betaFeedback

WorksPhysics and Astronomy

Axial gravitational-wave propagation and mass–gravity identifiability in conformal scalar–tensor lensing

Abstract

We investigate tensor-wave propagation in a specified massless, unscreened conformal scalar–tensor theory using a static Einstein–scalar exterior matched to a weakly self-gravitating matter lens. The odd-parity gravitational master potential has a scalar-charge correction at radial order R−4R^{-4} for fixed Einstein-frame mass. Its next R−5R^{-5} coefficient depends on ℓ(ℓ+1)−9\ell(\ell+1)-9, reversing sign between the quadrupole and higher multipoles; we test the expansion against the exact exterior potential. We connect this result to the locally normalized post-Newtonian light-deflection parameter and distinguish the Schwarzschild baseline at fixed Einstein-frame mass from that at fixed measured matter mass. A finite-radius WKB expansion reveals cancellation of the leading frequency-dependent potential moment for the quadrupole at fixed multipole, whereas a separate fixed-impact-parameter limit recovers an achromatic geometrical-optics action and a subleading axial contribution. Independent null-geodesic bending and numerical quadrature check these limits. For a point lens, achromatic observables constrain an effective mass rather than the lens mass and scalar coupling separately. An illustrative two-image Fisher calculation quantifies differential-phase identifiability after arrival-time and constant-phase nuisance parameters are released. The phenomenological phase experiment is separated from action-derived propagation quantities; model-specific dispersive observations require a complete matched image transfer function.

Introduction

The observation of gravitational waves (GWs) from compact binaries has opened a direct route to testing relativistic gravity in the radiative and strong-field regimes [1, 2, 3]. The multi-messenger event GW170817 has also established stringent low-redshift constraints on the speed of tensor waves [4, 5]. These achievements do not, however, determine the response of all additional degrees of freedom that arise in gravity theories beyond general relativity (GR). Scalar–tensor theories, including Horndeski gravity and its luminal subclasses, offer a well-defined setting in which scalar backgrounds, source generation and wave propagation must be treated consistently [6, 7, 8, 9, 10, 11, 12]. In particular, cosmological propagation functions describing tensor-speed and damping modifications cannot by themselves determine the local wave operator in the strongly inhomogeneous field of an intervening lens [13, 14, 15, 16].

Gravitational lensing adds a different observational configuration: multiple images are associated with the same source but follow different paths through the foreground potential. In geometric optics, their relative arrival time, magnification and Morse phase depend on the lens geometry, whereas diffraction and finite-frequency corrections require wave-optics information [17, 18, 19, 20, 21, 22, 23]. This distinction matters for future ET and LISA observations, whose sensitivities and source populations make resolved or overlapping lensed signals scientifically relevant [24, 25, 26, 27, 28, 29]. It is nevertheless incorrect to infer detectability from the lensing time delay alone: both images must fall within the observing window and satisfy signal-to-noise and selection requirements [30]. The use of a simple point lens is best understood as a transparent benchmark rather than a realistic cluster or galaxy mass model.

Theoretical progress has made the relevant physical distinction sharper. The short-wave expansion of scalar and tensor perturbations can be performed on general backgrounds, including nontrivial mixing and corrections beyond leading geometrical optics [31, 32, 33]. Menadeo and

Correspondence to: [email protected]; [email protected] Zumalacárregui derived lens-induced beyond-geometric-optics effects for a Brans–Dicke benchmark and emphasized their severe suppression under existing weak-field bounds [34]. More recently, Menadeo, Giardino and Zumalacárregui studied dispersion on inhomogeneous backgrounds, including screened and non-minimally coupled settings [35]; related approaches examine wave-optics lensing in modified gravity [36]. A recent EPJC perturbation analysis with angularly structured scalar hair demonstrates that axial–scalar closure depends on the parity content of the matter background, rather than on spherical symmetry of the stress tensor alone [37]. Our radial-scalar exterior belongs to the distinct parity-decoupled case. Consequently, a generic 1/f1/f phase, a principal null cone shared by modes, or a phenomenological image-difference parameter is not by itself a new Horndeski prediction. A model-specific result must follow from the chosen action, a physically admissible lens background and a consistently defined source-to-detector transfer calculation.

A central difficulty is that several apparently similar calculations answer distinct questions. The cosmological effective-field-theory description is organized by homogeneous-background functions [11, 38], while local lensing depends on gradients and boundary conditions. A scalar flux equation may display a 1/f1/f WKB correction without representing a physical tensor polarization. Likewise, a fixed-ℓ\ell radial partial wave is not automatically a finite-impact-parameter lensed wavepacket, and a potential difference is not automatically an observable GR-subtracted phase. The spherical Einstein–massless-scalar exterior provides a controlled laboratory in which odd-parity gravitational perturbations can be separated from the even-parity scalar sector [39, 40]. Its limitations must also be stated: an isolated scalar-charged exterior is not a self-consistent description of every extended astrophysical lens, and a constant-scalar vacuum black-hole branch need not generate the postulated anomalous source phase [41].

Here we focus on a specified, massless and unscreened conformal scalar–tensor subclass, not an unconstrained four-function Horndeski model. The objective is to determine which pieces of a proposed GW-lensing test are actually implied by this action and which remain conditional diagnostics. Our contributions are fourfold. First, we connect the Jordan-frame Cavendish coupling, the Einstein-frame mass normalization and the spherical scalar exterior, construct the physical odd-parity tensor master potential, and derive its multipole-dependent R−5R^{-5} correction, verified against the exact potential. Second, we distinguish its fixed-multipole radial WKB moment from the fixed-impact-parameter eikonal limit, checking the latter against an independent null-geodesic bending calculation. Third, we examine the model's leading achromatic lens geometry, the degeneracy of lens mass with a post-Newtonian parameter, and the conditions needed to observe both images. Fourth, using an explicitly phenomenological common-plus-differential 1/f1/f phase, we compare reduced five-parameter and equal-time/phase-marginalized nine-parameter Fisher calculations. The last calculation is a measurement-identifiability diagnostic, not a forecast for the fundamental scalar coupling [42, 43, 44].

The calculation establishes an exact odd-parity exterior sector, two complementary high-frequency limits and mass–gravity degeneracy relations. The separate phase-coefficient experiment supplies a test of statistical identifiability for future analyses that include both image transfer functions and complete waveform information.

The paper is organized in seven sections. § 2 develops the covariant perturbation framework; § 2.2 derives the axial exterior result and compares the two WKB limits with the scalar surrogate. § 4 analyzes the weak-field lens map and mass normalization, § 5 examines conditional two-image phase identifiability, and § 6 addresses the observing-window condition. § 7 summarizes the conclusions, followed by two technical appendices.

Notation and provenance. The Jordan scalar is ϕ\phi and the canonical Einstein-frame scalar is χ\chi. The dimensionless coupling QQ is a parameter of the specified action, not a measured phenomenological phase. GbareG_{\rm bare} denotes the bare tensor–sector constant; GNG_N is the locally measured Cavendish constant. For a Jordan-frame matter lens of mass MLM_L, rg=GNML/c2r_g = G_N M_L/c^2, while mEm_E is the independent Einstein-frame ADM mass expressed as a length. The FJNW exterior uses areal radius RR, coordinate rr, and parameters (B,ζ)(B,\zeta). r∗r_* is its optical radial coordinate, ℓ≥2\ell\ge2 the multipole index, ω\omega the angular frequency in Einstein-frame geometrized units and b=ℓ(ℓ+1)/ωb = \sqrt{\ell(\ell+1)}/\omega the high-multipole impact parameter. ff denotes the observed frequency in hertz. DOLD_{\rm OL}, DOSD_{\rm OS}, DLSD_{\rm LS} are angular-diameter distances and DlumD_{\rm lum} is the waveform luminosity distance. We specify the mass- and endpoint-matching conventions whenever making GR comparisons [6, 39, 41].

Covariant framework and tensor propagation

We work with the Jordan-frame Horndeski action and minimally coupled matter, using X=−gμν∇μϕ∇νϕ/2X = -g^{\mu\nu}\nabla_\mu\phi\nabla_\nu\phi/2 and metric signature (−,+,+,+)(-,+,+,+):

S=∫d4x−g∑i=25Li+Sm[gμν,ψm].(1)S = \int\mathrm{d}^{4}x \sqrt{-g} \sum_{i=2}^{5} \mathcal{L}_i + S_m[g_{\mu\nu},\psi_m]. \tag*{(1)}

The usual four densities, involving Gi(ϕ,X)G_i(\phi,X), define the covariant model [7, 9]. The familiar Bellini–Sawicki effective functions (αK,αB,αM,αT)(\alpha_K,\alpha_B,\alpha_M,\alpha_T) are defined on a homogeneous FLRW solution, not on a generic lens background [11]. In particular,

M∗2=2(G4−2XG4X+XG5ϕ−Hϕ˙XG5X),(2)M_*^2 = 2\left(G_4 - 2XG_{4X} + XG_{5\phi} - H\dot{\phi}XG_{5X}\right), \tag*{(2)}
αM=H−1d(ln⁡M∗2)/dt,(3)\alpha_M = H^{-1}\mathrm{d}(\ln M_*^2)/\mathrm{d}t, \tag*{(3)}
M∗2αT=2X[2G4X−2G5ϕ−(ϕ¨−Hϕ˙)G5X].(4)M_*^2\alpha_T = 2X\left[2G_{4X} - 2G_{5\phi} - (\ddot{\phi} - H\dot{\phi})G_{5X}\right]. \tag*{(4)}

The other two background functions constrain kineticity and braiding, but none of these quantities alone specifies the local lens profile or wave operator. We therefore do not introduce a general “ΥL(αi)\Upsilon_{\mathrm{L}}(\alpha_i)” lens-response formula. Locally, write ϕ=ϕˉ+δϕL+φ\phi= \bar{\phi} + \delta\phi_{\mathrm{L}} + \varphi and ds2=−(1+2Φ)c2dt2+(1−2Ψ)dx2ds^2 = -(1 + 2\Phi)c^2dt^2 + (1 - 2\Psi)d\mathbf{x}^2. For a fully coupled system after constraints have been eliminated, the propagating variables schematically obey

[KIJμν∇μ∇ν+AIJμ∇μ+UIJ]ψJ=0,(5)\left[K^{\mu\nu}_{IJ}\nabla_\mu\nabla_\nu+ A^\mu_{IJ}\nabla_\mu+ U_{IJ}\right]\psi^J = 0, \tag*{(5)}
ψI=(h+,h×,φ).\psi^I = (h_+,h_\times,\varphi).

The WKB principal symbol PIJ=KIJμνkμkνP_{IJ}=K^{\mu\nu}_{IJ}k_\mu k_\nu governs the characteristic and polarization branches, while transport and background gradients enter at the next orders [20, 31, 33]. The local matrices in (5) must be computed from the same action and background to obtain a general Horndeski lens-induced dispersion. We do not claim to have completed that computation.

Coupled metric–scalar perturbations

The restricted conformal model in equation (24) admits an exact Einstein-frame rewriting, which permits a check of terms absent from a longitudinal propagation equation. Defining gμνE=e2Qϕ/MPlgμνJg^E_{\mu\nu}=e^{2Q\phi/M_{\mathrm{Pl}}}g^J_{\mu\nu} and χ(ϕ)\chi(\phi) by

(dχdϕ)2=e−2Qϕ/MPl+6Q2,A(χ)=e−Qϕ(χ)/MPl,(6)\left(\frac{d\chi}{d\phi}\right)^2=e^{-2Q\phi/M_{\mathrm{Pl}}}+6Q^2,\qquad A(\chi)=e^{-Q\phi(\chi)/M_{\mathrm{Pl}}}, \tag*{(6)}

the action outside the matter source is

SE,ext=∫d4x−gE[MPl22RE−12gEμν∂μχ∂νχ].(7)S_{\mathrm{E,ext}}=\int d^4x\sqrt{-g_E}\left[\frac{M_{\mathrm{Pl}}^2}{2}R_E-\frac{1}{2}g_E^{\mu\nu}\partial_\mu\chi\partial_\nu\chi\right]. \tag*{(7)}

Matter, where present, couples to A2(χ)gμνEA^2(\chi)g^E_{\mu\nu}. The exact exterior field equations are

RμνE=MPl−2∇μχ∇νχ,□Eχ=0.(8)R^E_{\mu\nu}=M_{\mathrm{Pl}}^{-2}\nabla_\mu\chi\nabla_\nu\chi,\qquad\Box_E\chi=0. \tag*{(8)}

Let (gμνE,χ0)(g^E_{\mu\nu},\chi_0) solve equation (8) and consider gμνE↦gμνE+hμνEg^E_{\mu\nu}\mapsto g^E_{\mu\nu}+h^E_{\mu\nu} and χ0↦χ0+η\chi_0\mapsto\chi_0+\eta. Direct variation gives the following unfixed-gauge, covariant, coupled equations in the exterior:

δRμνE[hE]=MPl−2(∇μχ0)(∇νη)+MPl−2(∇νχ0)(∇μη),(9)\delta R^E_{\mu\nu}[h^E]=M_{\mathrm{Pl}}^{-2}(\nabla_\mu\chi_0)(\nabla_\nu\eta)+M_{\mathrm{Pl}}^{-2}(\nabla_\nu\chi_0)(\nabla_\mu\eta), \tag*{(9)}
□Eη=hEμν∇μ∇νχ0+(∇μhEμν−12∇νhE)∇νχ0.(10)\Box_E\eta=h_E^{\mu\nu}\nabla_\mu\nabla_\nu\chi_0+\left(\nabla_\mu h_E^{\mu\nu}-\frac{1}{2}\nabla^\nu h_E\right)\nabla_\nu\chi_0. \tag*{(10)}

where hE=gEμνhμνEh_E=g_E^{\mu\nu}h^E_{\mu\nu} and all derivatives refer to the background Einstein metric. The Ricci variation is explicitly

2δRμνE[hE]=∇ρ∇μhEρν+∇ρ∇νhEρμ−□EhμνE−∇μ∇νhE.(11)2\delta R^E_{\mu\nu}[h^E]=\nabla_\rho\nabla_\mu h_E{}^\rho{}_\nu+\nabla_\rho\nabla_\nu h_E{}^\rho{}_\mu-\Box_E h^E_{\mu\nu}-\nabla_\mu\nabla_\nu h_E. \tag*{(11)}

Equations (9)–(10) retain scalar–metric mixing, curved-background derivatives and their associated curvature commutators; no transverse derivatives have been discarded. They provide action-derived local equations but not yet a gauge-fixed physical tensor/scalar transfer matrix or a lensing phase. Under an infinitesimal background diffeomorphism, hμνE↦hμνE+2∇(μξν)h^E_{\mu\nu}\mapsto h^E_{\mu\nu}+2\nabla_{(\mu}\xi_{\nu)} and η↦η+ξρ∇ρχ0\eta\mapsto\eta+\xi^\rho\nabla_\rho\chi_0; this freedom must be fixed or handled by a gauge-invariant construction before interpreting the metric components as observable polarizations. In particular, a coordinate-gauge choice, physical polarization extraction, incident-state specification, boundary conditions and consistent GR subtraction are required before a phase can be assigned to an observable tensor mode.

