Einstein–Grisaru–Zanon Gravity Model

• Einstein–Grisaru–Zanon gravity is a fourth-order theory that incorporates leading (α')³ quartic-curvature corrections from superstring theory into general relativity.
• The model modifies the standard field equations with higher derivatives, allowing analysis of de Sitter dynamics, FLRW perturbations, and black-hole corrections.
• Dynamical-system analysis reveals a universal de Sitter instability, offering natural exit mechanisms in both inflationary and late-time cosmological contexts.

Searching arXiv for papers on Einstein–Grisaru–Zanon gravity and related stability/cosmology results. Search query: "Einstein-Grisaru-Zanon gravity" Einstein–Grisaru–Zanon (EGZ) gravity is a four-dimensional fourth-order gravity model defined by adding the leading $(\alpha')^3$ superstring correction to the Einstein–Hilbert action. In the recent literature it is treated as the low-energy effective gravitational sector of type II closed superstrings in four spacetime dimensions, with the correction encoded by a specific quartic-curvature invariant first found by Grisaru and Zanon. The model has been studied in perturbative black-hole backgrounds, in spatially flat FLRW cosmology, and in the context of Starobinsky inflation; it also admits an exact de Sitter solution whose generic instability has been established by a dynamical-system analysis (Delgado et al., 2024, Do, 13 Jun 2026, Toyama et al., 2024).

1. Action and string-theoretic origin

In four dimensions the EGZ action is written as

$$S_{\rm EGZ}[g]=\frac{M_{\rm Pl}^2}{2}\,\int d^4x\,\sqrt{-g}\;\Bigl[ R +\frac{\gamma}{M_{\rm Pl}^6}\,J \Bigr],$$

or equivalently as

$$S_{\rm EGZ}=\frac{M_p^2}{2}\,\int d^4x\;\sqrt{-g}\;\Bigl[R+\bar\gamma\,J\Bigr],\qquad \bar\gamma\equiv \gamma/M_p^6.$$

The quartic-curvature density is

$$J=\Bigl(R^{\mu\rho\sigma\nu}R_{\lambda\rho\sigma\tau}+\tfrac12\,R^{\mu\nu\rho\sigma}R_{\lambda\tau\rho\sigma}\Bigr)\,R_{\mu}{}^{\alpha\beta\lambda}\,R^{\tau}{}_{\alpha\beta\nu}.$$

In this formulation, $\gamma$ is a dimensionless effective coupling proportional to $(\alpha')^3$, and $J$ represents the first nontrivial $\alpha'$-correction from the four-loop $\beta$-function of the worldsheet $\sigma$-model (Delgado et al., 2024).

The same invariant is discussed in the inflationary literature under the label “Einstein–Grisaru–Zanon” or “Starobinsky–Grisaru–Zanon” gravity. In ten-dimensional string frame the tree-level effective action contains

$$S_{\rm EGZ}[g]=\frac{M_{\rm Pl}^2}{2}\,\int d^4x\,\sqrt{-g}\;\Bigl[ R +\frac{\gamma}{M_{\rm Pl}^6}\,J \Bigr],$$0

and compactification on a six-manifold yields a four-dimensional coefficient of the form

$$S_{\rm EGZ}[g]=\frac{M_{\rm Pl}^2}{2}\,\int d^4x\,\sqrt{-g}\;\Bigl[ R +\frac{\gamma}{M_{\rm Pl}^6}\,J \Bigr],$$1

up to numerical and volume-modulus factors. This places the four-dimensional coupling directly within the effective-field-theory expansion of closed superstring theory (Toyama et al., 2024).

2. Field equations and fourth-order structure

Variation of the EGZ action gives modified vacuum equations

$$S_{\rm EGZ}[g]=\frac{M_{\rm Pl}^2}{2}\,\int d^4x\,\sqrt{-g}\;\Bigl[ R +\frac{\gamma}{M_{\rm Pl}^6}\,J \Bigr],$$2

where $$S_{\rm EGZ}[g]=\frac{M_{\rm Pl}^2}{2}\,\int d^4x\,\sqrt{-g}\;\Bigl[ R +\frac{\gamma}{M_{\rm Pl}^6}\,J \Bigr],$$3 contains up to fourth derivatives of the metric. The fourth-order character is therefore intrinsic to the model rather than an artifact of a particular parametrization (Delgado et al., 2024).

