Regularised Proca Theory

• Regularised Proca Theory is a family of constructions for massive Abelian vector fields that preserve the constraint structure, avoiding extra degrees of freedom and Ostrogradsky instabilities.
• It employs methods like Stueckelberg completions, generalized Proca derivative self-interactions, and non-perturbative formulations (e.g., Proca-Nuevo) to maintain healthy dynamics.
• Key applications include regularizing Maxwell theory in curved spacetime, modeling black holes and cosmology, and ensuring consistent quantization in quantum field settings.

Regularised Proca Theory denotes a family of constructions built around the massive Abelian vector field rather than a single universally fixed model. In one usage, it refers to healthy derivative self-interactions of a massive spin-1 field that preserve the correct number of propagating degrees of freedom and avoid Ostrogradsky instabilities; in other usages, it refers to Stueckelberg completions, Proca fields used as a regulator for Maxwell theory in curved spacetime, and vector–tensor theories obtained by dimensional or geometric regularization procedures. A common thread is the preservation, deformation, or controlled recovery of the Proca constraint structure, so that the theory retains the appropriate massive-vector content or a well-defined massless limit (Jiménez et al., 2016, Schambach, 2017, Charmousis et al., 17 Apr 2025).

1. Canonical Proca sector and the origin of “regularisation”

The standard Proca field is a massive spin-1 field $A_\mu$ with field strength $$F_{\mu\nu}\equiv \partial_\mu A_\nu-\partial_\nu A_\mu$$ in flat spacetime, or $F_{\mu\nu}\equiv \nabla_\mu A_\nu-\nabla_\nu A_\mu$ in curved spacetime. With the convention $X\equiv -A_\mu A^\mu/2$, the canonical Lagrangian is

$$\mathcal{L}_{\text{Proca}}=-\frac14 F_{\mu\nu}F^{\mu\nu}+\frac12 m^2 A_\mu A^\mu .$$

A massive vector in four dimensions propagates three physical degrees of freedom: two transverse helicities and one longitudinal mode. The temporal component $A_0$ is non-dynamical and is removed by a constraint. Equivalently, the divergence of the equations of motion yields $\partial_\mu A^\mu=0$ for $m\neq 0$, or its covariant analogue in curved spacetime (Cadavid et al., 2019).

Within this setting, “regularisation” commonly means enforcing or preserving the constraint structure when additional interactions are introduced. The central requirement is that the Hessian or kinetic matrix remain degenerate in the right way, so that a primary constraint survives and the would-be fourth polarization does not propagate. In the generalized Proca literature this is achieved through carefully engineered derivative self-interactions; in other contexts it is achieved by Stueckelberg completion, disformal field redefinitions, or dimensional regularization procedures that turn otherwise topological or inconsistent sectors into well-defined vector–tensor dynamics (Jiménez et al., 2016, Hell, 2024).

2. Constraint-preserving completions: Stueckelberg, generalized Proca, and beyond generalized Proca

A standard route to a regularised massive vector theory is the Stueckelberg mechanism. One replaces the mass term by the gauge-invariant combination $mA_\mu-\partial_\mu\phi$, so that

$$\mathcal{L}_{\text{Stueck}}=-\frac14 F_{\mu\nu}F^{\mu\nu}+\frac12 (mA_\mu-\partial_\mu\phi)(mA^\mu-\partial^\mu\phi) .$$

This restores Abelian gauge invariance under $$F_{\mu\nu}\equiv \partial_\mu A_\nu-\partial_\nu A_\mu$$0, $$F_{\mu\nu}\equiv \partial_\mu A_\nu-\partial_\nu A_\mu$$1. In the BRST formulation, the Stueckelberg-modified Proca theory in arbitrary dimension admits nilpotent (anti-)BRST symmetries, while in $$F_{\mu\nu}\equiv \partial_\mu A_\nu-\partial_\nu A_\mu$$2 dimensions a modified Stueckelberg structure also supports discrete duality symmetries and nilpotent (anti-)co-BRST transformations (Rao et al., 2021).

Generalized Proca theory extends the canonical model by adding derivative self-interactions organized in levels $$F_{\mu\nu}\equiv \partial_\mu A_\nu-\partial_\nu A_\mu$$3–$$F_{\mu\nu}\equiv \partial_\mu A_\nu-\partial_\nu A_\mu$$4. Up to the quartic level, the standard pieces are

$$F_{\mu\nu}\equiv \partial_\mu A_\nu-\partial_\nu A_\mu$$5

$$F_{\mu\nu}\equiv \partial_\mu A_\nu-\partial_\nu A_\mu$$6

$$F_{\mu\nu}\equiv \partial_\mu A_\nu-\partial_\nu A_\mu$$7

The construction is based on Lorentz-invariant scalars built from $$F_{\mu\nu}\equiv \partial_\mu A_\nu-\partial_\nu A_\mu$$8 and first derivatives, together with degeneracy conditions on the flat-space Hessian. At the fourth level, the systematic analysis gives

$$F_{\mu\nu}\equiv \partial_\mu A_\nu-\partial_\nu A_\mu$$9

which ensure that only three vector degrees of freedom propagate (Cadavid et al., 2019).