Harmonic-gauge propagation and characteristics. One can go beyond the unreduced equations without assuming a longitudinal flux-weighted tensor channel. Impose the local background-covariant de Donder condition on the Einstein-frame metric perturbation,

∇μhˉμνE=0,hˉμνE=hμνE−12gμνEhE.(12)\nabla^\mu\bar{h}^E_{\mu\nu}=0,\qquad\bar{h}^E_{\mu\nu}=h^E_{\mu\nu}-\frac{1}{2}g^E_{\mu\nu}h_E. \tag*{(12)}

Commuting covariant derivatives in equation (11) and using equation (12) gives the coupled, gauge-fixed but not constraint-reduced exterior system

−12□EhμνE−RμρνσEhEρσ+Rρ(μEhν)Eρ=2MPl2(∇μχ0)(∇νη),(13)-\frac{1}{2}\Box_E h^E_{\mu\nu}-R^E_{\mu\rho\nu\sigma}h_E^{\rho\sigma}+R^E_{\rho(\mu}h^E_{\nu)}{}^\rho=\frac{2}{M_{\mathrm{Pl}}^2}(\nabla_\mu\chi_0)(\nabla_\nu\eta), \tag*{(13)}
□Eη−hEμν∇μ∇νχ0=0.(14)\Box_E\eta-h_E^{\mu\nu}\nabla_\mu\nabla_\nu\chi_0=0. \tag*{(14)}

The right-hand side of equation (13) has only one derivative of the perturbation η\eta; the background curvature terms have none. The scalar equation contains only an algebraic perturbation term in this gauge. Accordingly, for the rapidly varying amplitudes (aμν,b)eiθ/ϵ(a_{\mu\nu},b)e^{i\theta/\epsilon}, the two-derivative principal equations are

(gEρσkρkσ)aμν=0,(gEρσkρkσ)b=0,kμ=∇μθ,(15)(g_E^{\rho\sigma}k_\rho k_\sigma)a_{\mu\nu}=0,\qquad(g_E^{\rho\sigma}k_\rho k_\sigma)b=0,\qquad k_\mu=\nabla_\mu\theta, \tag*{(15)}

with the leading harmonic constraints kμ(aμν−gμνEa/2)=0k^\mu(a_{\mu\nu}-g^E_{\mu\nu}a/2)=0. After accounting for residual diffeomorphisms and constraints, the two tensor modes and one scalar therefore share the Einstein-metric null cone at principal order. Conformal rescaling preserves this null cone in the Jordan frame. This is a model-specific absence of geometric-optics speed splitting, not an absence of finite-frequency dispersion or mode conversion. The Riemann term and the scalar-gradient coupling in equation (13) remain essential when the WKB hierarchy is continued to subleading orders, despite the transverse projection of their leading derivative part in section 2.1. The source’s interior, the scattering boundary conditions and the tensor-polarization projection are not fixed by this local characteristic statement.

Screen-transverse transport. The derivative coupling on the right-hand side of equation (13) can be tested more directly than by inspecting the characteristic determinant. Along a null characteristic kμ=∇μθk_\mu=\nabla_\mu\theta, choose a locally parallel-transported screen basis of symmetric transverse–traceless tensors eAμνe^{\mu\nu}_{A}, A=+,×A=+,\times, satisfying

kμeAμν=0,gμνEeAμν=0,kρ∇ρeAμν=0.(16)k_{\mu}e^{\mu\nu}_{A}=0,\qquad g^{E}_{\mu\nu}e^{\mu\nu}_{A}=0,\qquad k^{\rho}\nabla_{\rho}e^{\mu\nu}_{A}=0. \tag*{(16)}

For the leading shortwave fields, take amplitudes varying on the background scale and write

hμνE=∑AAAeAμνeiθ/ϵ+non-TT components,h^{E}_{\mu\nu}=\sum_{A}A_{A}e^{\mu\nu}_{A}e^{i\theta/\epsilon}+\text{non-TT components},
η=Beiθ/ϵ.\eta=Be^{i\theta/\epsilon}.

The potentially leading scalar source projected on eAe_{A} obeys the algebraic identity

eAμν∇(μχ0∇ν)η=iϵBeiθ/ϵeAμν(∇μχ0)kν+O(1)=O(1).(17)e^{\mu\nu}_{A}\nabla_{(\mu}\chi_{0}\nabla_{\nu)}\eta=\frac{i}{\epsilon}Be^{i\theta/\epsilon}e^{\mu\nu}_{A}(\nabla_{\mu}\chi_{0})k_{\nu}+O(1)=O(1). \tag*{(17)}

Thus the derivative mixing term has no direct screen-TT projection at order ϵ−1\epsilon^{-1}. Projecting the same order of the wave operator on the left of equation (13), and taking the screen basis to be parallel transported along the ray, yields the standard geometric-optics law

2kρ∇ρAA+(∇ρkρ)AA=0,∇ρ(∣AA∣2kρ)=0.(18)2k^{\rho}\nabla_{\rho}A_{A}+(\nabla_{\rho}k^{\rho})A_{A}=0,\qquad\nabla_{\rho}(|A_{A}|^{2}k^{\rho})=0. \tag*{(18)}

This step assumes that the leading incident metric perturbation has been projected onto the physical screen-TT subspace and that the same eikonal branch is followed smoothly; it is not a global gauge-invariant construction of the detected polarization. To locate the first omitted terms, write hμνE=eiθ/ϵ(aμν(0)+ϵaμν(1)+⋯ )h^{E}_{\mu\nu}=e^{i\theta/\epsilon}(a^{(0)}_{\mu\nu}+\epsilon a^{(1)}_{\mu\nu}+\cdots) and η=eiθ/ϵ(b0+ϵb1+⋯ )\eta=e^{i\theta/\epsilon}(b_{0}+\epsilon b_{1}+\cdots). Define Dkz≡2kρ∇ρz+(∇ρkρ)zD_{k}z\equiv2k^{\rho}\nabla_{\rho}z+(\nabla_{\rho}k^{\rho})z and the curvature endomorphism

(Ua)μν≡−RμρνσEaρσ+Rρ(μEa ν)ρ.(19)(Ua)_{\mu\nu}\equiv-R^{E}_{\mu\rho\nu\sigma}a^{\rho\sigma}+R^{E}_{\rho(\mu}a^{\rho}_{\ \nu)}. \tag*{(19)}

The scalar equation at order ϵ−1\epsilon^{-1} implies Dkb0=0D_{k}b_{0}=0; if no leading scalar wave enters, b0b_{0} therefore remains zero along a smooth ray. The scalar equation at order ϵ0\epsilon^{0} then has the first inhomogeneous source,

iDkb1=a(0)μν∇μ∇νχ0(b0=0).(20)iD_{k}b_{1}=a^{(0)\mu\nu}\nabla_{\mu}\nabla_{\nu}\chi_{0}\qquad(b_{0}=0). \tag*{(20)}

For this tensor-only incident condition, the screen projection of the metric equation at the same order reads

eAμν[−i2Dkaμν(1)−12□eAμν(0)+(Ua(0))μν]=0.(21)e^{\mu\nu}_{A}\left[-\frac{i}{2}D_{k}a^{(1)}_{\mu\nu}-\frac{1}{2}\Box e^{(0)}_{A\mu\nu}+(Ua^{(0)})_{\mu\nu}\right]=0. \tag*{(21)}

The apparently leading ∇χ0 k b1\nabla\chi_{0}\,k\,b_{1} source vanishes after transverse projection just as in equation (17). This does not make equation (21) a completed physical phase calculation: the transport of all physical/gauge amplitudes, the curvature and polarization projections, ray boundary conditions and the matching GR baseline remain to be evaluated. A nonzero leading incident scalar b0b_{0} instead adds a projected ∇χ0∇b0\nabla\chi_{0}\nabla b_{0} source. These WKB order identities are consistent with geometric-optics studies of luminal scalar–tensor propagation [32, 33] and are not claimed as a new universal absence of mode interaction.

The order of the retained curvature terms can be checked directly. Let rg≡GNML/c2r_{g}\equiv G_{N}M_{L}/c^{2} and use the same weak-field scalar profile as equation (27). Since

dχdϕ∣ϕ=0=1+6Q2,\left.\frac{\mathrm{d}\chi}{\mathrm{d}\phi}\right|_{\phi=0}=\sqrt{1+6Q^{2}},
χ0−χ∞MPl=sQrgr+O(rg2/r2),sQ≡2Q1+6Q21+8Q2.(22)\frac{\chi_{0}-\chi_{\infty}}{M_{\mathrm{Pl}}}=s_{Q}\frac{r_{g}}{r}+O(r_{g}^{2}/r^{2}),\qquad s_{Q}\equiv\frac{2Q\sqrt{1+6Q^{2}}}{1+8Q^{2}}. \tag*{(22)}

Using the trace of equation (8), the leading exterior Einstein-frame Ricci scalar is

RE=(∇χ0)2MPl2=4Q2(1+6Q2)(1+8Q2)2rg2r4+O(rg3/r5).(23)R_{E}=\frac{(\nabla\chi_{0})^{2}}{M_{\mathrm{Pl}}^{2}}=\frac{4Q^{2}(1+6Q^{2})}{(1+8Q^{2})^{2}}\frac{r_{g}^{2}}{r^{4}}+O(r_{g}^{3}/r^{5}). \tag*{(23)}

The apparently leading ∇χ0 k b1\nabla\chi_{0}\,k\,b_{1} source vanishes after transverse projection just as in equation (17). This does not make equation (21) a completed physical phase calculation: the transport of all physical/gauge amplitudes, the curvature and polarization projections, ray boundary conditions and the matching GR baseline remain to be evaluated. A nonzero leading incident scalar b0b_{0} instead adds a projected ∇χ0∇b0\nabla\chi_{0}\nabla b_{0} source. These WKB order identities are consistent with geometric-optics studies of luminal scalar–tensor propagation [32, 33] and are not claimed as a new universal absence of mode interaction.

Axial perturbations and propagation diagnostics

The calculation below differs from the curvature-coupled phenomenological wave-optics model of [36]: its axial potential follows from the specified conformal action and an exact radial-scalar exterior. It also addresses a different limit from the coupled beyond-geometrical-optics propagation treatments of [34, 35]. Our independent analytic result is the mass- and radius-matched R−5R^{-5} coefficient in the odd-parity potential, together with its multipole sign change; the radial WKB, geodesic and identifiability calculations test the interpretation of that result. The local potential, the fixed-ℓ\ell WKB action, and a fully matched two-image transfer function are distinct quantities. Thus the new asymptotic coefficient is presented as a reproducible exterior tensor-mode prediction, not as a detector-level phase measurement.

Model and scalar-charge normalization

We choose the luminal conformal subclass

G2=X,G3=G5=0,G4=MPl22exp⁡(2QϕMPl).(24)G_{2}=X,\qquad G_{3}=G_{5}=0,\qquad G_{4}=\frac{M_{\mathrm{Pl}}^{2}}{2}\exp\left(\frac{2Q\phi}{M_{\mathrm{Pl}}}\right). \tag*{(24)}

with matter minimally coupled to the Jordan metric. It has M∗2=MPl2e2Qϕ/MPlM_*^2 = M_{\rm Pl}^2 e^{2Q\phi/M_{\rm Pl}}, αT=0\alpha_T = 0, αM=2Qϕ˙/(HMPl)\alpha_M = 2Q\dot{\phi}/(HM_{\rm Pl}), and αB=−αM\alpha_B = -\alpha_M. We take the local asymptotic field to be ϕˉL=0\bar{\phi}_L = 0 by choice of normalization, work to first order in the weak-field potential rg/r≪1r_g/r \ll1, and assume an unscreened nonrelativistic matter lens (not an isolated black-hole lens). Following standard scalar–tensor weak-field normalization [8], the Einstein-frame transformation in equation (6) gives dχ/dϕ∣ϕˉL=0=1+6Q2\left.d\chi/d\phi\right|_{\bar{\phi}_L=0} = \sqrt{1 + 6Q^2} and A(χ)=e−Qϕ(χ)/MPlA(\chi) = e^{-Q\phi(\chi)/M_{\rm Pl}}. The dimensionless weak-field scalar coupling is

α0≡MPldln⁡Adχ=−Q1+6Q2.\alpha_0 \equiv M_{\rm Pl}\frac{d\ln A}{d\chi} = -\frac{Q}{\sqrt{1+6Q^2}}.

In the massless, unscreened weak-field limit, the tensor Newton constant GbareG_{\rm bare} and locally measured Cavendish constant obey

GN=Gbare(1+2α02)=Gbare1+8Q21+6Q2.(25)G_N = G_{\rm bare}(1+2\alpha_0^2) = G_{\rm bare}\frac{1+8Q^2}{1+6Q^2}. \tag*{(25)}

The linearized scalar profile is

δϕLMPl=2Q1+6Q2GbareMLc2r.\frac{\delta\phi_L}{M_{\rm Pl}} = \frac{2Q}{1+6Q^2}\frac{G_{\rm bare}M_L}{c^2r}.

In terms of rg≡GNML/c2r_g \equiv G_NM_L/c^2 this becomes

δϕLMPl=2Q1+8Q2rgr+O(rg2/r2),\frac{\delta\phi_L}{M_{\rm Pl}} = \frac{2Q}{1+8Q^2}\frac{r_g}{r} + O(r_g^2/r^2),
δln⁡M∗2=ϵQrgr,ϵQ=4Q21+8Q2.(26)\delta\ln M_*^2 = \epsilon_Q\frac{r_g}{r}, \qquad\epsilon_Q = \frac{4Q^2}{1+8Q^2}. \tag*{(26)}

The replacement of GbareG_{\rm bare} by GNG_N must not be omitted when taking a finite-QQ upper envelope. This normalisation check does not by itself establish a complete coupled propagation phase.

Odd-parity gravitational master equation

A controlled, genuinely tensorial reduction is available if the exterior of the matter lens is static and spherically symmetric. For clarity, use Einstein-frame geometric units GE=c=1G_E = c = 1 and distinguish the Einstein-frame ADM mass in length units, mEm_E, from the locally measured Jordan-frame lens scale rg=GNML/c2r_g = G_NM_L/c^2. The exterior scalar equations (8) admit the Fisher–Janis–Newman–Winicour (FJNW) family [40]:

dsE2=−fζdt2+f−ζdr2+r2f1−ζdΩ2,(27)ds_E^2 = -f^\zeta dt^2 + f^{-\zeta}dr^2 + r^2f^{1-\zeta}d\Omega^2, \tag*{(27)}
f(r)=1−Br,B=2mEζ,f(r) = 1-\frac{B}{r}, \qquad B = \frac{2m_E}{\zeta},
R(r)=rf(1−ζ)/2,0<ζ≤1,(28)R(r) = rf^{(1-\zeta)/2}, \qquad0 < \zeta\le1, \tag*{(28)}
χ0−χ∞MPl=±1−ζ22ln⁡f(r).(29)\frac{\chi_0-\chi_\infty}{M_{\rm Pl}} = \pm\sqrt{\frac{1-\zeta^2}{2}}\ln f(r). \tag*{(29)}

Here RR is the circumferential (areal) radius, and r>Br > B is the exterior coordinate range. The formal continuation to r=Br = B is singular for ζ<1\zeta< 1; it is not treated as the interior of a physical galactic or stellar lens. The use of this background for an extended matter lens requires that the actual ray pass through its matter-free exterior and that an appropriate interior solution be matched at the source surface.

