Phase Portraits for Planar Systems

Given the linearity principle from the previous chapter, we may now compute the general solution of any planar system. There is a seemingly endless number of distinct cases, but we will see that these represent in the simplest possihie form nearly all of the types of solutions we will encounter in the higher dimensional case.

3.1 Real Distinct Eigenvalues

Consider Xl = AX and suppose that A has two real eigenvalues)", < A2. Assuming for the moment that Ai "" 0, there are. three cases to consider:

1. Al < 0 < A2;

2. Al <)..2 < 0;

3. 0 < Al < 1..2.

We give a specific example of each case; any system that falls into anyone of these three categories may be handled in a similar manner, as we show later.

Example. (Saddle) First consider the simple system X' = AX where

A= (~l ~2)

39

40 Chapter 3 Phase Portraits for Planar Systems

with Al < 0 < A2. This can be solved immediately since the system decouples into two unrelated first-order equations:

Xl = AIX yl = A2Y.

We already know how to solve these equations, but, having in mind what comes later, let's find the eigenvalues and eigenvectors. The characteristic equation is

so Al and A2 are the eigenvalues. An eigenvector corresponding to Al is (1,0) and to A:2 is (0, 1). Hence we find the general solution

Since Al < 0, the straight-line solutions of the form cu"lf(l, 0) lie on the x-axis and tend to (0,0) as t -+ 00. This axis is called the stable line. Since A2 > 0, the sol utions fJ eA2 t (0, I) lie on the y-axis and tend away from (0, 0) as

t _,. 00; this axis is the unstable line. All other solutions (with a, fJ ~ 0) tend to 00 in the direction of the unstable line, as t _,. 00, since X (t) comes closer and closer to (0, fJeA2t) as t increases. In backward time, these solutions tend"

to 00 in the direction of the stable line. •

In Figure 3.1 we have plotted the phase portrait of this system. The phase portrait is a picture of a collection of representative solution curves of the

Figure 3.1 Saddle phase portrait for x' '" -x,

y' = y.

3.1 Real Distinct Eigenvalues 41

system in ]R2, which we call the phase plane. The equilibrium point of a system of this type (eigenvalues satisfying A1 < 0 < A2) is called a saddle.

For a slightly more complicated example of this type, consider Xl = AX where

As we saw in Chapter 2, the eigenvalues of A are ±2. The eigenvector associated to A = 2 is the vector (3, 1); the eigenvector associated to A = -2 is (1, -1). Hence we have an unstable line that contains straight-line solutions of the form

each of which tends away from the origin as t -+ 00. The stable line contains the straight-line solutions

which tend toward the origin as t -+ 00. By the linearity principle, any other solution assumes the form

for some a, {3. Note that, if a -# 0, as t -+ 00, we have

X(t) ~ ae2t (~) = XI(t)

whereas, if {3 i=- 0, as t -+ -00,

Thus, as time increases, the typical solution approaches Xl (t) while, as time decreases, this solution tends toward X2(t), just as in the previous case. Figure 3.2 displays this phase portrait.

In the general case where A has a positive and negative eigenvalue, we always find a similar stable and unstable line on which solutions tend toward or away

42 Chapter 3 Phase Portraits for Planar Systems

Figure 3.2 Saddle phase portrait for x' = X + 3y,

yl = X- y.

from the origin. All other solutions approach the unstable line as t _,. 00, and tend toward the stable line as t _,. -00.

Example. (Sink) Now consider the case X' = AX where

but A1 < ,1".2 < O. As above we find two straight-line solutions and then the general solution:

Unlike the saddle case, now all solutions tend to (0,0) as t _,. 00. The question is: How do they approach the origin? To answer this, we compute the slope dyldx of a solution with fJ =1= o. We write

x(t) = ae>"lt y(t) = fJe).,2t

and. compute

dy = dyldt = A2fJe)..2f = A2fJ e()..2-i'l)t

dx dxl dt ,1".1 ae}q f Al a

Since ,1".2 - Al > 0, it follows that these slopes approach ±oo (provided fJ i=- 0). Thus these solutions tend to the origin tangentially to the y-axis. •

3 .. 1. Real Distinct Eigenvalues 43

(a)

(b)

Figure 3.3 Phase portraits for (a) a sink and (b) a source.

Since Al < ,1.2 < 0, we call Al the stronger eigenvalue and ,1.2 the weaker eigenvalue. The reason for this in this particular case is that the x-coordinates of solutions tend to 0 much more quickly than the y-coordinates. This accounts for why solutions (except those on the line corresponding to the Al eigenvector) tend to "hug" the straight-line solution corresponding to the weaker eigenvalue as they approach the origin.

The phase portrait for this system is displayed in Figure 3.. 3 a, In this case the equilibrium point is called a sink.

More generally, if the system has eigenvalues A I < ,1.2 < 0 with eigenvectors (Uj, U2) and (VI, V2), respectively, then the general solution is

The slope of this solution is given by

dy Al cuA I t U2 + ,1..2/3 eA2 t l'2

dx AICUA]tuI+A2fJeAztvl

(Al(UA1t U2 + A2fJ~.tV2) e~A2t

= Alae'\jtUj + A2fJeAztvl e"":A1t

Alae(AJ~Al)tu2 + AzfJl'2

= Alae('\I~Al)t UI + .A2fJVI '

which tends to the slope l'2/VI of the ,1.2 eigenvector, unless we have f3 = O. If f3 = 0, our solution is the straight-line solution corresponding to the eigenvalue A 1. Hence all solutions (except those on the straight line corresponding

44 Chapter 3 Phase Portraits for Planar Systems

to the stronger eigenvalue) tend to the origin tangentially to the straight-line solution corresponding to the weaker eigenvalue in this case as well.

Example. (Source) When the matrix

satisfies 0 < A2 < AI, our vector field may be regarded as the negative of the previous example. The general solution and phase portrait remain the same, except that all solutions now tend away from (0,0) along the same paths. See Figure 3.3b. •

Now one may argue that we are presenting examples here that are much too simple. While this is true, we will soon see that any system of differential equations whose matrix has real distinct eigenvalues can be manipulated into the above special forms by changing coordinates.

Finally, a special case occurs if one of the eigenvalues is equal to O. As we have seen, there is a straight-line of equilibrium points in this case. If the other eigenvalue A is nonzero, then the sign of A determines whether the other solutions tend toward or away from these equilibria (see Exercises 10 and 11 at the end of this chapter).

3.2 Complex Eigenvalues

It may happen that the roots of the characteristic polynomial are complex numbers. In analogy with the rea] case, we call these roots complex eigenvalues. When the matrix A has complex eigenvalues, we no longer have straight line solutions. However, we can still derive the general solution as before by using a few tricks involving complex numbers and functions. The following examples indicate the general procedure.

Example. (Center) Consider X' = AX with

A = (0 /3)

-/3 0

and {3 i= O. The characteristic polynomial is A 2 + /32 = 0, so the eigenvalues are now the imaginary numbers ±i{3. Without worrying about the resulting complex vectors, we react just as before to find the eigenvector corresponding

3.2 Complex Eigenvalues 45

to A. = ifJ. We therefore solve

(-if3 fJ) (x) (0)

-fJ -ifJ Y = 0

or ifJx = fJy, since the second equation is redundant. Thus we find a complex eigenvector (1, i), and so the function

X(t) = eifJt C)

is a complex solution of X' = AX.

Now in general it is not polite to hand someone a complex solution to a real system of differential equations, but we can remedy this with the help of Euler's formula

eifJt = cos fJ t + i sin fJ t,

Using this fact, we rewrite the solution a~

X(t) = ( cos Bt + isinfJt ) = ( cos{3t + i sin fJt ).

i(cos fJt + i sin fJt) - sin fJt + i cos Bt

Better yet, by breaking X( t) into its real and imaginary parts, we have

XU) = XRe(t) + iXlm(t)

where

( cosfJt ) (Sin,Bt)

XRe(t) = . fJ ., Xlm(t) = . fJ .. ' .

- SIn t cos t

But now we see that both XRe( t) and Xlm (t) are (reall) solutions of the original system. To see this, we simply check

XRe(t) + iX{m(t) = X'(t) = AX(t)

= A(XRe(t) + iXlm(t» = AXRe + iAXlm(t).

46 Chapter 3 Phase Portraits for Planar Systems

Equating the real and imaginary parts of this equation yields XRe = AXRe and X{m = AXlm, which shows that both are indeed solutions. Moreover, since

the linear combination of these solutions

where Cl and C2 are arbitrary constants provides a solution to any initial value problem.

We claim that this is the general solution of this equation. To prove this, we need to show that these are the only solutions of this equation. Suppose that this is not the case. Let

be another solution. Consider the complexfunction j'(r) = Cu{t) + iv(t»ei,Bt. Differentiating this expression and using the fact that Y (t) is a solution of the equation yields f'(t) = 0. Hence u(t) + iv(t) is a complex constant times e-i,Bt. From this it follows directly that yet) is a linear combination of XRe(t) and Xlm(t).

Note that each of these solutions is a periodic function with period 2JT / f3.

Indeed, the phase portrait shows that all solutions lie on circles centered at the origin. These circles are traversed in the clockwise direction if f3 > 0, counterclockwise if f3 < O. See Figure 3.4. This type of system is called a center. •

Figure 3.4 Phase portrait for a center.

3.3 Repeated Eigenvalues 47

Example. (Spiral Sink, Spiral Source) More generally, consider X' = AX where

A:::: ( a f3)

-/3 a

and a, f3 ¥= O. The characteristic equation is now A 2 - 2a)" + a2 + f32> so the eigenvalues are A = a ± i{3. An eigenvector associated to a + if3 is determined by the equation

(a - (a + i(3»x + f3r = 0 ..

Thus (1, i) is again an eigenvector. Hence we have complex solutions of the form

X(t) = e(o+i.B)t (!)

at ( cosf3t) . at (. S.i.nf3t)

= e + te

-sinf3t cosf3t

= XRe(t) + iXlm(t)·

As above, both XRe(t) and Xlm(t) yield teal solutions of the system whose initial conditions are linearly independent. Thus we find the general. solution

r:t:t ( c.osPt) . e<t (Sinf3t)

X(t) = cle . + C2e •

- sin f3 t cos f3 t

Without the term e();t, these solutions would wind periodically around circles centered at the origin. The eat term converts solutions into spirals that either spiral into the origin (when a < 0) or away from the origin (a > 0), In these cases the equilibrium point is called a spiral sink or spiral sourc~ respectively. See Figure 3.5. •

3.3 Repeated Eigenvalues

The only remaining cases occur when A has repeated real eigenvalues. One simple case occurs when A is a diagonal matrix of the form

A=(~ ~).

48 Chapter 3 Phase Portraits for Planar Systems

Figure 3.5 Phase portraits for a spiral sink. and a spiral source.

The eigenvalues of A are both equal to A. In this case every nonzero vector is an eigenvector since

AV=AV

for any V E ]R2. Hence solutions are of the form

Each such solution lies on a straight line through (0,0)' and either tends to (0,0) (in < 0) or away from (0,0) (if A > 0). So this is an easy case.

A more interesting case occurs when

A = (~ ~).

Again both eigenvalues are equal to A, but now there is only one linearly independent eigenvector given by (1,0). Hence we have one straight-line solution

At (1)

Xl(t) = ae O.

To find other solutions, note that the system can be written

X'=AX+Y y'=Ay.

3.4 Changing Coordinates 49

Thus, if Y # 0, we must have

r(t) = (3eAt•

Therefore the differential equation for x(t) reads

This is a nonautonomous, first-order differential equation for x(t). One might first expect solutions of the form d't, but the nonautonomous term is also in this form .. As you perhaps saw in calculus, the best option is to guess a solution of the form

for some constants a and J).. This technique is often called "the method of undetermined coefficients." Inserting this guess into the differential equation shows that J). = {3 while a is arbitrary. Hence the solution of the system may be written

This is in fact the general solution (see Exercise 12).

Note that, if,\ < 0, each term in this solution tends to 0 as t ---? 00. This is dear for the aeAt and j3eAt terms. For the term {3teM this is an immediate consequence of l'Hopital's rule. Hence all solutions tend to (0, 0) as t ---? 00. When ,\ > 0, all solutions tend away from (0,0). See Figure 3.6. In fact, solutions tend toward or away from the origin in a direction tangent to the eigenvector (1,0) (see Exercise 7).

3.4 Changing Coordinates

Despite differences in the associated phase portraits, we really have dealt with only three types of matrices in these past three sections:

where ),_ may equal J). in the first case.

Any 2 x 2 matrix that is in one of these three forms is said to be in canonical form. Systems in this form may seem rather special, but they are not. Given

50 Chapter 3 Phase Portraits for Planar Systems

Figure 3.6 Phase portrait for a system with repeated negative eigenva,lues.

any linear system Xl = AX, we can always "change coordinates" so that the new system's coefficient matrix is in canonical form and hence easily solved. Here is how to do this.

A linear map (or linear transformation) on JR.2 is a function T : JR.2 ---+ JR2 of the form

T (X) = (ax + bY).

Y cx+ dy

That is, T simply multiplies any vector by the 2 x 2 matrix:

We will thus think of the linear map and its matrix as being interchangeable, so that we also write

We hope no confusion will result from this slight imprecision.

