betaFeedback

WorksPhysics and Astronomy

Genesis in Minimally Modified Gravity

Abstract

We construct a Genesis scenario within the framework of a type-II minimally modified gravity. The background solution is past geodesically complete and approaches Minkowski spacetime in the asymptotic past. While the Hubble expansion rate grows, the matter sector satisfies the null-energy condition. Scalar perturbations are ghost-free and have no ultraviolet gradient instability, although the squared sound speed cR2c_\mathcal{R}^2 becomes negative within a finite infrared region in the Fourier space for a finite time interval. Numerical evolution, including superhorizon growth, yields a nearly scale-invariant, slightly red-tilted scalar spectrum over a broad region of parameter space. The tensor spectrum is blue, with a strongly suppressed tensor-to-scalar ratio. Finally, using the matter fluctuation δχδχ, whose asymptotic normalization is regular, we find that the background-dependent cubic couplings vanish in the far past at fixed nonzero momenta. The surviving matter–graviton and graviton self-interactions are Planck suppressed.

Prepared for submission to JHEP

Introduction

Genesis scenarios describe a universe that approaches Minkowski spacetime in the asymptotic past before entering an expanding phase with an increasing Hubble expansion rate. The scale factor tends to a nonzero constant while the Hubble expansion rate tends to zero in the infinite past, avoiding an initial curvature singularity. The original Galilean Genesis construction demonstrated that such an evolution can be supported by a scalar field with higher-derivative interactions [1, 2]. During Genesis, ϵ=−H˙/H2<0\epsilon= -\dot{H}/H^2 < 0. In a spatially flat universe governed by general relativity (GR), this requires violating the null-energy condition (NEC) [3, 4]. Realizing this evolution without pathological perturbations is therefore a central challenge.

A suitable background alone does not guarantee an acceptable primordial power spectrum. In the original Genesis scenario, a nearly scale-invariant spectrum can be generated by an additional spectator field [1, 2], while generalized Genesis models also allow nearly scale-invariant curvature perturbations to be generated directly [5]. The relevant observational requirements concern the scalar amplitude, spectral index and running, as well as the tensor spectrum and primordial non-Gaussianities. We use the results of Planck 2018 [6, 7] as an observational reference. Since the curvature perturbations need not be conserved outside the horizon during Genesis, its subsequent evolution must be included when determining the spectrum at the end of this phase.

The stability of nonsingular cosmologies has been studied extensively in scalar–tensor theories. No-go results in generalized Galileon and Horndeski theories show that, under their stated assumptions, a nonsingular background cannot remain free of instabilities throughout its evolution [8, 9]. These conclusions depend on the asymptotic behavior of the perturbation coefficients and on the theory under consideration. Stable Genesis solutions have, for example, been constructed in beyond-Horndeski and DHOST theories [10, 11, 12, 13]. Linear stability is also distinct from perturbative control: a time-dependent strong-coupling scale must be compared with the characteristic energy scale of the background, rather than judged from its asymptotic value alone. Analyses at cubic and higher orders have identified Horndeski Genesis regimes in which this hierarchy can be maintained [14, 15, 16].

In this work, we consider VCDM, a theory of type-II minimally modified gravity [17]. (For the relation between VCDM and cuscuton [18], see [19].) Its constraint structure leaves only two propagating tensor degrees of freedom in the gravitational sector, at the cost of breaking time-diffeomorphism invariance. The auxiliary scalar introduces no additional propagating degree of freedom, while its potential allows the background expansion to be reconstructed [17, 20]. This freedom has already been used to construct bouncing cosmologies with a canonical matter scalar [21]. Here we investigate whether VCDM can support an asymptotically Minkowski Genesis phase with canonical NEC-respecting matter, a nearly scale-invariant curvature spectrum, and cubic interactions that remain controlled in the asymptotic past.

We construct the background using a power-law evolution of the conformal Hubble expansion rate and a logarithmic evolution of the matter scalar. The matter potential is obtained in terms of exponential integral functions, while the monotonic evolution of the auxiliary scalar determines the VCDM potential parametrically. The matter sector satisfies the null energy condition (NEC)

ρχ+pχ=χ˙2>0,(1)\rho_{\chi} + p_{\chi} = \dot{\chi}^{2} > 0, \tag*{(1)}

while the modified gravitational equations allow H˙/H2>0\dot{H}/H^{2} > 0. Thus, the effective NEC-violating expansion is generated by the modified gravitational dynamics, rather than by NEC violation in the matter sector.

We study the scalar and tensor perturbations using analytical approximations and numerical evolution. The curvature perturbation has a positive kinetic coefficient throughout Genesis, and its short-wavelength sound speed approaches unity. At finite momentum, cR2c_{\mathcal{R}}^{2} can become negative within a bounded infrared region. However, since cR2→1c_{\mathcal{R}}^{2} \to1 in the ultraviolet, this does not lead to the arbitrarily fast short-wavelength growth characteristic of an ultraviolet gradient instability. After accounting for the superhorizon evolution, the numerical scalar spectrum exhibits a nearly scale-invariant, slightly red-tilted regime over a broad range of model parameters. The tensor modes have the GR quadratic action, and their spectrum is blue, with a strongly suppressed tensor-to-scalar ratio for the parameters considered.

The cubic analysis requires particular care in the choice of perturbation variable. Nonlinear definitions of the curvature perturbation and their application to VCDM have been discussed in [22]. On the present background, the nonlinear transformation to the curvature perturbation can become singular in the deep Genesis limit. We therefore analyze the asymptotic interactions directly in terms of the matter fluctuation δχ\delta\chi, whose normalization remains regular in the fixed-momentum limit. Considering the pure scalar, mixed scalar–tensor and pure tensor cubic sectors, we find that the background-dependent cubic couplings vanish in the asymptotic past, while the surviving matter–graviton and graviton self-interactions remain Planck suppressed. This result holds at fixed nonzero momenta and away from the singular surface of the δχ\delta\chi velocity reduction. Thus, the cubic power counting reveals no strong-coupling scale that collapses to zero in this limit. However, it does not determine the full nonlinear effective field theory cutoff.

The paper is organized as follows. § 2 introduces the model and reconstructs the matter scalar potential and the VCDM potential from the Genesis background. § 3 analyzes linear perturbations, their stability, the scalar and tensor power spectra, and past geodesic completeness. § 4 presents the cubic analysis. § 5 summarizes the results. Appendix § 6 gives the phenomenological fit to the scalar spectrum. Appendix § 7 provides the reduced perturbation actions and asymptotic cubic power counting.

The VCDM model

General description

The VCDM construction is described in detail in [17]. Here, we briefly review the definitions required for a general understanding. The VCDM action reads

SVCDM=MPl22∫d4xNγ[(3)R+KijKij−K2−2V(φ)−2λiN∂iφ−32λ02−2λ0(K+φ)](2)S_{\mathrm{VCDM}} = \frac{M_{\mathrm{Pl}}^2}{2} \int\mathrm{d}^4x N\sqrt{\gamma}\left[{}^{(3)}R + K_{ij}K^{ij} - K^2 - 2V(\varphi) - 2\frac{\lambda^i}{N}\partial_i\varphi- \frac{3}{2}\lambda_0^2 - 2\lambda_0(K+\varphi)\right] \tag*{(2)}

where KijK_{ij} is the extrinsic curvature, (3)R{}^{(3)}R is the Ricci scalar associated with the metric γij\gamma_{ij} on the hypersurfaces of constant time, NN is the lapse function, the quantities λi\lambda^i and λ0\lambda_0 are Lagrange multipliers, and φ\varphi is an auxiliary scalar field. The nontrivial constraint structure leaves only two propagating tensor degrees of freedom, at the cost of breaking time-diffeomorphism invariance. In addition, we consider a canonical scalar field for the matter sector

Smatter=∫d4xNγ(−12gμν∂μχ∂νχ−U(χ)).(3)S_{\mathrm{matter}} = \int\mathrm{d}^4x N\sqrt{\gamma}\left(-\frac{1}{2}g^{\mu\nu}\partial_\mu\chi\partial_\nu\chi- U(\chi)\right). \tag*{(3)}

The total action thus reads

S=SVCDM+Smatter.(4)S = S_{\mathrm{VCDM}} + S_{\mathrm{matter}} . \tag*{(4)}

Genesis background

For a homogeneous and isotropic background, we take the following background quantities

N=Nˉ(t),Ni=0,γij=a2(t)δij,φ=ϕ(t),N = \bar{N}(t), \qquad N^i = 0, \qquad\gamma_{ij} = a^2(t)\delta_{ij}, \qquad\varphi= \phi(t),
λi=0,λ0=λˉ(t),χ=χˉ(t),(5)\lambda^i = 0, \qquad\lambda_0 = \bar{\lambda}(t), \qquad\chi= \bar{\chi}(t), \tag*{(5)}

where NiN^i is the shift vector. Varying the action (4) with respect to Nˉ\bar{N}, λˉ\bar{\lambda}, ϕ\phi, aa, χˉ\bar{\chi} leads, respectively, to

3H2=V+ϕλˉ+34λˉ2+χˉ˙22MPl2+UMPl2,(6)3H^2 = V + \phi\bar{\lambda} + \frac{3}{4}\bar{\lambda}^2 + \frac{\dot{\bar{\chi}}^2}{2M_{\mathrm{Pl}}^2} + \frac{U}{M_{\mathrm{Pl}}^2}, \tag*{(6)}
0=3H+ϕ˙+32λˉ,(7)0 = 3H + \dot{\phi} + \frac{3}{2}\bar{\lambda}, \tag*{(7)}
0=λˉ+Vϕ,(8)0 = \bar{\lambda} + V_{\phi}, \tag*{(8)}
2∂tHNˉ+3H2=V+ϕλˉ+34λˉ2−∂tλˉNˉ−χˉ˙22MPl2+UMPl2,(9)2\frac{\partial_t H}{\bar{N}} + 3H^2 = V + \phi\bar{\lambda} + \frac{3}{4}\bar{\lambda}^2 - \frac{\partial_t\bar{\lambda}}{\bar{N}} - \frac{\dot{\bar{\chi}}^2}{2M_{\mathrm{Pl}}^2} + \frac{U}{M_{\mathrm{Pl}}^2}, \tag*{(9)}
0=χˉ¨+3Hχˉ˙+U,χ,(10)0 = \ddot{\bar{\chi}} + 3H\dot{\bar{\chi}} + U_{,\chi}, \tag*{(10)}

where H≡∂ta⋅(aN)−1H \equiv\partial_t a \cdot(aN)^{-1} is the Hubble expansion rate, and VV and UU (and their derivatives) are all evaluated at the background values. Dots denote proper-time derivatives, that is f˙≡Nˉ−1∂tf\dot{f} \equiv\bar{N}^{-1}\partial_t f, and primes denote conformal-time derivatives d/dτ≡aNˉ−1∂t\mathrm{d}/\mathrm{d}\tau\equiv a\bar{N}^{-1}\partial_t.

