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.

ANALYTIC CONTINUATION 329<br />

We claim that Q is afundamental domain ofr. This means that statements (a)<br />

<strong>and</strong> (b) of the following theorem are true.<br />

16.19 Theorem Let r<strong>and</strong> Q be as above.<br />

(a) If qJl <strong>and</strong> qJ2 E r<strong>and</strong> qJl #: qJ2' then qJl(Q) n qJ2(Q) = 0·<br />

(b) UtperqJ(Q)=IJ+.<br />

(c) r contains all transformations qJ E G of the form<br />

qJ(Z) = az + b<br />

cz + d<br />

(1)<br />

for which a <strong>and</strong> d are odd integers, b<strong>and</strong> c are even.<br />

PROOF Let r 1 be the set of all qJ E G described in (c). It is easily verified that<br />

r 1 is a subgroup of G. Since (1 E r 1 <strong>and</strong> T E r 1, it follows that r c r l' To<br />

show that r = r 1, i.e., to prove (c), it is enough to prove that (a/) <strong>and</strong> (b)<br />

hold, where (a/) is the statement obtained from (a) by replacing r by r l' For<br />

if (a/) <strong>and</strong> (b) hold, it is clear that r cannot be a proper subset of r l'<br />

We shall need the relation<br />

1m Z<br />

1m qJ(z) = I cz + d 12 (2)<br />

which is valid for every qJ E G given by (1). The proof of (2) is a matter of<br />

straightforward computation, <strong>and</strong> depends on the relation ad - be = 1.<br />

We now prove (a/). Suppose qJl <strong>and</strong> qJ2 E r 1, qJl #: qJ2' <strong>and</strong> define qJ =<br />

qJl 1 0 qJ2' If Z E qJl(Q) n qJ2(Q), then qJl 1 (z) E Q n qJ(Q). It is therefore<br />

enough to show that<br />

Q n qJ(Q) = 0 (3)<br />

if qJ E r 1 <strong>and</strong> qJ is not the identity transformation.<br />

The proof of (3) splits into three cases.<br />

If c = 0 in (1), then ad = 1, <strong>and</strong> since a <strong>and</strong> d are integers, we have<br />

a = d = ± 1. Hence qJ(z) = Z + 2n for some integer n #: 0, <strong>and</strong> the description<br />

of Q makes it evident that (3) holds.<br />

If c = 2d, then c = ± 2 <strong>and</strong> d = ± 1 (since ad - be = 1). Therefore<br />

qJ(z) = u(z) + 2m, where m is an integer. Since (1(Q) c D(t; !), (3) holds.<br />

If c#:O <strong>and</strong> c #: 2d, we claim that I cz + d I > 1 for all Z E Q. Otherwise,<br />

the disc D( -die; 1/1 c I) would intersect Q. The description of Q shows that if<br />

IX #: -t is a real number <strong>and</strong> if D(IX; r) intersects Q, then at least one of the<br />

points -1, 0, 1 lies in D(IX; r). Hence lew + d I < 1, for w = -1 or 0 or 1. But<br />

for these w, cw + d is an odd integer whose absolute value cannot be less<br />

than 1. So I cz + d I > 1, <strong>and</strong> it now follows from (2) that 1m qJ(z) < 1m Z for

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

Saved successfully!

Ooh no, something went wrong!