Renormalized Gravitational Couplings

Quantum matter makes the gravitational action larger, not arbitrary. Ultraviolet divergences in the matter effective action are local diffeomorphism-invariant curvature integrals, so the bare gravitational action must contain the corresponding cosmological, Einstein, and curvature-squared terms. Their measured finite coefficients—not the regulator-dependent bare parameters—enter the semiclassical equation. Omitting an allowed counterterm generally makes the renormalized stress tensor or its metric variation inconsistent.

Required background. Einstein Equations supplies the curvature field equation, and Gravitational Actions and Boundary Terms supplies the variational problem and boundary data used below.

Helpful background. Quantum Fields in Curved Spacetime supplies the renormalized matter effective action, while Black-Hole Geometry and Mechanics supplies later stationary applications.

The local gravitational basis

In four dimensions a convenient bulk action through four derivatives is

$$S_{\mathrm{grav}}= \int d^4x\sqrt{-g}\left[ \frac{R-2\Lambda}{16\pi G} +\alpha R^2+\beta C_{\mu\nu\rho\sigma}C^{\mu\nu\rho\sigma} +\gamma E_4+\delta\Box R \right]+S_{\partial\mathcal M},$$

where $E_4=R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}-4R_{\mu\nu}R^{\mu\nu}+R^2$. In a boundaryless four-dimensional spacetime, the integrals of $E_4$ and $\Box R$ do not alter local bulk equations, but they matter for topology, boundaries, and dimensional continuation. The term $S_{\partial\mathcal M}$ must be chosen for the actual boundary conditions; the Einstein Gibbons–Hawking–York term alone does not make every higher-curvature variation well posed.

The coefficients have dimensions $[G]=L^2$, $[\Lambda]=L^{-2}$, and $[\alpha]=[\beta]=1$ in the site’s units. This dimension check already predicts which divergences can multiply which operators.

How matter divergences select the counterterms

For a Laplace-type matter operator, the regulated one-loop action has a short-distance expansion whose divergent four-dimensional part is schematically

$$W_{\mathrm{div}} =\frac{\mu^{n-4}}{n-4} \int d^nx\sqrt{-g}\left( a_0m^4+a_1m^2R+a_2^{(1)}R^2+a_2^{(2)}C^2+a_2^{(3)}E_4+a_2^{(4)}\Box R \right).$$

The numerical $a_i$ depend on spin, mass, and curvature coupling, but the local operator basis follows from covariance and dimensional analysis. Shapiro derives this structure from the heat kernel and effective action in Shapiro 2008, §§ 2–3.

For an operator of four-dimensional mass dimension $d_i$, write its coefficient in the $n$-dimensional action as

$$c_i^{\mathrm B}=\mu^{n-4}\left[c_i(\mu)+\frac{A_i}{n-4}\right], \qquad [c_i]=[A_i]=\mathrm{mass}^{4-d_i}.$$

Here $c_i$ denotes the coefficient of the operator, such as $-\Lambda/(8\pi G)$ or $1/(16\pi G)$, rather than $\Lambda$ or $G$ separately. The residues contain the needed powers of $m$ for lower-dimension operators. Choosing $A_i$ to cancel the corresponding pole leaves a finite $c_i(\mu)$. A change of scale moves finite local pieces between the matter action and $c_i(\mu)$, while a physical prediction remains scale independent to the retained order.

This is visible by splitting the regulated effective action as $W_{\mathrm{reg}}=W_{\mathrm{div}}+W_{\mathrm{fin}}$. The renormalized stress expectation follows from $-2(\sqrt{-g})^{-1}\delta W_{\mathrm{fin}}/\delta g^{\mu\nu}$ after the poles combine with the bare gravitational coefficients. Changing subtraction adds a finite $\Delta S_{\mathrm{local}}$ to $W_{\mathrm{fin}}$ and subtracts the same functional from the renormalized gravitational action. Their sum, its variation, and the stationary metric are unchanged. This cancellation is why the complete counterterm basis must be present before a state-dependent finite answer can be interpreted.