We start from an ansatz for the conformal Hubble expansion rate and subsequently reconstruct the corresponding VCDM potential VV. Working in conformal time τ≡∫tdt~ Nˉ(t~)/a(t~)\tau\equiv\int^t \mathrm{d}\tilde{t}\,\bar{N}(\tilde{t})/a(\tilde{t}), we take

H=h0(−τ−τB)d+1,a=eh0d(−τ−τB)−d,(11)\mathcal{H} = \frac{h_0}{(-\tau-\tau_B)^{d+1}}, \qquad a = e^{\frac{h_0}{d}(-\tau-\tau_B)^{-d}}, \tag*{(11)}

where H≡a′/a\mathcal{H} \equiv a'/a is the conformal Hubble expansion rate, τB<0\tau_B < 0 and d>0d > 0 are constant. The “slow-roll” parameter ϵ\epsilon is given by

ϵ=−H′H2=1−(1+d)(−τB)dh0(1+x)d,(12)\epsilon= -\frac{\mathcal{H}'}{\mathcal{H}^2} = 1 - \frac{(1+d)(-\tau_B)^d}{h_0}(1+x)^d, \tag*{(12)}

where we have introduced the dimensionless time x=τ/τBx = \tau/\tau_B. For later convenience, we define T≡1+xT \equiv1+x. We choose h0h_0 such that τ=x=0\tau= x = 0 marks the end of the Genesis and that τ<0\tau< 0 during the Genesis epoch, i.e. ϵ(0)=0\epsilon(0) = 0, which gives

h0(−τB)d=1+d.(13)\frac{h_0}{(-\tau_B)^d} = 1+d . \tag*{(13)}

The approximation a≃1a \simeq1 is valid only when (1+d)T−d/d≪1(1+d)T^{-d}/d \ll1 and therefore breaks down near the endpoint, where

a(0)=e1+dd.(14)a(0) = e^{\frac{1+d}{d}} . \tag*{(14)}

The endpoint Hubble scale is

H(0)=(1+d)e−1+dd(−τB)−1,(15)H(0)=(1+d)e^{-\frac{1+d}{d}}(-\tau_B)^{-1}, \tag*{(15)}

so at fixed dd, it is (−τB)−1(-\tau_B)^{-1} that sets the characteristic energy scale. We restrict our analysis to the Genesis phase. Its transition to a subsequent inflationary or reheating stage is left for a future work.

Evolution of the matter sector

To keep the analysis tractable, we assume a logarithmic evolution for the background matter field:

χˉ(τ)=χ1log⁡[M⋆(−τ−τB)]=χ1[log⁡(1+x)+log⁡(−M⋆τB)],(16)\bar{\chi}(\tau)=\chi_1\log\left[M_\star(-\tau-\tau_B)\right]=\chi_1\left[\log(1+x)+\log(-M_\star\tau_B)\right], \tag*{(16)}

where M⋆M_\star is an auxiliary mass scale that makes the logarithm dimensionless. Here we consider M⋆>0M_\star>0 and χ1>0\chi_1>0. Note that deep in the Genesis epoch the scalar field grows without bound, lim⁡x→∞χˉ=∞\lim_{x\to\infty}\bar{\chi}=\infty. For later purposes, it is useful to introduce

α=ρχ+pχMPl2H2=χˉ˙2MPl2H2=χ12MPl2(1+d)2(1+x)2d≡α0(1+x)2d>0.(17)\alpha=\frac{\rho_\chi+p_\chi}{M_{\mathrm{Pl}}^2H^2}=\frac{\dot{\bar{\chi}}^2}{M_{\mathrm{Pl}}^2H^2}=\frac{\chi_1^2}{M_{\mathrm{Pl}}^2(1+d)^2}(1+x)^{2d}\equiv\alpha_0(1+x)^{2d}>0. \tag*{(17)}

Unlike in GR, α\alpha is not directly tied to ϵ\epsilon. In the far past, α→∞\alpha\to\infty, while αH2=a2(ρχ+pχ)MPl−2→0\alpha H^2=a^2(\rho_\chi+p_\chi)M_{\mathrm{Pl}}^{-2}\to0. At the endpoint α(0)=α0\alpha(0)=\alpha_0, and below we focus on α0≪1\alpha_0\ll1.

Substituting the ansatz (16) into the field equation (10), we get an equation for the potential UU that reads

−χ1(−τ−τB)−2−2h0χ1(−τ−τB)−d−2+a(τ)2ddχU=0.(18)-\chi_1(-\tau-\tau_B)^{-2}-2h_0\chi_1(-\tau-\tau_B)^{-d-2}+a(\tau)^2\frac{\mathrm{d}}{\mathrm{d}\chi}U=0. \tag*{(18)}

This equation can be solved exactly using dχ=−χ1(−τ−τB)−1dτ\mathrm{d}\chi=-\chi_1(-\tau-\tau_B)^{-1}\mathrm{d}\tau as

U(χ)=χ12τB2(dy2(1+d))2/d(1dE1−2d(y)+yE−2d(y)),(19)U(\chi)=\frac{\chi_1^2}{\tau_B^2}\left(\frac{d y}{2(1+d)}\right)^{2/d}\left(\frac{1}{d}E_{1-\frac{2}{d}}(y)+yE_{-\frac{2}{d}}(y)\right), \tag*{(19)}

where Eν(y)=∫1∞e−ytt−νdtE_\nu(y)=\int_1^\infty e^{-yt}t^{-\nu}\mathrm{d}t. The additive constant is chosen so that U(y→∞)=0U(y\to\infty)=0, and

y=2(1+d)de−dxχ1(−τBM⋆)d.(20)y=\frac{2(1+d)}{d}e^{-\frac{dx}{\chi_1}}(-\tau_BM_\star)^d. \tag*{(20)}

During the Genesis phase, yy is restricted to 0<y≤2(1+d)/d0<y\le2(1+d)/d. In general, the potential has essentially the form of a step function. Deep in the Genesis phase, x→∞x\to\infty, the scalar field lies near the top of the step-like potential and rolls downhill with χˉ˙→0\dot{\bar{\chi}}\to0. The parameter dd controls the steepness of the potential, as shown in Figure 1.

Figure 1. Reconstructed matter potential U(y)U(y) for d=1/3d = 1/3 (green), 1/21/2 (orange), and 11 (blue). Solid curves correspond to the Genesis trajectory, 0<y≤2(1+d)/d0 < y \le2(1+d)/d, and the markers indicate the end of Genesis at y=2(1+d)/dy = 2(1+d)/d. Dashed curves show the mathematical continuation of U(y)U(y) beyond the Genesis domain.

At the end of the Genesis phase, the energy density is given by

ρχ(0)=12χˉ˙2+U(χ)∣τ=0(21)\rho_\chi(0)=\left.\frac{1}{2}\dot{\bar{\chi}}^2+U(\chi)\right|_{\tau=0} \tag*{(21)}
=χ12τB2[12e−2+2dd+(2+2d)E−2d(2+2dd)+1dE1−2d(2+2dd)].(22)=\frac{\chi_1^2}{\tau_B^2}\left[\frac{1}{2}e^{-\frac{2+2d}{d}}+\left(2+\frac{2}{d}\right)E_{-\frac{2}{d}}\left(\frac{2+2d}{d}\right)+\frac{1}{d}E_{1-\frac{2}{d}}\left(\frac{2+2d}{d}\right)\right]. \tag*{(22)}

The kinetic and potential contributions are of the same order. For the parameters used below, Ωχ(0)≡ρχ(0)/(3MPl2H(0)2)≪1\Omega_\chi(0)\equiv\rho_\chi(0)/(3M_{\mathrm{Pl}}^2H(0)^2)\ll1.

Reconstruction of the VCDM potential

To reproduce the chosen Hubble evolution, we reconstruct the VCDM potential following [20, 21]. The background equations determine the auxiliary field:

ϕ=32MPl2∫dτ χˉ˙2=32MPl2∫dτ χˉ′2a=−3χ122(−τB)MPl2∫dx e−1+ddx(1+x)−d(1+x)2=−3χ122MPl2d(−τ−τB)Ed−1d(y2)+ϕ0,(23)\begin{aligned} \phi&= \frac{3}{2M_{\mathrm{Pl}}^2}\int\mathrm{d}\tau\,\dot{\bar{\chi}}^2 = \frac{3}{2M_{\mathrm{Pl}}^2}\int\mathrm{d}\tau\,\frac{\bar{\chi}^{\prime2}}{a} \\ &= -\frac{3\chi_1^2}{2(-\tau_B)M_{\mathrm{Pl}}^2}\int\mathrm{d}x\,\frac{e^{-\frac{1+d}{d}x}(1+x)^{-d}}{(1+x)^2} \\ &= -\frac{3\chi_1^2}{2M_{\mathrm{Pl}}^2d(-\tau-\tau_B)}E_{\frac{d-1}{d}}\left(\frac{y}{2}\right)+\phi_0 , \tag*{(23)} \end{aligned}

where we used χˉ˙=χˉ′/a\dot{\bar{\chi}}=\bar{\chi}'/a and where ϕ0\phi_0 is an integration constant, which we will set to ϕ0=0\phi_0=0. As ϕ\phi is a strictly monotonic function, one can invert the relation to obtain x=x(ϕ)x=x(\phi). For generic dd the inversion is numerical only, while special cases such as d=1d=1 admit a closed form. Substituting ϕ(x)\phi(x) into the background equations gives

V=ϕ23−ρχMPl2=3χ144MPl4d2(−τ−τB)2E−1+dd(y2)2−χ12e−2(1+d)dx(1+x)−d2MPl2(−τ−τB)2−χ12MPl2τB2(dy2(1+d))2/d(1dE1−2/d(y)+yE−2/d(y)),(24)\begin{aligned} V &= \frac{\phi^2}{3}-\frac{\rho_\chi}{M_{\mathrm{Pl}}^2} \\ &= \frac{3\chi_1^4}{4M_{\mathrm{Pl}}^4d^2(-\tau-\tau_B)^2}E_{\frac{-1+d}{d}}\left(\frac{y}{2}\right)^2-\frac{\chi_1^2e^{-\frac{2(1+d)}{d}x}(1+x)^{-d}}{2M_{\mathrm{Pl}}^2(-\tau-\tau_B)^2} \\ &\quad-\frac{\chi_1^2}{M_{\mathrm{Pl}}^2\tau_B^2}\left(\frac{dy}{2(1+d)}\right)^{2/d}\left(\frac{1}{d}E_{1-2/d}(y)+yE_{-2/d}(y)\right), \tag*{(24)} \end{aligned}

where y=y(x)y=y(x) should be understood as a function of xx. For generic dd, V(ϕ)V(\phi) is reconstructed numerically. For 0<d<10<d<1, the potential and its first two derivatives remain regular in the asymptotic past, while the third derivative diverges. This does not affect nonzero-momentum perturbations, since the λi\lambda^i constraint implies δϕk=0\delta\phi_k=0 for k≠0k\ne0.

Linear perturbations

The VCDM action is invariant under foliation-preserving transformations t→t~(t)t \to\tilde{t}(t) and xi→x~i(t,x)x^i \to\tilde{x}^i(t,\mathbf{x}), but not under spatially dependent changes of time slicing. We use the spatial gauge freedom to remove the scalar off-diagonal metric perturbation. We thus perturb the lapse, shift, and spatial metric as

N=Nˉ(t)(1+δN),Ni=Nˉ(t)a(t)δij∂jβ,γij=a2(t)e2ζ(eh)ij,(25)N = \bar{N}(t)(1+\delta N), \qquad N^i = \frac{\bar{N}(t)}{a(t)}\delta^{ij}\partial_j\beta, \qquad\gamma_{ij} = a^2(t)e^{2\zeta}(e^h)_{ij}, \tag*{(25)}

where hijh_{ij} is transverse traceless: δij∂ihjk=0\delta^{ij}\partial_i h_{jk}=0, δijhij=0\delta^{ij}h_{ij}=0. We also expand the auxiliary fields {φ,λ0,λi}\{\varphi,\lambda_0,\lambda^i\} and the matter field χ\chi as

φ=φˉ(t)+δφ(t,x),λ0=λˉ(t)+δλ0(t,x),\varphi= \bar{\varphi}(t)+\delta\varphi(t,\mathbf{x}), \qquad\lambda_0 = \bar{\lambda}(t)+\delta\lambda_0(t,\mathbf{x}),
χ=χˉ(t)+δχ(t,x),λi=1a2δij∂jδλs(t,x).(26)\chi= \bar{\chi}(t)+\delta\chi(t,\mathbf{x}), \qquad\lambda^i = \frac{1}{a^2}\delta^{ij}\partial_j\delta\lambda_s(t,\mathbf{x}). \tag*{(26)}

The quadratic action for the tensor perturbation remains the same as in standard GR,

ST(2)=MPl28∫dτ d3k a2(∣hij′∣2−k2∣hij∣2).(27)S_T^{(2)} = \frac{M_{\mathrm{Pl}}^2}{8}\int d\tau\,d^3k\,a^2\left(\left|h'_{ij}\right|^2-k^2\left|h_{ij}\right|^2\right). \tag*{(27)}

