Isaacson energy--momentum tensor of high-frequency gravitational waves in general f(R,G) gravity

arXiv:2504.15709 · gr-qc · Submitted 2025-04-22 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: General Relativity and Quantum Cosmology. The day in general relativity and quantum cosmology.

Kai: We're looking at the paper: "Isaacson energy--momentum tensor of high-frequency gravitational waves in general f(R,G) gravity".

Mira: Here's the gist of the paper. We investigate the high-frequency backreaction of gravitational waves in the Hessian-degenerate sector of four-dimensional f(R, G) gravity defined by f(R, G) = F(R + G/M2).

Kai: The degeneracy of the Hessian reduces the Legendre representation to a single scalar degree of freedom and maps the theory into a scalar–Gauss–Bonnet subclass of Horndeski gravity.

Mira: Model Reduction and Mapping The paper restricts attention to the Hessian-degenerate family f(R, G) = F(X), X = R + G/M2, which admits a one-field Legendre representation rather than the two-field representation appropriate to a nondegenerate Hessian I.3.

Kai: The theory is consequently equivalent to a scalar–Gauss–Bonnet Horndeski theory with one scalar degree of freedom on the cosmological background III.1, and the Gauss–Bonnet coupling in (II.4) is dynamical rather than topological since the coefficient is given by ϕ/M2 II.4>.

Paper summary: Kai: So, we’re looking at this paper today, "Isaacson energy--momentum tensor of high-frequency gravitational waves in general f(R,G) gravity." Basically, they're taking a complicated theory and focusing on a specific part of it where the math simplifies things.

Mira: Exactly. The main point is that they look at the Hessian-degenerate sector of four-dimensional f(R, G) gravity, which is defined by f(R + G/M2) = F(X). This degeneracy means they reduce the theory down to just one scalar degree of freedom and map it onto a scalar–Gauss–Bonnet subclass of Horndeski gravity.

Lev: So the central thesis seems to be that this specific restriction lets them get a concrete result about how gravitational waves backreact on things, even in these modified gravity setups.

Kai: Right. The paper claims they derive the tensor quadratic action and find the modified kinetic and gradient coefficients are G T = M2 Pl phi + 4H phi dot M2 and F T = M2 Pl phi + four phi double dot M2 <ref:2504.15709#pg2,derive the tensor quadratic action and>.

Mira: That's the mapping they use, right. And then they use that to identify a scalar channel that comes from the cubic scalar–tensor–tensor interaction in this specific model.

Lev: The real kicker seems to be that this induced scalar source is generically nonzero and it's not determined solely by the tensor-speed deviation, which is interesting because we usually think those are linked.

Kai: And they show that this source can actually remain nonzero even on a rolling de Sitter background where the tensor speed squared, c squared T, is equal to one <ref:2504.15709#pg1>.

Mira: That’s what they found when they looked at the exact analytic rolling de Sitter solution, and in that case, both c squared T and c squared S are equal to one <ref:2504.15709#pg1>.

Lev: So for someone who only listens to the show, this means even if you have a background where things look perfectly standard from a tensor perspective, you still get this extra scalar source term.

Kai: And they give us that leading scalar source term as S GW R = h i dot j h i dot j divided by eight plus some terms that depend on the Hubble parameter H over the frequency omega.

Mira: That structure, with the division by eight, is a direct consequence of that cubic scalar–tensor–tensor coupling they use in their work.

Lev: From an engineering standpoint, if you were building hardware to test this, you’d be looking at how much overhead this extra scalar source adds to the real physical measurement compared to just the standard tensor contribution.

Kai: And they do compare these two response channels: the finite-wavenumber scalar response and the direct tensor contribution.

Mira: They find that these are distinct channels, and a universal ratio between them cannot be assigned without specifying how you project that finite-k scalar perturbation onto the homogeneous sector.

Lev: That makes sense because one is a wave effect at a specific frequency, while the other is an average energy density contribution to the background constraint. They aren't directly comparable without more assumptions on how they interact in practice.

Kai: So, what does this mean for us? The paper "Isaacson energy--momentum tensor of high-frequency gravitational waves in general f(R,G) gravity" is really about showing that even in these restricted gravity theories, the gravitational wave backreaction isn't just a simple modification of the standard tensor stress.

Mira: It points to a persistent scalar channel generated by that cubic interaction, even when the background itself seems to be behaving like standard de Sitter space for the tensor speed.

Lev: It means any future experiments looking at high-frequency waves in these modified gravity regimes need to account for this extra scalar source, not just the standard tensor effects.

Kai: And they also noted a limitation, which is important: the condition that makes the Hessian degenerate does not establish global ghost freedom for all f(R,G) models.

Mira: That’s a caveat they put there. They stress that you still need to check the stability of these systems on the background by looking at those kinetic and gradient coefficients, G T and F T.

Lev: And they also have to remember this analysis is restricted to high-frequency regimes, not necessarily full nonlinear evolution of all sectors.

Kai: So, to wrap up, the main conclusion of "Isaacson energy--momentum tensor of high-frequency gravitational waves in general f(R,G) gravity" is that the nonzero leading scalar source persists due to the cubic scalar–tensor–tensor coupling even when c squared T equals one on a rolling de Sitter background <ref:2504.15709#pg1>.

Mira: And they found that this finite-wavenumber scalar response scales as R (two) k divided by omega squared S in the geometric optics regime <ref:2504.15709#pg1>.

Lev: That structure, that scaling with omega squared S, is what you’d need to consider if you were trying to run a simulation on real hardware and trying to isolate the effect of this source term <ref:2504.15709#pg1>.

Kai: So, when we think about this paper, it's not just about finding new terms; it's about defining precisely which parts of the theory produce these distinct response channels and how they compare.

