Hertz Potentials and Metric Reconstruction

Extreme Weyl scalars efficiently encode radiative curvature, but many calculations require a metric perturbation. Hertz reconstruction produces a vacuum metric from an adjoint Teukolsky field in a specified radiation gauge. It is not, by itself, a globally regular or unique inverse of curvature extraction: stationary Kerr-parameter variations lie in the missing kernel, generic sources require additional pieces, and radiation gauges can carry strings or jumps. The correct object is therefore a decomposition with an explicit domain, gauge, completion, and source treatment.

Required background. Frames, Coframes, and Anholonomy supplies null-tetrad components, spin weights, and tetrad gauge.

Helpful background. Black-Hole Geometry and Mechanics supplies Kerr parameter variations and horizon regularity; PDEs and Green Functions for Gravitation supplies adjoint operators and boundary-value problems; Formulations of General Relativity supplies linearized Einstein gauge and constraints.

Adjoint-operator construction

Let $\mathcal E[h]_{ab}=\delta G_{ab}[h]$ be the linearized Einstein operator on vacuum Kerr. Let $\mathcal C_s$ extract the master curvature: $\mathcal C_{+2}[h]=\delta\Psi_0$ and $\mathcal C_{-2}[h]=\rho_{NP}^{-4}\delta\Psi_4$. For the corresponding Teukolsky operator, decoupling gives

$$\mathcal T_s\mathcal C_s =\mathcal P_s\mathcal E,$$

where $\mathcal P_s$ prepares the projected source. Use the complex bilinear spacetime pairing, contracting tensor indices with the background metric and integrating with $dV_g$; pair scalar spin-weighted fields with their dual weights. Compactly supported test sections remove boundary terms from integration by parts. Taking these formal adjoints gives

$$\mathcal E^\dagger\mathcal P_s^\dagger =\mathcal C_s^\dagger\mathcal T_s^\dagger.$$

The linearized Einstein operator is formally self-adjoint after boundary terms are controlled. Therefore, if a Hertz potential $\Phi_H$ satisfies $\mathcal T_s^\dagger\Phi_H=0$, then

$$h^{\mathrm{rec}}_{ab} =2\operatorname{Re} \left(\mathcal P_s^\dagger\Phi_H\right)_{ab}$$

solves the source-free linearized Einstein equation. Indeed, applying $\mathcal E$ gives twice the real part of $\mathcal C_s^\dagger\mathcal T_s^\dagger\Phi_H=0$. This is the local adjoint construction introduced by Wald 1978, pp. 203–206, with explicit radiation-gauge operators originating in Chrzanowski 1975, pp. 2042–2062. The compact support is imposed on the test sections, not on every physical Hertz potential. A global adjoint domain requires separate boundary conditions.

Formal adjointness does not establish an inverse. The potential must be chosen so that applying the reconstruction operator and then extracting curvature returns the desired $\delta\Psi_0$ or $\delta\Psi_4$. At mode level this inversion uses fourth-order Teukolsky–Starobinsky identities and fails or changes form at exceptional parameters. It also inherits the mode’s radial and angular boundary conditions.

Ingoing and outgoing radiation gauges

With the site’s Kinnersley ordering and $(-,+,+,+)$ signature, define the traceless ingoing radiation gauge by

$$\ell^a h^{\mathrm{IRG}}_{ab}=0, \qquad g^{ab}h^{\mathrm{IRG}}_{ab}=0,$$

and the traceless outgoing radiation gauge by

$$n^a h^{\mathrm{ORG}}_{ab}=0, \qquad g^{ab}h^{\mathrm{ORG}}_{ab}=0.$$

In the explicit spin-weighted normalization used below, an $s=-2$ Hertz field generates IRG and an appropriately rescaled $s=+2$ field generates ORG. The adjectives name the null leg removed by the gauge and historical radiation adaptation; they do not guarantee smoothness on every boundary. The gauge and potential definitions are given in Berens–Gravely–Lupsasca 2024, arXiv v4 (2025), Eqs. (1.18)–(1.24), p. 7, PDF.

