Higher-Order Curvature-Scalar Gravity

• Higher-Order Curvature-Scalar Gravity describes theories that extend Einstein-Hilbert action with nonlinear curvature invariants and scalar fields to capture richer dynamics.
• These models include formulations like f(R), Gauss-Bonnet, and Lovelock-type frameworks with applications in inflation, bounce cosmology, and black hole phenomenology.
• Precise tuning and symmetry constraints are essential to avoid ghost modes and instabilities, and the theories are studied in both classical and discrete quantum-gravity contexts.

In the literature surveyed here, higher-order curvature-scalar gravity functions as an umbrella description for gravitational theories in which the Einstein-Hilbert action is extended by nonlinear curvature invariants and, in many formulations, by explicit scalar fields or scalar degrees of freedom generated through duality transformations. Typical ingredients include $R^2$, $R^3$, $R\square R$, $RR_{\mu\nu}R^{\mu\nu}$, Gauss-Bonnet couplings, and general $\mathcal{L}(\text{Riemann})$ contractions; their applications range from inflation and nonsingular bounces to black holes, holography, braneworld localization, and discrete quantum-gravity constructions (Bueno et al., 2016, Corelli et al., 20 Oct 2025, Moreno et al., 2023).

1. Scope and representative structures

A broad formulation is provided by general $\mathcal{L}(\text{Riemann})$ theories in arbitrary dimension,

$$S=\int_{\mathcal{M}} d^D x \sqrt{|g|}\,\left[\mathcal{L}(R_{\mu\nu\rho\sigma},g^{\alpha\beta})+L_{\rm matter}\right],$$

whose field equations can be written in terms of

$$\mathcal{E}_{\mu\nu}\equiv P_{\mu}^{\ \sigma\rho\lambda} R_{\nu\sigma\rho\lambda} - \frac{1}{2}g_{\mu\nu}\mathcal{L} - 2\nabla^{\alpha}\nabla^{\beta} P_{\mu\alpha\beta\nu} = \frac{1}{2} T_{\mu\nu},$$

with $$P^{\mu\nu\rho\sigma}\equiv \partial \mathcal{L}/\partial R_{\mu\nu\rho\sigma}$$. Bueno, Cano, Min, and Visser showed that such theories can be classified on maximally symmetric backgrounds by four constants $(a,b,c,e)$, yielding a spectrum that may contain a massless graviton, a ghost-like massive spin-2 mode, and a scalar mode. The corresponding masses are

$R^3$0

$R^3$1

so the absence of the massive spin-2 ghost requires $R^3$2, while removal of the scalar requires $R^3$3 (Bueno et al., 2016).

Within this broad class, several concrete higher-order curvature-scalar frameworks recur across the literature.

Framework	Representative action	Scalar-sector role
General $R^3$4 gravity	$R^3$5	May propagate a scalar mode depending on $R^3$6
$R^3$7 bounce model	$R^3$8	Effective matter coupling through $R^3$9 (2206.12423)
$R\square R$0-dGB theory	$R\square R$1	Two nonminimally coupled scalar fields after scalarization (Corelli et al., 20 Oct 2025)
HOCG black-hole framework	$R\square R$2	Scalar coupled to higher-curvature invariants (Filho et al., 15 Sep 2025)

The scalar sector is not always inserted by hand. In some theories it is explicit, as in $R\square R$3 models or Einstein-dilaton-Gauss-Bonnet-type constructions; in others it emerges through conformal or Legendre transformations of higher-curvature terms. This suggests that “curvature-scalar” is best understood as a structural relation rather than a single Lagrangian ansatz.

2. Scalarization, dual frames, and formulation dependence

For $R\square R$4 theories, the scalar-tensor correspondence is mathematically explicit. In the regular-spacetime analysis of $R\square R$5 gravity, a conformal transformation

$R\square R$6

together with

$R\square R$7

maps the Jordan-frame action to Einstein gravity plus a scalar field with potential

$R\square R$8

The same work shows that the equivalence extends from the action to the field equations, and even to their projection on a lower-dimensional hypersurface, provided the conformal transformation is regular (Chakraborty et al., 2016).