Because the background scalar is even under parity, a first-order odd-parity (axial) gravitational perturbation has no odd scalar-field partner. For ℓ≥2\ell\ge2 the gauge-invariant axial master function obeys the known static-spherical reduction [39, 40],

d2Zℓodddr∗2+[ω2−Vℓodd(R)]Zℓodd=0,dr∗dr=f−ζ,(30)\frac{d^2Z_\ell^{\rm odd}}{dr_*^2}+\left[\omega^2-V_\ell^{\rm odd}(R)\right]Z_\ell^{\rm odd}=0,\qquad\frac{dr_*}{dr}=f^{-\zeta}, \tag*{(30)}

with (using the Misner–Sharp mass $m(R)=\frac{R}{2}[1-g_E^{RR}])

Vℓodd=fζ[ℓ(ℓ+1)R2−6m(R)R3+4π(ρE−pr,E)].(31)V_\ell^{\rm odd}=f^\zeta\left[\frac{\ell(\ell+1)}{R^2}-\frac{6m(R)}{R^3}+4\pi\left(\rho_E-p_{r,E}\right)\right]. \tag*{(31)}

For a canonical, static, radial massless scalar, the radial pressure equals the energy density, pr,E=ρEp_{r,E}=\rho_E. Inserting R(r)R(r) then gives the explicit physical axial master potential

Vℓodd(r)=f2ζ−1r2[ℓ(ℓ+1)−3ζBr+3(1−ζ)2B24r2f].(32)V_\ell^{\rm odd}(r)=\frac{f^{2\zeta-1}}{r^2}\left[\ell(\ell+1)-\frac{3\zeta B}{r}+\frac{3(1-\zeta)^2B^2}{4r^2f}\right]. \tag*{(32)}

For ζ=1\zeta= 1 and B=2mEB = 2m_E, equation (33) reduces exactly to the Schwarzschild Regge–Wheeler potential (1−2mE/R)[ℓ(ℓ+1)/R2−6mE/R3](1-2m_E/R)[\ell(\ell+1)/R^2-6m_E/R^3]. This physical odd-parity tensor channel retains radial and angular curvature, unlike the longitudinal scalar surrogate in equation (57).

Matching the asymptotic scalar charge in equation (22) to (30) gives, for the weakly self-gravitating matter lens of equation (27),

mE=1+6Q21+8Q2rg,ζ2=1+6Q21+8Q2.(33)m_E=\frac{1+6Q^2}{1+8Q^2}r_g,\qquad\zeta^2=\frac{1+6Q^2}{1+8Q^2}. \tag*{(33)}

These are exterior asymptotic matching relations and do not specify the lens interior or hold for an object with a different effective scalar charge. The relation to the Jordan-frame post-Newtonian light-deflection parameter is particularly simple:

s≡ζ2=1+γPPN2,CQ=s−1−1=1−γPPN1+γPPN.(34)s\equiv\zeta^2=\frac{1+\gamma_{\rm PPN}}{2},\qquad C_Q=s^{-1}-1=\frac{1-\gamma_{\rm PPN}}{1+\gamma_{\rm PPN}}. \tag*{(34)}

This identity follows algebraically from the locally normalized coupling in (59); it is specific to the unscreened weak-field matter source considered here. In particular, a comparison at fixed Einstein-frame mass mEm_E is not interchangeable with one at fixed locally measured matter mass MLM_L. Expand the axial potential at fixed mEm_E and fixed areal radius RR (rather than comparing two different radial coordinates). One obtains

Vℓodd(R)=LR2−2(L+3)mER3+(9+32)mE2R4+O(mE3R5),(35)V_\ell^{\mathrm{odd}}(R) = \frac{L}{R^2} - 2(L+3)\frac{m_E}{R^3} + \left(9+\frac{3}{2}\right)\frac{m_E^2}{R^4} + O\left(\frac{m_E^3}{R^5}\right), \tag*{(35)}
ΔVℓodd(R)=Vℓodd−VℓSchw=6Q21+6Q2mE2R4+O(mE3R5).(36)\Delta V_\ell^{\mathrm{odd}}(R) = V_\ell^{\mathrm{odd}} - V_\ell^{\mathrm{Schw}} = \frac{6Q^2}{1+6Q^2}\frac{m_E^2}{R^4} + O\left(\frac{m_E^3}{R^5}\right). \tag*{(36)}

Here and below L=ℓ(ℓ+1)L = \ell(\ell+ 1). At the next radial order, the first explicit multipole dependence of the scalar-charge difference appears:

ΔVℓodd(R)∣mE=3CQmE2R4+L−93CQmE3R5+O(mE4R6).(37)\left.\Delta V_\ell^{\mathrm{odd}}(R)\right|_{m_E} = 3C_Q\frac{m_E^2}{R^4} + \frac{L-9}{3}C_Q\frac{m_E^3}{R^5} + O\left(\frac{m_E^4}{R^6}\right). \tag*{(37)}

This asymptotic coefficient is obtained by reverting the exact relation R=rf(1−ζ)/2R = r f(1-\zeta)/2 before expanding the potential; expanding directly at equal coordinate rr would give a different coefficient. The R−5R^{-5} term is negative for ℓ=2\ell= 2 and positive for ℓ≥3\ell\ge3 when Q≠0Q \ne0, whereas the displayed R−4R^{-4} term is multipole independent. This relative sign change gives a direct benchmark for independent numerical implementations of the axial equation. Figure 1 compares the exact exterior potentials with the predicted next-order limits using high-precision root finding for R(r)R(r). At Q=0.3Q = 0.3 and R=20mER = 20m_E, adding the R−5R^{-5} term reduces the relative asymptotic error from 1.69% to 0.006% for ℓ=2\ell= 2 and from 2.00% to 0.364% for ℓ=3\ell= 3. These are convergence checks of the exterior perturbation potential, rather than estimates of a detected signal.

Figure 1. Multipole-dependent next-order axial potential at fixed Einstein ADM mass (Q=0.3Q = 0.3). Solid: exact-minus-leading correction multiplied by (R/mE)5(R/m_E)^5; dotted: analytic limits (L−9)CQ/3(L - 9)C_Q/3 from Eq. (37).

Thus a parity-decoupled physical axial propagation potential contains an O(Q2)O(Q^2) correction relative to the Schwarzschild reference at fixed Einstein-frame ADM mass. This is not a claimed O(Q2)O(Q^2) observed lens-induced phase: physical two-image phases require a mapping from axial partial waves to a lensed wavepacket, the even-parity sector, consistent Jordan-frame source/detector conventions, and GR baseline subtraction with fixed observables. The result also relies on an exterior spherical ray, whereas a cluster-scale lens is generally extended and nonspherical.

Mass normalization and a model-specific consistency relation

The axial potential difference in Eq. (36) can be expressed entirely in the weak-field PPN parameter when the Einstein-frame ADM mass is held fixed:

ΔVℓodd(R)∣mE=31−γPPN1+γPPNmE2R4+O(mE3R5).(38)\left.\Delta V_\ell^{\mathrm{odd}}(R)\right|_{m_E} = 3\frac{1-\gamma_{\mathrm{PPN}}}{1+\gamma_{\mathrm{PPN}}}\frac{m_E^2}{R^4} + O\left(\frac{m_E^3}{R^5}\right). \tag*{(38)}

This relation is independent of the multipole ℓ\ell at the displayed radial order; the full potential itself remains ℓ\ell dependent. The parameter relation is a consistency check between the weak-field matter coupling and the strong-background exterior perturbation calculation, not a new independent measurement of γPPN\gamma_{\mathrm{PPN}}.

For comparison, if the same potential is re-expanded at fixed rg=GNML/c2r_g = G_{\mathrm{N}}M_L/c^2 and the Schwarzschild reference is instead assigned mass rgr_g, the two expansions give

ΔVℓodd(R)∣rg=(L+3)(1−γPPN)rgR3+(9s2+3s−12)rg2R4+O(rg3R5).(39)\left.\Delta V_\ell^{\mathrm{odd}}(R)\right|_{r_g} = (L+3)(1-\gamma_{\mathrm{PPN}})\frac{r_g}{R^3} + (9s^2+3s-12)\frac{r_g^2}{R^4} + O\left(\frac{r_g^3}{R^5}\right). \tag*{(39)}

with s=(1+γPPN)/2s = (1+\gamma_{\mathrm{PPN}})/2. The leading R−3R^{-3} term is a consequence of using different Einstein-frame masses for a fixed Jordan-frame matter mass. Equation (39) is a mass-convention diagnostic on a common Einstein-frame areal-radius coordinate, not an observed Jordan-frame lens comparison: a physical equal-mass test also requires a common operational radius and source/observer calibration. The two formulas must therefore not be combined as if they described the same subtraction.

Figure 2 illustrates the equal-mEm_E potential difference evaluated from the exact exterior solution. The weak-field coefficients in Eqs. (38) and (39) are independently checked by symbolic expansion in the accompanying verification code.

Figure 2. Relative ℓ=2\ell=2 axial-potential difference from Schwarzschild at fixed ADM mass and circumferential radius. Exact curves follow Eq. (32); the weak-exterior form is Eq. (36).

Finite-radius WKB phase of the axial mode

The local potential difference in equation (36) is not by itself a phase. We can, however, take one controlled step further within the exact exterior odd-parity master equation: define the finite-interval radial WKB action at fixed multipole ℓ\ell, fixed asymptotically normalized Einstein-frame frequency ω\omega, and fixed Einstein-frame areal radii Ra<RbR_a < R_b,

ΦℓWKB(ω;Ra,Rb;ζ)=∫RaRbJζ(R)ω2−Vℓodd dR,Jζ(R)=dr∗dR.(40)\Phi_\ell^{\mathrm{WKB}}(\omega; R_a, R_b; \zeta) = \int_{R_a}^{R_b} J_\zeta(R)\sqrt{\omega^2 - V_\ell^{\mathrm{odd}}}\,\mathrm{d}R, \qquad J_\zeta(R) = \frac{\mathrm{d}r_*}{\mathrm{d}R}. \tag*{(40)}

The square root is taken on the oscillatory branch; we require ω2>max⁡Vℓodd\omega^2>\max V_{\ell}^{\mathrm{odd}} throughout the interval, ωRa≫ℓ\omega R_a\gg\ell, and RaR_a outside any matter surface. In the FJNW exterior,

Jζ(R)=f(r)−(1+ζ)/21+(1−ζ)B/[2rf(r)],R=rf(r)(1−ζ)/2.(41)J_{\zeta}(R)=\frac{f(r)^{-(1+\zeta)/2}}{1+(1-\zeta)B/[2rf(r)]},\qquad R=rf(r)^{(1-\zeta)/2}. \tag*{(41)}

(40) is a radial partial-wave WKB diagnostic, not the asymptotic scattering phase (which also requires boundary conditions) and not the observed difference between two gravitationally lensed images. It is nevertheless defined from the physical odd-parity operator, rather than from the longitudinal scalar surrogate.

At fixed ℓ\ell the large-ω\omega difference relative to equal-mass Schwarzschild data on the same areal-radius interval has the expansion

ΔΦℓWKB=ω∫RaRbΔJ dR−12ω∫RaRbΔ(JVℓ) dR+O(ω−3),(42)\Delta\Phi_{\ell}^{\mathrm{WKB}}=\omega\int_{R_a}^{R_b}\Delta J\,\mathrm{d}R-\frac{1}{2\omega}\int_{R_a}^{R_b}\Delta(JV_{\ell})\,\mathrm{d}R+O(\omega^{-3}), \tag*{(42)}
ΔJ=Jζ−J1,\Delta J=J_{\zeta}-J_{1},
Δ(JVℓ)=JζVℓ,ζodd−J1Vℓ,1odd.\Delta(JV_{\ell})=J_{\zeta}V_{\ell,\zeta}^{\mathrm{odd}}-J_{1}V_{\ell,1}^{\mathrm{odd}}.

The first term is proportional to frequency and belongs to the achromatic optical-path part. The ω−1\omega^{-1} term is the first formal frequency-dependent contribution to this fixed-ℓ\ell radial action. Crucially, the potential must be multiplied by the optical Jacobian before comparing backgrounds: subtracting Vℓ,ζ−Vℓ,1V_{\ell,\zeta}-V_{\ell,1} alone gives a different answer.

Let CQ≡ζ−2−1=2Q2/(1+6Q2)C_Q\equiv\zeta^{-2}-1=2Q^2/(1+6Q^2), m≡mEm\equiv m_E, and L≡ℓ(ℓ+1)L\equiv\ell(\ell+1). Expanding at fixed mm and areal radius R≫mR\gg m gives

ΔJ=−CQm22R2−8CQm33R3+O(m4/R4),(43)\Delta J=-\frac{C_Qm^2}{2R^2}-\frac{8C_Qm^3}{3R^3}+O(m^4/R^4), \tag*{(43)}
Δ(JVℓ)=−CQ(L−6)m22R4−2CQ(2L−9)m33R5+O(m4/R6).(44)\Delta(JV_{\ell})=-\frac{C_Q(L-6)m^2}{2R^4}-\frac{2C_Q(2L-9)m^3}{3R^5}+O(m^4/R^6). \tag*{(44)}

Here J1Vℓ,1odd=L/R2−6m/R3J_1V_{\ell,1}^{\mathrm{odd}}=L/R^2-6m/R^3 exactly. Combining equations (42)–(44), with Δn≡Ra−n−Rb−n\Delta_n\equiv R_a^{-n}-R_b^{-n}, yields

ΔΦℓWKB=−CQωm22Δ1+CQ(L−6)m24ωΔ3+O(ωm3/Ra2)+O(m3/(ωRa4))+O(ω−3).(45)\begin{aligned} \Delta\Phi_{\ell}^{\mathrm{WKB}}={}&-\frac{C_Q\omega m^2}{2}\Delta_1+\frac{C_Q(L-6)m^2}{4\omega}\Delta_3\\ &+O(\omega m^3/R_a^2)+O(m^3/(\omega R_a^4))+O(\omega^{-3}). \tag*{(45)} \end{aligned}

For the lowest gravitational multipole ℓ=2\ell=2 (L=6L=6), the displayed m2/ωm^2/\omega term cancels even though the local ΔVℓ\Delta V_{\ell} in equation (37) is nonzero. The next exterior term in the same fixed-ℓ\ell WKB potential moment gives

ΔΦℓ,ω−1WKB∣ℓ=2=CQm34ωΔ4+O(m4/(ωRa5))+O(ω−3).(46)\left.\Delta\Phi_{\ell,\omega^{-1}}^{\mathrm{WKB}}\right|_{\ell=2}=\frac{C_Qm^3}{4\omega}\Delta_4+O(m^4/(\omega R_a^5))+O(\omega^{-3}). \tag*{(46)}

This is a cancellation in the radial WKB action under the stated matching convention; it is neither a universal cancellation of gravitational-wave dispersion nor a prediction of zero image-differential phase. The expansion ω→∞\omega\to\infty was taken at fixed ℓ\ell; it must not be interchanged without justification with the high-ℓ\ell limit ℓ∼ωb\ell\sim\omega b relevant for a finite-impact-parameter lensing wavepacket. Any source/observer master-field conversion, interior matching, even-parity response and endpoint phases must also be specified before extracting an observable.

