Gravity Beyond General Relativity

General relativity is extraordinarily successful, but its field equations are not the only generally covariant equations one can write, nor are the metric and its two tensor polarizations the only possible gravitational variables. This volume asks a more demanding question than “what terms can be added to Einstein’s equation?” For each proposed extension it identifies the fields and symmetries, counts the physical degrees of freedom, tests the initial-value problem and stability, states the cutoff and solution branch, recovers general relativity in a controlled regime, and only then derives compact-object, radiative, or cosmological predictions. The goal is not a catalogue of model names. It is a reusable method for deciding what a gravitational theory says, where it is predictive, and which parts of it can be compared.

Helpful background. Einstein Equations supplies the geometric field equation and Bianchi consistency used as the baseline; Gravitational Actions and Boundary Terms supplies the distinction between a bulk density and a complete variational problem; Formulations of General Relativity supplies the constraint, gauge, and initial-value language used to compare dynamical systems.

What counts as a gravitational theory?

Start with a familiar comparison: Newtonian gravity assigns a force to a mass distribution; GR makes the spacetime metric dynamical. An extension might add one scalar field. Its first test is then concrete: does the scalar carry a healthy wave, does a star source it, and does its force disappear in the regime where GR already works? A scalar can propagate without being excited by a particular black hole; conversely, a formal static profile can exist on an unstable background. The distinction between a field, a physical mode, and a source response will recur throughout the volume.

If this is your first encounter with modified gravity, use one running example before comparing many theory names: the Brans–Dicke action, the kinetic-sign test, and compact-body scalar charges. Together they show how an action becomes an equation, an extra degree of freedom, and a measurable response. Return to the more general notation below as a compact checklist for that calculation.

A useful specification is the tuple

$$\mathfrak T=\bigl(M,\Phi,\mathcal G,S,\mathcal B,\mathcal O\bigr).$$

Here $M$ includes the spacetime dimension and global structure; $\Phi$ is the complete set of dynamical fields; $\mathcal G$ contains gauge identifications and exact global symmetries; $S$ is the action with the boundary completion appropriate to the chosen data; $\mathcal B$ is the admissible initial, boundary, regularity, and branch data; and $\mathcal O$ specifies the physical observables and the metric to which rods, clocks, and matter couple. A local field redefinition can change the appearance of several entries without changing dimensionless observables. Changing the matter metric, discarding a boundary condition, crossing a singular transformation, or selecting a different global branch can instead change the physical theory.

This complete specification matters because a Lagrangian density alone does not answer basic dynamical questions. For a local covariant Lagrangian $D$-form $\mathbf L$,

$$\delta\mathbf L =\mathbf E_A\,\delta\Phi^A +d\boldsymbol\Theta(\Phi,\delta\Phi).$$

The bulk equations are $\mathbf E_A=0$, but the surface term $\boldsymbol\Theta$ determines whether the stated boundary data make the variation well posed and helps define charges. Gauge symmetry produces constraints and redundancies, so the number of components in $\Phi$ is not the number of propagating modes. Higher derivatives may describe a controlled effective correction, a constrained degenerate system, or an unwanted extra mode; those cases cannot be distinguished by counting derivatives in the written equation alone. The covariant phase-space identity and its boundary qualification are developed in Iyer and Wald 1994, §§2–3, pp. 3–8.

Pages use the site metric signature $(-,+,+,+)$, curvature convention, and $c=\hbar=k_{\mathrm B}=1$ with $G$ shown explicitly. Perturbations use $e^{-i\omega t}$ unless a source is translated as a complete convention package. Whenever it affects a conclusion, a page also names the physical matter metric, independent connection or spin connection, scalar kinetic convention, gauge, branch, Green function, boundary conditions, Planck-mass normalization, and cutoff. These choices are part of the claim, not notation to be filled in afterward.

The consistency ladder

A candidate model is carried through the same sequence of questions.

1. Definition. Which fields vary, which structures are fixed, what are the exact symmetries, how does matter couple, and which boundary data define the variational problem?
2. Physical modes. After all constraints and gauge identifications, how many degrees of freedom propagate on the stated background?
3. Local dynamics. Are the kinetic and gradient matrices healthy? Is the gauge-fixed system strongly hyperbolic with a propagated constraint subsystem? Are tachyonic timescales physically acceptable?
4. Predictive domain. Which dimensionless ratios organize the expansion, where does perturbation theory fail, and where must new operators or new degrees of freedom enter?
5. Branches and limits. Which regular solution branch is selected, and do its fields, sources, and observables approach general relativity uniformly in a domain that remains below the cutoff?
6. Predictions. What charges, multipoles, characteristic cones, fluxes, propagation laws, or cosmological response functions can be exported to an operational test?

