Asymptotic Safety

Perturbative quantum gravity is non-renormalizable: Newton’s constant $G$ has mass dimension $[G] = -2$ in $d = 4$, leading to divergences that require infinitely many counterterms. Weinberg (1979) proposed a way out: if the renormalization group flow possesses an ultraviolet fixed point with a finite number of relevant directions, the theory is “asymptotically safe” and predictive despite being non-renormalizable in the perturbative sense.

Introduce dimensionless couplings $\tilde{g}_i = g_i\,k^{d_i}$ where $k$ is the RG scale and $d_i$ is the mass dimension of $g_i$. The RG flow equations are:

A UV fixed point $\tilde{g}_i^*$ satisfies $\beta_i(\tilde{g}^*) = 0$. The theory is asymptotically safe if: (i) such a fixed point exists, and (ii) its UV critical surface (the set of trajectories attracted to $\tilde{g}^*$ as $k \to \infty$) is finite-dimensional.

The main tool for investigating asymptotic safety is the Wetterich equation for the effective average action $\Gamma_k$. This functional interpolates between the bare action ($k \to \infty$) and the full quantum effective action ($k \to 0$). The exact flow equation is:

where $\Gamma_k^{(2)}$ is the second functional derivative (the inverse propagator) and $\mathcal{R}_k$ is an IR regulator that suppresses modes with momenta $p^2 < k^2$. This is a functional differential equation: exact but requires truncation for practical computations.

For gravity, one writes $\Gamma_k[g_{\mu\nu}]$ and must handle the gauge and diffeomorphism invariance carefully using the background field method with metric decomposition $g_{\mu\nu} = \bar{g}_{\mu\nu} + h_{\mu\nu}$.

The simplest truncation retains only the cosmological constant $\Lambda$ and Newton’s constant $G$:

Defining dimensionless couplings $\tilde{g} = G_k\,k^2$ and $\tilde{\lambda} = \Lambda_k / k^2$, Reuter (1998) found the beta functions:

where the anomalous dimension of Newton’s constant is $\eta_N = -2\tilde{g}/(3\pi(1 - 2\tilde{\lambda}))$. These equations admit a non-Gaussian UV fixed point at:

The nature of the fixed point is determined by the stability matrix $B_{ij} = \partial\beta_i/\partial\tilde{g}_j\big|_{\tilde{g}^*}$. Its eigenvalues $-\theta_I$ (with the conventional sign) determine the critical exponents. For the Einstein-Hilbert truncation:

The positive real part means both directions are UV-attractive (relevant). The imaginary part produces a spiraling approach to the fixed point. Near the fixed point, the deviations scale as:

The UV critical surface is 2-dimensional in the Einstein-Hilbert truncation, meaning two free parameters ($G$ and $\Lambda$) suffice for UV completion, exactly as in the classical theory.

The crucial question is whether the fixed point survives in more complete truncations. Including higher-derivative terms $R^2, R_{\mu\nu}R^{\mu\nu}, C_{\mu\nu\rho\sigma}C^{\mu\nu\rho\sigma}, \ldots$, the effective average action becomes:

Studies up to $f(R)$ truncations with polynomials of order $N \sim 70$ consistently find: (i) the UV fixed point persists, (ii) the number of relevant directions remains finite (typically 3), and (iii) the critical exponents show convergence. The dimensionality of the UV critical surface appears to be $\dim(\mathcal{S}_{\rm UV}) = 3$, corresponding to $G, \Lambda$, and one higher-derivative coupling.

If asymptotic safety is realized, Newton’s constant runs with energy. At high energies $E \gg M_P$, the dimensionful coupling scales as:

This implies gravity becomes weaker at short distances, potentially resolving the singularities of classical general relativity. For black holes, the running $G(r)$ modifies the Schwarzschild metric near $r = 0$, replacing the singularity with a regular de Sitter core.

We integrate the beta functions of the Einstein-Hilbert truncation, visualize the RG flow in the $(\tilde{\lambda}, \tilde{g})$ plane showing the UV fixed point, the running of Newton’s constant, the anomalous dimension, and the spiraling approach governed by complex critical exponents: