Bounce and Singularity-Resolution Claims

A bounce is a minimum of a declared quantity with respect to a declared time variable. It may be a minimum expectation value of volume or a turning point of an effective scale factor. A regular no-boundary saddle is not a bounce: it is a distinct state-preparation proposal often discussed under singularity avoidance. These results are valuable, but they are not interchangeable with finite curvature, extendible causal geodesics, unique evolution, stable inhomogeneities, or derivation from a full quantum-gravity theory. This page supplies a common set of tests and shows exactly where the logical implications fail. Research-status statements are checked through 30 August 2026.

Required background. No-Boundary and Tunneling Wavefunction Proposals supplies the complex-saddle, contour, and probability questions used below. Loop Quantum Cosmology and Effective Difference Dynamics supplies the constraint, physical inner product, scalar clock, and difference and effective dynamics.

Helpful background. Cauchy Horizons, Chronology Horizons, and Loss of Global Hyperbolicity, Semiclassical States, Continuum Limits, and Classical Recovery, and Singularities, Initial Conditions, and Predictive Limits supply the classical and semiclassical criteria.

What a bounce establishes

The word bounce becomes precise only after four choices are stated:

1. the theory and truncation;
2. the state or family of solutions;
3. the clock or evolution parameter;
4. the observable that reaches a nonzero minimum.

For example, solvable loop quantum cosmology (sLQC) treats a spatially flat, homogeneous Friedmann–Lemaître–Robertson–Walker (FLRW) model with a free massless scalar $\phi$ as relational time. For a normalized physical state in the volume-operator domain, the volume expectation value has the form

$$\langle \hat V\rangle_\phi =V_+e^{\alpha\phi}+V_-e^{-\alpha\phi}, \qquad \alpha=\sqrt{12\pi G}, \qquad V_\pm>0.$$

It therefore has the nonzero minimum

$$\langle \hat V\rangle_{\min}=2\sqrt{V_+V_-}.$$

The same model supplies a physical inner product, unitary positive-frequency evolution in $\phi$, and the spectral bound

$$\rho_{\mathrm{sup}} =\frac{3}{8\pi\gamma^2G\lambda^2} \approx0.41\rho_{\mathrm{Pl}}$$

Here $\gamma$ is the Barbero–Immirzi parameter and $\lambda$ is the sLQC discreteness length set by the area-gap convention. The bound is a supremum of the density spectrum on the physical Hilbert space, not the expectation-value bounce density of every state; widely dispersed states can bounce at lower density. The volume formula and spectral bound are exact statements inside that model, not merely properties of a fitted effective trajectory Ashtekar, Corichi, and Singh 2008, § IV, Eqs. (4.4)–(4.5), and § V.A, Eqs. (5.1)–(5.5).

The result does not yet identify a four-dimensional effective metric across the quantum regime, prove that all curvature or tidal observables are finite in other matter models, or control discarded anisotropic and inhomogeneous modes. A bounce of $\langle\hat V\rangle_\phi$ licenses a relational-observable claim; every spacetime claim needs additional data.

Singularity criteria form a partial order

Classical singularity theorems conclude that at least one causal geodesic is incomplete under their stated hypotheses. They do not require every curvature scalar to diverge, and they do not prescribe a microscopic continuation Hawking and Penrose 1970, abstract and theorem, pp. 529 and 537–538. In a quantum regime there may be no single classical metric at all, so the proposed replacement criterion must be named rather than assumed.

Two criteria can be ordered only after both claims’ theory, state or solution domain, observables, quantifiers, regularity class, and conventions have been defined in a common model. Under those stated hypotheses, call criterion $C_2$ stronger than $C_1$ precisely when $C_2$ logically entails $C_1$; otherwise the criteria are incomparable. The useful branches are these.

Quantum-state branch

• Constraint solution. A formal wavefunction or distribution solves the declared constraint.
• Physical state. The solution is normalizable in a positive physical inner product on a stated operator domain.
• Relational continuation. A declared clock and self-adjoint generator give norm-preserving evolution on a stated interval.
• Observable bounce. A specified relational observable has a positive minimum for a declared state class.

