betaFeedback

WorksPhysics and Astronomy

Metric perturbations in Einstein-Hilbert gravity: An alternative perturbative formulation

Abstract

We derive the ordinary gauge-fixed quadratic operators of the four-dimensional Einstein–Hilbert action in exponential and inverse-metric-density variables on an arbitrary reference background. With a single nonlinear background-harmonic gauge transported through the exact field map, the density operator is the trace-reversed Lichnerowicz operator, while the exponential operator contains an additional algebraic Ricci-curvature term. This difference arises from the nonlinear transformation of the background tadpole and vanishes on Ricci-flat backgrounds; a cosmological-constant check gives the corresponding cancellation at Einstein saddles. The construction retains the exact determinant contribution and specifies the associated Faddeev–Popov operator. At the first interacting order on a flat background, we show that the two gauge-fixed cubic actions differ by a term proportional to the free field equation, and independently evaluate the same nonzero three-point coefficient in both variables, with its normalization matched to the Einstein-gravity interaction. An exact Schwarzschild configuration satisfies the same background-harmonic gauge and is stationary under unrestricted exterior variations of the gauge-fixed action. Its density and exponential reconstructions are checked together with the bulk and boundary terms; a separate radial variation illustrates the loss of an equation after imposing a restricted ansatz. These results provide an explicit comparison of established field coordinates at the classical and tree levels; quantum loop calculations additionally require a common measure, regulator and source prescription.

Introduction

The choice of metric variables determines how the infinitely many interactions of Einstein–Hilbert gravity appear in a perturbative calculation. Nonlinear field redefinitions reorganize these interactions, but a comparison on a curved reference background must also retain the first variation of the action and specify how the gauge functional is transported. We study this comparison for the trace-reversed exponential variable vv and the inverse-metric-density fluctuation σ1\sigma_1 in four-dimensional Einstein gravity, taking the cosmological constant to vanish except for a short quadratic consistency check.

The density representation has a long history [1]. Grishchuk, Petrov and Popova developed an exact field formulation on an arbitrary reference spacetime [2]. Cheung and Remmen obtained simplified action representations, including a cubic formulation with an auxiliary field and tree-level recursion relations; their construction also covers arbitrary curved backgrounds [3, 4]. Cho, Kim and Lee derived general-order expansions in density and metric variables, with amplitude checks and a discussion of curved-background extensions [5]. Gauge, measure and parametrization dependence also enter the off-shell one-loop analysis of Ohta, Percacci and Pereira [6]. Our focus is the explicit comparison of the curved quadratic operators and flat cubic interactions for the two specified coordinates using one transported gauge. We keep the density tadpole, exact determinant term and ghost transformation together in this comparison. The field variables, the existence of background formulations and the general field-redefinition equivalence are established ingredients. Earlier simplifications of the Einstein–gravity vertices also provide an external benchmark for the cubic normalization [7].

We use a second-order formalism in which the action is expressed entirely through the metric and its derivatives. The reference metric g0μνg_{0\mu\nu} need not be flat or satisfy the vacuum equations. In the density coordinate the exact action separates the background action, a term linear in σ1\sigma_1 and a nonlinear derivative contribution. This is a decomposition of one metric action.

The usual additive metric fluctuation is defined by

gμν=g0μν+γμν.(1)g_{\mu\nu}=g_{0\mu\nu}+\gamma_{\mu\nu}. \tag*{(1)}

The fields vv and σ1\sigma_1 also vanish at the reference configuration, but their nonlinear relation to γ\gamma must be retained. Their exact definitions are given in Sec. II. The alternative formulation in the title refers to these coordinates for Einstein gravity.

The central quadratic comparison can be stated directly. For the background-divergence gauge at unit gauge parameter, write the degree-two action as (2κ2)−1⟨u,Hu⟩(2\kappa^2)^{-1}\langle u,\mathcal{H}u\rangle, where κ2=32πG\kappa^2=32\pi G and the pairing uses g0g_0. We obtain

Hσ1=PΔL,Hv=PΔL−A.(2)\mathcal{H}_{\sigma_1}=P\Delta_L,\qquad\mathcal{H}_{v}=P\Delta_L-A. \tag*{(2)}

Here PP is four-dimensional trace reversal and AA is the algebraic Ricci operator,

(Pu)μν=uμν−12g0μνg0ρλuρλ,(Au)μν=R0μρuρν+R0νρuρμ.(3)(Pu)_{\mu\nu}=u_{\mu\nu}-\frac{1}{2}g_{0\mu\nu}g_0^{\rho\lambda}u_{\rho\lambda},\qquad(Au)_{\mu\nu}=R_{0\mu}{}^\rho u_{\rho\nu}+R_{0\nu}{}^\rho u_{\rho\mu}. \tag*{(3)}

and ΔL\Delta_L is the background Lichnerowicz operator defined in Eq. (29). The −A-A contribution is the ordinary second-derivative chain rule: the quadratic part of the field map multiplies the background tadpole. It vanishes on a Ricci-flat background. The operators in Eq. (2) are the quadratic forms of the same gauge-fixed action in the two coordinates; no background field equation has been imposed in deriving them. Section V B includes the cosmological volume term and shows how the same tadpole argument gives equality at an Einstein saddle.

The comparison extends to the first interacting order on a flat background. Section V displays the identity relating the cubic actions through the free field equation, including the nonlinear pullback of the gauge term. Appendix B expands both cubic interactions directly, evaluates their nonzero three-point coefficients and matches the result to the published Einstein-gravity interaction with common external-state conventions. The Schwarzschild construction tests exact metric reconstruction beyond a weak-field expansion [8]. Its harmonic representative satisfies the same nonlinear gauge, and the full exterior Ricci tensor, density action and boundary current are evaluated explicitly. The two checks address the perturbative interaction and a finite classical stationary configuration of the gauge-fixed action, respectively.

The resulting formulation supplies compatible curved-background quadratic operators, an explicit ghost sector and flat cubic interactions in both coordinates. It also identifies the terms that would be lost by dropping the tadpole or changing the nonlinear gauge independently in the two schemes. The calculation is classical and tree-level. At the functional-integral level, the coordinate Jacobian, measure and external sources must be transported consistently; a regulated loop calculation requires additional choices not made here.

The paper is organized as follows. Section II gives the exact action and field definitions, and Sec. III identifies the nonlinear background-harmonic gauge and fixes the functional-integral conventions. Sections IV and V derive the two quadratic operators and the cubic comparison. Section VI presents the harmonic Schwarzschild configuration and the independent radial variation, followed by the conclusions in Sec. VII. Appendices A–C contain the ghost derivation, detailed cubic calculation and radial curvature and variation identities, respectively.

Einstein–Hilbert Lagrangian in curved space-time in terms of the background metric and perturbation field

We begin with the gravitational action with vanishing cosmological constant,

S=−14χ∫d4x−g gμνRμν,χ=4πG,κ2=8χ=32πG.(4)S=-\frac{1}{4\chi}\int\mathrm{d}^4x\sqrt{-g}\,g^{\mu\nu}R_{\mu\nu},\qquad\chi=4\pi G,\qquad\kappa^2=8\chi=32\pi G. \tag*{(4)}

We use units c=ℏ=1c=\hbar=1, signature (+−−−)(+---), and

Rμν=∂ρΓμνρ−∂νΓμρρ+ΓρλρΓμνλ−ΓνλρΓμρλ.(5)R_{\mu\nu}=\partial_\rho\Gamma^\rho_{\mu\nu}-\partial_\nu\Gamma^\rho_{\mu\rho}+\Gamma^\rho_{\rho\lambda}\Gamma^\lambda_{\mu\nu}-\Gamma^\rho_{\nu\lambda}\Gamma^\lambda_{\mu\rho}. \tag*{(5)}

All spacetime indices are Greek. Background raising and lowering and the derivative ∇≡∇(0)\nabla\equiv\nabla^{(0)} refer to g0g_0, which is held fixed unless explicitly stated otherwise. Integrations by parts in the bulk assume compactly supported variations or boundary conditions with the appropriate gravitational boundary action.

The exponential metric variable hμνh_{\mu\nu} is defined by

gμν=g0μρ(eh)ρν,hμν=g0μρhρν,−g−g0=e12tr⁡h.(6)g_{\mu\nu}=g_{0\mu\rho}(e^h)^\rho{}_\nu,\qquad h^\mu{}_\nu=g_0^{\mu\rho}h_{\rho\nu},\qquad\frac{\sqrt{-g}}{\sqrt{-g_0}}=e^{\frac{1}{2}\operatorname{tr}h}. \tag*{(6)}

Thus hh agrees with γ\gamma only to first order. Introduce its trace reversal and the normalized inverse metric density,

vμν=hμν−12g0μνtr⁡h,Vμν=vμρg0ρν,v_{\mu\nu}=h_{\mu\nu}-\frac{1}{2}g_{0\mu\nu}\operatorname{tr}h,\qquad V^\mu{}_\nu=v^{\mu\rho}g_{0\rho\nu},
ω=−g−g0,σμν=ωgμν=g0μν+σ1μν=(e−V)μρg0ρν.(7)\omega=\frac{\sqrt{-g}}{\sqrt{-g_0}},\qquad\sigma^{\mu\nu}=\omega g^{\mu\nu}=g_0^{\mu\nu}+\sigma_1^{\mu\nu}=(e^{-V})^\mu{}_\rho g_0^{\rho\nu}. \tag*{(7)}

The volume ratio makes σμν\sigma^{\mu\nu} a tensor under simultaneous coordinate transformations of the full and background metrics. The inverse matrix is denoted by τμν\tau_{\mu\nu}:

τμρσρν=δμν,τμν=gμρ0(eV)ρν,τ1μν=τμν−gμν0.(8)\tau_{\mu\rho}\sigma^{\rho\nu}=\delta^\nu_\mu,\qquad\tau_{\mu\nu}=g^0_{\mu\rho}(e^V)^\rho{}_\nu,\qquad\tau_{1\mu\nu}=\tau_{\mu\nu}-g^0_{\mu\nu}. \tag*{(8)}

The inverse-density shift τ1\tau_1 is not the background lowering σ1μν=gμρ0gνλ0σ1ρλ\sigma_{1\mu\nu}=g^0_{\mu\rho}g^0_{\nu\lambda}\sigma_1^{\rho\lambda}. This distinction is necessary at every nonlinear order.

Set

Qμν=σ1μρgρν0=I+X,Xμν=σ1μρgρν0,H=ln⁡det⁡Q.(9)Q^\mu{}_\nu=\sigma_1^{\mu\rho}g^0_{\rho\nu}=\mathbb{I}+X,\qquad X^\mu{}_\nu=\sigma_1^{\mu\rho}g^0_{\rho\nu},\qquad H=\ln\det Q. \tag*{(9)}

In four dimensions,

det⁡Q=ω2,H=−tr⁡V=tr⁡h,gμν=eH/2τμν.(10)\det Q=\omega^2,\qquad H=-\operatorname{tr}V=\operatorname{tr}h,\qquad g_{\mu\nu}=e^{H/2}\tau_{\mu\nu}. \tag*{(10)}

We use the positive volume-ratio branch with nonsingular σ\sigma of Lorentzian inertia. The exponential and logarithm are mutually inverse in a sufficiently small neighborhood of the background; no global exponential chart for all Lorentzian metrics is assumed. The exact density reconstruction can be used for finite configurations on its nonsingular domain. The expansions relating the two schemes are

σ1μν=−vμν+12vμρgρλ0vλν+O(v3),\sigma_1^{\mu\nu}=-v^{\mu\nu}+\frac{1}{2}v^{\mu\rho}g^0_{\rho\lambda}v^{\lambda\nu}+O(v^3),
τ1μν=−σ1μν+σ1μρσ1ρν+O(σ13).(11)\tau_{1\mu\nu}=-\sigma_{1\mu\nu}+\sigma_{1\mu\rho}\sigma_1^\rho{}_\nu+O(\sigma_1^3). \tag*{(11)}

Both are field redefinitions of the same metric. The inverse in Eq. (8), rather than a truncated series, defines a finite density configuration.

To write the action covariantly, introduce the difference of connections Cρμν=Γρμν[g]−Γρμν[g0]C^\rho{}_{\mu\nu}=\Gamma^\rho{}_{\mu\nu}[g]-\Gamma^\rho{}_{\mu\nu}[g_0]. Direct substitution of Eq. (10) gives

Cρμν=12σρλ(∇μτλν+∇ντλμ−∇λτμν)+14(δμρHν+δνρHμ−σρλτμνHλ),Hμ=∂μH.(12)C^\rho{}_{\mu\nu}=\frac{1}{2}\sigma^{\rho\lambda}(\nabla_\mu\tau_{\lambda\nu}+\nabla_\nu\tau_{\lambda\mu}-\nabla_\lambda\tau_{\mu\nu}) +\frac{1}{4}(\delta^\rho_\mu H_\nu+\delta^\rho_\nu H_\mu-\sigma^{\rho\lambda}\tau_{\mu\nu}H_\lambda),\qquad H_\mu=\partial_\mu H. \tag*{(12)}

The steps in reducing the action are useful for fixing the sign of the determinant term. The curvature-difference identity is

Rμν[g]=Rμν0+∇ρCρμν−∇νCρμρ+CρρλCλμν−CρνλCλμρ.(13)R_{\mu\nu}[g]=R^0_{\mu\nu}+\nabla_\rho C^\rho{}_{\mu\nu}-\nabla_\nu C^\rho{}_{\mu\rho}+C^\rho{}_{\rho\lambda}C^\lambda{}_{\mu\nu}-C^\rho{}_{\nu\lambda}C^\lambda{}_{\mu\rho}. \tag*{(13)}

Metric compatibility of the full metric and ∂ρln⁡ω=Cλρλ\partial_\rho\ln\omega=C^\lambda{}_{\rho\lambda} imply

∇ρσμν=−Cμρλσλν−Cνρλσμλ+Cλρλσμν,Cλρλ=12Hρ.(14)\nabla_\rho\sigma^{\mu\nu}=-C^\mu{}_{\rho\lambda}\sigma^{\lambda\nu}-C^\nu{}_{\rho\lambda}\sigma^{\mu\lambda}+C^\lambda{}_{\rho\lambda}\sigma^{\mu\nu},\qquad C^\lambda{}_{\rho\lambda}=\frac{1}{2}H_\rho. \tag*{(14)}

Multiplying Eq. (13) by σμν\sigma^{\mu\nu} and integrating its two derivative terms by parts gives a divergence and a quadratic connection expression. In particular, the derivatives acting on σ\sigma are evaluated with Eq. (14); they must not be discarded. The result is

ωR[g]=Lσ+∇ρWρ,Wρ=σμνCρμν−σμρCνμν.(15)\omega R[g]=\mathcal{L}_\sigma+\nabla_\rho W^\rho,\qquad W^\rho=\sigma^{\mu\nu}C^\rho{}_{\mu\nu}-\sigma^{\mu\rho}C^\nu{}_{\mu\nu}. \tag*{(15)}

where the bulk Lagrangian is [3, 5]

Lσ=σμνRμν0+(∇μσαβ)(∇ντρβ)(14σμνδαρ−12σμρδαν)+18σμνHμHν.(16)\mathcal{L}_\sigma=\sigma^{\mu\nu}R^0_{\mu\nu}+(\nabla_\mu\sigma^{\alpha\beta})(\nabla_\nu\tau_{\rho\beta})\left(\frac{1}{4}\sigma^{\mu\nu}\delta^\rho_\alpha-\frac{1}{2}\sigma^{\mu\rho}\delta^\nu_\alpha\right)+\frac{1}{8}\sigma^{\mu\nu}H_\mu H_\nu. \tag*{(16)}

For clarity, the connection expression and its reduction to density derivatives are

Lder≡σμν(CρμρCλνλ−CρμλCλνρ)=−14σμνταρτβλ(∇μσαβ)(∇νσρλ)+12ταρ(∇λσαβ)(∇βσλρ)+18σμνHμHν.(17)\begin{aligned} \mathcal{L}_{\mathrm{der}}\equiv\sigma^{\mu\nu}(C^\rho{}_{\mu\rho}C^\lambda{}_{\nu\lambda}-C^\rho{}_{\mu\lambda}C^\lambda{}_{\nu\rho}) \\ &=-\frac{1}{4}\sigma^{\mu\nu}\tau_{\alpha\rho}\tau_{\beta\lambda}(\nabla_\mu\sigma^{\alpha\beta})(\nabla_\nu\sigma^{\rho\lambda})+\frac{1}{2}\tau_{\alpha\rho}(\nabla_\lambda\sigma^{\alpha\beta})(\nabla_\beta\sigma^{\lambda\rho})+\frac{1}{8}\sigma^{\mu\nu}H_\mu H_\nu. \tag*{(17)} \end{aligned}

This follows by inserting Eq. (12) and using

∇μταβ=−ταρτβλ∇μσρλ,Hμ=ταβ∇μσαβ.(18)\nabla_\mu\tau_{\alpha\beta}=-\tau_{\alpha\rho}\tau_{\beta\lambda}\nabla_\mu\sigma^{\rho\lambda},\qquad H_\mu=\tau_{\alpha\beta}\nabla_\mu\sigma^{\alpha\beta}. \tag*{(18)}

In the second term of Eq. (17), one inverse-density factor has canceled against σ\sigma. Reversing that cancellation and using the first identity in Eq. (18) gives precisely the two derivative contractions in Eq. (16). This derivation fixes the bulk action together with its boundary current.

In particular, 12σμν∂μln⁡ω ∂νln⁡ω=18σμνHμHν\frac{1}{2}\sigma^{\mu\nu}\partial_{\mu}\ln\omega\,\partial_{\nu}\ln\omega=\frac{1}{8}\sigma^{\mu\nu}H_{\mu}H_{\nu}. The coefficient is 1/8 when H=tr⁡hH=\operatorname{tr}h, not 1/2. A boundary term cannot independently reverse this sign. For g=e2ψg0g=e^{2\psi}g_{0}, Eq. (16) reduces to e2ψ[R0+6(∇ψ)2]e^{2\psi}[R_{0}+6(\nabla\psi)^{2}], consistently with the Einstein–Hilbert conformal transformation after integration by parts.

The background split now takes the simple form

Sbulk=S[g0]−14χ∫d4x−g0[σ1μνR0μν+Lder[σ1;g0]].(19)S_{\mathrm{bulk}}=S[g_{0}]-\frac{1}{4\chi}\int d^{4}x\sqrt{-g_{0}}\left[\sigma_{1}^{\mu\nu}R_{0\mu\nu}+\mathcal{L}_{\mathrm{der}}[\sigma_{1};g_{0}]\right]. \tag*{(19)}

Here Lder\mathcal{L}_{\mathrm{der}} is the derivative part of Eq. (16). It begins quadratically in σ1\sigma_{1} and contains infinitely many interactions through τ\tau and HH. The linear term is the background tadpole. Covariant derivative commutators still supply curvature terms in the quadratic action. Equation (19) separates contributions of a single metric action; neither the nonlinear derivative terms nor a coordinate-dependent part of them constitute a separate free classical theory.

Single-variable action, gauge fixing and functional determinants

The trace-dependent term in Eq. (16) can already be written entirely in terms of the density fluctuation:

H=ln⁡det⁡(I+σ1g0),∂μH=tr⁡[(I+X)−1∇μX].(20)H=\ln\det(\mathbb{I}+\sigma_{1}g_{0}),\qquad\partial_{\mu}H=\operatorname{tr}\left[(\mathbb{I}+X)^{-1}\nabla_{\mu}X\right]. \tag*{(20)}

Thus the action is a single-variable functional before any gauge fixing. Alternatively, inserting σ1(v)\sigma_{1}(v) gives its exponential representation. Since vv and σ1\sigma_{1} are related by Eq. (7), they cannot be integrated over as independent variables to cancel the trace term. An auxiliary identity does not justify deleting a field-dependent contribution from the action.

For perturbation theory choose one four-component gauge condition,

Fμ[σ1]=∇νσ1μν,SGF=18χα∫d4x−g0 g0μνFμFν.(21)F^{\mu}[\sigma_{1}]=\nabla_{\nu}\sigma_{1}^{\mu\nu},\qquad S_{\mathrm{GF}}=\frac{1}{8\chi\alpha}\int d^{4}x\sqrt{-g_{0}}\,g_{0\mu\nu}F^{\mu}F^{\nu}. \tag*{(21)}

The linear metric relation is γμν=−σ1μν+12g0μνσ1+O(σ12)\gamma_{\mu\nu}=-\sigma_{1\mu\nu}+\frac{1}{2}g_{0\mu\nu}\sigma_{1}+O(\sigma_{1}^{2}), where σ1=g0μνσ1μν\sigma_{1}=g_{0\mu\nu}\sigma_{1}^{\mu\nu}. Hence FF is the de Donder condition at linear order, up to an overall sign. We use α=1\alpha=1 below. No trace constraint or second gauge condition is imposed. In the first perturbative scheme the same gauge functional means F[σ1(v)]F[\sigma_{1}(v)], including its nonlinear terms.

The geometric meaning of this condition is exact. With the connection difference CC of Eq. (12), metric compatibility and ∂νln⁡ω=Cρνρ\partial_{\nu}\ln\omega=C^{\rho}_{\rho\nu} imply

Fμ=∇νσμν=−ωgρλCρλμ=−ωgρλ(Γρλμ[g]−Γρλμ[g0]).(22)F^{\mu}=\nabla_{\nu}\sigma^{\mu\nu}=-\omega g^{\rho\lambda}C^{\mu}_{\rho\lambda}=-\omega g^{\rho\lambda}\left(\Gamma^{\mu}_{\rho\lambda}[g]-\Gamma^{\mu}_{\rho\lambda}[g_{0}]\right). \tag*{(22)}

Indeed, contracting Eq. (14) cancels its connection-trace terms. Since ∇g0−1=0\nabla g_{0}^{-1}=0, the divergences of σ\sigma and σ1\sigma_{1} coincide. Thus F=0F=0 is the nonlinear background-harmonic (wave-map) condition relative to g0g_{0}. For a flat Cartesian reference metric it becomes

∂ν(−g gμν)=0⟺gρλΓρλμ[g]=0⟺□gxμ=0.(23)\partial_{\nu}\left(\sqrt{-g}\,g^{\mu\nu}\right)=0\quad\Longleftrightarrow\quad g^{\rho\lambda}\Gamma^{\mu}_{\rho\lambda}[g]=0\quad\Longleftrightarrow\quad\Box_{g}x^{\mu}=0. \tag*{(23)}

This is the established harmonic-coordinate condition for the gothic inverse metric [9]. Equation (22) is covariant under simultaneous coordinate transformations of gg and g0g_{0}; the middle expression in Eq. (23) alone is a condition on a particular coordinate chart. The reference connection must be retained in general coordinates. We keep precisely the gauge-fixing weight in Eq. (21): replacing FμF^{\mu} by −Fμ/ω-F^{\mu}/\omega preserves its zero set but changes the nonlinear gauge-fixing action and the off-shell ghost operator.

The Faddeev–Popov operator is defined by varying this gauge under a quantum diffeomorphism at fixed background [1, 10]:

(M[σ1]ξ)μ=δξFμ[σ1],M[0]μρ=−□0δρμ−R0μρ.(24)(\mathcal{M}[\sigma_{1}]\xi)^{\mu}=\delta_{\xi}F^{\mu}[\sigma_{1}],\qquad\mathcal{M}[0]^{\mu}{}_{\rho}=-\Box_{0}\delta^{\mu}_{\rho}-R_{0}^{\mu}{}_{\rho}. \tag*{(24)}

Here □0=g0μν∇μ∇ν\Box_{0}=g_{0}^{\mu\nu}\nabla_{\mu}\nabla_{\nu}. Appendix Appendix A gives the full operator, its ghost action and the first interaction explicitly. Inverses require boundary or Feynman prescriptions. Residual transformations and background isometries, if present, must be treated separately; no globally unique gauge representative is assumed.

