betaFeedback

Reading v1View latest version

WorksPhysics and Astronomy

Circuit Complexity of Scalar and Tensor Perturbations in Luminal Einstein–Gauss–Bonnet Inflation

Abstract

We investigate the circuit complexity of primordial scalar and tensor perturbations in the luminal branch of Einstein–Gauss–Bonnet inflation. Starting from the quadratic perturbation actions, we construct the two-mode Hamiltonians and derive the evolution equations for the squeezing amplitudes and angles. Although the tensor speed is fixed to unity, the Gauss–Bonnet (GB) coupling remains dynamical through the scalar and tensor pump fields. We track the squeezing dynamics from the subhorizon to the superhorizon regime and evaluate the Nielsen wave-function complexity of the resulting two-mode squeezed states. The GB correction produces a clear separation in the squeezing amplitudes, while the unwrapped squeezing angles remain nearly phase locked. Consequently, the circuit complexity is controlled mainly by the amplitude evolution. A positive coupling enhances the scalar complexity but slightly suppresses the tensor complexity for the benchmarks considered. These opposite trends show that scalar and tensor Gaussian states retain distinct information about the same modified inflationary background even when tensor propagation is luminal.

Introduction

Circuit complexity quantifies the minimum cost of preparing a target quantum state from a reference state with a prescribed set of gates. Its geometric formulation converts this optimization into the search for a minimal path on a unitary manifold [1, 2]. It has been extended to Gaussian quantum field theories through wave-function and covariance-matrix methods [3, 4, 5, 6]. These approaches have clarified how the result depends on the reference state, gate set, cost function, and penalty factors [7, 8, 9, 10]. Cosmological perturbations provide a natural setting because expansion converts the initial vacuum into a two-mode squeezed Gaussian state [11, 12]. Circuit complexity can then measure the cost of preparing the evolved perturbation state and distinguish histories that lead to similar late-time spectra [13, 14]. The full Gaussian state contains information that is not fixed by the power spectrum alone. The squeezing amplitude rkr_k controls the occupation of correlated mode pairs through nk=sinh⁡2rkn_k=\sinh^2 r_k, while the squeezing angle φk\varphi_k records their phase-space orientation. Previous inflationary analyses show that the dominant state-space separation is governed by the growth of rkr_k, whereas φk\varphi_k supplies the phase information required by phase-sensitive complexity measures [15, 16, 17]. We use this physical interpretation to determine how the EGB background changes the preparation cost of scalar and tensor Gaussian states.

The full Gaussian state contains information that is not fixed by the power spectrum alone. During inflation, the squeezing amplitude controls the occupation of correlated mode pairs, while the squeezing angle records their phase-space orientation. Large superhorizon squeezing can make equal-time correlators resemble a classical stochastic distribution, although the underlying opposite-momentum modes retain quantum correlations [18]. A state-preparation measure is therefore useful because it responds to both the amplification and the phase structure of the perturbations. Related operator-based constructions likewise show that cosmological complexity is sensitive to squeezing and to the time-dependent rotation of the state [19]. This motivates treating complexity as a quantum-information diagnostic that complements, rather than replaces, conventional spectra and correlation functions.

Einstein–Gauss–Bonnet (EGB) inflation couples the inflation to the curvature invariant G=R2−4RμνRμν+RμνρσRμνρσ\mathcal{G}=R^2-4R_{\mu\nu}R^{\mu\nu}+R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}. Because a constant GB coefficient is topological in four dimensions, a scalar-dependent coupling is required to modify the background and perturbations. Such couplings arise in low-energy effective descriptions of gravity and support slow-roll, constant-roll, and graceful-exit inflationary solutions [20, 21, 22, 23, 24, 25]. Their effects on the scalar and tensor spectra, spectral indices, and tensor-to-scalar ratio have been studied analytically and numerically in a broad class of potentials and coupling functions [25, 26, 27]. In particular, the coupling modifies the usual inflationary consistency relation and may generate tensor spectra that are not available in minimally coupled slow-roll models [28]. The same interaction is contained in the general second-order scalar–tensor framework, where the quadratic scalar and tensor actions provide the no-ghost and gradient-stability conditions [29].

At quadratic order, the coupling modifies the scalar kinetic coefficient and sound speed as well as the tensor kinetic coefficient and tensor speed [30, 31, 32]. A viable perturbative background must keep both kinetic coefficients and both squared propagation speeds positive throughout the evolution. Requiring luminal tensor propagation imposes an additional relation between the coupling function and the inflationary background, while still allowing nontrivial scalar and tensor pump fields [33]. We therefore focus on the luminal branch cT2=1c_T^2=1. This condition removes the direct tensor-speed correction but does not restore general relativity. The tensor kinetic coefficient qTq_T and pump field zT′/zTz'_T/z_T remain time dependent. The scalar sector is modified independently through QsQ_s, csc_s, and zs′/zsz'_s/z_s. Consequently, the two sectors provide distinct probes of the same higher-curvature background even when their tensor propagation speed is fixed to unity.

This structure motivates a circuit-complexity analysis. Scalar and tensor modes experience the same GB background but follow different squeezing trajectories. Their power spectra describe fluctuation amplitudes, whereas circuit complexity probes the distance between the initial vacuum and the complete Gaussian target state. The latter retains information from both the squeezing amplitude and squeezing angle. Comparing the two sectors therefore tests whether a luminal modified-gravity background leaves distinct quantum-state preparation costs even when the tensor propagation speed is unchanged. We derive the scalar and tensor two-mode Hamiltonians, solve their squeezing equations, and insert the results into a common geometric F2F_2 wave-function cost. The comparison isolates the role of the two pump fields and provides a state-space diagnostic complementary to conventional perturbation observables.

The paper is organized as follows. Section II presents the EGB quadratic actions, two-mode Hamiltonians, and scalar and tensor squeezing dynamics. Section III introduces the wave-function circuit complexity and applies it to the numerical squeezing solutions. Section IV summarizes the results and discusses their scope.