Relational continuation does not imply that the chosen observable bounces, and an observable bounce does not supply an effective spacetime.

Effective-geometry branch

• Metric and regularity class. State the effective metric and whether the claim concerns a $C^0$, $C^1$, $C^2$, or smoother geometry.
• Pointwise curvature and tides. Test scalar invariants and parallel-propagated tidal components relevant to the claimed observers. The latter are Riemann-tensor components measured in an orthonormal frame carried without rotation along the observer’s geodesic. Bounded $a$, $H$, volume, or density does not settle this test.
• Local causal extension. Determine whether timelike and null geodesics can cross the event in a specified differentiability-class extension. Roughly, the geodesic equation uses the metric and its first derivatives, whereas ordinary pointwise curvature uses second derivatives; this difference permits extendible geodesics even when curvature diverges.
• Completeness. Determine whether every inextendible causal geodesic in the resulting maximal spacetime has unbounded proper or affine parameter. Crossing one event locally does not establish this global condition.
• Existence and uniqueness of extension. Ask whether the dynamics selects one continuation without extra boundary data and whether global hyperbolicity survives. The existence of at least one extension does not establish uniqueness.

These steps are mostly incomparable. Bounded polynomial invariants do not guarantee bounded parallel-propagated tides or completeness; a local extension is not completeness; and the existence of an extension is not its uniqueness.

Robustness and full-theory branch

Anisotropy, spatial gradients, perturbations, and backreaction must remain controlled on an open set of admissible data. Covariance or anomaly freedom must be demonstrated for the structure actually used to infer spacetime. Finally, a full-theory claim requires a controlled embedding, derivation, or convergence and recovery of the low-curvature regime. These tests enlarge the quantified domain or strengthen the evidence; they are not automatic later consequences of a homogeneous bounce.

State-preparation branch

A regular complex saddle can define candidate boundary-state data. A path-integral representation additionally needs its contour, contributing saddles, intersection numbers, and fluctuation determinants; a probability claim then needs normalization and a physical measure or decoherent-histories rule. This branch does not imply Lorentzian continuation through a singularity.

“Singularity resolution” should therefore be followed by the branch and criterion actually established.

Compare two quantum-cosmology proposals criterion by criterion

The following comparison uses the same questions for two unlike constructions. Each proposal cell begins with its status so that absent structure cannot be mistaken for negative evidence. On a narrow screen, every row reflows into a labeled record.

When the table is wider than the available space, use the Left and Right arrow keys to pan; Home and End move to its edges. On phone-sized screens, each row becomes a labeled record.

What each proposal establishes, conditions, or leaves undefined

Test	Flat massless-scalar sLQC	No-boundary saddle	Strongest comparison claim
Object and domain	Established in the reduced model. Physical states of improved-dynamics, spatially flat FLRW sLQC; observables include relational volume and matter density.	Established at the semiclassical-definition level. A wavefunction of a boundary three-geometry is built from a weighted set of selected regular complex saddles. If a path integral is invoked, its contour fixes the contributing set.	The objects and domains are different; they do not define one shared “bounce object.”
Clock and continuation	Established. The massless scalar φ is monotonic and supplies unitary positive-frequency relational evolution on the chosen self-adjoint domain.	Not applicable to the saddle alone. Regularity removes an initial boundary in the complex geometry; it does not supply a Lorentzian clock or evolution through a big-bang surface.	sLQC establishes relational continuation. Saddle regularity is a state-preparation statement, not a weaker version of the same result.
Inner product and probability	Established. A physical inner product is part of the construction, so its expectation values and density bound are physical-model statements.	Conditional. A saddle weight is an amplitude. A positive normalized probability still needs a measure, physical inner product or decoherent-histories rule, and a declared coarse graining.	Only probabilities formed with the declared physical product or history rule are licensed.
Local regularity	Established for selected observables. Relational volume has a nonzero minimum and the density spectrum is bounded. Curvature regularity in other effective models is a separate question.	Established for the selected complex saddle. It can be regular at its closing point, but that does not establish finite Lorentzian curvature for every inferred classical history.	Each proposal passes a different local criterion; neither result implies generic curvature regularity.
Causal geometry	Not established by the bounce theorem. Geodesic and proper-time claims require an effective metric compatible with the dynamics.	Not defined on the complex saddle. A real Lorentzian branch and its domain must be justified before causal geodesics can be discussed.	Neither entry alone establishes causal completeness or unique extension.
Perturbations and inhomogeneity	Not established generically. Homogeneous exact solvability does not prove stability under BKL modes, gradients, or generic nonlinear perturbations.	Conditional and contour-dependent. Thimbles and fluctuation weights depend on the contour and boundary prescription; inverse-Gaussian weights are a failure mode for some Lorentzian contours.	A broad claim remains conditional until the retained modes and backreaction pass their own controls.
Full theory	Not established. The reduced dynamics has not been systematically derived from full loop quantum gravity, and quantization ambiguities remain relevant.	Not established. A contour-independent, nonperturbative definition on full superspace with an agreed probability interpretation is unavailable.	Neither proposal currently licenses full-theory singularity resolution.

