Spring 2007

EE222: Solutions to Homework 2

1. Draw the phase portrait of a reaction-diusion system
x 1
x 2

= 2(x2 x1 ) + x1 (1 x21 )

= 2(x2 x1 ) + x2 (1 x22 ).

List the equilibria and their types. Does the system have limit cycles? (Hint: Show that this is a gradient
system.)

Solution. By solving x 1 = 0, x 2 = 0, we get that the system has three equilibria: p0 = (0, 0),
p1 = (1, 1) and p2 = (1, 1). Linearizing the system at the equilibria, it is easy to see that
p0 is a saddle, while p1 and p2 are stable nodes. The unstable and stable spaces of p0 are given
by: x1 = x2 (|x1 | < 1) and x1 + x2 = 0, respectively.
In fact, if we denote the vector eld of the system by f = (f1 , f2 )T , then
f2
f1
=2=
,
x2
x1
which implies that f is a gradient vector eld. By integrating f1 relative to x1 , dierentiating
the obtained function with respect to x2 and equating the result with f2 , we get that f =
grad V , where
1
1
V (x1 , x2 ) = (x21 + x22 ) + (x41 + x42 ) 2x1 x2 .
2
4
x = 2 (y x) + x (1 x2)
2
y = 2 (y x) + y (1 y )
4

4
4

0
x

Figure 1: The phase portrait of f in problem 1.

Therefore, by exercise 5. from Homework 1, the system has no closed orbits. (Alternatively,
one could observe that the divergence of f is 2 3(x21 + x22 ) and use Bendixsons theorem.)
1

The phase portrait of f is shown in Fig. 1.

2. Shown below are phase portraits of some vector elds in the plane. Determine which of them are correct
and which are incorrect. Modify the incorect ones, not by deleting any orbits, but by changing their stability
type or adding new orbits.

Solution. Phase portrait (a) is correct, but heres how the given trajectories can be extended.

(b)

(c) Let P be the region bounded by four saddle connections. We claim that P must contain
an eqilibrium. Since P is invariant, for any p P , (p) and (p) can be: (i) an equilibrium,
or: (ii) a limit cycle or: (iii) a nite connected union of saddle connections. If (ii), then
the region bunded by the limit cycle must contain an equilibrium. If (iii), apply the same
reasoning to the region Q bounded by these saddle connections (if Q = P , reverse the time).

3. Draw phase portraits of the following 1-dimensional systems as changes:

(a) x = 2 x x3
(b) x = 2 x + 2x3 x5 , for dierent .

Solution. (a) Let f (x) = 2 xx3 . The equilibria are x0 () = 0, x1 () = and x2 () = .

By looking at f (xi ()) for i = 0, 1, 2, we see that = 0 is the bifurcation value and obtain
the following bifurcation diagram (Fig. 2).
x1

x0
0
x2
Figure 2: Problem 3.(a)
(b) Let f (x) = 2 x + 2x3 x5 . The solutions to f (x) = 0 are: x1 () = 0, x2 (), x3 (),
x4 () = x2 () and x5 () = x3 (), where:



x2 () = + || 1 + , x3 () = || 1 + .
We also have:

f (x) = 2 + 6x2 5x4 .

To investigate stability of equilibria of x = f (x), we need to consider the following cases:

Case 1: < 1 (Fig. 3).
The only equilibrium is x1 () and since f (x1 ()) = 2 , it is stable for = 0. If = 0,
then f (x) = x5 , which also has a stable equilibrium at 0.
Case 2: 1 < < 0.
The equilibria are x1 (), . . . , x5 (), if 0, and only x1 () if < 0. By checking the
sign of f (xi ()), for i = 1, . . . 4, we obtain:
if 0, the only equilibrium x1 () = 0 is stable;
if > 0, then: x1 (), x2 (), x4 () are all stable, and x3 (), x5 () are unstable.
Case 3: = 1 or = 0 (Fig. 5).

Figure 3: Problem 2(b), Case 1.

x2
x3
x1

0
x5
x4
Figure 4: Problem 2(b), Case 2.

x2 () =

x2 = x3

x1

x1

x4 = x2

x4 = x5

Figure 5: Problem 2(b), Case 3, cases = 1 (left) and = 0 (right).

Case 4: > 0.
The only equilibria are x1 (), x2 () and x4 () = x2 (), for all R. By checking
the sign of f at these points we obtain that x1 () is unstable, while the other ones are
stable for all (Fig. 6).
x2

x2

x1
0

x4

x4
Figure 6: Problem 2(b), Case 4.

4. Let f be a smooth vector eld on the annulus

A = {(x, y) R2 : 1 x2 + y 2 4}.
Assume f points inward along the boundary of A, and that for every 0 2, the radial segment (in polar
coordinates)
S = {(r, ) : 1 r 2, = }
is a local cross section in the sense dened in class; that is, at every point p S , the angle between f (p)
and S is not zero.
(a) Let p S0 be an arbitrary point. Show that the orbit of f starting at p returns to S0 after some positive
time.
(b) Show that every continuous function : [1, 2] [1, 2] has a xed point.
(c) For p S0 , let (p) be the point of rst return to S0 of the orbit starting at p. Show that : S0 S0
has a xed point.
(d) Show that there exists a closed orbit of f in A.