Mira: It shows that the structure inside a specifically degenerate f(R, G) sector can lead to results that are quite different from what you might expect in simpler models.

Lev: And for future work, I think the next step would be testing those kinetic and gradient coefficients on real systems to see how much they deviate from the idealized values derived here.

Conclusion: Kai: So we're looking at the conclusion of this paper, "Isaacson energy--momentum tensor of high-frequency gravitational waves in general f(R,G) gravity." It boils down to showing how high frequency gravitational waves backreact on spacetime in these specific modified gravity theories.

Mira: Yeah, the authors basically take a complicated setup where you've got the Ricci scalar and Gauss–Bonnet invariant mixed together and they find that even when the tensor speed is set to one on a de Sitter background, there's still this extra scalar source term sticking around.

Lev: That’s what I found interesting too. The paper says that this nonzero source isn't just an artifact of the tensor dispersion relation; it comes from that cubic scalar–tensor–tensor interaction, which is a different mechanism entirely.

Kai: So, for someone just listening to the show, what does this mean? It means that even in a background where gravity looks like standard de Sitter space on large scales, high frequency waves still generate a measurable scalar field effect that isn't just the standard tensor contribution.

Mira: Right. The paper is saying you can't just look at the tensor response and assume everything is accounted for; there’s this persistent scalar channel you need to consider.

Lev: From an engineering side, it means when we try to build detectors or simulate these systems, we can't ignore this extra source term because it scales differently with the frequency and how fast the background is rolling.

Kai: And the authors are very clear about what they *didn't* do; they say this analysis is only on a specific rolling de Sitter solution, not every possible universe. They also note that the condition making it degenerate doesn't automatically guarantee global ghost freedom for all these f(R,G) models.

Mira: That’s a fair point about the limitations. The paper has to be careful about those assumptions underpinning the whole thing because that's where things get shaky in modified gravity research.

Lev: And they stress that stability needs to be checked by looking at those kinetic and gradient coefficients, G T and F T, not just assuming everything is fine based on the background solution alone.

Kai: So the main point is this specific sector of f(R,G) gravity allows for a unique way gravity couples waves back onto the scalar field even under these seemingly simple conditions.

Mira: Exactly. It gives us a concrete example where we see how those cubic couplings manifest in observable response channels at high frequencies.

Lev: We’ve got this specific result now that we can use to test if our real hardware setup can actually measure the predicted scalar source term, which is what we need for real validation.

Farzad Milani

Technical and Vocational University (TVU)

gr-qc

Submitted: 2025-04-22

Updated: 2026-11-01

Comments: v6: new title and a new analysis; supersedes v1-v5. 14 pages. Code: doi:10.5281/zenodo.23811207

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 80/100

The gist: We investigate the high-frequency backreaction of gravitational waves in the Hessian-degenerate sector of four-dimensional f(R, G) gravity defined by f(R, G) = F(R + G/M2).

Key concepts

Hessian-degenerate sector
This refers to a specific family of f(R, G) gravity models where the Hessian matrix of the function f(R + G/M2) has rank one. This mathematical property is crucial because it reduces the theory's complexity, allowing it to be mapped onto a simpler scalar–Gauss–Bonnet Horndeski theory with only one scalar degree of freedom.
Horndeski gravity mapping
The degenerate f(R, G) model is mathematically equivalent to a specific subclass of Horndeski gravity. This mapping involves defining specific functions (G2, G3, G4, G5) that describe the theory's dynamics. This allows researchers to use established tools from Horndeski gravity to analyze the tensor and scalar sectors.
Tensor speed deviation ($c^2_T$)
This parameter measures how the speed of gravitational waves deviates from the speed of light in this modified gravity theory. It is determined by kinetic and gradient coefficients derived from the Horndeski mapping. In this model, $c^2_T = rac{FT}{GT}$, and it can be different from one even on a rolling de Sitter background.
Scalar source generation
The paper demonstrates that a genuine scalar channel can be generated by the cubic interaction between the scalar field and gravitational waves. This means that gravitational waves act as a source for a new type of scalar perturbation, which is not simply related to the tensor speed deviation, even when $c^2_T = 1.

Terminology

Summary

We investigate the high-frequency backreaction of gravitational waves in the Hessian-degenerate sector of four-dimensional f(R, G) gravity defined by f(R, G) = F(R + G/M2). p. 1

Where the paper says it

Abstract We investigate the high-frequency backreaction of gravitational waves in the Hessian-degenerate sector of four-dimensional f(R, G) gravity defined by f(R, G) = F(R + G/M2 ). The degeneracy of the Hessian reduces the Legendre representation to a single scalar degree of freedom and maps the theory into a scalar–Gauss–Bonnet subclass of Horndeski gravity.

The degeneracy of the Hessian reduces the Legendre representation to a single scalar degree of freedom and maps the theory into a scalar–Gauss–Bonnet subclass of Horndeski gravity. p. 1
Where the paper says it

Abstract We investigate the high-frequency backreaction of gravitational waves in the Hessian-degenerate sector of four-dimensional f(R, G) gravity defined by f(R, G) = F(R + G/M2 ). The degeneracy of the Hessian reduces the Legendre representation to a single scalar degree of freedom and maps the theory into a scalar–Gauss–Bonnet subclass of Horndeski gravity.

p. 1
Where the paper says it

Abstract We investigate the high-frequency backreaction of gravitational waves in the Hessian-degenerate sector of four-dimensional f(R, G) gravity defined by f(R, G) = F(R + G/M2 ). The degeneracy of the Hessian reduces the Legendre representation to a single scalar degree of freedom and maps the theory into a scalar–Gauss–Bonnet subclass of Horndeski gravity.