The sLQC entry is an exact symmetry-reduced quantum result with structural uncertainty in its generalization. The no-boundary entry is a semiclassical state-preparation proposal with saddle, fluctuation, and measure uncertainty. Hartle and Hawking define the compact regular proposal in Hartle and Hawking 1983, § I, Eq. (1.7), and § III, Eq. (3.1). A direct saddle definition and its classical-history interpretation are developed in Halliwell, Hartle, and Hertog 2019, § I and § IV.A, Eqs. (4.16)–(4.17), Open PDF. When a Lorentzian path-integral contour is used, a different thimble and problematic fluctuation sign can result Feldbrugge, Lehners, and Turok 2017, § III.B.1, Eq. (48), and § V, pp. 38–39, Open PDF. These constructions answer different questions, so a single numerical “resolution score” would be misleading.

A bounce can coexist with divergent curvature

A short counterexample exposes the most common false implication. Consider a spatially flat FLRW line element and the local scale factor

$$ds^2=dt^2-a^2(t)d\mathbf x^2, \qquad a(t)=a_b+A\lvert t\rvert^q, \qquad a_b,A>0, \qquad 1<q<2.$$

This is an adversarial metric, not a proposed microscopic LQC solution. It has a nonzero minimum at $t=0$, with

$$\dot a=Aq\,\operatorname{sgn}(t)\lvert t\rvert^{q-1}, \qquad \ddot a=Aq(q-1)\lvert t\rvert^{q-2}, \qquad \dot H=\frac{Aq(q-1)\lvert t\rvert^{q-2}}{a}-H^2$$

away from $t=0$. Hence $H=\dot a/a\to0$, while $\ddot a$ and $\dot H$ diverge. With the site’s (+---) metric and curvature convention,

$$\begin{aligned} R&=-6(\dot H+2H^2) \sim-\frac{6Aq(q-1)}{a_b}\lvert t\rvert^{q-2},\\ R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma} &=12\left[(\dot H+H^2)^2+H^4\right] \sim12\left[\frac{Aq(q-1)}{a_b}\right]^2\lvert t\rvert^{2q-4}. \end{aligned}$$

Both invariants diverge. If ordinary flat-FLRW Einstein equations are used to define the stress tensor, then $\rho=3H^2/(8\pi G)\to0$ while $p=-(2\dot H+3H^2)/(8\pi G)\to-\infty$. Thus the geometry fails pointwise curvature and tidal regularity even though $a$, $H$, and $\rho$ are bounded. Barrow’s global sudden-singularity solutions exhibit the same separation between finite scale factor and expansion rate and divergent acceleration Barrow 2004, Eqs. (2)–(14), pp. L79–L82.