For a spatially flat FLRW metric,

$$S_{\rm EGZ}[g]=\frac{M_{\rm Pl}^2}{2}\,\int d^4x\,\sqrt{-g}\;\Bigl[ R +\frac{\gamma}{M_{\rm Pl}^6}\,J \Bigr],$$4

the $$S_{\rm EGZ}[g]=\frac{M_{\rm Pl}^2}{2}\,\int d^4x\,\sqrt{-g}\;\Bigl[ R +\frac{\gamma}{M_{\rm Pl}^6}\,J \Bigr],$$5-component becomes

$$S_{\rm EGZ}[g]=\frac{M_{\rm Pl}^2}{2}\,\int d^4x\,\sqrt{-g}\;\Bigl[ R +\frac{\gamma}{M_{\rm Pl}^6}\,J \Bigr],$$6

while a second space-diagonal equation involves up to $$S_{\rm EGZ}[g]=\frac{M_{\rm Pl}^2}{2}\,\int d^4x\,\sqrt{-g}\;\Bigl[ R +\frac{\gamma}{M_{\rm Pl}^6}\,J \Bigr],$$7 (Delgado et al., 2024).

A separate derivation based on the reduced FLRW Lagrangian

$$S_{\rm EGZ}[g]=\frac{M_{\rm Pl}^2}{2}\,\int d^4x\,\sqrt{-g}\;\Bigl[ R +\frac{\gamma}{M_{\rm Pl}^6}\,J \Bigr],$$8

produces two Euler–Lagrange equations after setting $$S_{\rm EGZ}[g]=\frac{M_{\rm Pl}^2}{2}\,\int d^4x\,\sqrt{-g}\;\Bigl[ R +\frac{\gamma}{M_{\rm Pl}^6}\,J \Bigr],$$9: a Friedmann-type equation that is third order in $$S_{\rm EGZ}=\frac{M_p^2}{2}\,\int d^4x\;\sqrt{-g}\;\Bigl[R+\bar\gamma\,J\Bigr],\qquad \bar\gamma\equiv \gamma/M_p^6.$$0, and a scale-factor equation that is fourth order. In that analysis the Friedmann equation agrees with Eqs. (23–24) of Campos Delgado and Ketov, whereas the second equation exhibits a mismatch, described as a “gap,” in the coefficients of the highest-derivative terms. The gap is specifically associated with higher-order derivative pieces and does not alter the de Sitter solution because those terms vanish there (Do, 13 Jun 2026).

3. FLRW dynamics and the exact de Sitter branch

The model admits a constant-$$S_{\rm EGZ}=\frac{M_p^2}{2}\,\int d^4x\;\sqrt{-g}\;\Bigl[R+\bar\gamma\,J\Bigr],\qquad \bar\gamma\equiv \gamma/M_p^6.$$1 de Sitter solution. In the reduced-variable notation one sets

$$S_{\rm EGZ}=\frac{M_p^2}{2}\,\int d^4x\;\sqrt{-g}\;\Bigl[R+\bar\gamma\,J\Bigr],\qquad \bar\gamma\equiv \gamma/M_p^6.$$2

so that both field equations collapse to

$$S_{\rm EGZ}=\frac{M_p^2}{2}\,\int d^4x\;\sqrt{-g}\;\Bigl[R+\bar\gamma\,J\Bigr],\qquad \bar\gamma\equiv \gamma/M_p^6.$$3

The corresponding line element and Ricci scalar are

$$S_{\rm EGZ}=\frac{M_p^2}{2}\,\int d^4x\;\sqrt{-g}\;\Bigl[R+\bar\gamma\,J\Bigr],\qquad \bar\gamma\equiv \gamma/M_p^6.$$4

Because all time derivatives of $$S_{\rm EGZ}=\frac{M_p^2}{2}\,\int d^4x\;\sqrt{-g}\;\Bigl[R+\bar\gamma\,J\Bigr],\qquad \bar\gamma\equiv \gamma/M_p^6.$$5 beyond the first vanish, the “gap” terms in the alternative derivation drop out, and the de Sitter solution coincides with that obtained in the original EGZ paper (Do, 13 Jun 2026).