Model Reduction and Mapping

The paper restricts attention to the Hessian-degenerate family f(R, G) = F(X), X = R + G/M2, which admits a one-field Legendre representation rather than the two-field representation appropriate to a nondegenerate Hessian I.3. The theory is consequently equivalent to a scalar–Gauss–Bonnet Horndeski theory with one scalar degree of freedom on the cosmological background III. p. 2

Where the paper says it

The main results are: 1. the locked theory is equivalent to a scalar–Gauss–Bonnet Horndeski theory with one scalar degree of freedom on the cosmological background; 2.

1, and the Gauss–Bonnet coupling in (II. p. 4
Where the paper says it

Generic f(R, G) theories can possess ghost modes on anisotropic backgrounds, while the number of scalar modes changes on FLRW backgrounds [4]. The Gauss–Bonnet coupling in (II.

4) is dynamical rather than topological since the coefficient is given by ϕ/M2 II. p. 4
Where the paper says it

4) is a standard scalar–Gauss–Bonnet interaction. In four dimensions a constant coefficient multiplying G is topological [9, 10], but the present coefficient ϕ/M2 is dynamical and therefore contributes to the field equations.

4>. The Horndeski representation for this model involves specific functions, such as G2 = −M2Pl 2V(ϕ), G3 = 0, G4 = M2Pl 2ϕ, and G5 = −2M2Pl/Mln XH III. p. 3
Where the paper says it

These formulas are valid for the general Horndeski class and do not require a canonical X term in G2; the kinetic term X = −∇µϕ∇µϕ/2 is simply the variable on which the Horndeski functions may depend. The scalar kinetic structure in the present model arises from the nonminimal curvature mixing.

5>.

Tensor Sector Dynamics

The tensor sector has modified kinetic and gradient coefficients derived from the Horndeski mapping III.3, specifically GT = M2Plϕ + 4Hϕ˙M2! p. 1

Where the paper says it

The degeneracy of the Hessian reduces the Legendre representation to a single scalar degree of freedom and maps the theory into a scalar–Gauss–Bonnet subclass of Horndeski gravity. We derive the tensor quadratic action and obtain the modified kinetic and gradient coefficients QT = ϕ + 4Hϕ/M˙ 2 and FT = ϕ+4ϕ/M¨ 2 in units where the overall factor M2 Pl is extracted.

and FT = M2Plϕ + 4ϕ¨M2! p. 1
Where the paper says it

The degeneracy of the Hessian reduces the Legendre representation to a single scalar degree of freedom and maps the theory into a scalar–Gauss–Bonnet subclass of Horndeski gravity. We derive the tensor quadratic action and obtain the modified kinetic and gradient coefficients QT = ϕ + 4Hϕ/M˙ 2 and FT = ϕ+4ϕ/M¨ 2 in units where the overall factor M2 Pl is extracted.

III.10>. These coefficients determine the tensor speed deviation c 2T = FT/GT = ϕ + 4phi/M¨ 2 ϕ + 4Hphi/M˙ 2 III. p. 1
Where the paper says it

We then use the general Horndeski scalar–tensor–tensor cubic action to identify the induced scalar channel generated by a stochastic transverse-traceless gravitational-wave background. The scalar source is generically nonzero and, importantly, it is not determined solely by the tensor-speed deviation: it can remain nonzero even on a rolling de Sitter background for which c 2 T = 1.

11>. The quadratic action for the tensor perturbation is S(2)T = 1/8 Z dt d3x a3 GT h˙ijh˙ij − FT a squared ∂khij∂khij C. p. 7
Where the paper says it

Quadratic action The tensor perturbation is transverse and traceless, ∂ihij = 0, hii = 0. 1) The quadratic action is S (2) T = 1 8 Z dt d3x a3 GT h˙ ijh˙ ij − FT a 2 ∂khij∂khij .

2, and the leading dispersion relation in the WKB regime is ω squared = c 2T k squared a squared IV. p. 7
Where the paper says it

3) In the WKB regime, hij = Aij (t)e iθ/ϵ, ϵ ∼ λ L ≪ 1, (IV. 4) and the leading dispersion relation is ω 2 = c 2 T k 2 a 2 .

4, which leads to a canonical amplitude scaling as Ak ∝ 1/a√GT cT = 1/a(GTFT /MPl) 1/4 C. p. 7
Where the paper says it

4) and the leading dispersion relation is ω 2 = c 2 T k 2 a 2 . 5) The canonical amplitude scales as Ak ∝ 1 a √ GT cT = 1 a(GTFT ) 1/4 .

7>.

Scalar Source Generation

The second key result is the existence of a genuine scalar channel sourced by the cubic scalar–tensor–tensor interaction V.2, which is not reducible to a measure of the tensor-speed deviation VI. p. 1

Where the paper says it

We then use the general Horndeski scalar–tensor–tensor cubic action to identify the induced scalar channel generated by a stochastic transverse-traceless gravitational-wave background. The scalar source is generically nonzero and, importantly, it is not determined solely by the tensor-speed deviation: it can remain nonzero even on a rolling de Sitter background for which c 2 T = 1.

1. On a rolling de Sitter solution, both c 2T and c 2S equal unity, but the induced scalar source remains nonzero V. p. 2
Where the paper says it

the cubic scalar–tensor–tensor interaction generates a nonzero scalar source even when cT = 1; 5. on an exact rolling de Sitter solution the scalar propagation is luminal, c 2 S = 1, while the leading scalar source remains nonzero, S GW R = ⟨h˙ ijh˙ ij ⟩/8 + O(H/ω); the finite-k response scales as R (2) k ∼ SGW R,k /ω2 S .