IRG is often useful near the future horizon after a horizon-regular tetrad transformation, while ORG is often useful toward future null infinity. Yet a Kinnersley component may still contain powers of $\Delta$, and a radiation-gauge metric may be singular on a line or discontinuous surface. Regularity must be checked in the intended chart and tetrad; the gauge’s name is not evidence.

Curvature inversion and reconstruction check

Suppose a vacuum mode of $\delta\Psi_4$ has labels $(\ell,m,\omega)$ and the chapter phase $e^{-i\omega t+im\phi}$. A consistent reconstruction proceeds in four steps:

1. solve the appropriate fourth-order inversion relation for the mode of $\Phi_H$, including the Starobinsky constant and the conjugate mode required to obtain a real metric;
2. apply the IRG or ORG second-order reconstruction operator;
3. take the real part and add any allowed completion and gauge pieces;
4. reapply the curvature-extraction operator and verify the original Weyl mode.

The last step is indispensable. A generic solution of the adjoint Teukolsky equation produces a vacuum metric, but not necessarily the metric whose extreme curvature equals the starting field. The distinction and the fourth-order relations are explicit in Berens–Gravely–Lupsasca 2024, arXiv v4 (2025), p. 7, PDF.

Mode inversion is local to the stated subextremal vacuum problem. Complex frequencies, algebraically special modes, stationary modes, extremality, and changes of asymptotic boundary conditions require separate analysis. No mode formula here asserts completeness of a spectral expansion or existence of a globally smooth radiation gauge.

A local static quadrupole

A complete example is possible in elementary functions. Set $a=0$, $\mu=GM\gt0$, $f=1-2\mu/r$ and $\Delta=r^2f$. Work in the local vacuum region $r\gt2\mu$, $0\lt\theta\lt\pi$. We will reconstruct a static $\ell=2,m=0$ applied tidal field, not an isolated asymptotically flat perturbation or a quasinormal mode. The inversion below has a nonzero fourth-order coefficient even at $\omega=0$; no $1/\omega$ formula is used.

The Schwarzschild Kinnersley vectors and needed covectors are

$$\begin{aligned} D=\ell^a\partial_a&=f^{-1}\partial_t+\partial_r,\\ n^a\partial_a&=\tfrac12\partial_t-\tfrac f2\partial_r,\\ m^a\partial_a&=\frac{\partial_\theta+i\csc\theta\partial_\phi}{\sqrt2r},\\ \ell_a dx^a&=-dt+dr/f, &m_a dx^a&=\frac r{\sqrt2}(d\theta+i\sin\theta\,d\phi). \end{aligned}$$

Define $\mathcal L_j=\partial_\theta-i\csc\theta\partial_\phi+j\cot\theta$ and its coefficientwise conjugate $\overline{\mathcal L}_j$. We now use an explicitly normalized operator $\mathcal S_0^\dagger$ and potential $\Phi$, rather than identifying them by symbol with the abstract $\mathcal P_s^\dagger,\Phi_H$ above:

$$\begin{aligned} (\mathcal S_0^\dagger\Phi)_{ab} ={}&-\frac{\ell_a\ell_b}{4r^2} \overline{\mathcal L}_1\overline{\mathcal L}_2\Phi -\frac{m_am_b}{2}(D-2/r)D\Phi\\ &+\frac{\ell_{(a}m_{b)}}{\sqrt2r} (D\overline{\mathcal L}_2-2\overline{\mathcal L}_2/r)\Phi,\\ h_{ab}&=2\operatorname{Re}(\mathcal S_0^\dagger\Phi)_{ab}. \end{aligned}$$

Symmetrization has weight one half. The displayed derivatives act on $\Phi$, not on the exterior tetrad factors. This is the Schwarzschild specialization of Berens–Gravely–Lupsasca 2024, arXiv v4 (2025), Eq. (3.10a), p. 27, PDF. Operator normalization matters: multiplying a decoupled equation by a function $q$ replaces its source map $\mathcal P$ by $q\mathcal P$ and its adjoint by $\mathcal P^\dagger\circ M_q$, where $M_q$ multiplies the potential by $q$ before differentiation. Such weights cannot be replaced by a constant rescaling of a quoted Hertz formula.