Local field redefinitions, Jacobian factors, and ghost determinants

For a chosen starting measure μ[σ1]Dσ1\mu[\sigma_1]\mathcal{D}\sigma_1, the formal coordinate change gives

Z[J]=∫DvDcˉDc μ[σ1(v)]J[v]exp⁡{i(S[σ1(v)]+SGF[σ1(v)]+Sgh[σ1(v),cˉ,c]+J⋅σ1(v))},Z[J]=\int\mathcal{D}v\mathcal{D}\bar{c}\mathcal{D}c\,\mu[\sigma_1(v)]\mathcal{J}[v]\exp\left\{i\left(S[\sigma_1(v)]+S_{\mathrm{GF}}[\sigma_1(v)]+S_{\mathrm{gh}}[\sigma_1(v),\bar{c},c]+J\cdot\sigma_1(v)\right)\right\},
J[v]=∣Det⁡Symδσ1δv∣.(25)\mathcal{J}[v]=\left|\operatorname{Det}_{\mathrm{Sym}}\frac{\delta\sigma_1}{\delta v}\right|. \tag*{(25)}

The equality presumes the same regulated integration problem, with its source and any defining restrictions transported. The map is local and has derivative −I-\mathrm{I} on symmetric components at the background. Its functional Jacobian nevertheless needs a measure and regularization prescription. It is not the Faddeev–Popov determinant, which acts on vector gauge parameters. Replacing J⋅σ1(v)J\cdot\sigma_1(v) by a source linear in vv changes the off-shell functional. These distinctions matter for loop calculations and off-shell effective actions [6]; they do not alter the classical field redefinition or the tree-level comparison made here. We do not evaluate a regulated functional determinant in this paper.

First perturbative scheme: perturbative action in terms of the vμνv^{\mu\nu} field

In the first scheme the independent symmetric field is vμνv^{\mu\nu}. The action is obtained by substituting σ=e−Vg0−1\sigma=e^{-V}g_0^{-1}, τ=g0eV\tau=g_0e^V and H=−tr⁡VH=-\operatorname{tr}V into Eq. (16). The gauge and ghost actions are pulled back through the same map. The series needed through cubic order is

σ1μν=−vμν+12(vg0v)μν−16(vg0vg0v)μν+O(v4),\sigma_1^{\mu\nu}=-v^{\mu\nu}+\frac{1}{2}(vg_0v)^{\mu\nu}-\frac{1}{6}(vg_0vg_0v)^{\mu\nu}+O(v^4),
τ1μν=vμν+12(vg0−1v)μν+16(vg0−1vg0−1v)μν+O(v4).(26)\tau_{1\mu\nu}=v_{\mu\nu}+\frac{1}{2}(vg_0^{-1}v)_{\mu\nu}+\frac{1}{6}(vg_0^{-1}vg_0^{-1}v)_{\mu\nu}+O(v^4). \tag*{(26)}

Matrix products in the first line use contravariant vv, and those in the second use its background lowering. In particular the trace term in the action remains σμν∂μtr⁡V∂νtr⁡V/8\sigma^{\mu\nu}\partial_\mu\operatorname{tr}V\partial_\nu\operatorname{tr}V/8.

Linear and quadratic terms on a curved reference metric

Write S(n)S^{(n)} for the term homogeneous of degree nn, including its Taylor factor. The term linear in vv is

Sv(1)=14χ∫d4x−g0 vμνR0μν.(27)S_v^{(1)}=\frac{1}{4\chi}\int\mathrm{d}^4x\sqrt{-g_0}\,v^{\mu\nu}R_{0\mu\nu}. \tag*{(27)}

It vanishes for a vacuum reference metric. An expansion about a general background retains this term; such a background is not an unsourced stationary point. For a symmetric tensor uμνu_{\mu\nu} define

(Pu)μν=uμν−12g0μνu,u=g0ρλuρλ,⟨u,w⟩=∫d4x−g0 uμνwμν.(28)(Pu)_{\mu\nu}=u_{\mu\nu}-\frac{1}{2}g_{0\mu\nu}u,\qquad u=g_0^{\rho\lambda}u_{\rho\lambda},\qquad\langle u,w\rangle=\int\mathrm{d}^4x\sqrt{-g_0}\,u^{\mu\nu}w_{\mu\nu}. \tag*{(28)}

In four dimensions P2=IP^2=\mathrm{I}. With the curvature convention of Eq. (5), equivalently [∇μ,∇ν]aρ=R0ρλμνaλ[\nabla_\mu,\nabla_\nu]a^\rho=R_0{}^\rho{}_{\lambda\mu\nu}a^\lambda, the Lichnerowicz operator is

(ΔLu)μν=−□0uμν−2R0μρνσuρσ+R0μρuρν+R0νρuρμ.(29)(\Delta_Lu)_{\mu\nu}=-\Box_0u_{\mu\nu}-2R_{0\mu\rho\nu\sigma}u^{\rho\sigma}+R_{0\mu}{}^\rho u_{\rho\nu}+R_{0\nu}{}^\rho u_{\rho\mu}. \tag*{(29)}

Trace reversal commutes with this operator. At α=1\alpha=1 the gauge-fixed quadratic action is

(S+SGF)v(2)=116χ⟨v,PΔLv⟩−18χ∫d4x−g0 R0μνvμρg0ρλvλν=116χ⟨v,Hvv⟩,Hv=PΔL−A,\begin{aligned} (S+S_{\mathrm{GF}})_v^{(2)}=\frac{1}{16\chi}\langle v,P\Delta_Lv\rangle-\frac{1}{8\chi}\int\mathrm{d}^4x\sqrt{-g_0}\,R_{0\mu\nu}v^{\mu\rho}g_{0\rho\lambda}v^{\lambda\nu} \\ &=\frac{1}{16\chi}\langle v,\mathcal{H}_vv\rangle,\qquad\mathcal{H}_v=P\Delta_L-A, \end{aligned}
(Au)μν=R0μρuρν+R0νρuρμ.(30)(Au)_{\mu\nu}=R_{0\mu}{}^\rho u_{\rho\nu}+R_{0\nu}{}^\rho u_{\rho\mu}. \tag*{(30)}

The first term follows from the quadratic derivative action, derived in Sec. V; the second follows from the quadratic term in σ1(v)\sigma_1(v) multiplying R0μνR_{0\mu\nu} in Eq. (19). It would be lost by dropping the density tadpole before changing variables. Equivalently, after integration by parts,

(S+SGF)v(2)=116χ∫d4x−g0[(∇λvμν)(∇λvμν)−12(∇v)(∇v)−2R0μρνσvμνvρσ].(31)(S+S_{\mathrm{GF}})^{(2)}_v=\frac{1}{16\chi}\int d^4x\sqrt{-g_0}\left[(\nabla_\lambda v_{\mu\nu})(\nabla^\lambda v^{\mu\nu})-\frac{1}{2}(\nabla v)(\nabla v)-2R_{0\mu\rho\nu\sigma}v^{\mu\nu}v^{\rho\sigma}\right]. \tag*{(31)}

Here v=g0μνvμνv=g_{0\mu\nu}v^{\mu\nu}. The Ricci contributions in ΔL\Delta_L cancel the explicit term inherited from the tadpole, but the Riemann-curvature coupling remains. Equations (30)–(31) are background-covariant quadratic forms, rather than a flat-space replacement of a curved kinetic term.

Propagator and interactions

The inverse of the quadratic form, with an appropriate boundary prescription, determines the propagator. In a Cartesian flat reference metric, Hv=P(−□)\mathcal{H}_v=P(-\square) and the Feynman propagator for the unnormalized field is

⟨vμν(p)vρσ(q)⟩0=(2π)4δ(4)(p+q)8χip2+i0Pμν,ρσ,Pμν,ρσ=12(ημρηνσ+ημσηνρ−ημνηρσ).(32)\langle v^{\mu\nu}(p)v^{\rho\sigma}(q)\rangle_0=(2\pi)^4\delta^{(4)}(p+q)\frac{8\chi i}{p^2+i0}P^{\mu\nu,\rho\sigma}, \qquad P^{\mu\nu,\rho\sigma}=\frac{1}{2}\left(\eta^{\mu\rho}\eta^{\nu\sigma}+\eta^{\mu\sigma}\eta^{\nu\rho}-\eta^{\mu\nu}\eta^{\rho\sigma}\right). \tag*{(32)}

The Fourier convention is v(x)=∫d4p(2π)4v(p)eip⋅xv(x)=\int\frac{d^4p}{(2\pi)^4}v(p)e^{ip\cdot x}. The trace term in PP is necessary to invert the Einstein–Hilbert quadratic form. Rescaling vv by κ=8χ\kappa=\sqrt{8\chi} gives the usual canonical normalization. On a curved background the inverse is instead that of Hv\mathcal{H}_v, with its curvature terms and prescribed boundary conditions; no translation-invariant momentum formula is implied.

The interaction terms follow by expanding the complete action, including the trace term and the pullback F[σ1(v)]F[\sigma_1(v)] of the gauge functional. The latter is nonlinear in vv even though F[σ1]F[\sigma_1] is linear in σ1\sigma_1. The same substitution generates the ghost interactions from Appendix A. Appendix B displays the cubic comparison and an explicit three-point check. This construction reorganizes the interactions of Einstein gravity without introducing an additional field.

Second perturbative scheme: perturbative action in terms of the σ1μν\sigma_1^{\mu\nu} field

In the second scheme σ1μν\sigma_1^{\mu\nu} is independent, and the full action is Eq. (16) with

τ=(g0−1+σ1)−1,H=ln⁡det⁡(I+X),X=σ1g0.(33)\tau=(g_0^{-1}+\sigma_1)^{-1},\qquad H=\ln\det(\mathbb{I}+X),\qquad X=\sigma_1g_0. \tag*{(33)}

The perturbative series is

τ1μν=−σ1μν+σ1μρσ1ρν−σ1μρσ1ρλσ1λν+O(σ14),H=tr⁡X−12tr⁡X2+13tr⁡X3+O(X4).(34)\tau_{1\mu\nu}=-\sigma_{1\mu\nu}+\sigma_{1\mu\rho}\sigma_1{}^\rho{}_\nu-\sigma_{1\mu\rho}\sigma_1{}^\rho{}_\lambda\sigma_1{}^\lambda{}_\nu+O(\sigma_1^4), \qquad H=\operatorname{tr}X-\frac{1}{2}\operatorname{tr}X^2+\frac{1}{3}\operatorname{tr}X^3+O(X^4). \tag*{(34)}

The inverse and logarithm in Eq. (33) define the exact functional. A truncation of Eq. (34) defines only a chosen perturbative order.

