27.08.2014 Views

Real and Complex Analysis (Rudin)

Real and Complex Analysis (Rudin)

Real and Complex Analysis (Rudin)

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

HARMONIC FUNCTIONS 235<br />

Note: This theorem provides the solution of a boundary value problem (the<br />

Dirichlet problem): A continuous functionfis given on T <strong>and</strong> it is required to find<br />

a harmonic function F in U "whose boundary values are f." The theorem exhibits<br />

a solution, by means of the Poisson integral of f, <strong>and</strong> it states the relation<br />

between f <strong>and</strong> F more precisely. The uniqueness theorem which corresponds to<br />

this existence theorem is contained in the following result.<br />

11.9 Theorem Suppose u is a continuous real function on the closed unit disc 0,<br />

<strong>and</strong> suppose u is harmonic in U. Then (in U) u is the Poisson integral of its<br />

restriction to T, <strong>and</strong> u is the real part of the holomorphic function<br />

1 I" e il + z .<br />

f(z) = - -.- u(e") dt<br />

2n -" e" - z<br />

(z E U). (1)<br />

PROOF Theorem 10.7 shows that f E H(U). If Ul = Re f, then (1) shows that<br />

Ul is the Poisson integral of the boundary values of u, <strong>and</strong> the theorem will<br />

be proved as soon as we show that u = u 1•<br />

Put h = u - Ul. Then h is continuous on 0 (apply Theorem 11.8 to u 1), h<br />

is harmonic in U, <strong>and</strong> h = 0 at all points of T. Assume (this will lead to a<br />

contradiction) that h(zo) > 0 for some Zo E U. Fix E so that 0 < E < h(zo), <strong>and</strong><br />

define<br />

g(z) = h(z) + Ei Z 12 (z EO). (2)<br />

Then g(zo) ~ h(zo) > E. Since g E qO) <strong>and</strong> since g = E at all points of T,<br />

there exists a point z 1 E U at which g has a local maximum. This implies that<br />

gxx ~ 0 <strong>and</strong> gyy ~ 0 at Zl. But (2) shows that the Laplacian of g is 4E > 0, <strong>and</strong><br />

we have a contradiction.<br />

Thus u - Ul ~ O. The same argument shows that U 1 - u ~ O. Hence u =<br />

u 1, <strong>and</strong> the proof is complete. / / / /<br />

11.10 So far we have considered only the unit disc U = D(O; 1). It is clear that<br />

the preceding work can be carried over to arbitrary circular discs, by a simple<br />

change of variables. Hence we shall merely summarize some of the results:<br />

If u is a continuous real function on the boundary of the disc D(a; R) <strong>and</strong> if u<br />

is defined in D(a; R) by the Poisson integral<br />

·9 1 I" R2 - r2 .<br />

u(a + re' ) = -2 2 2R (0 ) 2 u(a + Re'~ dt<br />

n -" R - r cos - t + r<br />

then u is continuous on D(a; R) <strong>and</strong> harmonic in D(a; R).<br />

If u is harmonic (<strong>and</strong> real) in an open set n <strong>and</strong> if D(a; R) c n, then u<br />

satisfies (1) in D(a; R) <strong>and</strong> there is a hoi om orphic function f defined in D(a; R)<br />

whose real part is u. This f is uniquely defined, up to a pure imaginary additive<br />

constant. For if two functions, holomorphic in the same region, have the same<br />

real part, their difference must be constant (a corollary of the open mapping<br />

theorem, or the Cauchy-Riemann equations).<br />

(1)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!