2, specifically S GW R = 1/8 ⟨h˙ijh˙ij⟩ [1 + O(H/ω)] V. p. 1
Where the paper says it

The scalar source is generically nonzero and, importantly, it is not determined solely by the tensor-speed deviation: it can remain nonzero even on a rolling de Sitter background for which c 2 T = 1. For an exact analytic rolling de Sitter solution, we find c 2 T = c 2 S = 1 and obtain a nonzero leading scalar source, S GW R = ⟨h˙ ijh˙ ij ⟩/8 + O(H/ω).

14>.

Comparison of Response Channels

The finite-wavenumber scalar response is controlled by the inverse scalar WKB operator, giving R(2)k ∼ S GW R,k / ω 2S at finite wavenumber VI.9, which is not identified with a homogeneous Hubble shift VI. p. 1

Where the paper says it

The finite-wavenumber scalar response scales as R (2) k ∼ SGW R,k /ω2 S in the geometric-optics regime. We do not identify this finite-k response with a universal homogeneous Hubble shift; such a conversion requires an explicit zero-mode projection and averaging prescription.

3. The direct tensor contribution to the Friedmann constraint scales as δH(2)T = ρGW / M2PlF1,H = QT 4F1,H ⟨h˙ijh˙ij⟩ VI. p. 7
Where the paper says it

5) The canonical amplitude scales as Ak ∝ 1 a √ GT cT = 1 a(GTFT ) 1/4 . Effective GW stress tensor Averaging the quadratic tensor action over several wavelengths gives ρGW = M2 Pl 4 ϕ + 4Hϕ˙ M2 !

13>, and a universal ratio between the scalar response and the direct homogeneous tensor backreaction cannot be assigned without specifying how the finite-k scalar perturbation is projected onto the homogeneous sector VI. p. 1
Where the paper says it

The finite-wavenumber scalar response scales as R (2) k ∼ SGW R,k /ω2 S in the geometric-optics regime. We do not identify this finite-k response with a universal homogeneous Hubble shift; such a conversion requires an explicit zero-mode projection and averaging prescription.

4>.

Conclusion on Control

The analysis concludes that while c 2T = 1 in the rolling de Sitter control background, a nonzero leading scalar source persists due to the cubic scalar–tensor–tensor coupling V.16. The finite-k response is distinct from the direct tensor contribution, and this distinction is a result of the specific organization and backreaction interpretation inside the specifically degenerate f(R, G) sector VII. p. 3

Where the paper says it

Section VI derives the scalar response and its comparison with the direct tensor backreaction. Sections VII and VIII discuss the physical interpretation and limitations.

Scope and Limitations

Several qualifications are essential, including noting that the condition ∆f = 0 does not establish global ghost freedom VIII, and stability must be checked on the background through the kinetic and gradient coefficients VIII. The analysis is also restricted to the high-frequency regime and does not constitute a full nonlinear evolution of the coupled tensor, scalar, and background sectors VIII. p. 1

Where the paper says it

We emphasize that the result does not constitute a proof of global ghost freedom or a complete nonlinear treatment of the theory. INTRODUCTION Gravitational waves (GWs) carry energy and momentum and, once averaged over scales long compared with their wavelength, act as an effective source for the background geometry.

p. 1
Where the paper says it

We emphasize that the result does not constitute a proof of global ghost freedom or a complete nonlinear treatment of the theory. INTRODUCTION Gravitational waves (GWs) carry energy and momentum and, once averaged over scales long compared with their wavelength, act as an effective source for the background geometry.

The explicit analytic background test is based on a rolling de Sitter solution and therefore does not establish the same scalar-source coefficient for every possible cosmological history VIII. p. 6
Where the paper says it

18) Stability of the cosmological perturbations requires, within this perturbative description, GT > 0, FT > 0, GS > 0, FS > 0. Exact rolling de Sitter solution An exact analytic background useful for testing the scalar channel is H = H0, ϕ(t) = ϕ0 + Φe H0t .

The conversion of the finite-k scalar response into a homogeneous Hubble shift is prescription-dependent VIII. p. 1
Where the paper says it

The finite-wavenumber scalar response scales as R (2) k ∼ SGW R,k /ω2 S in the geometric-optics regime. We do not identify this finite-k response with a universal homogeneous Hubble shift; such a conversion requires an explicit zero-mode projection and averaging prescription.

The comparison with general relativity is instructive, as in GR there is no additional Horndeski scalar channel of this type VII>. p. 1
Where the paper says it

In general relativity this idea is captured by the Isaacson high-frequency expansion, in which the quadratic part of the metric perturbation is interpreted as an effective stress tensor [1, 2]. The same separation between rapidly varying wave degrees of freedom and slowly varying background fields is useful in modified gravity, but the averaging problem becomes more subtle because the gravitational action can contain additional propagating degrees of freedom and time-dependent effective gravitational couplings [3].

Appendix Details

The background equations are derived directly from the locked scalar–Gauss–Bonnet action S = M2Pl squared Z d4x √−g ϕR + ϕG/M2 − V (ϕ) A.1, yielding the authoritative equation (M2 + 4H 2)ϕ¨ + (8HH˙ − HM2 − 4H 3)ϕ˙ + 2M2H ϕ˙ = 0 A. p. 5

Where the paper says it

7) The Friedmann equation can be written as 3H 2ϕ = 1 2 V − 3Hϕ˙ − 12H3ϕ˙ M2 . 8) (M2 + 4H 2 )ϕ¨ + (8HH˙ − HM2 − 4H 3 )ϕ˙ + 2M2Hϕ˙ = 0, (III.