Figure 3 compares the exact radial potential moments with their exterior expansions. For an independent check we evaluate equation (41) through its local momentum, using the exact FJNW R(r)R(r) relation and VℓV_{\ell}, and separately integrate the two moments in equation (43). At Q=0.3Q=0.3, Ra=80mR_a=80m, Rb=2000mR_b=2000m, the numerical weighted-potential moments in units of m−1m^{-1} are −1.4700×10−9-1.4700\times10^{-9} for ℓ=2\ell=2 and −2.3558×10−7-2.3558\times10^{-7} for ℓ=3\ell=3. The distinct radial scaling and the ℓ=2\ell=2 cancellation agree with equation (45). The finite-ω\omega local-momentum quadrature at ωm=5\omega m=5 agrees with the leading ω−1\omega^{-1} term to within the explicitly tabulated remainder in the accompanying code.

Figure 3. Jacobian-weighted axial potential moment for R∈[Ra,2000mE]R\in[R_a,2000m_E] (Q=0.3Q=0.3). Solid: exact exterior; dashed: leading weak-exterior term. The quadrupole starts at a higher radial order than ℓ≥3\ell\geq3.

Fixed-impact-parameter eikonal limit and bending

The preceding fixed-ℓ\ell limit is not the impact-parameter limit of a lensed wavepacket. For a large-multipole sequence, set

b=ℓ(ℓ+1)/ω,Hζ(R)=Aζ(R)/R2,Wζ(R)=−6Aζ(R)mζ(R)/R3,Aζ=fζ.(47)b = \sqrt{\ell(\ell+1)}/\omega, \qquad H_{\zeta}(R) = A_{\zeta}(R)/R^2, \qquad W_{\zeta}(R) = -6A_{\zeta}(R)m_{\zeta}(R)/R^3, \qquad A_{\zeta} = f^{\zeta}. \tag*{(47)}

The exact axial potential is then Vℓodd=ω2b2Hζ+WζV_{\ell}^{\mathrm{odd}} = \omega^2 b^2 H_{\zeta} + W_{\zeta}. On an exterior finite annulus Ra<R<RbR_a < R < R_b obeying Kζ(R,b)≡1−b2Hζ(R)>κ>0K_{\zeta}(R,b) \equiv1-b^2H_{\zeta}(R) > \kappa> 0 (and the analogous inequality in the Schwarzschild reference), the same physical axial equation gives the controlled fixed-bb expansion

ΦannWKB=ω∫RaRbJζKζ dR−12ω∫RaRbJζWζKζ dR+O(ω−3).(48)\Phi_{\mathrm{ann}}^{\mathrm{WKB}} = \omega\int_{R_a}^{R_b} J_{\zeta}\sqrt{K_{\zeta}}\,\mathrm{d}R -\frac{1}{2\omega}\int_{R_a}^{R_b}\frac{J_{\zeta}W_{\zeta}}{\sqrt{K_{\zeta}}}\,\mathrm{d}R + O(\omega^{-3}). \tag*{(48)}
ΔΦannWKB=ωΔIgeo(b)−ΔIodd(b)2ω+O(ω−3).(49)\Delta\Phi_{\mathrm{ann}}^{\mathrm{WKB}} = \omega\Delta I_{\mathrm{geo}}(b) -\frac{\Delta I_{\mathrm{odd}}(b)}{2\omega}+O(\omega^{-3}). \tag*{(49)}

where the II’s denote precisely the first and second radial integrals in equation (48) and Δ\Delta subtracts the equal-mEm_E Schwarzschild annulus. The first integral is the geometrical optical path, while the second is an axial-master-field 1/ω1/\omega correction on a turning-point-free segment. This is not an asymptotic lens transfer function: the expansion of K−1/2K^{-1/2} is nonuniform at a closest-approach turning point, and an Airy/uniform matching construction plus source and receiver data would be necessary. In particular, the ℓ=2\ell=2 cancellation in equation (46) cannot be carried over to this distinct ℓ∼ωb\ell\sim\omega b limit.

To check the leading optical term against a full ray geometry, consider an equatorial null geodesic in the exact FJNW exterior. Conformal rescaling preserves null curves, and the asymptotic normalization A(χ∞)=1A(\chi_{\infty})=1 allows the asymptotic Einstein-frame impact parameter to be used here. If r0r_0 is the exterior turning radius, the conserved energy and angular momentum imply

b=r0f(r0)(1−2ζ)/2,α^(b;ζ)+π=2∫r0∞b drr2f1−ζ1−b2f2ζ−1/r2.(50)b=r_0f(r_0)^{(1-2\zeta)/2}, \qquad \hat{\alpha}(b;\zeta)+\pi =2\int_{r_0}^{\infty}\frac{b\,\mathrm{d}r}{r^2f^{1-\zeta}\sqrt{1-b^2f^{2\zeta-1}/r^2}}. \tag*{(50)}

This null-geodesic result is an achromatic geometrical-optics property, not a spin-two finite-frequency dispersive phase; the FJNW geometry and its lensing properties are established in the literature [40, 45]. Nevertheless, it permits a useful exact-to-asymptotic check of the earlier point-lens interpretation.

For x=mE/b≪1x=m_E/b\ll1, set ϵ=B/r0\epsilon=B/r_0 and u=r0/ru=r_0/r. Then equation (50) becomes

α^+π=2∫01(1−ϵ)(1−2ζ)/2(1−ϵu)ζ−1 du1−u2[(1−ϵu)/(1−ϵ)]2ζ−1.(51)\hat{\alpha}+\pi =2\int_0^1 \frac{(1-\epsilon)^{(1-2\zeta)/2}(1-\epsilon u)^{\zeta-1}\,\mathrm{d}u} {\sqrt{1-u^2[(1-\epsilon u)/(1-\epsilon)]^{2\zeta-1}}}. \tag*{(51)}

Expanding this integrand in ϵ\epsilon and using the elementary integrals

∫01du1−u2(1+u)=1,∫01du1−u2(1+u)2=23,\int_0^1\frac{\mathrm{d}u}{\sqrt{1-u^2}(1+u)}=1, \qquad \int_0^1\frac{\mathrm{d}u}{\sqrt{1-u^2}(1+u)^2}=\frac{2}{3},

yields

α^=2ζϵ+(πζ2−π16−2ζ2+ζ)ϵ2+O(ϵ3),(52)\hat{\alpha}=2\zeta\epsilon+\left(\pi\zeta^2-\frac{\pi}{16}-2\zeta^2+\zeta\right)\epsilon^2+O(\epsilon^3), \tag*{(52)}
ϵ=2xζ+4ζ−2ζ2x2+O(x3).(53)\epsilon=\frac{2x}{\zeta}+\frac{4\zeta-2}{\zeta^2}x^2+O(x^3). \tag*{(53)}

Therefore, at fixed Einstein-frame ADM mass mEm_E and asymptotic bb,

α^(b;ζ)=4mEb+π4(16−ζ−2)mE2b2+O(mE3/b3),(54)\hat{\alpha}(b;\zeta)=\frac{4m_E}{b}+\frac{\pi}{4}(16-\zeta^{-2})\frac{m_E^2}{b^2}+O(m_E^3/b^3), \tag*{(54)}
α^(b;Q)−α^(b;0)=−πQ22(1+6Q2)mE2b2+O(mE3/b3).(55)\hat{\alpha}(b;Q)-\hat{\alpha}(b;0) =-\frac{\pi Q^2}{2(1+6Q^2)}\frac{m_E^2}{b^2}+O(m_E^3/b^3). \tag*{(55)}

The leading 4mE/b4m_E/b term is equal to GR in this Einstein-ADM-mass-matched comparison; scalar charge first enters this bending comparison at second post-Minkowskian order, not at a frequency-dependent 1/ω1/\omega order. In the distinct, physically measured Jordan-frame mass convention mE=[(1+6Q2)/(1+8Q2)]rgm_E=[(1+6Q^2)/(1+8Q^2)]r_g, the leading term becomes 2(1+γPPN)rg/b2(1+\gamma_{\mathrm{PPN}})r_g/b, agreeing with equation (61). One must not mix these two mass-matching conventions when declaring a leading-order correction. In particular, neither equation (56) nor the annular coefficient in equation (50) is an observed two-image GW chromatic response.

Direct integration of equation (51) at Q=0.1,0.3,0.6Q=0.1,0.3,0.6 independently approaches the coefficient −π(ζ−2−1)/4-\pi(\zeta^{-2}-1)/4 as b/mEb/m_E increases; Figure 4 gives the GR-subtracted, impact-rescaled comparison. The same audit independently tests equation (50) over ℓ=20,40,80,160\ell=20,40,80,160, two impact parameters and three couplings in exterior, turning-point-free intervals. These tests show that the optical-subtracted axial WKB residual scales as 1/ω1/\omega within the defined annulus; they neither perform a partial-wave sum nor furnish a two-image likelihood.

FJNW null bending: Einstein-ADM-mass matched

Figure 4. FJNW-minus-Schwarzschild null-geodesic bending, scaled by (b/mE)2(b/m_E)^2, for equal Einstein ADM mass and impact parameter. Dashed: analytic asymptote −π(ζ−2−1)/4-\pi(\zeta^{-2}-1)/4.

Comparison with a scalar surrogate

For comparison with the physical odd-parity master equation (31), consider the following longitudinal flux-weighted scalar equation:

h′′+(ln⁡F)′h′+kL2h=0,F=M∗2,kL=2π(1+zL)f/c.(56)h''+(\ln F)'h'+k_L^2h=0, \qquad F=M_*^2, \qquad k_L=2\pi(1+z_L)f/c. \tag*{(56)}

Here a prime means a derivative along the straight ray, and zLz_L is the lens redshift. The rescaling u=Fhu=\sqrt{F}h yields a formal potential VF=12(ln⁡F)′′+14[(ln⁡F)′]2V_F=\frac{1}{2}(\ln F)^{\prime\prime}+\frac{1}{4}[(\ln F)^\prime]^2. In first-order WKB, with vanishing derivative boundary term, δΨ1D=−[8kL]−1∫[(ln⁡F)′]2 dz\delta\Psi_{\mathrm{1D}}=-[8k_L]^{-1}\int[(\ln F)^\prime]^2\,\mathrm{d}z. For r2=b2+z2r^2=b^2+z^2 and equation (27),

δΨ1D(f,b)=−crg28(1+zL)fb3κQ,κQ=Q4(1+8Q2)2,δΨ=β(f0/f).(57)\delta\Psi_{\mathrm{1D}}(f,b)=-\frac{c r_g^2}{8(1+z_L)f b^3}\kappa_Q, \qquad \kappa_Q=\frac{Q^4}{(1+8Q^2)^2}, \qquad \delta\Psi=\beta(f_0/f). \tag*{(57)}

Here f0f_0 is a chosen reference frequency and β\beta a dimensionless phenomenological phase coefficient. These results assume longitudinal dynamics and cannot be promoted to a Horndeski tensor prediction. Indeed, applying the same F(x)F(x) to a distinct three-dimensional flat-space scalar action leads after the rescaling u=Fhu=\sqrt{F}h to

V3D=12∇2ln⁡F+14∣∇ln⁡F∣2,δΨ3Dsc=4δΨ1D,(58)V_{\mathrm{3D}}=\frac{1}{2}\nabla^2\ln F+\frac{1}{4}\lvert\nabla\ln F\rvert^2, \qquad \delta\Psi^{\mathrm{sc}}_{\mathrm{3D}}=4\delta\Psi_{\mathrm{1D}}, \tag*{(58)}

for the exterior 1/r1/r profile and b>0b>0. The factor four is a counterexample to surrogate robustness, not a physical theory error bar. The standard weak-field massless scalar–tensor result [8, 41], with the same locally normalized coupling, is

γPPN=1+4Q21+8Q2,κQ=(1−γPPN)216.(59)\gamma_{\mathrm{PPN}}=\frac{1+4Q^2}{1+8Q^2}, \qquad \kappa_Q=\frac{(1-\gamma_{\mathrm{PPN}})^2}{16}. \tag*{(59)}

For the illustrative scale estimate we impose ∣γPPN−1∣≤10−4\lvert\gamma_{\mathrm{PPN}}-1\rvert\leq10^{-4}, not a new Cassini confidence bound. The Cassini measurement [46] and this algebraic identity provide a conditional surrogate scale, not an observational limit on a full tensor coupling. Figure 5 shows only the frequency dependence of the assumed scalar-surrogate phase. The plotting symbol ΥL\Upsilon_L labels the benchmark curves and is not the action coupling QQ or a measured Δβ\Delta\beta; the latter is introduced as an independent parameter in section 5. We now turn from this conditional phase illustration to the achromatic lens geometry that follows directly from the specified weak-field matter coupling.

Weak-field lens geometry and mass identifiability

The lens geometry used to compare two paths is itself a physical prediction of the specified unscreened model. To first post-Newtonian order around a weakly self-gravitating, static, approximately isolated mass, write the Jordan-frame metric in asymptotically isotropic coordinates as

dsJ2=−(1−2u)c2dt2+(1+2γPPNu)dx2+O(u2),u=GNMLc2r.(60)\mathrm{d}s_J^2=-(1-2u)c^2\mathrm{d}t^2+(1+2\gamma_{\mathrm{PPN}}u)\mathrm{d}\boldsymbol{x}^2+O(u^2), \qquad u=\frac{G_NM_L}{c^2r}. \tag*{(60)}

Null rays in this metric have refractive index n=1+(1+γPPN)u+O(u2)n=1+(1+\gamma_{\mathrm{PPN}})u+O(u^2), yielding the physical small-angle deflection

α^(b)=2(1+γPPN)GNMLc2b+O(u2).(61)\hat{\alpha}(b)=\frac{2(1+\gamma_{\mathrm{PPN}})G_NM_L}{c^2b}+O(u^2). \tag*{(61)}

For an ideal thin point lens with angular-diameter distances DOLD_{\mathrm{OL}}, DOSD_{\mathrm{OS}}, DLSD_{\mathrm{LS}}, the corresponding Einstein radius obeys

(REPN)2=2(1+γPPN)GNMLDOLDLSc2DOS,REPNREGR=(1+γPPN)/2.(62)(R_E^{\mathrm{PN}})^2=\frac{2(1+\gamma_{\mathrm{PPN}})G_NM_LD_{\mathrm{OL}}D_{\mathrm{LS}}}{c^2D_{\mathrm{OS}}}, \qquad \frac{R_E^{\mathrm{PN}}}{R_E^{\mathrm{GR}}}=\sqrt{(1+\gamma_{\mathrm{PPN}})/2}. \tag*{(62)}

At fixed dimensionless source position y≡βsrc/θEPNy\equiv\beta_{\mathrm{src}}/\theta_E^{\mathrm{PN}} the image positions x±=(y±y2+4)/2x_\pm=(y\pm\sqrt{y^2+4})/2 and signed magnifications take their usual point-lens forms. Define the positive delay-shape function T(y)≡yy2+4/2+ln⁡[(y2+4+y)/(y2+4−y)]T(y)\equiv y\sqrt{y^2+4}/2+\ln[(\sqrt{y^2+4}+y)/(\sqrt{y^2+4}-y)]. The achromatic image arrival-time difference, derived from the same Fermat potential, is

