Black-Hole Hair in Scalar–Tensor Gravity

A scalar field outside a stationary black hole is not forbidden merely because the spacetime is asymptotically flat and the horizon is regular. A no-hair result follows only after specifying the action, the symmetry inherited by the scalar, the asymptotic branch, and the regularity condition used at the horizon. This page proves a representative shift-symmetric theorem, then shows exactly why a scalar–Gauss–Bonnet coupling lies outside one of its hypotheses.

Required background. Einstein Equations supplies the field equations, Killing horizons, and curvature identities used in the argument.

Helpful background. Gravitational Actions and Boundary Terms supplies the variational treatment of derivative couplings; Formulations of General Relativity supplies the distinction between coordinate singularities and invariant horizon regularity; Gravity Beyond General Relativity supplies the mode and branch checks that must follow an existence result; Relativistic Matter and Compact Objects supplies the regular-center and matching conditions used when matter surrounds the hole.

A current-based no-hair statement

Consider a four-dimensional, asymptotically flat, static and spherically symmetric black hole,

$$ds^2=-h(r)\,dt^2+\frac{dr^2}{f(r)}+r^2d\Omega^2,$$

with a connected, nonextremal Killing horizon at $r=r_H$. Let the scalar share the spacetime symmetries, $\phi=\phi(r)$, and approach a constant at infinity. Suppose the scalar action is invariant under $\phi\mapsto\phi+c$, so its equation is a conserved-current equation,

$$\nabla_aJ^a=0.$$

The representative theorem is deliberately narrow:

If the metric and the invariant current norm are regular at the horizon, the radial current has the factorized form $J^r=f\phi'\mathcal F$, and $\mathcal F$ is finite, continuous, and nonzero on the branch connecting the horizon to the canonical weak-field vacuum, then $\phi$ is constant throughout the exterior.

The factorization hypothesis holds for the analytic shift-symmetric Galileon/Horndeski branch considered by Hui and Nicolis when the canonical kinetic term remains present. Their current proof and its asymptotic step are given in Hui and Nicolis 2013, equations (4)–(12). Broader uniqueness and no-hair theorems use different hypotheses—often an integral identity, a sign condition on a potential, or an energy condition—and must not be substituted for this statement without rechecking their assumptions; see Chruściel, Costa, and Heusler 2012, §§ 2–3.3 and § 6.

Proof by horizon regularity

Spherical symmetry removes the angular current components. Staticity and time-reversal invariance of this configuration also set $J^t=0$, leaving only $J^r$. Current conservation gives

$$0=\frac{1}{r^2\sqrt{h/f}} \frac{d}{dr}\left(r^2\sqrt{\frac{h}{f}}J^r\right),$$

and therefore

$$r^2\sqrt{\frac{h}{f}}J^r=C.$$

The invariant norm of this purely radial current is

$$J_aJ^a=g_{rr}(J^r)^2 =\frac{C^2}{r^4h}.$$

Because $h\to0$ at the Killing horizon, a nonzero $C$ makes this scalar invariant diverge. Regularity therefore fixes $C=0$, so $J^r=0$ everywhere in the connected exterior.

Now use the assumed branch structure,

$$0=J^r=f\phi'\mathcal F.$$

Away from the horizon $f\gt0$, while $\mathcal F$ neither vanishes nor diverges on the branch. Hence $\phi'=0$. Continuity extends the result to the horizon, and the asymptotic constant fixes the same value throughout the exterior. The proof uses no observational bound and says nothing yet about perturbative stability; it is a local mode-and-boundary argument.

There are two independent checks. First, the integration constant has dimension of $r^2J^r$, and the norm reconstructed from it is coordinate invariant. Second, in the canonical theory $J^r=-f\phi'$ up to the declared current sign, so $J^r=0$ immediately yields the constant-scalar general-relativistic branch.

Where an evasion enters

Every genuine evasion must identify a failed hypothesis rather than merely display a nonconstant profile.

Curvature-sourced shift symmetry. Consider the normalization

$$S=\frac{1}{16\pi G}\int d^4x\sqrt{-g} \left[ R-\frac12(\nabla\phi)^2+\alpha\phi\mathcal G \right],$$

where

$$\mathcal G =R_{abcd}R^{abcd}-4R_{ab}R^{ab}+R^2$$

is the Gauss–Bonnet density. A constant shift changes the action only by a topological boundary term, but the corresponding local current contains a curvature contribution that does not vanish with $\phi'$. Equivalently, its Horndeski representation uses a nonanalytic function $G_5\propto-\alpha\ln\lvert X\rvert$. Thus the step $J^r=f\phi'\mathcal F$ with regular, nonzero $\mathcal F$ fails. The evasion and the resulting regular hairy solutions are developed in Sotiriou and Zhou 2014, §§ II–III, equations (9)–(27), PDF.