14>. The cubic coefficients are collected in Appendix D, where for the rolling de Sitter solution, b1 = b2 = GS/8 D. p. 2
Where the paper says it

the cubic scalar–tensor–tensor interaction generates a nonzero scalar source even when cT = 1; 5. on an exact rolling de Sitter solution the scalar propagation is luminal, c 2 S = 1, while the leading scalar source remains nonzero, S GW R = ⟨h˙ ijh˙ ij ⟩/8 + O(H/ω); the finite-k response scales as R (2) k ∼ SGW R,k /ω2 S .

11>. The tensor WKB relation is given by ⟨∂khij∂khij⟩ a squared = 1/c 2T ⟨h˙ 2ij⟩ + O(H/ω) D. p. 7
Where the paper says it

Quadratic action The tensor perturbation is transverse and traceless, ∂ihij = 0, hii = 0. 1) The quadratic action is S (2) T = 1 8 Z dt d3x a3 GT h˙ ijh˙ ij − FT a 2 ∂khij∂khij .

14>.

Dimensional Conventions

The dimensional analysis uses natural units where [H] = [ϕ˙] = [M] = [MPl] = M, and the effective tensor energy density is given by ρGW = M2Pl 4 QT ⟨h˙ijh˙ij⟩ E.3>. The second-order background constraint yields δH(2)T = ρGW / M2PlF1,H = QT 4F1,H ⟨h˙ijhdotij⟩ E. p. 8

Where the paper says it

10) The separation is not unique; only the total averaged second-order equation is covariantly conserved. In the FLRW frame, ∇µ t GW µ0 = − c˙T cT ρGW + O(H/ω), (IV.

5>. The corrected background equation terms each have dimension M4 E. p. 5
Where the paper says it

Background equations For a spatially flat FLRW background, ds2 = −dt2 + a 2 (t)dx 2 , H = a˙ a , (III. 6) the scalar equation is V,ϕ = 12H 2 + 6H˙ + 24H2 (H2 + H˙ ) M2 .

6>.

Green Function Derivation

The retarded Green function is defined as GR(k;t, t′) = Θ(t − t′) uk(t′)vk(t) − vk(t′)uk(t) Wk(t') VI.3, and the corresponding retarded response is R 2 k (D. p. 1

Where the paper says it

For an exact analytic rolling de Sitter solution, we find c 2 T = c 2 S = 1 and obtain a nonzero leading scalar source, S GW R = ⟨h˙ ijh˙ ij ⟩/8 + O(H/ω). The finite-wavenumber scalar response scales as R (2) k ∼ SGW R,k /ω2 S in the geometric-optics regime.

18). The homogeneous scalar equation for a Fourier mode is u¨k + 3H + G˙S GS! p. 1
Where the paper says it

The finite-wavenumber scalar response scales as R (2) k ∼ SGW R,k /ω2 S in the geometric-optics regime. We do not identify this finite-k response with a universal homogeneous Hubble shift; such a conversion requires an explicit zero-mode projection and averaging prescription.

u˙ k + c 2S k squared a squared uk = 0 F.1, leading to the WKB inversion of the differential operator giving R(2)k ∼ S GW R,k ω 2S [1 + O(H/ωS)] D. p. 7
Where the paper says it

3) In the WKB regime, hij = Aij (t)e iθ/ϵ, ϵ ∼ λ L ≪ 1, (IV. 4) and the leading dispersion relation is ω 2 = c 2 T k 2 a 2 .

18>.

Final Source Expression

On rolling de Sitter, the nonzero source at c 2T = 1 is thus a direct consequence of the cubic scalar–tensor–tensor coupling and is not controlled solely by tensor dispersion D.16. The sourced scalar equation is 1/a 3GS d/dt a 3GSR˙(2) − c 2S∇ 2R(2)a squared = S GW R D. p. 5

Where the paper says it

Background equations For a spatially flat FLRW background, ds2 = −dt2 + a 2 (t)dx 2 , H = a˙ a , (III. 6) the scalar equation is V,ϕ = 12H 2 + 6H˙ + 24H2 (H2 + H˙ ) M2 .

17>. The leading source is S GW R = b1 + b2 2GS ⟨h˙ 2ij⟩ a squared + O(H/ω) D. p. 1
Where the paper says it

The scalar source is generically nonzero and, importantly, it is not determined solely by the tensor-speed deviation: it can remain nonzero even on a rolling de Sitter background for which c 2 T = 1. For an exact analytic rolling de Sitter solution, we find c 2 T = c 2 S = 1 and obtain a nonzero leading scalar source, S GW R = ⟨h˙ ijh˙ ij ⟩/8 + O(H/ω).

16>. The nonzero value of (V. p. 1
Where the paper says it

We then use the general Horndeski scalar–tensor–tensor cubic action to identify the induced scalar channel generated by a stochastic transverse-traceless gravitational-wave background. The scalar source is generically nonzero and, importantly, it is not determined solely by the tensor-speed deviation: it can remain nonzero even on a rolling de Sitter background for which c 2 T = 1.

14) despite c 2T = 1 is one of the principal results of this work VI. p. 2
Where the paper says it

The main results are: 1. the locked theory is equivalent to a scalar–Gauss–Bonnet Horndeski theory with one scalar degree of freedom on the cosmological background; 2.

3>. The finite-wavenumber scalar response is controlled by the inverse scalar WKB operator, giving R(2)k ∼ S GW R,k/ω 2S at finite wavenumber D. p. 1
Where the paper says it

For an exact analytic rolling de Sitter solution, we find c 2 T = c 2 S = 1 and obtain a nonzero leading scalar source, S GW R = ⟨h˙ ijh˙ ij ⟩/8 + O(H/ω). The finite-wavenumber scalar response scales as R (2) k ∼ SGW R,k /ω2 S in the geometric-optics regime.