Quadratic action and its relation to the first scheme

The first variation can be obtained directly from the inverse-metric variation. At fixed background,

δσμν=ω(δgμν−12gμνgρλδgρλ),δS=−14χ∫d4x−g0 Rμν[g]δσμν.(35)\delta\sigma^{\mu\nu}=\omega\left(\delta g^{\mu\nu}-\frac{1}{2}g^{\mu\nu}g_{\rho\lambda}\delta g^{\rho\lambda}\right),\qquad\delta S=-\frac{1}{4\chi}\int d^4x\sqrt{-g_0}\,R_{\mu\nu}[g]\delta\sigma^{\mu\nu}. \tag*{(35)}

To see the cancellation of the scalar-curvature term explicitly, take the trace with τμν\tau_{\mu\nu} and invert the first identity:

δgμν=ω−1(δσμν−12σμντρλδσρλ).(36)\delta g^{\mu\nu}=\omega^{-1}\left(\delta\sigma^{\mu\nu}-\frac{1}{2}\sigma^{\mu\nu}\tau_{\rho\lambda}\delta\sigma^{\rho\lambda}\right). \tag*{(36)}

Since Gρρ=−RG^\rho{}_\rho=-R in four dimensions, the Einstein–Hilbert variation becomes

−gGμνδgμν=−g0(Gμν−12gμνGρρ)δσμν=−g0Rμνδσμν.(37)\sqrt{-g}G_{\mu\nu}\delta g^{\mu\nu}=\sqrt{-g_0}\left(G_{\mu\nu}-\frac{1}{2}g_{\mu\nu}G^\rho{}_\rho\right)\delta\sigma^{\mu\nu}=\sqrt{-g_0}R_{\mu\nu}\delta\sigma^{\mu\nu}. \tag*{(37)}

This proves the second identity in Eq. (35). At the reference configuration,

Sσ1(1)=−14χ∫d4x−g0 σ1μνR0μν.(38)S_{\sigma_1}^{(1)}=-\frac{1}{4\chi}\int\mathrm{d}^4x\sqrt{-g_0}\,\sigma_1^{\mu\nu}R_{0\mu\nu}. \tag*{(38)}

For completeness the quadratic operator follows from the ordered Palatini identity for a metric variation pμνp_{\mu\nu}:

δRμν[p]=12(∇ρ∇μp νρ+∇ρ∇νp μρ−□0pμν−∇μ∇νp).(39)\delta R_{\mu\nu}[p]=\frac{1}{2}\left(\nabla_\rho\nabla_\mu p^\rho_{\ \nu}+\nabla_\rho\nabla_\nu p^\rho_{\ \mu}-\Box_0p_{\mu\nu}-\nabla_\mu\nabla_\nu p\right). \tag*{(39)}

Here p=g0μνpμνp=g_0^{\mu\nu}p_{\mu\nu}. The required commutator on a mixed tensor is

[∇ρ,∇μ]p νλ=R0λμλp νλ−R0νρμλp λρ.(40)[\nabla_\rho,\nabla_\mu]p^\lambda_{\ \nu}=R_{0\lambda\mu}{}^\lambda p^\lambda_{\ \nu}-R_{0\nu\rho\mu}{}^\lambda p^\rho_{\ \lambda}. \tag*{(40)}

Using it in the two ordered terms in Eq. (39) yields

δRμν[p]=12[(ΔLp)μν+∇μ∇ρpρν+∇ν∇ρpρμ−∇μ∇νp].(41)\delta R_{\mu\nu}[p]=\frac{1}{2}\left[(\Delta_Lp)_{\mu\nu}+\nabla_\mu\nabla^\rho p_{\rho\nu}+\nabla_\nu\nabla^\rho p_{\rho\mu}-\nabla_\mu\nabla_\nu p\right]. \tag*{(41)}

The linear density-to-metric relation p=−Pσ1p=-P\sigma_1 gives ∇ρpρν=−Fν+12∇νσ1\nabla^\rho p_{\rho\nu}=-F_\nu+\frac{1}{2}\nabla_\nu\sigma_1 and p=σ1p=\sigma_1. The scalar double derivatives therefore cancel. Moreover, ΔL(g0f)=−g0□0f\Delta_L(g_0f)=-g_0\Box_0f and tr⁡(ΔLu)=−□0tr⁡u\operatorname{tr}(\Delta_Lu)=-\Box_0\operatorname{tr}u, so that [P,ΔL]=0[P,\Delta_L]=0 on any reference background. Hence

δRμν[−Pσ1]=−12[(PΔLσ1)μν+∇μFν+∇νFμ].(42)\delta R_{\mu\nu}[-P\sigma_1]=-\frac{1}{2}\left[(P\Delta_L\sigma_1)_{\mu\nu}+\nabla_\mu F_\nu+\nabla_\nu F_\mu\right]. \tag*{(42)}

Consequently the quadratic Einstein–Hilbert action and its gauge-fixed form are

Sσ1(2)=116χ⟨σ1,PΔLσ1⟩−18χ⟨F,F⟩,S_{\sigma_1}^{(2)}=\frac{1}{16\chi}\langle\sigma_1,P\Delta_L\sigma_1\rangle-\frac{1}{8\chi}\langle F,F\rangle,
(S+SGF)σ1(2)=116χ⟨σ1,Hσ1σ1⟩,Hσ1=PΔL.(43)(S+S_{\mathrm{GF}})_{\sigma_1}^{(2)}=\frac{1}{16\chi}\langle\sigma_1,\mathcal{H}_{\sigma_1}\sigma_1\rangle,\qquad\mathcal{H}_{\sigma_1}=P\Delta_L. \tag*{(43)}

The pairing for FF is the corresponding vector pairing with g0g_0. As a check, direct expansion of the bulk Lagrangian gives

Lσ(2)=−14(∇μσαβ)(∇μσ1αβ)+12(∇μσαβ)(∇ασ1μβ)+18(∇μσ1)(∇μσ1).(44)\mathcal{L}_{\sigma}^{(2)}=-\frac{1}{4}(\nabla_\mu\sigma^{\alpha\beta})(\nabla^\mu\sigma_{1\alpha\beta})+\frac{1}{2}(\nabla_\mu\sigma^{\alpha\beta})(\nabla_\alpha\sigma_1{}^\mu{}_\beta)+\frac{1}{8}(\nabla_\mu\sigma_1)(\nabla^\mu\sigma_1). \tag*{(44)}

After integration by parts, the difference between the mixed derivative term and F2F^2 is −∫d4x−g0 σ1αβ[∇μ,∇α]σ1μβ-\int\mathrm{d}^4x\sqrt{-g_0}\,\sigma_1^{\alpha\beta}[\nabla_\mu,\nabla_\alpha]\sigma_1{}^\mu{}_\beta. Evaluating the commutator as in Eq. (40) gives

∫d4x−g0[(∇μσ1αβ)(∇ασ1μβ)−FμFμ]=∫d4x−g0[R0μρνσ1μρσ1νσ−R0μνσ1μρσ1ρ ν].(45)\int\mathrm{d}^4x\sqrt{-g_0}\left[(\nabla_\mu\sigma_1^{\alpha\beta})(\nabla_\alpha\sigma_1{}^\mu{}_\beta)-F_\mu F^\mu\right]=\int\mathrm{d}^4x\sqrt{-g_0}\left[R_{0\mu\rho\nu}\sigma_1^{\mu\rho}\sigma_1^{\nu\sigma}-R_{0\mu\nu} \sigma_1^{\mu\rho}\sigma_{1\rho}^{\ \nu}\right]. \tag*{(45)}

This identity supplies the curvature potential in Eq. (43) and explains why setting g0=ηg_0=\eta too early misses part of the comparison.

Equations (30) and (43) imply

Hv=Hσ1−A.(46)\mathcal{H}_v=\mathcal{H}_{\sigma_1}-A. \tag*{(46)}

The term −A-A is the ordinary second-derivative chain rule applied to a nonlinear change of field coordinates: the second derivative of the map multiplies the first derivative of the action. It vanishes on a Ricci-flat reference metric. The two quadratic forms then coincide after the linear identification σ1=−v\sigma_1=-v. For a general background they are different off-shell Hessians of the same action. The source and measure qualifications in Sec. III apply to a comparison of their generating functionals.

In flat space Eq. (43) gives the propagator (32), with vv replaced by σ1\sigma_1. Equality of the flat propagators is therefore exact, including the trace numerator and normalization. It is not a statement that arbitrary unsourced curved-background Hessians agree.

A cosmological-constant consistency check

The role of the tadpole can also be checked for

SΛ=−14χ∫d4x−g(R−2Λ).(47)S_\Lambda=-\frac{1}{4\chi}\int\mathrm{d}^4x\sqrt{-g}(R-2\Lambda). \tag*{(47)}

Only this subsection includes Λ\Lambda; the flat cubic and Schwarzschild calculations below have Λ=0\Lambda= 0. In four dimensions, with X=σ1g0X = \sigma_{1}g_{0} and V=vg0V = vg_{0}, the two volume expansions are

ω(σ1)=det⁡(I+X)=1+12tr⁡X+18(tr⁡X)2−14tr⁡X2+O(X3),ω(v)=e−tr⁡V/2=1−12tr⁡V+18(tr⁡V)2+O(V3).(48)\begin{aligned} \omega(\sigma_{1}) &= \sqrt{\det(\mathbb{I}+X)} = 1 + \frac{1}{2}\operatorname{tr}X + \frac{1}{8}(\operatorname{tr}X)^{2} - \frac{1}{4}\operatorname{tr}X^{2} + O(X^{3}),\\ \omega(v) &= e^{-\operatorname{tr}V/2} = 1 - \frac{1}{2}\operatorname{tr}V + \frac{1}{8}(\operatorname{tr}V)^{2} + O(V^{3}). \tag*{(48)} \end{aligned}

Writing dVˉ=d4x−g0\mathrm{d}\bar{V} = \mathrm{d}^{4}x\sqrt{-g_{0}}, the cosmological term therefore contributes

Scos,σ1(2)=Λ16X∫dVˉ[(tr⁡X)2−2tr⁡X2],Scos,v(2)=Λ16X∫dVˉ(tr⁡V)2.(49)S_{\mathrm{cos},\sigma_{1}}^{(2)} = \frac{\Lambda}{16\mathcal{X}}\int\mathrm{d}\bar{V}\left[(\operatorname{tr}X)^{2} - 2\operatorname{tr}X^{2}\right], \qquad S_{\mathrm{cos},v}^{(2)} = \frac{\Lambda}{16\mathcal{X}}\int\mathrm{d}\bar{V}(\operatorname{tr}V)^{2}. \tag*{(49)}

Define (Tu)μν=g0μνg0ρλuρλ(Tu)_{\mu\nu} = g_{0\mu\nu}g_{0}^{\rho\lambda}u_{\rho\lambda}, so P=I−T/2P = \mathbb{I} - T/2. With the same gauge and quadratic normalization as above,

Hσ1,Λ=P(ΔL−2ΛI),Hv,Λ=PΔL−A+ΛT.(50)\mathcal{H}_{\sigma_{1},\Lambda} = P(\Delta_{L} - 2\Lambda\mathbb{I}), \qquad\mathcal{H}_{v,\Lambda} = P\Delta_{L} - A + \Lambda T. \tag*{(50)}

The first variation in the density coordinate is now

Sσ1,Λ(1)=−14X∫dVˉ(R0μν−Λg0μν)σ1μν.(51)S_{\sigma_{1},\Lambda}^{(1)} = -\frac{1}{4\mathcal{X}}\int\mathrm{d}\bar{V}(R_{0\mu\nu} - \Lambda g_{0\mu\nu})\sigma_{1}^{\mu\nu}. \tag*{(51)}

Its pullback through Eq. (11) supplies the quadratic contribution

−18X∫dVˉ(R0μν−Λg0μν)(vg0v)μν=116X⟨v,(−A+2ΛI)v⟩.(52)-\frac{1}{8\mathcal{X}}\int\mathrm{d}\bar{V}(R_{0\mu\nu} - \Lambda g_{0\mu\nu})(vg_{0}v)^{\mu\nu} = \frac{1}{16\mathcal{X}}\langle v,(-A + 2\Lambda\mathbb{I})v\rangle. \tag*{(52)}

This agrees with the difference of the independently expanded operators,

Hv,Λ−Hσ1,Λ=−A+2ΛI.(53)\mathcal{H}_{v,\Lambda} - \mathcal{H}_{\sigma_{1},\Lambda} = -A + 2\Lambda\mathbb{I}. \tag*{(53)}

At an Einstein saddle R0μν=Λg0μνR_{0\mu\nu} = \Lambda g_{0\mu\nu}, one has A=2ΛIA = 2\Lambda\mathbb{I}, and both operators reduce to P(ΔL−2ΛI)P(\Delta_{L} - 2\Lambda\mathbb{I}). The cancellation is the expected second-variation chain rule at a stationary point, and extends the Ricci-flat consistency check without introducing a new equivalence principle.

