Perturbative General Relativity Methods

• Perturbative general relativity is a framework that expands Einstein’s equations around a fixed background to extract gravitational dynamics while handling gauge freedom.
• It employs diverse methods such as metric, post-Newtonian, and amplitude-based approaches tailored to specific observables and regimes.
• Its formulations uncover nonlinear resummation schemes essential for analyzing compact binaries, cosmological perturbations, and modified gravity scenarios.

Searching arXiv for the provided core and supporting papers on perturbative general relativity. Search query: (Adamo, 2015) perturbative general relativity two-dimensional CFT, (Nakamura, 2010) higher-order gauge-invariant perturbation theory, (Bjerrum-Bohr et al., 2018) amplitudes classical general relativity, (Krasnov et al., 2020) chiral perturbation theory GR, (Cristofoli et al., 2019) post-Minkowskian Hamiltonians general relativity Perturbative general relativity is the collection of methods that study Einstein gravity by expanding either the metric, the equations of motion, or observables around a chosen background spacetime or around a chosen small parameter such as $G_N$, $v/c$, wavelength, or a deformation parameter. Across these formulations, the central objective is to extract physically meaningful gravitational dynamics while controlling gauge freedom, nonlinear self-interaction, and the relation between local field variables and invariant observables. The subject ranges from higher-order gauge-invariant metric perturbation theory on generic backgrounds (Nakamura, 2010), to post-Newtonian and post-Minkowskian treatments of the two-body problem (Bjerrum-Bohr et al., 2018, Cristofoli et al., 2019, Kuntz, 2020), to curvature- and amplitude-based reformulations (Adamo, 2015, Krasnov et al., 2020, Holten, 2022), and to analyses of where perturbation theory fails or must be reorganized in strong-field or modified-gravity settings [(Gumrukcuoglu et al., 2011); (Molano et al., 2020); (Corman et al., 2024); (Allahyari et al., 2018)].

1. Background expansions and the meaning of “perturbative” gravity

In the conventional approach, one begins from the Einstein–Hilbert action

$$S[g]=\frac{1}{16\pi G_{\mathrm{N}}}\int_{M}\mathrm{d}^{d}x\,\sqrt{-g}\,R\,,$$

with field equations

$R = 0 = R_{\mu\nu}\,,$

and expands the metric as $g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$ or around another chosen background. The resulting perturbation theory generates Feynman rules, infinitely many interaction vertices, and large sets of gauge-dependent intermediate expressions, even when final amplitudes or observables are compact (Adamo, 2015).

This metric-expansion viewpoint is only one realization of perturbative GR. In the post-Newtonian expansion, the organizing assumption is weak gravity and small velocity, with $v^2\sim GM/r$, and the output is an effective Hamiltonian or potential for bound motion (Bjerrum-Bohr et al., 2018, Cristofoli et al., 2019). In the post-Minkowskian expansion, one expands in powers of Newton’s constant $G_N$ without assuming small velocity, making it natural for relativistic scattering and for Hamiltonians extracted from graviton-mediated amplitudes (Bjerrum-Bohr et al., 2018, Cristofoli et al., 2019). In long-wavelength cosmological perturbation theory, the small parameter can instead be the gradient scale $\epsilon\sim 1/(HL)$, while perturbation amplitudes need not be small (Gumrukcuoglu et al., 2011).

This breadth of organizing principles suggests that “perturbative GR” is not reducible to a single formalism. A plausible implication is that the subject is better understood as a family of approximation and resummation schemes adapted to different observables: local field evolution, gauge-invariant curvature, scattering amplitudes, worldline effective actions, or wave-generation problems. The common structure is expansion around controlled data rather than exact solution of the full nonlinear Einstein equations.

2. Gauge structure and higher-order perturbation theory

A central structural issue is gauge freedom. In general-relativistic perturbation theory, the physical spacetime $(\mathcal M,\bar g_{ab})$ and the background spacetime $(\mathcal M_0,g_{ab})$ are distinct manifolds, so perturbations depend on a choice of identification map between them. With a gauge choice $v/c$0, a tensor field $v/c$1 is expanded as

$v/c$2

and under a change of gauge $v/c$3, the first two perturbative orders satisfy

$v/c$4

$v/c$5

(Nakamura, 2010).

The key premise for higher-order gauge-invariant perturbation theory is that the linear metric perturbation can be decomposed as

$v/c$6

with $v/c$7 gauge invariant and $v/c$8 carrying the pure-gauge part. Once this decomposition is available, gauge-invariant variables for arbitrary tensor fields follow uniformly: $v/c$9

$$S[g]=\frac{1}{16\pi G_{\mathrm{N}}}\int_{M}\mathrm{d}^{d}x\,\sqrt{-g}\,R\,,$$0

(Nakamura, 2010).