The metric is nevertheless $C^1$. For an affine parameter $\lambda$, causal type $\varepsilon=1$ (timelike) or $0$ (null), and conserved comoving momentum $P^i$, the flat-FLRW first integrals can be written

$$\left(\frac{dt}{d\lambda}\right)^2 =\varepsilon+\frac{P^2}{a^2}, \qquad \frac{dx^i}{d\lambda}=\frac{P^i}{a^2}.$$

They contain $a$ but not $\ddot a$, so their first-order data remain finite and causal geodesics can be continued across the event Fernández-Jambrina and Lazkoz 2004, Eqs. (5)–(8), pp. 1–2, Open PDF. This local result does not prove uniqueness among all possible $C^1$ extensions. The standard strong-curvature diagnostics integrate an appropriate parallel-propagated curvature contraction once along a causal geodesic for the Królak test and twice for the Tipler test Singh 2012, § IV.A, Eqs. (41)–(45). For a comoving orthonormal tidal component here, the divergent factor behaves as $\lvert t\rvert^{q-2}$, but its single proper-time integral is finite:

$$\int_0^\epsilon t^{q-2}\,dt =\frac{\epsilon^{q-1}}{q-1}<\infty.$$

The corresponding Tipler double integral is finite as well. Thus this event is weak under both tests even though pointwise curvature and tidal components diverge. It passes “nonzero minimum,” “finite density,” “local causal extension,” and “weak integrated tides”; it fails pointwise curvature regularity and says nothing about extension uniqueness, quantum-state evolution, inhomogeneous stability, or a full theory. One example therefore disproves two tempting implications at once:

$$a_{\min}>0\ \nRightarrow\ \text{finite curvature}, \qquad \text{geodesic extension}\ \nRightarrow\ \text{finite curvature}.$$

LQC evidence and counterevidence

The exact sLQC result should not be weakened: within its physical Hilbert space and declared scalar clock, the relational bounce and density bound are analytic results. The broad claim should not be strengthened either. Effective LQC with particular scalar potentials can have bounded $H$ while the Ricci scalar diverges Cailleteau et al. 2008, pp. 2–4, Open PDF.

The distinction is not simply “divergence equals failure.” In effective Bianchi I LQC with vanishing anisotropic stress, density, expansion, and shear can be bounded while curvature invariants remain capable of diverging. Finite-volume pressure singularities in that analysis are weak and geodesically extendible, whereas perfect fluids with finite $w>-1$ satisfy stronger boundedness results Singh 2012, §§ III–IV, especially Eqs. (32)–(35) and § IV.B. The hypotheses determine which criterion survives.

Current contrary evidence adds a different failure mode. In one covariance-compatible emergent-metric completion, Bojowald, Díaz, and Duque find a bounce at $\bar\rho=\rho_Q/8$ but a Ricci divergence at $\bar\rho=\rho_Q/2$, reached at finite proper time Bojowald, Díaz, and Duque 2026, §§ 3.2–3.3, Eqs. (52)–(61). Their § 4 gives another covariance-compatible completion with the same background canonical dynamics but a nonsingular reconstructed background metric Bojowald, Díaz, and Duque 2026, § 4, Eqs. (62)–(65) and discussion after Eq. (65). A perturbative coefficient in that construction diverges near the bounce, so regular evolution of the inhomogeneous perturbations is not established. The adversarial pair shows that identical background canonical trajectories can yield singular and nonsingular reconstructed background metrics in their emergent-modified-gravity framework; it does not provide two fully controlled inhomogeneous completions. This published 2026 result is model-specific counterevidence to “background bounce implies curvature or geodesic resolution.” It does not overturn the exact relational-observable theorem in sLQC, and it does not establish that every LQC construction is singular.

The right conclusion is conditional: some symmetry-reduced LQC models rigorously resolve selected relational-observable and strong-singularity criteria; other matter choices, effective metrics, covariance implementations, or quantization choices can fail curvature or finite-proper-time tests. A publication-level claim names the model and criterion on both sides of that sentence.

An adversarial resolution protocol

Apply the following protocol before promoting any bounce or regular-saddle result.

