Alternatives and Tests of the Inflationary Paradigm

A nearly scale-invariant scalar spectrum does not uniquely identify the primordial background. Accelerated expansion, matter-like contraction, entropy-to-curvature conversion, thermal string fluctuations, and a modified initial quantum state can imitate part of the same two-point signal. A fair test must compare complete scenarios: their background smoothing, initial state, perturbative stability, transition to the hot Big Bang, scalar and tensor correlations, non-Gaussian shapes, and relics under the same late-time likelihood.

Required background. The Einstein Equation supplies Einstein Equations for expansion, contraction, bounce, and energy-condition tests. Classical Field Modes and Canonical Variables supplies Flat-Spacetime Quantum Field Theory for vacuum modes, stability, and Bogoliubov transformations.

Helpful background. Relativistic Matter and Compact Objects supplies matter-model and energy-condition language, while Relativistic Cosmology supplies the FLRW and primordial-spectrum conventions. Primordial Scalar and Tensor Spectra states the canonical inflationary benchmark being tested.

One scale-invariant spectrum, two backgrounds

An idealized constant-equation-of-state phase in flat Einstein gravity has

$$a(\eta)\propto\lvert\eta\rvert^q, \qquad q=\frac{2}{1+3w},$$

away from $w=-1/3$. If the canonical pump field is proportional to $a$, its mode equation is

$$v_k''+\left[k^2-\frac{q(q-1)}{\eta^2}\right]v_k=0.$$

The Bunch–Davies solution is a Hankel function of order

$$\nu=\left\lvert q-\frac12\right\rvert,$$

and at a fixed long-wavelength time

$$\mathcal P_{\mathcal R}(k)\propto k^{3-2\nu}.$$

Scale invariance, $\nu=3/2$, has two branches:

$$q=-1 \quad\hbox{or}\quad q=2.$$

The first is de Sitter-like expansion; the second is matter-like contraction. The equality is a useful counterexample to the claim that $n_s\simeq1$ alone proves inflation. The contracting branch relies on a growing mode and still has to pass through a bounce without changing its spectrum. Brandenberger and Peter derive this duality and the matter-bounce mode evolution in Brandenberger and Peter 2017, §§II.A–II.B, printed pp. 3–6, eqs. (1)–(22) (Open PDF).

For scalar curvature, $q=-1$ is understood as the quasi-de Sitter limit with a nonvanishing clock; exact de Sitter has $z=0$. Likewise, the matter branch requires a microphysical canonical oscillator with stable propagation and should not be identified with a literal pressureless hydrodynamic fluid whose sound speed vanishes.

The background also carries a sharp stability discriminator. Shear behaves as $\rho_\sigma\propto a^{-6}$, whereas a component with constant $w$ obeys $\rho_w\propto a^{-3(1+w)}$. During contraction,

$$\frac{\rho_\sigma}{\rho_w} \propto a^{3(w-1)}.$$

Matter contraction therefore amplifies relative shear as $a^{-3}$. An ekpyrotic phase with $w\gt 1$ suppresses it, but the simplest single-field adiabatic ekpyrotic curvature mode is strongly blue; producing the observed nearly scale-invariant curvature commonly requires entropy perturbations and a controlled conversion. Smoothing the background and generating the scalar spectrum are separate tasks.

The transition is part of the model

A nonsingular bounce requires $H$ to pass from negative to positive with $\dot H\gt 0$. In spatially flat Einstein gravity,

$$\dot H=-4\pi G(\rho+p),$$

so the bounce requires null-energy-condition violation. Positive spatial curvature or modified gravitational equations change this conclusion, but then their extra degrees of freedom and constraints belong to the declared model. A background solution alone is not enough: the quadratic action must remain free of ghost and gradient instabilities over the wavenumber range of interest, and perturbations must remain small enough for the expansion to be credible.

At $H=0$, the Hubble radius diverges and $k/(aH)$ is undefined. One cannot justify bounce matching by applying an inflationary “horizon crossing” rule at that instant. The regular perturbation variables, constraint equations, matching hypersurface, and any modified-gravity junction conditions must be specified and evolved through the transition. The distinction between a nonsingular background and a stable, predictive perturbation theory is surveyed in Brandenberger and Peter 2017, §III, printed pp. 11–14, and §V, printed pp. 18–22 (Open PDF).