The same gauge issue becomes more intricate when perturbed motion is part of the problem. In self-force theory, a small compact body sources a metric perturbation and is accelerated relative to the background metric $$S[g]=\frac{1}{16\pi G_{\mathrm{N}}}\int_{M}\mathrm{d}^{d}x\,\sqrt{-g}\,R\,,$$1, but is in free fall in the effective metric

$$S[g]=\frac{1}{16\pi G_{\mathrm{N}}}\int_{M}\mathrm{d}^{d}x\,\sqrt{-g}\,R\,,$$2

The object’s center-of-mass worldline is a geodesic of $$S[g]=\frac{1}{16\pi G_{\mathrm{N}}}\int_{M}\mathrm{d}^{d}x\,\sqrt{-g}\,R\,,$$3 up to the neglected order, even though it is accelerated relative to the background (Pound, 2015). Two perturbative organizations coexist: the self-consistent approximation, where the worldline is not expanded, and the Gralla–Wald approximation, where both metric and worldline are Taylor-expanded. Their relation can be made explicit through second order, and smooth gauge transformations alter both the metric perturbations and the representative worldline (Pound, 2015).

This makes gauge not merely a technical nuisance but part of the ontology of perturbative GR. Higher-order perturbations, self-force, and long-time orbital evolution all require formulations in which governing equations, effective metrics, and source motion are transformed coherently rather than treated as separate problems.

3. Alternative perturbative variables: amplitudes, worldsheet fields, chiral fields, and curvature

A recurring theme in recent work is that the standard expansion of the Einstein–Hilbert action is often a poor language for the actual simplicity of observables. One striking example is the tree-level graviton $$S[g]=\frac{1}{16\pi G_{\mathrm{N}}}\int_{M}\mathrm{d}^{d}x\,\sqrt{-g}\,R\,,$$4-matrix, which can be written in the Cachazo–He–Yuan form

$$S[g]=\frac{1}{16\pi G_{\mathrm{N}}}\int_{M}\mathrm{d}^{d}x\,\sqrt{-g}\,R\,,$$5

localized on the scattering equations

$$S[g]=\frac{1}{16\pi G_{\mathrm{N}}}\int_{M}\mathrm{d}^{d}x\,\sqrt{-g}\,R\,,$$6

This moduli-space organization, with factorization interpreted through sphere degenerations, suggests a formulation of perturbative GR in terms of a two-dimensional chiral CFT rather than off-shell Feynman diagrams (Adamo, 2015). In that formulation, the curved-space worldsheet action is

$$S[g]=\frac{1}{16\pi G_{\mathrm{N}}}\int_{M}\mathrm{d}^{d}x\,\sqrt{-g}\,R\,,$$7

with

$$S[g]=\frac{1}{16\pi G_{\mathrm{N}}}\int_{M}\mathrm{d}^{d}x\,\sqrt{-g}\,R\,,$$8

and after the field redefinition

$$S[g]=\frac{1}{16\pi G_{\mathrm{N}}}\int_{M}\mathrm{d}^{d}x\,\sqrt{-g}\,R\,,$$9

the action becomes

$R = 0 = R_{\mu\nu}\,,$0

The free OPEs then allow exact computation of the current algebra anomaly, and anomaly cancellation is equivalent to

$R = 0 = R_{\mu\nu}\,,$1

(Adamo, 2015).

A different simplification uses first-order chiral variables. Starting from the chiral Einstein–Cartan action

$R = 0 = R_{\mu\nu}\,,$2

with

$R = 0 = R_{\mu\nu}\,,$3

one can gauge-fix so that the connection-to-connection propagators vanish: $R = 0 = R_{\mu\nu}\,,$4 The free Lagrangian then decouples into simple first-order sectors, and the effective perturbation theory contains only cubic and quartic vertices, with “special legs” that may not contract with each other (Krasnov et al., 2020). This preserves the polynomial low-valence structure of first-order gravity without the usual burden of propagating auxiliary connection lines.

A third reformulation uses curvature rather than metric perturbations. The Riemann tensor is decomposed as a Weyl part plus Ricci/Einstein pieces, and in vacuum one has

$R = 0 = R_{\mu\nu}\,,$5

Einstein’s equations and the Bianchi identities can then be reorganized into the exact vacuum curvature-wave equation

$R = 0 = R_{\mu\nu}\,,$6

In the linearized limit,

$R = 0 = R_{\mu\nu}\,,$7

in vacuum, and the electric part $R = 0 = R_{\mu\nu}\,,$8 yields two transverse polarization states and the quadrupole law in curvature form (Holten, 2022).

Taken together, these formulations indicate that perturbative GR admits multiple natural variables. Metric perturbations remain standard, but worldsheet fields, chiral first-order fields, on-shell amplitudes, and Weyl curvature can all serve as the primary perturbative language. This suggests that the complexity of Einstein gravity is representation-dependent rather than absolute.