The scalar quadratic action differs from that of a canonical matter field because of the nontrivial constraint structure. It is convenient to introduce the comoving curvature perturbation,

R≡ζ−Hχˉ′δχ.(28)\mathcal{R}\equiv\zeta-\frac{\mathcal{H}}{\bar{\chi}'}\delta\chi. \tag*{(28)}

At linear order, this combination is invariant under xμ→xμ+ξμx^\mu\to x^\mu+\xi^\mu. This kinematic invariance does not, however, restore the broken time-slicing symmetry of the action. We define the background logarithmic derivatives

η≡ϵ˙ϵH,βα≡α˙αH.(29)\eta\equiv\frac{\dot{\epsilon}}{\epsilon H}, \qquad\beta_\alpha\equiv\frac{\dot{\alpha}}{\alpha H}. \tag*{(29)}

For k≠0k\ne0, solving the linear constraint equations gives

δλ0=δλs=δφ=0,(30)\delta\lambda_0=\delta\lambda_s=\delta\varphi=0, \tag*{(30)}
β=3αH22k2χˉ′δχ,(31)\beta=\frac{3\alpha\mathcal{H}^2}{2k^2\bar{\chi}'}\delta\chi, \tag*{(31)}
δN=−2k2αH2ζ−H3α(6+βα−2ϵ)+a2χˉ′MPl−2U,χH2αχˉ′δχ+δχ′χˉ′,(32)\delta N=-\frac{2k^2}{\alpha\mathcal{H}^2}\zeta-\frac{\mathcal{H}^3\alpha(6+\beta_\alpha-2\epsilon)+a^2\bar{\chi}'M_{\mathrm{Pl}}^{-2}U_{,\chi}}{\mathcal{H}^2\alpha\bar{\chi}'}\delta\chi+\frac{\delta\chi'}{\bar{\chi}'}, \tag*{(32)}
δχ=−χˉ′H4k2R+2αHR′4k2+H2α(6+α−2ϵ).(33)\delta\chi=-\frac{\bar{\chi}'}{\mathcal{H}}\frac{4k^2\mathcal{R}+2\alpha\mathcal{H}\mathcal{R}'}{4k^2+\mathcal{H}^2\alpha(6+\alpha-2\epsilon)}. \tag*{(33)}

Substituting these solutions back into the action yields the quadratic action for R\mathcal{R},

SR(2)=MPl22∫dτ d3k z2(∣Rk′∣2−cR2k2∣Rk∣2),(34)S_{\mathcal{R}}^{(2)}=\frac{M_{\mathrm{Pl}}^2}{2}\int d\tau\,d^3k\,z^2\left(\left|\mathcal{R}'_k\right|^2-c_{\mathcal{R}}^2k^2\left|\mathcal{R}_k\right|^2\right), \tag*{(34)}

where

z2=2a2α2k2+3αH24k2+6αH2+αH2(α−2ϵ)(35)z^2 = 2a^2\alpha\frac{2k^2 + 3\alpha\mathcal{H}^2}{4k^2 + 6\alpha\mathcal{H}^2 + \alpha\mathcal{H}^2(\alpha- 2\epsilon)} \tag*{(35)}
cR2=8k4+k2H2B1+H4B28k4+k2H2A1+H4A2(36)c_{\mathcal{R}}^2 = \frac{8k^4 + k^2\mathcal{H}^2B_1 + \mathcal{H}^4B_2}{8k^4 + k^2\mathcal{H}^2A_1 + \mathcal{H}^4A_2} \tag*{(36)}

with

B1=2[α(−4βα+2ϵ+α)+4ϵ(6+βα−2ϵ+η)](37)B_1 = 2\left[\alpha(-4\beta_\alpha+ 2\epsilon+ \alpha) + 4\epsilon(6 + \beta_\alpha- 2\epsilon+ \eta)\right] \tag*{(37)}
B2=α[−(α−2ϵ)2+36ϵ+6(α(−1−βα+ϵ)+2ϵ(1−ϵ+η))](38)B_2 = \alpha\left[-(\alpha- 2\epsilon)^2 + 36\epsilon+ 6\left(\alpha(-1 - \beta_\alpha+ \epsilon) + 2\epsilon(1 - \epsilon+ \eta)\right)\right] \tag*{(38)}
A1=2α(12+α−2ϵ),(39)A_1 = 2\alpha(12 + \alpha- 2\epsilon), \tag*{(39)}
A2=3α2(6+α−2ϵ).(40)A_2 = 3\alpha^2(6 + \alpha- 2\epsilon). \tag*{(40)}

For later purposes, it is convenient to introduce the dimensionless parameters H~=−τBH\widetilde{\mathcal{H}} = -\tau_B\mathcal{H} and κ=−kτB\kappa= -k\tau_B. Then the linear mode equation depends only on κ\kappa, α0\alpha_0, and dd.

Stability

Evolution of zz

For ϵ≤0\epsilon\le0 and α>0\alpha> 0, both the numerator and denominator of z2z^2 are positive. Hence z2>0z^2 > 0 throughout Genesis, and the scalar sector is ghost-free. At fixed nonzero κ\kappa and 0<d<10 < d < 1, one has α→∞\alpha\to\infty but αH~→0\alpha\widetilde{\mathcal{H}} \to0 in the far past. Therefore,

z2(x→∞)≃a2α,(41)z^2(x \to\infty) \simeq a^2\alpha, \tag*{(41)}

and z2z^2 diverges as x→∞x \to\infty. This approximation also holds throughout Genesis for modes satisfying

κ≫max⁡(α,α,−αϵ)H~.(42)\kappa\gg\max(\alpha, \sqrt{\alpha}, \sqrt{-\alpha\epsilon})\widetilde{\mathcal{H}}. \tag*{(42)}

The strong-coupling issue and its absence are discussed in section 4.

During the early Genesis phase, z2z^2 decreases, leading to the later superhorizon growth of R\mathcal{R}. Near the end of Genesis, zz increases with time, so R\mathcal{R} is suppressed, but it does not necessarily become constant. Figure 2 exhibits this behavior.

Figure 2. Evolution of z(x)z(x) with respect to xx. Upper left: varying α0\alpha_0 at κ=10−4\kappa=10^{-4} and d=1/2d=1/2; upper right: varying κ\kappa at d=1/2d=1/2 and α0=10−5\alpha_0=10^{-5}; bottom: varying dd at α0=10−5\alpha_0=10^{-5} and κ=10−4\kappa=10^{-4}.

Infrared region with cR2<0c_{\mathcal{R}}^2 < 0

The squared sound speed depends on momentum and can become negative over a finite range of κ\kappa. Since the denominator of equation (36) is positive, the boundaries of this region are determined by the zeros of the numerator,

κ±2=H~216[−B1±B12−32B2],max⁡(0,κ−2)<κ2<κ+2.(43)\kappa_{\pm}^2 = \frac{\widetilde{\mathcal{H}}^2}{16}\left[-B_1 \pm\sqrt{B_1^2 - 32B_2}\right], \qquad\max(0,\kappa_-^2) < \kappa^2 < \kappa_+^2. \tag*{(43)}

This region exists when B12−32B2>0B_1^2 - 32B_2 > 0 and κ+2>0\kappa_+^2 > 0.

A negative cR2c_{\mathcal{R}}^2 does not by itself imply that the canonical mode is non-oscillatory, since its evolution is governed by

ω2(κ,x)=cR2κ2−z,xxz.(44)\omega^2(\kappa,x) = c_{\mathcal{R}}^2\kappa^2 - \frac{z_{,xx}}{z}. \tag*{(44)}

Therefore, the regions with cR2<0c_{\mathcal{R}}^2 < 0 and ω2<0\omega^2 < 0 do not generally coincide.

In the far past, for α≫max⁡(1,∣ϵ∣,∣βα∣)\alpha\gg\max(1,|\epsilon|,|\beta\alpha|), the upper boundary behaves as

κ+2≃α0(1+d)22(1+x)−2.(45)\kappa_+^2 \simeq\frac{\alpha_0(1+d)^2}{2}(1+x)^{-2}. \tag*{(45)}

The expansion requires α0(1+x)d≫1\alpha_0(1+x)^d \gg1, equivalently α≫∣ϵ∣\alpha\gg|\epsilon|, which is satisfied in the stated regime. Hence the negative-cR2c_{\mathcal{R}}^2 region shrinks toward zero momentum. Thus, every fixed nonzero mode eventually lies outside this region in the deep Genesis phase. For κ2≫α(6+α−2ϵ)H~2\kappa^2 \gg\alpha(6+\alpha-2\epsilon)\widetilde{H}^2, the large-momentum expansion gives

cR2κ2=κ2+(B1−A1)H~28+O(κ−2),(46)c_{\mathcal{R}}^2\kappa^2=\kappa^2+\frac{(B_1-A_1)\widetilde{H}^2}{8}+\mathcal{O}(\kappa^{-2}), \tag*{(46)}

and hence cR2→1c_{\mathcal{R}}^2 \to1 in the ultraviolet. The negative region is therefore an infrared effect rather than an ultraviolet gradient instability. Figure 3 shows the evolution of cR2c_{\mathcal{R}}^2. Although cR2c_{\mathcal{R}}^2 can exceed unity at finite momentum, this can arise from a mass-like correction and does not by itself imply superluminal propagation. Furthermore, even superluminal propagation would not violate causality in the sense of past and future defined by the VCDM preferred frame.

Figure 3. Evolution of cR2(x)c_{\mathcal{R}}^2(x). Upper left: varying α0\alpha_0 at d=1d = 1 and κ=10−4\kappa= 10^{-4}; upper right: varying κ\kappa at d=1d = 1 and α0=10−5\alpha_0 = 10^{-5}; bottom: varying dd at κ=10−5\kappa= 10^{-5} and α0=10−3\alpha_0 = 10^{-3}. At finite momentum, cR2c_{\mathcal{R}}^2 may be negative or exceed unity.

Evolution of the curvature perturbation and power spectrum

For the canonically normalized variable u=MPlzRu = M_{\mathrm{Pl}}z\mathcal{R}, the mode equation is

d2udx2+(cR2κ2−z,xxz)u=0.(47)\frac{\mathrm{d}^2u}{\mathrm{d}x^2} + \left(c_{\mathcal{R}}^2\kappa^2 - \frac{z_{,xx}}{z}\right)u = 0. \tag*{(47)}

For the numerical spectra, we define u^k=uk/−τB\hat{u}_k = u_k/\sqrt{-\tau_B}. No closed-form solution is available, so we use analytic approximations to set initial conditions and then solve the full equation numerically.

We impose initial conditions sufficiently far in the past so that ∣cR2−1∣≪1\lvert c_{\mathcal{R}}^2 - 1\rvert\ll1, ∣z,xx/z∣≪κ2\lvert z_{,xx}/z\rvert\ll\kappa^2, and so that the evolution is adiabatic. In this limit, we have that z,xx⋅z−1≃0z_{,xx}\cdot z^{-1} \simeq0 and cR2≃1c_{\mathcal{R}}^2 \simeq1. Therefore, the canonical mode approaches a Minkowski plane wave, that is

u∣x→∞≃−τBeiκx2κ,dudx∣x→∞≃i−τBκ2eiκx.(48)\left.u\right|_{x\to\infty} \simeq\sqrt{-\tau_B}\frac{e^{i\kappa x}}{\sqrt{2\kappa}}, \qquad\left.\frac{\mathrm{d}u}{\mathrm{d}x}\right|_{x\to\infty} \simeq i\sqrt{-\tau_B}\sqrt{\frac{\kappa}{2}}e^{i\kappa x}. \tag*{(48)}

Equation (48) specifies the asymptotic Minkowski vacuum. In the numerical calculation, the evolution begins at a finite time xix_i inside the adiabatic oscillatory regime. We therefore use

u^k(xi)=12Ωi,(49)\hat{u}_k(x_i) = \frac{1}{\sqrt{2\Omega_i}}, \tag*{(49)}
u^k,x(xi)=[iΩi−Ω,x(xi)2Ωi]u^k(xi).(50)\hat{u}_{k,x}(x_i) = \left[i\Omega_i - \frac{\Omega_{,x}(x_i)}{2\Omega_i}\right]\hat{u}_k(x_i). \tag*{(50)}

where we defined

Ωi≡Ω(xi),(51)\Omega_i \equiv\Omega(x_i), \tag*{(51)}
Ω(x)≡ω2(x).(52)\Omega(x) \equiv\sqrt{\omega^2(x)}. \tag*{(52)}

These initial data reduce to the plane-wave form in equation (48) as xi→∞x_i \to\infty. In the parameter range considered here, 0<d<10 < d < 1, the same asymptotic regime gives z2≃a2αz^2 \simeq a^2\alpha, and hence ∣R∣=∣u/(MPlz)∣∼x−d|\mathcal{R}| = |u/(M_{\mathrm{Pl}}z)| \sim x^{-d}.