A five-dimensional braneworld realization of this correspondence uses $R\square R$9. After introducing an auxiliary field and performing the conformal transformation $RR_{\mu\nu}R^{\mu\nu}$0, the theory becomes Einstein gravity plus a scalar field $RR_{\mu\nu}R^{\mu\nu}$1. In that setting, fermion localization properties are identical in the $RR_{\mu\nu}R^{\mu\nu}$2 and scalar-tensor descriptions: increasing $RR_{\mu\nu}R^{\mu\nu}$3 localizes massless chiral modes toward the TeV brane, while massive KK fermions localize toward the Planck brane, and the radion-fermion coupling for massive KK modes is strongly suppressed on the visible brane (Mitra et al., 2017).

Formulation dependence is especially sharp in Palatini and metric-affine theories. In the Palatini treatment of Higgs inflation with higher-order gravity, the Jordan-frame action

$RR_{\mu\nu}R^{\mu\nu}$4

is restricted to terms built from the symmetric part of the Ricci tensor. In this sector there are no new gravitational degrees of freedom; the scalar perturbation spectrum is unchanged, while the tensor perturbation spectrum is suppressed by the higher-order curvature couplings. The Einstein-frame potential becomes

$RR_{\mu\nu}R^{\mu\nu}$5

and the tensor-to-scalar ratio is

$RR_{\mu\nu}R^{\mu\nu}$6

(Annala, 2021).

That simplification is not generic. In metric-affine Ricci-based gravity, the claim that higher-order curvature theories are generically ghost-free is explicitly rejected. When projective symmetry is broken and the action depends on the full Ricci tensor rather than only $RR_{\mu\nu}R^{\mu\nu}$7, the non-projectively invariant sector propagates ghost-like degrees of freedom. The pathology can be avoided either by imposing projective symmetry,

$RR_{\mu\nu}R^{\mu\nu}$8

or by enforcing a torsion-free constraint $RR_{\mu\nu}R^{\mu\nu}$9 (Jiménez et al., 2019). A common misconception is therefore that “Palatini” or “metric-affine” automatically removes higher-derivative pathologies; the cited results show that this is true only in specially constrained sectors.

3. Cosmology: second-order FLRW sectors, inflation, and bounces

A major line of work isolates higher-curvature theories whose FLRW equations remain second order. “Cosmological Gravities” are defined as higher-curvature theories whose equations of motion for FLRW configurations are of second order in time derivatives for the scale factor, just as in Einstein gravity. An all-orders family in $\mathcal{L}(\text{Riemann})$0 is

$\mathcal{L}(\text{Riemann})$1

with a unique nontrivial equivalence class at each curvature order. In these theories, the linearized equations for scalar cosmological perturbations contain no more than two time derivatives, and explicit four-dimensional examples were given up to fifth order in the curvature (Moreno et al., 2023).

An earlier FLRW construction based on Lovelock tensors reached a related conclusion from a different direction. By constructing higher-order terms linear in $\mathcal{L}(\text{Riemann})$2, the resulting Lagrangians can be reduced by integration by parts to dependence only on $\mathcal{L}(\text{Riemann})$3 and $\mathcal{L}(\text{Riemann})$4, so no scalar degrees of freedom appear in the cosmological model. The resulting dynamics takes the form of $\mathcal{L}(\text{Riemann})$5 cosmology, with

$\mathcal{L}(\text{Riemann})$6

and inflationary behavior of the scale factor can arise without introducing ad hoc inflaton fields (Kan et al., 2012).

Inflationary model building in higher-order curvature-scalar gravity has proceeded along several distinct routes. One route extends Starobinsky inflation by including all scalar-curvature corrections up to second order: $\mathcal{L}(\text{Riemann})$7 In the Einstein frame this becomes a two-field model. Phase-space analysis exhibits a slow-roll attractor and a stable spiral at the origin for $\mathcal{L}(\text{Riemann})$8, scalar and isocurvature perturbations can be separated into $\mathcal{L}(\text{Riemann})$9 and $\mathcal{L}(\text{Riemann})$0, and observational agreement with Planck 2018, BICEP3/Keck, and BAO is obtained for approximately $\mathcal{L}(\text{Riemann})$1 and $\mathcal{L}(\text{Riemann})$2 (Rodrigues-da-Silva et al., 2022).