Scalar and tensor squeezing in EGB inflation

We consider a canonical inflation coupled to the GB invariant,

S=∫d4x−g[MPl22R−12(∂ϕ)2−V(ϕ)−ξ(ϕ)G],S=\int d^4x\sqrt{-g}\left[\frac{M_{\mathrm{Pl}}^2}{2}R-\frac{1}{2}(\partial\phi)^2-V(\phi)-\xi(\phi)\mathcal{G}\right],
G=R2−4RμνRμν+RμνρσRμνρσ.(1)\mathcal{G}=R^2-4R_{\mu\nu}R^{\mu\nu}+R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}. \tag*{(1)}

The background equations and the perturbation constraints have been derived in Refs. [22, 23, 32]. We use their final perturbation coefficients. Our coupling convention is −ξG-\xi\mathcal{G}. For the convention −ξ~G/2-\tilde{\xi}\mathcal{G}/2, one must set ξ~=2ξ\tilde{\xi}=2\xi.

Quadratic actions and two-mode Hamiltonians

The scalar quadratic action is

Ss(2)=12∫dt d3x a3Qs[R˙2−cs2a2(∂iR)2].(2)S_s^{(2)}=\frac{1}{2}\int dt\,d^3x\,a^3Q_s\left[\dot{\mathcal{R}}^2-\frac{c_s^2}{a^2}(\partial_i\mathcal{R})^2\right]. \tag*{(2)}

This result follows from the ADM action after the lapse and shift constraints are eliminated [30, 31, 32]. Define Δ=8Hξ˙/(MPl2−8Hξ˙)\Delta=8H\dot{\xi}/(M_{\mathrm{Pl}}^2-8H\dot{\xi}). In our convention, the two coefficients are

Qs=ϕ˙2+12ΔH3ξ˙H2(1−Δ/2)2,(3)Q_s=\frac{\dot{\phi}^2+12\Delta H^3\dot{\xi}}{H^2(1-\Delta/2)^2}, \tag*{(3)}
cs2=1+16ΔHξ˙H˙+4Δ2H2(ξ¨−Hξ˙)ϕ˙2+12ΔH3ξ˙.(4)c_s^2=1+\frac{16\Delta H\dot{\xi}\dot{H}+4\Delta^2H^2(\ddot{\xi}-H\dot{\xi})}{\dot{\phi}^2+12\Delta H^3\dot{\xi}}. \tag*{(4)}

The quantities QsQ_s and csc_s are fixed by the background solution H(t)H(t) and ϕ(t)\phi(t), with ξ˙=ξ,ϕϕ˙\dot{\xi}=\xi_{,\phi}\dot{\phi} and ξ¨=ξ,ϕϕϕ˙2+ξ,ϕϕ¨\ddot{\xi}=\xi_{,\phi\phi}\dot{\phi}^2+\xi_{,\phi}\ddot{\phi}. Thus, csc_s is evaluated from the background rather than prescribed independently. For the tensor perturbation, we use hij=a2(eγ)ijh_{ij}=a^2(e^\gamma)_{ij} with ∂iγij=0\partial_i\gamma_{ij}=0 and γii=0\gamma_{ii}=0. The known quadratic action is

ST(2)=18∫dt d3x a3qT[γ˙ij2−cT2a2(∂kγij)2].(5)S_T^{(2)}=\frac{1}{8}\int dt\,d^3x\,a^3q_T\left[\dot{\gamma}_{ij}^2-\frac{c_T^2}{a^2}(\partial_k\gamma_{ij})^2\right]. \tag*{(5)}

where

qT=MPl2−8Hξ˙,cT2=MPl2−8ξ¨MPl2−8Hξ˙.(6)q_T=M_{\mathrm{Pl}}^2-8H\dot{\xi},\qquad c_T^2=\frac{M_{\mathrm{Pl}}^2-8\ddot{\xi}}{M_{\mathrm{Pl}}^2-8H\dot{\xi}}. \tag*{(6)}

Both coefficients are fixed by the same inflationary background. To separate the tensor pump-field correction from a modified tensor propagation speed, we restrict the numerical analysis to the luminal tensor-speed branch

ξ¨=Hξ˙,cT2=1.(7)\ddot{\xi}=H\dot{\xi},\qquad c_T^2=1. \tag*{(7)}

This is a theoretical restriction on the inflationary background rather than an inference from a late-time observation. It is compatible with the gravitational-wave speed bound associated with GW170817, but no late-time evolution or observational analysis is assumed here. The GB correction remains nontrivial because qTq_T can vary with time and modify the tensor pump field. These coefficients are standard results of GB perturbation theory [22, 23, 32]. The conditions Qs>0Q_s>0, cs2>0c_s^2>0, qT>0q_T>0, and cT2>0c_T^2>0 exclude ghost and short-scale gradient instabilities.

For later use, define

δGB=8Hξ˙MPl2,Δ=δGB1−δGB,ϵ=−H˙H2.(8)\delta_{\mathrm{GB}}=\frac{8H\dot{\xi}}{M_{\mathrm{Pl}}^2},\qquad\Delta=\frac{\delta_{\mathrm{GB}}}{1-\delta_{\mathrm{GB}}},\qquad\epsilon=-\frac{\dot{H}}{H^2}. \tag*{(8)}

A star denotes evaluation at the reference-mode horizon crossing. Thus, δGB,∗\delta_{\mathrm{GB},*} specifies the GB strength when the reference mode crosses the horizon. Using the background equations on the luminal branch, the perturbation coefficients may be written as

Qs=MPl22ϵ(1−δGB)+3δGB2/[2(1−δGB)](1−Δ/2)2,(9)Q_s=M_{\mathrm{Pl}}^2\frac{2\epsilon(1-\delta_{\mathrm{GB}})+3\delta_{\mathrm{GB}}^2/[2(1-\delta_{\mathrm{GB}})]}{(1-\Delta/2)^2}, \tag*{(9)}
cs2=1−2ϵδGB2/(1−δGB)2ϵ(1−δGB)+3δGB2/[2(1−δGB)],(10)c_s^2=1-\frac{2\epsilon\delta_{\mathrm{GB}}^2/(1-\delta_{\mathrm{GB}})}{2\epsilon(1-\delta_{\mathrm{GB}})+3\delta_{\mathrm{GB}}^2/[2(1-\delta_{\mathrm{GB}})]}, \tag*{(10)}
qT=MPl2(1−δGB),cT2=1.(11)q_T=M_{\mathrm{Pl}}^2(1-\delta_{\mathrm{GB}}),\qquad c_T^2=1. \tag*{(11)}

