Quadratic Curvature Gravity

Curvature-squared terms make the metric equations fourth order when they are treated as an exact classical theory. Around flat spacetime their propagator contains, besides the massless graviton, one massive scalar and one massive spin-2 pole. The scalar can have a healthy residue, whereas the massive spin-2 pole has the opposite residue to the graviton. That spectrum is a statement about the full higher-derivative theory. In a low-energy expansion whose cutoff lies below both masses, the same operators instead produce small corrections and their heavy poles must not be promoted to physical initial data.

Required background. Gravitational Actions and Boundary Terms supplies the action variation and boundary qualifications, while Curvature supplies the Weyl and Ricci decompositions used to choose an independent quadratic basis.

Helpful background. Einstein Equations supplies the linearized gauge constraints; Formulations of General Relativity supplies the physical-polarization count; Quantum Gravity provides the separate low-energy EFT interpretation of these operators.

The independent quadratic operators

Use metric signature $(-,+,+,+)$, the site’s curvature convention, and $\bar M_P^2=(8\pi G)^{-1}$. In four dimensions the identities

$$C_{abcd}C^{abcd} =R_{abcd}R^{abcd}-2R_{ab}R^{ab}+\frac13R^2$$

and

$$\mathcal G =R_{abcd}R^{abcd}-4R_{ab}R^{ab}+R^2$$

show that a constant multiple of one quadratic invariant can be traded for the others plus the Euler density $\mathcal G$. On a manifold and variational problem with the corresponding boundary completion, the local four-dimensional bulk spectrum can therefore be represented by

$$S=\int d^4x\sqrt{-g}\left[ \frac{\bar M_P^2}{2}R +\frac{R^2}{6f_0^2} -\frac{C_{abcd}C^{abcd}}{2f_2^2} \right].$$

Here $f_0$ and $f_2$ are dimensionless. This normalization is the $(-,+,+,+)$ translation of the quadratic basis summarized in Salvio 2018, § 2.1, equations (2.1)–(2.9). The signs of the Einstein and Weyl terms must be translated together when a source uses the opposite signature and curvature convention.

The scalar pole from R squared

The trace calculation can be recycled from metric $f(R)$ gravity. Factor the Einstein coefficient and write

$$\frac{\bar M_P^2}{2} \left(R+\frac{R^2}{3f_0^2\bar M_P^2}\right).$$

Thus the coefficient called $a$ in $R+aR^2$ is $a=(3f_0^2\bar M_P^2)^{-1}$. Linearizing the vacuum trace gives

$$(\Box-M_0^2)\,\delta R=0, \qquad M_0^2=\frac{f_0^2\bar M_P^2}{2}.$$

Equivalently, an auxiliary scalar converts the fourth-order scalar sector into a second-order scalar coupled to Einstein gravity. On the branch with $f_0^2\gt0$, its kinetic term and mass squared are positive. The auxiliary-field derivation is given in Salvio 2018, § 2.2, equations (2.13)–(2.20) and, in a general quadratic action, in Hindawi, Ovrut, and Waldram 1996, §§ III–IV, equations (3.1)–(4.16).

This result provides a normalization check: the limit $f_0\rightarrow\infty$ sends $M_0$ above every fixed low-energy scale and removes the $R^2$ response. Taking $f_0\rightarrow0$ does the opposite and is not an Einstein limit.

The spin-2 pole and its residue

The Weyl-squared term acts only on the spin-2 sector at quadratic order around $g_{ab}=\eta_{ab}+h_{ab}$. After gauge modes and constrained components are removed, the transverse-traceless kinetic operator factorizes into a massless and a massive factor. Define $z=-k^2$ using the site’s Fourier convention $e^{-i\omega t}$, so a physical massive pole occurs at $z=M^2$. Apart from an overall positive normalization, the spin-2 propagator contains

$$\frac{1}{z(1-z/M_2^2)}P^{(2)} =\left(\frac1z-\frac{1}{z-M_2^2}\right)P^{(2)}, \qquad M_2^2=\frac{f_2^2\bar M_P^2}{2}.$$

The two residues have opposite signs. Choosing the massless graviton to have positive norm therefore makes the massive spin-2 excitation a negative-residue mode in the exact local theory. The pole decomposition and complete $2+5+1$ polarization count are displayed in Salvio 2018, §§ 2.3 and 3.1, Table 1 and equations (2.21)–(2.40), (3.1)–(3.3).

This is more precise than saying “four derivatives imply a ghost.” Degenerate theories can evade the generic count, and a truncated EFT does not resolve a pole above its cutoff. But for the nondegenerate local action written above, the partial fraction makes the relative residue unavoidable.

Exact theory or effective expansion

There are two different questions.

Exact quadratic theory. If the action is assumed valid through and above $M_0$ and $M_2$, its solution space contains the massless graviton, five massive spin-2 polarizations, and one scalar. The negative spin-2 residue then requires an interpretation of stability, unitarity, or boundary conditions beyond the ordinary positive-norm particle picture.

Low-energy EFT. If curvature-squared terms are retained only as the first correction in an expansion with $E\ll\Lambda\lesssim M_0,M_2$, the massive roots are outside the validity domain. One expands amplitudes and solutions in $E^2/M_i^2$, supplies only the low-energy initial data, and uses field redefinitions or order reduction at fixed accuracy. Exciting the massive pole would demand frequencies at which omitted operators are equally important.

The distinction also controls the general-relativistic limit. Sending the quadratic coefficients to zero while holding momenta fixed and below the heavy masses recovers the Einstein propagator. Following a solution whose frequency scales like $M_i$ while taking that limit does not. Empirical likelihoods and numerical bounds are therefore not inferred from pole counting here; they belong to Tests of Gravity. A claimed microscopic cure of the residue problem belongs to Quantum Gravity.

Common pitfalls

Read a mass from the denominator without checking the residue. Pole location diagnoses dispersion and mass; the source-saturated residue diagnoses the kinetic sign. Both are required.

Count above-cutoff roots as EFT particles. A polynomial truncation knows about its own formal high-frequency roots but not whether they persist after the omitted tower of operators is restored.

Exercises

Factor the spin-2 denominator and determine both residues. Then use the linearized constraints to count the polarizations of the full quadratic theory and contrast them with the low-energy EFT count.

Solution

Writing the denominator as $z(1-z/M_2^2)$ gives $1/z-1/(z-M_2^2)$, so the massive and massless residues have opposite signs. A massless spin-2 field has two helicities, a massive spin-2 field has five, and the $R^2$ sector supplies one scalar, for eight modes in the exact theory. Below both heavy masses, perturbative boundary data exclude the massive roots; only the two graviton helicities remain as propagating EFT data, with their dynamics corrected order by order.

References

• Hindawi, Ahmed, Burt A. Ovrut, and Daniel Waldram. 1996. “Consistent Spin-Two Coupling and Quadratic Gravitation.” Physical Review D 53: 5583–5596. DOI.
• Salvio, Alberto. 2018. “Quadratic Gravity.” Frontiers in Physics 6: 77. DOI.