Generalized Starobinsky Potentials

• Generalized Starobinsky potentials are inflationary scalar potentials that extend the classic R+R² model through adjustable deformation parameters, preserving a Starobinsky limit.
• They encompass various formulations such as R^(2p), power-law R^β, α-deformed, supergravity-inspired, and cubic curvature models to modulate the inflationary plateau.
• Observational constraints from Planck, BAO, and related data typically narrow the model space, often steering deformations back toward the robust standard Starobinsky limit.

Generalized Starobinsky potentials are inflationary scalar potentials that deform the Einstein-frame scalar dual of the $R+R^2$ model while retaining the original Starobinsky form as a special limit. The undeformed potential is

$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$

and the literature uses “generalized Starobinsky” for several non-equivalent but closely related constructions: higher-curvature models with $R^{2p}$ or $R^\beta$, $\alpha$-deformed plateau models, supergravity realizations with modified Kähler or superpotential sectors, cubic $R^3$ corrections, and dynamical scenarios in which the coefficient of $R^2$ is itself field-dependent. In each case, the deformation parameter is chosen so that a distinguished limit reproduces the standard plateau model (Cedeño et al., 2023, Zambrano et al., 11 May 2026, Diamandis et al., 2014, Gialamas et al., 6 May 2025, Chaichian et al., 2022).

1. Canonical form and model classes

The common organizing principle is that the Starobinsky plateau is preserved only in a specific limit of a broader parameterized family. Different papers use different deformation parameters and different Jordan-frame starting points, so “generalized Starobinsky potential” is a family resemblance rather than a single universally normalized formula.

Class	Representative deformation	Starobinsky limit
$R^{2p}$ model	action contains $R^{2p}$	$p=1$
Power-law $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$0 model	$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$1 correction to Einstein gravity	$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$2
$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$3-Starobinsky potential	$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$4	$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$5
$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$6-Starobinsky model	$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$7 replaces $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$8 in the exponential	$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$9
Supergravity deformation	extra $R^{2p}$0 superpotential term or modified no-scale superpotential	deformation parameter $R^{2p}$1
Cubic curvature correction	$R^{2p}$2	$R^{2p}$3

In the $R^{2p}$4 formulation, the power $R^{2p}$5 controls the flatness and asymptotics of the inflationary plateau, and $R^{2p}$6 gives the original Starobinsky potential (Cedeño et al., 2023). In the power-law $R^{2p}$7 formulation, $R^{2p}$8 is the corresponding Starobinsky limit, while the $R^{2p}$9-Starobinsky potential motivated by brane inflation instead uses $R^\beta$0 as the limit that reproduces the standard exponential plateau (Chakravarty et al., 2014, Costa et al., 2020). The $R^\beta$1-Starobinsky and power-law $R^\beta$2-Starobinsky models make the exponential slope shallower or steeper through $R^\beta$3, with $R^\beta$4 restoring the original model (Zambrano et al., 11 May 2026). Supergravity constructions introduce additional parameters through Kähler potentials, superpotentials, or nilpotent sectors, and cubic curvature models add a first-order deformation to the plateau through an $R^\beta$5 term (Diamandis et al., 2014, Aldabergenov, 2020, Gialamas et al., 6 May 2025).

2. Functional deformations of the plateau

A central branch of the subject starts from generalized $R^\beta$6 actions. For the $R^\beta$7 model, the Jordan-frame action contains a generalized higher-order curvature correction of the form $R^\beta$8, and after a conformal transformation the Einstein-frame potential depends on the exponent $R^\beta$9. The inferred value of $\alpha$0 is statistically consistent with $\alpha$1, and the model is designed precisely to test how little the power can deviate from the $\alpha$2 case before the plateau prediction changes (Cedeño et al., 2023).

The power-law Starobinsky model replaces the quadratic curvature term with an $\alpha$3 deformation. In the Einstein frame this yields a more general exponential potential, and small departures from $\alpha$4 steepen the large-field region. An early analysis emphasized that this steepening can dramatically enhance the tensor signal: for $\alpha$5 the model gives $\alpha$6 for $\alpha$7, and for $\alpha$8 it gives $\alpha$9 for $R^3$0 if running is allowed (Chakravarty et al., 2014). This does not imply that such values remain observationally preferred; it shows that the power-law parameterization is a highly sensitive deformation of the plateau.

A separate deformation, motivated through brane inflation, is the $R^3$1-Starobinsky potential

$R^3$2

Here $R^3$3 recovers the original Starobinsky potential through the limit $R^3$4. The model admits a wider range of solutions for $R^3$5, but current data constrain $R^3$6 to remain close to zero (Costa et al., 2020).

Recent numerical work has also studied the $R^3$7-Starobinsky, power-law Starobinsky, and power-law $R^3$8-Starobinsky potentials. The $R^3$9-Starobinsky model uses

$R^2$0

the power-law Starobinsky model uses an Einstein-frame potential generated by an $R^2$1 deformation, and the combined power-law $R^2$2-Starobinsky model merges both parameters. In all three cases, the standard Starobinsky limit is recovered for special parameter values, and the combined model is the most general scenario considered in that study (Zambrano et al., 11 May 2026).