ΔtPN=2(1+γPPN)GNMLc3(1+zL)T(y),ΔtPNΔtGR=(1+γPPN)/2.(63)\Delta t_{\mathrm{PN}}=\frac{2(1+\gamma_{\mathrm{PPN}})G_NM_L}{c^3}(1+z_L)T(y), \qquad \frac{\Delta t_{\mathrm{PN}}}{\Delta t_{\mathrm{GR}}}=(1+\gamma_{\mathrm{PPN}})/2. \tag*{(63)}

The formulas hold at first post-Newtonian order in a massless unscreened model with weakly self-gravitating lens matter; they are not a full dynamical scalar–tensor lens solution. At fixed yy and lens mass, the relative time-delay shift is (γPPN−1)/2(\gamma_{\mathrm{PPN}}-1)/2, whereas the relative Einstein-radius shift is (1+γPPN)/2−1\sqrt{(1+\gamma_{\mathrm{PPN}})/2}-1. Under the illustrative condition ∣γPPN−1∣≤10−4\lvert\gamma_{\mathrm{PPN}}-1\rvert\leq10^{-4}, the delay shift is no larger than 5×10−55\times10^{-5} and the Einstein-radius shift no larger than approximately 2.51×10−52.51\times10^{-5} in magnitude. This is an action-linked, leading O(Q2)O(Q^2) achromatic lensing channel, distinct from the O(Q4)O(Q^4) corrections predicted by the scalar phase surrogates. It is not independently informative about QQ without an external calibration of lens mass, profile and distances. For consistency, if the same weak-field surrogate phase expression is evaluated at the PN-corrected impact parameters b±=REPN∣x±∣b_\pm=R_E^{\mathrm{PN}}\lvert x_\pm\rvert instead of the GR reference parameters, its coefficient changes as

CdPN(ML,γPPN)=CdGR(ML)(21+γPPN)3/2,d∈{1D,3D}.(64)C_d^{\mathrm{PN}}(M_L,\gamma_{\mathrm{PPN}})=C_d^{\mathrm{GR}}(M_L)\left(\frac{2}{1+\gamma_{\mathrm{PPN}}}\right)^{3/2}, \qquad d\in\{1\mathrm{D},3\mathrm{D}\}. \tag*{(64)}
Illustrative 1/f phase curves in the ET and LISA frequency bands

Figure 5. Illustrative 1/f1/f phase curves in the ET (a) and LISA (b) frequency bands. The labels ΥL=0.10,0.115,0.01\Upsilon_L = 0.10, 0.115, 0.01 specify illustrative surrogate inputs, not inferred values of the coupling QQ. The zero GR-excess reference is absent from the logarithmic vertical scale.

This is only a geometric consistency rescaling of the surrogate, not a recovered Horndeski propagation kernel. In the near-GR limit the correction to the surrogate’s leading Q4Q^4 term is O(Q6)O(Q^6). It is inappropriate, however, to use GR impact parameters together with an unrestricted finite-QQ envelope without acknowledging this geometrical difference.

Achromatic mass–gravity degeneracy

The first-PN rescaling in section 4 does not imply that the lens geometry alone measures the fundamental scalar coupling. In a point-lens model with known angular-diameter distances and a constant weak-field γPPN\gamma_{\mathrm{PPN}}, introduce the effective lensing mass

Mlens=ML(1+γPPN)/2,\mathcal{M}_{\mathrm{lens}} = M_L(1+\gamma_{\mathrm{PPN}})/2,
θE2=4GNMlensDLSc2DOLDOS,\theta_E^2 = \frac{4G_N\mathcal{M}_{\mathrm{lens}}D_{\mathrm{LS}}}{c^2D_{\mathrm{OL}}D_{\mathrm{OS}}},
Δt=4GNMlensc3(1+zL)T(y).(65)\Delta t = \frac{4G_N\mathcal{M}_{\mathrm{lens}}}{c^3}(1+z_L)\mathcal{T}(y). \tag*{(65)}

The usual point-lens equation is βsrc=θ−θE2/θ\beta_{\mathrm{src}}=\theta-\theta_E^2/\theta; its image positions and signed magnifications depend on θE\theta_E and the source angle βsrc\beta_{\mathrm{src}}, hence on Mlens\mathcal{M}_{\mathrm{lens}} rather than separately on (ML,γPPN)(M_L,\gamma_{\mathrm{PPN}}). The same is true of the image-arrival delay because both the gravitational and geometric pieces enter through the same Fermat potential. This statement holds whether the unknown source angle or the dimensionless y=βsrc/θEy=\beta_{\mathrm{src}}/\theta_E is used as a nuisance parameter: one can vary MLM_L and γPPN\gamma_{\mathrm{PPN}} while holding Mlens\mathcal{M}_{\mathrm{lens}} and βsrc\beta_{\mathrm{src}} fixed.

Rank-deficiency statement. Within the stated first-PN, static, isolated point-lens model, and without an independent dynamical mass estimate, the collection of achromatic image angles, magnification ratios and time delays cannot identify MLM_L and γPPN\gamma_{\mathrm{PPN}} separately. To see this algebraically, set m=ln⁡MLm=\ln M_L, g=γPPNg=\gamma_{\mathrm{PPN}} and λm=ln⁡Mlens=m+ln⁡[(1+g)/2]\lambda_m=\ln\mathcal{M}_{\mathrm{lens}}=m+\ln[(1+g)/2]. Any lens-only likelihood of the form Llens(λm,βsrc,…)\mathcal{L}_{\mathrm{lens}}(\lambda_m,\beta_{\mathrm{src}},\ldots) depends on (m,g)(m,g) only through λm\lambda_m. Conditional on fixed other parameters, if IλmI_{\lambda_m} denotes its Fisher information for λm\lambda_m, the corresponding block is

Γ(m,g)∝(1(1+g)−1(1+g)−1(1+g)−2),\Gamma_{(m,g)} \propto \begin{pmatrix} 1 & (1+g)^{-1}\\ (1+g)^{-1} & (1+g)^{-2} \end{pmatrix},
det⁡Γ(m,g)=0,δm=−δg1+g.(66)\det\Gamma_{(m,g)}=0,\qquad\delta m=-\frac{\delta g}{1+g}. \tag*{(66)}

The last equality gives its exact local null direction. Additional image-angle or time-delay measurements of the same ideal point lens can improve the estimate of λm\lambda_m, but cannot turn this block into a full-rank two-parameter measurement. In particular, the point-lens ratio

ΔtRE2=1+zLcDOSDOLDLST(y)(67)\frac{\Delta t}{R_E^2}=\frac{1+z_L}{c}\frac{D_{\mathrm{OS}}}{D_{\mathrm{OL}}D_{\mathrm{LS}}}\mathcal{T}(y) \tag*{(67)}

is independent of MLM_L and γPPN\gamma_{\mathrm{PPN}} once distances and yy are specified. This mass–potential degeneracy is related to the well-known need for independent stellar kinematics or other mass information in strong-lens tests of the PPN parameter [47]. Equation (66) specializes the established mass–potential degeneracy to the present point-lens identifiability experiment.

External mass information. Suppose a lensing analysis delivers an approximately Gaussian independent estimate of λm\lambda_m with standard deviation sλs_\lambda, and an external dynamical or other non-lensing measurement delivers an independent Gaussian prior on mm with standard deviation sms_m. Linear error propagation, or inversion of the two-measurement Fisher matrix, then yields

σg=(1+g)sλ2+sm2,\sigma_g=(1+g)\sqrt{s_\lambda^2+s_m^2},
σg∣g=1=2sλ2+sm2≥2sm.(68)\sigma_g\big|_{g=1}=2\sqrt{s_\lambda^2+s_m^2}\ge2s_m. \tag*{(68)}

The uncertainties here are fractional logarithmic mass errors and a local Gaussian approximation to gg; they are not the result of fitting an observed lens. For illustration, if sm=10−2s_m=10^{-2}, 10−310^{-3} or 10−410^{-4}, the ideal sλ→0s_\lambda\to0 floors are, respectively, σg=2×10−2\sigma_g = 2 \times10^{-2}, 2×10−32 \times10^{-3} or 2×10−42 \times10^{-4}. A statistical target σg≲10−5\sigma_g \lesssim10^{-5} around GR would require sm≲5×10−6s_m \lesssim5 \times10^{-6} even if the achromatic lens strength were measured perfectly. These idealized floors make no claim about achievable stellar-kinematic accuracies or extragalactic constraints; realistic extended lenses introduce further model and distance uncertainties. Existing extra-galactic analyses break the degeneracy by combining lensing with dynamics or additional distance information [47].

In the same specified conformal theory, Q2=(1−g)/(8g−4)Q^2 = (1-g)/(8g-4) with g∈(1/2,1]g \in(1/2,1]. Near Q=0Q=0, 1−g=4Q2+O(Q4)1-g=4Q^2+O(Q^4) and hence ∂g/∂Q=0\partial g/\partial Q=0. A regular Gaussian conversion σQ=σg/∣∂g/∂Q∣\sigma_Q=\sigma_g/|\partial g/\partial Q| therefore fails at GR. If one includes a validated chromatic phase prediction in the future, its additional parameter dependence could in principle break the achromatic mass degeneracy, but the scalar surrogates in section 3.5 are not such a validated signal. The existing Δβ\Delta\beta Fisher calculation cannot, by itself, be converted into a bound on QQ.

Source-branch consistency

The source-generation question cannot be settled by assigning an arbitrary f−1f^{-1} coefficient to the binary. For the same conformal action, the exterior Einstein-frame equations (8) admit the exact branch

χ=χ∞=constant,RμνE=0,\chi=\chi_\infty=\text{constant},\qquad R^E_{\mu\nu}=0,
gμνJ=A2(χ∞)gμνE.(69)g^J_{\mu\nu}=A^2(\chi_\infty)g^E_{\mu\nu}. \tag*{(69)}

Consequently every vacuum GR solution is also a solution on this constant-scalar branch, up to the constant metric rescaling. A vacuum black-hole binary selected on this branch, with no incoming scalar radiation and constant asymptotic scalar, has the same vacuum binary evolution and tensor waveform dynamics as its GR counterpart; the action does not require a new scalar-generation f−1f^{-1} phase for that source. This known GR branch is used here as a source-model consistency check, not claimed as a new no-hair theorem. It is an existence statement about a specified solution branch, not a theorem that all black-hole binaries in arbitrary scalar environments coincide with GR. Cosmological scalar drift, nontrivial initial scalar data and matter environments require separate treatment [8, 41].

By contrast, the lens considered below is a weakly self-gravitating matter distribution. Variation of the matter action Sm[A2(χ)gE,ψm]S_m[A^2(\chi)g^E,\psi_m] gives schematically in the lens interior

□Eχ=−α(χ)MPlTE,α(χ)=MPldln⁡Adχ.(70)\Box_E\chi=-\frac{\alpha(\chi)}{M_{\mathrm{Pl}}}T_E,\qquad\alpha(\chi)=M_{\mathrm{Pl}}\frac{d\ln A}{d\chi}. \tag*{(70)}

so matter can source a quasistatic scalar profile even though a distant vacuum binary is on the constant-scalar branch. Here TET_E is the Einstein-frame matter stress-tensor trace, with conventions fixed by this variation. The point-mass lens is therefore a thin-lens idealization of ordinary matter, not a claim that an isolated black hole of the same mass carries the scalar charge in equation (27). For a distant lens, it can be useful to approximate the black-hole generation region by the constant-scalar vacuum branch while retaining a matter-sourced scalar profile near the lens. This is a scale-separation approximation, not an exact global solution with a strictly constant scalar in one open region and a nontrivial scalar everywhere else; the lens’s physical GR-subtracted tensor transfer function remains uncomputed. For a resolved image pair, the physical quantity ultimately required is the frequency-dependent difference δϕ+−(f)=arg⁡[T+(f)/T−(f)]−arg⁡[T+GR(f)/T−GR(f)]\delta\phi_{+-}(f)=\arg[T_+(f)/T_-(f)]-\arg[T^{\mathrm{GR}}_+(f)/T^{\mathrm{GR}}_-(f)], defined with a common source, observer and image-labeling prescription. No equation in the present spherical or reduced Fisher analysis evaluates this quantity.

This distinction changes how the following Fisher comparison should be interpreted. The unrestricted common source-phase coefficient γ−1\gamma_{-1} is a phenomenological stress-test nuisance, not a prediction of equation (25) for the constant-scalar vacuum black-hole branch. On that branch it is consistent to set the additional source-generation coefficient to zero, so the advertised exact single-image source-versus-lens degeneracy need not arise. A two-image difference still offers a separately defined path comparison, but there is no demonstrated model-specific gain over an ordinary event. For neutron-star or mixed binaries, nontrivial matter sensitivities and source-generation waveforms must instead be derived before any coupling-level comparison.

The mass-normalized lens map thus determines the achromatic baseline, while the source-branch analysis shows why a same-power source phase cannot be assumed to occur in every binary. The next section isolates the statistical identifiability question without treating such a phenomenological phase as an action-derived signal.

Two-image phase identifiability and Fisher diagnostics

To test a potential degeneracy without assuming a specific binary-generation mechanism, introduce an unrestricted same-power common phase. As shown in section 4.2, this is not compulsory for a vacuum black-hole source on the constant-scalar branch. The complex waveform is written as

h~j(f)=FjGR(f;λ)Tcom(f)h~GR(f;θs)exp⁡[i(γ−1+βj)f0f].(71)\tilde h_j(f)=F_j^{\mathrm{GR}}(f;\lambda)T_{\mathrm{com}}(f)\tilde h_{\mathrm{GR}}(f;\theta_s) \exp\left[i\left(\gamma_{-1}+\beta_j\right)\frac{f_0}{f}\right]. \tag*{(71)}

where γ−1\gamma_{-1} describes a common source-side (or common-propagation) same-power phase, while βj\beta_j denotes the excess phase along lens path jj. Other source corrections are not claimed to vanish; the model isolates a possible exact same-power degeneracy. For one image, only cj=γ−1+βjc_j=\gamma_{-1}+\beta_j can be measured, because ∂γ−1h~j=∂βjh~j=i(f0/f)h~j\partial_{\gamma_{-1}}\tilde h_j=\partial_{\beta_j}\tilde h_j=i(f_0/f)\tilde h_j. The single-image Fisher matrix for these two unrestricted coefficients has rank one.

For two images define

ccom=γ−1+12(β++β−),Δβ=β+−β−.(72)c_{\mathrm{com}}=\gamma_{-1}+\frac{1}{2}(\beta_++\beta_-),\qquad\Delta\beta=\beta_+-\beta_-. \tag*{(72)}

In the local Gaussian-noise Fisher approximation [42, 43, 44], if noise from the two data segments is independent and the GR transfer functions are fitted, the shared phase factor cancels formally from the image ratio. With all other parameters held fixed, set Ij=(if0h~j/f∣if0h~j/f)>0I_j = (if_0\tilde{h}_j/f \mid if_0\tilde{h}_j/f) > 0. Then

Γ(ccom,Δβ)=(I++I−(I+−I−)/2(I+−I−)/2(I++I−)/4),(73)\Gamma_{(c_{\mathrm{com}},\Delta\beta)} = \begin{pmatrix} I_+ + I_- & (I_+ - I_-)/2 \\ (I_+ - I_-)/2 & (I_+ + I_-)/4 \end{pmatrix}, \tag*{(73)}
det⁡Γ=I+I−>0.\det\Gamma= I_+I_- > 0.

and

σΔβ2=I+−1+I−−1.(74)\sigma_{\Delta\beta}^{2} = I_+^{-1} + I_-^{-1}. \tag*{(74)}

This is a quantitative rank/identifiability benefit, not evidence that a lensed pair offers a stronger coupling bound than an unlensed event. A full comparison must use a shared action-level generation and propagation model with intrinsic, lens and detector nuisance parameters. These are not marginalized in (75). Importantly, observing both images requires an astrophysically feasible image delay.