A primary no-go argument makes the scope of that stability claim precise. In Kobayashi 2016, §II, pp. 2–3, eqs. (8)–(15), and §IV, p. 5 (Open PDF), a flat Horndeski background exists for all cosmic times, $a$ stays bounded away from zero, and scalar/tensor kinetic and spatial-gradient coefficients are smooth and strictly positive. With regular constraint elimination, positivity implies $\dot\xi\gt a\mathcal F_T$, where $\xi=a\mathcal G_T^2/\Theta$; $\mathcal G_T$ and $\mathcal F_T$ are the tensor kinetic and gradient coefficients, and $\Theta$ is the scalar constraint-mixing coefficient. If $\xi$ remains continuous and $\int a\mathcal F_Tdt$ diverges toward both time ends, $\xi$ must pass through zero, contradicting finite $\Theta$ and positive $a,\mathcal G_T$. Changing these hypotheses, including a vanishing quadratic coefficient or singular constraint reduction, requires a separate analysis. This argument does not exclude every positive-curvature, beyond-Horndeski, or finite-duration effective description.

For example, the explicit L4-Horndeski bounce of Ijjas and Steinhardt 2017, §§4–6, printed pp. 291–294, especially eqs. (29)–(33) (Open PDF) has finite limiting sound speeds and nonnegative quadratic coefficients at the zero crossing of its constraint coefficient $\gamma$. Both scalar and tensor quadratic coefficients vanish there. This crossing can occur before the bounce at $H=0$; the two events must not be identified. It therefore identifies a loophole boundary; it does not by itself establish weak coupling, nonlinear stability, or ultraviolet control at the crossing.

Ekpyrotic or cyclic constructions face an additional composition question: which entropy mode is converted, when, and with what nonlinearity? Repeated cycles can also reprocess an inherited spectrum, so a one-cycle calculation does not automatically define an indefinitely cyclic prediction. Matter bounces must control growing anisotropy before the new-physics bounce; ekpyrotic models trade that problem for conversion and bounce dynamics.

Inflationary extensions and initial states

Canonical single-field slow roll supplies a sharply normalized benchmark: adiabatic curvature is conserved on the attractor, leading non-Gaussianity is small, and $r=-8n_t$ at first slow-roll order. It does not constitute the whole inflationary paradigm. Multiple light fields can generate correlated isocurvature and super-Hubble curvature transfer; noncanonical dynamics can introduce a sound horizon, $r\simeq16\epsilon_Hc_s$, and enhanced equilateral or orthogonal bispectra. A null result for one of these signatures constrains a parameter regime, not all inflation.

Nor does a modified initial state require a different background. Write a normalized scalar mode as

$$v_k=\alpha_kv_k^{\rm BD}+\beta_kv_k^{{\rm BD}*}, \qquad \lvert\alpha_k\rvert^2-\lvert\beta_k\rvert^2=1.$$

For the mode convention used in this chapter, the late two-point spectrum is multiplied at leading order by

$$\mathcal P_{\mathcal R}(k) =\mathcal P_{\mathcal R}^{\rm BD}(k) \lvert\alpha_k-\beta_k\rvert^2.$$

An arbitrary choice is not admissible: ultraviolet adiabaticity, finite backreaction, the effective-theory cutoff, and a preparation mechanism must be checked. Initial-state modifications may correlate oscillatory power-spectrum features with folded non-Gaussianity; Baumann discusses the latter in Baumann 2012, §23.3.3, printed p. 97, eq. (340) (Open PDF). The trans-Planckian and initial-data limitations of inflation are reviewed in Ellis, Vennin, and Wands 2025/2026, §23.8, printed pp. 27–28 (Official PDF).

Emergent and string-gas scenarios use a different starting point. In a string-gas example, thermal fluctuations in a quasi-static high-temperature phase source metric perturbations before an exit to radiation domination. The proposal must supply stabilized extra dimensions, thermodynamic control, a nonsingular exit, and a regime in which the relevant fluctuations can be calculated. It should not be treated as merely another choice of inflaton potential. The mechanism and its characteristic scalar/tensor discussion are given in Brandenberger and Peter 2017, §II.E, printed pp. 8–10, and §IV, printed pp. 14–18 (Open PDF).

