110.

421 DYNAMICAL SYSTEMS

Week 7 Lecture 2 Notes

1. Invertible S 1 -maps.

Let’s return to maps on the circle, and try to gain more general information than by using
simply rigid rotations. Again, think of S 1 as the identification space S 1 = R/Z, given by the
level sets of the map
π ∶ R → S 1 , π(x) = [x] .
One easy way to think about [x] is to simply take any real number and disregard the integer
part. Thus [2.13] = .13, and [e] = e − 2. We will call π a projection of R onto S 1
Proposition 1. For any continuous map f ∶ S 1 → S 1 , there exist an associated map F ∶ R →
R, called a lift of f to R, where
f ○ π = π ○ F, equivalently f ([x]) = [F (x)] .

Some Notes:

● One way to see this is vis the commutative diagram

F
R ÐÐÐ→ R
× ×
π× ×π
× ×
Ö Ö
f
S 1 ÐÐÐ→ S1
.
● The lift F is unique up to an additive constant (sort of like how the anti-derivative
of a function is unique only up to an additive constant, right?)
● The quantity
deg(f ) = F (x + 1) − F (x)
is well-defined for all x ∈ R and is called the degree of f .
● If f is a homeomorphism, then ∣deg(f )∣ = 1.
● The structure of F is quite special. It looks like the sum of a periodic function with
the line y = (deg(f )) x. This is due to the structure of the projection π.

So just how much information about f can we learn by the study of the lifts of f ? Certainly,
maps on R are fairly easy to study. And maps with the structure of the lifts F may be easier
still. If we can use these lifts to say fairly general things about how an f may behave, this
would be quite important. For example, this quantity deg(f ) is defined solely by a choice
of lift f . We will see just what information deg(f ) conveys. For a moment, let’s first take a
look at why some of the assertions we just made are true.

Date: March 13, 2014.

1
2

● Lifts always exist. Simply construct one using the definition. I leave this as an
exercise.
● F is unique up to a constant.
Proof. Suppose F is another lift. Then
[F (x)] = f ([x]) = [F (x)] , ∀x ∈ R.
This is just another way of saying that π ○ F = f ○ π = π ○ F , ∀x ∈ R. But then F − F
ia always an integer! (why?) But F − F is the difference between two continuous
functions, and hence is continuous. But a continuous function on R that take values
in the integers is necessarily constant. 
● deg(f ) is well defined.
Proof. Here deg(f ) = F (x + 1) − F (x) is a continuous function on R that takes values
in the integers (it must, due to the projection π). Thus it also is a constant for all
x ∈ R. 
● If f is a homeomorphism, then ∣deg(f )∣ = 1.
Proof. Suppose that ∣deg(f )∣ > 1. Then ∣F (x + 1) − F (x)∣ > 1. And since F (x +
1) − F (x) is continuous, by the Intermediate Value Theorem, ∃y ∈ (x, x + 1) where
∣F (y) − F (x)∣ = 1. But then f ([y]) = f ([x]) for some y =/ x. Thus f cannot be
one-to-one and hence cannot be a homeomorphism.
Now suppose that ∣deg(f )∣ = 0. Then F (x + 1) = F (x), ∀x, and hence F is not
one-to-one on the interval (x, x + 1). But then neither is f , and again, f cannot be
a homeomorphism. 
● F (x) − xdeg(f ) is periodic.
Proof. It is certainly continuous (why?) To see that it is periodic (of period-1), simply
evaluate this function at x + 1:
F (x + 1) − (x + 1)deg(f ) = (F (x) + deg(f )) − (x + 1)deg(f )
= F (x) − xdeg(f ).

Example 2. Let f (x) = x. This is the “identity” map on S 1 , since all points are taken
to themselves. A suitable lift for f is the map F (x) = x on R. To see this, make sure the
definition works. Question: Are there any other lifts for f ? What about the map F (x) = x+a
for a a constant? Are there any restrictions on the constant a? The answer is yes. For a to
be an acceptable constant, we would need the definition of a lift of be satisfied. Thus
[F (x)] = [x + a] = f ([x]) = [x].
 3

So the condition that a must satisfy is [x + a] = [x] on S 1 . Hence, a ∈ Z. A new question:

For a real number a ∈/ Z, can F (x) = x + a serve as a lift of a circle map? What sort of circle
map?
Example 3. Let f (x) = xn . Thinking of x as the complex number x = e2πiθ , for θ ∈ R, then
n
f (x) = f (e2πiθ ) = (e2πiθ ) = e2πi(nθ) .
Hence a suitable lift is obviously F (x) = nx (I say obviously, since the variable in the exponent
is the lifted variable!) Question: This is a degree n map. For which values of n does the
map f have an inverse” And what does the map f actually do for different values of n?
Example 4. Let f be a general degree-r map. Then F (1) − F (0) = r = deg(f ). Suppose
that F (0) = 0. Then F (1) = r and if, for example, r > 1, it is now easy to see that there will
certainly be a y ∈ (0, 1), where F (y) = 1. This was a fact that we used in the proof above
to show that f cannot be a homeomorphism. In this case, where r > 1, at every point in
y ∈ (0, 1) where F (y) ∈ Z, we will have π ○ F (y) = [F (y)] = 0 on S 1 . This won’t happen when
r = 1. When r = 0, the map F will be periodic, which is definitely not one-to-one. Question:
What happens when r < 0? Draw some representative examples to see.
Definition 5. Suppose that f ∶ S 1 → S 1 is invertible. Then

(1) if deg(f ) = 1, f is orientation preserving.

(2) if deg(f ) = −1, f is orientation reversing.

Recall from Calculus III that orientation is a choice of direction in the parameterization
of a space (really, it exists outside of any choice of coordinates on a space, but once you put
coordinates on a space, you have essentially chosen an orientation for that space. This is
true at least for those spaces that actually are orientable, that is (Moebius Band?) On R, it
is the choice of direction for the symbol “>”. On a surface, it is a choice of side. In R3 , one
can use the Right Hand Rule. Etc. On S 1 , orientation preserving really means that after one
applies the map, points to the right of a designated point remain on that side. Orientation
reversing will flip a very small neighborhood of a point.
Circle maps may or may not have periodic points. And given an arbitrary homeomor-
phism, without regard to any other specific properties of the map, one would expect that we
can construct maps with lots of periodic points of any period. However, it turns out that
circle homeomorphisms are quite restricted. because they must be one-to-one and onto, only
certain things can happen. To explain, we will need another property of circle homeomor-
phisms to help us.
Proposition 6. Let f ∶ S 1 → S 1 be an orientation preserving homeomorphism, with F ∶ R →
R a lift. Then the quantity
F n (x) − x
ρ(F ) ∶= lim
∣n∣→∞ n
4

(1) exists ∀x ∈ R,
(2) is independent of the choice of x and is defined up to an additive integer, and
(3) is rational iff f has a periodic point.

Given these qualities, the additional quantity ρ(f ) = [ρ(F )] is well-defined and is called
the rotation number of f .
Some notes:

● ρ (Rα ) = α. (You should be able to actually calculate this directly using the defini-
tion.)
● ρ represents in a way the average rotation of points in a circle homeomorphism.
●
Proposition 7. If ρ(f ) = 0, then f has a fixed point.
Another way of saying that if there is no average rotation of the circle map, then
somewhere a point doesn’t move under f . This is like the Intermediate value Theorem
on a closed, bounded interval of R where a map is positive at one end point and
negative at the other.
● If f has a q-periodic point, then for a lift F , we have F q (x) = x+p for some p ∈ Z. For
example, let f = R 6 . Then a suitable lift for f would necessarily satisfy F 7 (x) = x+6,
7
∀x∈ R.
●
Proposition 8. Let f ∶ S 1 → S 1 be an orientation preserving homeomorphism. Then
all periodic points must have the same period.

This last point is quite restrictive. Essentially, if an orientation preserving homeomorphism

has a fixed point, it cannot have periodic points of any other period, say. Note that this is
not true of a orientation reversing map. For example, the map which flips the unit circle in
R2 across the y-axis, will fix the two points (0, 1) and (0, −1), while every other point is of
order two.
Exercise 1. For Rα ∶ S 1 → S 1 a circle rotation, show ρ(Rα ) = α.
Exercise 2. Show that any lift of the rotation R 6 must satisfy F 7 (x) = x + 6, ∀x∈ R, and
7
explicitly construct two such lifts.

This is enough for circle homeomorphisms for now. And ends our work in Chapter 4.
There is a great section on frequency locking on page 141. Look it over at your leisure. We
won’t work through it in the course, but it is very interesting. Dynamically, it represents a
situation where a linear flow on the torus (with its uncoupled ODEs) becomes the limiting
system to a system of coupled ODEs, representing a nonlinear flow. Question: For this to
be the case, must the resulting linear flow on the torus be a rational flow?