Scalar-Field Collapse Theorems

Spherical symmetry turns the Einstein–massless-scalar system into a $1+1$-dimensional nonlinear characteristic problem while retaining genuine gravitational collapse: energy disperses, forms trapped spheres and black holes, or—on exceptional data—can produce a visible singular endpoint. The Christodoulou theorem series is valuable precisely because it proves all three possibilities in one fixed matter model. Its censorship conclusion is an instability statement in a specified characteristic data topology, not a license to extrapolate spherical results to vacuum gravity or to every rough function space.

Required background. Analysis for Mathematical Relativity supplies bounded-variation and characteristic estimates; timelike congruences and spatial projection supplies null expansions; and the Einstein equation fixes the scalar stress tensor and energy condition.

Helpful background. Causality and Global Structure distinguishes visible singular boundaries from incomplete curves, and PDEs and Green Functions for Gravitation supports the reduced hyperbolic system.

The spherical Einstein–scalar system

Let the quotient by $SO(3)$ have double-null coordinates $(u,v)$ and write

$$g=-\Omega^2\,du\,dv+r^2d\omega^2, \qquad R_{\mu\nu}=8\pi G\,\partial_\mu\phi\,\partial_\nu\phi, \qquad \Box_g\phi=0.$$

The notation $du\,dv$ is the symmetric cross term: $g_{uv}=g_{vu}=-\Omega^2/2$, so $g^{uv}=g^{vu}=-2\Omega^{-2}$. Both $u$ and $v$ increase toward the future along their respective null families. Here $4\pi r^2$ is the area of an orbit sphere. The physical Hawking mass $m$ is defined invariantly by

$$1-\frac{2Gm}{r}=g^{ab}\partial_a r\,\partial_b r.$$

Explicitly, the right side is $-4\Omega^{-2}\partial_ur\,\partial_vr$. The two expansions in the positive coordinate-null normalization are $\theta_u=2\partial_ur/r$ and $\theta_v=2\partial_vr/r$; normalized null vectors multiply them by positive factors. This makes the invariant mass condition directly comparable with the geometric definition of trapping.

The sign of the right side divides the quotient into regular, marginal, and trapped regions. With $\partial_u r<0$ chosen ingoing, a regular exterior has $\partial_v r>0$; a trapped sphere has $\partial_v r<0$ as well, and $2Gm/r>1$.

In a consistent double-null normalization, the null Raychaudhuri constraints take the form

$$\partial_u(\Omega^{-2}\partial_u r) =-4\pi G r\Omega^{-2}(\partial_u\phi)^2, \qquad \partial_v(\Omega^{-2}\partial_v r) =-4\pi G r\Omega^{-2}(\partial_v\phi)^2.$$

Their right sides have a fixed sign. Together with monotonicity identities for $m$, this makes spherical collapse unusually tractable: energy flux, focusing, and the trapped condition can all be expressed through scalar functions on the quotient. Birkhoff’s theorem would remove vacuum dynamics in spherical symmetry; the scalar field is what makes this a nonlinear collapse model rather than a disguised Schwarzschild calculation.

For example, differentiating the mass definition and using the reduced Einstein equations gives the exact flux identities

$$\begin{aligned} \partial_um&=-8\pi r^2\Omega^{-2} (\partial_vr)(\partial_u\phi)^2,\\ \partial_vm&=-8\pi r^2\Omega^{-2} (\partial_ur)(\partial_v\phi)^2. \end{aligned}$$

In the untrapped region $\partial_ur\lt0\lt\partial_vr$, they imply $\partial_um\le0$ and $\partial_vm\ge0$. The signs change roles once $\partial_vr$ becomes negative; one must not carry an exterior monotonicity claim unchanged into the trapped region. There is no explicit $G$ on the right because $m$ is physical mass; the flux law for the geometrized length $Gm$ has an extra factor of $G$.

Existence, dispersion, and black-hole formation

Christodoulou constructed a global bounded-variation theory for characteristic data on an outgoing null cone with a regular center. In that class the reduced equations have a unique maximal future development, and breakdown is tied to the center $r=0$ rather than an arbitrary coordinate failure Christodoulou 1993, main existence and continuation theorems, pp. 1131–1220.

The same framework separates two robust regimes.

Weak-data dispersion. If an invariant characteristic size is sufficiently small, no trapped sphere forms; the solution is future complete and approaches Minkowski behavior along the outgoing region. “Small” refers to the theorem’s scale-sensitive flux and variation bounds, not merely the pointwise amplitude of $\phi$.

Concentrated-data collapse. If enough mass is compressed between two spheres relative to their radius change, a later sphere satisfies $2Gm/r>1$. A representative criterion compares the mass gain $m(v_2)-m(v_1)$ with $r(v_2)$ and the relative annular thickness. The exact inequality and its constants are proved in Christodoulou 1991, §§5–8, pp. 339–373. Once a trapped sphere forms, the exterior contains a black-hole region in the completed global classification, and the singularity theorem supplies incomplete causal geodesics under its separate global hypotheses.