Beyond generalized Proca theory arises when flat-space total divergences are retained during covariantization. The key observation is that terms that are pure divergences in flat space cease to be pure divergences once $F_{\mu\nu}\equiv \nabla_\mu A_\nu-\nabla_\nu A_\mu$0, because covariant derivatives do not commute. At the fourth level this produces curvature-dependent interactions such as $F_{\mu\nu}\equiv \nabla_\mu A_\nu-\nabla_\nu A_\mu$1, and the complete sector can be written as

$F_{\mu\nu}\equiv \nabla_\mu A_\nu-\nabla_\nu A_\mu$2

or equivalently

$F_{\mu\nu}\equiv \nabla_\mu A_\nu-\nabla_\nu A_\mu$3

with $F_{\mu\nu}\equiv \nabla_\mu A_\nu-\nabla_\nu A_\mu$4. A salient result is that the genuinely new fourth-level beyond-generalized-Proca interactions are parity-even, while parity-violating $F_{\mu\nu}\equiv \nabla_\mu A_\nu-\nabla_\nu A_\mu$5-contractions at this level reduce to $F_{\mu\nu}\equiv \nabla_\mu A_\nu-\nabla_\nu A_\mu$6-type pieces or total divergences (Cadavid et al., 2019).

3. New constrained classes and multi-vector extensions

A distinct constrained completion is Proca-Nuevo, or “Procanuevo,” which replaces the finite generalized Proca hierarchy by a nonlinearly completed dRGT-like matrix structure. The composite tensor is

$F_{\mu\nu}\equiv \nabla_\mu A_\nu-\nabla_\nu A_\mu$7

with $F_{\mu\nu}\equiv \nabla_\mu A_\nu-\nabla_\nu A_\mu$8 and $F_{\mu\nu}\equiv \nabla_\mu A_\nu-\nabla_\nu A_\mu$9, and the interaction is

$X\equiv -A_\mu A^\mu/2$0

Its Hessian is nonperturbatively degenerate, with a universal null eigenvector $X\equiv -A_\mu A^\mu/2$1, so a primary second-class constraint is present and, together with the secondary constraint, leaves three propagating modes in four dimensions. The theory is not equivalent to generalized Proca: tree-level $X\equiv -A_\mu A^\mu/2$2 amplitudes differ for all parameter choices, so there is no local, Lorentz-invariant field redefinition relating the two (Rham et al., 2020).

Extended Proca-Nuevo interpolates between generalized Proca operators and Proca-Nuevo ones. On flat spacetime and on fixed curved backgrounds it preserves the primary constraint, but when mixed dynamically with gravity the constraint is broken in a Planck scale suppressed way. The theory nevertheless admits covariantized models with ghost-free cosmological solutions. On FLRW backgrounds the constraint is exact at background and linearized level, and the paper derives tensor, vector, and scalar dispersion relations together with stability and subluminality conditions. In a specific set-up it also exhibits explicit hot Big Bang solutions with a late-time self-accelerating epoch and $X\equiv -A_\mu A^\mu/2$3 in the special model (Rham et al., 2021).

The multi-Proca generalization extends healthy derivative self-interactions to a set of massive vector fields $X\equiv -A_\mu A^\mu/2$4 with internal rotational symmetry. The construction separates antisymmetric tensors $X\equiv -A_\mu A^\mu/2$5 from symmetric tensors $X\equiv -A_\mu A^\mu/2$6, organizes interactions order by order in derivatives, and distinguishes direct extensions of single-vector generalized Proca terms from genuine multi-Proca interactions with no single-field analogue. In this framework one again requires that $X\equiv -A_\mu A^\mu/2$7 remain auxiliary, and in the multi-field case one must impose both primary Hessian degeneracy and the vanishing of a secondary Hessian. Cosmologically, the theory supports temporal, triad, and combined isotropic configurations (Jiménez et al., 2016).

