Two analysis problems
2005.09.27
Mathematics

Problem 1: Consider a bounded domain U with C2 boundary. Let f denote a harmonic function in C2(U) that is continuous up to the boundary of U, which we denote by S. Let T be a subset of S, and suppose both f and the normal derivative of f vanish identically on T. Show that f is identically 0 on U.

Proof (for which I am indebted to Spyros): The case in 2 dimensions is obvious: consider the function grad f. It represents a complex analytic function on U. Continuity up to boundary guarantees an analytic extension to a larger bounded set V, which, by the Riemann mapping theorem has a holomorphic map to the unit disc. So, by the pull back, we can safely assume that we are dealing with a complex analytic function f on the unit disc that evaluates to zero on a line segment containing the origin, which implies that it has an accumulation of zeroes, and must equal zero throughout.

The case in higher dimensions uses the formulation of weak solutions. We extend U to a larger domain V by "protruding" a piece outside of T. Extend f to a function on V by setting f_V = f on U, 0 on VU Due to the fact that the normal derivative of f on T is 0, fV is a W1,2 function, and weakly satisfies the harmonicity condition on V. But it is know that a weakly harmonic function is also a strongly harmonic one, and we know that any harmonic function that is identically zero on an open subset must be identically zero. So f is zero on U.

Problem 2: Let {fn} denote a sequence of continuous functions on the interval [0,1] converging point-wise to a function f, which has L1 norm 0. Is f necessarily continuous? If not, does there exists points in [0,1] on which f is continuous?

Answer: f doesn't have to be continuous. Let fn be defined as follows: f_n = 1-nx on [0,1/n] and 0 otherwise They are a family of continuous functions, and point-wise converges to f = 1 at 0 and 0 otherwise which is not continuous.

The answer to the second part is affirmative, however, and here I give a proof to that fact.

We start by considering the following sets A_{N,epsilon} = {x; |f_m(x) - f_n(x)|leq epsilon, mgeq N, ngeq N} We notice that AN,ε are closed sets, and An,ε is contained in An+1,ε. Further more, for fixed ε, by point-wise convergence, each x in [0,1] belongs to some AN,ε. So we have that the union of all AN,ε is [0,1]. We can then apply the Baire Category Theorem to this countable collection and obtain that at least one of AN,ε must be second Baire category. Since AN,ε are closed, being in second Baire category is equivalent to saying that the set of interior points is non-empty.

Let VN,ε be the interior points of AN,ε. It is by definition open, and for each fixed ε there exists at least one N such that VN,ε is non-empty. I claim that for each fixed ε, Vε, the union over N of all VN,ε, is dense and open in [0,1]. The open property is obvious by countable union. It suffices to show density.

Suppose Vε is not dense in [0,1], then its complement in [0,1] is closed and contains an interval, which makes it second category. Let BN denote the intersection of VεC with AN,ε. Baire Category implies one of BN is second category, and contains an interior point. By definition, the interior point must be contained in Vε, contradicting the assumption.

Consider the family V1/k where k is a natural number. It is a countable family of open and dense subsets of [0,1], and by the Baire Category Theorem again we have that its intersection V is dense in [0,1].

We finish by showing that the function f must be continuous on V.

Let x be a point in V. For each ε0 > 0, we set ε = ε0/3. Choose k such that 1/k < ε. Then x also belongs to V1/k. Which implies that there exists some Nε(x) and an open neighborhood Bε(x) such that |f(y) - f_n(y)|leq frac{1}{k}, ngeq N_epsilon(x), yin B_epsilon(x) Choose some m > Nε(x). By continuity of fm, there exists a neighborhood U(x) s.t. |f_m(x) - f_m(y)| leq epsilon, yin U(x) Then, for y in the open neighborhood U(x) intersect Bε(x), we have |f(x)-f(y)| leq |f(x) - f_m(x)| + |f_m(x) - f_m(y)| + |f_m(y) - f(y)| < frac{2}{k} + epsilon leq epsilon_0 So f is continuous at x.

Posted at 16:43:07 EDT by W comment

blogCentralFront Page
2009.11.20 00:41:20 GMT Feynman's Messenger Lectures online Just found out something rather cool: Microsoft Research, through Project Tuva, is publishing videos of Richard Feynman's Messenger Lectures. Go watch.
2009.11.18 11:05:07 GMT Alcohol consumption Different cultures certainly have different views on alcohol. For example, at Hertford College Oxford, wine is allowed if reasonably drunk and 4) A small amount of beer or lager will be allowed wher
2009.11.16 19:17:31 GMT Luc visits; Willie doesn't check e-mail Holy cow! I just realized that I spent a day at work without checking e-mail! Okay, to be honest, today I was hosting Luc Nguyen, who we invited to speak on his work about the regularity near the sing
2009.11.15 18:19:32 GMT Chicken soup Chicken soup is not just good for the soul. It has been scientifically proven to mitigate inflammation. Maybe mommy's chicken soup was the reason that the same bug that took Pin out of commission for
2009.11.10 17:58:53 GMT Sayonara, e-nibbles; hullo, Gee-Mi-Ni It's final: e-nibbles is no more. e-nibbles was my trusty Dell D600 which I purchased summer after my Junior year in college through the Student Computer Initiative. Immediately after receiving the ob
2009.09.30 10:12:57 BST Ahhh! Cruft discovered in pre-print. Ack, I should've known better. I stayed up a bit later on Monday night than I intended to. I was asked, by Claude, last week, about whether certain cases (in particular the Born-Infeld model) not cove
2009.09.28 18:30:27 BST Spiders spiders everywhere Wow! Third post today, and here I thought I have been neglecting my blog. Anyway, it turns out that I am not the only person to have noticed the large number of spiders in Britain this autumn. Going o
2009.09.28 15:12:09 BST Causality of generalized wave-maps--paper on arXiv Oh, almost forgot. New paper on arXiv. Gary Gibbons showed via explicit computations using eigenvalues that the Skyrmion equation obeys the dominant energy condition. In my paper, I proved the dominan
2009.09.28 14:42:39 BST The evolution debate as an illustration of speciation I was reading some article or another in Wired, which happens to be about dinosaurs. And of course, the religious kooks came out of the woodwork to attack evolution on the comment board. And it occurr
2009.09.02 12:42:44 BST New beginnings: first days at Cambridge Heh. Did you, dear reader, notice the change on the date-stamp for the previous entry? It was posted in British Standard Time. Yes, I am now taking a position in the Department of Pure Mathematics and