Notes by Willie Wong (wongwwy -at- member.ams.org). Comments and suggestions welcome.
Statement: $(M,g)$ globally hyperbolic smooth, then Cauchy problem for the linear wave equation has unique global solutions. What happens if the metric is not smooth? Colombini, Spagnola, Lerner: there exists $g \in \cap_{\alpha\in (0,1)} C^{0,\alpha} \setminus C^{0,1}$ such that there are no solutions of the linear wave equations in the distributional sense.
If $g$ is Lipschitz, the wave equation always have unique local solutions. For "global" question, a statement depends on what is meant by "globally hyperbolic".
In $C^2$, we introduce normal coordinates. ($C^{1,1}$ you can solve geodesic equation, but the coordinates may not be as good; in $C^2$ you can use implicit function.)
So for rough metrics, the standard causality theory based on normal coordinates and local behaviour need to be modified.
Einstein equation $(\Sigma, g, K)$ where $g\in H^{s+1}$ and $K\in H^{s}$ data. Solve vacuum equations.
Theorem: MGHE exists if $s = 2$. (Paper[s] on arXiv, 2011; partially joint work with James Grant)
Focus on part 2. for which we have Theorem: (a) all standard statements of Causality (cf. arxiv:11006706) not concerning geodesics and normal coordinates are true for $C^{0,1}$ metrics (b) several standard facts become non-facts for general $C^{0,\alpha}$ with $\alpha < 1$.
Main Idea: sandwich $g$ between sequences of smooth metrics and see how much one can retain under convergence (in the topology of uniform convergence on compact sets).
Sandwich in the sense of Causal sandwich; consider partial ordering on the space of metrics by containment of the cone of future timelike vectors. So the main question: if $\hat{g}_i \searrow g$ and $\check{g}_i \nearrow g$, is
A point $p$ is said to be time-like relative to $q$ if there exists a differentiable time-like curve connecting the two. (Differentiable can be replaced with Lipschitz with a.e. derivative timelike with no loss under assumption of $C^{0,1}$ metrics. Since when limits are taken of time-like curves we want them to fall inside the class of causal curvest; we don't need the reverse.)
A causal curve is a Lipschitz curve whose derivative (which is known to exists a.e.) is future directed causal where it exists.
We have the obvious inclusions (where all sets defined in some small compact neighborhood of a point)
The question is how much of these can be made to be equality. Claim: for Lipschitz metrics, the two $\subseteq$ can be replaced by $=$.
Example:
The surface $u = 0$ is null. The other null is given by $\frac{du}{dx} = |u|^\lambda$. For $\lambda < 1$ you get bifurcation of the ODE solution. You get what is called a "causal bubble", where the difference of $\mathcal{J}$ and $\mathcal{I}$ has, in some sense positive measure.
Joint with Igor Rodnianski.
Recall Penrose-Hawking theorem. (Assume Cosmological space-times, so the topology of spatial slice assume to be $\mathbb{T}^3$.) Strong energy condition + trace of the second fundamental form being uniformly bounded above by $-C$ then we have a singularity.
Fundamental question: is geodesic incompleteness in Penrose-Hawking type theorems a real singularity (e.g. curvature blow-up) or something else (e.g. Cauchy horizon)?
Study Einstein-Stiff-fluid and Einstein -scalar-field. Stiff-fluid has $p = \rho$ in fluid. Scalar field embeds into irrotational stiff-fluid under the assumption that the scalar field has a time-like gradient.
(Generalised) Kasner solutions
with $p = \bar{p}t^{-2}$ and $u^\mu = \delta_0^\mu$. Restrictions from constraints $\sum q_i = 1$ and $\sum q_i^2 = 1 - 2 \bar{p}$. FLRW is isotrpic, so $q = 1/3$.
Study small perturbations of FLRW (some smallness of the deviation of $q$ is assumed; c.f. "rotation of Kasner axes") and see what happens when solving backwards to the big-bang singularity.
Results:
On the BKL problem, it is true in symmetry classes (esp. Bianchi IX by Ringstrom). There are counterexamples: Taub, spikes [Weaver-Rendall, Lim] stiff fluid/scalar-field.)
Closely related results:
Theorem (Rodnianski-Speck): Consider near-FLRW (in Sobolev sense) data for the Einstein-stiff-fluid on three-torus. We have
Main ideas of the proof:
Toy problem:
and Toy FLRW: $\tilde{\phi} = \ln t$ as the background.
Global Cauchy problem under $T^2$ symmetry, Gowdy, or plane. Metrics with weak regularity.
Include: interacting impulsive waes, matter spacetimes with shock waves.
Challenges: define Einstein equations in a weak sense (delta singularities) and existence theory for such weak equations.
(Can't type fast enough...)
Joint with Eichmair, Lee, and Schoen
Theorem: $3 \leq n < 8$, $(M^n,g,k)$ AF, DEC, ... then $E \geq |P|$.
Schoen-Yau proved the time-symmetric case in the same dimensions; also showed rigidity: if $E = 0$ we have Euclidean. ('79, '81, '87)
Schoen-Yau '81 proved the space-time case in dimension 3 $E \geq 0$ with rigidity: if $E = 0$ then isometric embeddable in Minkowski space. Eichmair generalised to $3 \leq n < 8$.
Witten '81 $n = 3$ using spinors. AF DEC gives $E \geq |P|$.
Bartnik '87 and others $n \geq 3$ spin.
Equality case by Beig-Chrusciel (96) Chrusciel-Maerten ('98) for spin.
Riemannian PMT:
The proof of the space-time case is similar. Main difficulties are in steps 1 and 3.
For the space-time theorem requires more than just the hamiltonian constraint. Step 1 is more complicated as such. Step 2 follows from previous works of Eichmair and Metzger. In Step 3 we cannot "minimise" since mass is not variational in the non-symmetric context.
$E < |P|$ gives MOTS barriers. (Step 2)
Joint work with C. Lubbe
Einstein-Euler with cosmological constant. Use $(+---)$ signature. Recall FRW cosmologies. Remark: a perfect fluid model is FRW iff the fluid is shear-, rotation-, and acceleration-free.
A class of equations of states were shown to have future stability for FLRW with de Sitter like cosmological constant by Rod-Speck; personal failure excludes radiation EOS $p = \frac13 \rho$.
Theorem: small perturbations in radiation cosmologies with spatial curvature $+1$ lead to future global existence, future geodescially complete, and remains close to the background FRW.
Remark: spatial curvature $-1,0$ can be treated with bit more technical detail.
Use conformal rescalings: radiation case is conformal, $\mathrm{tr} T = 0$. Denote conformal weight by $\Theta$. $\implies$ Friedrich equations.
arXiv: 1111.4691 [gr-qc]
Joint work with Stefan Wenger, Dan Lee, Sajjad Lakzian.
If $M_i$ are compact oriented Riemannian manifolds, $d_F$ the intrinsic flat distance is zero iff there exists an orientation preserving isometry between them. Based on Federer-Flemming (Annals 1960) + Ambrosio-Kirchhein (spelling? Acta 2000).
Definition:
$d_{Flat}$ measures a filling volume between the two sets. (See Federer, Geometric Measure Theory)