Generalized Hybrid Metric-Palatini Gravity

• Generalized hybrid metric-Palatini theories are modified gravity models that combine both metric and Palatini curvature terms into a unified action.
• The scalar-tensor representation introduces two gravitational scalars with nonminimal curvature couplings, influencing cosmological dynamics and weak-field limits.
• Applications range from predicting de Sitter cosmologies and modified light propagation to enhancing compact star masses relative to general relativity.

Searching arXiv for recent and foundational papers on generalized hybrid metric-Palatini gravity. Generalized hybrid metric-Palatini theories are modified gravitational theories whose action depends on both the metric Ricci scalar $R$ and a Palatini scalar $\mathcal R$ built from an independent torsionless connection. They extend the standard hybrid construction $R+f(\mathcal R)$ to an arbitrary function $f(R,\mathcal R)$, thereby combining metric and Palatini variational sectors in a single framework. In scalar-tensor form, the generalized theory is dynamically equivalent to a two-scalar system with a nonminimal curvature coupling and a characteristic kinetic structure, and it has been used in cosmology, weak-field gravity, compact objects, junction problems, topological defects, and braneworlds (Capozziello et al., 2015, Rosa et al., 2020).

1. Definition and variational structure

In four dimensions, a standard form of the generalized hybrid metric-Palatini action is

$$S=\frac{1}{2\kappa^2}\int d^4x \sqrt{-g}\, f(R,\mathcal R) + S_M,$$

with $\kappa^2=8\pi G$, $R\equiv g^{\mu\nu}R_{\mu\nu}(g)$ the metric Ricci scalar, and $$\mathcal R\equiv g^{\mu\nu}\mathcal R_{\mu\nu}(\hat\Gamma)$$ the Palatini scalar formed from an a priori independent connection $\hat\Gamma^\lambda_{\mu\nu}$ (Golsanamlou et al., 2023). In five dimensions, the same construction is used with $$S=(1/2\kappa^2)\int_\Omega d^5x\sqrt{-g}\,f(R,\mathcal R)+S_m(g_{MN},\chi)$$, coordinates $\mathcal R$0, and signature $\mathcal R$1 (Rosa et al., 2020).

Varying the action with respect to the metric and independent connection yields

$\mathcal R$2

together with

$\mathcal R$3

so that the independent connection is Levi-Civita with respect to the conformally related metric $\mathcal R$4 (Golsanamlou et al., 2023). In consequence, the Palatini curvature can be rewritten in terms of metric curvature and derivatives of $\mathcal R$5.

The standard hybrid subclass is obtained by restricting to $\mathcal R$6. The generalized extension allows any $\mathcal R$7 and, in the scalar-tensor form used in several papers, introduces two gravitational scalar degrees of freedom instead of one (Rosa et al., 2020). This is the central structural distinction from metric $\mathcal R$8, Palatini $\mathcal R$9, and non-generalized hybrid models.

The literature also contains broader variational generalizations. A class of scalar-tensor theories with non-metricity $R+f(\mathcal R)$0 unifies metric, Palatini, and hybrid metric-Palatini gravitational actions with non-minimal interaction, is closed under a generalized conformal group acting independently on the metric, connection, and scalar field, and is on-shell equivalent to a purely metric scalar-tensor theory (Borowiec et al., 2020). By contrast, the bimetric reformulation of Goenner’s variational principle can be physically distinct from both the metric and the metric-affine ones, even for the Einstein-Hilbert action (Koivisto, 2011).

2. Scalar-tensor representations and frame structure

A common scalar-tensor representation introduces auxiliary variables and defines

$R+f(\mathcal R)$1

leading to

$R+f(\mathcal R)$2

After eliminating the Palatini curvature through the conformal relation, this becomes

$R+f(\mathcal R)$3

(Rosa et al., 2017, Rosa et al., 2021). In five dimensions an analogous construction yields

$R+f(\mathcal R)$4

which is the form used for thick branes (Rosa et al., 2020).