The decoupling calculation makes the failed step explicit. In this paragraph $M=GM_{\mathrm{phys}}$ is the geometrized mass, a length, while $\alpha$ has dimensions of length squared and $\phi$ is dimensionless. On Schwarzschild, $f=h=1-2M/r$ and $\mathcal G=48M^2/r^6$. To first order in $\alpha$, the scalar equation is

$$\Box\phi=-\alpha\mathcal G, \qquad \frac{d}{dr}\left(r^2f\phi'\right) =-\frac{48\alpha M^2}{r^4}.$$

One integration gives

$$r^2f\phi'=\frac{16\alpha M^2}{r^3}+C.$$

Regularity at $r=2M$ now fixes $C=-2\alpha/M$, not $C=0$, because the curvature source supplies the other horizon term. At infinity,

$$\phi=\phi_\infty+\frac{2\alpha}{Mr}+O(r^{-2}).$$

The expansion requires $\lvert\alpha\rvert/M^2\ll1$: the scalar is first order in this ratio and its metric backreaction starts at second order. The scalar charge is therefore secondary: at this perturbative order it is fixed by $M$ and $\alpha$, not freely specifiable. Taking $\alpha\to0$ removes the profile and recovers the controlled general-relativistic limit. The perturbative solution and absence of first-order metric backreaction appear in Sotiriou and Zhou 2014, § III.E, equations (43)–(49), PDF.

Other failed hypotheses. A scalar of the form $\phi=qt+\psi(r)$ can have a stationary stress tensor while failing symmetry inheritance. A potential or a non-shift-symmetric matter coupling replaces the conserved-current equation by a sourced equation. A branch on which $\mathcal F=0$, an extremal or singular horizon, de Sitter rather than flat asymptotics, or nonanalytic derivative couplings also requires a separate argument. Each case may admit hair, but none contradicts the theorem just proved.

Existence is not viability

A regular background solution is only the beginning of the analysis. One must still count perturbative modes, determine the signs of their kinetic and gradient terms, establish hyperbolicity, locate the strong-coupling cutoff, and impose boundary conditions for the physical perturbations. Some formal profiles cross a singular effective metric or sit on a branch whose quadratic kinetic coefficient vanishes. Such a profile is not a healthy counterexample to a no-hair theorem.

Conversely, rejecting every hairy solution because one branch is unstable is also invalid. Stability is background- and parameter-dependent. The precise claim established here is only that the current proof excludes hair on its regular analytic branch and that scalar–Gauss–Bonnet coupling bypasses the factorization step.

Common pitfalls

“Shift symmetric” automatically means the theorem applies. A symmetry up to a topological boundary term can alter the local current. Its analytic structure must be inspected.

A divergent coordinate derivative proves a singular scalar. Regularity is tested with scalars and horizon-regular coordinates. Here the current norm, not $\phi'$ by itself, fixes the integration constant.

Any time-dependent scalar makes the black hole time dependent. Derivative interactions can make the stress tensor stationary even when the scalar does not inherit the Killing symmetry. That is a different hypothesis class.

Exercises

Reproduce the current proof for a static, spherical exterior. Then identify the exact failed step for the linear scalar–Gauss–Bonnet coupling and compute the leading scalar charge on a Schwarzschild background in the normalization used above.

Solution

Current conservation integrates to

$$r^2\sqrt{\frac{h}{f}}J^r=C.$$

Since $J_aJ^a=C^2/(r^4h)$, regularity where $h=0$ requires $C=0$. For a factorized analytic current $J^r=f\phi'\mathcal F$ with $f\gt0$ and $\mathcal F\ne0$ outside the horizon, this implies $\phi'=0$.

For $\alpha\phi\mathcal G$, the current includes a curvature term independent of $\phi'$, so the factorization step is unavailable. The scalar equation on Schwarzschild is

$$\frac{d}{dr}(r^2f\phi')=-\frac{48\alpha M^2}{r^4}.$$

Thus

$$r^2f\phi'=\frac{16\alpha M^2}{r^3}+C.$$

Cancellation of the numerator at $r=2M$ gives $C=-2\alpha/M$. Expanding at large radius yields $\phi=\phi_\infty+2\alpha/(Mr)+O(r^{-2})$, so the leading charge is $2\alpha/M$. It vanishes continuously with $\alpha$, as required by the general-relativistic limit. Choosing any other $C$ creates a horizon singularity and is an apparent, not regular, evasion.

References

• Chruściel, Piotr T., João Lopes Costa, and Markus Heusler (2012), “Stationary Black Holes: Uniqueness and Beyond,” Living Reviews in Relativity 15, 7. DOI.
• Hui, Lam, and Alberto Nicolis (2013), “No-Hair Theorem for the Galileon,” Physical Review Letters 110, 241104. DOI; Open PDF.
• Sotiriou, Thomas P., and Shuang-Yong Zhou (2014), “Black Hole Hair in Generalized Scalar-Tensor Gravity: An Explicit Example,” Physical Review D 90, 124063. DOI; Open PDF.