Isometries of the space
02 Dec 2016
Let’s recall what distance is.
You define a norm on a vector space
that induces a metric. Note that metric does not require the underlying
set to be a vector space, nor does it have to come from a norm. In this note,
however, we will focus on
finite-dimensional normed vector spaces.
In particular, we want to find isometries of
the Banach space .
The symmetry group of the space—known as
the Euclidean group—consisting of isometries of the Euclidean space,
has as subgroups the translation group and the orthogonal group .
In other words, translations, rotations, and reflections
preserve the Euclidean distance.
What is the symmetry group of the space?
Translations of course preserve the metric, so
should be there. What else?
Notice that the symmetry group of an space has the symmetry group
of its unit ball as a subgroup. The symmetry group of the unit ball
(i.e., the diamond) consists of signed permutations of its vertices.
That group is known known as the Hyperoctahedral group.
There are no other symmetries, so we may conclude that only
translations and signed permutations preserve the metric.
If you don’t like handwavy arguments and
would rather prefer a more technical proof,
check out this very nice post on isometries of vector -norms
and read the forum discussion.