At intermediate times the effective mass term cannot be neglected. Near the non-oscillatory transition, we use the canonical turning point ω2=0\omega^2 = 0, which generally differs from Hubble radius crossing κ=H~\kappa= \widetilde{\mathcal{H}}. At small α0\alpha_0 and with both TT and κ\kappa fixed, the exact coefficients give

ω2=κ2−2+3(1+d)2T−d(1−T−d)T2+O(α0).(53)\omega^2 = \kappa^2 - \frac{2 + 3(1+d)^2T^{-d}(1-T^{-d})}{T^2} + O(\alpha_0). \tag*{(53)}

The leading Hankel form further requires

α≪1,αϵ2≪1,x≫1,3(1+d)2T−d(1−T−d)≪2.(54)\alpha\ll1,\qquad\alpha\epsilon^2 \ll1,\qquad x \gg1,\qquad3(1+d)^2T^{-d}(1-T^{-d}) \ll2. \tag*{(54)}

Under these conditions, we have

ω2=cR2κ2−z,xxz≃κ2−2x2.(55)\omega^2 = c_{\mathcal{R}}^2\kappa^2 - \frac{z_{,xx}}{z} \simeq\kappa^2 - \frac{2}{x^2}. \tag*{(55)}

The corresponding Hankel solution gives the leading scale-invariant spectrum. Under the same assumptions, with a≃1a \simeq1 and (1+d)T−d/d≪1(1+d)T^{-d}/d \ll1, the growing curvature mode scales as ∣R∣∼x−1−d|\mathcal{R}| \sim x^{-1-d}. Toward the end of the Genesis phase, the modes approach a finite amplitude, although R\mathcal{R} is not exactly conserved.

The left panel of Figure 4 shows the numerical evolution of ∣R∣|\mathcal{R}| for d=0.3d = 0.3, α0=10−22\alpha_0 = 10^{-22}, and κ=10−7\kappa= 10^{-7}. One can visually observe the close agreement between the asymptotic behavior with its analytical estimates. We evaluate ∣R∣|\mathcal{R}| at xf=10−7x_f = 10^{-7}, close to the end of the Genesis phase at x=0x = 0. The power spectrum is

Figure 4. Left: evolution of R\mathcal{R} for d=0.3d = 0.3, α0=10−22\alpha_0 = 10^{-22}, and κ=10−7\kappa= 10^{-7}. The black curves show the x−1−dx^{-1-d} and x−dx^{-d} asymptotic scalings. Right: scalar power spectra for α0=10−22\alpha_0 = 10^{-22} and d=0.3d = 0.3, 0.50.5 and 0.70.7.

PR=κ32π2τB2MPl2∣u^k(xf)z(xf)∣2.(56)\mathcal{P}_{\mathcal{R}} = \frac{\kappa^3}{2\pi^2\tau_B^2 M_{\mathrm{Pl}}^2}\left|\frac{\hat{u}_k(x_f)}{z(x_f)}\right|^2 . \tag*{(56)}

The spectrum is evaluated at xfx_f after including the superhorizon evolution. Matching to a later phase requires both R\mathcal{R} and its conjugate momentum. For d=0.3d = 0.3, α0=10−23\alpha_0 = 10^{-23}, and κ=10−7\kappa= 10^{-7}, the power at xfx_f differs from its x=0x = 0 value by only 2.27×10−82.27 \times10^{-8}, although (dln⁡∣R∣/dln⁡a)0=0.08727(\mathrm{d}\ln|\mathcal{R}|/\mathrm{d}\ln a)_0 = 0.08727. Initial data are imposed only in the adiabatic oscillatory regime, where

ω2>0,∣cR2−1∣≪1,∣z,xxz∣/κ2≪1,∣ω,x/ω2∣≪1.(57)\omega^2 > 0,\qquad\left|c_{\mathcal{R}}^2 - 1\right| \ll1,\qquad\left|\frac{z_{,xx}}{z}\right|/\kappa^2 \ll1,\qquad\left|\omega_{,x}/\omega^2\right| \ll1 . \tag*{(57)}

The right panel of Figure 4 shows the scalar power spectrum of several values of dd. Their overall amplitude can be adjusted with τB\tau_B. The infrared spectrum is strongly blue, while an intermediate range is red.

We scan this intermediate regime on an 80380^3 grid in (κ,α0,d)(\kappa, \alpha_0, d):

10−30≤α0≤10−21,0.1≤d≤0.4,10−9≤κ≤10−4.(58)10^{-30} \leq\alpha_0 \leq10^{-21},\qquad0.1 \leq d \leq0.4,\qquad10^{-9} \leq\kappa\leq10^{-4}. \tag*{(58)}

Appendix § 6 gives a three-parameter phenomenological fit to the power spectrum. The scalar spectral index and its running can be obtained directly from the fit,

nsfit−1=dln⁡PRfitdln⁡κ,αsfit=dnsfitdln⁡κ.(59)n_s^{\mathrm{fit}} - 1 = \frac{\mathrm{d}\ln\mathcal{P}_{\mathcal{R}}^{\mathrm{fit}}}{\mathrm{d}\ln\kappa},\qquad\alpha_s^{\mathrm{fit}} = \frac{\mathrm{d}n_s^{\mathrm{fit}}}{\mathrm{d}\ln\kappa}. \tag*{(59)}

Since κ=−kτB\kappa= -k\tau_B, this is equivalent to differentiating with respect to ln⁡k\ln k at fixed τB\tau_B. Figure 5 compares the fit derivatives with derivatives obtained directly from the numerical spectrum. Here, we have αs=dns/dln⁡k\alpha_s = \mathrm{d}n_s/\mathrm{d}\ln k. Across the displayed range, the fit differs from the numerical derivatives by at most 1.5×10−31.5 \times10^{-3} in nsn_s and 1.3×10−31.3 \times10^{-3} in αs\alpha_s. For reference, the quoted Planck constraints [6] are

Figure 5. Spectral index nsn_s (left) and running αs\alpha_s (right) for α0=10−23\alpha_0 = 10^{-23}. Dots are derivatives of the numerical spectrum, without smoothing; Dashed curves are obtained from the phenomenological fit. The running panel uses d=0.30d = 0.30, 0.350.35 and 0.400.40. Gray bands show the quoted Planck 68% intervals.

ns=0.9649±0.0042,αs=−0.0045±0.0067.(60)n_s = 0.9649 \pm0.0042,\qquad\alpha_s = -0.0045 \pm0.0067. \tag*{(60)}

The quoted Planck values are separate 68% marginal constraints: nsn_s is obtained without running, whereas the αs\alpha_s constraint allows running. The mapping between κ\kappa and the observed pivot depends on τB\tau_B and the post-Genesis evolution, so these values are shown only as a reference.

Tensor modes

Because of the constraint structure of VCDM, tensor modes are unaffected, and their quadratic action is equation (27), the same as in GR. For the canonically normalized mode uλ=aMPlhλ/2u^\lambda= aM_{\rm Pl}h^\lambda/2,

d2uλdx2+(κ2−(1+d)21+(1+x)d(1+x)2+2d)uλ=0.(61)\frac{\mathrm{d}^2u^\lambda}{\mathrm{d}x^2}+\left(\kappa^2-(1+d)^2\frac{1+(1+x)^d}{(1+x)^{2+2d}}\right)u^\lambda=0. \tag*{(61)}

For either polarization, define u^T=uλ/−τB\widehat{u}_T=u^\lambda/\sqrt{-\tau_B} and normalize eijλeijσ∗=δλσe^\lambda_{ij}e^{\sigma*}_{ij}=\delta_{\lambda\sigma}. The total power sums the two equal polarizations. The power spectrum for the gravitational waves is given by

Ph=4κ3π2τB2MPl2∣u^T(0)a(0)∣2.(62)\mathcal{P}_h=\frac{4\kappa^3}{\pi^2\tau_B^2M_{\rm Pl}^2}\left|\frac{\widehat{u}_T(0)}{a(0)}\right|^2. \tag*{(62)}

Figure 6 shows the numerical spectrum for several values of dd. With this normalization, an independent numerical evolution reproduces Figure 6.

Figure 6. Tensor power spectrum for d=0.2d=0.2, 0.50.5, and 0.70.7. The black dashed line shows the κ2\kappa^2 scaling.

At leading order, the tensor-to-scalar ratio is r=Ph/PR∝α0≪1r=\mathcal{P}_h/\mathcal{P}_{\mathcal{R}}\propto\alpha_0\ll1, and we therefore observe that rr is strongly suppressed.

Geodesic completeness

To discuss geodesic completeness, we consider the tensor and scalar propagations separately. The tensor action (27) can be written as

ST(2)=−MPl28∫dτ d3x a2ημν∂μhij∂νhij,(63)S_T^{(2)}=-\frac{M_{\rm Pl}^2}{8}\int\mathrm{d}\tau\,\mathrm{d}^3x\,a^2\eta^{\mu\nu}\partial_\mu h_{ij}\partial_\nu h_{ij}, \tag*{(63)}

where ημν\eta^{\mu\nu} is the Minkowski metric. Therefore, as expected, the effective metric for the tensor perturbation is past geodesically complete. Indeed, a→1a \to1 as τ→−∞\tau\to-\infty, so both integrals ∫a2dτ\int a^2\mathrm{d}\tau and ∫a2dτ/a2+P2\int a^2\mathrm{d}\tau/\sqrt{a^2+P^2} diverge, where PP is the conserved comoving momentum.

The scalar sector is subtler because its kinetic coefficient and sound speed depend on momentum. A variable-dependent effective metric cannot by itself define spacetime completeness, and VCDM preserves only foliation-preserving diffeomorphisms. We therefore also use the quadratic action in a regular variable δχ\delta\chi, which is derived in appendix § 7.§ 1.