The original cosmological treatment also gives a local time-dependent expansion around $$S_{\rm EGZ}=\frac{M_p^2}{2}\,\int d^4x\;\sqrt{-g}\;\Bigl[R+\bar\gamma\,J\Bigr],\qquad \bar\gamma\equiv \gamma/M_p^6.$$6 and describes it as displaying a slow time-dependence induced by the $$S_{\rm EGZ}=\frac{M_p^2}{2}\,\int d^4x\;\sqrt{-g}\;\Bigl[R+\bar\gamma\,J\Bigr],\qquad \bar\gamma\equiv \gamma/M_p^6.$$7-term. This suggests that the exact de Sitter branch is not the only cosmological behavior encoded by the modified Friedmann system, even though the constant-$$S_{\rm EGZ}=\frac{M_p^2}{2}\,\int d^4x\;\sqrt{-g}\;\Bigl[R+\bar\gamma\,J\Bigr],\qquad \bar\gamma\equiv \gamma/M_p^6.$$8 solution is the analytically simplest one (Delgado et al., 2024).

In the inflationary context, the same quartic-curvature structure produces an extended “Starobinsky equation” with up to fourth time-derivatives of $$S_{\rm EGZ}=\frac{M_p^2}{2}\,\int d^4x\;\sqrt{-g}\;\Bigl[R+\bar\gamma\,J\Bigr],\qquad \bar\gamma\equiv \gamma/M_p^6.$$9. That formulation emphasizes the role of EGZ corrections as a controlled higher-curvature deformation rather than as an unrelated phenomenological modification (Toyama et al., 2024).

4. Dynamical-system analysis and instability of de Sitter

The stability problem can be reformulated as an autonomous dynamical system by introducing the e-fold time $$J=\Bigl(R^{\mu\rho\sigma\nu}R_{\lambda\rho\sigma\tau}+\tfrac12\,R^{\mu\nu\rho\sigma}R_{\lambda\tau\rho\sigma}\Bigr)\,R_{\mu}{}^{\alpha\beta\lambda}\,R^{\tau}{}_{\alpha\beta\nu}.$$0 and the dimensionless variables

$$J=\Bigl(R^{\mu\rho\sigma\nu}R_{\lambda\rho\sigma\tau}+\tfrac12\,R^{\mu\nu\rho\sigma}R_{\lambda\tau\rho\sigma}\Bigr)\,R_{\mu}{}^{\alpha\beta\lambda}\,R^{\tau}{}_{\alpha\beta\nu}.$$1

with

$$J=\Bigl(R^{\mu\rho\sigma\nu}R_{\lambda\rho\sigma\tau}+\tfrac12\,R^{\mu\nu\rho\sigma}R_{\lambda\tau\rho\sigma}\Bigr)\,R_{\mu}{}^{\alpha\beta\lambda}\,R^{\tau}{}_{\alpha\beta\nu}.$$2

Using the fourth-order field equation to solve for $$J=\Bigl(R^{\mu\rho\sigma\nu}R_{\lambda\rho\sigma\tau}+\tfrac12\,R^{\mu\nu\rho\sigma}R_{\lambda\tau\rho\sigma}\Bigr)\,R_{\mu}{}^{\alpha\beta\lambda}\,R^{\tau}{}_{\alpha\beta\nu}.$$3 and the Friedmann equation as a constraint yields a closed autonomous system in $$J=\Bigl(R^{\mu\rho\sigma\nu}R_{\lambda\rho\sigma\tau}+\tfrac12\,R^{\mu\nu\rho\sigma}R_{\lambda\tau\rho\sigma}\Bigr)\,R_{\mu}{}^{\alpha\beta\lambda}\,R^{\tau}{}_{\alpha\beta\nu}.$$4 (Do, 13 Jun 2026).

The isotropic fixed point corresponding to de Sitter is

$$J=\Bigl(R^{\mu\rho\sigma\nu}R_{\lambda\rho\sigma\tau}+\tfrac12\,R^{\mu\nu\rho\sigma}R_{\lambda\tau\rho\sigma}\Bigr)\,R_{\mu}{}^{\alpha\beta\lambda}\,R^{\tau}{}_{\alpha\beta\nu}.$$5

Linearization around this point gives the Jacobian

$$J=\Bigl(R^{\mu\rho\sigma\nu}R_{\lambda\rho\sigma\tau}+\tfrac12\,R^{\mu\nu\rho\sigma}R_{\lambda\tau\rho\sigma}\Bigr)\,R_{\mu}{}^{\alpha\beta\lambda}\,R^{\tau}{}_{\alpha\beta\nu}.$$6

with characteristic equation

$$J=\Bigl(R^{\mu\rho\sigma\nu}R_{\lambda\rho\sigma\tau}+\tfrac12\,R^{\mu\nu\rho\sigma}R_{\lambda\tau\rho\sigma}\Bigr)\,R_{\mu}{}^{\alpha\beta\lambda}\,R^{\tau}{}_{\alpha\beta\nu}.$$7

