Massive Vector and Higher-Spin Test Fields

Adding mass changes which field components are physical. A Proca vector has three local polarizations, whereas a Maxwell field has two; a Fierz–Pauli massive tensor has five on a Ricci-flat background. These counts do not tell us whether a particular spherical multipole exists, whether the equations separate on Kerr, or whether a black hole is stable against that field. We derive the constraints that establish the counts, then compare what analytic separation and numerical evolution actually establish. The fields are classical and propagate on a fixed metric: backreaction, nonlinear massive-gravity models and quantum-state interpretation are outside this comparison.

Required background. PDEs and Green Functions for Gravitation supplies constrained hyperbolic systems, characteristic structure, and spectral boundary problems.

Helpful background. Black-Hole Geometry and Mechanics supplies stationary and rotating backgrounds; Formulations of General Relativity supplies constraint propagation and hyperbolicity; Frames, Coframes, and Anholonomy supplies the local-frame language required by spin fields.

Proca dynamics and its constraint

Let $\mu_V\gt0$ be a constant inverse Compton length. A real, minimally coupled vector follows from

$$S_A=\int d^4x\sqrt{-g}\left( -\frac14F_{\mu\nu}F^{\mu\nu} -\frac12\mu_V^2A_\mu A^\mu \right), \qquad F_{\mu\nu}=2\nabla_{[\mu}A_{\nu]}.$$

Variation gives the four component equations

$$E^\nu\equiv\nabla_\mu F^{\mu\nu}-\mu_V^2A^\nu=0.$$

The double divergence of an antisymmetric tensor vanishes: commuting the two derivatives gives Ricci contractions with $F$, and a symmetric Ricci tensor contracts to zero. Thus, even before imposing $E^\nu=0$,

$$C\equiv\nabla_\mu A^\mu, \qquad \nabla_\nu E^\nu=-\mu_V^2 C.$$

Consequently $C=0$ is a derived Lorenz constraint, not a gauge choice. Commuting derivatives once in the field equation gives

$$E^\nu=(PA)^\nu-\nabla^\nu C, \qquad (PA)^\nu\equiv(\Box-\mu_V^2)A^\nu-R^\nu{}_{\rho}A^\rho.$$

On a Ricci-flat background the explicit Ricci term vanishes, but the connection in $\Box A^\nu$ still contains curvature. This is the vacuum Proca system used in Rosa and Dolan 2012, §II A, p. 3, Eqs. (2)–(3), PDF.

Why the constraint propagates

Solving the four reduced wave equations $PA=0$ is not sufficient with arbitrary initial data. The same off-shell identity implies

$$\nabla_\nu(PA)^\nu=(\Box-\mu_V^2)C.$$

Therefore constraint violations themselves satisfy a homogeneous scalar wave equation. On a spacelike Cauchy slice $\Sigma$ with future unit normal $n^\mu$, require both $C|_\Sigma=0$ and $n^\mu\nabla_\mu C|_\Sigma=0$. With the reduced equations and $C=0$ on the slice, the original normal equation $n_\nu E^\nu=0$ supplies the second condition. This normal equation is the Proca version of Gauss’s constraint. Uniqueness for the scalar Cauchy problem then preserves $C=0$ in the domain of dependence. If a timelike boundary is introduced, its conditions must also preserve this subsidiary equation.

Four wave components would require eight freely specifiable initial functions. The two compatible constraint conditions leave six, or three field–momentum pairs. In flat space the same count is particularly transparent. A plane wave $A^\mu=\epsilon^\mu e^{-i\omega t+ikz}$ obeys

$$\omega^2=k^2+\mu_V^2, \qquad p_\mu\epsilon^\mu=0, \qquad p^\mu=(\omega,0,0,k).$$

Two polarizations point along $x$ and $y$. A normalized third is

$$\epsilon_L^\mu=\frac{1}{\mu_V}(k,0,0,\omega), \qquad \epsilon_{L\mu}\epsilon_L^\mu=1.$$

It has a spatial component along the direction of travel and is called longitudinal. There is no gauge transformation that removes it at nonzero mass.