In this explicit normalization, $\Phi$ has spin weight $-2$ and boost weight $-2$ (GHP weight $\{-4,0\}$), and satisfies the homogeneous spin-minus-two Teukolsky equation. Its conjugate has opposite spin weight, even when a particular component happens to be real. The source’s mostly-plus Weyl contractions have the opposite sign to this site’s contractions. The translated curvature formulas are therefore

$$\begin{aligned} \delta\Psi_0&=-\tfrac14D^4\bar\Phi,\\ r^4\delta\Psi_4&=-\tfrac1{16} \mathcal L_{-1}\mathcal L_0\mathcal L_1\mathcal L_2\bar\Phi +\tfrac{3\mu}{4}\partial_t\Phi. \end{aligned}$$

The factors two in the real metric, one quarter and one sixteenth in curvature recovery belong to the same convention Berens–Gravely–Lupsasca 2024, arXiv v4 (2025), Eqs. (1.4), (1.18), (1.22), pp. 3, 7, PDF.

Choose a real curvature coefficient $E_0$ of dimension $L^{-2}$ and set

$$\Phi=-\frac{E_0}{6}\Delta^2\sin^2\theta.$$

Here $E_0$ is not an orbital energy or an independently matched STF tidal moment. At $s=-2,\ell=2,m=\omega=0$, the radial and angular equations are

$$\begin{aligned} \Delta R''-\Delta'R'-4R&=0,\\ \frac1{\sin\theta}(\sin\theta S')' +(2-4\cot^2\theta)S&=0. \end{aligned}$$

They are solved by $R=\Delta^2$ and $S=\sin^2\theta$. Moreover,

$$\frac{d^4\Delta^2}{dr^4}=24, \qquad \mathcal L_{-1}\mathcal L_0\mathcal L_1\mathcal L_2\sin^2\theta =24\sin^2\theta.$$

Thus neither fourth-order inversion coefficient vanishes; the associated squared Starobinsky coefficient is $24^2=576$. The zero frequency does not obstruct this particular polynomial inversion.

Recovering the curvature

Applying the second-order operator, without leaving unevaluated derivatives, gives

$$\begin{aligned} h_{ab}dx^a dx^b=\frac{E_0}{3}\big[& r^2(3\cos^2\theta-1)(f\,dt-dr)^2\\ &+4r^3\sin\theta\cos\theta(f\,dt-dr)d\theta\\ &+r^2(r^2-2\mu^2)\sin^2\theta (d\theta^2-\sin^2\theta\,d\phi^2)\big]. \end{aligned}$$

Direct contraction gives $\ell^ah_{ab}=0$ and $g^{ab}h_{ab}=0$. A separate connection-variation calculation gives $\delta R_{ab}=0$ and recovers the extreme curvatures

$$\delta\Psi_0=E_0\sin^2\theta, \qquad \delta\Psi_4=\frac{E_0f^2}{4}\sin^2\theta.$$

To perform this check directly, compute $\delta\Gamma$ and $\delta R^a{}_{bcd}$ from the displayed metric, remembering that lowering the first index also varies the metric:

$$\delta R_{abcd}=g_{ae}\delta R^e{}_{bcd}+h_{ae}R^e{}_{bcd}.$$

Omitting the second term would not give the curvature variation of $g+h$. Because the background and its perturbation are Ricci-flat here, $\delta C_{abcd}=\delta R_{abcd}$. Contract with the background principal tetrad and use the site’s negative Weyl contractions. First-order tetrad adjustments do not change these extreme projections on a type-D background. The results agree with the independently applied fourth-order Hertz formulas above.