Interactions and the flat cubic comparison

The density gauge term is exactly quadratic, whereas the inverse density and HH generate higher interactions in the gravitational action. The first interacting order makes the relation between the two schemes explicit. On a Cartesian flat background define

Uvμν=vμρηρλvλν,σ1μν=−vμν+12Uvμν+O(v3).(54)U_{v}^{\mu\nu} = v^{\mu\rho}\eta_{\rho\lambda}v^{\lambda\nu}, \qquad\sigma_{1}^{\mu\nu} = -v^{\mu\nu} + \frac{1}{2}U_{v}^{\mu\nu} + O(v^{3}). \tag*{(54)}

Let S3,vS_{3,v} and S3,σ1S_{3,\sigma_{1}} denote the cubic parts of S+SGFS + S_{\mathrm{GF}} in the two variables, and let S2S_{2} be their common quadratic functional. Taylor expansion about the flat vacuum, whose first variation vanishes, gives

S3,v[v]−S3,σ1[−v]=−12∫d4xδS2[v]δvμνUvμν=−116X∫d4x(P(−□)v)μνUvμν.(55)\begin{aligned} S_{3,v}[v] - S_{3,\sigma_{1}}[-v] &= -\frac{1}{2}\int\mathrm{d}^{4}x\frac{\delta S_{2}[v]}{\delta v^{\mu\nu}}U_{v}^{\mu\nu}\\ &= -\frac{1}{16\mathcal{X}}\int\mathrm{d}^{4}x\bigl(P(-\Box)v\bigr)_{\mu\nu}U_{v}^{\mu\nu}. \tag*{(55)} \end{aligned}

The functional derivative is with respect to the independent contravariant field, and the equality holds for integrated actions with the stated boundary convention. It includes the cubic term generated by F[σ1(v)]2F[\sigma_{1}(v)]^{2}; omitting that term changes the comparison.

For three external free solutions satisfying P(−□)vi=0P(-\Box)v_{i} = 0, every term in the multilinear part of the right-hand side vanishes. With the common linear identification σ1i=−vi\sigma_{1i} = -v_{i}, identical canonical normalization and momentum conservation, the physical transverse-traceless three-point coefficients therefore agree. This is the first interacting-order realization of local field-redefinition equivalence in the chosen gauge [11, 12]. The individual off-shell vertices remain different. Equation (55) is the flat-vacuum identity; on a general off-shell curved background, additional contributions from the first variation must be retained.

Appendix B derives all six density cubic terms, the exponential gauge contribution and Eq. (55). Its explicit complex-momentum example evaluates the same nonzero three-point coefficient directly in both variables and matches its normalization to the Einstein-gravity interaction. This checks the interaction without discarding the determinant contribution or introducing a second independent gravitational field.

Schwarzschild configurations and the background-harmonic gauge

We construct a finite Schwarzschild configuration in the background-harmonic gauge of Sec. III. The density and exponential variables reconstruct the same metric exactly, and the configuration is stationary under unrestricted variations of the gauge-fixed action in the exterior. We then distinguish this statement from variation within a restricted radial ansatz.

Exact density and exponential fields

Let m=GM>0m=GM>0, so that rs=2mr_s=2m. In coordinates (t,ρ,θ,ϕ)(t,\rho,\theta,\phi), choose the fixed flat reference metric

ds02=dt2−dρ2−ρ2dΩ2,dΩ2=dθ2+sin⁡2θ dϕ2.(56)ds_0^2=dt^2-d\rho^2-\rho^2d\Omega^2,\qquad d\Omega^2=d\theta^2+\sin^2\theta\,d\phi^2. \tag*{(56)}

The Schwarzschild exterior has the harmonic-radius representation [13]

ds2=f(ρ) dt2−f(ρ)−1dρ2−(ρ+m)2dΩ2,f(ρ)=ρ−mρ+m,ρ>m.(57)ds^2=f(\rho)\,dt^2-f(\rho)^{-1}d\rho^2-(\rho+m)^2d\Omega^2,\qquad f(\rho)=\frac{\rho-m}{\rho+m},\qquad\rho>m. \tag*{(57)}

Its areal radius is r=ρ+mr=\rho+m. The associated Cartesian functions (t,ρsin⁡θcos⁡ϕ,ρsin⁡θsin⁡ϕ,ρcos⁡θ)(t,\rho\sin\theta\cos\phi,\rho\sin\theta\sin\phi,\rho\cos\theta) form a harmonic coordinate system. Below we verify the background-covariant gauge directly in spherical coordinates, retaining the nonzero reference connection.

Set u=m/ρu=m/\rho, with 0<u<10<u<1. The volume ratio and normalized inverse-density fluctuation are

ω=−g−g0=(1+u)2,\omega=\frac{\sqrt{-g}}{\sqrt{-g_0}}=(1+u)^2,
σ1μν=diag⁡((1+u)31−u−1,u2,0,0).(58)\sigma_1^{\mu\nu}=\operatorname{diag}\left(\frac{(1+u)^3}{1-u}-1,u^2,0,0\right). \tag*{(58)}

Consequently,

Q νμ=σμλg0λν=diag⁡((1+u)31−u,1−u2,1,1),Q^\mu_{\ \nu}=\sigma^{\mu\lambda}g_{0\lambda\nu}=\operatorname{diag}\left(\frac{(1+u)^3}{1-u},1-u^2,1,1\right),
det⁡Q=(1+u)4=ω2,H=4ln⁡(1+u).(59)\det Q=(1+u)^4=\omega^2,\qquad H=4\ln(1+u). \tag*{(59)}

The inverse density is

τμν=diag⁡(1−u(1+u)3,11−u2,−ρ2,−ρ2sin⁡2θ).(60)\tau_{\mu\nu}=\operatorname{diag}\left(\frac{1-u}{(1+u)^3},\frac{1}{1-u^2},-\rho^2,-\rho^2\sin^2\theta\right). \tag*{(60)}

Multiplication by eH/2=ωe^{H/2}=\omega reproduces every component of Eq. (57). All eigenvalues of QQ are positive on the exterior patch, so its principal real matrix logarithm gives

V νμ=−(ln⁡Q) νμ=diag⁡(ln⁡1−u(1+u)3,−ln⁡(1−u2),0,0).(61)V^\mu_{\ \nu}=-(\ln Q)^\mu_{\ \nu}=\operatorname{diag}\left(\ln\frac{1-u}{(1+u)^3},-\ln(1-u^2),0,0\right). \tag*{(61)}

The nonzero contravariant components are therefore

vtt=ln⁡1−u(1+u)3,vρρ=ln⁡(1−u2).(62)v^{tt}=\ln\frac{1-u}{(1+u)^3},\qquad v^{\rho\rho}=\ln(1-u^2). \tag*{(62)}

In particular, V ρρ=vρρg0ρρ=−vρρV^\rho_{\ \rho}=v^{\rho\rho}g_{0\rho\rho}=-v^{\rho\rho}. Since tr⁡V=−H\operatorname{tr}V=-H, the metric exponential is

h=V−12Itr⁡V=diag⁡(ln⁡f,−ln⁡f,2ln⁡(1+u),2ln⁡(1+u)).(63)h=V-\frac{1}{2}\mathbb{I}\operatorname{tr}V=\operatorname{diag}\left(\ln f,-\ln f,2\ln(1+u),2\ln(1+u)\right). \tag*{(63)}

Thus g=g0ehg=g_0e^h independently reconstructs the same metric. Both fields tend to zero as ρ→∞\rho\to\infty, and no weak-field truncation has been used.

Bibliography

Exact gauge condition and bulk action

For the fixed reference metric (56), the only potentially nonzero component of Fμ=∇νσ1μνF^\mu= \nabla_\nu\sigma_1^{\mu\nu} is radial. Since the angular components of σ1μν\sigma_1^{\mu\nu} vanish, the spherical connection gives

Fρ=∂ρσ1ρρ+2ρσ1ρρ=∂ρ(m2ρ2)+2m2ρ3=0,Ft=Fθ=Fϕ=0.(64)F^\rho= \partial_\rho\sigma_1^{\rho\rho} + \frac{2}{\rho}\sigma_1^{\rho\rho} = \partial_\rho\left(\frac{m^2}{\rho^2}\right) + \frac{2m^2}{\rho^3} = 0,\qquad F^t = F^\theta= F^\phi= 0. \tag*{(64)}

This is the exact nonlinear condition (22). In the exponential coordinate it is satisfied by F[σ1(v)]F[\sigma_1(v)], with the complete field map retained.

The determinant contribution to the exact density action is nonzero:

Hρ=−4mρ(ρ+m),18σμνHμHν=−2m2(ρ−m)ρ4(ρ+m).(65)H_\rho= -\frac{4m}{\rho(\rho+m)},\qquad\frac{1}{8}\sigma^{\mu\nu}H_\mu H_\nu= -\frac{2m^2(\rho-m)}{\rho^4(\rho+m)}. \tag*{(65)}

Direct evaluation of the other two contractions in Eq. (17) gives

−14σμνταρτβλ(∇μσαβ)(∇νσρλ)+12ταρ(∇λσαβ)(∇βσλρ)=4m2ρ3(ρ+m).(66)-\frac{1}{4}\sigma^{\mu\nu}\tau_{\alpha\rho}\tau_{\beta\lambda}(\nabla_\mu\sigma^{\alpha\beta})(\nabla_\nu\sigma^{\rho\lambda}) +\frac{1}{2}\tau_{\alpha\rho}(\nabla_\lambda\sigma^{\alpha\beta})(\nabla_\beta\sigma^{\lambda\rho}) =\frac{4m^2}{\rho^3(\rho+m)}. \tag*{(66)}

The flatness of g0g_0 then implies

Lσ=2m2ρ4.(67)\mathcal{L}_\sigma= \frac{2m^2}{\rho^4}. \tag*{(67)}

The boundary current in Eq. (15) is

Wμ=(0,−2m(ρ−m)ρ3,0,0),∇μWμ=1ρ2∂ρ(ρ2Wρ)=−2m2ρ4.(68)W^\mu= \left(0,-\frac{2m(\rho-m)}{\rho^3},0,0\right),\qquad\nabla_\mu W^\mu= \frac{1}{\rho^2}\partial_\rho(\rho^2W^\rho)=-\frac{2m^2}{\rho^4}. \tag*{(68)}

Hence Lσ+∇μWμ=0\mathcal{L}_\sigma+\nabla_\mu W^\mu=0, consistently with ωR[g]=0\omega R[g]=0. The nonzero bulk density is compatible with the vanishing Ricci scalar because the exact action identity includes this divergence.

Unrestricted stationarity

With primes denoting derivatives with respect to ρ\rho, the independent Ricci components of Eq. (57) are

Rtt=f(f′′2+f′ρ+m),Rρρ=−1f(f′′2+f′ρ+m),Rθθ=1−f−(ρ+m)f′,Rϕϕ=sin⁡2θ Rθθ.(69)\begin{aligned} R_{tt} &= f\left(\frac{f''}{2}+\frac{f'}{\rho+m}\right),& R_{\rho\rho} &= -\frac{1}{f}\left(\frac{f''}{2}+\frac{f'}{\rho+m}\right),\\ R_{\theta\theta} &= 1-f-(\rho+m)f',& R_{\phi\phi} &= \sin^2\theta\,R_{\theta\theta}. \tag*{(69)} \end{aligned}

All off-diagonal components vanish. Substitution of f′=2m/(ρ+m)2f'=2m/(\rho+m)^2 and f′′=−4m/(ρ+m)3f''=-4m/(\rho+m)^3 gives Rμν[g]=0R_{\mu\nu}[g]=0 throughout the exterior.

At fixed g0g_0, the four-dimensional density reconstruction implies

δln⁡ω=12ταβδσαβ,δgμν=ω−1(δσμν−12σμνταβδσαβ).(70)\delta\ln\omega=\frac{1}{2}\tau_{\alpha\beta}\delta\sigma^{\alpha\beta},\qquad \delta g^{\mu\nu}=\omega^{-1}\left(\delta\sigma^{\mu\nu}-\frac{1}{2}\sigma^{\mu\nu}\tau_{\alpha\beta}\delta\sigma^{\alpha\beta}\right). \tag*{(70)}

For arbitrary smooth variations with compact support in the exterior, the Einstein–Hilbert variation therefore becomes

δS=−14χ∫d4x−g0 Rμν[g] δσμν=0.(71)\delta S=-\frac{1}{4\chi}\int d^4x\sqrt{-g_0}\,R_{\mu\nu}[g]\,\delta\sigma^{\mu\nu}=0. \tag*{(71)}

The gauge-fixing term varies as

δSGF=14χα∫d4x−g0 Fμ∇νδσμν,Fμ=g0μνFν.(72)\delta S_{\mathrm{GF}}=\frac{1}{4\chi\alpha}\int d^4x\sqrt{-g_0}\,F_\mu\nabla_\nu\delta\sigma^{\mu\nu},\qquad F_\mu=g_{0\mu\nu}F^\nu. \tag*{(72)}

Equation (64) consequently establishes δ(S+SGF)=0\delta(S+S_{\mathrm{GF}})=0, in particular at α=1\alpha=1. These variations are neither restricted to spherical symmetry nor required to preserve the gauge condition. The locally invertible field map gives the same stationarity in the exponential variable. With vanishing ghost and antighost fields, the configuration is also stationary for the classical ghost-extended action. These statements concern the exterior bulk variational problem; boundary action values require separately specified boundary conditions and terms.