4. Curved-spacetime renormalization, non-minimal couplings, and disformal regularisation

In curved spacetime, the fluctuation operator of a vector model is naturally classified by the degeneracy of its principal symbol. The non-degenerate vector field, the Abelian gauge field, and the Proca field fall into distinct classes; generalized Proca is more intricate because the principal part is degenerate while the mass-like tensor $X\equiv -A_\mu A^\mu/2$8 can itself depend on curvature. For the standard Proca field, a “massive Ward identity” converts the one-loop calculation into minimal traces. The divergent part of the one-loop effective action is

$X\equiv -A_\mu A^\mu/2$9

so the required counterterms are purely gravitational: cosmological constant, Einstein–Hilbert, and curvature-squared terms. There is no divergence proportional to $$\mathcal{L}_{\text{Proca}}=-\frac14 F_{\mu\nu}F^{\mu\nu}+\frac12 m^2 A_\mu A^\mu .$$0 nor to $$\mathcal{L}_{\text{Proca}}=-\frac14 F_{\mu\nu}F^{\mu\nu}+\frac12 m^2 A_\mu A^\mu .$$1 in the free Proca loop on a curved background (Ruf et al., 2018).

For generalized Proca with a background-dependent symmetric positive-definite mass tensor $$\mathcal{L}_{\text{Proca}}=-\frac14 F_{\mu\nu}F^{\mu\nu}+\frac12 m^2 A_\mu A^\mu .$$2, standard heat-kernel methods are not directly applicable. The solution is a Stueckelberg trick combined with a Weyl transformation and a bimetric reformulation. The fluctuation operator becomes block-minimal and non-degenerate in the vector–scalar sector, and the final divergences are local but non-polynomial in $$\mathcal{L}_{\text{Proca}}=-\frac14 F_{\mu\nu}F^{\mu\nu}+\frac12 m^2 A_\mu A^\mu .$$3 and its derivatives through an effective metric $$\mathcal{L}_{\text{Proca}}=-\frac14 F_{\mu\nu}F^{\mu\nu}+\frac12 m^2 A_\mu A^\mu .$$4, the difference tensor $$\mathcal{L}_{\text{Proca}}=-\frac14 F_{\mu\nu}F^{\mu\nu}+\frac12 m^2 A_\mu A^\mu .$$5, and irreducible multimetric tensor integrals $$\mathcal{L}_{\text{Proca}}=-\frac14 F_{\mu\nu}F^{\mu\nu}+\frac12 m^2 A_\mu A^\mu .$$6 (Ruf et al., 2018).

Not all non-minimal gravitational couplings are consistent. For a Proca field with $$\mathcal{L}_{\text{Proca}}=-\frac14 F_{\mu\nu}F^{\mu\nu}+\frac12 m^2 A_\mu A^\mu .$$7, the Ricci-tensor term generically couples the longitudinal vector and tensor graviton modes so that both become strongly coupled at the same scale. The inconsistency is removed by a vector-type disformal transformation,

$$\mathcal{L}_{\text{Proca}}=-\frac14 F_{\mu\nu}F^{\mu\nu}+\frac12 m^2 A_\mu A^\mu .$$8

which maps the dangerous Ricci-tensor coupling into a Ricci-scalar coupling plus higher-order, Planck-suppressed vector self-interactions. In the resulting disformal frame, only the longitudinal mode becomes strongly coupled, while tensor and transverse modes remain weakly coupled up to the Planck scale; the same transformation also removes the runaway modes reported on FRW backgrounds (Hell, 2024).

5. Black holes, cosmology, and regularized Gauss–Bonnet realizations

A particularly direct notion of “regularised Proca” appears in Einstein–Proca theories with curvature couplings that regulate the asymptotics induced by the mass term. In the four-dimensional extended Proca theory

$$\mathcal{L}_{\text{Proca}}=-\frac14 F_{\mu\nu}F^{\mu\nu}+\frac12 m^2 A_\mu A^\mu .$$9

the Einstein-tensor coupling preserves second-order field equations and modifies the effective mass in an $A_0$0 background to $A_0$1. For asymptotically $A_0$2 solutions one has $A_0$3, so $A_0$4 asymptotically. This is the mechanism by which the higher-order term regularizes the effect of the Proca mass term. The theory admits asymptotically flat, AdS, and Lifshitz black holes, and for $A_0$5 also admits particle-like solitons with regular and non-trivial geometry everywhere (Babichev et al., 2017).

A different regularization scheme begins from the Gauss–Bonnet invariant in Weyl geometry. In four dimensions, the finite regularized density is

$A_0$6

and the action

$A_0$7

maps into the generalized Proca class with

$A_0$8

Static spherically symmetric solutions carry a primary hair $A_0$9 that deforms the metric and an additional integration constant $\partial_\mu A^\mu=0$0 that is hidden in the seed solution but becomes a second, independent primary hair after a disformal transformation. In the disformed geometry this extra hair acts as an effective cosmological constant even with no bare cosmological constant term (Charmousis et al., 17 Apr 2025).

