Is the Sixth Mode of Non-Minimal Proca Gravity a Hidden Ghost?

A massive vector field is the simplest new ingredient one can add to gravity. Theorists often couple it directly to spacetime curvature, so that how strongly the field is felt depends on how curved space is. A recent analysis of such a theory on an expanding universe found something it should not have: one extra propagating mode that does not exist in flat space, and that mode does not travel at all. Its propagation speed is exactly zero (On the cosmological degrees of freedom of Proca field with non-minimal coupling to gravity). A zero-speed mode means the standard first-order (linear) calculation, which keeps only the leading small-fluctuation terms, can say nothing about whether this mode is harmless, grows exponentially, or carries negative energy. We propose to settle the question with the full nonlinear theory.

The claim is this. A mode that is absent around one background and present around another is the classic sign of a constraint that holds only on special backgrounds. The best-known example is the Boulware–Deser ghost of massive gravity. We conjecture that the sixth mode is a constraint-loss ghost of this kind, rather than a strongly coupled but healthy mode. If so, the nonlinear energy of the theory is unbounded from below wherever the vector's time component is nonzero. That is an Ostrogradsky-type pathology (a negative-energy degree of freedom) that the linear analysis cannot see. The alternative is that nonlinear terms bound the energy, as a screening mechanism would. These are sharp, opposite outcomes, and a Hamiltonian constraint count combined with explicit numerical evolution can tell them apart.

Background and Where This Sits

The theory is a massive (Proca) vector $A_\mu$ with mass $m$, coupled to the curvature scalar with strength $\alpha$ and to the Ricci tensor with strength $\beta$. Hell showed that around flat space it carries five modes: three vector modes and two graviton polarizations. She also showed that the $\beta$ coupling drags gravitational waves into strong coupling at a length scale $L_{\beta} \sim (\beta/M_{\rm pl}^2 m^2)^{1/3}$ and produces runaway negative-frequency modes on curved backgrounds. Her fix was a field-dependent rescaling of the metric, $g_{\mu\nu}\to g_{\mu\nu}-\beta A_\mu A_\nu/M_{\rm pl}^2$ (a "disformal transformation"), which removes the $\beta$ coupling (Unveiling the Inconsistency of the Proca Theory with Nonminimal Coupling to Gravity).

De Felice and Hell then went to Friedmann and anisotropic Bianchi I backgrounds. They found the sixth mode with $c_s^2=0$, together with tachyonic and gradient instabilities in the anisotropic case. They explicitly left the nonlinear Hamiltonian uncomputed. That open question motivates this proposal; we take the nonlinear step they did not.

Two further lines of work constrain the design. The Kalb–Ramond analysis shows that disformal rescalings cure Ricci couplings but fail for Riemann couplings (On the Kalb-Ramond field with non-minimal coupling to gravity). So field redefinitions are not a universal cure, and we must test rather than assume them. The study of new general relativity shows that mode counts can jump discontinuously between parameter branches at linear order (Revisiting stability in new general relativity). We read that as the same symptom of background-dependent constraints, and it supplies a second test case for our method.

On method, Cline and Hell exposed a ghost in a fourth-order scalar theory by introducing auxiliary fields and diagonalizing the kinetic matrix (Pathologies of dimension-zero scalar fields). We reuse that technique on reduced nonlinear systems. The disformal-Proca literature shows that matter inhomogeneities such as dark-matter halos spoil tunings made on smooth backgrounds (Vector disformal transformation of generalized Proca theory). For that reason our numerical tests include an inhomogeneous direction rather than staying homogeneous.

Research Program

The first stage is a full Dirac–Bergmann constraint analysis of the nonlinear theory in the split of spacetime into space plus time. This procedure counts true degrees of freedom by finding which equations are constraints rather than evolution equations. In pure Proca theory, $A_0$ has no time derivative and generates a constraint that removes one mode. The curvature couplings bring in second time derivatives of the metric multiplied by $A_\mu A_\nu$. We will compute whether the matrix of second derivatives of the Lagrangian with respect to velocities stays degenerate when $A_0 \neq 0$. The prediction is that it does not. In that case the $A_0$ constraint becomes a second-class pair only on the $A_0=0$ surface, and away from it the theory has six degrees of freedom. The algebra will be done in computer algebra (Maple, following the workflow of LLMs with in-context learning for Algorithmic Theoretical Physics, without relying on the agent). We will also check it independently by hand-reducing to the minisuperspace models below.