For unit-sphere harmonics, ${}_{2}Y_{20}={}_{-2}Y_{20}=N\sin^2\theta$ with $N=\sqrt{15/(32\pi)}$. Choosing $E_0=NA$ gives target mode amplitude $\delta\Psi_0=A\,{}_2Y_{20}$ and $\delta\Psi_4=(Af^2/4)\,{}_{-2}Y_{20}$. This unit-sphere notation is distinct from the chapter’s polar-norm-one separation; the physical curvature is unchanged. The displayed $h_{tt}$ grows like $r^2$, making its externally applied, nonisolated character explicit. No horizon or infinity regularity claim is inferred from the label IRG.

Completion pieces and nonuniqueness

Curvature extraction has a kernel. In an asymptotically flat, regular vacuum Kerr exterior, the physically central missing stationary axisymmetric pieces are infinitesimal variations of the black hole’s mass and angular momentum:

$$h^{\mathrm{comp}}_{ab} =\delta M\,\frac{\partial g_{ab}}{\partial M} +\delta J\,\frac{\partial g_{ab}}{\partial J},$$

up to a gauge perturbation $\mathcal L_Xg_{ab}$. They change the Coulombic curvature and charges but have $\delta\Psi_0=\delta\Psi_4=0$ in a principal tetrad. Thus a radiative scalar cannot determine $\delta M$ or $\delta J$.

The nonuniqueness can be seen without a global kernel classification. On Schwarzschild, put $\delta\mu=G\delta M$ and $\delta j=G\delta J$. Two explicit parameter directions have the nonzero components

$$h^{(M)}_{tt}=\frac{2\delta\mu}{r},\qquad h^{(M)}_{rr}=\frac{2\delta\mu}{rf^2},\qquad h^{(J)}_{t\phi}=h^{(J)}_{\phi t} =-\frac{2\delta j}{r}\sin^2\theta.$$

Each satisfies $\delta R_{ab}=0$ and has both extreme Weyl scalars zero, while its $1/r$ asymptotic coefficient changes the indicated charge. Adding either to the tidal metric leaves its two extreme curvatures unchanged. The parameter representatives displayed here need not themselves preserve IRG; requiring a common radiation gauge entails a separate allowed gauge transformation and boundary check.

These examples establish missing charge information, not that mass and spin exhaust the kernel on every possible domain. The modern reconstruction account discusses the missing pieces and its limitations in Berens–Gravely–Lupsasca 2024, arXiv v4 (2025), § 1.3, pp. 11–12, PDF. Completion must be fixed from independent charges, matching conditions or jumps across a source. A complete kernel theorem would additionally need its precise vacuum, regularity, topology and boundary hypotheses; the local construction above does not supply it.

Gauge freedom remains after the curvature data are fixed. Two reconstructions can differ by $\mathcal L_Xg_{ab}$ and completion while agreeing on the extreme Weyl scalars. A claim of uniqueness must therefore state the gauge class, charges, function space, boundaries, and source domain.

Sources, strings, and correctors

Classic Chrzanowski–Cohen–Kegeles reconstruction directly gives a vacuum metric in a radiation gauge. For a point particle, one may reconstruct separately in vacuum regions on either side of the worldline and then determine completion by matching. A globally imposed radiation gauge is generally singular near a generic source: half-string gauges place a singular ray on one side, while a no-string construction glues two regular-sided gauges at the price of a jump across a surface.

For a general sourced linearized equation, a broader decomposition has the form

$$h_{ab} =2\operatorname{Re}(\mathcal P_s^\dagger\Phi_H)_{ab} +h^{\mathrm{comp}}_{ab} +\mathcal L_Xg_{ab} +x_{ab}.$$

The corrector $x_{ab}$ is determined by transport equations and carries precisely the source components incompatible with a traceless radiation gauge. In particular, a pure IRG reconstructed metric obeys $(\mathcal Eh)_{\ell\ell}=0$, so it cannot represent a source with $T_{\ell\ell}\ne0$ without an extra piece. Hollands and Toomani develop a sourced Kerr Cauchy scheme with reconstructed, zero-mode, gauge, and corrector terms and state its decay and domain hypotheses Hollands–Toomani 2024, §§1.1–1.3 and §3.