Solution. (a) First of all, observe that since f points inward along the boundary of A, A is
positively invariant relative to the ow t of f . Let be the angular coordinate in the polar
coordinate system on A. Let p = (r, ) A be an arbitrary point and assume it belongs to
S , for some . Then by assumption that S is a local cross section for the ow t , we obtain
d
(t (r, )) = 0.
dt
Since both and t are smooth on A, the left hand side is either always > 0 or always < 0.
Assume the former. Since A is compact, dtd (t (r, )) has a minimum m > 0 on A. By the
Fundamental Theorem of Calculus,
(t (r, )) (r, ) mt
5

which becomes greater than 2 for suciently large t > 0. Therefore, the orbit of f starting
at p S0 returns to S0 after some positive time.
(b) If (1) = 1 or (2) = 2, there is nothing to prove. So assume (1) > 1 and (2) < 2,
and dene g(s) = (s) s. Then g is continuous, g(1) > 0, and g(2) < 0. Therefore, by the
Intermediate Value Theorem, there exists a point s0 (1, 2) such that g(s0 ) = 0. Clearly, s0
is a xed point for .
(c) Observe that by the Implicit Function Theorem is a smooth, hence continuous function.
Therefore, since S0 = [1, 2] {0} [1, 2], we can take = and use part (b).
(d) Let p S0 be a xed point of from (c). Then, by denition of , p = (p) = (p) for
some > 0, i.e., the orbit of p is closed.

5. Consider the following parametrized family of dierential equations:

x =

1 x + 2 y + x3 x2 y.

(a) What does the ow near (0, 0) look like when 1 = 2 = 0?

(b) What bifurcations occur at 1 = 0 and at 2 = 0 (for 1 < 0)?
(c) Use Bendixsons theorem and index theory to rule out parameter regions where there are no periodic
orbits.
(d) Conjecture the full bifurcation diagram.

Solution. (a) For the phase portrait when 1 = 2 = 0, see Fig. 7.

(b) The equilibria of the system are:

p0 () = (0, 0), p1 () = ( 1 , 0), p2 () = ( 1 , 0),

where = (1 , 2 ).

At p0 (), the eigenvalues of the linearization are [2 22 + 41 ]/2, so:
2

1 < 42 : if 2 > 0, then p0 () is a source (Fig. 8), and if 2 < 0, it is a sink (Fig. 9);
2

42 < 1 < 0: if 2 < 0, then p0 () is a stable node (Fig. 10), and if 2 > 0, then it is
an unstable node (Fig. 11);
if 1 > 0 and 2 R, then p0 () is a saddle.
At p1,2 (), the eigenvalues of the linearization are:

1 + 2 (1 + 2 )2 81
.
2
6

x=y
3
2
y=x yx
2

4
2

1
x

Figure 7: The phase portrait in problem 5 for 1 = 2 = 0.

x=y
3
2
y = m1 x + m2 y + x y x

m1 = 16
m2 = 2

8
8

0
x

Figure 8: The phase portrait in problem 5(b) for 1 < 22 /4 and 2 > 0.

x=y
3
2
y = m1 x + m2 y + x y x

m1 = 16
m2 = 2

8
8

0
x

Figure 9: The phase portrait in problem 5(b) for 1 < 22 /4 and 2 < 0.

x=y
3
2
y = m1 x + m2 y + x y x

m1 = 16
m2 = 10

8
8

0
x

Figure 10: The phase portrait in problem 5(b) for 2 /4 < 1 < 0 and 2 < 0.

x=y
3
2
y = m1 x + m2 y + x y x

m1 = 16
m2 = 10

8
8

0
x

Figure 11: The phase portrait in problem 5(b) for 2 /4 < 1 < 0 and 2 > 0.
Since 1 < 0, both p1 () and p2 () are saddles.

To summerize, at 1 = 0, when 2 > 0, two saddles ( , 0) and the unstable equlibirium point (0, 0) merges into the unstable equlibrium (0, 0). At 1 = 0, when 2 < 0, two

saddles ( , 0) and the stable equlibirium point (0, 0) merges into the unstable equlibrium
(0, 0). i.e. a pitchfork bifurcation appears.
At 2 = 0, when 1 < 0, the eigenvalue of the linearization at (0, 0) cross the jw-axis (i.e.
from Co when 2 < 0 to Co+ when 2 > 0) and not through the origin. In fact, a stable limit
cycle and an unstable equlibrium arise from a stable equilibrium, i.e. a (supercritical) Hopf
bifurcation appears.
(c) Since the divergence of f is div f = 2 x2 , we get that div f 0 for 2 0, so f has
no closed orbits when 2 0. It is not dicult to check that if 1 0, the only equilibrium,
namely, 0, is a saddle, hence has index 1, which guarantees nonexistence of closed orbits
(recall that the index of a smooth vector eld on the disk bounded by a closed orbit is 1).
Therefore, f has no closed orbits when 1 0 or 2 0.

6. Let f : R2 R2 be a dieomorphism; in other words, f is invertible, and both f and f 1 are smooth
maps. Let p0 , p1 , . . . , pN 1 be a period N cycle of f ; that is, pk = f k (p0 ) for 1 k N 1 and f N (p0 ) = p0 .
Prove that the matrices
Df N (pk )
have the same spectrum for k = 0, . . . , N 1.

Solution. By the Chain Rule,

Df N (pk ) = Ak1 Ak2 . . . A0 AN 1 AN 2 . . . Ak+1 Ak ,
where Ai = Df (pi ). Since f is a dieomorphism, each matrix Ai is invertible, so we get
N
Df N (pk1 ) = A1
k1 Df (pk )Ak1 .

Therefore, for all k = 1, 2, . . . , N, Df N (pk1) and Df N (pk ) are similar, thus have the same
spectrum.

10