Gravitational Self-Force and Regularization

Gravitational self-force theory describes a compact body whose mass $m$ is small compared with the curvature-producing mass scale $M$, while retaining the body’s backreaction on the spacetime and orbit. The central result is not that an object pushes on itself. It is that matched small-body expansions separate a locally singular field, which belongs to the body’s multipolar structure, from a smooth regular field that changes its effective external geometry. This chapter develops that statement into analytical and computational methods through second perturbative order.

Helpful background. The Einstein Equation supplies the nonlinear field equation being expanded; Weak Metric Perturbations and Gauge Freedom supplies perturbative gauge transformations; Presymplectic Potentials and Currents supplies the relation between local fields, boundary data, and conserved charges.

The small-body expansion

Let $\mathcal R$ denote the background curvature radius and introduce

$$\varepsilon=\frac{Gm}{\mathcal R}\ll1.$$

The body’s size shrinks with its gravitational radius $Gm$ while its compactness remains finite. An inner expansion resolves the body on that scale; an outer expansion replaces it by a singular multipolar field around a representative worldline. For a black-hole background with $\mathcal R=O(GM)$, this small parameter is of the same order as the mass ratio $m/M$. In their common buffer region,

$$Gm\ll s\ll\mathcal R,$$

matching identifies the body’s mass, spin, and higher multipoles and determines its motion. The limiting worldline is a background geodesic. At first order, a nonspinning monopolar body moves as a geodesic of

$$\tilde g_{\mu\nu} =g_{\mu\nu} +\varepsilon h_{\mu\nu}^{\mathrm R(1)}.$$

At second order the effective metric includes $\varepsilon^2h_{\mu\nu}^{\mathrm R(2)}$ together with the correct perturbative interpretation of the worldline and its arguments.

The point-particle description appears only after this limiting construction. A worldline source such as

$$T^{\mu\nu}(x) =m\int u^\mu u^\nu \delta_4(x,z(\tau))\,d\tau$$

uses the invariant delta distribution, normalized by $\int\delta_4(x,z)f(x)\sqrt{-g}\,d^4x=f(z)$. It is defined by integration against smooth test tensors. Its retarded solution diverges at the worldline, so substituting it directly into a geodesic equation is meaningless. The mathematical task is to isolate the sourced singular structure and retain a field smooth enough for the derivative required by the force or observable. Poisson, Pound, and Vega develop the Green-function, point-particle, and small-body parts of this construction in Poisson, Pound, and Vega 2011, §§ 14–24, while Pound and Wardell 2021, § 5, pp. 40–54 emphasize the modern matched and puncture formulations.

The chapter sequence

The six pages form a dependency chain, not six interchangeable regularization recipes.

1. Matched Expansions and the Small-Body Limit begins with a finite compact body, constructs inner and outer expansions, matches them in the buffer region, and derives the geodesic limit and first-order effective-metric motion.
2. Detweiler–Whiting Singular and Regular Fields gives the local scalar, electromagnetic, and gravitational Green-function decompositions. It separates the definition of the split, smooth homogeneity of the regular field, effective-metric motion, and the singular no-force statement.
3. Mode-Sum Regularization projects the retarded force onto finite angular modes, subtracts the analytically predicted large-$\ell$ singular sequence, and controls the remaining tail, one-sided limits and precision. A reproducible circular Schwarzschild scalar calculation demonstrates the method; gravitational gauge qualifications are stated separately.
4. Effective-Source and Puncture Methods moves a finite-order singular approximation into the source before numerical solution. A time-domain scalar wave calculation tests source differentiability, worldtubes, boundaries and refinement against the frequency-domain result. A published gravitational benchmark explains what the tensor extension additionally requires.
5. Gauge-Invariant Self-Force Observables replaces coordinate acceleration by redshift, frequency ratios, spin precession, tidal eigenvalues, and scattering data, always with the admissible gauge class and asymptotic normalization stated.
6. Second-Order Self-Force and Nonlinear Regularization confronts quadratic singular products, second-order punctures and gauges, mass and spin completion, multiscale waveforms, and the boundary between validated quasicircular results and the generic-Kerr frontier.

The order matters. A puncture is interpretable only after the exact singular/regular split is understood; a gauge-invariant comparison requires a regularized field and matched asymptotic clock; a second-order source inherits every first-order convention and adds nonlinear ones.

Singular and regular fields

For a linear scalar model,

$$L\Phi_{\mathrm{ret}}=-4\pi\rho, \qquad \Phi_{\mathrm{ret}}=\Phi_{\mathrm S}+\Phi_{\mathrm R}.$$

The singular field obeys the same inhomogeneous equation and reproduces the retarded field’s local divergence. Therefore

$$L\Phi_{\mathrm R}=0$$

near the worldline, and $\Phi_{\mathrm R}$ is smooth there when the singular terms have been matched to sufficient order. The vector and tensor cases require their own Green functions, curvature couplings, gauge conditions, and force laws; indices cannot simply be attached to the scalar result.

For gravity in a smooth vacuum background and a sufficiently regular first-order gauge,

$$h_{\mu\nu}^{\mathrm{ret}} =h_{\mu\nu}^{\mathrm S} +h_{\mu\nu}^{\mathrm R},$$

where $h^{\mathrm R}=\varepsilon h^{\mathrm R(1)}$ includes the small mass-ratio factor. The acceleration of a nonspinning monopolar body, through first order, is

$$a^\mu =-\frac12 \left(g^{\mu\nu}+u^\mu u^\nu\right) \left( 2h_{\nu\rho;\sigma}^{\mathrm R} -h_{\rho\sigma;\nu}^{\mathrm R} \right)u^\rho u^\sigma.$$

The projector guarantees $u_\mu a^\mu=0$. Detweiler and Whiting give the local decomposition and effective-field interpretation in Detweiler and Whiting 2003, §§ III–VI, especially eqs. (17)–(24) and (39)–(41). Later small-body matching supplies the gravitational no-force statement with a derivational basis: the singular field matches the body’s local multipolar geometry, while the regular field supplies the external tidal environment.

This split has a precise domain. Its Hadamard representation is local to a convex normal neighborhood. Global caustics and multiple null geodesics are not determined by local coefficients. A finite-order puncture approximates $h^{\mathrm S}$ only near the worldline and leaves finite differentiability. A singular gauge vector can invalidate the smooth-worldline argument even when a reconstructed metric is adequate away from the particle.

Two computational realizations

Mode sum and effective source implement the same local subtraction at different stages.

In mode sum, each retarded angular mode is finite on the worldline, but its large-$L$ behavior, with $L=\ell+\tfrac12$, is

$$F_{\alpha,\pm}^{\mathrm{ret},\ell} =A_{\alpha,\pm}L +B_\alpha +\frac{C_\alpha}{L} +O(L^{-2}).$$

The self-force is obtained from the convergent sum

$$F_\alpha^{\mathrm{self}} =\sum_{\ell=0}^{\infty} \left( F_{\alpha,\pm}^{\mathrm{ret},\ell} -A_{\alpha,\pm}L -B_\alpha -\frac{C_\alpha}{L} \right)-D_\alpha.$$

The $A$, $B$, $C$, and $D$ data depend on the field, component, harmonic basis, extension, and gauge. A wrong $A$ leaves linear growth, a wrong $B$ leaves a constant, and a wrong $C$ leaves a logarithmically divergent partial sum. Higher regularization parameters accelerate the inverse-power tail. The foundational prescription is derived by Barack and Ori 2000, pp. 2–4, eqs. (3)–(6).

In an effective-source method, a local puncture $\Phi_{\mathrm P}$ is subtracted before solving:

$$\Phi_{\mathrm{res}} =\Phi_{\mathrm{ret}}-W\Phi_{\mathrm P},$$ $$L\Phi_{\mathrm{res}} =-4\pi\rho-L(W\Phi_{\mathrm P}) \equiv S_{\mathrm{eff}}.$$

The delta distribution cancels analytically against the distributional image of the puncture. The resulting extended source can be evolved in multiple dimensions, but its differentiability limits numerical order. A smooth window localizes it, or a worldtube matches residual and retarded variables across an interface.

The methods permit useful cross-checks because their numerical errors differ. Mode sum is sensitive to angular tails, one-sided limits and cancellation precision. Effective source is sensitive to spatial and temporal resolution, puncture differentiability, tube or window placement, and boundary treatment. The two downloadable examples compare the same covariant Schwarzschild scalar radial force at $r_0/(GM)=6$ and $10$, with retarded boundary conditions. This is a coordinate-component comparison, not a gauge-invariant gravitational result. Their wave solutions are independent, but their higher angular subtraction coefficients are shared; agreement does not independently validate those coefficients. An invariant comparison tests an additional physical question after the required coordinate, gauge and normalization checks.

From gauge-dependent force to observables

Under an infinitesimal diffeomorphism,

$$h_{\mu\nu}^{\mathrm R} \mapsto h_{\mu\nu}^{\mathrm R} +2\nabla_{(\mu}\xi_{\nu)},$$

and the coordinate worldline moves. Consequently, a coordinate force component is not invariant. Circular motion instead supplies the helical vector

$$k^\mu=t^\mu+\Omega\phi^\mu, \qquad u^\mu=Uk^\mu.$$

The redshift relation

$$U(\Omega) =\left(-\tilde g_{\mu\nu}k^\mu k^\nu\right)^{-1/2}$$

uses the effective-metric proper time and the conservative circular problem. It is invariant under smooth helically symmetric transformations that preserve the asymptotic clock and mass normalization. At first order and fixed $\Omega$, write $U=U_0+\delta U+O(\varepsilon^2)$, with

$$\delta U =\frac12U_0h_{uu}^{\mathrm R}.$$

For eccentric Schwarzschild motion, the pair $(\Omega_r,\Omega_\phi)$ replaces coordinate $(p,e)$ as comparison variables. The ratio $K=\Omega_\phi/\Omega_r$ is a useful derived observable; specifying both frequencies already specifies $K$. Generic Kerr motion adds a polar frequency. Spin precession and tidal eigenvalues add independent information. Unbound motion supplies a scattering angle at fixed asymptotic energy and angular momentum.

No observable is invariant without qualifications. A rescaling of asymptotic time changes $U$ and $\Omega$. A no-string radiation gauge requires side averaging and completion. A tidal eigenvalue needs two controlled derivatives of the regular field, while a self-force needs one. A scattering angle needs a shared Poincaré frame at both infinities. These conditions are part of the definition, not technical corrections applied after comparison.

Second order and waveform phase

At second order,

$$\delta G[h^{(2)}] =8\pi G T^{(2)} -\delta^2G[h^{(1)},h^{(1)}].$$

Here $h^{(n)}$ and $T^{(n)}$ are coefficients of $\varepsilon^n$, and $\delta^2G$ includes the quadratic Taylor coefficient. The physical first-order perturbation $\varepsilon h^{(1)}$ scales as $Gm/s$, so the physical quadratic source $\varepsilon^2\delta^2G$ behaves as $G^2m^2/s^4$ and is not locally integrable on a spatial slice. The physical second-order puncture begins at $G^2m^2/s^2$ and cancels the nonintegrable terms to define a residual effective source. Gauge transformations also acquire terms $\mathcal L_{\xi_1}h^{(1)}$ and $\mathcal L_{\xi_1}^2g$, while monopole and dipole completion determine physical masses, angular momentum, and worldline choice.

These difficulties matter for waveform phase. An inspiral lasts $O(\varepsilon^{-1})$ orbital cycles. A fractional $O(\varepsilon)$ correction to the instantaneous frequency can therefore accumulate an $O(1)$ phase. First-postadiabatic waveforms require first-order conservative information and second-order dissipative information organized in a two-timescale expansion, not merely a second-order field snapshot.

Quasicircular Schwarzschild calculations have produced binding energy, flux, and nonspinning waveforms, and slowly spinning specialized extensions exist. The documented source contains quadratic mode coupling, slow evolution, first-order puncture products, and the second-order puncture, with separate validation Upton et al. 2026, §§ II–VI. The second-order page distinguishes these results from generic Kerr waveforms and includes a dated September 2026 comparison between Teukolsky and Lorenz-gauge flux calculations.

A hierarchy of checks

The chapter’s methods are accepted through checks that target different failure mechanisms.

Local analytical checks. The singular field must reproduce the delta normalization and correct local powers. Inner and outer expansions must match the same monopole and dipole. The regular field must be homogeneous and smooth to the needed derivative order. At second order, nonintegrable powers must cancel before numerical evaluation.

Numerical checks. Mode residuals must approach their predicted large-$\ell$ slope, tail fits must be stable under window changes, and increased precision must preserve quoted digits. Effective-source calculations separate spatial, time-step, relaxation and worldtube or window errors. Finite-domain methods also vary boundary placement; the example with horizon and null-infinity endpoints instead tests its hyperboloidal transition. Arithmetic precision needs a separate check when roundoff or cancellation approaches the claimed error. Constraint residuals, when present, must converge at a compatible rate.

Invariant checks. Two admissible gauges must yield the same $U(\Omega)$, frequency relation, or tidal eigenvalue after clock and mass normalization. Local dissipation must agree with horizon and infinity fluxes. Circular redshift and binding energy must satisfy the applicable first law.

Regime checks. Sending $\varepsilon\to0$ must recover background geodesic motion. Weak-field results should match post-Newtonian theory, while overlap with numerical relativity or effective-one-body calculations provides an external comparison. Near the light ring, separatrix, transient resonance, or plunge, coefficients can grow and the expansion may cease to be uniform even when $\varepsilon$ remains small.

Scope and routes onward

This chapter treats nonspinning monopolar motion as its baseline. Spin, quadrupole moments, tidal response, and material structure enter according to their mass-ratio scaling and cannot be hidden inside “regularization.” The scalar and electromagnetic examples clarify the decomposition, but their force laws are not gravitational results.

For weak-field comparable-mass dynamics, continue to Post-Newtonian Dynamics. For the mode equations, reconstruction, horizon conditions, and Kerr separability used by the numerical methods, continue to Black-Hole Perturbation Theory. For long inspirals and detector-facing phase models, the radiation-reaction and extreme-mass-ratio chapters supply multiscale orbital evolution and waveform construction.

The durable lesson is methodological: begin with a controlled small-body limit, separate local singular structure from global retarded boundary data, regularize before taking a worldline limit, compare relational observables in a declared gauge class, and state where the perturbative and numerical evidence ends.

Review the chapter

Consider a scalar model with $\Phi_{\mathrm{ret}}=\Phi_{\mathrm S}+\Phi_{\mathrm R}$ and a gravitational circular-orbit calculation using either mode sum or a puncture.

1. If $L\Phi_{\mathrm{ret}}=L\Phi_{\mathrm S}=-4\pi\rho$, what equation does $\Phi_{\mathrm R}$ satisfy near the worldline, and what additional fact is needed before evaluating its gradient there?
2. A regularized mode sequence behaves as $R_\ell=c_2/L^2+c_4/L^4+\cdots$. How does the truncation tail scale before and after subtracting the known $c_2/L^2$ term and restoring its exact sum?
3. An effective-source result changes when the worldtube radius and mesh spacing are reduced together. Why does this not isolate the error?
4. Two gauges give different $U$ at the same coordinate radius but the same $U(\Omega)$ after time normalization. Is this agreement or disagreement?
5. At second order, $(\nabla h^{(1)})^2\sim s^{-4}$. Explain why an ordinary pointwise source is insufficient and name two further ingredients needed before claiming a generic waveform.

Solution

Linearity gives

$$L\Phi_{\mathrm R} =L\Phi_{\mathrm{ret}}-L\Phi_{\mathrm S}=0.$$

Equal equations alone do not ensure a finite force. The local singular expansions must cancel to sufficient order that $\Phi_{\mathrm R}$ has a unique, sufficiently differentiable worldline limit; a force requires its first derivative.

Since $\sum_{\ell\gt N}L^{-p}=O(N^{1-p})$, the $c_2/L^2$ tail is $O(N^{-1})$. After subtracting it and restoring its analytic infinite sum, the remaining sequence begins at $L^{-4}$ and the tail is $O(N^{-3})$.

Changing tube radius and mesh spacing simultaneously mixes interface and discretization errors. They can cancel or reinforce, so no convergence order can be assigned. Hold the tube fixed while refining the mesh, then hold a resolved mesh fixed while varying the tube. A finite-domain boundary requires its own placement study, and arithmetic precision requires a separate check when cancellation or roundoff limits the result.

The differing coordinate-radius values are not a physical disagreement because coordinate radius is gauge dependent. Agreement of $U(\Omega)$ after matching the asymptotic clock is the correct invariant comparison. One must also match the mass convention, low-mode completion, and worldline prescription.

With spatial measure $s^2ds\,d\Omega$, the absolute integral of $s^{-4}$ behaves as $\int s^{-2}ds$ and diverges. A nonlinear puncture or a proved highly regular distributional formulation is required before a source is defined. Even after local regularization, a generic waveform needs, among other things, globally retarded boundary and asymptotic-frame control, low-multipole completion, convergent mode coupling, and generic Kerr orbital evolution including resonances.

References

• Barack, Leor, and Amos Ori. “Mode Sum Regularization Approach for the Self Force in Black Hole Spacetime.” Physical Review D 61, 061502(R) (2000). doi:10.1103/PhysRevD.61.061502.
• Detweiler, Steven, and Bernard F. Whiting. “Self-Force via a Green’s Function Decomposition.” Physical Review D 67, 024025 (2003). doi:10.1103/PhysRevD.67.024025.
• Poisson, Eric, Adam Pound, and Ian Vega. “The Motion of Point Particles in Curved Spacetime.” Living Reviews in Relativity 14, 7 (2011). doi:10.12942/lrr-2011-7.
• Pound, Adam, and Barry Wardell. “Black Hole Perturbation Theory and Gravitational Self-Force.” In Handbook of Gravitational Wave Astronomy, edited by Cosimo Bambi, Stavros Katsanevas, and Konstantinos D. Kokkotas, 1–119. Singapore: Springer (2021). doi:10.1007/978-981-15-4702-7_38-1.
• Upton, Samuel D., Barry Wardell, Adam Pound, Niels Warburton, and Leor Barack. “Effective Source for Second-Order Self-Force Calculations: Quasicircular Orbits in Schwarzschild Spacetime.” Physical Review D 113, 064013 (2026). doi:10.1103/f9j2-k64r.