A dimension-six extension focused on

$\mathcal{L}(\text{Riemann})$3

derives a four-dimensional autonomous system in the Jordan frame and linearizes in the cubic parameter $\mathcal{L}(\text{Riemann})$4. The slow-roll potential becomes

$\mathcal{L}(\text{Riemann})$5

with inflationary observables

$\mathcal{L}(\text{Riemann})$6

Compatibility with Planck, BICEP/Keck, and BAO requires $\mathcal{L}(\text{Riemann})$7, while negative $\mathcal{L}(\text{Riemann})$8 restores compatibility with recent ACT, Planck, and DESI results (Morais et al., 11 Sep 2025).

A more radical possibility is inflation in a higher-curvature dominated regime. In Geometric Inflation models,

$\mathcal{L}(\text{Riemann})$9

with

$$S=\int_{\mathcal{M}} d^D x \sqrt{|g|}\,\left[\mathcal{L}(R_{\mu\nu\rho\sigma},g^{\alpha\beta})+L_{\rm matter}\right],$$0

If $$S=\int_{\mathcal{M}} d^D x \sqrt{|g|}\,\left[\mathcal{L}(R_{\mu\nu\rho\sigma},g^{\alpha\beta})+L_{\rm matter}\right],$$1 has a GR regime, a transition or flat regime, and a steep inflating regime, then a minimally coupled scalar with quadratic potential can drive more than 100 e-folds with $$S=\int_{\mathcal{M}} d^D x \sqrt{|g|}\,\left[\mathcal{L}(R_{\mu\nu\rho\sigma},g^{\alpha\beta})+L_{\rm matter}\right],$$2 in explicit examples (Edelstein et al., 2020).

Bounce cosmology provides the complementary use of higher-curvature terms. In $$S=\int_{\mathcal{M}} d^D x \sqrt{|g|}\,\left[\mathcal{L}(R_{\mu\nu\rho\sigma},g^{\alpha\beta})+L_{\rm matter}\right],$$3 gravity with

$$S=\int_{\mathcal{M}} d^D x \sqrt{|g|}\,\left[\mathcal{L}(R_{\mu\nu\rho\sigma},g^{\alpha\beta})+L_{\rm matter}\right],$$4

and a power-law Hubble parameter

$$S=\int_{\mathcal{M}} d^D x \sqrt{|g|}\,\left[\mathcal{L}(R_{\mu\nu\rho\sigma},g^{\alpha\beta})+L_{\rm matter}\right],$$5

the conditions $$S=\int_{\mathcal{M}} d^D x \sqrt{|g|}\,\left[\mathcal{L}(R_{\mu\nu\rho\sigma},g^{\alpha\beta})+L_{\rm matter}\right],$$6, $$S=\int_{\mathcal{M}} d^D x \sqrt{|g|}\,\left[\mathcal{L}(R_{\mu\nu\rho\sigma},g^{\alpha\beta})+L_{\rm matter}\right],$$7, and $$S=\int_{\mathcal{M}} d^D x \sqrt{|g|}\,\left[\mathcal{L}(R_{\mu\nu\rho\sigma},g^{\alpha\beta})+L_{\rm matter}\right],$$8 can be satisfied, the equation-of-state parameter crosses the phantom divide $$S=\int_{\mathcal{M}} d^D x \sqrt{|g|}\,\left[\mathcal{L}(R_{\mu\nu\rho\sigma},g^{\alpha\beta})+L_{\rm matter}\right],$$9, and the bounce avoids the cosmological initial singularity (2206.12423). A related analysis in quadratic curvature gravity minimally coupled to a scalar field distinguishes two energy-momentum tensors. If only the scalar field is counted, NEC, WEC, and DEC remain satisfied while SEC is violated during the bounce. If an effective tensor is constructed to absorb higher-curvature corrections, all four energy conditions are violated near the bounce, emphasizing that the non-Einsteinian behavior resides in the gravitational sector rather than in exotic matter (Hashimoto et al., 24 Mar 2026).

