Curvature Corrections and Higher-Derivative Gravity

Curvature corrections can mean two very different things. They may define an exact higher-derivative classical theory with additional propagating modes, or they may be the first terms of a low-energy expansion whose extra formal roots lie beyond the cutoff. This chapter develops the calculations needed to tell those interpretations apart. It follows each operator from the action through its field equations, degrees of freedom, boundary conditions, corrected black-hole branches, and causal or ultraviolet consistency tests.

Helpful background. Einstein Equations supplies the GR seed and trace equations; Gravitational Actions and Boundary Terms supplies the variational boundary problem; Formulations of General Relativity supplies constraint, hyperbolicity, and boundary-value methods; and Curvature supplies the Riemann, Ricci, Weyl, and invariant decompositions used throughout.

Enter this chapter

Use metric signature $(-,+,+,+)$, explicit $G$, and the site’s curvature convention. A schematic local action is

$$S=\int d^4x\sqrt{-g}\left[ \frac{R}{16\pi G} +c_1R^2+c_2R_{ab}R^{ab}+c_3R_{abcd}R^{abcd}+\cdots \right].$$

The ellipsis is consequential. If the displayed polynomial is exact, its higher time derivatives normally enlarge the solution space. If instead it is truncated at order $(E/\Lambda)^n$, then operators represented by the ellipsis are only parametrically smaller while $E\ll\Lambda$; formal roots at $E\sim\Lambda$ are not controlled predictions.

Four questions organize the chapter.

1. Which fields and connections are varied, and which field redefinitions are invertible on the selected branch?
2. Which poles and constraints appear when the action is linearized about the stated background?
3. Which boundary, topology, dimension, and coupling hypotheses make a curvature density dynamical or nondynamical?
4. Which solution and observable remain controlled as the correction is removed or the cutoff is approached?

Power counting and the separation between low-energy prediction and microscopic completion are reviewed in Burgess 2004, §§ 2.3–2.7 and 3.1–3.3.

Route through the six pages

Read the pages in order because each changes one part of the interpretation.

1. f(R) Gravity and Its Scalar Representation derives the metric trace equation, performs the Legendre transformation, and shows why metric and Palatini variations do not propagate the same scalar. The equivalence is branch-local: $f''\ne0$, invertibility of $f'$, and a positive conformal factor are physical conditions.
2. Quadratic Curvature Gravity reduces the four-dimensional quadratic basis and factorizes the flat-background propagator. It identifies the massless graviton, a massive scalar, and the opposite-residue massive spin-2 pole, then distinguishes exact-theory and EFT counts.
3. Gauss–Bonnet and Topological Curvature Terms proves the critical-dimensional result rather than quoting a slogan. A constant coefficient on a smooth closed four-manifold gives an Euler invariant, while boundaries, $D\gt4$, and a field-dependent coupling evade a stated hypothesis.
4. Order Reduction in Gravitational EFT excludes high-frequency solutions without a regular low-energy expansion while retaining the correction through the declared order. Its oscillator calculation makes the initial-data restriction explicit.
5. Higher-Curvature Black-Hole Branches turns branch selection into a reproducible boundary-value problem. Its pinned executable benchmark follows scalar–Gauss–Bonnet hair from the GR point through a coupling fold and reports residual, charge, boundary, and independent weak-coupling checks.
6. Causality, Positivity, and UV Consistency compares eikonal time delays, heavy-state spectral positivity, and strong hyperbolicity at the 2026-08-19 evidence cutoff. The result is a set of conditional implications, not a dimension-independent sign slogan.

One operator, several physical questions

The same curvature term can occupy different rows of the following conceptual comparison.

Question	Exact local theory	Truncated EFT
Extra pole	A pole inside the assumed validity range belongs to the formal spectrum and its residue must be interpreted.	A pole at or above the cutoff is not admissible low-energy initial data.
Field redefinition	An invertible map preserves the exact solution space only with matter and boundary data transformed.	Operators proportional to lower-order equations can be exchanged through the retained order.
GR limit	A branch must approach the GR fields, charges, horizon, and boundary conditions.	Observables must converge at fixed $E/\Lambda\to0$ without following an above-cutoff root.
Numerical solution	Every resolved branch of the exact equations is a candidate, subject to stability and regularity.	Only the perturbative, cutoff-respecting branch is retained.
UV statement	The action claims validity through its additional modes.	Causality or positivity can instead diagnose the scale where new states or operators must enter.

This comparison prevents three common category errors. A healthy scalar representation of $R+R^2$ does not cure the massive spin-2 residue produced by $C^2$. A vanishing four-dimensional bulk variation of constant-coupling Gauss–Bonnet does not make $F(\phi)\mathcal G$ topological. And an order-reduced solution does not prove that the corresponding exact fourth-order theory is fundamental.

The Lovelock dimensional argument and boundary structure are developed in Padmanabhan and Kothawala 2013, §§ 2.2, 2.5, and 2.6, eqs. (12)–(68).

Checks to carry from page to page

For an analytic calculation, verify dimensions, relative signs, Bianchi or constraint consistency, a controlled GR limit, and a limit where the claimed equivalence fails. For a proof, list smoothness, dimension, orientation, boundary, and coupling hypotheses before the conclusion. For a computational branch, refine at least three times, distinguish an update norm from the field-equation residual, extract horizon and asymptotic data independently, and state how a turn is crossed. For a frontier claim, record the background, operator basis, cutoff, matter coupling, exact source locator, and evidence date.

No single check replaces the others. A positive flat-space residue does not establish strong hyperbolicity on a curved background. A converged boundary value does not prove stability. A spectral sign under a gapped weakly coupled completion does not remove the massless-graviton forward pole in another dimension. Conversely, failure of one formulation or extrapolation need not be a no-go theorem for every completion.

The chapter stops before empirical likelihoods and mutable numerical bounds; those belong to Tests of Gravity. It also stops before promoting the curvature expansion to a microscopic theory; candidate completions and stable low-energy quantum-gravity EFT belong to Quantum Gravity.

Review the chapter

Classify each modification before calculating: (a) constant-coupling $\alpha\mathcal G$ in four dimensions on a closed manifold, (b) $F(\phi)\mathcal G$ with a dynamical scalar, (c) an exact $R^2+C_{abcd}C^{abcd}$ theory, and (d) the same quadratic operators retained only through first order below their heavy masses. State the local modes or field equations that should be kept in each case.

Solution

Case (a) changes the Euler invariant but has no local four-dimensional bulk metric equation under the stated closed-manifold and constant-coefficient hypotheses. Case (b) is dynamical because derivatives act on $F$ and the scalar equation contains $F_{,\phi}\mathcal G$. Case (c) contains the massless graviton, one massive scalar, and an opposite-residue massive spin-2 mode in the generic nondegenerate theory. Case (d) retains only the low-energy graviton data and perturbative corrections; its above-cutoff roots are excluded or order-reduced rather than counted as EFT particles.

References

• Burgess, C. P. 2004. “Quantum Gravity in Everyday Life: General Relativity as an Effective Field Theory.” Living Reviews in Relativity 7: 5. DOI.
• Padmanabhan, T., and Dawood Kothawala. 2013. “Lanczos–Lovelock Models of Gravity.” Physics Reports 531: 115–171. DOI.