A slice of Swiss cheese has one-dimensional holes;
a block of Swiss cheese has two-dimensional holes.
What is the dimension of a point? From classical geometry we have the definition “A point is that which has no parts” — also sprach Euclid. A point has dimension zero, a line has dimension one, a plane has dimension two, and so on.
Missing the Point
A point has dimension zero. But what if the point is not there? The real line is one-dimensional. It is a connected manifold: we can join any two points in
by a path. This is not true of the punctured line
: this manifold falls into two disconnected components,
and
; the points
and
cannot be connected by a path within
.
We define the unit -sphere in
as
.
is the usual sphere in 3-space,
is the unit circle in 2-space and
is the “0-sphere”, a pair of points on the real line. Within the real line
the 0-sphere can be continuously distorted, or shrunk, to a single point. The technical term for this shrinkability is homotopy: we say that
is homotopic to a point in
. Within the punctured line,
, the 0-sphere cannot be shrunk in this way because of the gap or hole. We say that there is a zero-dimensional hole in
.
In the plane , both the 0-sphere
and the 1-sphere (a circle) can be continuously distorted, or shrunk, to a single point. In the punctured plane,
, this is no longer true. The two points of the 0-sphere can be smoothly moved to a single point, but the 1-sphere cannot be shrunk to a point because it surrounds the missing point. We say that
has a one-dimensional hole, even thought only a single point is removed. Obviously, this continues:
has a 2-dimensional hole, since the unit 2-sphere cannot be shrunk to a point, whereas circles and point-pairs can.
A Hole is not there
We see that the removal of a single point from results in an
-dimensional hole. The point may be zero-dimensional in
, but the gap that it leaves in
is an
-dimensional hole. The crux is that, in the punctured space
, the hole is “not there!” We have to study the properties of the manifold
while remaining within
. This is done by defining algebraic stuctures that enable us to identify and quantify holes of various dimensions in the manifold. Let’s start with a simple example.
Holes usually comprise more than a single point. We compare two manifolds, the plane annulus and the spherical shell
.
is like a sheet of paper with a hole punched in it. A loop within
that surrounds the hole cannot be contracted to a point within
; there is a one-dimensonal hole in
. By contrast,
is more like a Swiss cheese. Loops in
can all be shrunk to a point, but a sphere surrounding the hole cannot be so shrunk: it is not homotopic to a point. A thin slice of Swiss cheese has one-dimensional holes; a block of Swiss cheese has two-dimensional holes.
A Little History
Bernhard Riemann noticed that holes in a surface could be counted by checking how many cuts could be made while the surface remains in one piece. For a bounded surface like a disk or an annulus, each cut begins and ends on a boundary. Clearly, the disk falls into two pieces once a single cut is made (it has no holes) but the annulus remains connected after a cut from the inner to the outer boundary circle is made.
These ideas were greatly developed by Henri Poincaré in his paper Analysis Situs (1895) and five follow-up papers. He introduced the idea of homology, which generalizes Riemann’s ideas to higher dimensions. Poincaré called the number of holes of dimension the
-th Betti number, after Italian mathematician Enrico Betti. The zeroth Betti number,
, indicates the number of pieces. For a connected manifold,
. The first Betti number is the number of one-dimensional holes. For the 2-sphere,
; for the torus,
. The second Betti number is the number of cavities in the manifold. For both the 2-sphere and the spherical shell,
. In general,
gives the number of
-dimensional holes in the manifold.
Poincaré showed that the Betti numbers are topological invariants. They do not change under distortions of a shape as long as it is not cut or broken. Therefore, shapes that have different sets of Betti numbers cannot be homeomorphic or topologically equivalent. For example, the numbers
for a sphere are
while for a torus they are
. Several other examples are shown in the table above. The invariance greatly helps topologists to categorize topological spaces.