14.3 Modified Gravity Theories

Scalar-Tensor Theories

Fundamental Concepts of Scalar-Tensor Theories

The core idea behind scalar-tensor theories is straightforward: instead of treating gravity as purely geometric (as GR does), you introduce an additional scalar field that couples to the metric. This scalar field can evolve over cosmic time, and its dynamics can produce late-time acceleration without a cosmological constant.

• f(R) gravity modifies Einstein's field equations by replacing the Ricci scalar $R$ with a general function $f(R)$. This is actually equivalent to a specific class of scalar-tensor theories (you can show this through a conformal transformation).
• Brans-Dicke theory is the earliest and most studied scalar-tensor theory. It promotes Newton's constant $G$ to a dynamical field, so gravitational coupling strength can vary in space and time.
• More general scalar-tensor theories allow arbitrary coupling functions and scalar potentials, giving a wide landscape of models that can mimic $\Lambda$CDM expansion history while differing in their predictions for structure growth.

Mathematical Formulations and Key Features

The action for f(R) gravity is:

$S = \frac{1}{16\pi G} \int d^4x \sqrt{-g}\, f(R)$

When $f(R) = R - 2\Lambda$, you recover standard GR with a cosmological constant. Choosing other functional forms (e.g., $f(R) = R + \alpha R^2$) introduces new degrees of freedom.

The general scalar-tensor action takes the form:

$S = \int d^4x \sqrt{-g} \left[\frac{1}{2}f(\phi)R - \frac{1}{2}\omega(\phi)(\nabla\phi)^2 - V(\phi)\right]$

Here $f(\phi)$ controls the non-minimal coupling between the scalar field and curvature, $\omega(\phi)$ is the kinetic coupling, and $V(\phi)$ is the scalar potential.

For Brans-Dicke theory specifically, $f(\phi) = \phi$, $\omega(\phi) = \omega_{BD}/\phi$ (where $\omega_{BD}$ is a dimensionless constant), and $V(\phi) = 0$. GR is recovered in the limit $\omega_{BD} \to \infty$.

Key features shared across these theories:

• A time-varying effective gravitational constant $G_{\text{eff}}$, since the scalar field mediates an additional gravitational interaction
• Many models possess screening mechanisms (e.g., chameleon, symmetron, Vainshtein) that suppress deviations from GR in high-density environments, allowing them to pass local tests while still modifying gravity cosmologically
• In the Jordan frame, matter follows geodesics of the physical metric; in the Einstein frame (related by conformal transformation), the gravitational sector looks like GR plus a minimally coupled scalar. Physical predictions are frame-independent, but calculations are often easier in one frame or the other.

Observational Tests and Constraints

These theories are tightly constrained by multiple lines of evidence:

• Solar system tests: The Cassini spacecraft measured the Shapiro time delay with enough precision to require $\omega_{BD} > 40{,}000$, pushing Brans-Dicke theory very close to GR in the solar system. Screening mechanisms are what allow more general scalar-tensor models to evade this bound at cosmological scales.
• Gravitational waves: The near-simultaneous detection of GW170817 (gravitational waves) and GRB 170817A (gamma rays) from a binary neutron star merger constrained the speed of gravitational waves to match the speed of light to about one part in $10^{15}$. This single observation ruled out large classes of scalar-tensor and vector-tensor theories that predicted $c_{\text{gw}} \neq c$.
• Cosmological probes: CMB anisotropies, BAO measurements, and Type Ia supernovae constrain the expansion history. But the real discriminating power comes from growth-rate measurements (redshift-space distortions, weak lensing), because modified gravity models that match the same expansion history as $\Lambda$CDM typically predict different rates of structure formation.
• Some f(R) models (e.g., the Hu-Sawicki model) remain observationally viable and can produce self-accelerating solutions, though they do not eliminate the need for dark matter in galaxy clusters.

Modified Gravity Approaches

MOND and Its Foundations

MOND (Modified Newtonian Dynamics) takes a fundamentally different approach from scalar-tensor theories. Rather than modifying the action of GR, it proposes that Newtonian dynamics itself breaks down at very low accelerations.