Independent radial variation

For comparison with Appendix C, use the areal-radius ansatz

gμν=diag⁡(A,−B,−r2,−r2sin⁡2θ),g0μν=diag⁡(1,−1,−r2,−r2sin⁡2θ),(73)g_{\mu\nu}=\operatorname{diag}(A,-B,-r^2,-r^2\sin^2\theta),\qquad g_{0\mu\nu}=\operatorname{diag}(1,-1,-r^2,-r^2\sin^2\theta), \tag*{(73)}

where A(r),B(r)>0A(r), B(r)>0. After angular integration and removal of the radial total derivative displayed in Appendix C, the action is

Sred[A,B]=2πχ∫dt dr Lr,Lr=AB3[rB′+B(B−1)].(74)S_{\mathrm{red}}[A,B]=\frac{2\pi}{\chi}\int dt\,dr\,L_r,\qquad L_r=\sqrt{\frac{A}{B^3}}\left[rB'+B(B-1)\right]. \tag*{(74)}

Primes now denote radial derivatives with respect to rr. Independent variation gives

rB′+B(B−1)=0,rA′−A(B−1)=0.(75)rB'+B(B-1)=0,\qquad rA'-A(B-1)=0. \tag*{(75)}

The first equation integrates to r(1−B−1)=2mr(1-B^{-1})=2m, while the pair gives (ln⁡AB)′=0(\ln AB)'=0. Asymptotic normalization fixes A=1−2m/rA=1-2m/r and B=A−1B=A^{-1}. Imposing A=F(r)A=F(r) and B=F(r)−1B=F(r)^{-1} before variation instead gives

Lr=1−F−rF′=[r(1−F)]′.(76)L_r=1-F-rF'=[r(1-F)]'. \tag*{(76)}

whose restricted variation supplies no radial equation. The additional relation AB=1AB=1 is not a consequence of spherical symmetry alone. Its premature imposition removes an independent variation; this does not constitute a failure of the principle of symmetric criticality [14].

Relative to the flat reference in Eq. (73), the areal-radius Schwarzschild representative has

σ1μν=diag⁡(2mr−2m,2mr,0,0),Fμ=(0,2mr2,0,0).(77)\sigma_1^{\mu\nu}=\operatorname{diag}\left(\frac{2m}{r-2m},\frac{2m}{r},0,0\right),\qquad F^\mu=\left(0,\frac{2m}{r^2},0,0\right). \tag*{(77)}

The harmonic configuration (57) uses a different identification of the same physical geometry with the flat reference. A simultaneous coordinate transformation r=ρ+mr=\rho+m of both metrics in Eq. (73) would give background angular coefficients −(ρ+m)2-(\rho+m)^2 and −(ρ+m)2sin⁡2θ-(\rho+m)^2\sin^2\theta, rather than those in Eq. (56), and would preserve the nonzero gauge vector. The exact gauge-fixed stationary configuration is the representative defined by Eqs. (56) and (57).

Discussion and conclusions

We have compared two established choices of metric perturbation, the trace-reversed exponential field vv and the inverse-density fluctuation σ1\sigma_1, within the same second-order Einstein–Hilbert action. The covariant background split isolates a term linear in σ1\sigma_1 and a nonlinear derivative contribution. The exact inverse-density and determinant identities make this a single-variable action in either scheme. The determinant term is part of the gravitational dynamics and is retained throughout.

A single background-divergence gauge, identified exactly with a nonlinear background-harmonic condition, fixes the perturbative construction. Its explicit Faddeev–Popov operator supplies the free ghosts and their interaction with the density field. The curved-background quadratic operators obey Eq. (2): their Ricci-dependent difference is generated when the nonlinear field map acts on the background tadpole. On Ricci-flat backgrounds they agree under the linear field identification, and their flat propagator contains the usual trace reversal. Including the cosmological term changes the difference to −A+2ΔΠ-A+2\Delta\Pi, which vanishes at an Einstein saddle. At the first interacting order on a flat background, Eq. (55) relates the two cubic actions through the free field equation. Its multilinear difference vanishes for the common on-shell external fields, as checked by independent exponential and density calculations in Appendix B, with explicit normalization against the known Einstein-gravity three-point interaction.

The exterior Schwarzschild example uses a finite harmonic representative relative to a fixed flat reference. Both field coordinates reconstruct the metric exactly, the density divergence vanishes, and the complete exterior Ricci tensor is zero. The bulk density and boundary current satisfy the exact action identity. These results establish stationarity of the gauge-fixed action under unrestricted compactly supported exterior variations. The separate areal-radius calculation retains the independent two-function variation: imposing the relation between the radial functions beforehand removes an equation. The distinction between the two reference identifications also explains why a simultaneous coordinate transformation of the original metric pair cannot remove its nonzero gauge vector.

The two schemes organize the same theory differently. Their practical use in a particular higher-order calculation depends on the vertices, gauge and observable involved; no general computational advantage follows from the field redefinition alone. Extending the present classical and tree-level comparison to loops requires a common choice of functional measure, regulator and source prescription, with coordinate Jacobians treated consistently. Those choices are separate from the single Faddeev–Popov gauge construction used here.

On the gauge-fixing procedure

The reference metric is fixed in a quantum gauge transformation. Under the convention δξg=Lξg\delta_{\xi}g=\mathcal{L}_{\xi}g, the full metric density transforms as a density, while the reference volume in σμν=−ggμν/−g0\sigma^{\mu\nu}=\sqrt{-g}g^{\mu\nu}/\sqrt{-g_{0}} is held fixed. In background-covariant notation,

δξσμν=ξρ∇ρσμν−σρν∇ρξμ−σμρ∇ρξν+σμν∇ρξρ.(78)\delta_{\xi}\sigma^{\mu\nu}=\xi^{\rho}\nabla_{\rho}\sigma^{\mu\nu}-\sigma^{\rho\nu}\nabla_{\rho}\xi^{\mu}-\sigma^{\mu\rho}\nabla_{\rho}\xi^{\nu}+\sigma^{\mu\nu}\nabla_{\rho}\xi^{\rho}. \tag*{(78)}

This is distinct from a simultaneous coordinate transformation of gg and g0g_{0}, under which σ\sigma is an ordinary tensor. Its linear term at σ1=0\sigma_{1}=0 is −∇νξμ−∇μξν+g0μν∇⋅ξ-\nabla^{\nu}\xi^{\mu}-\nabla^{\mu}\xi^{\nu}+g_{0}^{\mu\nu}\nabla\cdot\xi.

For the single gauge Fμ=∇νσ1μνF^{\mu}=\nabla_{\nu}\sigma_{1}^{\mu\nu}, define

Nμν(σ1,c)=cρ∇ρσ1μν−σ1ρν∇ρcμ−σ1μρ∇ρcν+σ1μν∇ρcρ,N^{\mu\nu}(\sigma_{1},c)=c^{\rho}\nabla_{\rho}\sigma_{1}^{\mu\nu}-\sigma_{1}^{\rho\nu}\nabla_{\rho}c^{\mu}-\sigma_{1}^{\mu\rho}\nabla_{\rho}c^{\nu}+\sigma_{1}^{\mu\nu}\nabla_{\rho}c^{\rho},
(M[σ1]c)μ=−□0cμ−R0 ρ μcρ+∇νNμν(σ1,c).(79)(M[\sigma_{1}]c)^{\mu}=-\Box_{0}c^{\mu}-R_{0\ \rho}^{\ \mu}c^{\rho}+\nabla_{\nu}N^{\mu\nu}(\sigma_{1},c). \tag*{(79)}

For example, its field-independent part follows directly from

∇νδcσμν∣σ1=0=−□0cμ−∇ν∇μcν+∇μ∇νcν=−□0cμ−[∇ν,∇μ]cν=−□0cμ−R0 ρ μcρ.(80)\begin{aligned} \left.\nabla_{\nu}\delta_{c}\sigma^{\mu\nu}\right|_{\sigma_{1}=0}=-\Box_{0}c^{\mu}-\nabla_{\nu}\nabla^{\mu}c^{\nu}+\nabla^{\mu}\nabla_{\nu}c^{\nu} \\ &=-\Box_{0}c^{\mu}-[\nabla_{\nu},\nabla^{\mu}]c^{\nu}=-\Box_{0}c^{\mu}-R_{0\ \rho}^{\ \mu}c^{\rho}. \tag*{(80)} \end{aligned}

This explicit operator is the variation of the four-component gauge condition. An additional trace-gradient or determinant condition is neither needed nor included.

Up to a field-independent normalization, the Faddeev–Popov determinant is represented by anticommuting vector fields cμc^{\mu}, cˉμ\bar{c}_{\mu} with

Sgh=∫d4x−g0 cˉμ(M[σ1]c)μ.(81)S_{\mathrm{gh}}=\int d^{4}x\sqrt{-g_{0}}\,\bar{c}_{\mu}(M[\sigma_{1}]c)^{\mu}. \tag*{(81)}

A constant rescaling of cˉ\bar{c} has been used to choose this normalization. With the Fourier convention of Eq. (32), the flat ghost propagator, omitting the momentum delta function, is

⟨cμ(p)cˉν(−p)⟩0=iδνμp2+i0.(82)\langle c^{\mu}(p)\bar{c}_{\nu}(-p)\rangle_{0}=\frac{i\delta^{\mu}_{\nu}}{p^{2}+i0}. \tag*{(82)}

Integrating the last term of Eq. (79) by parts gives the explicit flat ghost–graviton interaction,

Sgh(3)=−∫d4x (∂νcˉμ)[cρ∂ρσ1μν−σ1ρν∂ρcμ−σ1μρ∂ρcν+σ1μν∂ρcρ].(83)S_{\mathrm{gh}}^{(3)}=-\int d^{4}x\,(\partial_{\nu}\bar{c}_{\mu})\left[c^{\rho}\partial_{\rho}\sigma_{1}^{\mu\nu}-\sigma_{1}^{\rho\nu}\partial_{\rho}c^{\mu}-\sigma_{1}^{\mu\rho}\partial_{\rho}c^{\nu}+\sigma_{1}^{\mu\nu}\partial_{\rho}c^{\rho}\right]. \tag*{(83)}

This specifies the vertex with its derivatives, tensor contractions and sign. On a curved reference metric, replacing these derivatives and the integration measure by their background-covariant versions gives the same interaction. The operator is affine in σ1\sigma_{1}, so the density ghost action has no higher graviton powers. In the exponential scheme one substitutes σ1(v)\sigma_{1}(v); this generates its higher ghost interactions with the same gauge choice. Functional inverses and determinants must be defined on the chosen domain with any residual zero modes handled separately.

Flat-background verification of the cubic expansion

This appendix checks the cubic interaction in the two field variables on a Cartesian flat reference metric. Write sμν≡σ1μνs^{\mu\nu}\equiv\sigma_{1}^{\mu\nu}, and lower its indices with ημν\eta_{\mu\nu}. The lowered field sμνs_{\mu\nu} is not the perturbation of the inverse density. Define

t=ημνsμν,Uμν=sμρs νρ,u=sμνsμν.(84)t=\eta_{\mu\nu}s^{\mu\nu},\qquad U_{\mu\nu}=s_{\mu\rho}s^{\rho}_{\ \nu},\qquad u=s^{\mu\nu}s_{\mu\nu}. \tag*{(84)}

The expansions needed at cubic order are

τμν=ημν−sμν+Uμν+O(s3),H=ln⁡det⁡(1+sη)=t−12u+O(s3).(85)\tau_{\mu\nu}=\eta_{\mu\nu}-s_{\mu\nu}+U_{\mu\nu}+O(s^{3}),\qquad H=\ln\det(1+s\eta)=t-\frac{1}{2}u+O(s^{3}). \tag*{(85)}

Here τ=σ−1\tau=\sigma^{-1}, σ=η−1+s\sigma=\eta^{-1}+s, and χ=4πG\chi=4\pi G with κ2=8χ\kappa^{2}=8\chi.

