No-Hair Theorems and Their Evasion

A no-hair theorem is an implication with a sharply delimited antecedent, not a ban on every field outside a black hole. The useful question is therefore diagnostic: for a proposed nonconstant field, which field equation, symmetry, regularity, topology, horizon, or asymptotic hypothesis falls outside the theorem? This page proves a representative minimally coupled scalar result, then traces two genuine evasions and one false counterexample. Existence of hair is kept separate from stability and from empirical preference.

Helpful background. Einstein Equations fix the gravitational and matter equations; Gravitational Actions and Boundary Terms are used when the scalar identity is integrated; Formulations of General Relativity clarify the domain of outer communications and symmetry reduction; Black-Hole Geometry and Mechanics provides Killing-horizon regularity; Black-Hole Perturbation Theory distinguishes existence from mode stability; and Relativistic Matter and Compact Objects supplies comparison cases with material surfaces.

A representative scalar no-hair statement

Consider four-dimensional Einstein gravity with one real, minimally coupled scalar,

$$S=\int d^4x\sqrt{-g}\left[ \frac{R}{16\pi G}-\frac{1}{2}\nabla_a\phi\nabla^a\phi-V(\phi) \right], \qquad \square\phi-V'(\phi)=0.$$

Assume all of the following hypotheses.

• The domain of outer communications is globally hyperbolic, static, asymptotically flat, and has a connected regular Killing horizon with compact cross-section.
• The scalar inherits the static symmetry, is $C^2$ in the exterior, is regular in a horizon-penetrating frame, and tends to a constant $\phi_\infty$ rapidly enough that the outer surface term vanishes.
• The horizon flux term is finite and vanishes in the limiting static slice. There is no inner material boundary or distributional source.
• The potential satisfies $V(\phi_\infty)=0$, as required for the stated asymptotically flat vacuum without a compensating cosmological constant, $V'(\phi_\infty)=0$, and

$$(\phi-\phi_\infty)V'(\phi)\geq0$$

throughout the exterior.

Under these conditions, the only solution is $\phi=\phi_\infty$. This is narrower than the stationary Einstein–Maxwell classification: the latter also needs dimension, asymptotic flatness, regularity, stationarity, horizon and topology conditions such as the $I^+$-regular framework collected in Chruściel, Costa, and Heusler 2012, §§ 2–3.3, pp. 5–18. It would be incorrect to transfer that electrovacuum uniqueness theorem unchanged to arbitrary scalar, vector, or non-Abelian matter.

The integral proof and its boundary terms

Take the spacetime slab between two static time slices, truncate it at a large timelike cylinder, and approach the future horizon by regular timelike surfaces. Multiply the scalar equation by $\phi-\phi_\infty$ and integrate by parts:

$$0=\int_{\partial\mathcal V} (\phi-\phi_\infty)n^a\nabla_a\phi\,d\Sigma -\int_{\mathcal V}\sqrt{-g}\, \left[\nabla_a\phi\nabla^a\phi +(\phi-\phi_\infty)V'(\phi)\right]d^4x.$$

The two time-slice contributions cancel by staticity. The outer term vanishes by the stated falloff. At the horizon, the static lapse tends to zero while the scalar and its freely falling derivative remain bounded, so the limiting flux vanishes. Those are substantive hypotheses; calling a coordinate component finite is not enough.

Because the scalar inherits the static Killing field, its gradient is spatial on a static slice and $\nabla_a\phi\nabla^a\phi\geq0$. The potential term is nonnegative by assumption. The remaining volume integral is therefore a sum of nonnegative functions with zero integral. Continuity forces $\nabla_a\phi=0$, and the asymptotic value fixes the constant. This proof architecture has four load-bearing pieces: a homogeneous scalar equation, symmetry inheritance, sign definiteness, and vanishing boundary terms.

The argument only establishes nonexistence within this theory class. It does not imply uniqueness of the metric, nonlinear stability, cosmic censorship, or formation from generic initial data.

Curvature coupling breaks the homogeneous equation

Set $V=0$ and add a shift-symmetric linear coupling $\alpha\phi\mathcal G$ to the Gauss–Bonnet invariant, with $\phi$ still canonically normalized as above. Thus $\phi$ has mass dimension one and $\alpha$ has mass dimension minus one. Rescaling to a dimensionless scalar $\varphi=\sqrt{4\pi G}\,\phi$ also rescales the coefficient in its scalar equation to $\widehat\alpha=\sqrt{4\pi G}\,\alpha$, of dimension length squared. The latter is the type of normalization used in the decoupling benchmark. In the canonical variables,

$$\square\phi+\alpha\mathcal G=0.$$

The same multiplication produces an extra bulk term $-\alpha(\phi-\phi_\infty)\mathcal G$, which has no required sign and cannot be discarded as a boundary contribution after its coefficient becomes a field. Equivalently, the radial shift current contains a piece that need not be proportional to $\phi'$. Thus the commonly used last step $J^r=\phi'F$ with finite nonzero $F$ fails for this operator. The exact current-based loophole is isolated in Sotiriou and Zhou 2014, § II.A–C, pp. 3–5, and their horizon expansion in Sotiriou and Zhou 2014, §§ III.C–D, pp. 5–7 shows that regularity selects secondary scalar hair and imposes a minimum-size condition.