4. Spherically symmetric solutions and black-hole phenomenology

Higher-order curvature-scalar gravity also generates a large class of static and spherically symmetric solutions. A six-derivative scalar-curvature theory with

$$\mathcal{E}_{\mu\nu}\equiv P_{\mu}^{\ \sigma\rho\lambda} R_{\nu\sigma\rho\lambda} - \frac{1}{2}g_{\mu\nu}\mathcal{L} - 2\nabla^{\alpha}\nabla^{\beta} P_{\mu\alpha\beta\nu} = \frac{1}{2} T_{\mu\nu},$$0

yields weak-field potentials of Yukawa type,

$$\mathcal{E}_{\mu\nu}\equiv P_{\mu}^{\ \sigma\rho\lambda} R_{\nu\sigma\rho\lambda} - \frac{1}{2}g_{\mu\nu}\mathcal{L} - 2\nabla^{\alpha}\nabla^{\beta} P_{\mu\alpha\beta\nu} = \frac{1}{2} T_{\mu\nu},$$1

with

$$\mathcal{E}_{\mu\nu}\equiv P_{\mu}^{\ \sigma\rho\lambda} R_{\nu\sigma\rho\lambda} - \frac{1}{2}g_{\mu\nu}\mathcal{L} - 2\nabla^{\alpha}\nabla^{\beta} P_{\mu\alpha\beta\nu} = \frac{1}{2} T_{\mu\nu},$$2

Stability restricts the parameters to

$$\mathcal{E}_{\mu\nu}\equiv P_{\mu}^{\ \sigma\rho\lambda} R_{\nu\sigma\rho\lambda} - \frac{1}{2}g_{\mu\nu}\mathcal{L} - 2\nabla^{\alpha}\nabla^{\beta} P_{\mu\alpha\beta\nu} = \frac{1}{2} T_{\mu\nu},$$3

so only Yukawa-type corrections are physically allowed. Non-Schwarzschild black holes might exist mathematically, but for reasonable values of the parameters their horizon radii are extremely small, $$\mathcal{E}_{\mu\nu}\equiv P_{\mu}^{\ \sigma\rho\lambda} R_{\nu\sigma\rho\lambda} - \frac{1}{2}g_{\mu\nu}\mathcal{L} - 2\nabla^{\alpha}\nabla^{\beta} P_{\mu\alpha\beta\nu} = \frac{1}{2} T_{\mu\nu},$$4, making macroscopic deviations from Schwarzschild unfeasible (Rodrigues-da-Silva et al., 2020).

The strong-field situation is more restrictive in Einsteinian cubic gravity. For the action

$$\mathcal{E}_{\mu\nu}\equiv P_{\mu}^{\ \sigma\rho\lambda} R_{\nu\sigma\rho\lambda} - \frac{1}{2}g_{\mu\nu}\mathcal{L} - 2\nabla^{\alpha}\nabla^{\beta} P_{\mu\alpha\beta\nu} = \frac{1}{2} T_{\mu\nu},$$5

odd-parity perturbations around static and spherically symmetric black holes reveal three propagating degrees of freedom rather than the single odd-parity mode of GR. At least one is always a ghost, and one mode has

$$\mathcal{E}_{\mu\nu}\equiv P_{\mu}^{\ \sigma\rho\lambda} R_{\nu\sigma\rho\lambda} - \frac{1}{2}g_{\mu\nu}\mathcal{L} - 2\nabla^{\alpha}\nabla^{\beta} P_{\mu\alpha\beta\nu} = \frac{1}{2} T_{\mu\nu},$$6

so static and spherically symmetric vacuum black holes supported by unsuppressed cubic curvature terms are excluded by ghost and Laplacian instabilities (Felice et al., 2023).

