Past exam of the mathematics course of the University of Cambridge / 2013 / iii / Paper 51 / 3

The harmonic equation, including critical points. All derivatives in this argument use the three-dimensional Levi-Civita connection of $\gamma$. Write $U_i={\nabla }_{i}U$, $|\nabla U{|}^2={\gamma }^{ij}{U}_i{U}_j$ and ${\Delta }_{\gamma }U={\gamma }^{ij}{\nabla }_i{\nabla }_{j}U$. The curvature relation for the static vacuum conformal spatial metric gives $R=2|\nabla U{|}^2$. Substituting it into the contracted Bianchi identity yieldsThe Hessian of a scalar is symmetric, so the last terms cancel:Where $\nabla U\ne 0$, this implies ${\Delta }_{\gamma }U=0$. On the interior of the critical set $\{\nabla U=0\}$, $U$ is locally constant and its Laplacian is also zero. Every other critical point is a limit of noncritical points, so continuity of the Laplacian givesThis avoids incorrectly dividing by a gradient at its zeros.

Integration by parts and rigidity. The usual whole-space energy identity isWith standard static asymptotic falloff $U=O(R^{-1})$ and $\nabla U=O(R^{-2})$, the last term is $O(R^{-1})$ and vanishes. Positivity of the spatial Riemannian metric then makes $U$ constant, and its limiting value fixes that constant to zero.

There is also a direct integration by parts proof requiring only the stated vanishing of $U$ at infinity, rather than an assumed flux decay rate. For a regular value $\epsilon >0$, the region ${\Omega }_{\epsilon }=\{U>\epsilon \}$ has compact closure. Completeness and the absence of an inner boundary ensure no missing boundary pieces; standard asymptotic flatness and $U\to 0$ on all ends keep this positive level set away from infinity. Its outward unit normal is $n=-\nabla U/|\nabla U|$. Multiplying the harmonic equation by $U$ and integrating givesBoth terms are nonpositive, so both vanish. A nonempty component with $U>\epsilon$ cannot have zero gradient throughout and boundary value $\epsilon$. Thus ${\Omega }_{\epsilon }$ is empty. Choose arbitrarily small regular values and apply the same argument to $-U$; it follows thatThis is vanishing harmonic function by level-set integration. It also makes explicit why an inner boundary would invalidate the conclusion.

From zero Ricci curvature to Minkowski spacetime. In dimension three the Schouten tensor is $S_{ij}=R_{ij}-R{\gamma }_{ij}/4$, so it vanishes here. The three-dimensional curvature reconstruction then gives $R_{ijpq}=0$: $\gamma$ is a flat Riemannian manifold. This use of three-dimensional curvature from the Ricci tensor is essential; zero Ricci tensor would not by itself imply flatness in four dimensions.

A complete connected flat spatial manifold has Euclidean universal cover. Standard asymptotic flatness, with an ordinary Euclidean end, excludes a nontrivial free Euclidean quotient: a nontrivial fixed-point-free Euclidean isometry contains a translational or screw component, and its cyclic quotient has at most quadratic volume growth, incompatible with a three-dimensional Euclidean end. Extra identifications cannot restore cubic growth. Thus the complete spatial metric is globally Euclidean, not just locally flat. With $U=0$ and the standard global time coordinate the four-metric becomesTherefore horizonless static vacuum rigidity gives Minkowski spacetime as the sole solution under these hypotheses.

Allowing horizons. A black hole exterior can be static and asymptotically flat without being Minkowski. The Schwarzschild spacetime is the basic example. Its lapse function vanishes at the event horizon, so $U$ is unbounded below there; the horizon introduces an inner end or boundary in the spatial reduction, and the previous energy argument no longer has its hypotheses. In fact for Schwarzschild,so this conformal quotient terminates at $r=2M$ and is not the complete nonsingular quotient used above.

With the usual regularity, connected-horizon and global hypotheses of the static vacuum black-hole uniqueness theorem, the nontrivial black-hole exterior is Schwarzschild. Rotating Kerr black holes are stationary but not static and therefore are not alternatives in this question. This statement concerns a regular domain outside the horizon; the Schwarzschild interior contains a curvature singularity. If “globally static” and “nonsingular everywhere” are retained literally, Schwarzschild does not meet them. Allowing horizons means relaxing the horizonless global hypotheses to the corresponding static exterior problem.