We now use conformal time dη=dt/ad\eta= dt/a. For the scalar mode, define vs=zsRv_s = z_s\mathcal{R}, zs=aQsz_s = a\sqrt{Q_s}, and hs=zs′/zsh_s = z_s'/z_s. For each tensor polarization, define vT=zTγλv_T = z_T\gamma_\lambda, zT=aqT/2z_T = a\sqrt{q_T/2}, and hT=zT′/zTh_T = z_T'/z_T. Both sectors can then be written as

SA(2)=12∫dη d3x[(vA′−hAvA)2−cA2(∂ivA)2].(12)S_A^{(2)} = \frac{1}{2}\int d\eta\,d^3x\left[(v_A'-h_Av_A)^2-c_A^2(\partial_i v_A)^2\right]. \tag*{(12)}

The index A=s,TA=s,T labels the two sectors and does not introduce an additional propagation speed. Specifically, (vA,zA,hA,cA)(v_A,z_A,h_A,c_A) denotes (vs,zs,zs′/zs,cs)(v_s,z_s,z_s'/z_s,c_s) for scalar perturbations and (vT,zT,zT′/zT,cT)(v_T,z_T,z_T'/z_T,c_T) for tensor perturbations. Thus, the numerical evolution uses the distinct pairs (cs,zs′/zs)(c_s,z_s'/z_s) and (cT,zT′/zT)(c_T,z_T'/z_T). The conjugate momentum is πA=vA′−hAvA\pi_A=v_A'-h_Av_A, and the Hamiltonian is

HA=12∫d3x[πA2+cA2(∂ivA)2+hA(πAvA+vAπA)].(13)H_A=\frac{1}{2}\int d^3x\left[\pi_A^2+c_A^2(\partial_i v_A)^2+h_A(\pi_Av_A+v_A\pi_A)\right]. \tag*{(13)}

We use a fixed reference frequency kk. This is the convention adopted for the squeezed-state construction in [16]. The modes kk and −k-k form one two-mode sector. Its Hamiltonian is

H^A,k=ΩA,k(c^A,k†c^A,k+c^A,k†c^A,−k+1)+ΛA,kc^A,k†c^A,−k†+ΛA,k∗c^A,kc^A,−k,(14)\hat{H}_{A,k}=\Omega_{A,k}\left(\hat{c}_{A,k}^{\dagger}\hat{c}_{A,k}+\hat{c}_{A,k}^{\dagger}\hat{c}_{A,-k}+1\right)+\Lambda_{A,k}\hat{c}_{A,k}^{\dagger}\hat{c}_{A,-k}^{\dagger}+\Lambda_{A,k}^{*}\hat{c}_{A,k}\hat{c}_{A,-k}, \tag*{(14)}
ΩA,k=k2(1+cA2),ΛA,k=k2(cA2−1)+ihA.(15)\Omega_{A,k}=\frac{k}{2}(1+c_A^2),\qquad\Lambda_{A,k}=\frac{k}{2}(c_A^2-1)+ih_A. \tag*{(15)}

The coefficient ΛA,k\Lambda_{A,k} describes pair production caused by a nonunit propagation speed, whereas hA=zA′/zAh_A=z_A'/z_A describes parametric amplification by the evolving background. The propagation speed changes the real part of ΛA,k\Lambda_{A,k}. The pump field changes its imaginary part.

The scalar coefficients follow from cA=csc_A=c_s and hA=zs′/zsh_A=z_s'/z_s. The tensor coefficients follow from cA=cTc_A=c_T and hA=zT′/zTh_A=z_T'/z_T. Even when cT2=1c_T^2=1, the tensor Hamiltonian need not reduce to its general-relativistic form. In that case, qTq_T can still vary and hTh_T can differ from a′/aa'/a. The scalar pump field zs=aQsz_s=a\sqrt{Q_s} is the GB generalization of the standard scalar pump field zz. In the general-relativistic limit, Qs→2MPl2ϵQ_s\to2M_{\rm Pl}^2\epsilon and zs→aMPl2ϵ=zz_s\to aM_{\rm Pl}\sqrt{2\epsilon}=z. By contrast, zT=aqT/2z_T=a\sqrt{q_T/2} is controlled by the tensor kinetic coefficient. Hence, zs′/zsz_s'/z_s and zT′/zTz_T'/z_T generally generate different squeezing dynamics even on the luminal branch cT2=1c_T^2=1. In particular, using dδGB/dN=(1−ϵ)δGBd\delta_{\rm GB}/dN=(1-\epsilon)\delta_{\rm GB} on this branch gives

zs′zs=aH(1+12dln⁡QsdN),(16)\frac{z_s'}{z_s}=aH\left(1+\frac{1}{2}\frac{d\ln Q_s}{dN}\right), \tag*{(16)}
zT′zT=aH[1−(1−ϵ)δGB2(1−δGB)].(17)\frac{z_T'}{z_T}=aH\left[1-\frac{(1-\epsilon)\delta_{\rm GB}}{2(1-\delta_{\rm GB})}\right]. \tag*{(17)}

The tensor squeezing can therefore differ from the general-relativistic result through the pump field, without invoking a nonluminal tensor speed.