Field equations then take the form of modified Einstein equations coupled to two scalar equations. In the four-dimensional scalar-tensor representation,

$R+f(\mathcal R)$5

while

$R+f(\mathcal R)$6

(Rosa et al., 2021). In this Jordan-like frame, $R+f(\mathcal R)$7 supplies the effective nonminimal coupling, and only $R+f(\mathcal R)$8 carries an explicit kinetic term.

Different sign conventions also appear. In the weak-field analysis, the scalar-tensor action is written as

$R+f(\mathcal R)$9

followed by field redefinitions and a conformal transformation to the Einstein frame (Rosa et al., 2021). In the Einstein frame of both generalized hybrid metric-Palatini gravity and the linear-$f(R,\mathcal R)$0 subclass of $f(R,\mathcal R)$1 theories, the two-scalar kinetic geometry takes the same form,

$f(R,\mathcal R)$2

which underlies a local reconstruction map between the two formulations (Ramírez, 6 May 2026).

The generalized hybrid metric-Palatini scalar-tensor framework has also been extended to scalar-modulated actions of the form

$f(R,\mathcal R)$3

and, for the tractable specialization $f(R,\mathcal R)$4, to

$f(R,\mathcal R)$5

The linear $f(R,\mathcal R)$6 subclass freezes the Palatini scalar and reduces to a single-scalar-tensor theory with shifted effective Planck mass and effective cosmological constant (Pereira et al., 12 Nov 2025).

3. Cosmology, light propagation, and dynamical systems

For FLRW spacetimes,

$f(R,\mathcal R)$7

the scalar-tensor equations become modified Friedmann, Raychaudhuri, and scalar-field equations (Rosa et al., 2017). In one standard notation,

$f(R,\mathcal R)$8

$f(R,\mathcal R)$9

The exact-solution program based on ansätze for $$S=\frac{1}{2\kappa^2}\int d^4x \sqrt{-g}\, f(R,\mathcal R) + S_M,$$0 or for the effective potential yields de Sitter solutions, power-law solutions, nonflat vacuum solutions, and non-vacuum exponentially expanding flat solutions. In particular, it is possible to obtain exponentially expanding solutions for flat universes even when the cosmology is not purely vacuum (Rosa et al., 2017).

The reconstruction problem is encoded in

$$S=\frac{1}{2\kappa^2}\int d^4x \sqrt{-g}\, f(R,\mathcal R) + S_M,$$1

For the de Sitter potential $$S=\frac{1}{2\kappa^2}\int d^4x \sqrt{-g}\, f(R,\mathcal R) + S_M,$$2, one reconstructed class is

$$S=\frac{1}{2\kappa^2}\int d^4x \sqrt{-g}\, f(R,\mathcal R) + S_M,$$3

with $$S=\frac{1}{2\kappa^2}\int d^4x \sqrt{-g}\, f(R,\mathcal R) + S_M,$$4 arbitrary, subject to invertibility restrictions (Rosa et al., 2017).

The FLRW phase space has also been studied with autonomous systems built from variables such as $$S=\frac{1}{2\kappa^2}\int d^4x \sqrt{-g}\, f(R,\mathcal R) + S_M,$$5, $$S=\frac{1}{2\kappa^2}\int d^4x \sqrt{-g}\, f(R,\mathcal R) + S_M,$$6, $$S=\frac{1}{2\kappa^2}\int d^4x \sqrt{-g}\, f(R,\mathcal R) + S_M,$$7, $$S=\frac{1}{2\kappa^2}\int d^4x \sqrt{-g}\, f(R,\mathcal R) + S_M,$$8, $$S=\frac{1}{2\kappa^2}\int d^4x \sqrt{-g}\, f(R,\mathcal R) + S_M,$$9, $\kappa^2=8\pi G$0, $\kappa^2=8\pi G$1, and $\kappa^2=8\pi G$2. For the models analyzed, no global attractors can be present because the invariant submanifold $\kappa^2=8\pi G$3 is also a singular boundary of the cosmological equations (Rosa et al., 2019). Two different classes of scale-factor behavior appear at fixed points: one with finite-time divergence and one that asymptotically approaches constant values (Rosa et al., 2019, Rosa, 2019).