Single-image versus two-image information. The phase information of a nominal unlensed signal at the same band and signal-to-noise ratio ρ0\rho_0 is

I0ph=ρ02∫W(f)(f0/f)2 df∫W(f) df,σideal=(I0ph)−1/2.(75)I_0^{\mathrm{ph}} = \rho_0^2 \frac{\int W(f)(f_0/f)^2\,\mathrm{d}f}{\int W(f)\,\mathrm{d}f}, \qquad\sigma_{\mathrm{ideal}} = \left(I_0^{\mathrm{ph}}\right)^{-1/2}. \tag*{(75)}

Here σideal\sigma_{\mathrm{ideal}} refers to a single identifiable phenomenological coefficient of f0/ff_0/f, with the other parameters held fixed; it is not a coupling uncertainty. In a one-image model with two freely varying same-power coefficients γ−1\gamma_{-1} and β\beta, the Fisher matrix has rank one and the individual β\beta variance is undefined. With an independent Gaussian prior of width σγ\sigma_\gamma on the common source coefficient, the marginal single-image phase uncertainty is

σβ,single2=σideal2+σγ2.(76)\sigma_{\beta,\mathrm{single}}^2 = \sigma_{\mathrm{ideal}}^2 + \sigma_\gamma^2. \tag*{(76)}

For two point-lens images, Ij=∣μj∣I0phI_j = |\mu_j|I_0^{\mathrm{ph}}, and the formal differential-phase error becomes

σΔβ,pairσideal=1∣μ+∣+1∣μ−∣=2.59002(y=1).(77)\frac{\sigma_{\Delta\beta,\mathrm{pair}}}{\sigma_{\mathrm{ideal}}} = \sqrt{\frac{1}{|\mu_+|}+\frac{1}{|\mu_-|}} = 2.59002 \qquad(y=1). \tag*{(77)}

Thus a pair does not automatically beat an ideally identifiable unlensed phase measurement in coefficient precision. Conversely, for an unrestricted common 1/f1/f source contribution it provides a finite difference estimate where one image cannot. In the restricted two-coefficient comparison, the crossover occurs when σγ/σideal=2.590022−1=2.3892\sigma_\gamma/\sigma_{\mathrm{ideal}} = \sqrt{2.59002^2-1} = 2.3892. The same algebra applies to both detector bands because both images use the same normalized spectral shape. The two observables β\beta and Δβ\Delta\beta are not the same model-dependent coupling: this crossover is a statistical information diagnostic only, not a claim of observational competitiveness for modified gravity. The resulting comparison is given in Table 1.

σideal\sigma_{\mathrm{ideal}}σΔβ\sigma_{\Delta\beta}σγ,cross\sigma_{\gamma,\mathrm{cross}}
ET1.67×10−31.67 \times10^{-3}4.33×10−34.33 \times10^{-3}3.99×10−33.99 \times10^{-3}
LISA8.73×10−78.73 \times10^{-7}2.26×10−62.26 \times10^{-6}2.09×10−62.09 \times10^{-6}

Table 1. Ideal phase errors and the source-prior crossover in the reduced model at y=1y=1. Columns refer, respectively, to the single-image error, image-difference error, and crossover prior width; none constrains QQ.

Reduced conditional waveform experiment

We use the analytical ET-D and LISA noise prescriptions and the standard point-lens magnification formula. The parameter errors are calculated from derivatives of the specified waveform and noise spectra. This is a reduced conditional calculation, not an event forecast or a fit to strain data.

For each resolved image j=±j=\pm, after removing a known standard-GR lens transfer phase, we use

h~j(f)=AMz5/6f−7/6Dlum∣μj(y)∣eiΨj(f),(78)\tilde{h}_j(f) = \mathcal{A}\frac{\mathcal{M}_z^{5/6}f^{-7/6}}{D_{\mathrm{lum}}}\sqrt{|\mu_j(y)|}e^{i\Psi_j(f)}, \tag*{(78)}
Ψj(f)=ΨN(f;Mz)+(ccom+sjΔβ2)f0f.\Psi_j(f) = \Psi_{\mathrm{N}}(f;\mathcal{M}_z) + \left(c_{\mathrm{com}}+\frac{s_j\Delta\beta}{2}\right)\frac{f_0}{f}.

Here s+=1s_+=1, s−=−1s_-=-1, and f0=10 Hzf_0=10\,\mathrm{Hz}. The Newtonian phase is

ΨN=3128(πGMzfc3)−5/3.\Psi_{\mathrm{N}} = \frac{3}{128}\left(\frac{\pi G\mathcal{M}_z f}{c^3}\right)^{-5/3}.

The nuisance parameter set is

θ=(ln⁡Dlum,ln⁡Mz,ccom,Δβ,y).(79)\boldsymbol{\theta} = (\ln D_{\mathrm{lum}},\ln\mathcal{M}_z,c_{\mathrm{com}},\Delta\beta,y). \tag*{(79)}

The image magnifications are

μ±=12±y2+22yy2+4,\mu_\pm= \frac{1}{2} \pm\frac{y^2+2}{2y\sqrt{y^2+4}},

with their absolute values used for amplitudes. In this experiment yy affects the image amplitudes, but the time-delay and Morse-phase derivatives are excluded, as are sky position, inclination, mass ratio, spins, coalescence time, coalescence phase, the full PN/IMR structure, image-identification uncertainty, and detector evolution between arrival times. These omissions must be held fixed when interpreting the covariance.

Let W(f)=f−7/3/Sn(f)W(f)=f^{-7/3}/S_n(f) and I0=∫fmin⁡fmax⁡W(f) dfI_0=\int_{f_{\min}}^{f_{\max}}W(f)\,\mathrm{d}f. Normalize the unlensed signal to the nominal high-SNR value ρ0\rho_0 and define daj(f)=∂θaln⁡h~jd_{aj}(f)=\partial_{\theta_a}\ln\tilde{h}_j. Then the independently reconstructed Fisher elements are

Γab=∑j=±ρ02∣μj∣1I0∫fmin⁡fmax⁡W(f)Re⁡ ⁣[daj(f)dbj∗(f)]df.(80)\Gamma_{ab}=\sum_{j=\pm}\rho_0^2|\mu_j|\frac{1}{I_0}\int_{f_{\min}}^{f_{\max}}W(f)\operatorname{Re}\!\left[d_{aj}(f)d_{bj}^*(f)\right]\mathrm{d}f. \tag*{(80)}

where the two image-noise realizations are taken as independent. The derivatives used in the code are

dln⁡Dlum,j=−1,dln⁡Mz,j=56−5i3ΨN,(81)d_{\ln D_{\mathrm{lum},j}}=-1,\qquad d_{\ln\mathcal{M}_{z,j}}=\frac{5}{6}-\frac{5i}{3}\Psi_N, \tag*{(81)}
dccom,j=if0f,dΔβ,j=isjf02f,(82)d_{c_{\mathrm{com},j}}=i\frac{f_0}{f},\qquad d_{\Delta\beta,j}=\frac{i s_j f_0}{2f}, \tag*{(82)}
dy,j=12∂yln⁡∣μj∣.(83)d_{y,j}=\frac{1}{2}\partial_y\ln\lvert\mu_j\rvert. \tag*{(83)}

The detector-frame inspiral cutoff follows the redshifted total binary mass, not the chirp mass. We use fmin⁡=1f_{\min}=1 Hz and fmax⁡=24.43f_{\max}=24.43 Hz for a 30+30M⊙30+30M_\odot source-frame equal-mass binary at zs=2z_s=2 (so Mz=78.35M⊙\mathcal{M}_z=78.35M_\odot), and fmin⁡=10−4f_{\min}=10^{-4} Hz, fmax⁡=7.328×10−4f_{\max}=7.328\times10^{-4} Hz for a 106+106M⊙10^6+10^6M_\odot source-frame system at the same redshift (Mz=2.61165×106M⊙\mathcal{M}_z=2.61165\times10^6M_\odot). The nominal unlensed SNRs ρ0=800\rho_0=800 (ET) and 5050 (LISA) are imposed diagnostic normalizations, not rates or guaranteed luminosity-distance predictions. The PSD prescriptions are the analytical fits cited in [24, 26] and reproduced in the code.

The explicitly inverted five-parameter results are given in Table 2. For LISA, an additional Gaussian yy prior of standard deviation 0.05 or 0.02 changes (σln⁡Dlum,σy)(\sigma_{\ln D_{\mathrm{lum}}},\sigma_y) from (0.02319,0.05791)(0.02319,0.05791) to (0.02003,0.03785)(0.02003,0.03785) or (0.01802,0.01890)(0.01802,0.01890), respectively, in the same reduced likelihood. The differential-phase uncertainty remains 2.2610×10−62.2610\times10^{-6} because in this factorized model the amplitude-only yy direction has no direct phase-gradient overlap. This numerical orthogonality is not generic when the lens time delay, finite-dimensional lens mass/profile, detector motion and source-model uncertainties are included.

Setupσln⁡Dlum\sigma_{\ln D_{\mathrm{lum}}}σln⁡Mz\sigma_{\ln\mathcal{M}_z}σϵcom\sigma_{\epsilon_{\mathrm{com}}}σΔβ\sigma_{\Delta\beta}σy\sigma_y
ET1.4481×10−31.4481 \times10^{-3}3.3535×10−53.3535 \times10^{-5}3.2610×10−33.2610 \times10^{-3}4.3263×10−34.3263 \times10^{-3}3.6197×10−33.6197 \times10^{-3}
LISA2.3191×10−22.3191 \times10^{-2}1.3003×10−31.3003 \times10^{-3}2.4735×10−62.4735 \times10^{-6}2.2610×10−62.2610 \times10^{-6}5.7915×10−25.7915 \times10^{-2}

Table 2. Table 2. Marginalized standard errors from the five-parameter Fisher matrix of Eq. (78), without a yy prior. Values refer to the reduced waveform experiment.

The analytic two-image relation in equation (74) reproduces the numerical σΔβ\sigma_{\Delta\beta} to the stated precision for this separable waveform. This provides a useful internal test of the calculation but does not elevate the conditional error to an astrophysical forecast. At the parameter level, a single event with freely varying common 1/f1/f generation and propagation amplitudes remains rank deficient; observing a resolved pair identifies the differential coefficient, but a model-specific unlensed-versus-lensed coupling comparison additionally requires an independently derived binary-generation map, which is absent from the current action-level analysis.

Arrival-time and constant-phase nuisances. The preceding five-parameter diagnostic assumes that the standard GR transfer phases, including the two geometric arrival times and image parity factors, can be removed without uncertainty. This is intentionally optimistic. Even when a lens is firmly identified, the measured signal contains the arrival-time phase 2πftj2\pi f t_j and a constant phase associated with coalescence and image Morse index. The Morse indices of standard geometric-optics images are discrete and may be determined by a trusted lens model; to quantify a deliberately conservative error budget, here we instead allow an independent continuous constant phase φj\varphi_j and an independent arrival-time parameter tjt_j for each image.

This is a nuisance stress test, not the default physically optimal lens likelihood.

For this calculation extend the argument of the waveform in equation (78) by

Ψj(f)⟼Ψj(f)+φj+2πftj,j=±,(84)\Psi_j(f)\longmapsto\Psi_j(f)+\varphi_j+2\pi f t_j,\qquad j=\pm, \tag*{(84)}

and introduce the nine parameters

θ9=(ln⁡Dlum,ln⁡Mz,ccom,Δβ,y,φ+,φ−,τ+,τ−),\boldsymbol{\theta}_9=(\ln D_{\mathrm{lum}},\ln\mathcal{M}_z,c_{\mathrm{com}},\Delta\beta,y,\varphi_+,\varphi_-,\tau_+,\tau_-),

where τj=2πfpivtj\tau_j=2\pi f_{\mathrm{piv}}t_j and fpiv=fmin⁡fmax⁡f_{\mathrm{piv} }=\sqrt{f_{\min}f_{\max}}. The new derivatives are ∂φjln⁡h~j=i\partial_{\varphi_j}\ln\tilde h_j=i and ∂τjln⁡h~j=if/fpiv\partial_{\tau_j}\ln\tilde h_j=if/f_{\mathrm{piv}}, with zero contribution to the opposite image. Source and lens parameters not appearing in the reduced waveform remain fixed. The normalizations of the two images and their frequency intervals are unchanged, and this calculation does not invoke a time-delay relation or a lens-geometry prior on t+−t−t_+-t_-.

The degradation admits an independent moment-level derivation. Let the normalized spectral weight be p(f)∝W(f)p(f)\propto W(f), with ∫p(f) df=1\int p(f)\,df=1, and define ⟨a,b⟩=∫p(f)a(f)b(f) df\langle a,b\rangle=\int p(f)a(f)b(f)\,df. Set q(f)=f/f0q(f)=f/f_0, u0(f)=1u_0(f)=1 and u1(f)=f/fpivu_1(f)=f/f_{\mathrm{piv}}. Then

Bab=⟨ua,ub⟩,B_{ab}=\langle u_a,u_b\rangle,
va=⟨q,ua⟩,v_a=\langle q,u_a\rangle,
J⊥=⟨q,q⟩−vTB−1v>0,a,b∈{0,1}.(85)J_\perp=\langle q,q\rangle-\boldsymbol{v}^{\mathsf{T}}B^{-1}\boldsymbol{v}>0,\qquad a,b\in\{0,1\}. \tag*{(85)}

J⊥J_\perp is the squared norm of the f−1f^{-1} phase direction after orthogonal projection away from the constant- and linear-frequency directions. For separable, identical normalized spectral shapes in the two images, their independent phase nuisances lead to

σΔβ,free φ,t2=1ρ02J⊥(1∣μ+∣+1∣μ−∣).(86)\sigma^2_{\Delta\beta,\mathrm{free}\,\varphi,t}=\frac{1}{\rho_0^2J_\perp}\left(\frac{1}{\lvert\mu_+\rvert}+\frac{1}{\lvert\mu_-\rvert}\right). \tag*{(86)}

Replacing the two-vector (u0,u1)(u_0,u_1) by u0u_0 alone gives the free-constant-phase result; omitting both nuisance directions returns equation (74). Equation (86) is valid for the defined conditional phase experiment, not a model-specific constraint on the covariant coupling QQ. The additional common chirp-mass phase in the nine-parameter calculation is shared between both images; the numerical inversion below verifies that it does not change the differential-mode result under this particular waveform factorization.

Table 3 shows the two-detector nuisance comparison. For each detector, the 9×99\times9 Fisher inverse and the independent Gram-matrix projection agree at numerical precision, and the frequency-integral results are stable under grid refinement. The two-sided nominal Gaussian 95% phase thresholds accordingly change from 8.48×10−38.48\times10^{-3} to 2.90×10−22.90\times10^{-2} (ET) and from 4.43×10−64.43\times10^{-6} to 2.64×10−52.64\times10^{-5} (LISA). These thresholds concern the phenomenological differential phase only. In a physical analysis, the GR image delays, parity hypotheses, detector responses at separated epochs, intrinsic waveform parameters and the source/lens priors would have to enter a common likelihood. The calculation quantifies the sensitivity loss caused by freeing these nuisance directions within the stated Fisher model.

