To illustrate the geometry, we
begin with the simplest elements in H(Rn), namely the one point sets. In the 2002 REU, John Mayberry (then at the University of
California at Fullerton, now at the University of Southern California), Audrey Powers (then at Agnes Scott College,
now at Emory University) and I examined the subspace of H(Rn) consisting of the
one-point sets. We completely classified the elements in H(R n) that satisfy the three versions
of the triangle equality in this subspace. The next theorem (Theorem 6 from [7])
provides the
classification.
Theorem 2: Let a, b be
points in Rn. Let
A = {a}, B = {b} and C be elements of H(Rn). Then C
satisfies ABC, ACB, or CAB
if and only if there exist
positive real numbers s and t so that
1. C
is a subset of (A+s) ∩ (B+t) and
2. there is a point c0 in
C with c0 in the intersection of ∂(A+s)
∩ ∂(B+t) with the Euclidean line through a and
b.
This theorem is illustrated by the
next applet. Any compact set in the shaded region that contains the black point
c0 on the ab axis is an element on the Hausdorff line
defined by A and B. The points a and b are directly
moveable as is the point c0, by moving the black point on the
slider.
One interesting consequence of
Theorem 2 is that there many be many elements at a particular location on a
Hausdorff line. For example, if the point c0 described in
Theorem 2 is on the Euclidean segment defined by a and b, note
that the intersection of A+s with B+t is the single
point set C = {c0}. In this case, there is exactly one
point at this location on the Hausdorff line defined by A and B.
However, if c0 lies
outside the Euclidean interval defined by a and b, then the
intersection of A+s with B+t is a disk with
positive radius. In this case, there are many different compact sets contained
in the intersection of A+s with B+t. The union of
any such set with {c0} lies s units from A and t
units from B. We will say that these sets lie on the same location
Hausdorff line defined by A and B.
Definition: Let A and B be distinct elements of H(Rn) and let r =
h(A, B). Two elements C1 and
C2 of H(Rn)
on the Hausdorff line L(A, B), through A and B lie at
the same location on L(A, B) if
h(A,
C1) = h(A, C2) and h(B,
C1) = h(B, C2).
Unlike Euclidean lines which have
unique points at each location, Theorem 2 tells us that Hausdorff lines may
contain an infinite number or a finite number of elements at a given location.
We will see more about this later.