Light propagation in these theories has been treated through the geodesic deviation equation. In FLRW, where the Weyl tensor vanishes, the generalized GDE takes the form

$\kappa^2=8\pi G$4

for the hybrid $\kappa^2=8\pi G$5 subclass, with corresponding modified Raychaudhuri and null-focusing conditions (Golsanamlou et al., 2023). The redshift-space null GDE becomes a second-order ODE,

$\kappa^2=8\pi G$6

and the observer area distance is

$\kappa^2=8\pi G$7

For the case $\kappa^2=8\pi G$8, the reported numerical behavior shows that both $\kappa^2=8\pi G$9 and $R\equiv g^{\mu\nu}R_{\mu\nu}(g)$0 exhibit a peak at a certain redshift, while at low redshift the behavior is close to $R\equiv g^{\mu\nu}R_{\mu\nu}(g)$1CDM (Golsanamlou et al., 2023).

4. Weak-field, strong-field, and compact stars

In the weak-field, slow-motion regime, the Einstein-frame analysis of the generalized theory produces coupled scalar Helmholtz equations and Jordan-frame metric potentials with two Yukawa contributions. For a spherical source,

$R\equiv g^{\mu\nu}R_{\mu\nu}(g)$2

$R\equiv g^{\mu\nu}R_{\mu\nu}(g)$3

so the generic phenomenology contains both attractive and repulsive Yukawa-type corrections (Rosa et al., 2021). Two special cases require separate treatment: a non-diagonalizable mass matrix, and a decoupled case $R\equiv g^{\mu\nu}R_{\mu\nu}(g)$4 yielding a single attractive Yukawa correction (Rosa et al., 2021).

A later extension, the hybrid metric-Palatini scalar-tensor theory with scalar modulation, yields analytic cosmological, strong-field, and weak-field sectors for the linear-$R\equiv g^{\mu\nu}R_{\mu\nu}(g)$5 subclass. In the weak field, the Newtonian potential becomes

$R\equiv g^{\mu\nu}R_{\mu\nu}(g)$6

and GR is recovered for heavy scalars or weak mixing (Pereira et al., 12 Nov 2025). In the strong field, the same subclass generalizes the Janis-Newman-Winicour and Buchdahl metrics and includes the Schwarzschild-de Sitter limit (Pereira et al., 12 Nov 2025).

Compact-star applications show that the extra scalar sector can substantially modify stellar equilibrium. For quark stars in generalized hybrid metric-Palatini gravity, the numerical solutions with MIT bag and CFL equations of state show that compact objects are more massive than their GR counterparts, with higher compactness and higher surface redshift (Aliannejadi et al., 2024). For MIT bag stars, the reported maxima reach $R\equiv g^{\mu\nu}R_{\mu\nu}(g)$7, $R\equiv g^{\mu\nu}R_{\mu\nu}(g)$8, and $R\equiv g^{\mu\nu}R_{\mu\nu}(g)$9, compared with the GR baseline $$\mathcal R\equiv g^{\mu\nu}\mathcal R_{\mu\nu}(\hat\Gamma)$$0, $$\mathcal R\equiv g^{\mu\nu}\mathcal R_{\mu\nu}(\hat\Gamma)$$1, and $$\mathcal R\equiv g^{\mu\nu}\mathcal R_{\mu\nu}(\hat\Gamma)$$2 (Aliannejadi et al., 2024). For CFL stars, the reported maxima reach $$\mathcal R\equiv g^{\mu\nu}\mathcal R_{\mu\nu}(\hat\Gamma)$$3, $$\mathcal R\equiv g^{\mu\nu}\mathcal R_{\mu\nu}(\hat\Gamma)$$4, and $$\mathcal R\equiv g^{\mu\nu}\mathcal R_{\mu\nu}(\hat\Gamma)$$5, compared with the GR baseline $$\mathcal R\equiv g^{\mu\nu}\mathcal R_{\mu\nu}(\hat\Gamma)$$6, $$\mathcal R\equiv g^{\mu\nu}\mathcal R_{\mu\nu}(\hat\Gamma)$$7, and $$\mathcal R\equiv g^{\mu\nu}\mathcal R_{\mu\nu}(\hat\Gamma)$$8 (Aliannejadi et al., 2024).

