28 2. GEOMETRIC PROPERTIES
PROOF. Choose z 1 , z 2 EE\ B(z 0 , r) and let
2 z - zo
f(z ) = r I
12
+ zo.
z - zo
Then f (z 1 ) , f (z 2 ) E f (E) n B(z 0 , r ). Next condition 1° implies that these points
can be joined by a connected set I in f(E) n B(z 0 , er). Hence 1-^1 (1) joins z 1 and
z 2 in E \ B(z 0 , r / c) and condition 2° holds. D
The connection cited above between linear local connectivity and the three-
point condition yields the following reformulation of Theorem 2.2.5.
COROLLARY 2.4.3 (Walker [165]). A Jordan domain D is a K-quasidisk if and
only if 8D is linearly locally connected with constant c, where K and c depend only
on each other.
In contrast to the three-point condit ion, the property of linear local connectivity
can b e applied to a sdoma in as well as to its boundary. For example, a sector D of
angle a is linearly locally connected with c = csc(a/ 2).
The converse of the Jordan curve theorem implies that a simply connected
domain D is a Jordan domain if and only if D is locally connected at ea ch point of
- 2
R (Newman [140]).
We have the following counterpart of this r esult for quasidisks.
THEOREM 2.4.4 (Gehring [49]). A simply connected domain D is a K -quasidisk
if and only if it is linearly locally connected with constant c, where K and c depend
only on each other.
Moreover D is a disk or ha lf-plane if and only if it is linearly locally connected
with constant c = l. See Langmeyer [ 110 ].
Finally we show that in general, t he notion of linear local connectivity is invari-
ant with respect to Mi:ibius transformations. Our proof depends on the following
result.- 2
LEMMA 2.4.5. Suppose that 1 < c < oo and that E 1 , E 2 c R are sets which
are separated by an annulus
{ z : a < I z - zo I < ac}.
If f is a Mobius transformation, then f(E 1 ) and f(E2) are separated by an annulus
{w: b < lw - wol < bg(c)}
where g(c) = c^112 + c^112 - l. This bound is sharp.
- 2
PROOF. By means of a preliminary similarity mapping we may assume that
a = 1 and z 0 = 0. Let
C1 = { z : lz l = 1}, C2 = { z : lz l = c}.
Since the inversion i(z) = c/z interchanges C 1 and C 2 , we m ay also assume that
(2.4.6) diamf(C1) ~ diamf(C2).
This implies that f(Ci) is a circle and j(C 2 ) is a circle or line. Hence there exists
a line L' through the center of f(Ci) which contains the center of f(C 2 ) if f(C 2 )
is a circle or is orthogonal to f(C 2 ) if f(C 2 ) is a line. Now L = f-^1 (L') is a line
through the origin which is orthogonal to the conce ntric ci rcles C 1 and C 2.