Interior Free Boundary Problems for the Laplace Equation ANDREW ACKER
Communicated by J. C. C. NITSCHE 1. Introduction and main results
Our results pertain to the following general free boundary problem with hydrodynamic applications. Problem 1. Given are a bounded, simply-connected region G in the plane ~2 (with closure C, and boundary 0G, the latter being a simple, closed curve) and a positive, continuous, real function a(p) defined in G. We define 1[M [[ = S~a2 (P) dx dy M
for any measurable set M e G . Given the value A, 0 < A < IIGIh we seek a simple closed curve F c G such that, if U(p) denotes the harmonic measure of •G in the doubly-connected region f2 bounded by F w 3G, then (1)
I1#11 = A
and
(2)
IVU(p)l=c.a(p)
on s for some c>O
(i.e., for each pes IVU(q)l-*c.a(p) as q--*p, qef2).
.r2: ~72U(p)=O
j,---- a6: u = 1
Fig. l
Archive for Rational Mechanics and Analysis, Volume 75, 9 by Springer-Verlag 1981
158
A. ACKER
Our existence result is Theorem 1. Assume that the set {(p,z)eG x N: 0 2 . a ( p ) + ( 1 - 2 ) . a ( q )
whenever p, qeG and 0 < 2 < 1 .
Then there exists at least one closed, convex curve F c G which satisfies (1) and (2). A number of authors have treated free boundary problems for the Laplace equation, corresponding to such physical phenomena as waves, wakes, jets, and cavitation. (See, for example, BEURLING [6], DANILJUK [8] and [9], FRIEDRICHS [10], GARABEDIAN, LEWY, & SCHIFFER [11], LAVRENT'EV [12], SERRIN [14], TEPPER [16] and WEINSTEIN [17].) However, our main purpose is to exhibit a novel method for the existence proof, which may even prove to be computationally useful (see Remark2). Moreover, the present results do not appear to follow from the literature (see Remark 3). The proof of Theorem 1 is based on the following free boundary optimization problem involving capacity. Problem 2. Let X be the set of all simple, closed curves F c G, and let X c denote the subset of X containing only convex curves. To a given F ~ X , we assign the capacity K--S[VU(p)] [dpl, 7
where U(p) is the harmonic measure of ~?G in the doubly-connected region f2 bounded by F uOG, and ~c(2 is an equip otential curve of U. (The notation (J(p),O,K A, etc. refers to other curves F , F , FA, etc. in X.) For any value 0 < A < HG II, we seek a curve Fa which minimizes capacity in Xc(A)= {FeX~: 1If2II =A}. Theorem 1 is a corollary of the following theorem. Theorem 2. (a) If G is convex, then Problem 2 has at least one solution for each value 0 < A < ]IGL (b) If the set {(p,z)eG x N: O
Free Boundary Problems
159
Remark 2. The author knows of no uniqueness result applicable to Problem 1 or 2 in the general case where {(p,z)EG x R : O
Remark 3. In a more frequently studied variation of Problem 1, one seeks, for any given c > 0 , a curve FeX such that
IVU(p)l=c.a(p)
(4)
on F.
Apparently, the most general existence results for this problem in the literature are due to BEURLING [-6] and DANILJUK [-9]. DANILJUK's existence result is incorrect, as the author has recently shown in [-18]. It is easy to compare Theorem 1 with BEURLING'S results in the case where G={p~R2: [pl
2. Proof of Theorem 2 (a) For At(0, IIGll) fixed, let k = i n f { K : FsXc(A)}>O, and let (F.)cXc(A) be a sequence such that lim K , = k. It sflffices to find a subsequence ( ~ ) c (F,) and a n~oo
curve
PeX~(A) such
that
lim F~= F
(i.e., any neighborhood of/~ contains ~ for all sufficiently large n~N), since then we have /( = lim K , = k. n~oc
To define (~) and F, notice that there exists a value p > 0 so small that for each n~N, G\O, contains a disc of radius 2p. Passing to a subsequence, we can assume (for fixed po~G) that
G\f2,~Bo(Po) = {P dR2
:lp-p0l
for all n~N. By applying the theorem of Ascoli-Arzel/l to the polar coordinate representations of these curves relative to Po, we see that a further subsequence
160
A. ACKER
(~) converges to a closed convex curve /~ ~ (~, whose interior complement has area (]IGH-A). Although we omit the details, one can show that i f / ~ r then /s as n--*o% in violation of our choice of (F,). Therefore f'eX~(A), and our proof is complete.
3. Area-preserving boundary perturbations which (to first order) diminish capacity We assume throughout this section that {(p, z)eG x IR: 0 <_z <_a(p)} is convex in IR3. In order to define the class of boundary perturbation transformations T~: X c ~ X r 0=
[?[=Sa(p)[dp[. .y