The quadratic actions for R\mathcal{R} and δχ\delta\chi are

SR2(2)=MPl22∫dτ d3k z2(∣Rk′∣2−cR2k2∣Rk∣2),(64)S_{\mathcal{R}^2}^{(2)}=\frac{M_{\mathrm{Pl}}^2}{2}\int\mathrm{d}\tau\,\mathrm{d}^3 k\,z^2\left(\left|\mathcal{R}'_k\right|^2-c_{\mathcal{R}}^2 k^2\left|\mathcal{R}_k\right|^2\right), \tag*{(64)}

and

Sδχ2(2)=12∫d3kdτ a2K(k,τ)(∣δχ′∣2−cδχ2(k,τ)k2∣δχ∣2).(65)S_{\delta\chi^2}^{(2)}=\frac{1}{2}\int\mathrm{d}^3 k\mathrm{d}\tau\,a^2 K(k,\tau)\left(\left|\delta\chi'\right|^2-c_{\delta\chi}^2(k,\tau)k^2\left|\delta\chi\right|^2\right). \tag*{(65)}

For k≠0k\ne0, z2>0z^2>0 throughout Genesis, while the kinetic coefficient KK of the δχ\delta\chi action has a pole at k⋆2=αH2/2k_\star^2=\alpha\mathcal{H}^2/2. This pole reflects the velocity reduction, not a singularity of the canonical phase-space evolution. Because δχ\delta\chi mixes the curvature coordinate and momentum, K<0K<0 does not imply a UV ghost instability (see [23]), and the region in the Fourier space where K<0K<0 needs not coincide with the region where cR2<0c_{\mathcal{R}}^2<0 or with the canonical non-oscillatory regime. Because the exact coefficients depend on momentum, the scalar sector cannot generally be represented by a single local effective metric. In the far past, an approximately massless δχ\delta\chi regime requires both k≫k⋆k\gg k_\star and k(−τ−τB)≫2(1+χ12/MPl2)k(-\tau-\tau_B)\gg\sqrt{2(1+\chi_1^2/M_{\mathrm{Pl}}^2)}. For R\mathcal{R}, the requirement is k2≫α(6+α−2ϵ)H2k^2\gg\alpha(6+\alpha-2\epsilon)\mathcal{H}^2. In these domains,

z2≃a2α,cR2k2≃k2+MR2,(66)z^2\simeq a^{2\alpha},\qquad c_{\mathcal{R}}^2k^2\simeq k^2+M_{\mathcal{R}}^2, \tag*{(66)}
K≃1,cδχ2k2≃k2.(67)K\simeq1,\qquad c_{\delta\chi}^2k^2\simeq k^2. \tag*{(67)}

Here MR2=(B1−A1)H2/8M_{\mathcal{R}}^2=(B_1-A_1)\mathcal{H}^2/8. The curvature mode is approximately massless only when ∣MR2∣≪k2\left|M_{\mathcal{R}}^2\right|\ll k^2. The corresponding asymptotic quadratic actions are

SR2(2)≃−MPl22∫d4x a2α(ημν∂μR∂νR+MR2R2),(68)S_{\mathcal{R}^2}^{(2)}\simeq-\frac{M_{\mathrm{Pl}}^2}{2}\int\mathrm{d}^4x\,a^{2\alpha}\left(\eta^{\mu\nu}\partial_\mu\mathcal{R}\partial_\nu\mathcal{R}+M_{\mathcal{R}}^2\mathcal{R}^2\right), \tag*{(68)}
Sδχ2(2)≃−12∫d4x a2ημν∂μδχ∂νδχ.(69)S_{\delta\chi^2}^{(2)}\simeq-\frac{1}{2}\int\mathrm{d}^4x\,a^2\eta^{\mu\nu}\partial_\mu\delta\chi\partial_\nu\delta\chi. \tag*{(69)}

Within these approximations, the matter kinetic metric approaches a2ημνa^2\eta_{\mu\nu}. For 0<d<10<d<1, the curvature kinetic metric has conformal factor a2αa^{2\alpha}, and ∫a2αdτ\int a^{2\alpha}\mathrm{d}\tau diverges in the past. Every fixed nonzero mode eventually enters this asymptotic regime, but the exact momentum-dependent scalar dynamics is not described by a single local metric at all wavelengths.

Cubic interactions

Strong coupling is a common concern in Genesis models (see, e.g., [14, 15, 16, 24, 25, 26, 27]). We therefore examine the cubic interactions. The linear definition ((28)) must be completed at second order. The temporal terms follow [22, 28], while the gradient term is obtained by transforming to uniform-field slices and restoring the spatial gauge. For scalar perturbations only, we obtain

R1=ζ1−Hχˉ′δχ1,(70)\mathcal{R}_1 = \zeta_1 - \frac{\mathcal{H}}{\bar{\chi}'}\delta\chi_1 , \tag*{(70)}
R2=ζ2−Hχˉ′δχ2+2Hχˉ′2δχ1δχ1′−2δχ1′χˉ′ζ1−βα2(Hδχ1χˉ′)2+ΔR2(∇).(71)\mathcal{R}_2 = \zeta_2 - \frac{\mathcal{H}}{\bar{\chi}'}\delta\chi_2 + \frac{2\mathcal{H}}{\bar{\chi}'^2}\delta\chi_1\delta\chi_1' - 2\frac{\delta\chi_1'}{\bar{\chi}'}\zeta_1 - \frac{\beta_\alpha}{2}\left(\frac{\mathcal{H}\delta\chi_1}{\bar{\chi}'}\right)^2 + \Delta\mathcal{R}_2^{(\nabla)} . \tag*{(71)}

With the convention δχ=δχ1+δχ2/2\delta\chi= \delta\chi_1 + \delta\chi_2/2, and similarly for ζ\zeta and R\mathcal{R}, the additional nonzero-momentum gradient term is

ΔR2(∇)=12(δij−∂i∂jΔ)Xij,(72)\Delta\mathcal{R}_2^{(\nabla)} = \frac{1}{2}\left(\delta^{ij} - \frac{\partial^i\partial^j}{\Delta}\right)X_{ij}, \tag*{(72)}
θ1=−δχ1χˉ′,(73)\theta_1 = -\frac{\delta\chi_1}{\bar{\chi}'}, \tag*{(73)}
Xij=∂iβ ∂jθ1+∂jβ ∂iθ1−∂iθ1 ∂jθ1.(74)X_{ij} = \partial_i\beta\,\partial_j\theta_1 + \partial_j\beta\,\partial_i\theta_1 - \partial_i\theta_1\,\partial_j\theta_1. \tag*{(74)}

Here H/χˉ′2=(1+d)(−τB)T1−d/χ12\mathcal{H}/\bar{\chi}'^2 = (1+d)(-\tau_B)T^{1-d}/\chi_1^2 diverges only for 0<d<10<d<1, while (χˉ′)−2∝T2(\bar{\chi}')^{-2}\propto T^2 grows for all d>0d>0. These growing coefficients make the nonlinear conversion nonuniform at fixed momentum, but they neither determine the full second-order solution nor establish physical strong coupling. The linear curvature mode still vanishes in the far past. Using R=R1+12R2\mathcal{R}=\mathcal{R}_1+\frac{1}{2}\mathcal{R}_2 to diagnose strong coupling can therefore be misleading.

Bouncing and Genesis models often avoid this problem by working in the flat gauge, but that gauge is unavailable in VCDM because time diffeomorphism invariance is broken. We instead analyze the regular variable δχ\delta\chi in a chosen spatial gauge. Appendix § 7.§ 1 gives its quadratic action and constraint structure. At fixed nonzero momenta, aK→1a\sqrt{K}\to1 in the far past, so δχ\delta\chi has a regular asymptotic normalization.

Absence of strong coupling below MPlM_{\mathrm{Pl}}

Taking δχ\delta\chi as the fundamental variable, the linear constraints are

δN=δχ′χˉ′+a2U,χχˉ′2δχ+Dζ,(75)\delta N = \frac{\delta\chi'}{\bar{\chi}'} + \frac{a^2U_{,\chi}}{\bar{\chi}'^2}\delta\chi+ \mathcal{D}\zeta, \tag*{(75)}
β=−3χˉ′2MPl2Δ−1δχ,(76)\beta= -\frac{3\bar{\chi}'}{2M_{\mathrm{Pl}}^2}\Delta^{-1}\delta\chi, \tag*{(76)}
ζ=(3−D)δχ′χˉ′−(3+D)a2U,χχˉ′2δχD(1+D),(77)\zeta= \frac{(3-\mathcal{D})\frac{\delta\chi'}{\bar{\chi}'}-(3+\mathcal{D})\frac{a^2U_{,\chi}}{\bar{\chi}'^2}\delta\chi}{\mathcal{D}(1+\mathcal{D})}, \tag*{(77)}

where

D≡2MPl2χˉ′2Δ.(78)\mathcal{D}\equiv\frac{2M_{\mathrm{Pl}}^2}{\bar{\chi}'^2}\Delta. \tag*{(78)}

At fixed nonzero momentum, the deep-Genesis limit x→∞x\to\infty gives

∣D∣=2MPl2k2τB2T2/χ12≫1,(79)|\mathcal{D}|=2M_{\mathrm{Pl}}^2 k^2\tau_B^2T^2/\chi_1^2\gg1 , \tag*{(79)}

and hence

δN∼O((χˉ′D)−1),β∼O((χˉ′D)−1),ζ∼O((χˉ′D)−1).(80)\delta N \sim\mathcal{O}\left((\bar{\chi}'\mathcal{D})^{-1}\right), \qquad\beta\sim\mathcal{O}\left((\bar{\chi}'\mathcal{D})^{-1}\right), \qquad\zeta\sim\mathcal{O}\left((\bar{\chi}'\mathcal{D})^{-1}\right). \tag*{(80)}

Thus, at fixed nonzero momentum, δN\delta N, β\beta, and ζ\zeta vanish in the far past. Appendix § 7.§ 2 shows that the pure-scalar cubic interactions are suppressed as x→∞x \to\infty.

Keeping only the leading terms, we obtain the cubic scalar interaction

Sδχ3(3)≃∫d4x a2{12(3ζ−δN)δχ′2−δχ′∂iβ ∂iδχ−δN+ζ2(∂kδχ)2}≃χ14MPl2(−τB)∫d4x 11+x[7(δχc′)2Δ−1δχc′−6δχc′∂iΔ−1δχc ∂iδχc+3(Δ−1δχc′)(∂iδχc)2].(81)\begin{aligned} S_{\delta\chi^3}^{(3)} \simeq\int d^4x\,a^2\left\{\frac{1}{2}(3\zeta-\delta N)\delta\chi'^2-\delta\chi'\partial_i\beta\,\partial_i\delta\chi-\frac{\delta N+\zeta}{2}(\partial_k\delta\chi)^2\right\} \\ &\simeq\frac{\chi_1}{4M_{\mathrm{Pl}}^2(-\tau_B)}\int d^4x\,\frac{1}{1+x}\left[7(\delta\chi_c')^2\Delta^{-1}\delta\chi_c'-6\delta\chi_c'\partial_i\Delta^{-1}\delta\chi_c\,\partial_i\delta\chi_c+3(\Delta^{-1}\delta\chi_c')(\partial_i\delta\chi_c)^2\right]. \tag*{(81)} \end{aligned}

Here, δχc=aK δχ\delta\chi_c=a\sqrt{K}\,\delta\chi is canonically normalized, with K\sqrt{K} acting mode by mode on the K>0K>0 branch. The approximation δχc≃aδχ\delta\chi_c\simeq a\delta\chi holds only asymptotically at fixed nonzero momentum and is not an exact local field redefinition at finite time. Since the cubic scalar interaction vanishes when x→∞x\to\infty, no cubic strong-coupling scale collapses to zero in this limit. This does not remove the ordinary matter–graviton interaction or establish finite-time, all-order control.

The mixed scalar–tensor interactions are analyzed in appendix § 7.§ 2, while the pure tensor cubic action is the GR one, which reads

Sh3(3)=MPl24∫d4x a2(hikhjℓ−12hijhkℓ)∂k∂ℓhij.(82)S_{h^3}^{(3)}=\frac{M_{\mathrm{Pl}}^2}{4}\int d^4x\,a^2\left(h_{ik}h_{j\ell}-\frac{1}{2}h_{ij}h_{k\ell}\right)\partial_k\partial_\ell h_{ij}. \tag*{(82)}

After canonical normalization, this interaction has the usual strength E/MPlE/M_{\mathrm{Pl}} and is weak for E≪MPlE\ll M_{\mathrm{Pl}}. The scalar–scalar–tensor interaction in the deep Genesis limit scales as

Sδχ2h(3)≃∫d4x 12hij∂iδχc∂jδχc,(83)S_{\delta\chi^2h}^{(3)}\simeq\int d^4x\,\frac{1}{2}h^{ij}\partial_i\delta\chi_c\partial_j\delta\chi_c, \tag*{(83)}

which is the only asymptotically non-vanishing contribution within the two-scalar/one-tensor sector. Finally, the tensor–tensor–scalar interactions vanish asymptotically, with the same O((χˉ′D)−1)\mathcal{O}((\bar{\chi}'\mathcal{D})^{-1}) suppression as the pure scalar sector.

The leading cubic interactions therefore scale as given in Table 1.

