Matching Perturbative and Full Nonlinear Dynamics

Two relativistic calculations cannot be matched by overlaying coordinate trajectories. Coordinates, time slicings, centers, and spin supplementary conditions differ across perturbation theory, effective descriptions, and numerical relativity. Matching must use quantities with a common operational definition—conserved charges, scattering angle, frequency–energy relations, asymptotic radiation in an aligned Bondi frame, or quasi-local horizon invariants—and it must occur in a demonstrable overlap regime.

Helpful background. Black-Hole Geometry and Mechanics supplies horizon charges; Black-Hole Perturbation Theory supplies controlled expansions; Formulations of General Relativity supplies gauge freedom; Relativistic Two-Body Dynamics supplies post-Newtonian, self-force, and effective descriptions; Computational Relativity supplies nonlinear evolution and extraction.

Choose a shared invariant

For conservative bound motion, candidates include the relation between binding energy and orbital frequency,

$$E_b(\Omega),$$

periastron advance as a function of invariant frequencies, and Detweiler’s redshift in the small-mass-ratio limit. For scattering, use

$$\chi(E,J),$$

the asymptotic deflection angle at fixed center-of-mass energy and angular momentum. For dissipative evolution, match Bondi fluxes and waveform modes only after fixing asymptotic time, rotation, boost, translation, and supertranslation freedom.

Scattering angles provide a particularly clean bridge between post-Minkowskian calculations, effective-one-body dynamics, self-force expansions, and numerical relativity Damour 2016, §§II–V, pp. 2–16. For quasi-circular binaries, the first law relates variations of total energy and angular momentum to invariant redshift variables Le Tiec, Blanchet, and Whiting 2012, Eqs. (2.1)–(2.47), pp. 2–13.

The overlap test

For a concrete example, suppose method A is a post-Newtonian expansion in $x=(GM\Omega)^{2/3}$, with total mass $M$ and fixed symmetric mass ratio $\nu=m_1m_2/M^2$, while method B is a full numerical solution. Post-Newtonian theory does not require $\nu\ll1$; a self-force comparison would instead need an independently controlled small mass ratio. Choose an interval where:

$$x\ll1$$

enough for neglected perturbative orders to be small, yet the numerical signal is long, resolved, and free of initial transients and boundary contamination. Compare

$$\Delta Q(x) =Q_{\rm NR}(x)-Q_{\rm pert}(x)$$

with numerical, truncation, and alignment uncertainties. If they can be modeled as independent standard deviations, their quadrature combination is

$$\sigma_{\rm tot}^2 =\sigma_{\rm NR}^2 +\sigma_{\rm trunc}^2 +\sigma_{\rm align}^2.$$

Correlated errors require covariance terms. Deterministic error bounds instead combine conservatively as $\delta_{\rm tot}\leq\delta_{\rm NR}+\delta_{\rm trunc}+\delta_{\rm align}$ for the bound on their sum; one must not silently interpret an omitted-order estimate as a Gaussian standard deviation. A successful overlap keeps the discrepancy within the justified uncertainty across an interval, not at one tuned point. Vary the interval and truncation order. If the discrepancy falls with the expected next power of $x$ while numerical error converges independently, the comparison discriminates approximation error from computation error.

Continuum inputs and benchmark ladder

For the nonlinear calculation, record the Einstein or Einstein–matter continuum system, initial-data family, gauge, boundaries, resolution sequence, horizon definitions, and extraction frame. For the approximate method, record expansion variables, resummation, calibration parameters, and omitted order.

Use a benchmark ladder:

1. exact test-mass geodesics for conservative scattering;
2. black-hole perturbation theory and self-force for small mass ratio;
3. post-Newtonian or post-Minkowskian relations at weak field;
4. independently converged numerical relativity in the strong-field overlap;
5. balance closure among ADM initial charges, Bondi radiation, and final horizon charges.

The numerical-relativity review describes the merger calculations and the interface with analytic waveform and dynamics models in Centrella et al. 2010, §§IV–VI, pp. 8–33.

Frame alignment without hiding disagreement

Waveform matching requires an asymptotic Bondi frame. A time and phase shift may remove conventional origins, but an arbitrary time-dependent warp can absorb physical disagreement. For precessing or recoiling systems, align rotations, boosts, and translations using declared geometric conditions, and account for supertranslations where relevant.

Report results before and after the allowed alignment. A coordinate separation $r_{\rm coord}(t)$ is unsuitable because a gauge transformation changes it without changing radiation or scattering. Likewise, a coordinate common-horizon time cannot be matched directly to an analytic “merger time.”

Convergence, failure modes, and rerun boundary

At least three effective resolutions or an equivalent tolerance sequence should converge each numerical invariant. Perturbative truncation should be estimated by order variation or comparison among known coefficients. A principal failure mode is calibration and validation on the same numerical data, which turns agreement into interpolation. Other limitations include junk radiation, finite extraction, inconsistent mass normalization, BMS-frame mismatch, and overlap intervals too narrow to separate slopes.

No simulation or approximate-model fit was rerun for this page. A later clean rerun must reproduce the limiting benchmark, regenerate invariant comparisons and uncertainty components, and retain machine-readable data and alignment transformations.

The matching problem ends here when the shared invariants establish entry into a common-horizon regime. The subsequent formation and evolution of that common horizon belong to Black-Hole Mergers and Common Horizons, not to a continued extrapolation of the encounter expansion.

Check the match

At one point in an overlap interval, $\Delta Q=0.010$. Assume, for this exercise, independent zero-mean Gaussian errors with standard deviations $\sigma_{\rm NR}=0.004$, $\sigma_{\rm trunc}=0.008$, and $\sigma_{\rm align}=0.002$. Is the discrepancy significant at a two-standard-deviation threshold? What if these numbers are only deterministic bounds?

Solution

The quadrature uncertainty is

$$\sigma_{\rm tot} =\sqrt{0.004^2+0.008^2+0.002^2} \simeq0.0092.$$

The discrepancy is about $1.09\sigma_{\rm tot}$, below the stated $2\sigma$ threshold. If the three numbers are only bounds, their conservative sum is $0.014$ and the discrepancy remains within that bound; a statistical significance cannot be assigned. Neither calculation validates a whole interval from a single point. One should vary the overlap interval and approximation order and account for correlations between sampled points.

References

• Centrella, Joan, John G. Baker, Bernard J. Kelly, and James R. van Meter, “Black-Hole Binaries, Gravitational Waves, and Numerical Relativity”, Reviews of Modern Physics 82, 3069–3119, 2010. Open PDF.
• Damour, Thibault, “Gravitational Scattering, Post-Minkowskian Approximation and Effective-One-Body Theory”, Physical Review D 94, 104015, 2016. Open PDF.
• Le Tiec, Alexandre, Luc Blanchet, and Bernard F. Whiting, “The First Law of Binary Black Hole Mechanics in General Relativity and Post-Newtonian Theory”, Physical Review D 85, 064039, 2012. Open PDF.