Results and limitations through 2026-08-19

At the 2026-08-19 evidence cutoff, broad primordial paradigms remain underdetermined by current two-point data. The evidence does not establish a universal verdict between all inflationary, bouncing, emergent, and initial-state formulations. Individual parameter regimes can be excluded or favored within specified likelihoods, while stability, matching, and ultraviolet completion remain unresolved for many alternatives. The competing positions differ not only in background history but also in what counts as controlled initial data and a valid transition to radiation.

Scenario or formulation	Evidence and supported mechanism	Scope, limitation, or missing test
Canonical single-field inflation	Analytic background and perturbation theory are controlled within a slow-roll effective field theory; it predicts adiabatic freeze-out and a leading scalar–tensor consistency relation.	The potential, initial conditions, ultraviolet completion, and reheating are not fixed by the paradigm; a broad family has no single value of $r$.
Multifield or noncanonical inflation	Covariant field-space and effective-action calculations support entropy transfer, sound-speed effects, and distinctive bispectrum templates in declared regimes.	Predictions depend on turns, masses, decay channels, sound-speed evolution, and priors; null isocurvature or non-Gaussianity results constrain only parts of the model class.
Matter bounce	Linear vacuum modes in matter contraction can produce a scale-invariant growing curvature mode, providing an analytic counterexample to uniqueness of $n_s\simeq1$.	Shear grows relative to matter; the bounce action, stability, nonlinearities, matching, and tensor/non-Gaussian amplitudes must all remain controlled.
Ekpyrotic or cyclic contraction	A very stiff contracting component can suppress relative shear; entropy mechanisms can generate and convert nearly scale-invariant fluctuations.	A single adiabatic field is blue, while conversion, non-Gaussianity, the bounce, and cycle-to-cycle processing are model dependent.
Emergent or string-gas scenario	Thermal string fluctuations offer a non-vacuum route to primordial spectra and potentially different tensor behavior.	Thermodynamic equilibrium, dimension stabilization, strong-coupling control, the exit, and a complete effective action are not generically established.
Initial-state modification	Bogoliubov modes give calculable features and folded correlations on an inflationary background.	Backreaction, ultraviolet admissibility, preparation, cutoff dependence, and look-elsewhere effects restrict arbitrary templates; no feature alone identifies its origin.

Two versioned null tests illustrate what the matrix can and cannot decide. The BK18 multicomponent likelihood combines BICEP/Keck with WMAP and Planck maps from 23 to 353 GHz and reports no primordial B-mode detection, with $r_{0.05}\lt0.036$ at 95% confidence for its standard nearly scale-invariant tensor template BICEP/Keck Collaboration 2021, likelihood analysis, Fig. 4, and conclusion (Open PDF). The official NASA LAMBDA BK18 products provide bandpowers, bandpasses, component spectra, a tabulated marginalized $r$ likelihood, a CosmoMC package, and chains; the one-dimensional $r$ curve is not the full multifrequency likelihood. This result constrains that tensor and foreground model, not every inflationary realization.

For a named alternative, Bodnia et al. 2024, §§2–3, printed pp. 3–10, §4, pp. 11–18, and §5, pp. 19–20 (Open PDF) use Planck/WMAP maps, simulations, artifact controls, and a neural-network search and find no statistically relevant low-variance circles or Hawking points. That null applies to the tested conformal-cyclic templates, masks, simulations, and pipeline; it does not disprove every cyclic or conformal-cyclic construction.

The inflationary extensions and current observational discriminants are reviewed in Ellis, Vennin, and Wands 2025/2026, §§23.7–23.9, printed pp. 22–29 (Official PDF). The bouncing mechanisms, signatures, and failure modes in the matrix use Brandenberger and Peter 2017, §§II–VI, printed pp. 3–22 (Open PDF). The latter is a 2017 synthesis, so its model examples are used for mechanisms and diagnostic axes, not as a standalone current observational verdict.

The interpretation could change after a robust primordial-tensor detection with a measured tilt, a distinctive non-Gaussian or isocurvature shape, a parity-violating primordial signal, or a demonstrated stability theorem or no-go result that closes a declared model class over its full parameter regime.