The eigenvalues are

$$J=\Bigl(R^{\mu\rho\sigma\nu}R_{\lambda\rho\sigma\tau}+\tfrac12\,R^{\mu\nu\rho\sigma}R_{\lambda\tau\rho\sigma}\Bigr)\,R_{\mu}{}^{\alpha\beta\lambda}\,R^{\tau}{}_{\alpha\beta\nu}.$$8

Because $$J=\Bigl(R^{\mu\rho\sigma\nu}R_{\lambda\rho\sigma\tau}+\tfrac12\,R^{\mu\nu\rho\sigma}R_{\lambda\tau\rho\sigma}\Bigr)\,R_{\mu}{}^{\alpha\beta\lambda}\,R^{\tau}{}_{\alpha\beta\nu}.$$9 is strictly positive, generic perturbations grow as $\gamma$0, so the de Sitter point is a repeller. The zero mode reflects residual time-shift symmetry, and the negative eigenvalue gives one decaying direction. A central result of the analysis is that none of the eigenvalues depends on the magnitude of $\gamma$1, so the instability is universal: it holds whether the de Sitter branch is interpreted as an inflationary phase or as a late-time accelerating phase (Do, 13 Jun 2026).

5. Relation to Einstein–Bel–Robinson gravity and Starobinsky inflation

A persistent point of comparison is Einstein–Bel–Robinson (EBR) gravity, defined by

$\gamma$2

where $\gamma$3. For Ricci-flat backgrounds one has

$\gamma$4

so the perturbative Schwarzschild corrections in EGZ and EBR coincide under $\gamma$5. In FLRW, however, the two theories yield genuinely different modified Friedmann equations (Delgado et al., 2024).

That distinction is explicit in flat FRW. The Grisaru–Zanon invariant reduces to

$\gamma$6

whereas the Bel–Robinson tensor squared becomes

$\gamma$7

At leading order in slow roll, both invariants agree in their $\gamma$8 and $\gamma$9 terms, but they differ at subleading orders in $(\alpha')^3$0 (Toyama et al., 2024).

In Starobinsky inflation, unitarity, ghost-freedom, and causality are analyzed through an effective function

$(\alpha')^3$1

together with the conditions

$(\alpha')^3$2

and

$(\alpha')^3$3

Using the maximal $(\alpha')^3$4 during Starobinsky inflation gives

$(\alpha')^3$5

Restoring $(\alpha')^3$6, the conclusion is that $(\alpha')^3$7 cannot exceed $(\alpha')^3$8–$(\alpha')^3$9, modulo volume factors. For the maximal $J$0, the quantum shifts are

$J$1

placing the quartic-curvature contribution at the level of the $J$2 classical corrections in the Starobinsky expansion (Toyama et al., 2024).

6. Black-hole sector, phenomenology, and physical implications

In the static, spherically symmetric sector one considers

$J$3

and expands around the Einstein–Schwarzschild solution to first order in $J$4. The resulting metric functions satisfy $J$5 at $J$6, unlike in pure GR, and this leads to post-Newtonian deviations (Delgado et al., 2024).

The photon-sphere radius and far-observer shadow radius are, to $J$7,

$J$8

and

$J$9

The analysis notes that deviations from $\alpha'$0 can in principle be constrained by EHT observations, although current uncertainties are too large for a meaningful bound on $\alpha'$1. The horizon radius and Hawking temperature also receive $\alpha'$2 corrections, and requiring the correction to remain subleading yields

$\alpha'$3

For $\alpha'$4 this becomes the robust bound

$\alpha'$5

(Delgado et al., 2024).

The physical interpretation of the unstable de Sitter branch depends on cosmological epoch. For late-time acceleration, the instability means that the EGZ de Sitter solution cannot serve as a late-time attractor on its own, so the model must be supplemented, for instance by a dark-energy fluid, to obtain a stable accelerating regime today. For the early universe, by contrast, the repeller property is presented as beneficial because it guarantees a natural exit from the de Sitter phase without fine-tuning, thereby avoiding eternal inflation and multiverse issues. More generally, the dynamical-system treatment shows that a fourth-order cosmological system can be reduced to the eigenvalue problem of a small Jacobian matrix, making stability analysis tractable despite the higher-derivative field equations (Do, 13 Jun 2026).