3. Supergravity, superconformal, and dynamical realizations

Generalized Starobinsky potentials are deeply tied to supergravity. In the superconformal construction, the Starobinsky model can be represented as a conformally invariant theory with spontaneous symmetry breaking, and the supergravity generalization is obtained from three chiral multiplets. A broad class of generalized potentials follows from choosing

$R^2$3

with $R^2$4 a real holomorphic function, so that along the inflationary trajectory one obtains potentials of the form $R^2$5 (Kallosh et al., 2013). In old-minimal supergravity, the Starobinsky model is dual to a no-scale model with an $R^2$6-term potential; in new-minimal supergravity, it is equivalent to standard supergravity coupled to a massive vector multiplet with a $R^2$7-term potential. Both formulations admit higher-order corrections that threaten the flatness of the inflaton potential (Farakos et al., 2013). A more general old-minimal framework uses a single holomorphic potential $R^2$8 and a single non-holomorphic potential $R^2$9, dualizable into standard matter-coupled supergravity with two chiral superfields (Ketov, 2013).

A particularly explicit deformation arises in $R^{2p}$0 supergravity with non-minimal superpotentials. Starting from two chiral multiplets $R^{2p}$1 and $R^{2p}$2, the superpotential is generalized to

$R^{2p}$3

The Starobinsky case corresponds to $R^{2p}$4, while $R^{2p}$5 introduces a quadratic dependence in $R^{2p}$6. After canonical normalization, the resulting scalar potential is the Starobinsky potential multiplied by a factor that is exponential in the inflaton field and dominates for large inflaton values. The paper states that the standard Starobinsky potential is recovered only when $R^{2p}$7, and that sufficient inflation requires $R^{2p}$8. Even then, the tensor-to-scalar ratio can only rise slightly above the Starobinsky value, up to about $R^{2p}$9 (Diamandis et al., 2014).

Nilpotent-superfield realizations produce another important family. For

$R^{2p}$0

with $R^{2p}$1, the models accommodate Starobinsky-like inflation with

$R^{2p}$2

For $R^{2p}$3, viable hilltop inflation is possible, with $R^{2p}$4 and $R^{2p}$5 close to the same expressions (Aldabergenov, 2020). A different dynamical mechanism promotes the $R^{2p}$6 coefficient to a function $R^{2p}$7 of a shift-symmetric scalar, giving an Einstein-frame potential

$R^{2p}$8

When the scalar condenses at $R^{2p}$9 and $p=1$0, the model reduces to the classic Starobinsky potential; the absence of ghost modes imposes the conditions

$p=1$1

(Chaichian et al., 2022). The cubic-curvature model

$p=1$2

has an Einstein-frame potential whose first-order expansion is identical to the potential obtained by modifying the superpotential in no-scale supergravity, with the identification $p=1$3 (Gialamas et al., 6 May 2025).

4. Higher-order corrections, steepening, and alternative inflationary regimes

A recurring result is that generalized Starobinsky potentials often preserve the plateau only approximately. In the $p=1$4 extension, the first-order potential is

$p=1$5

For $p=1$6, the potential acquires a rising tail at large $p=1$7; for $p=1$8, the plateau develops runaway behavior. The corresponding slow-roll expressions,

$p=1$9

show that negative $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$00 raises both $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$01 and $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$02 relative to the standard Starobinsky model (Gialamas et al., 6 May 2025).

The same theme appears in effective higher-order corrections to the Starobinsky potential. When the Jordan-frame potential is corrected by two real coefficients $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$03, the Einstein-frame potential can exhibit the Starobinsky plateau, a steep slope, and possibly an additional minimum, local maximum, or saddle point. Three types of inflationary behavior are identified: inflation on the plateau, at the local maximum (topological inflation), and at the saddle point. For $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$04 and $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$05, the potential may contain a second minimum away from the GR vacuum, and the paper states that this minimum is stable against both quantum tunnelling and thermal corrections (Artymowski et al., 2015).

The reinterpretation based on compactification of extra dimensions gives a different rationale for higher-order corrections. Starting from a $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$06-dimensional action with order-one coefficients $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$07, compactification yields a four-dimensional action with a large overall coefficient $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$08, and the paper argues that $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$09 can naturally explain the large coefficient of $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$10. In this picture only the coefficient of the linear $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$11 term, $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$12, must be tuned small, while higher-order terms need not be separately suppressed. The corresponding leading deformation changes $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$13, $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$14, and the runnings in a calculable way, and the quantum gravity scale is estimated as

$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$15

(Asaka et al., 2015).

These results support a general conclusion: generalized Starobinsky potentials are not merely nearby reparameterizations of a robust plateau. Several constructions produce exponential steepening at large field, shorten the inflationary plateau, or create extra extrema, so successful inflation often requires a restricted parameter region and sometimes explicit fine-tuning (Diamandis et al., 2014, Farakos et al., 2013, Artymowski et al., 2015).