Angular channels

Local polarization count is not a count of radial functions at every angular momentum. On Schwarzschild, a vector’s angular components are expanded using the unit sphere’s polar and axial harmonics,

$$D_A Y_{\ell m}, \qquad \varepsilon_A{}^B D_B Y_{\ell m},$$

where $D_A$ is the sphere derivative and $\varepsilon_{AB}$ its area form. For $\ell\ge1$, the axial amplitude decouples and two polar amplitudes remain coupled after applying the Lorenz constraint. For $\ell=0$, however, $Y_{00}$ is constant: both angular vector harmonics vanish. Only the time and radial components remain, and their constraint leaves one dynamical polar channel. See Rosa and Dolan 2012, §II B–C, pp. 3–4, Eqs. (4)–(17), PDF.

Field or sector	Physical count	Meaning and limitation
Massive vector, locally	Three polarizations	Two transverse and one longitudinal; no Maxwell gauge quotient.
Schwarzschild Proca, $\ell\ge1$	One axial and two polar channels	Different parity sectors decouple; the two polar equations generally do not.
Schwarzschild Proca, $\ell=0$	One polar channel	Angular vector harmonics vanish; three channels must not be imposed here.
Maxwell, locally	Two radiative polarizations	Gauge equivalence removes the extra pair; a static Coulomb field is not a third radiative mode.

For orientation, write $A_r=u(t,r)Y_{00}/(rf)$, with $f=1-2GM/r$ and $dr_*/dr=f^{-1}$. The Proca monopole equation is

$$\left[-\partial_t^2+\partial_{r_*}^2 -f\left(\mu_V^2+\frac{2(r-3GM)}{r^3}\right)\right]u=0.$$

This is the source’s Eq. (17), with $G$ restored. Its massless limit must be interpreted together with the reconstruction and gauge freedom; a surviving radial equation alone does not establish an extra photon. Nor does a negative region of this effective potential, by itself, prove an instability.

Two different zero-mass questions

Set the mass to zero in the original theory. The action becomes Maxwell’s, with $A_\mu\mapsto A_\mu+\nabla_\mu\Lambda$. The equation $\mu_V^2 C=0$ no longer enforces $C=0$; Lorenz gauge may instead be chosen. Gauge-equivalent potentials represent the same field, and there are two photon polarizations.

Take a limit of canonically normalized massive fields. Introduce a Stückelberg scalar $\pi$ by

$$A_\mu=B_\mu+\frac{\nabla_\mu\pi}{\mu_V}, \qquad B_\mu\mapsto B_\mu+\nabla_\mu\Lambda, \quad \pi\mapsto\pi-\mu_V\Lambda.$$

The joint transformation leaves $A_\mu$ unchanged. At finite mass it adds a redundancy, not a physical mode. Since second derivatives of a scalar commute, $F(A)=F(B)$, and the scalar Lagrangian inside the invariant volume integral becomes

$$\begin{aligned} \mathcal L={}&-\frac14F_{\mu\nu}(B)F^{\mu\nu}(B) -\frac12\nabla_\mu\pi\nabla^\mu\pi\\ &-\mu_V B^\mu\nabla_\mu\pi -\frac12\mu_V^2B_\mu B^\mu. \end{aligned}$$

Holding $B$ and $\pi$ finite while taking $\mu_V\to0$ gives a free Maxwell field plus a scalar with a positive kinetic energy. The scalar is not a third photon: it retains the normalized longitudinal degree of freedom of the limiting family, whereas setting the mass to zero in the original variables defines Maxwell theory. The change of variables itself is singular at zero mass.

Which distinction is observable depends on what is measured. With a prescribed current $J^\mu$, the longitudinal source coupling obeys

$$\int\sqrt{-g}\,\frac{J^\mu\nabla_\mu\pi}{\mu_V}\,d^4x =-\int\sqrt{-g}\,\frac{\pi\nabla_\mu J^\mu}{\mu_V}\,d^4x +\text{boundary term}.$$

It vanishes for a conserved current with vanishing boundary contribution. That statement does not remove the free scalar’s energy or prove a smooth limit for nonconserved sources, arbitrary initial states or interactions.

The Ricci-flat Fierz–Pauli tensor

Fix a smooth Ricci-flat background, a symmetric tensor $h_{\mu\nu}$, and a constant $\mu_T\gt0$. Define $h=g^{\mu\nu}h_{\mu\nu}$ and $C_\nu=\nabla^\mu h_{\mu\nu}$. Here $\delta G_{\mu\nu}[h]$ is the derivative at $\epsilon=0$ of the Einstein tensor of $g_{\mu\nu}+\epsilon h_{\mu\nu}$. A precise linear test model is

$$\mathcal E_{\mu\nu}\equiv\delta G_{\mu\nu}[h] +\frac{\mu_T^2}{2}(h_{\mu\nu}-g_{\mu\nu}h)=0.$$

The relative coefficient of the two mass terms is the Fierz–Pauli tuning. The parent equation and its reduction are given in Brito, Cardoso, and Pani 2013, §III B–C, pp. 6–7, Eqs. (19), (23), (26)–(28), PDF. We use this linear sector, not the paper’s nonlinear model or observational bounds.