4. Compact binaries, PM/PN theory, and reorganizations of the two-body problem

The two-body problem is the main arena in which perturbative GR becomes computationally demanding. In amplitude-based PM theory, one starts from Einstein gravity minimally coupled to massive matter,

$R = 0 = R_{\mu\nu}\,,$9

with

$g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$0

and extracts the classical long-range sector of scattering amplitudes by keeping only non-analytic dependence on the momentum transfer $g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$1, such as

$g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$2

(Bjerrum-Bohr et al., 2018). Tree amplitudes give 1PM, one-loop amplitudes give 2PM, and after Fourier transform or eikonalization one obtains, respectively, PN Hamiltonians or PM scattering angles (Bjerrum-Bohr et al., 2018).

A direct amplitude-to-Hamiltonian map uses the relativistic Lippmann–Schwinger equation. With

$g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$3

the potential kernel follows from the on-shell amplitude after Born subtraction: $g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$4 This yields a 2PM Hamiltonian for two non-spinning bodies that agrees with EFT matching approaches (Cristofoli et al., 2019).

Worldline EFT methods can also be reorganized internally. In the “half-solution” approach, one rewrites each point-particle action as

$g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$5

keeps the auxiliary einbeins $g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$6 while integrating out the linearized graviton, and obtains an effective action with only one exchange diagram in the worldline sector. The resulting algebraic equations,

$g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$7

resum an infinite class of standard NRGR matter-sector diagrams (Kuntz, 2020). The same framework exhibits an “effective two-body horizon” and, for circular motion, a Critical Innermost Circular Orbit defined by the disappearance of real redshift solutions (Kuntz, 2020).

A further reorganization replaces explicit diagrammatic expansion with a classical exact RG flow,

$g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$8

where

$g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$9

When iterated, the flow reproduces the first three PM orders and, with a local ansatz, recovers the harmonic-coordinate 1PN action without explicit three-graviton-vertex calculation (Gutiérrez et al., 31 Oct 2025). This suggests a route to non-perturbative approximations, though the paper presents the idea mainly as a proof of concept.

These developments show that in compact-binary dynamics the main difficulty is not simply expansion in small parameters, but the organization of nonlinear information. Amplitudes, worldline resummations, and RG flows all seek to isolate the genuinely hard sector of the two-body problem while avoiding large sets of gauge-dependent intermediate objects.

5. Black-hole, cosmological, and curvature perturbations

Perturbation theory on curved backgrounds remains a major branch of the subject. One line of work emphasizes that confirming a Kerr background does not suffice to confirm GR, because many modified theories share Kerr or Kerr–de Sitter as an exact solution while differing at the level of linearized perturbations. In metric $v^2\sim GM/r$0 gravity, for example, perturbations of Kerr satisfy

$v^2\sim GM/r$1

For $v^2\sim GM/r$2, one recovers the GR vacuum perturbation equation, while for $v^2\sim GM/r$3 the trace $v^2\sim GM/r$4 becomes an extra propagating scalar-like mode with

$v^2\sim GM/r$5

(0803.3433).

Another line develops closed-form perturbation methods for bosonic fields of spin

$v^2\sim GM/r$6

on isotropic spherically symmetric or conformally flat backgrounds. Using the generalized Regge–Wheeler operator

$v^2\sim GM/r$7

together with a Hadamard series

$v^2\sim GM/r$8

one can derive closed-form spin-weighted electrostatic potentials for Reissner–Nordström and fully dynamical potentials for FRW backgrounds (Suvorov et al., 2014). The method is explicitly not a replacement for Regge–Wheeler, Zerilli, or Teukolsky mode decompositions, but a complementary position-space fundamental-solution construction (Suvorov et al., 2014).

More recently, the Teukolsky strategy has been extended beyond vacuum Schwarzschild to asymptotically flat, spherically symmetric backgrounds. For

$v^2\sim GM/r$9

with $G_N$0 and $G_N$1 at infinity, and especially on the $G_N$2-symmetric subclass

$G_N$3

one obtains a modified Teukolsky equation for $G_N$4 and $G_N$5,

$G_N$6

together with a metric-reconstruction method in outgoing radiation gauge that does not rely on a Hertz potential (Guo et al., 1 Jan 2026). The same operator structure persists at second order, where the source acquires quadratic terms built from first-order perturbations (Guo et al., 1 Jan 2026).

At a broader conceptual level, some horizon phenomena are not visible at low perturbative order even when the metric perturbation is small. For Schwarzschild expanded in isotropic coordinates, the apparent-horizon condition first appears at fourth order in the potential expansion; the same is true for the de Sitter cosmological horizon in the isotropic weak-potential expansion (Allahyari et al., 2018). This does not show that perturbation theory is useless, but it does show that some observables are intrinsically nonlinear threshold conditions rather than linear-response quantities.