5 par.Phase freePhase/time freeRatio
ET4.33⋅10−34.33 \cdot10^{-3}8.81⋅10−38.81 \cdot10^{-3}1.48⋅10−21.48 \cdot10^{-2}3.433.43
LISA2.26⋅10−62.26 \cdot10^{-6}5.91⋅10−65.91 \cdot10^{-6}1.35⋅10−51.35 \cdot10^{-5}5.965.96

Table 3. Table 3. Marginalized difference-phase errors with image-specific phase (φ±\varphi_{\pm}) and arrival-time (t±t_{\pm}) nuisances. Ratio is relative to the five-parameter result.

Figure 6 presents an analytic two-dimensional Gaussian summary of the recomputed Fisher covariance. Its elliptical contours express the reduced local likelihood approximation and are not numerical posterior samples.

These Fisher errors are conditioned on having both images available. The observing-window calculation in the following section tests that prerequisite separately from the formal phase precision.

Time-delay selection and observing windows

For the standard point lens [17, 19] and source position y>0y > 0,

x±=y±y2+42,b±=RE∣x±∣(87)x_{\pm} = \frac{y \pm\sqrt{y^2 + 4}}{2}, \qquad b_{\pm} = R_{\mathrm{E}} \lvert x_{\pm} \rvert \tag*{(87)}

The Einstein radius, in terms of angular-diameter distances, is

RE2=4GNMLDOLDLSc2DOS.R_{\mathrm{E}}^2 = \frac{4G_{\mathrm{N}}M_{\mathrm{L}}D_{\mathrm{OL}}D_{\mathrm{LS}}}{c^2D_{\mathrm{OS}}}.

The GR-reference observer-frame image time delay is

Δt=4GNML(1+zL)c3T(y),\Delta t = \frac{4G_{\mathrm{N}}M_{\mathrm{L}}(1+z_{\mathrm{L}})}{c^3}\mathcal{T}(y),
T(y)=yy2+42+ln⁡(y2+4+yy2+4−y).(88)\mathcal{T}(y) = \frac{y\sqrt{y^2+4}}{2} + \ln\left(\frac{\sqrt{y^2+4}+y}{\sqrt{y^2+4}-y}\right). \tag*{(88)}

At zL=0.5z_{\mathrm{L}}=0.5, zs=2z_{\mathrm{s}}=2, y=1y=1, and ML=1014M⊙M_{\mathrm{L}}=10^{14}M_{\odot}, the GR-reference delay is 194.84 years (using M⊙=1.98847×1030M_{\odot}=1.98847\times10^{30} kg and the standard physical constants). Both images would not be recorded in a four-year mission. A two-image Fisher diagnostic for this particular lens is therefore only a hypothetical both-images-available sensitivity, not a realistic event forecast.

Finite observing-window envelope

An explicit observing-window restriction can be imposed without an event-rate model. Write the positive point-lens delay function as

T(y)=yy2+42+ln⁡(y2+4+yy2+4−y).(89)\mathcal{T}(y) = \frac{y\sqrt{y^2+4}}{2} + \ln\left(\frac{\sqrt{y^2+4}+y}{\sqrt{y^2+4}-y}\right). \tag*{(89)}

Using the GR-reference geometry, requiring the two images to arrive within a window TwinT_{\mathrm{win}} yields the necessary point-lens condition

ML≤Mwin(y)=c3Twin4GN(1+zL)T(y).(90)M_{\mathrm{L}} \le M_{\mathrm{win}}(y) = \frac{c^3T_{\mathrm{win}}}{4G_{\mathrm{N}}(1+z_{\mathrm{L}})\mathcal{T}(y)}. \tag*{(90)}

Under the assumptions of equation (64), the PN-corrected mass-window bound is the GR bound multiplied by 2/(1+γPPN)2/(1+\gamma_{\mathrm{PPN}}); for ∣γPPN−1∣≤10−4\lvert\gamma_{\mathrm{PPN}}-1\rvert\leq10^{-4} this differs by no more than approximately 5.01×10−55.01\times10^{-5} relative to the GR-reference mass window. For fixed lens and source redshifts and fixed yy, the Einstein radius scales as RE∝ML1/2R_{\mathrm{E}}\propto M_{\mathrm{L}}^{1/2} and each surrogate differential phase has ∣Δβd∣=Cd(ML)κQ\lvert\Delta\beta_d\rvert=C_d(M_{\mathrm{L}})\kappa_Q with Cd∝ML1/2C_d\propto M_{\mathrm{L}}^{1/2}. Combining this with the GR-reference scaling with equation (91) gives the surrogate-only, GR-referenced window-conditioned inequality

∣Δβd∣≤Cd(Mwin)ϵPPN216,∣1−γPPN∣≤ϵPPN,d=1D,3D.(91)\lvert\Delta\beta_d\rvert\leq C_d(M_{\mathrm{win}})\frac{\epsilon_{\mathrm{PPN}}^2}{16}, \qquad\lvert1-\gamma_{\mathrm{PPN}}\rvert\leq\epsilon_{\mathrm{PPN}}, \quad d=1\mathrm{D},3\mathrm{D}. \tag*{(91)}

At zL=0.5z_{\mathrm{L}}=0.5, zs=2z_{\mathrm{s}}=2, y=1y=1, and Twin=4T_{\mathrm{win}}=4 years, equation (91) gives Mwin=2.0530×1012M⊙M_{\mathrm{win}}=2.0530\times10^{12}M_{\odot}. For the illustrative ϵPPN=10−4\epsilon_{\mathrm{PPN}}=10^{-4}, the one-dimensional bound is 3.37×10−343.37\times10^{-34} and the distinct three-dimensional scalar-surrogate bound is 1.35×10−331.35\times10^{-33}, approximately 5.95×10−285.95\times10^{-28} times the formal LISA one-sigma phase uncertainty. Using the PN time delay and PN impact parameter together would multiply the quoted GR-reference envelope by [2/(1+γPPN)]2[2/(1+\gamma_{\mathrm{PPN}})]^2 at the corresponding limiting mass; the illustrative PN criterion changes this number by at most about 1.01×10−41.01\times10^{-4}. Neither this necessary time-window condition nor the fixed-SNR Fisher model predicts an astrophysical event rate or a complete tensor-mode phase. The bound illustrates the tradeoff imposed by observing the delayed image, and it would need to be recomputed using the full covariant propagation response before being used as a modified-gravity constraint.

Continuous-window image-pair selection. The mass cap in equation (91) is necessary, not sufficient, to observe two short-lived images. With the first image uniformly distributed inside a single continuous observing window [0,Twin][0,T_{\rm win}], and conditional on both magnifications exceeding threshold, the second image is also within the window with probability

Figure 6. Reduced five-parameter LISA Fisher covariance: marginalized Gaussian densities (diagonal) and 68%/95% contours (off-diagonal), with ln⁡Mz\ln\mathcal{M}_z marginalized.

Ppair=(1−Δt/Twin)+,(x)+≡max⁡(x,0).(92)P_{\rm pair} = (1-\Delta t/T_{\rm win})_{+}, \qquad(x)_{+} \equiv\max(x,0). \tag*{(92)}

For fixed yy and zLz_L, Δt∝ML\Delta t \propto M_L. Setting x=ML/Mwinx=M_L/M_{\rm win} gives Ppair=(1−x)+P_{\rm pair}=(1-x)_{+}. A scalar-surrogate information proxy, which assumes mass-independent conditional Fisher variance and ∣Δβ∣∝ML|\Delta\beta| \propto\sqrt{M_L}, is

W(x)=x(1−x),0≤x≤1,W(x)=x(1-x), \qquad0\le x\le1,
Wmax⁡=1/4(x=1/2).(93)W_{\max}=1/4 \quad(x=1/2). \tag*{(93)}

For Twin=4 yrT_{\rm win}=4\ {\rm yr}, y=1y=1 and zL=0.5z_L=0.5, Mwin=2.0530×1012M⊙M_{\rm win}=2.0530\times10^{12}M_{\odot}. A mass at the boundary has zero pair probability in this idealized short-signal model; half this mass gives a 2 yr delay and Ppair=1/2P_{\rm pair}=1/2. Equation (94) is neither an event rate nor an information forecast for the physical tensor coupling. It excludes a source population, detector duty cycle and long-duration waveforms [30].

Discussion and conclusions

The static Einstein–scalar exterior has a genuine odd-parity tensor master equation, not a prescribed scalar flux surrogate [39, 40]. Matching the exterior scalar charge to the conformal coupling and the local Newton constant yields Eq. (36). The independently verified next-order result, Eq. (37), exposes a multipole-dependent coefficient and a sign change between the quadrupole and higher modes, providing an additional analytic check of the exact tensor potential. Equations (38) and (39) make explicit how the radial order of the difference depends on whether the Schwarzschild baseline is matched at fixed Einstein-frame mass or fixed locally inferred matter mass. Its local O(Q2)O(Q^2) term does not imply an observable O(Q2)O(Q^2) interimage phase. At fixed ℓ\ell, the Jacobian must accompany the radial WKB action; at fixed impact parameter the high-ℓ\ell limit instead contains an order-ω\omega geometric term and a distinct subleading annular correction. The FJNW-minus-Schwarzschild bending benchmark is achromatic. None of these finite-radius calculations includes the turning-point matching and wave-packet summation necessary for a lens transfer function.

The local scalar coupling also changes the weak-field lens map. Its point-mass achromatic observables constrain ML(1+γPPN)/2M_L(1+\gamma_{\rm PPN})/2, not MLM_L and QQ separately; external mass information is needed. The constant-scalar vacuum black-hole source branch is GR-like, so the freely varying common source 1/f1/f coefficient in the illustrative Fisher calculation is a stress-test parameter rather than a source prediction. Once independent image arrival times and constant phases are released, the conditional difference-phase uncertainty grows by factors 3.43 (ET) and 5.96 (LISA). With y=1y = 1 and zL=0.5z_L = 0.5, a 1014M⊙10^{14}M_\odot point lens has a delay of approximately 195 yr, inconsistent with recording a resolved pair in a four-year window.

To construct a physical, GR-subtracted two-image dispersion signal one must solve the even-parity scalar–metric system, match a realistic matter-lens interior, implement uniform turning-point matching and multi-ℓ\ell wave-packet propagation, and fit generation and propagation within the same action. This paper establishes narrower exterior and identifiability checks, not a completed observational bound on Horndeski couplings. The accompanying conditional 1/f1/f illustration is therefore interpreted strictly as an information-geometry benchmark.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (Grant Nos. 12288102, 12573052 and 12475046) and by the Major Science and Technology Special Project of Xinjiang Uygur Autonomous Region (Grant No. 2022A03013-3).

Data and code availability

The source and verification package accompanying this manuscript contains the scripts and numerical tables for the axial potential, radial and fixed-impact WKB comparisons, FJNW null-geodesic deflection, reduced Fisher matrices and finite-window selection. It also contains an independent symbolic check of the mass-normalization identities. The data are theoretical diagnostics rather than observed-event measurements. A public persistent repository link has not been assigned.

The local Newton constant in Brans–Dicke theory

For completeness consider the Jordan-frame Brans–Dicke action [6]

S=116πG∗∫d4x−g[ϕR−ωBDϕ(∇ϕ)2]+Sm.(94)S = \frac{1}{16\pi G_*}\int d^4x\sqrt{-g}\left[\phi R-\frac{\omega_{\rm BD}}{\phi}(\nabla\phi)^2\right]+S_{\rm m}. \tag*{(94)}

The field equations obtained by varying the Jordan-frame action are

ϕGμν=8πG∗c4Tμν+ωBDϕ∇μϕ∇νϕ−ωBD2ϕgμν(∇ϕ)2+∇μ∇νϕ−gμν□ϕ,(95)\phi G_{\mu\nu}=\frac{8\pi G_*}{c^4}T_{\mu\nu}+\frac{\omega_{\rm BD}}{\phi}\nabla_\mu\phi\nabla_\nu\phi-\frac{\omega_{\rm BD}}{2\phi}g_{\mu\nu}(\nabla\phi)^2+\nabla_\mu\nabla_\nu\phi-g_{\mu\nu}\Box\phi, \tag*{(95)}
(2ωBD+3)□ϕ=8πG∗c4T.(96)(2\omega_{\rm BD}+3)\Box\phi=\frac{8\pi G_*}{c^4}T. \tag*{(96)}

The scalar equation follows from the scalar variation combined with the trace of the metric equation for constant ωBD\omega_{\rm BD}. Both scalar and tensor perturbations contribute to the locally measured gravitational force, so GNG_N cannot be identified solely with G∗/ϕˉG_*/\bar{\phi}.

Writing ϕ=ϕˉ+δϕ\phi=\bar{\phi}+\delta\phi and g00=−1+2U/c2g_{00}=-1+2U/c^2 about a constant background, the static weak-field equations for a nonrelativistic source ρ\rho take the form

∇2δϕ=−8πG∗(2ωBD+3)c2ρ,(97)\nabla^2\delta\phi=-\frac{8\pi G_*}{(2\omega_{\rm BD}+3)c^2}\rho, \tag*{(97)}
∇2(U−c2δϕ2ϕˉ)=−4πG∗ϕˉρ.(98)\nabla^2\left(U-\frac{c^2\delta\phi}{2\bar{\phi}}\right)=-\frac{4\pi G_*}{\bar{\phi}}\rho. \tag*{(98)}

Here the scalar equation and the appropriate linearized metric combination follow from the action, after suppressing time derivatives and higher weak-field terms. For a point mass, using ∇2(1/r)=−4πδ(3)(x)\nabla^2(1/r)=-4\pi\delta^{(3)}(\boldsymbol{x}), one obtains

δϕ=2G∗M(2ωBD+3)c2r,U=G∗Mϕˉr(1+12ωBD+3).(99)\delta\phi=\frac{2G_*M}{(2\omega_{\rm BD}+3)c^2r},\qquad U=\frac{G_*M}{\bar{\phi}r}\left(1+\frac{1}{2\omega_{\rm BD}+3}\right). \tag*{(99)}

Comparing U=GNM/rU=G_NM/r gives

GN=G∗ϕˉ2ωBD+42ωBD+3.(100)G_N=\frac{G_*}{\bar{\phi}}\frac{2\omega_{\rm BD}+4}{2\omega_{\rm BD}+3}. \tag*{(100)}

The local scalar charge is δϕ/ϕˉ=qL(GNM/c2r)\delta\phi/\bar{\phi}=q_L(G_NM/c^2r) with qL=(ωBD+2)−1q_L=(\omega_{\rm BD}+2)^{-1}. This is a weak-field normalization identity, not a derivation of the lens-induced bGO kernel of Mena­deo and Zumalacárregui [34]; the latter requires a matched coupled wave equation and the same GR-subtracted observable.