1. Freeze the claim. Record the theory, truncation, state, clock, inner product or measure, operator domain, effective metric if one exists, observable, and parameter regime.
2. Name the criterion. Decide whether the target is bounded observables, state continuation, curvature control, weak or strong tidal behavior, causal completeness, unique prediction, stability, or full-theory completion.
3. Change one vulnerable input. Vary the factor ordering, polymerization, self-adjoint domain, clock, matter potential, contour, or boundary condition. Then add shear, a gradient mode, or perturbative backreaction without changing the advertised observable.
4. Compute the matching diagnostic. Test norms and expectation values in the physical product; curvature in a parallel-propagated frame; causal curves in proper or affine parameter; and perturbations with constraint residuals and convergence checks. If the model supplies no effective metric, do not manufacture a geodesic conclusion.
5. Separate a weak event from a terminal one. A divergent curvature component may have finite tidal integrals and extendible geodesics. Conversely, incomplete geodesics can occur without divergent polynomial invariants. Evaluate the criterion that the claim actually names.
6. Report the surviving statement. A useful form is: “Under hypotheses $\mathcal H$, observable $\mathcal O$ remains controlled in domain $\mathcal D$; this establishes $X$ but does not establish $Y$.” Give the exact changed assumption responsible for every downgrade.

The most important inhomogeneous adversary is developed on BKL, Mixmaster, and Inhomogeneous Singularities. There, shear and curvature walls are dynamical sectors rather than decorative perturbations.

Common pitfalls

A nonzero scale factor means no singularity. A nonzero $a$ excludes one familiar route to a big bang, but $\dot H$, pressure, curvature, or tidal components may still diverge. The local counterexample above makes the distinction explicit.

A continued wavefunction means unitary physical evolution. A formal solution can continue while its physical inner product, probability rule, clock, or operator domain remains unspecified. Continuation becomes a physical claim only after those structures are supplied.

Finite curvature means deterministic continuation. Curvature bounds do not by themselves establish geodesic completeness, global hyperbolicity, or uniqueness of an extension. Check causal curves and the data needed beyond the candidate boundary.

One countermodel refutes an entire program. A counterexample refutes the universal implication it violates. It does not erase a theorem proved under different hypotheses, so every negative conclusion must retain the countermodel’s matter content, quantization, covariance assumptions, and regime.

Exercises

Locate the regularity thresholds

Keep $a(t)=a_b+A\lvert t\rvert^q$ with $a_b,A>0$, but compare $q=3/2$ with $q=5/2$. For each case, determine the leading behavior of $H$, $R$, and the Kretschmann scalar. Decide whether causal geodesics cross $t=0$ locally and whether the Królak and Tipler tidal integrals converge. Which value of $q$ separates divergent from finite pointwise curvature?

Solution

For any $q>1$, the derivative changes from negative to positive and

$$H\sim\frac{Aq}{a_b}\operatorname{sgn}(t)\lvert t\rvert^{q-1}\to0,$$

while

$$\begin{aligned} R&\sim-\frac{6Aq(q-1)}{a_b}\lvert t\rvert^{q-2},\\ R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma} &\sim12\left[\frac{Aq(q-1)}{a_b}\right]^2\lvert t\rvert^{2q-4}. \end{aligned}$$

For $q=3/2$, $H=O(\lvert t\rvert^{1/2})$ tends to zero, but $R=O(\lvert t\rvert^{-1/2})$ and the Kretschmann scalar is $O(\lvert t\rvert^{-1})$, so pointwise curvature diverges. The relevant parallel-propagated tidal components scale as $\lvert t\rvert^{-1/2}$; their single integral is proportional to $\lvert t\rvert^{1/2}$ and their double integral also converges. The event is therefore weak under both integral tests.

For $q=5/2$, $H=O(\lvert t\rvert^{3/2})$, $R=O(\lvert t\rvert^{1/2})$, and the Kretschmann scalar is $O(\lvert t\rvert)$, so all three tend to zero. In both cases $a$ is positive and $C^1$, and the first-order geodesic data remain finite; causal geodesics therefore admit a local continuation. The pointwise-curvature threshold is $q=2$: curvature diverges for $1<q<2$, is finite at $q=2$, and tends to zero for $q>2$. The single tidal integral already converges for every $q>1$, so finite integrated tides and finite pointwise curvature have different thresholds. None of these local statements proves completeness or uniqueness of extension.

Downgrade a relational claim correctly

A calculation proves unitary evolution in a scalar clock, a nonzero minimum of $\langle\hat V\rangle_\phi$, and a bounded density spectrum. It does not define an effective metric in the high-curvature regime. What is the strongest justified singularity-resolution statement, and which claims remain unavailable?