On the vacuum background the linearized Bianchi identity is $\nabla^\mu\delta G_{\mu\nu}=0$. Taking the divergence of the parent equation therefore gives $C_\nu=\nabla_\nu h$. Taking the trace then gives

$$g^{\mu\nu}\mathcal E_{\mu\nu} =\Box h-\nabla^\nu C_\nu-\frac32\mu_T^2h=0.$$

The two derivative terms cancel, so $h=0$, and consequently $C_\nu=0$. Neither relation was imposed as a gauge choice. On this transverse, trace-free subspace the linearized Einstein tensor reduces to

$$\delta G_{\mu\nu}[h] =-\frac12\left(\Box h_{\mu\nu} +2R_{\mu\alpha\nu\beta}h^{\alpha\beta}\right).$$

Thus the complete constrained test system is

$$\begin{gathered} (\Box-\mu_T^2)h_{\mu\nu} +2R_{\mu\alpha\nu\beta}h^{\alpha\beta}=0,\\ \nabla^\mu h_{\mu\nu}=0,\qquad h=0. \end{gathered}$$

Unlike Proca’s explicit Ricci coupling, this Riemann coupling survives in vacuum and describes a tidal interaction. In a flat massive rest frame $p^\mu=(\mu_T,0,0,0)$, transversality sets $h_{0\nu}=0$. The remaining symmetric spatial $3\times3$ matrix has six components; zero trace leaves five. This checks the physical polarization count without mistaking a differential constraint for a gauge identification.

On a non-Einstein background the varied Bianchi identity contains background Einstein-tensor terms. The preceding cancellation cannot simply be reused. Additional curvature couplings or fields change the system whose constraints must be checked. For still higher spin, commuting derivatives in a subsidiary condition likewise introduces curvature acting on every tensor or spinor index. Covariant notation alone is not a consistency proof: one must name the model and show that its constraints close. There is no single generic higher-spin equation being asserted here.

Hyperbolicity is not stability

For the displayed reduced Proca and tensor wave operators, the highest-derivative symbol is $g^{\alpha\beta}q_\alpha q_\beta$ times the identity on the relevant vector or tensor fiber. They are normally hyperbolic: mass and the explicit curvature terms do not change their null characteristic cone. This statement concerns wave fronts, not the speed of a monochromatic massive packet. In flat space its group velocity is $d\omega/dk=k/\omega$, which is less than one at nonzero mass.

The physical initial-value problem still requires compatible constraints. In the Proca case we proved their propagation above; in the tensor case the divergence and trace conditions must likewise accompany the reduced equations and compatible initial data. A global evolution or boundary-value claim additionally requires a specified spacetime region, regularity and boundary conditions. None of these local structural facts proves decay, positivity of Killing energy, or absence of exponentially growing modes. Compare Energy Estimates and Test-Field Stability.

The spin-2 Teukolsky equation belongs to massless gravitational perturbations, not to the massive Fierz–Pauli test system. Its scalar curvature variable and metric reconstruction must not be used to erase three massive polarizations. The distinction persists even if both problems use the same Kerr background.

Proca separation and missing branches

Kerr’s principal tensor is a special antisymmetric hidden-symmetry tensor, denoted $k_{ab}$ here. The FKKS construction uses its inverse combination with the metric to reconstruct a vector from a separated scalar:

$$\begin{gathered} A^a=B^{ab}\nabla_b Z,\qquad B^{ab}(g_{bc}+i\nu k_{bc})=\delta^a{}_c,\\ Z=R(r)S(\theta)e^{-i\omega t+im\varphi}. \end{gathered}$$

The parameter $\nu$ labels separation branches; it is not the field mass. Substitution reduces the Proca and Lorenz equations to linked radial and angular ODEs with common separation parameters. This construction, rather than independent separation of four coordinate components, is the essential idea. The general canonical-background derivation is in Frolov, Krtouš, Kubizňák, and Santos 2018, §III, pp. 2–3, Eqs. (9)–(23), PDF; the full ODE derivation is outside this comparison’s scope.