18>. The direct tensor contribution remains well defined VI. p. 4
Where the paper says it

The Horndeski formulas are formulated for arbitrary Gi(ϕ, X) and therefore apply to (III. 5) whenever the background is regular and the scalar gradient is timelike, so that the unitary-gauge scalar perturbation variable is well defined.

3>. The direct homogeneous tensor contribution remains well defined VI. p. 4
Where the paper says it

The Horndeski formulas are formulated for arbitrary Gi(ϕ, X) and therefore apply to (III. 5) whenever the background is regular and the scalar gradient is timelike, so that the unitary-gauge scalar perturbation variable is well defined.

3>. Both expressions have dimension M E.2>. Both expressions have dimension M4 E.6>. Both expressions have dimension M E.3>. The tensor perturbation hij is dimensionless because it is defined through the exponential spatial metric parametrization C. p. 1
Where the paper says it

In general relativity this idea is captured by the Isaacson high-frequency expansion, in which the quadratic part of the metric perturbation is interpreted as an effective stress tensor [1, 2]. The same separation between rapidly varying wave degrees of freedom and slowly varying background fields is useful in modified gravity, but the averaging problem becomes more subtle because the gravitational action can contain additional propagating degrees of freedom and time-dependent effective gravitational couplings [3].

6>.

--- Page 1 ---

Scalar Backreaction from Gravitational Waves in a Degenerate f(R, G) Sector

Farzad Milani∗

Department of Basic Sciences, Technical and Vocational University (TVU), Tehran, Iran. (Dated: October 2, 2026) arXiv:2504. p. 0

Where the paper says it

Scalar Backreaction from Gravitational Waves in a Degenerate f(R, G) Sector Farzad Milani∗ Department of Basic Sciences, Technical and Vocational University (TVU), Tehran, Iran. (Dated: October 2, 2026) 1 arXiv:2504.

15709v5 [gr-qc] 1 Oct 2026

--- Page 3 ---

Abstract

We investigate the high-frequency backreaction of gravitational waves in the Hessian-degenerate sector of four-dimensional f(R, G) gravity defined by f(R, G) = F(R + G/M2). The degeneracy of the Hessian reduces the Legendre representation to a single scalar degree of freedom and maps the theory into a scalar–Gauss–Bonnet subclass of Horndeski gravity.

--- Page 4 ---

I. INTRODUCTION

Gravitational waves (GWs) carry energy and momentum and, once averaged over scales long compared with their wavelength, act as an effective source for the background geometry I.1. In general relativity this idea is captured by the Isaacson high-frequency expansion, in which the quadratic part of the metric perturbation is interpreted as an effective stress tensor [1, 2]. p. 1

Where the paper says it

INTRODUCTION Gravitational waves (GWs) carry energy and momentum and, once averaged over scales long compared with their wavelength, act as an effective source for the background geometry. In general relativity this idea is captured by the Isaacson high-frequency expansion, in which the quadratic part of the metric perturbation is interpreted as an effective stress tensor [1, 2].

The same separation between rapidly varying wave degrees of freedom and slowly varying background fields is useful in modified gravity, but the averaging problem becomes more subtle because the gravitational action can contain additional propagating degrees of freedom and time-dependent effective gravitational couplings [3]. p. 1
Where the paper says it

In general relativity this idea is captured by the Isaacson high-frequency expansion, in which the quadratic part of the metric perturbation is interpreted as an effective stress tensor [1, 2]. The same separation between rapidly varying wave degrees of freedom and slowly varying background fields is useful in modified gravity, but the averaging problem becomes more subtle because the gravitational action can contain additional propagating degrees of freedom and time-dependent effective gravitational couplings [3].

The class of f(R, G) theories is particularly interesting because it combines the Ricci scalar with the four-dimensional Gauss–Bonnet invariant I. p. 1
Where the paper says it

The same separation between rapidly varying wave degrees of freedom and slowly varying background fields is useful in modified gravity, but the averaging problem becomes more subtle because the gravitational action can contain additional propagating degrees of freedom and time-dependent effective gravitational couplings [3]. The class of f(R, G) theories is particularly interesting because it combines the Ricci scalar with the four-dimensional Gauss–Bonnet invariant.

1. For a generic nonlinear function of R and G, the scalar sector can be richer than that of f(R) gravity, and perturbative stability can depend sensitively on the background and on the rank of the Hessian of f [4, 5]. p. 2
Where the paper says it

stability can depend sensitively on the background and on the rank of the Hessian of f [4, 5]. Consequently, a generic statement about “the” gravitational-wave backreaction of f(R, G) gravity is not appropriate without specifying a dynamically controlled subclass.

Consequently, a generic statement about “the” gravitational-wave backreaction of f(R, G) gravity is not appropriate without specifying a dynamically controlled subclass I. p. 2
Where the paper says it

stability can depend sensitively on the background and on the rank of the Hessian of f [4, 5]. Consequently, a generic statement about “the” gravitational-wave backreaction of f(R, G) gravity is not appropriate without specifying a dynamically controlled subclass.

1. Here we restrict attention to the Hessian-degenerate family f(R, G) = F(X), X = R + G/M2, (I. p. 2
Where the paper says it

Consequently, a generic statement about “the” gravitational-wave backreaction of f(R, G) gravity is not appropriate without specifying a dynamically controlled subclass. Here we restrict attention to the Hessian-degenerate family[6] f(R, G) = F(X), X = R + G M2 , (I.

1) where M is a constant mass scale I. p. 2
Where the paper says it

Here we restrict attention to the Hessian-degenerate family[6] f(R, G) = F(X), X = R + G M2 , (I. 1) where M is a constant mass scale.