That modern decomposition does not license the claim that one globally regular formula covers arbitrary distributions, all gauges, all boundaries, and extremality. For each use, state the source class, region, initial or boundary data, reconstruction gauge, corrector prescription, and charge completion.

Common pitfalls

Calling the Weyl scalar a complete metric variable. It captures radiative degrees of freedom but misses at least mass, angular momentum, and pure gauge. Supply and justify completion.

Using radiation gauge through a source without qualification. Generic matter violates an algebraic compatibility condition and produces strings or jumps. Reconstruct in vacuum subdomains or include a sourced corrector scheme.

Equating a formal adjoint map with a global inverse. The adjoint map generates solutions. Curvature recovery, mode exceptions, boundary behavior, regularity, and uniqueness require separate checks.

Exercises

A stationary perturbation changes the Kerr mass by $\delta M$ and a compact source has $T_{\ell\ell}\ne0$ in some neighborhood. What information is missing from extreme-curvature data in the first case? What compatibility obstruction does a pure reconstructed IRG metric face in the second? Give the decomposition that separates the two issues.

Solution

The parameter variation $h^{(M)}_{ab}=\delta M\,\partial_Mg_{ab}$ is a stationary vacuum perturbation with $\delta\Psi_0=\delta\Psi_4=0$. Curvature-only data therefore cannot determine its coefficient: an algorithm may choose a completion convention, but that default is not supplied by the extreme scalar. The missing mass must be determined from an independent charge or matching condition. The same reasoning applies to $\delta J$.

A traceless IRG reconstructed metric satisfies $\ell^ah_{ab}=0$ and has $(\mathcal Eh)_{\ell\ell}=0$ identically. With $\mathcal E[h]=\delta G[h]$, it cannot satisfy $(\mathcal Eh)_{\ell\ell}=8\pi G T_{\ell\ell}$ where the physical source is nonzero. Forcing the gauge through the source creates a singular or discontinuous representative rather than eliminating the incompatibility.

The diagnostic decomposition is

$$h_{ab}=h^{\mathrm{rec}}_{ab} +\delta M\,\partial_Mg_{ab} +\delta J\,\partial_Jg_{ab} +\mathcal L_Xg_{ab}+x_{ab}.$$

Here $h^{\mathrm{rec}}$ carries the Teukolsky radiative sector, the two parameter derivatives carry charge completion, $\mathcal L_Xg$ fixes the chosen gauge representative, and $x_{ab}$ carries source components incompatible with pure IRG. One must still state the vacuum subdomains, matching surface, boundary data, and regularity class; the decomposition is a framework, not an automatic global existence or uniqueness result.

References

• Berens, Roman, Trevor Gravely, and Alexandru Lupsasca. “Gravitational Waves on Kerr Black Holes I: Reconstruction of Linearized Metric Perturbations.” Classical and Quantum Gravity 41 (2024): 195004. DOI. Open PDF, arXiv:2403.20311v4, revised 10 October 2025.
• Chrzanowski, Paul L. “Vector Potential and Metric Perturbations of a Rotating Black Hole.” Physical Review D 11 (1975): 2042–2062. DOI.
• Hollands, Stefan, and Vahid Toomani. “Metric Reconstruction in Kerr Spacetime.” Classical and Quantum Gravity 42 (2025): 175008. DOI. Open PDF.
• Wald, Robert M. “Construction of Solutions of Gravitational, Electromagnetic, or Other Perturbation Equations from Solutions of Decoupled Equations.” Physical Review Letters 41 (1978): 203–206. DOI.