Passing one rung does not imply the next. Removing the Boulware–Deser mode from a massive spin-2 theory, for example, does not establish stability or causal propagation on every background. Finding a regular black-hole solution does not establish its perturbative stability. Recovering the Einstein action as a parameter tends to zero does not prove that the corresponding sourced solutions or observables approach their GR values. This separation of definition, consistency, solution existence, and prediction is the main discipline of the volume.

Effective field theory supplies a second discipline. In four spacetime dimensions, schematically,

$$\mathcal L_{\mathrm{eff}} =\mathcal L_{\mathrm{light}} +\sum_i\frac{c_i}{\Lambda^{\Delta_i-4}}\mathcal O_i, \qquad \epsilon_i(E)\sim c_i\left(\frac{E}{\Lambda}\right)^{\Delta_i-4}.$$

The relevant expansion variables may involve locally measured frequency, curvature, compactness, background gradients, or a redressed kinetic scale; they are not always a single coordinate energy. Redundant operators are removed only with field redefinitions, sources, boundary terms, and retained orders transformed consistently. A pole above the cutoff of a truncated higher-derivative action is not automatically a resolved particle of that effective theory. The gravitational power counting and its domain are set out in Burgess 2004, §§3.1–3.3, pp. 23–35, while higher-order field-redefinition qualifications are developed in Criado and Pérez-Victoria 2019, §§2–5.

The volume in fourteen chapters

The chapter order moves from common machinery to theory families, then to the nonlinear and observationally portable predictions shared across families. For a comparative survey of scalar, vector, bimetric, higher-derivative, and extra-dimensional families, see Clifton et al. 2012, §§3–5 (PDF).

Chapter	Central question	Essential boundary
Theory Space and Effective Descriptions	What data define a theory, an equivalent description, and a controlled approximation?	A changed field basis is not a changed matter coupling; an EFT truncation is not a microscopic completion.
Degrees of Freedom and Dynamical Consistency	Which modes propagate, and do they define a stable, well-posed predictive system?	Constraint counting, kinetic signs, hyperbolicity, strong coupling, and radiative stability are distinct tests.
Scalar–Tensor Gravity	How can one or more scalars modify gravity while retaining controlled dynamics?	Frame equivalence requires invertibility and transformed matter; degeneracy removes a mode but does not prove full viability.
Curvature Corrections and Higher-Derivative Gravity	When are curvature operators effective corrections, new fundamental dynamics, or topological terms?	Constant-coupling Gauss–Bonnet is topological in four dimensions only with its dimensional and boundary qualifications.
Vector–Tensor and Preferred-Foliation Gravity	What changes when a vector, æther, or preferred foliation carries gravitational degrees of freedom?	Unit constraints, mode cones, projectability, and the chosen constraint algebra must be stated model by model.
Torsion, Nonmetricity, and Metric–Affine Gravity	How do tetrads and independent connections redistribute or enlarge gravitational dynamics?	The TEGR curvature–torsion boundary identity does not make generic $f(T)$, $f(Q)$, or metric–affine models equivalent to GR.
Massive Spin-2 and Multimetric Gravity	How can an interacting massive spin-2 field avoid its nonlinear sixth mode?	The special dRGT constraint addresses one ghost; branch stability, causality, coupling, and cutoff remain separate.
Extra Dimensions and Braneworld Gravity	How do higher-dimensional fields, geometry, and junction data appear to lower-dimensional observers?	A Kaluza–Klein ansatz must be a consistent truncation, and an effective brane equation is not closed without bulk data.
Nonlocal and Infrared Modifications	Can inverse differential operators modify long-distance gravity without introducing arbitrary new histories?	A localized auxiliary system is equivalent to a nonlocal model only after its Green function and homogeneous data are fixed.
Low-Acceleration and Modified-Dynamics Frameworks	What can an acceleration-dependent phenomenology determine about an underlying relativistic theory?	Modified gravity, modified inertia, and an effective force law are logically different constructions.
Screening and the Recovery of General Relativity	How do nonlinear environments suppress an extra force while preserving predictivity?	A screening radius depends on source, environment, branch, boundary data, and cutoff; it is not a universal property of a theory name.
Black Holes and Compact Objects Beyond GR	Which non-GR compact solutions exist, bifurcate, carry charges, and remain viable?	Existence, regularity, thermodynamics, perturbative stability, and astrophysical interpretation must be checked separately.
Binary Dynamics and Gravitational Radiation Beyond GR	How do extra fields alter conservative motion, radiation generation, polarizations, and propagation?	One common theory parameter must be mapped consistently across near zone, wave zone, and propagation without double counting.
Cosmology Beyond General Relativity	How do modified background branches and perturbations change growth, slip, lensing, and gravitational-wave travel?	Background expansion, stability, nonlinear screening, and observable response functions are not interchangeable tests.