A distinct HOCG black-hole solution employs a scalar field coupled to higher-order curvature terms. In one formulation,

$$\mathcal{E}_{\mu\nu}\equiv P_{\mu}^{\ \sigma\rho\lambda} R_{\nu\sigma\rho\lambda} - \frac{1}{2}g_{\mu\nu}\mathcal{L} - 2\nabla^{\alpha}\nabla^{\beta} P_{\mu\alpha\beta\nu} = \frac{1}{2} T_{\mu\nu},$$7

and the static line element is built from

$$\mathcal{E}_{\mu\nu}\equiv P_{\mu}^{\ \sigma\rho\lambda} R_{\nu\sigma\rho\lambda} - \frac{1}{2}g_{\mu\nu}\mathcal{L} - 2\nabla^{\alpha}\nabla^{\beta} P_{\mu\alpha\beta\nu} = \frac{1}{2} T_{\mu\nu},$$8

This solution supports up to two horizons $$\mathcal{E}_{\mu\nu}\equiv P_{\mu}^{\ \sigma\rho\lambda} R_{\nu\sigma\rho\lambda} - \frac{1}{2}g_{\mu\nu}\mathcal{L} - 2\nabla^{\alpha}\nabla^{\beta} P_{\mu\alpha\beta\nu} = \frac{1}{2} T_{\mu\nu},$$9, admits Wald and Barrow entropies, exhibits Davies-type phase transitions where the heat capacity diverges, and has positive quasi-local energy and Gibbs free energy in the thermodynamic analysis. Geodesic deviation yields an explicit stability condition outside the event horizon (Nashed et al., 2 Sep 2025).

The phenomenology of a related static metric,

$$P^{\mu\nu\rho\sigma}\equiv \partial \mathcal{L}/\partial R_{\mu\nu\rho\sigma}$$0

has been analyzed in detail. Increasing $$P^{\mu\nu\rho\sigma}\equiv \partial \mathcal{L}/\partial R_{\mu\nu\rho\sigma}$$1 decreases the event horizon radius and increases the Cauchy horizon radius; the photon sphere is shifted to

$$P^{\mu\nu\rho\sigma}\equiv \partial \mathcal{L}/\partial R_{\mu\nu\rho\sigma}$$2

the shadow radius decreases, and quasinormal modes for scalar, vector, tensor, and spinorial perturbations become less damped. Weak-field and strong-deflection lensing are both modified, and comparison with EHT and Solar System data constrains $$P^{\mu\nu\rho\sigma}\equiv \partial \mathcal{L}/\partial R_{\mu\nu\rho\sigma}$$3: for M87* one finds $$P^{\mu\nu\rho\sigma}\equiv \partial \mathcal{L}/\partial R_{\mu\nu\rho\sigma}$$4, while the tightest Solar System bound comes from light deflection, $$P^{\mu\nu\rho\sigma}\equiv \partial \mathcal{L}/\partial R_{\mu\nu\rho\sigma}$$5 (Filho et al., 15 Sep 2025).

Combining $$P^{\mu\nu\rho\sigma}\equiv \partial \mathcal{L}/\partial R_{\mu\nu\rho\sigma}$$6 gravity with Einstein-dilaton-Gauss-Bonnet interactions produces a bi-scalar theory with second-order field equations,

$$P^{\mu\nu\rho\sigma}\equiv \partial \mathcal{L}/\partial R_{\mu\nu\rho\sigma}$$7

In that framework black holes are modified by $$P^{\mu\nu\rho\sigma}\equiv \partial \mathcal{L}/\partial R_{\mu\nu\rho\sigma}$$8 terms, retain the EdGB features of a minimum mass and multiple branches, and display a nontrivial suppression of Ricci-scalar divergence in the interior. Even so, the singularity structure and elliptic regions inside the horizon remain similar to those of pure EdGB gravity, so adding individual higher-order terms does not resolve the underlying ill-posedness (Corelli et al., 20 Oct 2025).