This is a large-data theorem in spherical symmetry. It is not a theorem that a single universal value of total mass separates dispersion from collapse: rescaling and the radial distribution of the incoming energy both matter.

Naked examples and their instability

The existence of smooth or controlled characteristic data whose central singularity sends signals to the regular region is not ruled out by the field equations. Christodoulou constructed continuously self-similar naked-singularity examples Christodoulou 1994, construction and main theorem, pp. 607–653. Their existence already shows why a naive statement—“reasonable matter never produces a visible singularity”—is false.

The censorship result is subtler. In Christodoulou’s bounded-variation characteristic data class, every data set producing a naked singular endpoint admits arbitrarily small prescribed perturbations that instead form a trapped surface before that endpoint becomes visible. The argument identifies a finite-dimensional family of perturbation directions and proves that the exceptional naked outcomes cannot occupy an open set in the relevant topology Christodoulou 1999, Theorems 1.1–1.3 and pp. 183–217. Liu and Li give a later direct stability-estimate version of the instability mechanism Liu–Li 2018, Theorem 1 and §§2–4.

The proof architecture is:

1. encode the initial scalar profile by an invariant characteristic quantity;
2. assume a first singular endpoint is visible and extract sharp asymptotics of $r$, $m$, and the scalar flux near it;
3. add a small perturbation whose integrated energy has a controlled sign;
4. propagate that excess by the monotonicity and Raychaudhuri equations;
5. verify $2Gm/r>1$ on a sphere lying to the past of the would-be naked endpoint.

The limitation is part of the theorem: perturbative instability in this spherical BV topology does not prove nonspherical stability of the black-hole outcome, nor does it settle what “generic” means in a weaker self-similar topology. A 2026 preprint claims nonlinear stability of a family of $C^{1,\alpha}$ continuously self-similar naked solutions within a rough spherically symmetric class Zheng 2026, main result and data-space definition. That is not a direct contradiction: the background family, admissible perturbations, norm, and regularity differ. The precise topology belongs in any censorship verdict.

Relation to vacuum formation

The scalar model supplies exact monotonicity through spherical symmetry and a positive matter flux. Vacuum gravitational-wave collapse has neither scalar matter nor a dynamical spherical sector; shear carries the focusing energy and the proof needs tensorial null structure and elliptic estimates. The scalar theorems therefore motivate but do not prove the vacuum result on the next page.

Likewise, formation of a trapped sphere is not a complete description of the singularity. Mass inflation, weak null boundaries, scalar curvature blowup, and $C^0$ or $C^2$ inextendibility ask different questions, developed in the next chapter.

Check the invariant threshold

Why can total initial Bondi mass alone not decide whether a spherical scalar pulse forms a trapped surface? Use the ratio $2Gm/r$ and the scaling symmetry to give two reasons.

Solution

Trapping is local in the quotient: it requires $2Gm(u,v)/r(u,v)>1$ on a particular sphere. The same total mass spread over a very large radial interval can keep this ratio small, whereas concentrating it into a small radius can make the ratio exceed one. Moreover, the massless Einstein–scalar equations admit a simultaneous length rescaling under which $m$ and $r$ scale together, leaving $2Gm/r$ unchanged while changing the numerical total mass. A valid criterion must therefore include concentration or a dimensionless flux profile, not a universal mass number.

References

• Christodoulou, Demetrios. “Bounded Variation Solutions of the Spherically Symmetric Einstein-Scalar Field Equations.” Communications on Pure and Applied Mathematics 46 (1993): 1131–1220. DOI.
• Christodoulou, Demetrios. “Examples of Naked Singularity Formation in the Gravitational Collapse of a Scalar Field.” Annals of Mathematics 140 (1994): 607–653. DOI.
• Christodoulou, Demetrios. “The Formation of Black Holes and Singularities in Spherically Symmetric Gravitational Collapse.” Communications on Pure and Applied Mathematics 44 (1991): 339–373. DOI.
• Christodoulou, Demetrios. “The Instability of Naked Singularities in the Gravitational Collapse of a Scalar Field.” Annals of Mathematics 149 (1999): 183–217. DOI.
• Liu, Jue, and Junbin Li. “A Robust Proof of the Instability of Naked Singularities of a Scalar Field in Spherical Symmetry.” Communications in Mathematical Physics 363 (2018): 561–578. DOI.
• Zheng, Weihao. “Nonlinear Stability of Continuously Self-Similar Naked Singularities for the Einstein–Scalar Field Equations I: Main Results.” arXiv:2605.16235 (2026). arXiv.