For an independent normalization check, define Φ=e2Qϕ/MPl\Phi=e^{2Q\phi/M_{\rm Pl}} in the conformal action (25). Its canonical Jordan-frame kinetic density becomes

X=−MPl28Q2Φ2(∇Φ)2,ω(Φ)=14Q2Φ.(101)X=-\frac{M_{\rm Pl}^2}{8Q^2\Phi^2}(\nabla\Phi)^2,\qquad\omega(\Phi)=\frac{1}{4Q^2\Phi}. \tag*{(101)}

Thus this model is a variable-ω\omega scalar–tensor theory, not globally the constant-ωBD\omega_{\rm BD} model. At the local asymptotic background Φ0=1\Phi_0=1 the effective weak-field parameter is ω0=1/(4Q2)\omega_0=1/(4Q^2), giving

γPPN=ω0+1ω0+2=1+4Q21+8Q2.(102)\gamma_{\rm PPN}=\frac{\omega_0+1}{\omega_0+2}=\frac{1+4Q^2}{1+8Q^2}. \tag*{(102)}

in agreement with equation (59). These identities require a massless unscreened scalar and weakly self-gravitating sources; they are not automatically valid for screened or strongly self-gravitating objects.

Horndeski notation and the shortwave hierarchy

For reference, the four familiar covariant Horndeski densities [9, 10] in our conventions are

L2=G2(ϕ,X),(103)\mathcal{L}_2 = G_2(\phi, X), \tag*{(103)}
L3=−G3(ϕ,X)□ϕ,(104)\mathcal{L}_3 = -G_3(\phi, X)\Box\phi, \tag*{(104)}
L4=G4R+G4X[(□ϕ)2−ϕ;μνϕ;μν],(105)\mathcal{L}_4 = G_4 R + G_{4X}\left[(\Box\phi)^2-\phi_{;\mu\nu}\phi^{;\mu\nu}\right], \tag*{(105)}
L5=G5Gμνϕ;μν−G5X6(□ϕ)3+G5X2□ϕϕ;μνϕ;μν−G5X3ϕ;μ;νϕ;ν;ρϕ;ρ;μ.(106)\begin{aligned} \mathcal{L}_5 ={}& G_5G_{\mu\nu}\phi^{;\mu\nu}-\frac{G_{5X}}{6}(\Box\phi)^3 \\ &+\frac{G_{5X}}{2}\Box\phi\phi_{;\mu\nu}\phi^{;\mu\nu} -\frac{G_{5X}}{3}\phi_{;\mu}^{;\nu}\phi_{;\nu}^{;\rho}\phi_{;\rho}^{;\mu}. \tag*{(106)} \end{aligned}

The background quantities αK\alpha_K and αB\alpha_B are standard FLRW projections and, in general, are insufficient to determine the local principal symbol (5). When the coefficients change on a scale much longer than the wavelength, write ψI=∑n≥0ϵnanIeiθ/ϵ\psi^I=\sum_{n\geq0}\epsilon^n a_n^I e^{i\theta/\epsilon}. The leading equation is KIJμνkμkνa0J=0K^{\mu\nu}_{IJ}k_\mu k_\nu a_0^J=0; the next orders involve derivatives of the same background operator and of the polarization eigenvectors. For a nonluminal branch, the correct characteristic need not coincide with the metric light cone. The diagonal, luminal, one-dimensional reduction (56) is an additional approximation, not a consequence of replacing the full Horndeski principal symbol with the metric null symbol. A complete calculation must recover (56) from the action or supplement it with the omitted same-order terms.

References

  1. [1]B. P. Abbott, R. Abbott, T. D. Abbott, M. R. Abernathy, F. Acernese, et al. Tests of General Relativity with GW150914. Physical Review Letters, 116(22):221101, July 2016. doi:10.1103/PhysRevLett.116.221101.DOI
  2. [2]B. P. Abbott, R. Abbott, T. D. Abbott, et al. Tests of General Relativity with GW170817. Physical Review Letters, 123(1):011102, 2019. doi:10.1103/PhysRevLett.123.011102.DOI
  3. [3]R. Abbott, T. D. Abbott, S. Abraham, et al. Tests of General Relativity with Binary Black Holes from the second LIGO-Virgo Gravitational-Wave Transient Catalog. Physical Review D, 103(12):122002, 2021. doi:10.1103/PhysRevD.103.122002.DOI
  4. [4]B. P. Abbott, R. Abbott, T. D. Abbott, et al. GW170817: Observation of Gravitational Waves from a Binary Neutron Star Inspiral. Physical Review Letters, 119(16):161101, 2017. doi:10.1103/PhysRevLett.119.161101.DOI
  5. [5]J. M. Ezquiaga and M. Zumalacárregui. Dark Energy After GW170817: Dead Ends and the Road Ahead. Physical Review Letters, 119(25):251304, 2017. doi:10.1103/PhysRevLett.119.251304.DOI
  6. [6]C. Brans and R. H. Dicke. Mach’s principle and a relativistic theory of gravitation. Physical Review, 124:925–935, 1961. doi:10.1103/PhysRev.124.925.DOI
  7. [7]G. W. Horndeski. Second-Order Scalar-Tensor Field Equations in a Four-Dimensional Space. International Journal of Theoretical Physics, 10(6):363–384, 1974. doi:10.1007/BF01807638.DOI
  8. [8]T. Clifton, P. G. Ferreira, A. Padilla, and C. Skordis. Modified Gravity and Cosmology. Physics Reports, 513(1-3):1–189, 2012. doi:10.1016/j.physrep.2012.01.001.DOI
  9. [9]T. Kobayashi, M. Yamaguchi, and J. Yokoyama. Generalized G-Inflation: Inflation with the Most General Second-Order Field Equations. Progress of Theoretical Physics, 126(3):511–529, 2011. doi:10.1143/PTP.126.511.DOI
  10. [10]D. Bettoni, S. Liberati, and L. Sindoni. Extended Library of Horndeski Theories. Journal of Cosmology and Astroparticle Physics, 2013(11):007, 2013. doi:10.1088/1475-7516/2013/11/007.
  11. [11]Emilio Bellini and Ignacy Sawicki. Maximal freedom at minimum cost: Linear large-scale structure in general modifications of gravity. Journal of Cosmology and Astroparticle Physics, 2014(07):050, 2014. doi:10.1088/1475-7516/2014/07/050.DOI
  12. [12]Ippocratis D. Saltas, Ignacy Sawicki, Luca Amendola, and Martin Kunz. Anisotropic stress as a signature of nonstandard propagation of gravitational waves. Physical Review Letters, 113:191101, 2014. doi:10.1103/PhysRevLett.113.191101.DOI
  13. [13]Lucas Lombriser and Andy Taylor. Breaking a dark degeneracy with gravitational waves. Journal of Cosmology and Astroparticle Physics, 2016(03):031, 2016. doi:10.1088/1475-7516/2016/03/031.DOI
  14. [14]Enis Belgacem, Yves Dirian, Stefano Foffa, and Michele Maggiore. Modified gravitational-wave propagation and standard sirens. Physical Review D, 98:023510, 2018. doi:10.1103/PhysRevD.98.023510.DOI
  15. [15]Atsushi Nishizawa. Generalized framework for testing gravity with gravitational-wave propagation. I. formulation. Physical Review D, 97:104037, 2018. doi:10.1103/PhysRevD.97.104037.DOI
  16. [16]Jose Maria Ezquiaga and Miguel Zumalacárregui. Dark energy in light of multi-messenger gravitational-wave astronomy. Frontiers in Astronomy and Space Sciences, 5:44, 2018. doi:10.3389/fspas.2018.00044.DOI
  17. [17]R. Takahashi and T. Nakamura. Wave Effects on the Gravitational Lensing of Gravitational Waves from Chirping Binaries. The Astrophysical Journal, 595(2):1039–1051, 2003. doi:10.1086/377430.DOI
  18. [18]L. Dai, T. Venumadhav, and K. Sigurdson. Effect of Lens Mass Distribution on Gravitational Wave Signals. Physical Review D, 95(4):044011, 2017. doi:10.1103/PhysRevD.95.044011.
  19. [19]M. Bartelmann. Gravitational Lensing. Classical and Quantum Gravity, 27(23):233001, 2010. doi:10.1088/0264-9381/27/23/233001.DOI
  20. [20]Jose Maria Ezquiaga and Miguel Zumalacárregui. Gravitational wave lensing beyond general relativity: Birefringence, echoes, and shadows. Physical Review D, 102(12):124048, 2020. doi:10.1103/PhysRevD.102.124048.DOI
  21. [21]Jose Maria Ezquiaga, Daniel E. Holz, Wayne Hu, Macarena Lagos, and Robert M. Wald. Phase effects from strong gravitational lensing of gravitational waves. Physical Review D, 103(6):064047, 2021. doi:10.1103/PhysRevD.103.064047.DOI
  22. [22]Srashti Goyal, Aditya Vijaykumar, Jose Maria Ezquiaga, and Miguel Zumalacárregui. Probing lens-induced gravitational-wave birefringence as a test of general relativity. Physical Review D, 108(2):024052, 2023. doi:10.1103/PhysRevD.108.024052.
  23. [23]S. Savastano, O. A. Hannuksela, E. Barausse, and T. Collett. Gravitational-Wave Lensing of Binary Systems. Physical Review D, 100(10):104035, 2019. doi:10.1103/PhysRevD.100.104035.
  24. [24]S. Hild, M. Abernathy, F. Acernese, et al. Sensitivity Studies for Third-Generation Gravitational Wave Observatories. Classical and Quantum Gravity, 28(9):094013, 2011. doi:10.1088/0264-9381/28/9/094013.DOI
  25. [25]P. Amaro-Seoane, H. Audley, S. Babak, et al. Laser Interferometer Space Antenna. arXiv e-prints, art. arXiv:1702.00786, February 2017.
  26. [26]Travis Robson, Neil J. Cornish, and Chang Liu. The construction and use of lisa sensitivity curves. Class. Quant. Grav., 36(10):105011, 2019. doi:10.1088/1361-6382/ab1101.DOI
  27. [27]O. Auclair, J. Baker, E. Barausse, et al. Cosmology with the Laser Interferometer Space Antenna. Living Reviews in Relativity, 26(1):5, 2023. doi:10.1007/s41114-023-00045-2.DOI
  28. [28]Margherita Grespan and Marek Biesiada. Strong gravitational lensing of gravitational waves: A review. Universe, 9(5):200, 2023. doi:10.3390/universe9050200.
  29. [29]Andreas Finke, Stefano Foffa, Francesco Iacovelli, Michele Maggiore, and Michele Mancarella. Probing modified gravitational wave propagation with strongly lensed coalescing binaries. Phys. Rev. D, 104(8):084057, October 2021. doi:10.1103/PhysRevD.104.084057.DOI
  30. [30]Shun-Sheng Li, Shude Mao, Yuetong Zhao, and Youjun Lu. Gravitational lensing of gravitational waves: a statistical perspective. Monthly Notices of the Royal Astronomical Society, 476:2220–2229, 2018. doi:10.1093/mnras/sty411.
  31. [31]R. A. Isaacson. Gravitational Radiation in the Limit of High Frequency. I. The Linear Approximation and Geometrical Optics. Physical Review, 166(5):1263–1271, 1968. doi:10.1103/PhysRev.166.1263.DOI
  32. [32]Charles Dalang, Pierre Fleury, and Lucas Lombriser. Scalar and tensor gravitational waves. Physical Review D, 103:064075, 2021. doi:10.1103/PhysRevD.103.064075.DOI
  33. [33]Alice Garofalo, Gianmassimo Tasinato, Carmelita Carbone, Daniele Bertacca, and Sabino Matarrese. Gravitational waves and geometrical optics in scalar-tensor theories. Journal of Cosmology and Astroparticle Physics, 2020(11):040, 2020. doi:10.1088/1475-7516/2020/11/040.
  34. [34]N. Menadeo and M. Zumalacárregui. Gravitational wave propagation beyond general relativity: Geometric optic expansion and lens-induced dispersion. Physical Review D, 111(10):104022, 2025. doi:10.1103/PhysRevD.111.104022.DOI
  35. [35]Nicola Menadeo, Serena Giardino, and Miguel Zumalacárregui. Gravitational-wave dispersion over inhomogeneous space-times: General relativity, screened theories of gravity and non-minimal dark energy. Journal of Cosmology and Astroparticle Physics, 2026(09):092, 2026. doi:10.1088/1475-7516/2026/09/092.DOI
  36. [36]Alice Garofolo and Gianmassimo Tasinato. Wave-optics gravitational wave lensing in modified gravity. Physical Review D, 114:064059, 2026. doi:10.1103/smzt-5v1s.
  37. [37]G. G. L. Nashed. Odd-parity perturbations of geometrically regular black holes with hedgehog scalar hair. The European Physical Journal C, 86:1289, 2026. doi:10.1140/epjc/s10052-026-16584-1.arxiv.org/abs/2609.00016
  38. [38]Macarena Lagos, Emilio Bellini, Johannes Noller, Pedro G. Ferreira, and Tessa Baker. A general theory of linear cosmological perturbations: stability conditions, the quasi-static limit and dynamics. Journal of Cosmology and Astroparticle Physics, 2018(03):021, 2018. doi:10.1088/1475-7516/2018/03/021.DOI
  39. [39]Max Karlinovini. Axial perturbations of general spherically symmetric space-times. Classical and Quantum Gravity, 19:2125–2140, 2002. doi:10.1088/0264-9381/19/8/305.DOI
  40. [40]Avijit Chowdhury and Narayan Banerjee. Echoes from a singularity. Physical Review D, 102:124051, 2020. doi:10.1103/PhysRevD.102.124051.DOI
  41. [41]Clifford M. Will. The confrontation between general relativity and experiment. Living Reviews in Relativity, 17:4, 2014. doi:10.12942/lrr-2014-4.DOI
  42. [42]C. Cutler and E. E. Flanagan. Gravitational Waves from Merging Compact Binaries: How Accurately Can One Extract the Binary’s Parameters from the Inspiral Waveform? Physical Review D, 49(6):2658–2697, 1994. doi:10.1103/PhysRevD.49.2658.DOI
  43. [43]Michele Vallisneri. Use and abuse of the fisher information matrix in the assessment of gravitational-wave parameter-estimation prospects. Physical Review D, 77(4):042001, 2008. doi:10.1103/PhysRevD.77.042001.DOI
  44. [44]Z. Carson, C.-J. Haster, K. Chatziioannou, et al. The Impact of Non-Tensorial Polarizations on the Detection and Characterization of Gravitational-Wave Signals. Classical and Quantum Gravity, 37(15):155005, 2020. doi:10.1088/1361-6382/ab9a1e.
  45. [45]Volker Perlick. Gravitational lensing from a spacetime perspective. Living Reviews in Relativity, 7:9, 2004. doi:10.12942/lrr-2004-9.DOI
  46. [46]Bruno Bertotti, Luciano Iess, and Paolo Tortora. A test of general relativity using radio links with the cassini spacecraft. Nature, 425:374–376, 2003. doi:10.1038/nature01997.DOI
  47. [47]Wuzheng Guo, Shuo Cao, Qiumin Wang, Yun Chen, Marek Biesiada, Tonghua Liu, and Yujie Lian. Extragalactic test of general relativity with time-delay gravitational lenses. European Physical Journal C, 86:650, 2026. doi:10.1140/epjc/s10052-026-15856-0.DOI

Paper details

vlnc:W_tzwahaf3
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY-SA 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