These compact-star results continue a broader hybrid metric-Palatini stellar program in which neutron, quark, and Bose-Einstein condensate stars in the non-generalized hybrid theory were found to be more massive than their GR counterparts, and in which a constant scalar field forces the matter equation of state into a bag-model form (Danila et al., 2016). This suggests that the mass-enhancement mechanism survives, and in generalized settings can be strengthened by the bi-scalar sector. A plausible implication is that the effective scalar contributions can mimic additional self-gravity without introducing exotic matter content in the fluid sector.

5. Junctions, branes, strings, and wormholes

The junction conditions of generalized hybrid metric-Palatini gravity have been derived both in geometric and scalar-tensor representations. For thin-shell matchings in the geometric form, one requires

$$\mathcal R\equiv g^{\mu\nu}\mathcal R_{\mu\nu}(\hat\Gamma)$$9

together with the shell equation

$\hat\Gamma^\lambda_{\mu\nu}$0

and the derivative-jump relation

$\hat\Gamma^\lambda_{\mu\nu}$1

(Rosa et al., 2021). These conditions are stronger than the GR Israel conditions.

Their applications include a star with Minkowski interior, thin shell, and Schwarzschild exterior for which the matching can only be performed at

$\hat\Gamma^\lambda_{\mu\nu}$2

the Buchdahl radius, with shell stresses

$\hat\Gamma^\lambda_{\mu\nu}$3

satisfying all energy conditions (Rosa et al., 2021). The same formalism yields a quasistar with an interior Schwarzschild black hole, a thick shell smoothly matched to a Schwarzschild exterior at the light ring, and a wormhole with a Schwarzschild-AdS exterior for which the null energy condition is obeyed (Rosa et al., 2021).

Wormholes had already been constructed in the scalar-tensor representation with a three-region architecture consisting of an interior containing the throat, a thin shell, and a vacuum Schwarzschild-AdS exterior, such that the matter NEC is verified for the entire spacetime (Rosa et al., 2018). With

$\hat\Gamma^\lambda_{\mu\nu}$4

and $\hat\Gamma^\lambda_{\mu\nu}$5 constant, one gets

$\hat\Gamma^\lambda_{\mu\nu}$6

while the matter combinations $\hat\Gamma^\lambda_{\mu\nu}$7 and $\hat\Gamma^\lambda_{\mu\nu}$8 can be made nonnegative near the throat and, after matching, everywhere (Rosa et al., 2018).

Generalized hybrid metric-Palatini gravity also supports cylindrically symmetric $\hat\Gamma^\lambda_{\mu\nu}$9 local strings in scalar-tensor form. Under boost invariance along $$S=(1/2\kappa^2)\int_\Omega d^5x\sqrt{-g}\,f(R,\mathcal R)+S_m(g_{MN},\chi)$$0 and $$S=(1/2\kappa^2)\int_\Omega d^5x\sqrt{-g}\,f(R,\mathcal R)+S_m(g_{MN},\chi)$$1, the metric reduces to

$$S=(1/2\kappa^2)\int_\Omega d^5x\sqrt{-g}\,f(R,\mathcal R)+S_m(g_{MN},\chi)$$2

For a constant potential $$S=(1/2\kappa^2)\int_\Omega d^5x\sqrt{-g}\,f(R,\mathcal R)+S_m(g_{MN},\chi)$$3, the paper finds the exact solution

$$S=(1/2\kappa^2)\int_\Omega d^5x\sqrt{-g}\,f(R,\mathcal R)+S_m(g_{MN},\chi)$$4

with deficit angle

$$S=(1/2\kappa^2)\int_\Omega d^5x\sqrt{-g}\,f(R,\mathcal R)+S_m(g_{MN},\chi)$$5

and

$$S=(1/2\kappa^2)\int_\Omega d^5x\sqrt{-g}\,f(R,\mathcal R)+S_m(g_{MN},\chi)$$6

(Silva et al., 2021). For the interacting potential $$S=(1/2\kappa^2)\int_\Omega d^5x\sqrt{-g}\,f(R,\mathcal R)+S_m(g_{MN},\chi)$$7, numerical integration yields several stable string configurations whose basic parameters depend essentially on the effective field potential and on the boundary conditions (Silva et al., 2021).