SectorCoupling
δχc3\delta\chi_c^3O(x−1)\mathcal{O}(x^{-1})
δχc2h\delta\chi_c^2hO(E/MPl)+O(x−2)\mathcal{O}(E/M_{\mathrm{Pl}})+\mathcal{O}(x^{-2})
δχch2\delta\chi_ch^2O(x−1)\mathcal{O}(x^{-1})
h3h^3O(E/MPl)\mathcal{O}(E/M_{\mathrm{Pl}})

Table 1. Leading cubic interactions scalings.

where E=k/aE=k/a is the physical momentum. Here hc,ij=aMPlhij/2h_{c,ij}=aM_{\mathrm{Pl}}h_{ij}/2; the time scalings assume fixed parameters and generic nonzero momenta, excluding parametrically soft internal momenta. Thus, all background-dependent cubic couplings vanish in the asymptotic past, while the remaining matter–graviton and graviton self-interactions are Planck suppressed. Therefore, the cubic strong-coupling scale is parametrically

Λcubic∼MPl.(84)\Lambda_{\mathrm{cubic}}\sim M_{\mathrm{Pl}}. \tag*{(84)}

Thus, at cubic order there is no new cutoff besides the Planck mass MPlM_{\mathrm{Pl}}.

Conclusion

We have constructed an asymptotically Minkowski Genesis cosmology in VCDM coupled to a canonical scalar field. The modified gravitational equations allow

H˙/H2>0,ρχ+pχ=χ˙2>0,(85)\dot{H}/H^2>0,\qquad\rho_\chi+p_\chi=\dot{\chi}^{2}>0, \tag*{(85)}

so that the effective NEC-violating expansion does not require NEC violation in the matter sector. The gravitational sector has only two propagating tensor degrees of freedom. The background metric, which also governs tensor propagation, is past geodesically complete.

The curvature perturbation has a positive kinetic coefficient throughout the Genesis epoch, and its short-wavelength sound speed approaches unity, cR2→1c_{\mathcal{R}}^2\to1. At finite momentum, cR2c_{\mathcal{R}}^2 can become negative within an infrared band. This is not an ultraviolet gradient instability and, in general, does not coincide with the non-oscillatory regime of the canonical mode, which is determined by the sign of ω2=cR2κ2−z,xx/z\omega^2=c_{\mathcal{R}}^2\kappa^2-z_{,xx}/z. For the parameter range considered, the region with cR2<0c_{\mathcal{R}}^2<0 shrinks toward zero momentum in the asymptotic past, while every fixed nonzero mode satisfies ω2→κ2>0\omega^2\to\kappa^2>0 and therefore becomes oscillatory.

The numerical scalar spectrum exhibits a nearly scale-invariant, slightly red-tilted regime over an intermediate range of scales. Under the controlled turning-point approximations of § 3, the relevant modes obey

d2udx2+(κ2−2x2)u∼0,(86)\frac{\mathrm{d}^2u}{\mathrm{d}x^2}+\left(\kappa^2-\frac{2}{x^2}\right)u\sim0, \tag*{(86)}

which accounts for the leading scale-invariant behavior. The curvature perturbation then undergoes non-oscillatory growth, including a superhorizon stage for the modes studied, and approaches a finite amplitude at the end of Genesis. The tensor sector has the GR quadratic action and produces a blue spectrum, with

Ph∝κ2,r∝α0≪1,(87)\mathcal{P}_h\propto\kappa^2,\qquad r\propto\alpha_0\ll1, \tag*{(87)}

for the parameter region studied.

The cubic analysis highlights the importance of a regular perturbation variable. The nonlinear transformation to the curvature perturbation can become singular in the deep Genesis limit, whereas the matter fluctuation δχ\delta\chi has a regular asymptotic normalization. At fixed nonzero momenta and away from the singular surface D=−1\mathcal{D}=-1 of the matter-variable velocity reduction, we find that the background-dependent cubic couplings vanish in the asymptotic past. The surviving matter–graviton and graviton self-interactions remain Planck suppressed. The resulting cubic power counting gives Λcubic∼MPl\Lambda_{\mathrm{cubic}} \sim M_{\mathrm{Pl}} in this limit, with no cubic strong-coupling scale that collapses to zero toward the past. This asymptotic result does not determine the full nonlinear effective field theory cutoff.

A complete early-Universe scenario will require a systematic analysis of higher-order interactions, together with an explicit matching of both the background evolution and perturbations onto a subsequent inflationary phase and/or reheating epoch. Since the VCDM, as one of type-II minimally modified gravity theories, does not have any extra gravitational degrees of freedom beyond the ordinary tensor gravitons with MPlM_{\mathrm{Pl}}-suppressed interactions, it serves as a promising framework that potentially avoids strong coupling of higher-order interactions in early-universe cosmological scenarios. We leave these important questions for future work.

Acknowledgments

We would like to thank Ryo Namba for helpful comments and discussions. The work of P.P. is supported by IBS under the project code IBS-R018-D3. A.G. is supported by the DFG under the Emmy-Noether program, project number 496592360. P.M. is supported by grants from the Research Grants Council of Hong Kong (Project Nos. CUHK 14304622, 14307923, and 14307724), the start-up grant from The Chinese University of Hong Kong, and the Direct Grant for Research from the Research Committee of The Chinese University of Hong Kong. The work of S.M. was supported in part by Japan Society for the Promotion of Science Grants-in-Aid for Scientific Research No. 24K07017 and the World Premier International Research Center Initiative (WPI), MEXT, Japan.

Phenomenological fit

The phenomenological template factors out the pieces whose scaling is already understood analytically and fits only the smooth remainder. In the scanned small-α0\alpha_0 regime, z2(xf)∝α0z^2(x_f) \propto\alpha_0 motivates an explicit α0−1\alpha_0^{-1} factor. Residual α0\alpha_0 dependence of the evolved mode is included in the fit error. The dimensional factor (τB2MPl2)−1(\tau_B^2 M_{\mathrm{Pl}}^2)^{-1} is also kept explicit, since, at fixed dimensionless κ\kappa, it fixes the overall energy scale. The remaining dependence on dd and κ\kappa is described by an effective growth function, which approximates the accumulated evolution from the oscillatory regime to the finite amplitude at the end of Genesis.

The resulting fit is

PRfit=14π2τB2MPl2α0exp⁡[−2(1+d)d+2J(d,κ)d+c0+c1d]Cνq.(88)\mathcal{P}_{\mathcal{R}}^{\mathrm{fit}}=\frac{1}{4\pi^2\tau_B^2M_{\mathrm{Pl}}^2\alpha_0}\exp\left[-\frac{2(1+d)}{d}+\frac{2J(d,\kappa)}{d}+c_0+c_1d\right]C_\nu^q. \tag*{(88)}

Here

Cν=22ν−3[Γ(ν)Γ(3/2)]2,(89)C_\nu=2^{2\nu-3}\left[\frac{\Gamma(\nu)}{\Gamma(3/2)}\right]^2, \tag*{(89)}

and we define

A(d)=3(1+d)2,ξ(d,κ)=κd.(90)A(d)=3(1+d)^2,\qquad\xi(d,\kappa)=\kappa^d. \tag*{(90)}

The effective index is

ν⋆=[94+Aξ(1−ξ)]1/2(91)\nu_\star= \left[\frac{9}{4} + A\xi(1-\xi)\right]^{1/2} \tag*{(91)}

and the auxiliary growth variable is

u⋆=A(1−ξ)ν⋆+3/2.(92)u_\star= \frac{A(1-\xi)}{\nu_\star+ 3/2}. \tag*{(92)}

The function J(d,κ)J(d,\kappa) is given by

J(d,κ)=3+Aarctan⁡(u⋆A)+32ln⁡(AA+u⋆2)−A(3+u⋆)A+u⋆2.(93)J(d,\kappa) = 3 + \sqrt{A}\arctan\left(\frac{u_\star}{\sqrt{A}}\right) + \frac{3}{2}\ln\left(\frac{A}{A+u_\star^2}\right) - \frac{A(3+u_\star)}{A+u_\star^2}. \tag*{(93)}

In the formal template limit κd→0\kappa^d \to0,

ν⋆→32,Cν⋆→1.(94)\nu_\star\to\frac{3}{2}, \qquad C_{\nu_\star} \to1. \tag*{(94)}

However, J(d,κ)J(d,\kappa) does not vanish in this limit. Instead,

J(d,κ)→J0(d),(95)J(d,\kappa) \to J_0(d), \tag*{(95)}

where

J0(d)=3(1+d)arctan⁡(1+d3)+32ln⁡[33+(1+d)2].(96)J_0(d) = \sqrt{3}(1+d)\arctan\left(\frac{1+d}{\sqrt{3}}\right) + \frac{3}{2}\ln\left[\frac{3}{3+(1+d)^2}\right]. \tag*{(96)}

The template therefore has a formally scale-independent infrared limit, but this should not be extrapolated beyond the scan. Only part of the scan is nearly scale-invariant. For example, d=0.1d=0.1, α0=10−23\alpha_0=10^{-23}, and κ=10−4\kappa=10^{-4} give ns≃0.48958n_s \simeq0.48958 from the full mode equation. The template fits amplitudes across the full scan and does not imply near scale invariance everywhere. The dd-dependent constant J0(d)J_0(d) contributes only to the overall amplitude. The factor Cν⋆C_{\nu_\star} has the form of the usual Hankel-amplitude factor, while the exponent qq parametrizes the residual departure from a pure Hankel approximation.

The numerical coefficients are

c0=0.0139557668608,(97)c_0 = 0.0139557668608, \tag*{(97)}
c1=−0.275286738213,(98)c_1 = -0.275286738213, \tag*{(98)}
q=1.32186006083.(99)q = 1.32186006083. \tag*{(99)}

They are obtained by a fit that minimizes the maximum logarithmic error. Explicitly, after removing the known contribution

ln⁡PRbase=−ln⁡(4π2TB2MPl2α0)−2(1+d)d+2J(d,κ)d,(100)\ln\mathcal{P}_{\mathcal{R}}^{\mathrm{base}} = -\ln\left(4\pi^2 T_B^2 M_{\mathrm{Pl}}^2\alpha_0\right) - \frac{2(1+d)}{d} + \frac{2J(d,\kappa)}{d}, \tag*{(100)}

we fit the residual

yi=ln⁡PR,inum−ln⁡PR,ibase(101)y_i = \ln\mathcal{P}_{\mathcal{R},i}^{\mathrm{num}} - \ln\mathcal{P}_{\mathcal{R},i}^{\mathrm{base}} \tag*{(101)}

by

yi≃c0+c1di+qln⁡Cν⋆,i.(102)y_i \simeq c_0 + c_1d_i + q\ln C_{\nu_{\star,i}}. \tag*{(102)}
α0\alpha_0ddκ\kappaPRfit\mathcal{P}^{\mathrm{fit}}_{\mathcal{R}}PRnum\mathcal{P}^{\mathrm{num}}_{\mathcal{R}}Δrel(%)\Delta_{\mathrm{rel}}(\%)
7.69265×10−227.69265 \times10^{-22}0.1607590.1607592.19671×10−72.19671 \times10^{-7}1.4064197×10161.4064197 \times10^{16}1.4067648×10161.4067648 \times10^{16}0.1520.152
1.09138×10−241.09138 \times10^{-24}0.1835440.1835443.02700×10−63.02700 \times10^{-6}2.2404214×10192.2404214 \times10^{19}2.2432384×10192.2432384 \times10^{19}0.1260.126
4.55227×10−224.55227 \times10^{-22}0.2518990.2518995.91769×10−85.91769 \times10^{-8}7.0641840×10177.0641840 \times10^{17}7.0622356×10177.0622356 \times10^{17}0.02760.0276
2.26168×10−252.26168 \times10^{-25}0.1645570.1645576.27290×10−66.27290 \times10^{-6}2.9927273×10192.9927273 \times10^{19}2.9948496×10192.9948496 \times10^{19}0.07090.0709

Table 2. Sample comparison between the phenomenological fit and the numerical spectrum. Here τB=−1\tau_B=-1 and MPl=1M_{\mathrm{Pl}}=1, as in the dimensionless fitting convention, and Δrel=100∣PRfit/PRnum−1∣\Delta_{\mathrm{rel}}=100\lvert\mathcal{P}^{\mathrm{fit}}_{\mathcal{R}}/\mathcal{P}^{\mathrm{num}}_{\mathcal{R}}-1\rvert.