The original Kerr calculation recovered two polarizations. Dolan recovered the third branch, whose Schwarzschild limit is axial: its $\nu$ diverges as rotation tends to zero, while $a\nu$ remains finite. Discarding divergent separation parameters would therefore discard a regular physical branch. This historically missing axial branch is not synonymous with the longitudinal plane-wave polarization. See Dolan 2018, §V, pp. 18–19, PDF.

Recovery of all three polarization branches is a major result, but it is not by itself a proof that a chosen set of separated modes spans arbitrary admissible initial data. Dolan also explicitly leaves the angular-spectrum investigation incomplete. Regular reconstruction, exceptional parameters, boundary domains and spectral completeness remain different questions. In particular, completeness of a resonance expansion is stronger than separation of the differential equation; see Completeness, Excitation, and Ringdown Content.

Separation and spectral results through 2026

The following comparison was assessed through 2026-09-07. It separates established constructions from their scope limits; it is not a claim that a literature search proves the nonexistence of further results.

Formulation and evidence	Established result	Scope or limitation
Proca, FKKS and Dolan	Separated reconstruction on the principal-tensor family; all three Kerr polarization branches recovered.	Branch recovery is not a general solution-space or resonance-completeness theorem. Low-multipole and singular-parameter limits need separate treatment.
Ricci-flat Fierz–Pauli, parent-equation reduction	Divergence and trace constraints give five massive polarizations and a curvature-coupled wave system.	Does not extend this exact parent to arbitrary matter backgrounds, nor prove black-hole stability.
Massive spin two, Dias et al.	Coupled frequency-domain computation of the special superradiant dipole on Kerr.	A specified azimuthal sector and finite parameter range; not the fastest instability over every sector.
Massive spin two, East–Siemonsen	Time-domain comparison finds the axisymmetric instability faster across much of the sampled domain.	Linear fixed-background results are distinct from the paper’s model-dependent nonlinear evolution.
Hidden-symmetry operators, Gray–Kubizňák	Explicit vector and vacuum tensor operator constructions.	The paper leaves direct separation using those operator sets open; operators alone do not prove complete massive tensor separation.

For the numerical comparison define $\alpha=GM\mu_T$ and $\chi=a/(GM)$, where $a$ is Kerr’s rotation length. Dias and collaborators impose regularity at the future horizon and poles and exponential decay at infinity. Their ten tensor components are implemented as nine coupled PDEs plus an algebraic trace relation. The displayed $m=1$ dipole survey covers $0.05\le\alpha\le0.8$ and $0\le\chi\lesssim0.99$; it is not an all-mode survey Dias, Lingetti, Pani, and Santos 2023, pp. 2–3, Fig. 1, and Appendix A, p. 6, PDF. The azimuthal reduction does not eliminate the remaining radial–angular coupling.

East and Siemonsen evolve the linear tensor system and compare $m=0,1,2,3$. The axisymmetric instability already exists in the Schwarzschild limit; it is not rotational superradiance. Across much of their sampled mass and spin range it grows faster than the superradiant sectors. Their $m=1,2$ modes dominate only in a narrow high-spin, high-mass region, requiring approximately $\chi\gtrsim0.95$. This is a finite numerical mode comparison, not a theorem ordering all modes East and Siemonsen 2023, §§II–III, pp. 2–3, Fig. 1, PDF.

The hidden-symmetry approach remains valuable, but Gray and Kubizňák 2024, §VI, p. 11, PDF explicitly distinguish constructing symmetry operators from obtaining direct separation: whether those operator sets suffice remains unresolved in that work. The cited results do not establish complete separation of the generic massive Fierz–Pauli Kerr system. Developments for massless linearized Einstein equations cannot settle that different massive equation merely by sharing the words “spin two.”

A complete massive-tensor separation theorem or a demonstrated obstruction would settle the separation question. Corrections to the constraints or hyperbolicity of the specified model, or an independently reproduced spectrum with a different mode hierarchy, could also change these conclusions. Observational exclusions and nonlinear saturation require additional model and evolution assumptions.

Exercises

1. A longitudinal mode with small field strength. For the normalized flat longitudinal polarization above, calculate $F_{03}$ and explain why its vanishing limit does not prove that the mode carries vanishing energy. Separately explain why Schwarzschild has only one Proca monopole channel.