In five dimensions, thick brane solutions are obtained from the scalar-tensor representation with two gravitational scalars, optionally supplemented by a matter scalar $$S=(1/2\kappa^2)\int_\Omega d^5x\sqrt{-g}\,f(R,\mathcal R)+S_m(g_{MN},\chi)$$8. The tensor perturbation equation reduces to a supersymmetric Schrödinger problem,

$$S=(1/2\kappa^2)\int_\Omega d^5x\sqrt{-g}\,f(R,\mathcal R)+S_m(g_{MN},\chi)$$9

with factorization $\mathcal R$00, implying $\mathcal R$01 and the absence of tachyonic tensor instabilities (Rosa et al., 2020). The graviton zero mode

$\mathcal R$02

is normalizable for the three classes of backgrounds studied there, and in the matter-supported Model 2 it can widen and split into two peaks, signaling internal structure in the localized graviton profile (Rosa et al., 2020).

6. Comparative structure, reconstruction, and contested limits

A recurrent clarification in this subject is that generalized hybrid metric-Palatini gravity is not merely metric $\mathcal R$03 plus a Palatini correction. In the comparative summary of the thick-brane analysis, metric $\mathcal R$04 is characterized by an extra scalar tied directly to $\mathcal R$05, Palatini $\mathcal R$06 by an extra scalar algebraically related to the matter trace and liable to gradient instabilities, standard hybrid $\mathcal R$07 by one hybrid scalar, and generalized $\mathcal R$08 by two gravitational scalars with effective coupling $\mathcal R$09 and a noncanonical kinetic term for $\mathcal R$10 (Rosa et al., 2020).

The generalized non-metric scalar-tensor framework sharpens this comparison by introducing generalized invariants and a frame-independent criterion for a non-dynamical scalar,

$\mathcal R$11

which selects the Palatini subclass (Borowiec et al., 2020). In that same framework, the metric Starobinsky embedding remains viable, but the minimal hybrid Starobinsky embedding has

$\mathcal R$12

so slow-roll fails and the paper’s estimate gives $\mathcal R$13, making that minimal hybrid realization observationally disfavored (Borowiec et al., 2020).

The relation to alternative variational principles is likewise nontrivial. Goenner’s variational principle, in which an independent metric generates the spacetime connection, is equivalent to the usual Palatini theory, but its bimetric reformulation is physically distinct from both the metric and the metric-affine ones, even for the Einstein-Hilbert action (Koivisto, 2011). This matters because generalized hybrid theories are often discussed alongside C-theories and other connection-based extensions; not every such reformulation preserves the same propagating content.

Recent reconstruction results further enlarge the theory space. In vacuum and on regular branches, generalized hybrid metric-Palatini gravity and $\mathcal R$14 theories with linear dependence on $\mathcal R$15 share the same Einstein-frame two-scalar field-space geometry, so the map between them reduces to matching Einstein-frame potentials (Ramírez, 6 May 2026). The forward reconstruction from $\mathcal R$16 to $\mathcal R$17 is governed by a Clairaut-type first-order PDE,

$\mathcal R$18

while the inverse reconstruction is not unique and is parametrized by the kinetic coupling (Ramírez, 6 May 2026). This suggests that the Einstein-frame two-scalar sector is, locally, a more primitive classification object than the original Jordan-frame action.

Taken together, these developments depict generalized hybrid metric-Palatini theories as a family of two-curvature, typically two-scalar gravitational models whose distinguishing features are the coexistence of metric and Palatini curvature sectors, the scalar-tensor coupling structure, and the possibility of controlling cosmological, weak-field, and localized gravitational phenomena through functions of both $\mathcal R$19 and $\mathcal R$20. The same body of work also shows that viability is highly model-dependent: some embeddings pass local tests through small couplings or large scalar masses, some cosmological models lack global attractors, and some inflationary realizations are ruled out in minimal form (Rosa et al., 2021, Rosa et al., 2019, Borowiec et al., 2020).