6. Perturbative GR in modified gravity, order reduction, and failure regimes

Perturbative methods are also used to compare modified theories with GR, but here the choice of perturbative scheme becomes decisive. In projectable Hořava–Lifshitz gravity with a scalar field, a fully nonlinear gradient expansion yields solutions continuous in the $G_N$7 limit, recovering not pure GR but

$G_N$8

with the dust arising as “dark matter as an integration constant” (Gumrukcuoglu et al., 2011). The same theory looks singular in the naive small-amplitude approach because the momentum constraint is linearized in a regime where that approximation ceases to be valid. The paper identifies two branches, a linear branch and a nonlinear branch, and argues that the apparent strong coupling belongs to the naive perturbative expansion rather than to the underlying classical theory (Gumrukcuoglu et al., 2011).

A different use of perturbation theory appears in vacuum $G_N$9 gravity. For analytic models of the form

$\epsilon\sim 1/(HL)$0

the perturbative-constraints or order-reduction framework expands the solution and the theory itself in powers of $\epsilon\sim 1/(HL)$1. In vacuum, one then finds recursively that

$\epsilon\sim 1/(HL)$2

order by order, so no perturbative correction beyond GR appears at any finite order within that analytic branch (Molano et al., 2020). This does not imply that $\epsilon\sim 1/(HL)$3 is equivalent to GR, but rather that genuinely new vacuum behavior can reside in non-analytic or singularly perturbed solution branches disconnected from the GR perturbative branch (Molano et al., 2020).

Strong-field modified-gravity simulations reveal a more operational limitation. In shift-symmetric Einstein-scalar-Gauss-Bonnet gravity,

$\epsilon\sim 1/(HL)$4

the order-by-order expansion around a fully nonlinear GR background uses

$\epsilon\sim 1/(HL)$5

with

$\epsilon\sim 1/(HL)$6

at leading nontrivial orders (Corman et al., 2024). The method is mathematically close to perturbative GR around a nonlinear background, but in binary inspiral it suffers secular error because the true modified theory changes the orbital dynamics while the perturbative background remains GR. The paper concludes that the order-by-order method “cannot faithfully track the solutions when the corrections to general relativity are non-negligible,” whereas the fixing-the-equations approach can remain consistent if its auxiliary fields are driven on timescales short compared with the physical timescales (Corman et al., 2024).

These examples show that perturbative GR is not merely a formal expansion technology; it is also a diagnostic for when an approximation scheme matches the dynamics it is meant to describe. Small instantaneous corrections do not guarantee long-time fidelity, and a perturbative branch of solutions need not approximate the full solution space of a theory.

7. Scope, limitations, and open directions

Several limitations recur across the literature. First, tree level and sphere correlators are the best understood sectors in worldsheet formulations; higher genus and full loop consistency remain less settled (Adamo, 2015). Second, many simplifications rely on special geometric structures such as type D, $\epsilon\sim 1/(HL)$7-symmetry, asymptotic flatness, or conformal flatness [(Suvorov et al., 2014); (Guo et al., 1 Jan 2026)]. Third, gauge-invariant higher-order formalisms on generic backgrounds still face global issues, especially the zero-mode problem associated with kernels of elliptic operators and the dependence on boundary conditions (Nakamura, 2010). Fourth, many perturbative schemes control only conservative dynamics, not radiation reaction, or only one sector of the full Einstein nonlinearity, as in the truncation that resums worldline couplings but omits bulk graviton self-interactions (Kuntz, 2020).

A further limitation is conceptual. Some observables are perturbatively accessible but coordinate dependent, such as potentials or Hamiltonians before translation to gauge-invariant scattering data (Bjerrum-Bohr et al., 2018). Others are defined by conditions that can be absent at low order, such as horizons or strong-field branch structure (Allahyari et al., 2018, Molano et al., 2020). The subject therefore demands continual separation of invariant content from representation-dependent intermediate quantities.

At the same time, the range of existing formulations indicates that perturbative GR is not exhausted by linearized metric waves around Minkowski space. It includes gauge-invariant curvature dynamics (Holten, 2022), effective worldline and amplitude descriptions of classical scattering (Bjerrum-Bohr et al., 2018, Cristofoli et al., 2019), chiral and worldsheet reformulations (Adamo, 2015, Krasnov et al., 2020), multipolar near-zone constructions of $\epsilon\sim 1/(HL)$8-body dynamics (Turyshev et al., 2013), and higher-order motion/gauge frameworks for self-force problems (Pound, 2015). This suggests that the subject’s unifying principle is not a particular perturbation variable, but the controlled extraction of nonlinear Einstein dynamics from expansions tailored to the observable and regime of interest.