The second stage determines the sign of the sixth mode's energy. Because the mode's linear gradient term vanishes, its fate is fixed by the leading term that does not vanish. We will extract that term in two ways. One is the cubic and quartic action on Friedmann backgrounds. The other is the exact reduced Hamiltonian on a Bianchi I background plus a single plane-wave inhomogeneity, a finite-dimensional nonlinear system where boundedness can be checked directly. The decisive object is the minimum of the Hamiltonian $H$ over the surface where the constraints hold, $\inf H|_{\mathcal C}$. If it equals $-\infty$, the theory has a ghost. If it is finite, the screening hypothesis survives.

The third stage is numerical evolution in plane-symmetric spacetimes, which depend on time and one spatial direction. Starting from Friedmann data with small seeded perturbations of the sixth mode, we will track three things over many Hubble times: the perturbation amplitude, the separate positive and negative parts of the energy, and whether the constraints stay satisfied. These runs will cover the pure-$\alpha$ theory, the $\beta$ theory, and Hell's disformal frame.

Baselines

The naive reference point is the linear Friedmann analysis of De Felice and Hell. It predicts nothing about the sixth mode, because $c_s^2=0$. Any definite statement from us about boundedness or growth goes beyond it. No reported quantitative stability result was found in the research file for this mode; the paper reports only the qualitative $c_s^2 = 0$ breakdown.

The strongest existing method is linear perturbation theory on Bianchi I, also from On the cosmological degrees of freedom of Proca field with non-minimal coupling to gravity. There the anisotropy gives the mode a direction-dependent speed and reveals tachyonic growth. We will rerun it on the same backgrounds and parameter grid we use nonlinearly. To beat it, we must show one of two things. Either the nonlinear Hamiltonian is unbounded in a region where the linear Bianchi I analysis calls the mode stable, or nonlinear terms stabilize a region the linear analysis calls tachyonic. Either result shows that the linear analysis gives the wrong verdict.

The key ablations isolate what drives the extra mode. We will set $\beta=0$ and keep $\alpha$, and the reverse. We will take the $A_0\to 0$ background limit. We will also run Hell's disformal frame, which is the strongest competitor at the level of methods. Every configuration uses the same reduced models, initial data, resolution, and integration time.

As a methodological control, the constraint code must reproduce known counts before we trust it on the open problem. It must give five modes for this theory around flat space, the branch counts of new general relativity, and the conformal-gravity ghost of On the ghost problem of conformal gravity.

Why not just…?

Why not just treat it as an effective field theory and assume higher terms are suppressed? That argument needs a range of energies where the linear terms dominate. With $c_s^2=0$, no gradient term stabilizes the mode at any wavelength, so no such range exists. This is exactly the failure De Felice and Hell point to, and it is why the low strong-coupling scale $L_\beta$ in Hell's paper matters even for light fields.

Why not just apply the disformal transformation? This is the real threat. If Hell's frame removes the sixth mode on Friedmann backgrounds, the problem has a simple solution for the $\beta$ coupling. Our ablation tests this first. Even in that case, the $\alpha$ coupling is untouched by the transformation. The Kalb–Ramond Riemann case also shows that such redefinitions can fail. So a positive result narrows the proposal to the $\alpha$ sector rather than ending it, and we would report it as such.

Why not just compute the cubic action on Friedmann backgrounds? It is cheaper, and we include it. But it only yields a strong-coupling scale. It cannot establish whether the energy is bounded below, which needs the constraint analysis and the exact reduced Hamiltonian. If the cubic and Hamiltonian verdicts agree everywhere, we will say the cheaper tool suffices.

Why not just set the couplings to zero? That avoids the question rather than answering it. The aim is to learn whether any non-minimally coupled massive vector is viable, so the zero-coupling theory serves as our control rather than as our answer.

Decisive Checks

The proposal succeeds if three things hold together. The constraint count must give six degrees of freedom whenever $A_0\neq0$. We must construct an explicit trajectory on the constraint surface along which $H\to-\infty$, in the reduced Bianchi I plus plane-wave model. And the numerical evolution must show negative energy growing at a rate faster than the expansion rate, with the constraints preserved to integration accuracy.

The conjecture is falsified if the reduced Hamiltonian has a finite minimum for all couplings, or if the sixth mode turns out to be a constraint artifact that disappears in a correct count. Either outcome is a clean, publishable answer to the question De Felice and Hell left open.

For phenomenology, any branch that survives must keep gravitational waves at light speed inside an inhomogeneous region. The test is the GW170817 bound $|c_T-c|/c\le10^{-15}$ quoted in Vector disformal transformation of generalized Proca theory, evaluated on our plane-symmetric solutions rather than on a smooth background alone.