Two-mode squeezed states and squeezing parameters

The pair-creation terms in Eq. (14) generate a two-mode squeezed state. For each sector, the squeeze operator is

S^A,k(rA,k,φA,k)=exp⁡{rA,k[e−2iφA,kc^A,kc^A,−k−e2iφA,kc^A,−k†c^A,k†]}.(18)\hat{S}_{A,k}(r_{A,k},\varphi_{A,k})=\exp\left\{r_{A,k}\left[e^{-2i\varphi_{A,k}}\hat{c}_{A,k}\hat{c}_{A,-k}-e^{2i\varphi_{A,k}}\hat{c}_{A,-k}^{\dagger}\hat{c}_{A,k}^{\dagger}\right]\right\}. \tag*{(18)}

Acting on the two-mode vacuum gives

∣ψA(η)⟩sq=1cosh⁡rA,k∑n=0∞[−e2iφA,ktanh⁡rA,k]n∣n;n⟩A,k,−k.(19)|\psi_A(\eta)\rangle_{\rm sq}=\frac{1}{\cosh r_{A,k}}\sum_{n=0}^{\infty}\left[-e^{2i\varphi_{A,k}}\tanh r_{A,k}\right]^n|n;n\rangle_{A,k,-k}. \tag*{(19)}

The rotation operator only contributes an overall phase when it acts on the initial vacuum. It is omitted here. This convention follows [16].

Define the complex pair amplitude λA,k=−e2iφA,ktanh⁡rA,k\lambda_{A,k}=-e^{2i\varphi_{A,k}}\tanh r_{A,k}. Substitution of Eq. (19) into the Schrödinger equation gives the Riccati equation

iλA,k′=ΛA,k+2ΩA,kλA,k+ΛA,k∗λA,k2.(20)i\lambda_{A,k}'=\Lambda_{A,k}+2\Omega_{A,k}\lambda_{A,k}+\Lambda_{A,k}^{*}\lambda_{A,k}^{2}. \tag*{(20)}

This equation is regular at the exact vacuum point λA,k=0\lambda_{A,k}=0 and provides an equivalent nonsingular representation of the squeezing dynamics. The real squeezing variables are recovered through rA,k=arctanh⁡∣λA,k∣r_{A,k}=\operatorname{arctanh}|\lambda_{A,k}| and φA,k=arg⁡(−λA,k)/2\varphi_{A,k}=\arg(-\lambda_{A,k})/2.

It is useful to introduce the real pair-production coefficient AA,k=k(cA2−1)/2\mathcal{A}_{A,k}=k(c_A^2-1)/2 and the rotation frequency ΩA,k=k(1+cA2)/2\Omega_{A,k}=k(1+c_A^2)/2. Using the definition of λA,k\lambda_{A,k}, Eq. (20) then gives the two coupled squeezing equations

rA,k′=AA,ksin⁡(2φA,k)−hAcos⁡(2φA,k),(21)r_{A,k}'=\mathcal{A}_{A,k}\sin(2\varphi_{A,k})-h_A\cos(2\varphi_{A,k}), \tag*{(21)}
φA,k′=−ΩA,k+[AA,kcos⁡(2φA,k)+hAsin⁡(2φA,k)]coth⁡(2rA,k).(22)\varphi_{A,k}'=-\Omega_{A,k}+\left[\mathcal{A}_{A,k}\cos(2\varphi_{A,k})+h_A\sin(2\varphi_{A,k})\right]\coth(2r_{A,k}). \tag*{(22)}

For scalar perturbations, (cA,hA,AA,k,φA,k)=(cs,zs′/zs,rs,k,φs,k)(c_A,h_A,\mathcal{A}_{A,k},\varphi_{A,k})=(c_s,z_s'/z_s,r_{s,k},\varphi_{s,k}). For each tensor polarization, they are (cT,zT′/zT,rT,k,φT,k)(c_T,z_T'/z_T,r_{T,k},\varphi_{T,k}). These substitutions contain the complete scalar and tensor systems.

We set MPl=H∗=1M_{\rm Pl}=H_*=1, ϵ∗=0.002\epsilon_*=0.002, and k=0.01k=0.01, and use δGB,∗=0\delta_{\rm GB,*}=0, δGB,0=0.025\delta_{\rm GB,0}=0.025, and 0.0050.005. The first value is the general-relativistic reference. All backgrounds satisfy Qs>0Q_s>0, qT>0q_T>0, cs2>0c_s^2>0, and cT2>0c_T^2>0. The variable x=−ln⁡(−kη)x=-\ln(-k\eta) runs from the subhorizon regime (x<0)(x<0) through horizon crossing (x≃0)(x\simeq0) to the superhorizon regime (x>0)(x>0); for scalar modes the more precise crossing condition is csk≃aHc_sk\simeq aH. We use x=−log⁡(−kη)x=-\log(-k\eta) as the evolution variable. Negative xx corresponds to the subhorizon regime, x≃0x \simeq0 approximately marks horizon crossing, and positive xx describes superhorizon evolution. For scalar perturbations, the more precise crossing condition is csk≃aHc_{\mathrm{s}} k \simeq aH.

Eq. (21) separates the two sources of squeezing in the EGB background. The term AA,k\mathcal{A}_{A,k} is generated by a nonunit propagation speed, while hA=zA′/zAh_A = z'_A/z_A describes parametric amplification by the evolving pump field. Both terms contribute in the scalar sector. On the luminal tensor branch, AT,k=0\mathcal{A}_{T,k}=0, so the tensor correction is carried entirely by hT=zT′/zTh_T = z'_T/z_T. In the scalar sector, both contributions can be present. In the luminal tensor sector, AT,k=0\mathcal{A}_{T,k}=0, so the GB correction enters entirely through the tensor pump field hT=zT′/zTh_T = z'_T/z_T.