Developed by Mordehai Milgrom in 1983, MOND introduces a characteristic acceleration scale:

$a_0 \approx 1.2 \times 10^{-10}\;\text{m/s}^2$

The modified force law replaces Newton's second law with:

$\mu\!\left(\frac{a}{a_0}\right) a = \frac{GM}{r^2}$

where $\mu(x)$ is an interpolating function satisfying $\mu(x) \to 1$ for $x \gg 1$ (Newtonian regime) and $\mu(x) \to x$ for $x \ll 1$ (deep-MOND regime). In the deep-MOND limit, the effective gravitational acceleration becomes $a = \sqrt{a_0\, g_N}$, where $g_N$ is the usual Newtonian value. This gives a flat rotation velocity $v^4 = G M a_0$, which is exactly the observed baryonic Tully-Fisher relation.

MOND's strengths and weaknesses:

• It successfully predicts flat galaxy rotation curves from the baryonic mass distribution alone, often with no free parameters beyond $a_0$
• It naturally explains the tight baryonic Tully-Fisher relation, which $\Lambda$CDM must treat as an emergent correlation
• It struggles with galaxy clusters, where MOND still requires some unseen mass (though less than GR needs)
• It has no satisfactory explanation for CMB acoustic peaks or large-scale structure formation without additional ingredients
• The Bullet Cluster, where the gravitational lensing center is offset from the baryonic mass, is difficult (though not impossible) to reconcile with MOND

TeVeS and Its Relativistic Extension

A major limitation of MOND as originally formulated is that it's non-relativistic, so it can't address gravitational lensing, cosmology, or gravitational waves. TeVeS (Tensor-Vector-Scalar gravity), developed by Jacob Bekenstein in 2004, was designed to fix this.

TeVeS introduces three dynamical fields:

1. Tensor field: the metric $g_{\mu\nu}$ (as in GR)
2. Vector field: a unit timelike vector $U_\alpha$ that defines a preferred frame
3. Scalar field: $\phi$, which generates the MOND-like behavior

Matter couples not to $g_{\mu\nu}$ directly but to a "physical metric" constructed from all three fields. In the non-relativistic, weak-field limit, TeVeS reduces to MOND, preserving its success with rotation curves.

However, TeVeS faces serious problems:

• It predicts gravitational lensing, but quantitative fits to cluster lensing data are poor without additional dark matter
• The GW170817 constraint on $c_{\text{gw}} = c$ is very difficult for TeVeS to satisfy in its original formulation, since the vector field generically causes gravitational waves to propagate at a different speed
• Cosmological structure formation in TeVeS does not match observed CMB and large-scale structure data well

These issues have led most of the community to view TeVeS as effectively ruled out in its original form, though modified versions continue to be explored.

DGP Model and Extra Dimensions

The DGP (Dvali-Gabadadze-Porrati) model takes yet another approach: gravity itself is fundamentally five-dimensional, but we live on a four-dimensional brane embedded in a 5D bulk spacetime.

The action combines 5D and 4D Einstein-Hilbert terms:

$S = M_5^3 \int d^5x \sqrt{-g^{(5)}}\, R^{(5)} + M_4^2 \int d^4x \sqrt{-g^{(4)}}\, R^{(4)}$

This setup introduces a crossover scale:

$r_c = \frac{M_4^2}{2 M_5^3}$

At distances $r \ll r_c$, gravity looks four-dimensional (recovering GR). At distances $r \gg r_c$, gravity "leaks" into the fifth dimension and weakens. For $r_c \sim H_0^{-1}$ (the current Hubble radius), this transition happens at cosmological scales.

The DGP model has two branches:

• Self-accelerating branch: Produces late-time acceleration without any dark energy component. This is the branch that generated the most excitement. However, it suffers from a ghost instability (a scalar mode with wrong-sign kinetic energy), which renders it theoretically problematic. It also predicts an expansion history that is now observationally disfavored.
• Normal branch: Does not self-accelerate, so it requires an additional dark energy component (like a cosmological constant) on the brane. This branch is ghost-free and remains viable, but it loses the original motivation of explaining acceleration purely through modified gravity.

Despite these issues, the DGP model has been influential. The Vainshtein screening mechanism, first understood in the DGP context, is now recognized as a general feature of many modified gravity theories. It explains how nonlinear effects can restore GR predictions near massive sources even when gravity is modified at large scales. The DGP framework also inspired the broader Galileon class of scalar field theories, which remain an active area of research.