Universality from Perturbation Modes

Universality in critical collapse is the loss of initial-data detail caused by attraction toward a threshold solution. Perturbation modes make that statement precise: decaying modes span directions along the critical surface, while one growing mode points away from it. Families that cross the same surface transversely then share the intermediate solution and critical exponents, although their threshold parameters and amplitudes differ.

Required background. The Einstein equation defines the nonlinear flow on constrained data, while differential equations supplies linearization, spectra, and stable manifolds.

Helpful background. Black-Hole Geometry and Mechanics supplies endpoint observables; Black-Hole Perturbation Theory develops physical mode decompositions; Formulations of General Relativity explains gauge and constraint modes; and Computational Relativity develops family comparisons.

Codimension from the mode spectrum

Treat gauge-fixed Einstein–matter evolution as a flow on initial data satisfying the constraints. Let $\mathcal C$ be the boundary between two open basins, such as dispersion and black-hole formation. A critical solution $Z_*$ is an attractor within $\mathcal C$ if nearby data on the boundary approach it.

For a continuously self-similar solution,

$$\delta Z(\tau,x)=\sum_i C_i e^{\lambda_i\tau}Z_i(x).$$

For a discretely self-similar solution, $Z_i$ is periodic in $\tau$ and $\lambda_i$ is a Floquet exponent. Gauge transformations, variations that violate the constraints, and changes of an arbitrary accumulation time are not physical instabilities. After removing them, exactly one independent real growing amplitude means the stable manifold of $Z_*$ has codimension one in the full allowed phase space. Degenerate angular modes and complex conjugate pairs can carry several real amplitudes even when one writes only one eigenvalue.

A generic one-parameter family intersects a codimension-one surface at isolated $p_*$. If there are two independent real growing amplitudes, one parameter cannot suppress both; reaching $Z_*$ generically needs two tunings. Conversely, observing one scaling law inside an imposed symmetry class proves nothing about modes excluded by that symmetry.

Why different families agree

Write a near-threshold family as

$$Z_{\mathcal F}(p,\tau) =Z_*+\kappa_{\mathcal F}(p-p_{*,\mathcal F}) e^{\lambda_0\tau}Z_0 +\sum_{i\geq1}C_{i,\mathcal F}e^{\lambda_i\tau}Z_i.$$

As $\tau$ grows, every term with $\operatorname{Re}\lambda_i<0$ decays. Family information survives mainly in the threshold location $p_{*,\mathcal F}$, the coefficient $\kappa_{\mathcal F}$, and an overall phase for DSS. The growing eigenvalue and dimensionless profile belong to $Z_*$, giving a common mass exponent $\gamma=1/\lambda_0$. Koike, Hara, and Adachi make this stable-manifold argument explicit for radiation-fluid collapse and derive the reciprocal exponent in Koike, Hara, and Adachi 1995, § 2b–2c, Eqs. (8)–(15), pp. 5171–5172.

This is universality across initial-data families within a fixed universality class. It does not require identical waveforms at infinity, because exterior regions retain family-specific information and need not enter the self-similar core. Nor does it imply universality across matter models. Models share a class only if their extra couplings become irrelevant in the shrinking-scale limit and they approach the same effective critical equations.

The phase-space account and one-mode argument are developed in Gundlach and Martín-García 2007, §§2.1–2.3, pp. 4–8. Their discussion of dimensionful couplings and universality classes is in Gundlach and Martín-García 2007, §2.6, pp. 9–10, PDF.

Relevant, irrelevant, and marginal directions

The terms “relevant” and “irrelevant” refer to growth under evolution toward smaller scales. A positive $\operatorname{Re}\lambda$ is relevant; a negative one is irrelevant. A zero-real-part mode requires nonlinear analysis and may parameterize a family of critical solutions, a symmetry, or a genuine marginal instability. Complex exponents produce spiraling approach or oscillatory departure.

Dimensionful terms can also be irrelevant by scaling. Let the shrinking local length be $\ell=L_0e^{-\tau}$. A scalar mass $\mu$ enters the rescaled Klein–Gordon equation through $(\mu\ell)^2$, the ratio of the mass term to terms containing two derivatives. It therefore disappears as $\ell\to0$, provided the field amplitudes remain in the assumed regime. Equivalently, for Compton length $L_\mu=\mu^{-1}$ the small parameter is $\ell/L_\mu$, a dimensionless ratio. A higher-derivative correction can scale oppositely and become important at small $\ell$. By contrast, a dimensionless fluid equation-of-state parameter can remain in the limiting equations and change $Z_*$ and $\gamma$.

The spectrum depends on the admitted perturbations. Spherical calculations exclude nonspherical harmonics; axisymmetry excludes $m\ne0$; a matter ansatz may exclude charge or vorticity. A critical solution is generic only relative to a stated phase space.

Tests and limitations

Test universality with several initial-data shapes, amplitudes, and widths that cross the threshold independently. After aligning scale and DSS phase, dimensionless central profiles should converge to one curve. The mass and curvature exponents should agree, while thresholds and intercepts may differ. A perturbative eigenvalue computation should reproduce the fitted exponent.

Failed universality can have several meanings: data may approach different critical solutions, a second unstable mode may intervene, the scaling window may be too short, a physical scale may become relevant, or gauge alignment may be wrong. Agreement in a single scalar exponent is weaker than agreement in the full dimensionless profile and spectrum.

Check your understanding

A spherical critical solution has one growing $\ell=0$ amplitude and one real growing physical $\ell=1$ multiplet, with gauge modes already removed. How many real amplitudes must be tuned in spherical and unrestricted data, and what universality claim is justified?

Solution

Within spherical data, one tuning suppresses the $\ell=0$ amplitude and the solution can be a codimension-one spherical threshold attractor. The $\ell=1$ sector has $2\ell+1=3$ independent real angular components with the same growth rate. Unrestricted data therefore require four tunings, including the spherical amplitude. Generic one-parameter families will not approach this solution arbitrarily closely. Universality remains a claim about the spherical phase space, not the unrestricted theory.

References

• Gundlach, Carsten, and José M. Martín-García. “Critical Phenomena in Gravitational Collapse.” Living Reviews in Relativity 10 (2007): 5. doi:10.12942/lrr-2007-5.
• Koike, Tatsuhiko, Takashi Hara, and Satoshi Adachi. “Critical Behavior in Gravitational Collapse of Radiation Fluid: A Renormalization Group Analysis.” Physical Review Letters 74 (1995): 5170–5173. doi:10.1103/PhysRevLett.74.5170.