5. Entropy, holography, and matter localization

Black-hole thermodynamics in higher-curvature gravity departs sharply from the pure Wald picture once dynamical processes are considered. For arbitrary higher-curvature gravity with scalar and electromagnetic fields, the entropy satisfying the linearized second law is

$$P^{\mu\nu\rho\sigma}\equiv \partial \mathcal{L}/\partial R_{\mu\nu\rho\sigma}$$9

where $(a,b,c,e)$0 is the Wald entropy density, $(a,b,c,e)$1 is the higher-curvature correction already known from previous work, and $(a,b,c,e)$2 is a new correction sourced by the minimal coupling interaction between gravity and the scalar field. The electromagnetic field does not contribute to this entropy correction at linear order. A standard misconception is therefore that only non-minimal curvature couplings modify the entropy relevant for the second law; the cited result shows that minimally coupled scalars can also do so (Wang et al., 2022).

In holography, three-dimensional new massive gravity,

$(a,b,c,e)$3

provides an explicit AdS/BCFT example in which higher-curvature terms can be absorbed into effective couplings. The effective Newton constant is

$(a,b,c,e)$4

the central charge becomes

$(a,b,c,e)$5

and the boundary entropy and entanglement entropy retain the Einstein-gravity form after the substitution $(a,b,c,e)$6 or, equivalently, $(a,b,c,e)$7. The holographic $(a,b,c,e)$8-theorem continues to hold under the null energy condition (Kwon et al., 2012).

Matter localization in warped extra dimensions shows a different manifestation of higher-order curvature-scalar effects. In a five-dimensional $(a,b,c,e)$9 braneworld, the bulk geometry becomes $R^3$00-dependent rather than constant-curvature, massless left- and right-chiral fermion zero modes are driven toward the TeV brane as $R^3$01 increases, and massive KK modes move toward the Planck brane. The effective radion-fermion coupling for zero modes increases with $R^3$02, but the ratio $R^3$03 decreases, suppressing collider visibility; the same localization pattern persists in the scalar-tensor dual description (Mitra et al., 2017).

6. Structural issues, discrete extensions, and unresolved questions

A recurring theme is that higher-order curvature-scalar gravity is highly non-generic: consistency depends on carefully chosen combinations, symmetry restrictions, or background truncations. In general $R^3$04 theories, the massive spin-2 sector is ghost-like unless the coefficients satisfy the Einstein-like constraints, and only specially tuned combinations such as Einsteinian cubic gravity, its quartic extensions, $R^3$05, or the new cubic term $R^3$06 remove the ghost-like spin-2 mode at linear order (Bueno et al., 2016). On the cosmological side, second-order FLRW dynamics can be engineered systematically, but this property does not automatically extend to all perturbative or strong-field sectors (Moreno et al., 2023).

Discrete quantum-gravity work provides an independent route to higher-order invariants. In causal set theory, the Benincasa-Dowker operator

$R^3$07

has the continuum limit

$R^3$08

Iterating the operator yields

$R^3$09

and, more generally,

$R^3$10

so higher-order Ricci-scalar invariants emerge directly from the discrete causal structure (Brito et al., 2023).

The main controversies are therefore not about whether higher-curvature terms can be written down, but about when they define physically viable theories. Metric-affine theories generically remain pathological without projective symmetry or additional constraints (Jiménez et al., 2019). Unsuppressed cubic corrections can exclude black-hole solutions through ghost and Laplacian instabilities (Felice et al., 2023). In higher-curvature-dominated inflation, perturbation theory and the calculation of cosmological observables remain technically difficult and are explicitly identified as open questions (Edelstein et al., 2020). In combined $R^3$11-Gauss-Bonnet theories, second-order field equations are retained, but singularity resolution and hyperbolicity problems are not achieved by adding finitely many individual higher-order terms (Corelli et al., 20 Oct 2025).

Taken together, these results indicate that higher-order curvature-scalar gravity is not a single theory but a research program organized around a tension: higher-curvature operators and scalar sectors greatly enlarge the space of cosmological and black-hole solutions, yet only specially structured models avoid ghost modes, gradient instabilities, entropy-law violations, or strong-field ill-posedness. This suggests that the enduring problems in the subject are not the construction of new invariants, but the identification of nonperturbative sectors in which the enlarged dynamics remains mathematically controlled and observationally viable.