1>. Its Hessian has rank one, fRR = FXX, fRG = FXX/M2, fGG = FXX/M4, so that ∆f ≡ fRRfGG − f 2RG = 0 I. p. 2
Where the paper says it

Its Hessian has rank one, fRR = FXX, fRG = FXX M2 , fGG = FXX M4 , (I. 2) so that ∆f ≡ fRRfGG − f 2 RG = 0.

3>. The theory consequently admits a one-field Legendre representation rather than the two-field representation appropriate to a nondegenerate Hessian I. p. 2
Where the paper says it

2) so that ∆f ≡ fRRfGG − f 2 RG = 0. 3) The theory consequently admits a one-field Legendre representation rather than the two-field representation appropriate to a nondegenerate Hessian.

3>.

Pruning redundant reasoning tokens

The central purpose of this paper is not to claim generic ghost freedom for all f(R, G) models I.1. Instead, we derive the tensor backreaction and the induced scalar channel in this controlled degenerate sector I. p. 2

Where the paper says it

The central purpose of this paper is not to claim generic ghost freedom for all f(R, G) models. Instead, we derive the tensor backreaction and the induced scalar channel in this controlled degenerate sector.

1>. The main results are:

  1. the locked theory is equivalent to a scalar–Gauss–Bonnet Horndeski theory with one scalar degree of freedom on the cosmological background I. p. 2

    Where the paper says it

    The main results are: 1. the locked theory is equivalent to a scalar–Gauss–Bonnet Horndeski theory with one scalar degree of freedom on the cosmological background; 2.

    2;

  2. the tensor kinetic and gradient coefficients are GT = M2Plϕ + 4Hϕ˙M2! p. 1

    Where the paper says it

    The degeneracy of the Hessian reduces the Legendre representation to a single scalar degree of freedom and maps the theory into a scalar–Gauss–Bonnet subclass of Horndeski gravity. We derive the tensor quadratic action and obtain the modified kinetic and gradient coefficients QT = ϕ + 4Hϕ/M˙ 2 and FT = ϕ+4ϕ/M¨ 2 in units where the overall factor M2 Pl is extracted.

    and FT = M2Plϕ + 4phi¨M2! I.4;

  3. the leading high-frequency tensor stress has ρGW = M2Pl 4 ϕ + 4Hϕ˙M2! p. 2

    Where the paper says it

    , FT = M2 Pl ϕ + 4ϕ¨ M2 ! the leading high-frequency tensor stress has ρGW = M2 Pl 4 ϕ + 4Hϕ˙ M2 !

    ⟨h˙ijh˙ij⟩, pGW = 1/3 ρGW I. p. 2
    Where the paper says it

    the leading high-frequency tensor stress has ρGW = M2 Pl 4 ϕ + 4Hϕ˙ M2 ! ⟨h˙ ijh˙ ij ⟩, pGW = 1 3 ρGW; (I.

    5;

  4. the cubic scalar–tensor–tensor interaction generates a nonzero scalar source even when cT = 1 I. p. 2

    Where the paper says it

    ⟨h˙ ijh˙ ij ⟩, pGW = 1 3 ρGW; (I. the cubic scalar–tensor–tensor interaction generates a nonzero scalar source even when cT = 1; 5.

    6;

and

on an exact rolling de Sitter solution the scalar propagation is luminal, c 2S = 1, while the leading scalar source remains nonzero, S GW R = ⟨h˙ijh˙ij⟩/8 + O(H/ω) I.6.

How the training signal is built

The analysis uses the general Horndeski perturbation formalism of Kobayashi, Yamaguchi and Yokoyama [7] and the general cubic scalar–tensor–tensor action of Gao et al. [8] III.1, III.2>. The scalar kinetic structure in the present model arises from the nonminimal curvature mixing III. p. 3

Where the paper says it

These formulas are valid for the general Horndeski class and do not require a canonical X term in G2; the kinetic term X = −∇µϕ∇µϕ/2 is simply the variable on which the Horndeski functions may depend. The scalar kinetic structure in the present model arises from the nonminimal curvature mixing.

3>. The paper is organized as follows, with Section II introducing the degenerate sector and its Legendre form IV, Section III giving the Horndeski mapping and the cosmological background III, Section IV deriving the tensor sector and its effective stress tensor IV, Section V treating the cubic scalar–tensor–tensor interaction V, and Sections VI through VIII discussing the physical interpretation and limitations VI-VIII>. p. 3
Where the paper says it

Section III gives the Horndeski mapping and the cosmological background. Section IV derives the tensor sector and its effective stress tensor.

Where the gains come from

The main results are derived by substituting the specific model into general formulas for Horndeski functions and cubic interactions III.1, III.2, V.1, V.3>, leading to the identification of the induced scalar channel generated by a stochastic transverse-traceless gravitational-wave background Abstract>. p. 1

Where the paper says it

We derive the tensor quadratic action and obtain the modified kinetic and gradient coefficients QT = ϕ + 4Hϕ/M˙ 2 and FT = ϕ+4ϕ/M¨ 2 in units where the overall factor M2 Pl is extracted. We then use the general Horndeski scalar–tensor–tensor cubic action to identify the induced scalar channel generated by a stochastic transverse-traceless gravitational-wave background.

The cubic scalar–tensor–tensor interaction generates a nonzero scalar source even when cT = 1 I. p. 2
Where the paper says it

⟨h˙ ijh˙ ij ⟩, pGW = 1 3 ρGW; (I. the cubic scalar–tensor–tensor interaction generates a nonzero scalar source even when cT = 1; 5.