Figure 1 shows the scalar and tensor squeezing amplitudes. In the upper panel, increasing δGB,∗\delta_{\mathrm{GB},*} changes both csc_{\mathrm{s}} and hs=zs′/zsh_{\mathrm{s}} = z'_{\mathrm{s}}/z_{\mathrm{s}}. The scalar curves therefore separate progressively after horizon exit, when the background pump overcomes the rapid subhorizon oscillation. A larger scalar amplitude means a larger mean pair number, ns,k=sinh⁡2rs,kn_{\mathrm{s},k} = \sinh^2 r_{\mathrm{s},k}, and hence stronger scalar-mode squeezing. In the lower panel, the luminal condition gives cT2=1c_{\mathrm{T}}^2 = 1 and thus AT,k=0\mathcal{A}_{\mathrm{T},k}=0. The tensor curves are separated only by hT=zT′/zTh_{\mathrm{T}} = z'_{\mathrm{T}}/z_{\mathrm{T}}, through qT=MPl2(1−δGB)q_{\mathrm{T}} = M_{\mathrm{Pl}}^2(1-\delta_{\mathrm{GB}}). The opposite or weaker ordering of the tensor curves is therefore a consequence of the different pump field, rather than a correction to the tensor propagation speed. The two panels demonstrate that luminal tensor propa-

Figure 1. Evolution of the scalar squeezing amplitude rs,kr_{\mathrm{s},k} (upper panel) and tensor squeezing amplitude rT,kr_{\mathrm{T},k} (lower panel) for different values of δGB,∗\delta_{\mathrm{GB},*}.

gation does not eliminate the GB modification of tensor squeezing. The squeezing amplitude also determines the mean occupation of correlated mode pairs through nA,k=sinh⁡2rA,kn_{A,k} = \sinh^2 r_{A,k}. The ordering in Figure 1 therefore shows that a positive GB coupling enhances scalar pair production but slightly suppresses tensor pair production for the backgrounds considered here. This opposite response originates from the different scalar and tensor pump fields.

The angle equation, Eq. (22), contains a large common rotation. Deep inside the horizon, cA2≃1c_A^2 \simeq1 and ∣hA∣≪k\lvert h_A\rvert\ll k, so AA,k≃0\mathcal{A}_{A,k} \simeq0, ΩA,k≃k\Omega_{A,k} \simeq k, and φA,k≃−k\varphi_{A,k} \simeq-k. After horizon exit, rA,kr_{A,k} grows and coth⁡(2rA,k)→1\coth(2r_{A,k}) \to1. The pump term then drives the angle toward the fixed orientations determined approximately by sin⁡(2φA,k)≃0\sin(2\varphi_{A,k}) \simeq0.

Figure 2 shows the unwrapped squeezing angles. The nearly coincident curves in the upper panel arise because the scalar angle is dominated by the common rotation −Ωs,k-\Omega_{\mathrm{s},k}; the direct correction from As,k\mathcal{A}_{\mathrm{s},k} remains small because the stable benchmark solutions keep cs2c_{\mathrm{s}}^2 close to unity. In the lower panel, cT2=1c_{\mathrm{T}}^2 = 1 makes AT,k\mathcal{A}_{\mathrm{T},k} vanish exactly, so the tensor angle receives no direct speed-dependent pair-production term. The residual GB dependence enters through hTh_{\mathrm{T}}, but it mainly changes the rate at which the solution approaches the superhorizon orientation. It has little effect on that final orientation. The overlap of the curves therefore reflects a common subhorizon phase rotation followed by superhorizon phase locking. It does not imply that the GB correction is absent, as the amplitude evolution in Figure 1 remains clearly coupling dependent. The squeeze angle is periodic, φA,k∼φA,k+nπ\varphi_{A,k} \sim\varphi_{A,k} + n\pi, because the state depends on e2iφA,ke^{2i\varphi_{A,k}}. The small coupling-dependent phase shift is extracted with the principal angular residual

Figure 2. Evolution of the unwrapped scalar squeezing angle φs,k\varphi_{\mathrm{s},k} (upper panel) and tensor squeezing angle φT,k\varphi_{\mathrm{T},k} (lower panel) for different values of δGB,∗\delta_{\mathrm{GB},*}.

ΔφA,k=12Arg⁡{exp⁡[2i(φA,k−φA,kGR)]}.(23)\Delta\varphi_{A,k} = \frac{1}{2}\operatorname{Arg}\left\{\exp\left[2i\left(\varphi_{A,k} - \varphi^{\mathrm{GR}}_{A,k}\right)\right]\right\}. \tag*{(23)}

Figure 3 displays the phase information hidden by the large common rotation in Figure 2. The upper panel measures the scalar phase shift produced jointly by the sound-speed correction and the modified scalar pump field. The lower panel isolates the tensor phase shift generated by hTh_T, since cT2=1c_T^2 = 1. The residuals are small compared with the accumulated unwrapped rotation, although they remain finite on the principal angular interval. The GB correction therefore produces a genuine phase shift without qualitatively changing the phase-locking behavior. The nearly coincident curves show that the common phase rotation dominates over the GB-dependent correction. The GB coupling therefore changes the amount of squeezing more efficiently than the final phase-space orientation. The finite residual nevertheless confirms that the scalar and tensor Gaussian states retain a small coupling-dependent phase correction. In the present stable EGB background, the dominant information is instead carried by the squeezing amplitude. Nevertheless, a nonzero ΔφA,k\Delta\varphi_{A,k} represents a genuine coupling-dependent change of the Gaussian state. It

Figure 3. Principal squeezing-angle residual relative to general relativity in the scalar sector, Δφs,k\Delta\varphi_{s,k} (upper panel), and tensor sector, ΔφT,k\Delta\varphi_{T,k} (lower panel).

must therefore be retained in phase-sensitive quantities, including the circuit complexity considered in subsequent work. The present construction is particularly suited to circuit complexity. The reference state and the target two-mode Gaussian state are already specified, and the latter is completely characterized by rA,kr_{A,k} and φA,k\varphi_{A,k}. The coefficient AA,kA_{A,k} describes pair production caused by a nonunit propagation speed, whereas hA=zA′/zAh_A = z'_A/z_A describes parametric amplification by the evolving background.