The same Weyl-geometry regularization program has also been implemented in three dimensions. There the regularized vector–tensor Gauss–Bonnet density is

$\partial_\mu A^\mu=0$1

which again sits in the generalized Proca class. The resulting theory admits asymptotically AdS$\partial_\mu A^\mu=0$2, static, circularly symmetric black holes with primary Proca hair, and can be generalized further by including the scalar-tensor regularized Gauss–Bonnet coupling or an electric charge (Alkac et al., 5 Aug 2025).

6. Quantum regularization, massless limits, and operator formulations

A mathematically distinct use of “regularised Proca” treats the massive vector as a regulator for Maxwell theory on curved spacetimes. On an arbitrary globally hyperbolic spacetime, the Proca operator $\partial_\mu A^\mu=0$3 is Green-hyperbolic and has well-defined advanced and retarded Green operators. The classical and quantum zero-mass limits exist only after restricting observables to co-closed test one-forms, thereby implementing gauge equivalence by exact distributional one-forms. To recover Maxwell dynamics in the limit one must also restrict the initial data so that the Lorenz-type constraint is well behaved and the current is conserved (Schambach, 2017).

Adiabatic regularization on a four-dimensional FLRW background sharpens this picture. The two transverse Proca polarizations obey the same mode equation as a conformally coupled scalar, whereas the longitudinal mode obeys the same mode equation as a minimally coupled scalar. After fourth-order adiabatic subtraction, the renormalized massless-limit identity is

$\partial_\mu A^\mu=0$4

This exhibits the discontinuity of the Proca massless limit: the longitudinal polarization does not disappear, but instead survives as a minimally coupled scalar. The paper explicitly interprets this extra contribution as a Stueckelberg-type field and verifies the result in de Sitter space (Marañón-González et al., 2023).

Regularization also appears in lattice formulations. By eliminating the non-dynamical $\partial_\mu A^\mu=0$5 and placing only the spatial components on a spatial lattice, one obtains a coupled system of harmonic oscillators with lattice dispersion $\partial_\mu A^\mu=0$6. In this setting, Nielsen’s approach gives a ground-state complexity that is an extensive mode sum of logarithmic frequency ratios, and a thermofield-double complexity with logarithmic time growth, whereas the Fubini–Study approach yields linear time growth on the SU(1,1)/U(1) geometry (Meng et al., 2021).

Finally, operator and dual formulations show that the regularised Proca idea is not confined to classical Lagrangian engineering. Quantisation of the $\partial_\mu A^\mu=0$7 supersymmetric spinning worldline with an extra oscillator multiplet reproduces the BV-extended spectrum of Proca theory together with a Stueckelberg field, and background consistency of the BRST charge yields the Proca equations (Carosi et al., 2021). A distinct Kalb–Ramond-based master action reproduces all purely fermionic 1PI functions of Proca electrodynamics to all orders, while also showing that generic derivative-current Kalb–Ramond couplings are non-renormalizable and break weak duality under radiative corrections (Gracia, 2023).

7. Conceptual synthesis and scope

Taken together, these constructions show that “Regularised Proca Theory” is best understood as a research program rather than a single model. In healthy effective field theory, it denotes generalized Proca, beyond generalized Proca, Proca-Nuevo, Extended Proca-Nuevo, and multi-Proca systems designed so that the primary and secondary constraints eliminate the unwanted fourth polarization. In gravitational physics, it includes curvature-dressed vector–tensor theories in which non-minimal couplings, disformal transformations, or Weyl-geometric regularization tame asymptotics, strong coupling, or topological triviality. In quantum field theory, it includes the use of a nonzero Proca mass as a regulator for Maxwell observables, adiabatic renormalization in cosmological backgrounds, and lattice or worldline constructions (Cadavid et al., 2019, Rham et al., 2020, Hell, 2024).

A recurrent misconception is that any non-minimal or higher-derivative extension of Proca is automatically healthy. The literature surveyed here points in the opposite direction. Health depends on a detailed degeneracy analysis, on the preservation or controlled deformation of the Proca constraint, and on the treatment of covariantization, background dependence, and renormalization. This suggests that the most stable use of the term is technical rather than generic: a “regularised” Proca theory is one in which the massive vector sector has been completed, covariantized, or quantized in a way that preserves the intended number of degrees of freedom and renders the dynamics well defined in the regime under study (Ruf et al., 2018, Rham et al., 2021).