The density cubic interaction

The cubic terms can be read off without truncating away the determinant contribution. The necessary derivative products are

(∂μσαβ)(∂ντρβ)=−∂μsαβ∂νsρβ+∂μsαβ∂νUρβ+O(s4),σμν∂μH∂νH=∂μt∂μt+sμν∂μt∂νt−∂μt∂μu+O(s4).(86)\begin{aligned} (\partial_\mu\sigma^{\alpha\beta})(\partial_\nu\tau_{\rho\beta})&=-\partial_\mu s^{\alpha\beta}\partial_\nu s_{\rho\beta}+\partial_\mu s^{\alpha\beta}\partial_\nu U_{\rho\beta}+O(s^4),\\ \sigma^{\mu\nu}\partial_\mu H\partial_\nu H&=\partial_\mu t\partial^\mu t+s^{\mu\nu}\partial_\mu t\partial_\nu t-\partial_\mu t\partial^\mu u+O(s^4). \tag*{(86)} \end{aligned}

In the first line, each quadratic product also multiplies the linear part of the explicit σ\sigma in Eq. (16). Collecting all terms of degree three and including the action normalization −1/(4χ)-1/(4\chi) gives

L3,s=116χsμν∂μsαβ∂νsαβ−18χsμρ∂μsαβ∂αsρβ−116χ∂μsαβ∂μUαβ+18χ∂ρsαβ∂αUρβ−132χsμν∂μt∂νt+132χ∂μt∂μu.(87)\begin{aligned} \mathcal{L}_{3,s}={}&\frac{1}{16\chi}s^{\mu\nu}\partial_\mu s^{\alpha\beta}\partial_\nu s_{\alpha\beta}-\frac{1}{8\chi}s^{\mu\rho}\partial_\mu s^{\alpha\beta}\partial_\alpha s_{\rho\beta}\\ &-\frac{1}{16\chi}\partial_\mu s^{\alpha\beta}\partial^\mu U_{\alpha\beta}+\frac{1}{8\chi}\partial_\rho s^{\alpha\beta}\partial_\alpha U_{\rho\beta}\\ &-\frac{1}{32\chi}s^{\mu\nu}\partial_\mu t\partial_\nu t+\frac{1}{32\chi}\partial_\mu t\partial^\mu u. \tag*{(87)} \end{aligned}

The last two terms follow from σμν∂μH∂νH/8\sigma^{\mu\nu}\partial_\mu H\partial_\nu H/8 in the bulk Lagrangian and are required off shell. No transverse-traceless condition has been used. The gauge choice of Appendix A,

Fμ[s]=∂νsμν,SGF[s]=18χ∫d4x FμFμ,(88)F^\mu[s]=\partial_\nu s^{\mu\nu},\qquad S_{\mathrm{GF}}[s]=\frac{1}{8\chi}\int d^4x\,F_\mu F^\mu, \tag*{(88)}

is exactly quadratic in ss and therefore contributes no cubic term in this variable. It gives

S2[s]=116χ∫d4x sμν(P(−□)s)μν,(Ps)μν=sμν−12ημνt.(89)S_2[s]=\frac{1}{16\chi}\int d^4x\,s^{\mu\nu}(P(-\Box)s)_{\mu\nu},\qquad(Ps)_{\mu\nu}=s_{\mu\nu}-\frac{1}{2}\eta_{\mu\nu}t. \tag*{(89)}

The trace-reversal operator satisfies P2=1P^2=1 in four dimensions.

Comparison with the exponential variable

For Vμν=vμρηρνV^\mu{}_\nu=v^{\mu\rho}\eta_{\rho\nu} and σ=e−Vη−1\sigma=e^{-V}\eta^{-1}, the field map is

sμν=−vμν+12Uvμν+O(v3),Uvμν=vμρηρλvλν.(90)s^{\mu\nu}=-v^{\mu\nu}+\frac{1}{2}U_v^{\mu\nu}+O(v^3),\qquad U_v^{\mu\nu}=v^{\mu\rho}\eta_{\rho\lambda}v^{\lambda\nu}. \tag*{(90)}

The same gauge is F[s(v)]F[s(v)]. Its cubic contribution is explicitly

SGF,v(3)=−18χ∫d4x (∂νvμν)(∂λUvμλ).(91)S_{\mathrm{GF},v}^{(3)}=-\frac{1}{8\chi}\int d^4x\,(\partial_\nu v^{\mu\nu})(\partial_\lambda U_v{}_\mu{}^\lambda). \tag*{(91)}

It must be retained when comparing the two gauge-fixed cubic actions. Expanding S2[s(v)]+S3,s[s(v)]S_2[s(v)]+S_{3,s}[s(v)] through degree three gives

S3,v[v]−S3,s[−v]=−12∫d4x δS2[v]δvμνUvμν=−116χ∫d4x (P(−□)v)μνUvμν.(92)\begin{aligned} S_{3,v}[v]-S_{3,s}[-v]&=-\frac{1}{2}\int d^4x\,\frac{\delta S_2[v]}{\delta v^{\mu\nu}}U_v^{\mu\nu}\\ &=-\frac{1}{16\chi}\int d^4x\,(P(-\Box)v)_{\mu\nu}U_v^{\mu\nu}. \tag*{(92)} \end{aligned}

This is the derivation of Eq. (55), with S3,s=S3,σ1S_{3,s}=S_{3,\sigma_1}. The equality is for integrated gauge-fixed actions with the same boundary convention. It follows from the quadratic Taylor term in Eq. (90); the first variation vanishes on the flat vacuum. Thus the difference is proportional to the free field equation. For three external solutions of that equation, every term in the multilinear difference vanishes. With the common linear identification s=−vs=-v, the physical transverse-traceless cubic coefficients therefore agree. This proves the on-shell cubic comparison without assuming a pre-existing expression for either three-graviton vertex.

Direct exponential expansion

The exponential interaction can also be calculated directly from the exact density action, without using Eq. (92). Set tv=ημνvμνt_v=\eta_{\mu\nu}v^{\mu\nu} and use

σ=η−1−v+12Uv+O(v3),τ=η+vlow+12Uv,low+O(v3),H=−tv.(93)\sigma=\eta^{-1}-v+\frac{1}{2}U_v+O(v^3),\qquad\tau=\eta+v_{\mathrm{low}}+\frac{1}{2}U_{v,\mathrm{low}}+O(v^3),\qquad H=-t_v. \tag*{(93)}

In the derivative part of Eq. (16), the cubic terms containing one derivative of UvU_v and one derivative of vv cancel pairwise upon contracting the background coefficients. For example, ∂ρUvαβ∂αvρβ=∂ρvαβ∂αUvρβ\partial^\rho U_v^{\alpha\beta}\partial_\alpha v_{\rho\beta}=\partial^\rho v^{\alpha\beta}\partial_\alpha U_{v\rho\beta} by relabeling the contracted indices. The remaining cubic density, including the transported gauge term, is

L3,vtot=−116χvμν∂μvαβ∂νvαβ+18χvμρ∂μvαβ∂αvρβ+132χvμν∂μtv∂νtv−18χ(∂νvμν)(∂λUvμλ).(94)\mathcal{L}^{\mathrm{tot}}_{3,v}=-\frac{1}{16\chi}v^{\mu\nu}\partial_\mu v^{\alpha\beta}\partial_\nu v_{\alpha\beta}+\frac{1}{8\chi}v^{\mu\rho}\partial_\mu v^{\alpha\beta}\partial_\alpha v_{\rho\beta} +\frac{1}{32\chi}v^{\mu\nu}\partial_\mu t_v\partial_\nu t_v-\frac{1}{8\chi}(\partial_\nu v^{\mu\nu})(\partial_\lambda U_v{}^\mu{}_\lambda). \tag*{(94)}

No free equation or tracelessness condition was used here. The last term is precisely Eq. (91); dropping it would change the off-shell comparison.

Explicit evaluation of the three-point coefficient

For an explicit nonzero check, take the following complex momenta and polarization vectors, in units of a fixed momentum scale:

k1=(1,0,0,1),e1=(0,1,−i,0),k2=(0,1,i,0),e2=(1,0,0,−1),k3=(−1,−1,−i,−1),e3=(0,1,i,0).(95)\begin{aligned} k_1&=(1,0,0,1),& e_1&=(0,1,-i,0),\\ k_2&=(0,1,i,0),& e_2&=(1,0,0,-1),\\ k_3&=(-1,-1,-i,-1),& e_3&=(0,1,i,0). \tag*{(95)} \end{aligned}

They satisfy ∑iki=0\sum_i k_i=0, ki2=ei2=ki⋅ei=0k_i^2=e_i^2=k_i\cdot e_i=0. Consequently ϵiμν=eiμeiν\epsilon_i^{\mu\nu}=e_i^\mu e_i^\nu are transverse and traceless. Introduce a common canonically normalized external field ϕμν(x)=∑iziϵiμνeiki⋅x\phi^{\mu\nu}(x)=\sum_i z_i\epsilon_i^{\mu\nu}e^{ik_i\cdot x}. Evaluate the density cubic action at s=κϕs=\kappa\phi and the exponential cubic action at v=−κϕv=-\kappa\phi. Define V3V_3 as the coefficient of z1z2z3z_1z_2z_3 in the action after removing the common momentum-conserving delta function; the vertex is iV3iV_3.

Each wave obeys □ϕμν=0\Box\phi^{\mu\nu}=0 and ∂μϕμν=0\partial_\mu\phi^{\mu\nu}=0. After integration by parts, the third term in Eq. (87) is proportional to □ϕαβ\Box\phi^{\alpha\beta} and the fourth to ∂α∂ρϕαβ\partial_\alpha\partial^\rho\phi^\alpha{}_\beta. Neither contributes. The last two terms vanish because t=0t=0. Thus only the first two terms remain. For factorized polarizations their coefficients are

V3(1)=−κ2∑(a,b,c)∈S3(ea⋅kb)(ea⋅kc)(eb⋅ec)2,V_3^{(1)}=-\frac{\kappa}{2}\sum_{(a,b,c)\in S_3}(e_a\cdot k_b)(e_a\cdot k_c)(e_b\cdot e_c)^2,
V3(2)=κ∑(a,b,c)∈S3(ea⋅kb)(eb⋅kc)(ea⋅ec)(eb⋅ec).(96)V_3^{(2)}=\kappa\sum_{(a,b,c)\in S_3}(e_a\cdot k_b)(e_b\cdot k_c)(e_a\cdot e_c)(e_b\cdot e_c). \tag*{(96)}

The sum is over the six permutations of the three wave labels, which are not spacetime indices. The signs include the two factors of ii from the derivatives. For Eq. (95), the relevant contractions are

e1⋅e3=−2,e1⋅e2=e2⋅e3=0,e2⋅k1=2,e2⋅k3=−2.(97)e_1\cdot e_3=-2,\qquad e_1\cdot e_2=e_2\cdot e_3=0,\qquad e_2\cdot k_1=2,\qquad e_2\cdot k_3=-2. \tag*{(97)}

Only (a,b,c)=(2,1,3),(2,3,1)(a,b,c)=(2,1,3),(2,3,1) contribute to V3(1)V_3^{(1)}, and every summand in V3(2)V_3^{(2)} contains a vanishing polarization contraction. Therefore

V3(1)=−κ2[2(2)(−2)(−2)2]=16κ,V3(2)=0.(98)V_3^{(1)}=-\frac{\kappa}{2}[2(2)(-2)(-2)^2]=16\kappa,\qquad V_3^{(2)}=0. \tag*{(98)}

Independently substitute v=−κϕv=-\kappa\phi into Eq. (94). Its trace and gauge terms vanish on these external waves. The first two terms become, respectively,

κ2ϕμν∂μϕαβ∂νϕαβ,−κϕμρ∂μϕαβ∂αϕρβ.(99)\frac{\kappa}{2}\phi^{\mu\nu}\partial_\mu\phi^{\alpha\beta}\partial_\nu\phi_{\alpha\beta},\qquad-\kappa\phi^{\mu\rho}\partial_\mu\phi^{\alpha\beta}\partial_\alpha\phi_{\rho\beta}. \tag*{(99)}

Their six-permutation sums are the two expressions in Eq. (96), giving 16κ16\kappa and zero directly. Thus both explicit evaluations give

V3,s=V3,v=16κ.(100)V_{3,s}=V_{3,v}=16\kappa. \tag*{(100)}

Equivalently, using unit coefficients for the unnormalized ss waves and the opposite vv waves gives 16/κ2=2/χ16/\kappa^2=2/\chi. The displayed value uses the explicit, unnormalized polarization vectors in Eq. (95); restoring a momentum scale EE multiplies it by E2E^2. Complex momenta allow a nonzero three-point coefficient, which real collinear massless momenta would not test.

