partition functions with the product of averages. This second exercise illustrates connected fluctuations; a finite Gaussian matrix ensemble is not automatically the double-scaled JT model, so do not claim numerical verification of that dictionary.

Produce the integral derivation, a notebook with numerical checks, and a clear statement of the averaging operation in each calculation. A laptop is adequate. Progress toward gravitational topology sums requires more training in moduli-space integration and random matrices, while fixed-theory reconstruction also benefits from quantum-information or mathematical-physics supervision.

A candidate extension is to study how a specified conditioning or smoothing prescription changes a connected observable in a model with a fully stated measure. Check existing work and the range of the dictionary before claiming novelty. Pivot if the calculation silently alternates between one Hamiltonian and an ensemble, or if a nonperturbative completion is needed but unspecified. The finite result should say exactly which amplitude has been reproduced and why it need not equal a product of one-boundary averages; it should not claim to settle factorization in every holographic theory.

JT gravity as the controlled benchmark

Saad, Shenker, and Stanford showed that the JT-gravity genus expansion is reproduced order by order by a double-scaled random-matrix integral Saad, Shenker, and Stanford 2019, §§2–5. This is a powerful nonperturbative organizing model: multi-boundary wormholes calculate ensemble correlations, and the matrix integral supplies a completion of the asymptotic topological expansion. It is also explicit about being ensemble-valued. Choosing one matrix realization produces a fixed spectrum but is extra information not encoded by the smooth genus expansion alone.

The lesson is not that all AdS/CFT duals are ensembles. It is that a gravitational sum over topologies may compute less state-specific information than a named boundary Hamiltonian. The comparison must be made observable by observable and at the same smoothing scale.

Results and open questions at 2026-08-20

Evidence	Result	Secure scope	Interpretive limit
JT topology expansion and matrix integral	Order-by-order match including multi-boundary correlations	Two-dimensional JT gravity and its spectral ensemble	Does not select one nonperturbative Hamiltonian
Baby-universe alpha sectors	Conditional amplitudes can factorize	A chosen simultaneous eigenstate/sector	Requires a physical rule selecting the sector
Spectral form-factor ramp	Random-matrix correlations after smoothing	Late-time spectral statistics	Not unique evidence for literal ensemble ontology
Functorial reconstruction theorem	Under five axioms, closed-manifold data reconstruct a compact unitary QFT	The stated source category and preparable sector	Gravity must satisfy the axioms and include charged sectors appropriately

The spectral ramp is an especially useful warning. Cotler and collaborators showed that black-hole spectra exhibit random-matrix-like correlations after ensemble, time, or energy smoothing Cotler et al. 2017, §§2–4. A single chaotic Hamiltonian can display such statistics. The ramp therefore supports universal spectral correlations but does not uniquely identify the averaging operation represented by a wormhole.

Conditioning on baby-universe data

Marolf and Maxfield formulate baby-universe observables whose simultaneous eigenstates, often called alpha states, condition the gravitational amplitudes Marolf and Maxfield 2020, §§2–5. An unconditioned state can correlate disconnected boundaries, while conditional amplitudes in a fixed sector factorize. This turns an apparent contradiction into a question about which state the path integral prepares.

Conditioning is an organizational mechanism, not yet a complete microscopic solution. A boundary theory needs a principle choosing the alpha sector, and the available boundary sources must be rich enough to prepare all physical states under discussion. Otherwise factorization may hold only in an invariant or restricted sector while charged states remain unseen.

The 2026 reconstruction result

McNamara and Wang prove that, under finiteness, reality, multiplicativity, continuity, and reflection-positivity assumptions, closed-manifold partition functions reconstruct a unique compact unitary functorial QFT McNamara and Wang 2026, Theorem 1.1 and §§4–5, 9, 11. Their analysis reframes many wormhole correlations as insufficient source data rather than proof of a fundamental ensemble.

This is a substantial theorem-level development but remains a preprint at the cutoff. Applying it to quantum gravity requires constructing the correct bordism or source category, verifying every axiom, and determining whether manifold-preparable states exhaust the physical Hilbert space. In particular, invariant data need not reconstruct charged sectors without suitable defects or boundaries.

Counterexamples, null results, and failed routes

The decisive counterexample to “a ramp proves an ensemble” is one fixed chaotic spectrum exhibiting the same smoothed statistics. The decisive counterexample to “a wormhole violates factorization in one theory” is that the bulk object may be an average or an unconditioned sector expectation rather than $Z_H(\beta_1)Z_H(\beta_2)$ for one $H$. Comparisons that switch objects midway are not paradoxes.

Other failed routes include treating an asymptotic genus series as a unique nonperturbative definition, ignoring negative modes or contour choices, and claiming that alpha-state conditioning selects a sector without specifying a preparation rule. Reflection positivity and source completeness are substantive conditions, not cosmetic axioms.

What would change the assessment

A fixed-theory microscopic construction that reproduces connected gravitational amplitudes, including nonperturbative corrections, while preserving factorization of truly independent boundaries would strongly advance the question. So would an explicit UV-complete gravitational source category satisfying the 2026 reconstruction hypotheses and demonstrably preparing all charged and invariant sectors.

The assessment would weaken if independent analysis found a failed axiom or hidden nonuniqueness in the reconstruction theorem, or if the relevant wormhole saddles proved absent from the correct contour. Any claim must identify the boundary observable before assigning an ensemble, sector, or fixed-theory interpretation.

Research checklist

• State whether $H$ is fixed, sampled, energy-averaged, or time-smoothed.
• Define the multi-boundary observable before comparing bulk and boundary answers.
• Specify topology weights, contour, negative modes, and nonperturbative completion.
• Separate unconditional baby-universe expectation values from alpha-sector amplitudes.
• Test reflection positivity, source completeness, and charged-state preparation.
• Distinguish a spectral universality class from a literal microscopic ensemble.
• Compare like with like: the same theory, sources, sector, and smoothing operation.

Source and data map

JT gravity and its matrix integral provide the controlled positive example; baby-universe sectors provide the conditional-factorization mechanism; spectral statistics provide positive but nonunique evidence; and the 2026 theorem supplies the newest structural constraint. No independent reproduction was completed for this assessment. A reproduction should publish the source category, axiom checks, topology sum, contour, sector projection, and a fixed-versus-averaged comparison of identical observables.

References

• Cotler, J. S., et al. “Black Holes and Random Matrices.” Journal of High Energy Physics 05 (2017) 118. doi:10.1007/JHEP05(2017)118.
• Marolf, D., and H. Maxfield. “Transcending the Ensemble.” Journal of High Energy Physics 08 (2020) 044. doi:10.1007/JHEP08(2020)044.
• McNamara, J., and Z. Wang. “Wormholes as Red Herrings: Reflection Positivity and the Reconstruction of Unitary Quantum Field Theories.” arXiv:2607.01322v1 (2026). arXiv.
• Saad, P., S. H. Shenker, and D. Stanford. “JT Gravity as a Matrix Integral.” arXiv:1903.11115v1 (2019). arXiv.