ScalingStacks

6. Unique Continuation [04UC]

Original official author HTML, exact retained edition. Historical TeX conversion verdicts remain unchanged. Cited-edition alignment and mathematical self-containment are not assessed.

Complete original source context · Original author HTML

6. Unique Continuation

For our proof of unique continuation we rely on the following classical result:

Theorem 6.1.

Assume that Ω⊂ℝn\Omega\subset\mathbb{R}^{n} is a connected open set and u∈Wl​o​c1,2​(Ω)u\in W^{1,2}_{loc}(\Omega) is a weak solution to the equation

∂i(ai​j​(x)​uj)+bi​(x)​ui+c⁡(x)​u=0,\partial_{i}(a^{ij}(x)u_{j})+b^{i}(x)u_{i}+c(x)u=0,

where ai​j​(x)a^{ij}(x) is Lipschitz and uniformly elliptic and bi​(x),c​(x)b^{i}(x),c(x) are bounded measurable. If u=0u=0 on some open subset of Ω\Omega, then u≡0u\equiv 0 in Ω\Omega.

A proof can be found in Hörmander’s book [H], Theorem 17.2.617.2.6. In [AS], the authors use the same theorem to prove unique continuation for fully nonlinear uniformly elliptic equations.

We will apply this result to the difference of uu and vv, which solves a linear equation where uu and vv are sufficiently regular. Indeed, suppose uu and vv are C2C^{2} in a neighborhood of xx and let wtw_{t} be the convex combination t​u+(1−t)​vtu+(1-t)v. Let (Wt)i​j(W_{t})^{ij} be the matrix of cofactors for D2​wtD^{2}w_{t}. Then by expanding 0=∫01dd​t​detD2​wt​𝑑t0=\int_{0}^{1}\frac{d}{dt}\det D^{2}w_{t}dt we get

ai​j​(x)​(u−v)i​j=0,a^{ij}(x)(u-v)_{ij}=0,

where

ai​j​(x)=∫01(Wt)i​j​(x)​𝑑t.a^{ij}(x)=\int_{0}^{1}(W_{t})^{ij}(x)dt.

The regularity theory of Caffarelli [C2] allows us to use this observation at points of strict convexity for solutions to the Monge-Ampère equation:

Theorem 6.2.

Assume

detD2​u=f in ​Ω,u|∂Ω=0\det D^{2}u=f\quad\text{ in }\Omega,\quad\quad u|_{\partial\Omega}=0

where f∈C1,α​(Ω)f\in C^{1,\alpha}(\Omega) is strictly positive. Then

u∈C3,α​(Ω).u\in C^{3,\alpha}(\Omega).

The proof of unique continuation follows easily from these observations and our main theorem.

Proof of Theorem 1.3:.

Let Σu\Sigma_{u} and Σv\Sigma_{v} be the singular sets of uu and vv respectively, and let A=Ω−(Σu∪Σv)A=\Omega-(\Sigma_{u}\cup\Sigma_{v}). Since AA is dense in Ω\Omega, it suffices to show that u=vu=v on AA.

By Caffarelli’s theory ([C1]), AA is an open set, and by Theorem 1.1, AA is connected. By Theorem 6.2, the difference u−vu-v satisfies the linear equation

ai​j​(x)​(u−v)i​j=0a^{ij}(x)(u-v)_{ij}=0

on AA, where ai​ja^{ij} are locally uniformly elliptic and C1,αC^{1,\alpha} in AA. The conclusion follows from Theorem 6.1. ∎

Original mathematics by the credited authors. Source-backed reader collection; mathematical self-containment is not assessed.