The rotation number is invariant under
topological conjugacy, and even monotone topological
semiconjugacy: if and are two homeomorphisms of the circle and : h\circ f = g\circ h for a monotone continuous map of the circle into itself (not necessarily homeomorphic) then and have the same rotation numbers. It was used by Poincaré and
Arnaud Denjoy for topological classification of homeomorphisms of the circle. There are two distinct possibilities. • The rotation number of is a
rational number (in the lowest terms). Then has a
periodic orbit, every periodic orbit has period , and the order of the points on each such orbit coincides with the order of the points for a rotation by . Moreover, every forward orbit of converges to a periodic orbit. The same is true for
backward orbits, corresponding to iterations of , but the limiting periodic orbits in forward and backward directions may be different. • The rotation number of is an
irrational number . Then has no periodic orbits (this follows immediately by considering a periodic point of ). There are two subcases. :# There exists a dense orbit. In this case is topologically conjugate to the
irrational rotation by the angle and all orbits are
dense. Denjoy proved that this possibility is always realized when is twice continuously differentiable. :# There exists a
Cantor set invariant under . Then is a unique minimal set and the orbits of all points both in forward and backward direction converge to . In this case, is semiconjugate to the irrational rotation by , and the semiconjugating map of degree 1 is constant on components of the complement of . The rotation number is
continuous when viewed as a map from the group of homeomorphisms (with topology) of the circle into the circle. ==See also==