Comparing predictions with one likelihood

A fair comparison holds fixed the late-time data treatment. Each scenario must deliver, over the same scale range, the primordial quantities required by the transfer calculation:

$$\left\{ \mathcal P_{\mathcal R}(k), \mathcal P_t(k), \mathcal P_S(k), \mathcal C_{\mathcal RS}(k), B_{\mathcal R}(k_1,k_2,k_3),\ldots \right\}.$$

They must then be propagated with the same late-time cosmology, recombination and structure-transfer treatment, foreground and calibration model, data selection, likelihood, and prior policy. Matching only $n_s$ while changing the late-time nuisance model is not a test of the primordial scenario.

The most discriminating combinations are relational rather than single numbers: $r$ together with $n_t$; isocurvature amplitude together with its correlation; bispectrum amplitude together with squeezed, equilateral, orthogonal, or folded shape; tensor chirality or parity violation; features coherent across two- and three-point functions; and relic predictions such as primordial black holes or gravitational-wave backgrounds. A ghost or a gradient instability within the model’s claimed effective-theory domain can invalidate a prediction even outside the observed band. Its growth time and the ultraviolet cutoff must still be checked; an instability extrapolated above that cutoff, or ordinary gravitational growth of structure, is not by itself such a verdict.

Exercises

In the idealized constant-$w$ calculation, compare de Sitter-like expansion with matter-like contraction. Show why their scalar tilts can agree, then give three reasons the two formulations are not observationally or theoretically equivalent.

Solution

In the formal quasi-de Sitter limit, $w\to-1$ gives $q\to-1$ and $\nu=\lvert-1-1/2\rvert=3/2$. For matter contraction, $w=0$ gives $q=2$ and $\nu=\lvert2-1/2\rvert=3/2$. Hence both idealized branches yield

$$\mathcal P_{\mathcal R}\propto k^{3-2\nu}=k^0.$$

They differ first in background stability: during matter contraction, $\rho_\sigma/\rho_m\propto a^{-3}$ grows, whereas accelerated expansion dilutes curvature and ordinary shear. Second, the matter branch uses a growing curvature mode that must be transmitted through a specified bounce; the inflationary attractor uses a constant long-wavelength mode and reheats without crossing $H=0$. Third, their tensor spectra, tensor tilts, non-Gaussian shapes, and relic production need not agree. One must compare those complete outputs with the same late-time likelihood and also require a ghost-free, gradient-stable bounce. Equal $n_s$ is therefore necessary for neither formulation and sufficient for neither.

References

• Baumann, D., “TASI Lectures on Inflation,” arXiv:0907.5424v2 (2012), Open PDF.
• BICEP/Keck Collaboration, “Improved Constraints on Primordial Gravitational Waves Using Planck, WMAP, and BICEP/Keck Observations through the 2018 Observing Season,” Physical Review Letters 127, 151301 (2021), DOI, Open PDF.
• Bodnia, E., V. Isenbaev, K. Colburn, J. Swearngin, and D. Bouwmeester, “The Quest for CMB Signatures of Conformal Cyclic Cosmology,” Journal of Cosmology and Astroparticle Physics 2024(05), 009 (2024), DOI, Open PDF.
• Brandenberger, R., and P. Peter, “Bouncing Cosmologies: Progress and Problems,” Foundations of Physics 47, 797–850 (2017), DOI, Open PDF.
• Ellis, J., V. Vennin, and D. Wands, “Inflation,” in F. Takahashi et al. (Particle Data Group), Review of Particle Physics, International Journal of Modern Physics A 41, 2630011 (2026), revised August 2025, Official PDF.
• Ijjas, A., and P. J. Steinhardt, “Fully Stable Cosmological Solutions with a Non-singular Classical Bounce,” Physics Letters B 764, 289–294 (2017), DOI, Open PDF.
• Kobayashi, T., “Generic Instabilities of Nonsingular Cosmologies in Horndeski Theory: A No-Go Theorem,” Physical Review D 94, 043511 (2016), DOI, Open PDF.
• NASA Legacy Archive for Microwave Background Data Analysis. n.d. “BICEP2/Keck Array Data Products.” Official BK18 products.