The parameters are fixed by minimizing the maximal absolute logarithmic error,

min⁡c0,c1,q,t t,∣c0+c1di+qln⁡Cν⋆,i−yi∣≤t,(103)\min_{c_0,c_1,q,t}\ t,\qquad\left|c_0+c_1d_i+q\ln C_{\nu^\star,i}-y_i\right|\leq t, \tag*{(103)}

which is a linear-programming problem once the template functions are fixed.

The numerical scan uses an 80×80×8080 \times80 \times80 grid,

10−30≤α0≤10−21,0.1≤d≤0.4,10−9≤κ≤10−4.(104)10^{-30}\leq\alpha_0\leq10^{-21},\qquad0.1\leq d\leq0.4,\qquad10^{-9}\leq\kappa\leq10^{-4}. \tag*{(104)}

In the data file, the column denoted by alpha corresponds to α0\alpha_0. The fit should not be extrapolated outside this range. The precomputed numerical table, fitting notebook, Wolfram notebook, and validation plots are available in the companion GitHub repository1. The table stored in the repository should be interpreted as the dimensionless quantity τB2MPl2PR\tau_B^2M_{\mathrm{Pl}}^2\mathcal{P}_{\mathcal{R}}; the physical amplitude follows from the explicit factor (τB2MPl2)−1(\tau_B^2M_{\mathrm{Pl}}^2)^{-1} in equation (A.1).

The verification set holds out complete (d,κ)(d,\kappa) pairs while retaining all α0\alpha_0 values. After removing the leading α0−1\alpha_0^{-1} scaling, this mainly tests interpolation in (d,κ)(d,\kappa). For the scalar amplitude, the mean relative error is about 0.11%0.11\%, and the maximum remains below 0.28%0.28\% on both the training and verification sets.

Absence of asymptotic cubic strong coupling

Reduced action and scalar perturbations

The λ0\lambda_0 equation derived from the VCDM action (2) gives

λ0=−23(K+φ).(105)\lambda_0=-\frac{2}{3}(K+\varphi). \tag*{(105)}

Substituting this result back into the action gives

SVCDM=MPl22∫d4x Nγ[(3)R+KijKij−13K2+43φK+23φ2−2V(φ)−2λi∂iφN].(106)S_{\mathrm{VCDM}}=\frac{M_{\mathrm{Pl}}^2}{2}\int d^4x\,N\sqrt{\gamma}\left[{}^{(3)}R+K_{ij}K^{ij}-\frac{1}{3}K^2+\frac{4}{3}\varphi K+\frac{2}{3}\varphi^2-2V(\varphi)-2\lambda^i\frac{\partial_i\varphi}{N}\right]. \tag*{(106)}

At linear order, δN\delta N, β\beta, and ζ\zeta are nondynamical. Varying them gives the constraints

δN=δχ′χˉ′+a2U,χχˉ′2δχ+Dζ,β=−3χˉ′2MPl2Δ−1δχ,ζ=(3−D)δχ′χˉ′−(3+D)a2U,χχˉ′2δχD(1+D).(107)\delta N=\frac{\delta\chi'}{\bar{\chi}'}+\frac{a^2U_{,\chi}}{\bar{\chi}'^2}\delta\chi+\mathcal{D}\zeta,\qquad \beta=-\frac{3\bar{\chi}'}{2M_{\mathrm{Pl}}^2}\Delta^{-1}\delta\chi,\qquad \zeta=\frac{(3-\mathcal{D})\frac{\delta\chi'}{\bar{\chi}'}-(3+\mathcal{D})\frac{a^2U_{,\chi}}{\bar{\chi}'^2}\delta\chi}{\mathcal{D}(1+\mathcal{D})}. \tag*{(107)}
Relative error on the full grid and verification set

Figure 7. (a) Relative error on the full grid. (b) Relative error on the verification set. Accuracy of the phenomenological fit for the scalar power spectrum. The left panel shows the error over the full scan, while the right panel shows the held-out verification points. These tests indicate that the template interpolates the scanned (d,κ)(d,\kappa) region accurately. They do not constitute a validation of extrapolations outside the scan range.

where

D≡2MPl2χˉ′ 2Δ.(108)\mathcal{D} \equiv\frac{2M_{\mathrm{Pl}}^2}{\bar{\chi}'^{\,2}}\Delta. \tag*{(108)}

At D=−1\mathcal{D}=-1, the algebraic constraint solution at fixed (δχ,δχ′)(\delta\chi,\delta\chi') loses rank; consistency requires

2χˉ′δχ′=a2U,χδχ.2\bar{\chi}'\delta\chi'=a^2U_{,\chi}\delta\chi.

For nonzero momentum, the canonical Hamiltonian system remains regular; this is a singularity of the velocity reduction, not of the physical linear evolution. Using the constraint equations, the quadratic action reads in Fourier space

Sδχ(2)=12∫dτ∫d3k a2K(k,τ)[∣δχ′∣2−cδχ2(k,τ)k2∣δχ∣2],(109)S_{\delta\chi}^{(2)}=\frac{1}{2}\int d\tau\int d^3k\,a^2K(k,\tau)\left[|\delta\chi'|^2-c_{\delta\chi}^2(k,\tau)k^2|\delta\chi|^2\right], \tag*{(109)}

where

K=(κ~2+3)2κ~2(κ~2−1),(110)K=\frac{(\tilde{\kappa}^2+3)^2}{\tilde{\kappa}^2(\tilde{\kappa}^2-1)}, \tag*{(110)}
cδχ2=κ~2−13+κ~2−4MPl2(1+x)−2d(κ~2−1)(κ~2+3)κ~2χ12(−2(1+d)2(κ~2−1)κ~2+(1+d)(1+x)d(6+d(κ~2−3)(κ~2−1)+κ~2(κ~2−11))+(1+x)2d(3+κ~2(κ~2−6))).(111)c_{\delta\chi}^2=\frac{\tilde{\kappa}^2-1}{3+\tilde{\kappa}^2}-\frac{4M_{\mathrm{Pl}}^2(1+x)^{-2d}}{(\tilde{\kappa}^2-1)(\tilde{\kappa}^2+3)\tilde{\kappa}^2\chi_1^2}\left(-2(1+d)^2(\tilde{\kappa}^2-1)\tilde{\kappa}^2+(1+d)(1+x)^d\left(6+d(\tilde{\kappa}^2-3)(\tilde{\kappa}^2-1)+\tilde{\kappa}^2(\tilde{\kappa}^2-11)\right)+(1+x)^{2d}\left(3+\tilde{\kappa}^2(\tilde{\kappa}^2-6)\right)\right). \tag*{(111)}

In Fourier space, D=−κ~2\mathcal{D}=-\tilde{\kappa}^2 with κ~2=2MPl2k2/χˉ′ 2\tilde{\kappa}^2=2M_{\mathrm{Pl}}^2k^2/\bar{\chi}'^{\,2}. The velocity-reduced δχ\delta\chi action has a pole at

k⋆(x)=χ12MPl(−τB)(1+x).(112)k_\star(x)=\frac{\chi_1}{\sqrt{2}M_{\mathrm{Pl}}(-\tau_B)(1+x)}. \tag*{(112)}

For 0<k<k⋆0<k<k_\star, the velocity-reduced coefficient KK is negative. Because δχ\delta\chi mixes the curvature coordinate and momentum, this does not signal an additional ghost; the regular curvature or canonical phase-space description should be used, and this interval need not coincide with the region with CR2<0\mathcal{C}_{R}^{2}<0 or the canonical non-oscillatory regime. At fixed nonzero momentum, x→∞x\to\infty implies κ~→∞\tilde{\kappa}\to\infty, and the leading terms are

K≃1+7κ~2,cδχ2≃1−4MPl2+χ12χ12κ~2.(113)K\simeq1+\frac{7}{\tilde{\kappa}^{2}},\qquad c_{\delta\chi}^{2}\simeq1-4\frac{M_{\mathrm{Pl}}^{2}+\chi_{1}^{2}}{\chi_{1}^{2}\tilde{\kappa}^{2}}. \tag*{(113)}

Neglecting the sound-speed correction also requires κ~2≫4(1+MPl2/χ12)\tilde{\kappa}^{2}\gg4(1+M_{\mathrm{Pl}}^{2}/\chi_{1}^{2}), while κ~≫1\tilde{\kappa}\gg1 controls only the kinetic correction, and subleading background terms must also be small.

Cubic action

Let A=(n,ζ,β)\mathcal{A}=(n,\zeta,\beta), with n≡δNn\equiv\delta N, denote the scalar auxiliary fields. The transverse shift vanishes at first order, Si(1)=0S_{i}^{(1)}=0. We expand A=A(1)+A(2)+⋯\mathcal{A}=\mathcal{A}^{(1)}+\mathcal{A}^{(2)}+\cdots, where A(n)\mathcal{A}^{(n)} denotes the nnth-order contribution, including the factorial factor. Away from singular constraint surfaces,

Sred(3)=S3[A(1)]+δS2δA∣A(1)A(2)=S3[A(1)].(114)\begin{aligned} S_{\mathrm{red}}^{(3)}=S_{3}[\mathcal{A}^{(1)}]+\left.\frac{\delta S_{2}}{\delta\mathcal{A}}\right|_{\mathcal{A}^{(1)}}\mathcal{A}^{(2)} \\ &=S_{3}[\mathcal{A}^{(1)}]. \tag*{(114)} \end{aligned}

Hence only the linear solutions are needed for the complete reduced cubic action. After using the background equations and discarding total derivatives, the complete cubic action for the scalar sector is

Sδχ3(3)=∫dτ d3x a2{2MPl2ζ,iδN,jδij+MPl2(δN+ζ)(∂ζ)2+MPl28(3ζ−δN)BjiBij+12(3ζ−δN)δχ2−δχ′∂iβ ∂iδχ+χˉ′δχ(δN2−3δNζ+92ζ2)−χˉ′(3ζ−δN)∂iβ ∂iδχ+χˉ′′2δN2(3ζ−δN)−12(δN+ζ)(∂δχ)2−a2[16U,χχχδχ3+12U,χχ(δN+3ζ)δχ2+U,χ(3δNζ+92ζ2)δχ]}.(115)S_{\delta\chi^{3}}^{(3)}=\int d\tau\,d^{3}x\,a^{2}\left\{2M_{\mathrm{Pl}}^{2}\zeta_{,i}\delta N_{,j}\delta_{ij}+M_{\mathrm{Pl}}^{2}(\delta N+\zeta)(\partial\zeta)^{2}+\frac{M_{\mathrm{Pl}}^{2}}{8}(3\zeta-\delta N)B^{i}_{j}B^{j}_{i} +\frac{1}{2}(3\zeta-\delta N)\delta\chi^{2}-\delta\chi^{\prime}\partial_{i}\beta\,\partial_{i}\delta\chi+\bar{\chi}^{\prime}\delta\chi\left(\delta N^{2}-3\delta N\zeta+\frac{9}{2}\zeta^{2}\right) -\bar{\chi}^{\prime}(3\zeta-\delta N)\partial_{i}\beta\,\partial_{i}\delta\chi+\frac{\bar{\chi}^{\prime\prime}}{2}\delta N^{2}(3\zeta-\delta N)-\frac{1}{2}(\delta N+\zeta)(\partial\delta\chi)^{2} -a^{2}\left[\frac{1}{6}U_{,\chi\chi\chi}\delta\chi^{3}+\frac{1}{2}U_{,\chi\chi}(\delta N+3\zeta)\delta\chi^{2}+U_{,\chi}\left(3\delta N\zeta+\frac{9}{2}\zeta^{2}\right)\delta\chi\right]\right\}. \tag*{(115)}

where

Bji≡2(∂i∂j−13δjiΔ)β.(116)B^{i}_{j}\equiv2\left(\partial^{i}\partial_{j}-\frac{1}{3}\delta^{i}_{j}\Delta\right)\beta. \tag*{(116)}