Normalization against the Einstein-gravity three-point interaction

For an external benchmark, use the cubic Einstein-gravity Lagrangian in Eq. (10) of Bern and Grant [7]. Denote its canonically normalized graviton by qμνq_{\mu\nu} to distinguish it from the exponential metric variable hh used here. Its linear metric normalization is gμν=ημν+κqμν+O(q2)g_{\mu\nu}=\eta_{\mu\nu}+\kappa q_{\mu\nu}+O(q^2), with the same κ2=32πG\kappa^2=32\pi G and signature (+−−−)(+---). Their Weinberg curvature convention has the opposite sign to Eq. (5), so their +2−gR/κ2+2\sqrt{-g}R/\kappa^2 is our Einstein–Hilbert action. With indices restored, their cubic expression is

L3,GR[q]=κ[12qμνqρσ∂μ∂νqρσ+qνμ(∂λqμρ)(∂νqρλ)].(101)\mathcal{L}_{3,\mathrm{GR}}[q]=\kappa\left[\frac{1}{2}q^{\mu\nu}q^{\rho\sigma}\partial_\mu\partial_\nu q_{\rho\sigma}+q^{\nu\mu}(\partial^\lambda q_{\mu}{}^\rho)(\partial_\nu q_{\rho\lambda})\right]. \tag*{(101)}

This is an on-shell benchmark in the reference’s field and gauge conventions, not an identification of its full off-shell action with ours. We evaluate the displayed Lagrangian with our Fourier waves eik⋅xe^{ik\cdot x} and the vertex convention iV3iV_3; consequently each pair of derivatives supplies i2=−1i^2=-1.

For q=ϕq=\phi and arbitrary factorized transverse null polarizations, its two terms give

V3,GR[ϕ]=−κ2∑(a,b,c)∈S3(ea⋅kb)2(eb⋅ec)2−κ∑(a,b,c)∈S3(eb⋅kc)(ec⋅ka)(ea⋅eb)(ea⋅ec)=−κJ2,V_{3,\mathrm{GR}}[\phi]=-\frac{\kappa}{2}\sum_{(a,b,c)\in S_3}(e_a\cdot k_b)^2(e_b\cdot e_c)^2-\kappa\sum_{(a,b,c)\in S_3}(e_b\cdot k_c)(e_c\cdot k_a)(e_a\cdot e_b)(e_a\cdot e_c)=-\kappa J^2,
J=(e1⋅e2)(e3⋅k1)+(e2⋅e3)(e1⋅k2)+(e3⋅e1)(e2⋅k3).(102)J=(e_1\cdot e_2)(e_3\cdot k_1)+(e_2\cdot e_3)(e_1\cdot k_2)+(e_3\cdot e_1)(e_2\cdot k_3). \tag*{(102)}

The last equality uses momentum conservation and ei⋅ki=0e_i\cdot k_i=0. The corresponding sums in Eq. (96) give +κJ2+\kappa J^2. The apparent sign difference is fixed by the external-field identification: s=κϕs=\kappa\phi and v=−κϕv=-\kappa\phi both give g=η−κϕlow+O(ϕ2)g=\eta-\kappa\phi_{\mathrm{low}}+O(\phi^2) for traceless waves. Therefore the matching reference field is q=−ϕlowq=-\phi_{\mathrm{low}}, and

iV3,s=iV3,v=iV3,GR[−ϕ]=iκJ2=16iκfor xref‘Eq. (B12)‘.(103)iV_{3,s}=iV_{3,v}=iV_{3,\mathrm{GR}}[-\phi]=i\kappa J^2=16i\kappa\quad\text{for {xref}`Eq. (B12)`}. \tag*{(103)}

No polarization rescaling or coupling is stripped in this equality: each external tensor is exactly eiμeiνe_i^\mu e_i^\nu with the field sign just specified, and the common (2π)4δ(4)(∑iki)(2\pi)^4\delta^{(4)}(\sum_i k_i) is omitted. Rescaling eie_i by cic_i multiplies every result by ∏ici2\prod_i c_i^2. As a further check, replacing any eie_i by kik_i makes JJ vanish using ki⋅kj=0k_i\cdot k_j=0. Multilinearity then implies invariance under ei↦ei+aikie_i\mapsto e_i+a_i k_i, providing the on-shell Ward check. These are validations of the known Einstein-gravity interaction.

The analytic comparison includes the full off-shell cubic actions, their field-redefinition identity, direct on-shell evaluations in both variables and the normalization check above. As additional algebraic checks, the metric reconstructed from σ\sigma was differentiated directly to form its Christoffel symbols and Ricci scalar, and its quadratic and cubic coefficients were compared with the density expressions using exact symbolic arithmetic. The direct exponential formula was also checked for traceful off-shell fields, and the amplitude comparison for gauge-shifted and complex-conjugate external data.

Radial curvature, variation and boundary identities

This appendix supplies the algebra underlying the two-function variation in Sec. VI. For the metric in Eq. (73), the nonzero independent Ricci components computed with Eq. (5) are

Rtt=A′′2B−A′4B(A′A+B′B)+A′rB,Rrr=−A′′2A+A′4A(A′A+B′B)+B′rB,Rθθ=1−1B+r2B(B′B−A′A),Rϕϕ=sin⁡2θ Rθθ.(104)\begin{aligned} R_{tt} &= \frac{A''}{2B}-\frac{A'}{4B}\left(\frac{A'}{A}+\frac{B'}{B}\right)+\frac{A'}{rB},\\ R_{rr} &= -\frac{A''}{2A}+\frac{A'}{4A}\left(\frac{A'}{A}+\frac{B'}{B}\right)+\frac{B'}{rB},\\ R_{\theta\theta} &= 1-\frac{1}{B}+\frac{r}{2B}\left(\frac{B'}{B}-\frac{A'}{A}\right), \qquad R_{\phi\phi}=\sin^2\theta\,R_{\theta\theta}. \tag*{(104)} \end{aligned}

Their contraction gives

r2AB R=(r2A′AB)′−2Lr,(105)r^2\sqrt{AB}\,R=\left(\frac{r^2A'}{\sqrt{AB}}\right)'-2L_r, \tag*{(105)}

where LrL_r is defined in Eq. (74). After angular integration, the prefactor of the left-hand side in the Einstein–Hilbert action is −π/χ-\pi/\chi. Removing the displayed radial total derivative therefore yields Sred=(2π/χ)∫dt dr LrS_{\mathrm{red}}=(2\pi/\chi)\int dt\,dr\,L_r.

The independent Euler–Lagrange derivatives are

EA(Lr)=Lr2A,EB(Lr)=AB2AB2[A(B−1)−rA′],EA=∂∂A−ddr∂∂A′,(106)\mathcal{E}_A(L_r)=\frac{L_r}{2A}, \qquad\mathcal{E}_B(L_r)=\frac{\sqrt{AB}}{2AB^2}\left[A(B-1)-rA'\right], \qquad\mathcal{E}_A=\frac{\partial}{\partial A}-\frac{d}{dr}\frac{\partial}{\partial A'}, \tag*{(106)}

with the analogous definition of EB\mathcal{E}_B. Setting them to zero for A,B>0A,B>0 gives Eq. (75).

The restriction A=FA=F, B=F−1B=F^{-1} instead gives Lr=1−F−rF′=[r(1−F)]′L_r=1-F-rF'=[r(1-F)]'. It also sets ω=1\omega=1 and H=0H=0. Evaluating the connection difference in Eq. (12), the boundary current in Eq. (15) has only the radial component

Wr=F′+2(F−1)r,R=1r2(r2Wr)′,Lσ=0.(107)W^r=F'+\frac{2(F-1)}{r}, \qquad R=\frac{1}{r^2}(r^2W^r)', \qquad\mathcal{L}_\sigma=0. \tag*{(107)}

The second identity follows directly by contracting Eq. (104) after the restriction. The last then follows from ωR=Lσ+∇ρWρ\omega R=\mathcal{L}_\sigma+\nabla_\rho W^\rho and the reference volume element. These identities hold for arbitrary F(r)F(r) on the nonsingular patch, not only for Schwarzschild. They explain why the restricted action cannot determine the Schwarzschild profile; the independent variations in Eq. (106) are required.

References

  1. [1]L. D. Faddeev and V. N. Popov, Covariant quantization of the gravitational field, Usp. Fiz. Nauk 111, 427 (1973), english translation: Sov. Phys. Usp. 16, 777–788 (1974), doi:10.1070/PU1974v016n06ABEH004089.
  2. [2]L. P. Grishchuk, A. N. Petrov, and A. D. Popova, Exact theory of the (Einstein) gravitational field in an arbitrary background space-time, Commun. Math. Phys. 94, 379 (1984).DOI
  3. [3]C. Cheung and G. N. Remmen, Twofold symmetries of the pure gravity action, J. High Energy Phys. 01 (2017), 104, arXiv:1612.03927 [hep-th].DOI
  4. [4]C. Cheung and G. N. Remmen, Hidden simplicity of the gravity action, J. High Energy Phys. 09 (2017), 002, arXiv:1705.00626 [hep-th].DOI
  5. [5]K. Cho, K. Kim, and K. Lee, Perturbations of general relativity to all orders and the general nᵗʰ order terms, J. High Energy Phys. 03 (2023), 112, arXiv:2209.11424 [hep-th].DOI
  6. [6]N. Ohta, R. Percacci, and A. D. Pereira, Gauges and functional measures in quantum gravity I: Einstein theory, J. High Energy Phys. 06 (2016), 115, arXiv:1605.00454 [hep-th].DOI
  7. [7]Z. Bern and A. K. Grant, Perturbative gravity from QCD amplitudes, Phys. Lett. B 457, 23 (1999), arXiv:hep-th/9904026.DOI
  8. [8]M. J. Duff, Quantum tree graphs and the Schwarzschild solution, Phys. Rev. D 7, 2317 (1973).DOI
  9. [9]L. Blanchet, Gravitational radiation from post-newtonian sources and inspiralling compact binaries, Living Rev. Relativ. 17, 2 (2014).
  10. [10]L. D. Faddeev and V. N. Popov, Feynman diagrams for the Yang–Mills field, Phys. Lett. B 25, 29 (1967).
  11. [11]S. Kamefuchi, L. O’Raifeartaigh, and A. Salam, Change of variables and equivalence theorems in quantum field theories, Nucl. Phys. 28, 529 (1961).
  12. [12]J. S. R. Chisholm, Change of variables in quantum field theories, Nucl. Phys. 26, 469 (1961).DOI
  13. [13]P. Fromholz, E. Poisson, and C. M. Will, The Schwarzschild metric: It’s the coordinates, stupid!, Am. J. Phys. 82, 295 (2014), arXiv:1308.0394 [gr-qc].DOI
  14. [14]M. E. Fels and C. G. Torre, The principle of symmetric criticality in general relativity, Class. Quantum Grav. 19, 641 (2002), arXiv:gr-qc/0108033.DOI

Paper details

vlnc:W_auxm9t77
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