Circuit complexity of the squeezed perturbations

Circuit complexity measures the shortest allowed circuit that prepares a target state from a chosen reference state. Its geometric formulation represents the circuit by a path on the manifold of unitary transformations and minimizes a prescribed cost along that path [1, 2]. This construction has been extended to Gaussian states and quantum field theory using wave functions, covariance matrices, and continuous gate sets [3, 4, 5, 6]. Related studies have examined the role of the reference state, penalty factors, gate scale, and alternative Gaussian-state geometries [7, 8, 9, 10]. Circuit complexity has also been applied to squeezed cosmological perturbations, where the Bunch–Davies vacuum provides a natural reference and the evolved two-mode squeezed state provides the target [13, 16].

For each sector A=s,TA = s,T, we take the unsqueezed vacuum at the initial time as the reference state and the state in Eq. (19) as the target. We adopt the unpenalized Nielsen F2F_2 cost evaluated from the Gaussian wave function. In the conventions of the present squeeze operator, the mode complexity is [13]

C2,A(k)=12{[ln⁡∣1+e−2iφA,ktanh⁡rA,k1−e−2iφA,ktanh⁡rA,k∣]2+[tan⁡−1(2sin⁡(2φA,k)sinh⁡rA,kcosh⁡rA,k)]2}1/2.(24)\mathcal{C}_{2,A}(k) = \frac{1}{\sqrt{2}}\left\{\left[\ln\left|\frac{1+e^{-2i\varphi_{A,k}}\tanh r_{A,k}}{1-e^{-2i\varphi_{A,k}}\tanh r_{A,k}}\right|\right]^2 + \left[\tan^{-1}\left(2\sin(2\varphi_{A,k})\sinh r_{A,k}\cosh r_{A,k}\right)\right]^2\right\}^{1/2}. \tag*{(24)}

This expression keeps both pieces of information carried by the squeezed state. The logarithmic term is mainly controlled by the squeezing amplitude, while the bounded angular term records the orientation of the squeezed ellipse in phase space. In the superhorizon phase-locked regime, where sin⁡(2φA,k)≃0\sin(2\varphi_{A,k}) \simeq0, Eq. (24) reduces approximately to C2,A(k)≃2 rA,k\mathcal{C}_{2,A}(k) \simeq\sqrt{2}\,r_{A,k}. The late-time ordering of the circuit-complexity curves is therefore expected to follow the ordering of the squeezing amplitudes. The finite angular residual changes the detailed curve but does not reverse this ordering in the present parameter range. The formula is convention dependent: changing the gate set, cost function, reference frequency, or covariance-matrix prescription changes the numerical normalization and may remove the explicit angle dependence. Eq. (24) is therefore used consistently for both scalar and tensor sectors.

The numerical curves are obtained by inserting the solutions {rs,k,φs,k}\{r_{s,k},\varphi_{s,k}\} and {rT,k,φT,k}\{r_{T,k},\varphi_{T,k}\} from Eqs. (21) and (22) into Eq. (24). The same reference state, cost function, comoving mode, and GB benchmarks are used in both sectors. This common prescription makes the scalar–tensor comparison sensitive only to their different propagation and pump coefficients. Since the unwrapped angles in Figure 2 are nearly phase locked, the dominant separation is expected to follow the squeezing amplitudes, while the angular term provides a smaller bounded correction.

Figure 4 presents the final circuit-complexity results. In the upper panel, the scalar complexity remains close to zero deep inside the horizon because the state is close to the unsqueezed reference vacuum. Small oscillations appear as the pump term begins to compete with the mode frequency. After horizon exit, rs,kr_{s,k} grows and the complexity increases rapidly. Positive δGB,∗\delta_{\mathrm{GB},*} raises the scalar pumping for the chosen backgrounds and produces a larger scalar squeezing amplitude. The ordering C2,s(0.005)>C2,s(0.0025)>C2,s(0)\mathcal{C}_{2,s}(0.005)>\mathcal{C}_{2,s}(0.0025)>\mathcal{C}_{2,s}(0) at late times therefore follows the amplitude ordering in Figure 1. The squeezing-angle term changes the detailed shape but remains subleading because the scalar angles are nearly locked.

Figure 4. Evolution of the Nielsen F2F_2 circuit complexity for scalar perturbations (upper panel) and tensor perturbations (lower panel). The horizontal axis −log⁡(−kη)-\log(-k\eta) runs from the subhorizon regime on the left to the superhorizon regime on the right. The dashed blue curves denote the general-relativistic limit δGB,∗=0\delta_{\mathrm{GB},*}=0, while the orange and magenta curves correspond to δGB,∗=0.0025\delta_{\mathrm{GB},*}=0.0025 and 0.0050.005, respectively. The scalar complexity is enhanced by increasing positive GB coupling, whereas the tensor complexity is slightly suppressed.

The lower panel shows the tensor complexity. All three curves again remain small in the subhorizon regime and grow after horizon exit. Their separation is smaller than in the scalar sector. On the luminal branch, cT2=1c_T^2=1 implies AT,k=0\mathcal{A}_{T,k}=0, so the GB dependence enters only through qTq_T and hT=zT′/zTh_T=z'_T/z_T. For the positive benchmarks used here, this pump-field correction slightly reduces rT,kr_{T,k} and hence lowers the late-time complexity, giving C2,T(0.005)<C2,T(0.0025)<C2,T(0)\mathcal{C}_{2,T}(0.005)<\mathcal{C}_{2,T}(0.0025)<\mathcal{C}_{2,T}(0). The opposite scalar and tensor orderings are dynamical: they result from the different background dependence of zsz_s and zTz_T, rather than from a nonluminal tensor speed. The same GB coupling can therefore make the scalar Gaussian state more costly and the tensor Gaussian state less costly to prepare relative to their corresponding vacua. The comparatively smooth superhorizon growth of C2,A\mathcal{C}_{2,A} reflects the cumulative action of the scalar and tensor pump fields. In the scalar sector, the GB coupling modifies both csc_s and hs=zs′/zsh_s=z'_s/z_s. In the luminal tensor sector, cT2=1c_T^2=1 removes the direct speed-dependent contribution, while the time dependence of qTq_T continues to modify hT=zT′/zTh_T=z'_T/z_T. The opposite scalar and tensor orderings therefore arise from their different pump-field responses to the same EGB background.