The table also locates the scope boundary. This volume derives actions, constraints, solution branches, screening behavior, and theory-side predictions. Tests of Gravity develops instruments, source and environment systematics, likelihoods, mutable bounds, and empirical verdicts. Quantum Gravity studies microscopic candidate programmes and the stable low-energy quantum-gravity treatment. Relativistic Two-Body Dynamics, Gravitational Radiation, and Relativistic Cosmology supply the general dynamical methods to which the theory-specific results are handed. The Black Hole Atlas orients compact objects, Research carries dated result records, Computational Relativity carries numerical methods and reproducible execution practice, and Reference provides reusable definitions and conventions.

Four routes through the material

Build or compare a theory. Read the first two chapters in order, then the relevant family chapter. Complete the comparative theory record before using a background solution as evidence of viability. This route is designed for a student or researcher entering a new model class.

Follow strong-field predictions. Combine the scalar, curvature, vector, or massive-spin-2 chapter with screening, compact objects, and binary radiation. Track the same matter metric, branch, cutoff, and object-dependent charge through every step. This is the shortest route from an action to a portable compact-binary prediction.

Follow cosmological predictions. Begin with theory space and consistency, choose a family, then read screening and cosmology. Keep background evolution, linear perturbations, nonlinear structure, and gravitational-wave propagation as separate calculations connected by shared parameters rather than by a single phenomenological label.

Interrogate a mathematical claim. Follow field redefinitions, Hamiltonian constraints, hyperbolicity, Lovelock and Gauss–Bonnet structure, teleparallel equivalence, ghost-free massive gravity, Green-function prescriptions, and no-hair theorems. At every theorem, list the dimension, regularity, topology, boundary conditions, gauge or formulation, and coupling assumptions before using the conclusion elsewhere.

Check your preparation

You are given a four-dimensional action

$$S[g,\phi,\psi] =\int_M d^4x\sqrt{-g}\left[ \frac{R}{16\pi G}-\frac12(\nabla\phi)^2-V(\phi) \right] +S_m[A^2(\phi)g_{\mu\nu},\psi].$$

Explain why writing this integral is not yet a complete theory specification. List the additional data needed before one can count degrees of freedom, compare it with GR, or predict a compact-binary signal.

Solution

First declare which fields are varied and whether the connection is the Levi–Civita connection of $g_{\mu\nu}$. Specify the spacetime dimension, topology, regularity class, time orientation, gauge equivalence, and any global sectors. Complete the action with the boundary terms appropriate to the chosen initial, outer, and possible horizon data; otherwise the surface variation is not fixed. State the scalar potential and coupling function together with their parameter domain, and identify $\widetilde g_{\mu\nu}=A^2g_{\mu\nu}$ as the metric seen by matter, so that rods, clocks, stress tensors, and asymptotic masses are unambiguous.

Next choose a background and solution branch, give its center or horizon and asymptotic conditions, and state the EFT cutoff and every small parameter. Constraint counting then shows two tensor modes plus one scalar on a regular background with a nondegenerate scalar kinetic term; this count must be revisited at singular points or on branches where the kinetic matrix changes rank. A GR limit requires a path such as $A\to1$ together with a controlled scalar background and source coupling, not merely deletion of a term in the written action. A compact-binary prediction additionally needs the bodies’ equations of state or effective worldline data, scalar charges and sensitivities, conservative equations of motion, tensor and scalar fluxes, wave-zone boundary conditions, polarization normalization, and propagation law. Only after those theory-side quantities are derived can detector response, likelihoods, and empirical conclusions be handed to Tests of Gravity.

References

• Burgess, C. P. (2004), “Quantum Gravity in Everyday Life: General Relativity as an Effective Field Theory,” Living Reviews in Relativity 7, article 5. DOI; Open PDF.
• Clifton, Timothy, Pedro G. Ferreira, Antonio Padilla, and Constantinos Skordis (2012), “Modified Gravity and Cosmology,” Physics Reports 513, 1–189. DOI; Open PDF.
• Criado, Juan Carlos, and Manuel Pérez-Victoria (2019), “Field Redefinitions in Effective Theories at Higher Orders,” JHEP 03, 038. DOI; Open PDF.
• Iyer, Vivek, and Robert M. Wald (1994), “Some Properties of Noether Charge and a Proposal for Dynamical Black Hole Entropy,” Physical Review D 50, 846–864. DOI; Open PDF.

Further reading

• Heisenberg, Lavinia (2019), “A Systematic Approach to Generalisations of General Relativity and Their Cosmological Implications,” Physics Reports 796, 1–113. DOI; Open PDF.