5. Observational constraints and numerical analyses

Current observational analyses overwhelmingly constrain deformations toward the Starobinsky limit, though the precise preferred point depends on the parameterization and dataset. In the $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$16 model, a Bayesian analysis using Planck 2018 TT, TE, EE+lowE+lensing and BAO data from BOSS, 6dFGS, and SDSS DR7, together with reheating information, finds

$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$17

with

$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$18

and $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$19. The paper emphasizes that reheating constraints tighten the allowable range of $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$20 relative to the case with no reheating (Cedeño et al., 2023).

For the power-law Starobinsky model constrained with Planck-2018, BICEP3, and BAO data, the MCMC analysis gives

$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$21

with derived values

$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$22

This result permits slight deviations from the $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$23 model while keeping $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$24 close to 2 (Saini et al., 6 Feb 2025).

Direct numerical evolution of the generalized $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$25 potential also leads to near-Starobinsky best fits. One study reports that $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$26 reproduces

$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$27

and that models remain viable for $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$28 (Meza et al., 2021). A related semiclassical analysis solved the perturbation equations with the improved uniform approximation and the phase-integral method up to third-order in deviation, finding that the third-order phase-integral method reproduces the numerical result for $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$29 with relative errors of $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$30 in $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$31 and $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$32 in $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$33, while keeping the prediction inside the $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$34 confidence-level region in the $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$35 plane (Rojas, 2022).

Comparative numerical work on the $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$36-Starobinsky, power-law Starobinsky, and power-law $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$37-Starobinsky models finds that, for certain choices of parameters, the $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$38-Starobinsky model and the power-law $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$39-Starobinsky model are favored by Planck 2018 observations. By contrast, the power-law Starobinsky model is favored only when $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$40 is very close to 2 (Zambrano et al., 11 May 2026). In the brane-motivated $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$41-Starobinsky model, current CMB and BAO data yield

$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$42

fully compatible with zero, and the Bayesian Information Criterion gives $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$43, positively preferring the minimal Starobinsky model (Costa et al., 2020).

A dataset-dependent contrast appears in the $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$44 extension. Using ACT+DESI+Planck, the allowed range for the cubic parameter is

$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$45

for $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$46–$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$47, and the paper states that the standard Starobinsky model $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$48 is outside the $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$49 preferred region of the latest data (Gialamas et al., 6 May 2025). This suggests that the status of the undeformed plateau can depend on whether one analyzes reheating-constrained Planck+BAO datasets, Planck+B-mode+BAO combinations, or ACT-centered combinations.

6. UV completion, string realizations, and quantum-cosmology extensions

Attempts to embed generalized Starobinsky potentials in string theory expose a structural tension between plateau shape and matter couplings. A detailed Type IIB analysis studies the volume modulus, bulk fibre moduli, and blow-up modes. The volume modulus has the correct universal Yukawa/conformal coupling $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$50 to matter fermions but no plateau at large field values; fibre moduli have a potential very similar to Starobinsky inflation and a natural suppression of higher-curvature corrections through a term $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$51, but they do not reproduce the required matter coupling; blow-up modes have both the wrong potential and the wrong coupling. The paper concludes that embedding Starobinsky inflation into string theory seems rather hard (Brinkmann et al., 2023).

A different string-theoretic route uses axion monodromy in non-geometric flux compactifications. There the backreacted, uplifted F-term axion-monodromy potential interpolates between quadratic and Starobinsky-like form. In the large-field regime, after canonical normalization, the backreacted potential takes the Starobinsky-like form

$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$52

while in the small-field regime it is quadratic. The same construction highlights a tension between single-field inflation and a controlled UV approximation, expressed through the scale hierarchy

$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$53

(Blumenhagen et al., 2015).

Generalized Starobinsky potentials also enter loop quantum cosmology. For the Starobinsky potential

$$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$54

the hybrid and dressed metric formalisms produce different background-dependent masses for perturbation modes at the bounce. In the hybrid formalism the scalar mass is positive for $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$55, whereas in the dressed metric formalism it is negative for $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$56 when $$V_{\rm St}(\phi)=\frac{3}{4}M^2\left(1-e^{-\sqrt{\frac{2}{3}}\phi}\right)^2,$$57. The paper further states that similar sign properties extend to exponential and hyperbolic cosine generalizations, affecting vacuum selection and the well-posedness of the mode problem near the bounce (Iteanu et al., 2022).

Taken together, these developments show that generalized Starobinsky potentials serve as a precise diagnostic of how inflationary plateaus respond to curvature corrections, supergravity data, reheating assumptions, UV completion attempts, and quantum-gravity modifications. The literature consistently finds that viable deformations exist, but it also consistently finds that the original plateau remains a highly restrictive attractor: observationally, many parameterizations are driven back toward the Starobinsky limit, while theoretically, uncontrolled corrections often steepen the potential or spoil the coupling structure required for a complete embedding.