The GB dependence of the circuit complexity originates from two coupled sources: the propagation coefficient cAc_A and the pump field hA=zA′/zAh_A=z'_A/z_A. These effects are not dynamically independent, because both enter the coupled evolution equations for rA,kr_{A,k} and φA,k\varphi_{A,k}. Nevertheless, the luminal tensor sector isolates the pump-field channel: cT2=1c_T^2=1 gives AT,k=0\mathcal{A}_{T,k}=0, while the time-dependent qTq_T keeps hTh_T nontrivial. For scalar perturbations, both contributions are present because the GB coupling modifies csc_s and QsQ_s. For tensor perturbations, cT2=1c_T^2=1 gives ΔC2,T(prop)=0\Delta\mathcal{C}_{2,T}^{(\mathrm{prop})}=0, while the nonzero ΔC2,T(pump)\Delta\mathcal{C}_{2,T}^{(\mathrm{pump})} is generated by the time-dependent coefficient qTq_T. Hence, luminal tensor propagation does not imply general-relativistic tensor-state evolution. The quantity in Eq. (24) is a state-space diagnostic and is not itself a cosmological observable. Its physical value here lies in comparing Gaussian-state evolution under a fixed circuit prescription. A direct relation to power spectra, non-Gaussianity, or late-time observables would require a separate analysis. A Lyapunov exponent would require an independent analysis of the sensitivity to nearby initial conditions or of the asymptotic Lyapunov coefficients. We therefore interpret the present result as a pump-field-induced separation of Gaussian state-preparation costs, rather than as evidence for enhanced quantum chaos.

Conclusions and outlook

We studied scalar and tensor two-mode squeezed states in the luminal branch of EGB inflation and used them to evaluate circuit complexity. The quadratic actions lead to different scalar and tensor kinetic coefficients, propagation speeds, and pump fields. Imposing cT2=1c_T^2 = 1 removes the direct tensor-speed correction, but the time-dependent tensor coefficient qTq_T continues to modify zT/zTz_T/z_T. Luminal propagation therefore does not make the tensor quantum dynamics identical to general relativity.

The GB coupling affects the squeezing amplitudes more strongly than the squeezing angles. Deep inside the horizon, both sectors share a large approximately free phase rotation. After horizon exit, the pump terms amplify the states and drive their angles toward nearly common phase-locked orientations. The principal angular residuals remain nonzero but small. The final wave-function complexity is therefore governed mainly by rA,kr_{A,k}, with a smaller phase-sensitive correction from φA,k\varphi_{A,k}.

For the positive benchmark couplings considered, increasing δGB,∗\delta_{\mathrm{GB},*} enhances the scalar squeezing amplitude and circuit complexity but slightly suppresses their tensor counterparts. The opposite ordering originates from the different functions QsQ_s and qTq_T and from their associated pump fields. Unlike a resonant sound speed or a modified dispersion relation, the dominant EGB correction in the present parameter range is not a strong direct modification of the mode frequency. It is instead a cumulative pump-field effect. Since the closed-state Krylov complexity satisfies KA,k=sinh⁡2rA,kK_{A,k} = \sinh^2 r_{A,k}, the same results imply enhanced scalar Krylov spreading and suppressed tensor Krylov spreading. Circuit and Krylov complexity therefore give the same parameter ordering here, while only the circuit prescription retains the finite squeezing angle correction.

The numerical values depend on the background solution, reference state, gate set, and F2F_2 cost. Circuit complexity should therefore be interpreted as a comparative state-space diagnostic rather than as a direct observable. The main result is the separation between direct propagation effects and pump-field effects, together with the opposite scalar and tensor responses to the same EGB coupling. Future work may test whether the scalar–tensor complexity splitting persists in other stable EGB backgrounds and may extend the analysis to open-system Krylov dynamics, decoherence, and conventional inflationary correlators.

Acknowledgments

This work is supported by the Guizhou Provincial Natural Science Foundation (Grant No. 2025JYJYJ178), the Natural Science Foundation of Guizhou Minzu University (Grant No. GZMUBS7), the National Natural Science Foundation of China (Grant Nos. 12665009 and 12265010), the Hunan Provincial Department of Education Project (Grant No. 25B0480), and the Guizhou Provincial Department of Science and Technology (Grant Nos. MS[2025]219 and CXTD[2025]030).

References

  1. [1]M. A. Nielsen, A geometric approach to quantum circuit lower bounds, Quant. Inf. Comput. 6, 213 (2006), arXiv:quant-ph/0502070.DOI
  2. [2]M. R. Dowling and M. A. Nielsen, The geometry of quantum computation, Quant. Inf. Comput. 8, 0861 (2008), arXiv:quant-ph/0701004.arxiv.org/abs/quant-ph/0701004
  3. [3]R. Jefferson and R. C. Myers, Circuit complexity in quantum field theory, JHEP 10, 107, arXiv:1707.08570 [hep-th].DOI
  4. [4]S. Chapman, M. P. Heller, H. Marrochio, and F. Pastawski, Toward a Definition of Complexity for Quantum Field Theory States, Phys. Rev. Lett. 120, 121602 (2018), arXiv:1707.08582 [hep-th].
  5. [5]S. Chapman, J. Eisert, L. Hackl, M. P. Heller, R. Jefferson, H. Marrochio, and R. C. Myers, Complexity and entanglement for thermofield double states, SciPost Phys. 6, 034 (2019), arXiv:1810.05151 [hep-th].DOI
  6. [6]L. Hackl and R. C. Myers, Circuit complexity for free fermions, JHEP 07, 139, arXiv:1803.10638 [hep-th].DOI
  7. [7]A. Bhattacharyya, A. Shekar, and A. Sinha, Circuit complexity in interacting QFTs and RG flows, JHEP 10, 140, arXiv:1808.03105 [hep-th].DOI
  8. [8]D. W. F. Alves and G. Camilo, Evolution of complexity following a quantum quench in free field theory, JHEP 06, 029, arXiv:1804.00107 [hep-th].DOI
  9. [9]H. A. Camargo, P. Caputa, D. Das, M. P. Heller, and R. Jefferson, Complexity as a novel probe of quantum quenches: universal scalings and purifications, Phys. Rev. Lett. 122, 081601 (2019), arXiv:1807.07075 [hep-th].DOI
  10. [10]T. Ali, A. Bhattacharyya, S. Shajidul Haque, E. H. Kim, and N. Moynihan, Time Evolution of Complexity: A Critique of Three Methods, JHEP 04, 087, arXiv:1810.02734 [hep-th].DOI
  11. [11]L. P. Grishchuk and Y. V. Sidorov, Squeezed quantum states of relic gravitons and primordial density fluctuations, Phys. Rev. D 42, 3413 (1990).pubmed.ncbi.nlm.nih.gov/10012742
  12. [12]A. Albrecht, P. Ferreira, M. Joyce, and T. Prokopec, Inflation and squeezed quantum states, Phys. Rev. D 50, 4807 (1994), arXiv:astro-ph/9303001.DOI
  13. [13]A. Bhattacharyya, W. Chemissany, S. S. Haque, J. Murugan, and B. Yan, The Multi-faceted Inverted Harmonic Oscillator: Chaos and Complexity, SciPost Phys. Core 4, 002 (2021), arXiv:2007.01232 [hep-th].DOI
  14. [14]J.-L. Lehners and J. Quintin, Quantum Circuit Complexity of Primordial Perturbations, Phys. Rev. D 103, 063527 (2021), arXiv:2012.04911 [hep-th].DOI
  15. [15]L.-H. Liu and A.-C. Li, Complexity of non-trivial sound speed in inflation, Phys. Dark Univ. 37, 101123 (2022), arXiv:2102 12014 [gr-qc].DOI
  16. [16]T. Li and L.-H. Liu, Cosmological complexity of the modified dispersion relation, Phys. Lett. B 854, 138728 (2024), arXiv:2309.01595 [gr-qc].DOI
  17. [17]T. Li and L.-H. Liu, Inflationary Krylov complexity, JHEP 04, 123, arXiv:2401.09307 [hep-th].DOI
  18. [18]J. Martin and V. Vennin, Quantum Discord of Cosmic Inflation: Can We Show that CMB Anisotropies are of Quantum-Mechanical Origin?, Phys. Rev. D 93, 023505 (2016), arXiv:1510.04038 [astro-ph.CO].DOI
  19. [19]S. Haque, C. Jana, and B. Underwood, Operator complexity for quantum scalar fields and cosmological perturbations, Phys. Rev. D 106, 063510 (2022), arXiv:2110.08356 [hep-th].arxiv.org/abs/2110.08356
  20. [20]S. Kawai and J. Soda, Evolution of fluctuations during graceful exit in string cosmology, Phys. Lett. B 460, 41 (1999), arXiv:gr-qc/9903017.DOI
  21. [21]M. Satoh and J. Soda, Higher Curvature Corrections to Primordial Fluctuations in Slow-roll Inflation, JCAP 09, 019, arXiv:0806.4594 [astro-ph].DOI
  22. [22]Z.-K. Guo and D. J. Schwarz, Power spectra from an inflaton coupled to the Gauss-Bonnet term, Phys. Rev. D 80, 063523 (2009), arXiv:0907.0427 [hep-th].DOI
  23. [23]Z.-K. Guo and D. J. Schwarz, Slow-roll inflation with a Gauss-Bonnet correction, Phys. Rev. D 81, 123520 (2010), arXiv:1001.1897 [hep-th].DOI
  24. [24]P. Kanti, R. Gannouji, and N. Dadhich, Gauss-Bonnet Inflation, Phys. Rev. D 92, 041302 (2015), arXiv:1503.01579 [hep-th].DOI
  25. [25]Z. Yi, Y. Gong, and M. Sabir, Inflation with Gauss-Bonnet coupling, Phys. Rev. D 98, 083521 (2018), arXiv:1804.09116 [gr-qc].DOI
  26. [26]S. Koh, B.-H. Lee, W. Lee, and G. Tumurtushaa, Observational constraints on slow-roll inflation coupled to a Gauss-Bonnet term, Phys. Rev. D 90, 063527 (2014), arXiv:1404.6069 [gr-qc].
  27. [27]K. Mudrunka and K. Nakayama, Probing Gauss-Bonnet-corrected inflation with gravitational waves, JCAP 05, 069, arXiv:2312.15766 [astro-ph.CO].
  28. [28]M. Satoh, Slow-roll Inflation with the Gauss-Bonnet and Chern-Simons Corrections, JCAP 11, 024, arXiv:1008.2724 [astro-ph.CO].
  29. [29]T. Kobayashi, M. Yamaguchi, and J. Yokoyama, Generalized G-inflation: Inflation with the most general second-order field equations, Prog. Theor. Phys. 126, 511 (2011), arXiv:1105.5723 [hep-th].DOI
  30. [30]J.-c. Hwang and H. Noh, Classical evolution and quantum generation in generalized gravity theories including string corrections and tachyon: Unified analyses, Phys. Rev. D 71, 063536 (2005), arXiv:gr-qc/0412126.DOI
  31. [31]A. De Felice and S. Tsujikawa, Inflationary non-Gaussianities in the most general second-order scalar-tensor theories, Phys. Rev. D 84, 083504 (2011), arXiv:1107.3917 [gr-qc].DOI
  32. [32]P.-X. Jiang, J.-W. Hu, and Z.-K. Guo, Inflation coupled to a Gauss-Bonnet term, Phys. Rev. D 88, 123508 (2013), arXiv:1310.5579 [hep-th].DOI
  33. [33]S. D. Odintsov, V. K. Oikonomou, and F. P. Fronimos, Rectifying Einstein-Gauss-Bonnet Inflation in View of GW170817, Nucl. Phys. B 958, 115135 (2020), arXiv:2003.13724 [gr-qc].arxiv.org/abs/2003.13724

Paper details

vlnc:W_bfmzmasp
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 2

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