Solution

The calculation establishes physical-state continuation and boundedness of the declared relational observables on the stated Hilbert-space domain. It may accurately be called relational-observable singularity resolution in that model. Curvature regularity, tidal strength, proper-time or affine completeness, and unique spacetime extension are unavailable because each requires an effective metric. Stability against omitted anisotropic and inhomogeneous sectors and derivation from the full theory also remain separate tests.

Test shear during contraction

In a classical contracting regime, let a perfect-fluid density scale as $\rho_w\propto a^{-3(1+w)}$ and the shear contribution as $\sigma^2\propto a^{-6}$. Find the scaling of $\sigma^2/\rho_w$. For which $w$ does initially small shear grow relative to the fluid as $a$ decreases?

Solution

The ratio is

$$\frac{\sigma^2}{\rho_w}\propto a^{-6+3(1+w)}=a^{-3(1-w)}.$$

During contraction, $a$ decreases. The ratio grows for $w<1$, stays constant for $w=1$, and decreases for $w>1$. An isotropic bounce with ordinary matter therefore needs an explicit shear-stability test; exact isotropy cannot be assumed to remain representative. This scaling is a classical entrance diagnostic, not a substitute for evolving the quantum-corrected anisotropic system.

Turn a saddle weight into a probability claim

Suppose a semiclassical calculation gives two saddle contributions $\Psi_s\sim A_s e^{-I_s}$. Explain why comparing $\lvert\Psi_1\rvert^2$ and $\lvert\Psi_2\rvert^2$ is not yet a normalized probability prediction.

Solution

The definition must first determine which saddles contribute and with what weights and phases; if they came from a path integral, that includes the contour and its thimbles. One must then specify the physical inner product or history measure, the variables being conditioned on, the coarse graining, and—if histories interfere—a decoherence condition. The prefactors $A_s$, fluctuation determinants, negative modes, and normalization over the allowed configuration space also matter. Without those ingredients, the calculation supplies semiclassical amplitudes or relative weights, not a unique Born probability for cosmological histories.

The chapter overview contains the structure diagram and validity and failure diagram. They are embedded there once so that their shared chapter-level context is not repeated on every article.

For the chapter-wide comparison of assumptions, counterevidence, falsifiers, and claim ceilings, see the claim-domain table.

References

• Ashtekar, A., A. Corichi, and P. Singh. “Robustness of Key Features of Loop Quantum Cosmology.” Physical Review D 77 (2008): 024046. DOI. Open PDF.
• Barrow, J. D. “Sudden Future Singularities.” Classical and Quantum Gravity 21 (2004): L79–L82. DOI. Open PDF.
• Bojowald, M., M. Díaz, and E. I. Duque. “Singularities in Loop Quantum Cosmology.” Physical Review D 113 (2026): 084060. DOI. Open PDF.
• Cailleteau, T., A. Cardoso, K. Vandersloot, and D. Wands. “Singularities in Loop Quantum Cosmology.” Physical Review Letters 101 (2008): 251302. DOI. Open PDF.
• Feldbrugge, J., J.-L. Lehners, and N. Turok. “Lorentzian Quantum Cosmology.” Physical Review D 95 (2017): 103508. DOI. Open PDF.
• Fernández-Jambrina, L., and R. Lazkoz. “Geodesic Behavior of Sudden Future Singularities.” Physical Review D 70 (2004): 121503(R). DOI. Open PDF.
• Halliwell, J. J., J. B. Hartle, and T. Hertog. “What Is the No-Boundary Wave Function of the Universe?” Physical Review D 99 (2019): 043526. DOI. Open PDF.
• Hartle, J. B., and S. W. Hawking. “Wave Function of the Universe.” Physical Review D 28 (1983): 2960–2975. DOI.
• Hawking, S. W., and R. Penrose. “The Singularities of Gravitational Collapse and Cosmology.” Proceedings of the Royal Society A 314 (1970): 529–548. DOI.
• Singh, P. “Curvature Invariants, Geodesics, and the Strength of Singularities in Bianchi-I Loop Quantum Cosmology.” Physical Review D 85 (2012): 104011. DOI. Open PDF.