Solution

Lowering the time index gives $A_0=-(k/\mu_V)e^{-i\omega t+ikz}$ and $A_3=(\omega/\mu_V)e^{-i\omega t+ikz}$. Hence

$$F_{03}=-\frac{i(\omega^2-k^2)}{\mu_V}e^{-i\omega t+ikz} =-i\mu_V e^{-i\omega t+ikz}.$$

Although $F_{03}$ tends to zero, components of $A$ grow as $1/\mu_V$ at fixed nonzero $k$. The mass contribution to energy contains $\mu_V^2(|A_0|^2+|\mathbf A|^2)/2$, which need not vanish. The Stückelberg scalar makes that finite longitudinal energy explicit. The complex wave is a mode representation of a real field; a real wave or its time-averaged quadratic energy gives the same conclusion.

For the monopole, $D_A Y_{00}=0$ removes both angular vector harmonics. The remaining time and radial functions are linked by the constraint, leaving one polar wave equation. This restricted spherical sector is not the whole local solution space, so it need not contain all three plane-wave polarizations.

2. Why the tensor mass is tuned. In flat space replace the mass term in the parent equation by $\mu_T^2(h_{\mu\nu}-\kappa\eta_{\mu\nu}h)/2$. Derive its divergence and trace conditions. Why is $\kappa=1$ special?

Solution

The divergence gives $\partial^\mu h_{\mu\nu}=\kappa\partial_\nu h$. Substituting it into the trace gives

$$2(1-\kappa)\Box h+(1-4\kappa)\mu_T^2h=0.$$

At $\kappa=1$, the derivative term cancels and nonzero mass enforces $h=0$, followed by transversality. Away from that tuning the trace has a differential equation rather than this algebraic elimination; one cannot retain the five-polarization claim. Counting an extra scalar does not determine its energy sign: that requires the quadratic action or Hamiltonian. The finite-mass derivation also cannot be continued by dividing by $\mu_T^2$ at zero mass.

3. What does a separated ansatz prove? A paper gives two regular massive-vector branches with quasinormal boundary conditions. What has it established, what remains before a completeness claim, and why does the result not automatically apply to massive spin two?

Solution

Direct substitution into every original component and the derived constraint establishes the two branches represented by the ansatz. It does not account for the third physical polarization. Search exceptional separation-parameter limits and reconstruct the field before discarding a branch; the historically missing Kerr Proca branch was axial in the Schwarzschild limit, not simply “the longitudinal mode.”

Even after all three are recovered, verify regular reconstruction, low multipoles, boundary domains and any exceptional frequencies. A claim to span arbitrary initial data needs its own completeness argument; a quasinormal-mode expansion also has to address continuum or branch-cut contributions. Agreement with an independent PDE calculation tests particular modes, not that general theorem.

A massive tensor has five polarizations and an explicit Riemann coupling. The vector ansatz does not supply its divergence and trace constraints, its reconstruction, or its boundary-value problem. Each must be checked for the tensor operator itself.

References

• Brito, Richard, Vitor Cardoso, and Paolo Pani. “Massive Spin-2 Fields on Black Hole Spacetimes: Instability of the Schwarzschild and Kerr Solutions and Bounds on the Graviton Mass.” Physical Review D 88 (2013): 023514. DOI. Open PDF.
• Dias, Oscar J. C., Giuseppe Lingetti, Paolo Pani, and Jorge E. Santos. “Black Hole Superradiant Instability for Massive Spin-2 Fields.” Physical Review D 108 (2023): L041502. DOI. Open PDF.
• Dolan, Sam R. “Instability of the Proca Field on Kerr Spacetime.” Physical Review D 98 (2018): 104006. DOI. Open PDF.
• East, William E., and Nils Siemonsen. “Instability and Backreaction of Massive Spin-2 Fields around Black Holes.” Physical Review D 108 (2023): 124048. DOI. Open PDF.
• Frolov, Valeri P., Pavel Krtouš, David Kubizňák, and Jorge E. Santos. “Massive Vector Fields in Rotating Black-Hole Spacetimes: Separability and Quasinormal Modes.” Physical Review Letters 120 (2018): 231103. DOI. Open PDF.
• Gray, Finnian, and David Kubizňák. “Homogeneous Symmetry Operators in Kerr–NUT–AdS Spacetimes.” Physical Review D 109 (2024): 084027. DOI. Open PDF.
• Rosa, João G., and Sam R. Dolan. “Massive Vector Fields on the Schwarzschild Spacetime: Quasinormal Modes and Bound States.” Physical Review D 85 (2012): 044043. DOI. Open PDF.