6, and the rolling de Sitter solution provides a particularly clean demonstration III. p. 1
Where the paper says it

The scalar source is generically nonzero and, importantly, it is not determined solely by the tensor-speed deviation: it can remain nonzero even on a rolling de Sitter background for which c 2 T = 1. For an exact analytic rolling de Sitter solution, we find c 2 T = c 2 S = 1 and obtain a nonzero leading scalar source, S GW R = ⟨h˙ ijh˙ ij ⟩/8 + O(H/ω).

4>. The resulting leading scalar source is S GW R = b1 + b2 2GS ⟨h˙ 2ij⟩ + O(H/ω) D. p. 1
Where the paper says it

The scalar source is generically nonzero and, importantly, it is not determined solely by the tensor-speed deviation: it can remain nonzero even on a rolling de Sitter background for which c 2 T = 1. For an exact analytic rolling de Sitter solution, we find c 2 T = c 2 S = 1 and obtain a nonzero leading scalar source, S GW R = ⟨h˙ ijh˙ ij ⟩/8 + O(H/ω).

16, which is a direct consequence of the cubic scalar–tensor–tensor coupling and is not controlled solely by tensor dispersion D. p. 1
Where the paper says it

We then use the general Horndeski scalar–tensor–tensor cubic action to identify the induced scalar channel generated by a stochastic transverse-traceless gravitational-wave background. The scalar source is generically nonzero and, importantly, it is not determined solely by the tensor-speed deviation: it can remain nonzero even on a rolling de Sitter background for which c 2 T = 1.

16>.

Physical Interpretation

The finite-wavenumber scalar response is determined directly from the sourced scalar equation, and this quantity is a finite-k perturbative response and is not, by itself, a homogeneous shift of the Hubble parameter VI.12>. The direct tensor contribution enters the homogeneous background constraint through the averaged GW energy density VI. p. 1

Where the paper says it

INTRODUCTION Gravitational waves (GWs) carry energy and momentum and, once averaged over scales long compared with their wavelength, act as an effective source for the background geometry. In general relativity this idea is captured by the Isaacson high-frequency expansion, in which the quadratic part of the metric perturbation is interpreted as an effective stress tensor [1, 2].

3>. These are distinct response channels, and a universal ratio between them cannot be assigned without specifying how the finite-k scalar perturbation is projected onto the homogeneous sector VI. p. 1
Where the paper says it

The finite-wavenumber scalar response scales as R (2) k ∼ SGW R,k /ω2 S in the geometric-optics regime. We do not identify this finite-k response with a universal homogeneous Hubble shift; such a conversion requires an explicit zero-mode projection and averaging prescription.

4>.

Limitations

The condition ∆f = 0 does not establish global ghost freedom VIII, and stability must be checked on the background through the kinetic and gradient coefficients VIII. The analysis is also restricted to the high-frequency regime and does not constitute a full nonlinear evolution of the coupled tensor, scalar, and background sectors VIII. The analytic rolling de Sitter solution is a control background, not a universal cosmological history VIII. p. 6

Where the paper says it

18) Stability of the cosmological perturbations requires, within this perturbative description, GT > 0, FT > 0, GS > 0, FS > 0. Exact rolling de Sitter solution An exact analytic background useful for testing the scalar channel is H = H0, ϕ(t) = ϕ0 + Φe H0t .

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved understanding of modified gravity backreaction in degenerate sectors: The system can now accurately model gravitational wave effects in f(R, G) theories by using a single-field Legendre representation equivalent to a scalar Gauss-Bonnet Horndeski theory.

  2. Precise calculation of the induced scalar source: The AI can determine that the leading scalar source remains nonzero even on a rolling de Sitter background for which cT = 1, specifically calculating it as S GW R = ⟨h˙ijh˙ij ⟩/8 + O(H/ω) using the cubic scalar tensor tensor interaction.

  3. Discrimination between response channels: The system can distinguish between the finite-k scalar response, which scales as R(2)k S GW R,k / ω 2S, and the direct homogeneous tensor backreaction, which is controlled by δH(2)T = ρGW / M2PlF1,H.

  4. Analysis of coupling dependence: The AI can analyze how the cubic interaction source is not reducible to a measure of the tensor-speed deviation and demonstrate this using the result that on rolling de Sitter, both cT and cS equal unity, yet the induced scalar source remains nonzero.

  5. Modeling wave propagation in modified gravity: The system can predict that for a sub-horizon Fourier mode, the response is R(2)k S GW R,k ω 2S [1 + O(H/ωS)] and distinguish this from a homogeneous Hubble shift.

Abstract

We construct the averaged energy--momentum tensor of high-frequency gravitational waves, in the sense of Isaacson, for a general function of the Ricci scalar and the Gauss--Bonnet invariant on de Sitter, extending a recent f(R) result. At linear order, every such theory behaves like f(R) with a renormalised second derivative. At second order, tensor waves carry the Isaacson stress times the effective Planck mass, while scalar waves pick up an unsuppressed, traceless and non-null Gauss--Bonnet term. In the degenerate classes F(R+λG) and R-2Λ+F(G), its size does not depend on F. For an isotropic ensemble, it lowers the energy that scalar waves contribute to the background by a factor between 1/3 and 1; for a single beam, the averaged energy violates the dominant energy condition. In wave packets, conservation fixes an accompanying gradient term. Both terms are gauge invariant at leading order, and the main result is confirmed in four ways: two independent implementations of the field equations, an exact special case and a direct variation of the action. Off de Sitter, theories with a non-degenerate Hessian have no high-frequency regime. The degenerate class keeps the de Sitter structure with ordinary slow-roll corrections. Still, stability of its scalar mode in the radiation and matter eras makes the Gauss--Bonnet correction negligible at late times.

Sources

Related papers