This is an evasion, not a contradiction: the minimally coupled field equation hypothesis has changed. It is also not sufficient by itself to prove a physically acceptable black hole. A candidate must still possess a regular horizon, admissible asymptotics, a controlled effective-theory regime, and a stable perturbation spectrum. Sotiriou and Zhou explicitly distinguish failure of the no-hair argument from construction of the nonlinear solution.

Symmetry noninheritance opens another route

A complex massive scalar may have

$$\Phi=e^{-i\omega t+im\varphi}\,\psi(r,\theta)$$

while its stress tensor is stationary and axisymmetric. The field itself does not inherit the two Killing symmetries, so the positivity proof above does not apply. Regularity at a rotating horizon imposes the synchronization condition $\omega=m\Omega_H$, which removes flux through the generator $\chi=\partial_t+\Omega_H\partial_\varphi$. Kerr black holes with such scalar hair branch from the superradiant threshold; the field ansatz, synchronization, charges, and limiting families are given in Herdeiro and Radu 2014, Eqs. (1)–(3), pp. 1–4.

Again, the exact failed hypothesis is symmetry inheritance. The spacetime remains stationary, but the scalar is invariant only under the helical combination. This distinction also blocks a careless extension of a real-static-scalar theorem to all stationary stress tensors.

Apparent counterexamples that remain excluded

A formal nonconstant solution is not hair if the putative horizon is singular or absent. The Fisher–Janis–Newman–Winicour massless-scalar family is static and asymptotically flat, but its would-be horizon is a naked curvature singularity. The Bocharova–Bronnikov–Melnikov–Bekenstein configuration has a regular extremal metric but a conformally coupled scalar that diverges at the horizon. The first fails horizon existence and regularity; the second changes the coupling and fails the bounded-field hypothesis. Neither refutes the representative theorem.

Other theorem boundaries should be named rather than hidden:

Proposed hair	Exact hypothesis changed	Additional question still required
Scalar–Gauss–Bonnet profile	homogeneous minimal scalar equation and sign identity	regular branch, cutoff, perturbative stability
Synchronized complex scalar	scalar symmetry inheritance	superradiant and nonlinear stability
AdS condensate	asymptotic flatness and vanishing outer term	boundary ensemble and admissible falloff
Non-Abelian gauge hair	Abelian matter structure and uniqueness reduction	node sector and unstable modes
Singular scalar profile	horizon regularity	reject as a regular black hole

The broad classification and its known matter-dependent failures are emphasized in Chruściel, Costa, and Heusler 2012, §§ 1.1 and 5, pp. 3–4 and 27–30. Empirical likelihoods and numerical bounds remain outside this theorem question and are treated by Tests of Gravity.

Exercises

A static, asymptotically flat solution has a nonconstant real scalar satisfying $\square\phi=0$, with $\phi\to0$ at infinity. Near the candidate horizon, $\phi\sim q\ln(r-r_h)$. Is this a counterexample to the theorem? Then repeat the diagnosis after replacing the equation by $\square\phi+\alpha\mathcal G=0$ while requiring a regular scalar.

Solution

The logarithmic solution is not a counterexample: its scalar gradient and stress tensor diverge in a horizon-regular frame, so the horizon regularity hypothesis and the vanishing inner boundary term fail. With the Gauss–Bonnet source, a regular nonconstant profile can evade the proof because the integrated identity gains a sign-indefinite curvature term. One must still solve the horizon condition and check asymptotics, cutoff control, and stability; changing the equation removes the theorem but does not establish viability.

References

• Chruściel, Piotr T., João Lopes Costa, and Markus Heusler, “Stationary Black Holes: Uniqueness and Beyond,” Living Reviews in Relativity 15 (2012) 7. DOI. Open PDF.
• Herdeiro, Carlos A. R., and Eugen Radu, “Kerr Black Holes with Scalar Hair,” Physical Review Letters 112 (2014) 221101. DOI. Open PDF.
• Sotiriou, Thomas P., and Shuang-Yong Zhou, “Black Hole Hair in Generalized Scalar-Tensor Gravity: An Explicit Example,” Physical Review D 90 (2014) 124063. DOI. Open PDF.