Metric variation and conservation

The curvature-squared terms produce conserved geometric tensors. For example, define

$$H^{(R^2)}_{\mu\nu} =\frac{1}{\sqrt{-g}}\frac{\delta}{\delta g^{\mu\nu}} \int d^4x\sqrt{-g}\,R^2.$$

With variation by $g^{\mu\nu}$,

$$H^{(R^2)}_{\mu\nu} =2RR_{\mu\nu}-\frac12g_{\mu\nu}R^2 +2g_{\mu\nu}\Box R-2\nabla_\mu\nabla_\nu R.$$

Diffeomorphism invariance gives $\nabla^\mu H^{(R^2)}_{\mu\nu}=0$ without using a matter equation. The analogous Weyl-squared variation is also conserved. Hu and Verdaguer display the renormalized semiclassical action and its two independent curvature-squared variations in Hu and Verdaguer 2008, § 3.1, Eqs. (3.7)–(3.10).

This conservation is an independent sign check. The algebraic terms have divergence $2R_{\mu\nu}\nabla^\mu R$, while

$$\Box\nabla_\nu R-\nabla_\nu\Box R =R_{\nu\mu}\nabla^\mu R$$

makes the derivative terms contribute $-2R_{\mu\nu}\nabla^\mu R$. They cancel. A sign error in either derivative term spoils the cancellation. The trace is another useful check: in four dimensions, $g^{\mu\nu}H^{(R^2)}_{\mu\nu}=6\Box R$.

Renormalization conditions and predictive content

Absorbing a divergence does not determine a finite coupling. One must specify renormalization conditions: for example, the measured long-distance $G$ and $\Lambda$, plus finite values of $\alpha$ and $\beta$ at a scale $\mu$. A semiclassical prediction then keeps these conditions fixed while changing the quantum state or geometry.

In flat space, all curvature-squared variations vanish and the equation tests only the renormalized cosmological term. On an Einstein space, several quadratic tensors become algebraically dependent, so that single background cannot determine every finite coefficient. Comparing less symmetric geometries or momentum-dependent response supplies the missing discrimination.

Power counting gives another check on truncation. A term such as $R^3/M^2$ is smaller than a curvature-squared term by roughly $(ML_{\mathrm{curv}})^{-2}$. If this ratio is not small, retaining only the displayed four-derivative basis is inconsistent even when every coefficient has been renormalized exactly. Renormalizability of the source and validity of a finite derivative expansion are distinct requirements.

The construction controls local counterterms. Nonlocal state-dependent terms in the effective action remain on the quantum side and cannot be fitted away by a finite polynomial. Fixed-background calculation of those terms belongs to Quantum Fields in Curved Spacetime; this chapter asks whether the resulting source and geometry solve one renormalized equation. Microscopic quantum-gravity operators beyond the declared cutoff remain outside the theory.

Exercise

Show that $H^{(R^2)}_{\mu\nu}$ vanishes on a four-dimensional constant-curvature geometry with $R_{\mu\nu}=Rg_{\mu\nu}/4$ and constant $R$. What does this imply about measuring $\alpha$ on that background?

Solution

Constant $R$ removes both derivative terms. The algebraic part becomes $2R(Rg_{\mu\nu}/4)-\frac12g_{\mu\nu}R^2=(R^2/2-R^2/2)g_{\mu\nu}=0$. Thus the $R^2$ coupling does not affect the bulk field equation on this exact background. A measurement restricted to that family cannot determine $\alpha$; one needs perturbations, a less symmetric geometry, boundary information, or another observable sensitive to its variation.

References

• Hu, B. L., and Verdaguer, E. 2008. “Stochastic Gravity: Theory and Applications.” Living Reviews in Relativity 11, 3. DOI.
• Shapiro, I. L. 2008. “Effective Action of Vacuum: Semiclassical Approach.” Classical and Quantum Gravity 25, 103001. Open PDF.