The two-scalar/one-tensor terms are

Sδχ2h(3)=∫dτ d3x a2{12hij∂iδχ ∂jδχ−MPl2hij(2∂iδN ∂jζ+∂iζ ∂jζ)+MPl24[−(3ζ−δN)h′ijBij+Bij∂kβ ∂khij]}.(117)S_{\delta\chi^{2}h}^{(3)}=\int d\tau\,d^{3}x\,a^{2}\left\{\frac{1}{2}h^{ij}\partial_{i}\delta\chi\,\partial_{j}\delta\chi-M_{\mathrm{Pl}}^{2}h^{ij}(2\partial_{i}\delta N\,\partial_{j}\zeta+\partial_{i}\zeta\,\partial_{j}\zeta) +\frac{M_{\mathrm{Pl}}^{2}}{4}\left[-(3\zeta-\delta N)h^{\prime ij}B_{ij}+B^{ij}\partial^{k}\beta\,\partial_{k}h_{ij}\right]\right\}. \tag*{(117)}

The one-scalar/two-tensor terms are

Sδχh2(3)=MPl22∫dτ d3x a2(δN+ζ)R~(2)[h]+MPl28∫dτ d3x a2[(3ζ−δN)hjihi′j−2hji∂kβ ∂khij].(118)S_{\delta\chi h^{2}}^{(3)}=\frac{M_{\mathrm{Pl}}^{2}}{2}\int d\tau\,d^{3}x\,a^{2}(\delta N+\zeta)\widetilde{R}^{(2)}[h] +\frac{M_{\mathrm{Pl}}^{2}}{8}\int d\tau\,d^{3}x\,a^{2}\left[(3\zeta-\delta N)h^{i}_{j}h^{\prime j}_{i}-2h^{i}_{j}\partial^{k}\beta\,\partial_{k}h^{j}_{i}\right]. \tag*{(118)}

where $\widetilde{\mathcal{R}}^{(2)}[h] \equiv 3_{\mathcal{O}(h^2)}isthequadraticintrinsic−curvaturecontribution,alocalsumofis the quadratic intrinsic-curvature contribution, a local sum ofh,\partial^2 handand(\partial h)^2.Thereisnodirectmatter. There is no direct matter\delta\chi h^2vertexbecausevertex because\det(e^h)=1andand\partial_i\bar{\chi}=0$.

For pure tensors the complete action reads

Sh3(3)=MPl24∫dτ d3x a2(hikhjℓ−12hijhkℓ)∂k∂ℓhij(119)S_{h^3}^{(3)} = \frac{M_{\mathrm{Pl}}^2}{4}\int\mathrm{d}\tau\,\mathrm{d}^3x\,a^2\left(h_{ik}h_{j\ell}-\frac{1}{2}h_{ij}h_{k\ell}\right)\partial_k\partial_\ell h_{ij} \tag*{(119)}

up to a spatial boundary term.

Equations (B.11)–(B.15) give the possible classes of cubic operators shown in Table 3, where A∈{n,ζ,kβ}A\in\{n,\zeta,k\beta\} and ∂\partial denotes either a time or spatial derivative.

SectorCubic classes
δχ3\delta\chi^3MPl21∂2A3,(∂δχ)2A,χˉ′δχ A2,χˉ′2A3,U,χχχδχ3,U,χχAδχ2,U,χA2δχ\frac{M_{\mathrm{Pl}}^2}{1}\partial^2 A^3,\quad(\partial\delta\chi)^2A,\quad\bar{\chi}'\delta\chi\,A^2,\quad\bar{\chi}'^2A^3,\quad U_{,\chi\chi\chi}\delta\chi^3,\quad U_{,\chi\chi}A\delta\chi^2,\quad U_{,\chi}A^2\delta\chi
δχ2h\delta\chi^2hh(∂δχ)2,MPl2h ∂2A2h(\partial\delta\chi)^2,\quad M_{\mathrm{Pl}}^2h\,\partial^2A^2
δχh2\delta\chi h^2MPl2A ∂2h2M_{\mathrm{Pl}}^2A\,\partial^2h^2
h3h^3MPl2h2∂2hM_{\mathrm{Pl}}^2h^2\partial^2h

Table 3. Possible classes of cubic operators.

Using the constraints (B.3), at fixed nonzero momentum the deep-Genesis limit x→∞x\to\infty gives ∣D∣≫1|\mathcal{D}|\gg1. In this limit, δN\delta N, β\beta, and ζ\zeta are suppressed by (χˉ′D)−1(\bar{\chi}'\mathcal{D})^{-1}:

δN≃4(χˉ′D)−1δχ′−2a2U,χχˉ′(χˉ′D)−1δχ≃O(x−1)δχ′+O(x−2)δχ,(120)\delta N \simeq4(\bar{\chi}'\mathcal{D})^{-1}\delta\chi' - 2a^2\frac{U_{,\chi}}{\bar{\chi}'}(\bar{\chi}'\mathcal{D})^{-1}\delta\chi\simeq\mathcal{O}(x^{-1})\delta\chi' + \mathcal{O}(x^{-2})\delta\chi, \tag*{(120)}
β=−3(χˉ′D)−1δχ≃O(x−1)δχ,(121)\beta=-3(\bar{\chi}'\mathcal{D})^{-1}\delta\chi\simeq\mathcal{O}(x^{-1})\delta\chi, \tag*{(121)}
ζ≃−(χˉ′D)−1δχ′−a2U,χχˉ′(χˉ′D)−1δχ≃O(x−1)δχ′+O(x−2)δχ.(122)\zeta\simeq-(\bar{\chi}'\mathcal{D})^{-1}\delta\chi' - a^2\frac{U_{,\chi}}{\bar{\chi}'}(\bar{\chi}'\mathcal{D})^{-1}\delta\chi\simeq\mathcal{O}(x^{-1})\delta\chi' + \mathcal{O}(x^{-2})\delta\chi. \tag*{(122)}

Thus, at fixed nonzero momentum, δN\delta N, β\beta, and ζ\zeta vanish asymptotically. The leading pure-scalar cubic interaction is (∂δχ)2A(\partial\delta\chi)^2A, while the surviving interaction is Planck suppressed.

References

  1. [1]P. Creminelli, A. Nicolis and E. Trincherini, Galilean Genesis: An Alternative to inflation, JCAP 11 (2010) 021 [1007.0027].
  2. [2]P. Creminelli, K. Hinterbichler, J. Khoury, A. Nicolis and E. Trincherini, Subluminal Galilean Genesis, JHEP 02 (2013) 006 [1209.3768].DOI
  3. [3]P. Creminelli, M.A. Luty, A. Nicolis and L. Senatore, Starting the Universe: Stable Violation of the Null Energy Condition and Non-standard Cosmologies, JHEP 12 (2006) 080 [hep-th/0606090].DOI
  4. [4]V.A. Rubakov, The Null Energy Condition and its violation, Phys. Usp. 57 (2014) 128 [1401.4024].
  5. [5]S. Nishi and T. Kobayashi, Generalized Galilean Genesis, JCAP 03 (2015) 057 [1501.02553].
  6. [6]Planck collaboration, Planck 2018 results. X. Constraints on inflation, Astron. Astrophys. 641 (2020) A10 [1807.06211].
  7. [7]Planck collaboration, Planck 2018 results. IX. Constraints on primordial non-Gaussianity, Astron. Astrophys. 641 (2020) A9 [1905.05697].
  8. [8]M. Libanov, S. Mironov and V. Rubakov, Generalized Galileons: instabilities of bouncing and Genesis cosmologies and modified Genesis, JCAP 08 (2016) 037 [1605.05992].DOI
  9. [9]T. Kobayashi, Generic instabilities of nonsingular cosmologies in Horndeski theory: A no-go theorem, Phys. Rev. D 94 (2016) 043511 [1606.05831].DOI
  10. [10]R. Kolevatov, S. Mironov, N. Sukhov and V. Volkova, Cosmological bounce and Genesis beyond Horndeski, JCAP 08 (2017) 038 [1705.06626].DOI
  11. [11]S. Mironov, V. Rubakov and V. Volkova, Genesis with general relativity asymptotics in beyond Horndeski theory, Phys. Rev. D 100 (2019) 083521 [1905.06249].DOI
  12. [12]A. Ilyas, M. Zhu, Y. Zheng and Y.-F. Cai, Emergent Universe and Genesis from the DHOST Cosmology, JHEP 01 (2021) 141 [2009.10351].DOI
  13. [13]M. Zhu and Y. Zheng, Improved DHOST Genesis, JHEP 11 (2021) 163 [2109.05277].DOI
  14. [14]Y. Ageeva, O. Evseev, O. Melichev and V. Rubakov, Toward evading the strong coupling problem in Horndeski genesis, Phys. Rev. D 102 (2020) 023519 [2003.01202].DOI
  15. [15]Y. Ageeva, P. Petrov and V. Rubakov, Horndeski genesis: consistency of classical theory, JHEP 12 (2020) 107 [2009.05071].DOI
  16. [16]H. Gil Choi, P. Petrov and M. Yamaguchi, Can Horndeski Genesis be nonpathological?, JHEP 08 (2025) 044 [2503.02626].DOI
  17. [17]A. De Felice, A. Doll and S. Mukohyama, A theory of type-II minimally modified gravity, JCAP 09 (2020) 034 [2004.12549].DOI
  18. [18]N. Afshordi, D.J.H. Chung and G. Geshnizjani, Cuscuton: A Causal Field Theory with an Infinite Speed of Sound, Phys. Rev. D 75 (2007) 083513 [hep-th/0609150].DOI
  19. [19]A. De Felice, K.-i. Maeda, S. Mukohyama and M.C. Pookkillath, Comparison of two theories of Type-IIa minimally modified gravity, Phys. Rev. D 106 (2022) 024028 [2204.08294].DOI
  20. [20]A. De Felice, S. Mukohyama and M.C. Pookkillath, Addressing H₀ tension by means of VCDM, Phys. Lett. B 816 (2021) 136201 [2009.08718].DOI
  21. [21]A. Ganz, P. Martens, S. Mukohyama and R. Namba, Bouncing cosmology in VCDM, JCAP 04 (2023) 060 [2212.13561].DOI
  22. [22]A. Ganz, P. Martens, S. Mukohyama and R. Namba, Bispectrum from inflation/bouncing Universe in VCDM, Int. J. Mod. Phys. A 40 (2025) 2540007 [2407.02882].arxiv.org/abs/2407.02882
  23. [23]A.E. Gümrükçüoğlu, S. Mukohyama and T.P. Sotiriou, Low energy ghosts and the Jeans’ instability, Phys. Rev. D 94 (2016) 064001 [1606.00618].DOI
  24. [24]Y.A. Ageeva, O.A. Evseev, O.I. Melichev and V.A. Rubakov, Horndeski Genesis: strong coupling and absence thereof, EPJ Web Conf. 191 (2018) 07010 [1810.00465].arxiv.org/abs/1810.00465
  25. [25]Y. Ageeva, P. Petrov and V. Rubakov, Nonsingular cosmological models with strong gravity in the past, Phys. Rev. D 104 (2021) 063530 [2104.13412].DOI
  26. [26]Y. Cai, J. Xu, S. Zhao and S. Zhou, Perturbative unitarity and NEC violation in genesis cosmology, JHEP 10 (2022) 140 [2207.11772].DOI
  27. [27]S. Akama and S. Hirano, Primordial non-Gaussianity from Galilean Genesis without strong coupling problem, Phys. Rev. D 107 (2023) 063504 [2211.03838].DOI
  28. [28]D.H. Lyth and Y. Rodriguez, Non-Gaussianity from the second-order cosmological perturbation, Phys. Rev. D 71 (2005) 123508 [astro-ph/0502578].DOI

Paper details

vlnc:W_x72qpm3y
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents