Baby Universes and Alpha Sectors

Baby-universe language packages multi-boundary gravitational amplitudes into expectation values of commuting operators on an auxiliary Hilbert space. Simultaneous eigenstates—alpha states—assign definite values to the effective couplings or boundary-creating operators. Amplitudes factorize when conditioned on a sharp alpha sector, while an unconditioned superposition or mixture can produce connected correlations. This is a precise mechanism in specified models, not automatic proof that every boundary theory has hidden random couplings.

Required background. Quantum Information supplies conditional states, commuting observables, superselection sectors, and covariance.

Helpful background. Semiclassical Gravity supplies topology-changing amplitudes; Quantum Fields in Curved Spacetime supplies generating functionals; Black-Hole Information supplies wormhole correlations; Quantum Gravity supplies the Hilbert-space and UV-completion questions.

Boundary-creating operators

Introduce a baby-universe Hilbert space $\mathcal H_{\mathrm{BU}}$ and an operator $\widehat Z[J]$ that creates an asymptotic boundary with source data $J$. A normalized state $\lvert\Psi\rangle\in\mathcal H_{\mathrm{BU}}$ assigns the multi-boundary amplitude

$$Z_\Psi[J_1,\ldots,J_n] = \langle\Psi\rvert \widehat Z[J_1]\cdots\widehat Z[J_n] \lvert\Psi\rangle.$$

In the simplest baby-universe construction, the $\widehat Z[J]$ form a commuting family closed under adjoints. With the required operator-domain and spectral assumptions, they admit a simultaneous spectral decomposition:

$$\widehat Z[J]\lvert\alpha\rangle =Z_\alpha[J]\lvert\alpha\rangle.$$

The labels $\alpha$ behave like fixed effective couplings because every boundary amplitude in that sector uses the same eigenvalues. The Hilbert-space construction, commuting observables, and alpha-state interpretation are developed in Marolf and Maxfield 2020, §§2–5.

Calling $\alpha$ a “coupling” is an interpretation of the eigenvalue data. One must still specify the state, the inner product, the operator domain, and whether distinct alpha values are superselected or can be coherently manipulated.

Conditional factorization

For a sharp normalized alpha state,

$$Z_\alpha[J_1,J_2] = \langle\alpha\rvert \widehat Z[J_1]\widehat Z[J_2] \lvert\alpha\rangle =Z_\alpha[J_1]Z_\alpha[J_2].$$

Thus conditional factorization is exact. For a state with probability density $p(\alpha)$ in the common eigenbasis,

$$Z_\Psi[J_1,J_2] = \int d\alpha\,p(\alpha) Z_\alpha[J_1]Z_\alpha[J_2],$$

and its connected part is the covariance of the eigenvalues. This is mathematically the same distinction as fixed sample versus ensemble average, but the probability distribution now arises from the baby-universe state.

For a continuous spectrum, a perfectly sharp $\lvert\alpha\rangle$ can be a generalized eigenstate rather than a normalizable Hilbert-space vector. Exact conditioning is then understood as a sector representation or an idealized limit. A normalizable wavepacket with finite spread generally leaves nonzero covariance. This distinction matters when translating formal sector factorization into an achievable preparation.

Repeated multi-boundary experiments can update knowledge of $\alpha$ or condition the state toward an eigen-sector. The operational replica interpretation and its distinction from a single-copy entropy are developed in Marolf and Maxfield 2021, §§2–6. Measurement language must not be used loosely: the accessible boundary observables and the preparation protocol determine whether such conditioning is physically available.

Because all boundary-creating operators are diagonal in the same basis, their expectation values depend on the diagonal weights $p(\alpha)=\lvert\langle\alpha\vert\Psi\rangle\rvert^2$. Phases between alpha components are invisible to this commuting family. Consequently, the observed amplitudes can look like a classical ensemble even when the underlying baby-universe state is pure. Distinguishing coherence would require another operator that does not commute with the boundary family, if the theory admits and operationally exposes one.

What alpha sectors do and do not resolve

Alpha sectors explain how a nonfactorizing unconditioned amplitude can coexist with factorizing conditional amplitudes. They also organize wormhole contributions as moments of shared eigenvalues. But they do not by themselves answer which alpha state a proposed boundary theory selects, whether the spectrum is continuous or discrete, or how the baby-universe Hilbert space arises from microscopic degrees of freedom.

There is a particularly sharp competing conjecture: in a consistent holographic quantum gravity, the baby-universe Hilbert space relevant to fixed AdS boundary conditions may be one-dimensional. Then every boundary-creating operator is proportional to the identity and factorization is automatic. Arguments for this “baby-universe hypothesis” and its swampland implications are given in McNamara and Vafa 2020, §§2–4. It is not a general theorem derived from the semiclassical path integral.

A multi-dimensional $\mathcal H_{\mathrm{BU}}$ with a sharp alpha state and a one-dimensional $\mathcal H_{\mathrm{BU}}$ can give identical factorized amplitudes in limited tests. Distinguishing them requires observables sensitive to sector variance, state preparation, or transitions—not just one-boundary partition functions.

The center of the observable algebra is therefore consequential. If alpha labels live in the center, local boundary operations cannot create coherent transitions between them, and conditioning resembles superselection. If additional physical operators connect sectors, treating $p(\alpha)$ as immutable classical ignorance would be wrong. Any claim of conditional factorization must name which algebra defines the center.

Results and open questions as of 2026-08-19

This discussion reflects results available through 2026-08-19.

Evidence or proposal	Controlled scope or assumptions	Limitation or adverse conclusion
Baby-universe operator algebra, Marolf and Maxfield 2020, §§2–5	Gravitational amplitudes represented on $\mathcal H_{\mathrm{BU}}$ with commuting boundary operators	Does not uniquely select the state or identify a microscopic boundary theory
Operational replicas and Page curves, Marolf and Maxfield 2021, §§2–6	Multi-copy permutation observables and conditioned baby-universe data	Swap entropy need not equal a single unknown state’s von Neumann entropy without assumptions
One-dimensional baby-universe hypothesis, McNamara and Vafa 2020, §§2–4	Holographic and swampland consistency arguments	A conjectured UV constraint, not established for every gravitational model

The ensemble-like position treats the unconditioned state as physically primary. The conditional position says a definite alpha sector describes the actual theory even if semiclassical calculations average over it. The one-dimensional position removes the sector multiplicity altogether. These competing positions attach different Hilbert spaces and operational meanings to the same connected amplitudes, and their general validity remains unresolved.

The adverse check is a two-sector model. If $Z_a[J]\ne Z_b[J]$ and both have nonzero probability, the connected two-boundary covariance is nonzero. Projecting onto $a$ makes it vanish immediately without changing either eigenvalue. Thus factorization follows from conditioning, not from cancelling the wormhole contribution inside the unconditioned state.

The conclusions could change after a microscopic theory derives the dimension and state of $\mathcal H_{\mathrm{BU}}$, after an operational protocol measures sector variance without ensemble assumptions, or after a consistency theorem proves or falsifies the one-dimensional baby-universe hypothesis for the relevant boundary conditions.

Exercises

Let $Z_a=1$, $Z_b=5$, and let the unconditioned state assign probability $p=1/4$ to $a$. Compute the two-boundary connected covariance, then condition on sector $b$.

Solution

For a two-valued variable, the variance is $p(1-p)(Z_a-Z_b)^2=(1/4)(3/4)(-4)^2=3$. This equals $\langle Z^2\rangle-\langle Z\rangle^2$, the connected two-boundary amplitude. After conditioning on $b$, $Z$ is deterministically $5$, so $\langle Z^2\rangle=25=\langle Z\rangle^2$ and the covariance vanishes. The eigenvalue did not change; only the state over sectors did.

References

• Marolf and Maxfield 2020, Donald Marolf and Henry Maxfield, “Transcending the Ensemble: Baby Universes, Spacetime Wormholes, and the Order and Disorder of Black Hole Information,” Journal of High Energy Physics 08 (2020) 044.
• Marolf and Maxfield 2021, Donald Marolf and Henry Maxfield, “The Page Curve and Baby Universes,” International Journal of Modern Physics D 30 (2021) 2142027.
• McNamara and Vafa 2020, Jacob McNamara and Cumrun Vafa, “Baby Universes, Holography, and the Swampland,” arXiv:2004.06738 (2020).