Initial-State and Boundary Proposals

The Wheeler–DeWitt constraint does not select one cosmological wave function. Initial-state and boundary proposals supply additional data through regularity, outgoing flux, reflection, a contour in lapse or metric space, or a dynamical selection principle. Their predictions cannot be compared until the integration cycle, saddle weights, perturbation state, inner product, and probability measure are stated in the same convention.

Helpful background. Flat-Spacetime Quantum Field Theory supplies path integrals and vacuum choices; Einstein Equations supplies cosmological saddles; Quantum Information supplies state and probability concepts; Quantum Gravity supplies evidential comparison; Causality and Global Structure supplies boundary interpretation; and Quantum Fields in Curved Spacetime supplies perturbation states.

A boundary condition is extra physics

For a minisuperspace coordinate $q$ and lapse $N$, a schematic wave function is

$$\Psi(q_1)=\int_{\mathcal C_N}dN \int_{q(0)\in\mathcal B}^{q(1)=q_1}\mathcal Dq\, e^{iS[q,N]}.$$

The boundary class $\mathcal B$ and complex contour $\mathcal C_N$ define the object. Deforming a contour without crossing singularities may leave it unchanged; choosing a different convergent cycle can select different saddles. Writing “sum over compact geometries” or “outgoing modes” does not specify the measure, gauge fixing, negative modes, or perturbation integration.

In a WKB region,

$$\Psi\approx C_+e^{iS}+C_-e^{-iS},$$

and the superspace current assigns opposite flux to the two terms. “Outgoing” is meaningful only after defining the boundary, normal orientation, and clock. A real standing wave has zero net current even though it contains expanding and contracting components.

No-boundary and tunneling proposals

The no-boundary proposal assigns amplitudes using regular compact geometries with only the final three-geometry as a boundary. In simple saddle treatments, a Euclidean cap joins a Lorentzian expanding history. Hartle and Hawking formulate the original prescription and its semiclassical probabilities Hartle and Hawking 1983, pp. 2961–2968.

The tunneling proposal imposes outgoing flux at singular boundaries of superspace, intending to describe a universe nucleating into expansion. Vilenkin gives a current-based formulation and compares semiclassical weights Vilenkin 1988, §§ II–IV. In elementary de Sitter minisuperspace, slogans about opposite signs of a Euclidean action often summarize the contrast, but those signs depend on contour, lapse convention, and which saddle contributes.

Neither proposal is merely a boundary value of a regular ordinary differential equation. The gravitational action is unbounded along some Euclidean directions, so a real Euclidean integral is not automatically convergent. Modern Lorentzian and Picard–Lefschetz analyses emphasize that the integration cycle determines which complex saddles and perturbation weights contribute Lehners 2023, §§ 3.6, 4.1 and appendix C.

Reflective and dynamical alternatives

A reflective condition sets current through a chosen minisuperspace boundary to zero. It can make the constraint operator self-adjoint on a selected domain and generate a standing wave. The physical meaning depends on whether the boundary represents zero volume, large anisotropy, a potential wall, or a breakdown of variables. Reflection avoids an outgoing-versus-incoming choice but does not determine the boundary’s location or measure.

Dynamical proposals seek an attractor, normalizable stationary state, decoherent ensemble, or stochastic equilibrium. They shift the selection question from boundary data to evolution, but still require a clock, state space, and measure. If the dynamics is imposed on a background time, it may not solve the timeless cosmological problem.

A density matrix can replace a pure wave function, and consistent-histories probabilities can replace a naive $|\Psi|^2$. That extension may regulate interference between branches, yet the decoherence functional and coarse graining become part of the proposal.

Measures and perturbative stability

A semiclassical weight for a background is not an observational probability until conditionalization and the measure over histories are fixed. Volume weighting, observer selection, and conditioning on late data can overwhelm the saddle weight. These are distinct from mathematical contour choices and should not be folded into one unexplained “probability.”

Perturbations provide an adverse check. A background saddle that appears regular can weight fluctuations with an inverse Gaussian, making short modes unsuppressed. The result may change when the contour, boundary term, or allowed off-shell geometries change. Stability must be demonstrated for gauge-invariant perturbations with their normalization and final boundary data specified.

Results and open questions through 2026-08-20

This discussion reflects results available through 2026-08-20. No-boundary, tunneling, reflective, and dynamical proposals remain genuinely competing. Each is mathematically concrete in selected models; none has a universally accepted gravitational path integral, probability measure, and perturbation prescription. The general selection problem remains unresolved.

Proposal or result	Evidence at the cutoff	Scope and limitation
No-boundary saddles	Regular compact complex solutions yield calculable semiclassical wave functions	Integration cycle, negative modes, and perturbation stability are disputed
Tunneling flux	Outgoing-current conditions sharply select WKB branches in minisuperspace	Superspace boundary, clock orientation, and measure are model-dependent
Reflective domain	Self-adjoint boundary conditions conserve probability current	Physical reason for reflection and full-theory extension remain open
Dynamical selection	Attractors, decoherence, or stochastic states can suppress alternatives	Requires a dynamics, time variable, coarse graining, and normalizable measure

Some authors regard Lorentzian contour analysis as selecting the relevant no-boundary cycle; others argue that different admissible cycles define different states rather than one correction. A third position treats the proposals as effective initial-state models to be compared observationally. Conclusions depend on the contour, topology class, perturbation boundary condition, and conditional measure.

The conclusions could change if a nonperturbative gravitational path integral fixes a unique integration cycle and positive physical inner product, or if a theorem excludes one proposal’s perturbative state; the interpretation could also change with an observable likelihood that discriminates the boundary states after the same conditioning and foreground assumptions.

Exercise

At a large-volume boundary, let $\Psi=C_+e^{iS}+C_-e^{-iS}$ with each branch normalized to current magnitude $j_0$. Compute the net outward current and state the conditions represented by pure outgoing and reflective choices.

Solution

Neglecting rapidly oscillating interference in the WKB regime, the two phase branches carry opposite currents. With $e^{iS}$ defined as outward,

$$j_{\mathrm{out}}=j_0\left(|C_+|^2-|C_-|^2\right).$$

Pure outgoing flux sets $C_-=0$. A reflective condition requires zero net current, so $|C_+|=|C_-|$, with the relative phase fixed by the chosen self-adjoint boundary condition. Equal magnitudes do not mean the state lacks expanding and contracting components; it is a standing-wave combination. Reversing the normal or clock convention swaps the labels but not the flux calculation.

References

• Hartle, James B., and Stephen W. Hawking (1983), “Wave Function of the Universe,” Physical Review D 28, 2960–2975, DOI: 10.1103/PhysRevD.28.2960.
• Lehners, Jean-Luc (2023), “Review of the No-Boundary Wave Function,” Physics Reports 1022, 1–82, arXiv:2303.08802.
• Vilenkin, Alexander (1988), “Quantum Cosmology and the Initial State of the Universe,” Physical Review D 37, 888–897, DOI: 10.1103/PhysRevD.37.888.