Now suppose that T is invertible. This means that the matrix T has an inverse matrix S that satisfies TS = ST = I where I is the 2 x 2 identity matrix. It is traditional to denote the inverse of a matrix T by T~ 1. As is easily checked, the matrix

1 (d -b)

S = det T -c a

3.4 Changing Coordinates 51

serves as T-1 if det T f. O. If det T = 0, then we know from Chapter 2 that there are infinitely many vectors (x, y) for which

T(;)=(~).

Hence there is no inverse matrix in this case, for we would need

for each such vector. We have shown:

P.roposition. The 2 x 2 matrix T is invertible ifand only if det T f. O. •

Now, instead of considering a linear system X' = AX, suppose we consider a different system

for some invertible linear map T. Note that if Y (t) is a solution of this new system, then X(t) = net) solves X' = AX. Indeed, we have

(TY(t))' = TY'(t)

= T(T-'AT)Y(t) = A(TY(t»

as required. That is, the linear map T converts solutions of Y'= (y-l AT) Y to solutions of X' = AX. Alternatively, y-l takes solutions of X' = AX to solutions of Y' = (T-I A T) Y.

We therefore think of T as a change of coordinates that converts a given linear system into one whose coefficient matrix is different. What we hope to be able to do is find a linear map T that converts the given system into a system of the form y' = (T-1AT)Y that is easily solved. And, as you may have guessed, we can always do this by finding a linear map that converts a given linear system to one in canonical form.

Example. (R ea I E 1 9 enva I ues) Suppose the matrix A has two real, distinct eigenvalues Aj and A2 with associated eigenvectors VI and V2• Let T be the matrix whose columns are V] and V2. Thus TEj = Vj for j = 1,2 where the

52 Chapter 3 Phase Portraits for Planar Systems

Ej form the standard basis of lR_2 • Also, T~ 1 Vj = Ej. Therefore we have

(T~IAT)Ej = T-1AVj = T~lp'jVj) = AjT~IVj

= AjEj.

Thus the matrix T-1 AT assumes the canonical form

and the corresponding system is easy to solve.

•

Example. As a further specific example, suppose

(~l

A=

1

The characteristic equation is A 2 + 3A + 2, which yields eigenvalues A =~ 1 and A = ~ 2. An eigenvector corresponding to A = ~ 1 is given by solving

(A + I) (;) = (i -i) (;) = (~)

which yields an eigenvector (1, I). Similarly an eigenvector associated to A = ~2 is given by (0,1).

We therefore have a pair of straight ~ line solutions, each tending to the origin as t --+ 00. The straight-line solution corresponding to the weaker eigenvalue lies along the line y = x; the straight-line solution corresponding to the stronger eigenvalue lies on the y-axis. All other solutions tend to the origin tangentially to the line y= x.

To put this system in canonical form, we choose T to be the matrix whose columns are these eigenvectors:

T= G i)

so that

T-1 = ( I 0) ~l 1 .

3.4 Changing Coordinates 53

Finally, we compute

so y-l A Y is in canonical form. The general solution of the system Y' (T-1 AT)Y is

so the general solution of X' = AX is

TY(t) = G ~) (ae-t (~) + pe-2r (~)) == ae-f C) + fie-2t (~) .

Thus the linear map T converts the phase portrait for the system

r _ (-1 O. )

Y - 0 -2 Y

to that of X' = AX as shown in Figure 3.7.

•

Note that we really do not have to go through the step of converting a specific system to one in canonical form; once we have the eigenvalues and eigenvectors, we can simply write down the general solution. We take this extra step because, when we attempt to classify all possible linear systems, the canonical form of the system will greatly simplify this process.

T

-

Figure 3.7 The change of variables Tin the case of a (real) sink.

54 Chapter 3 Phase Portraits for Planar Systems

Example. (Complex Eigenvalues) Now suppose that the matrix A has complex eigenvalues a ± ifJ with /3 ::j: O. Then we may find a cornplex eigenvector Vt + iV2 corresponding to a + i/3, where both VI and V2 are real vectors. We claim that VI and V2 are linearly independent vectors in ]R2. If this were not the case, then we would have VI = cVz for some C E lit But then we have

But we also have

So we conclude that A V2 = (a + i/3) V2. This is a contradiction since the left-hand side is a real vector while the right is complex.

Since Vt + iV2 is an eigenvector associated to a + i/3, we have

Equating the real and imaginary components of this vector equation, we find

AV1 = aVI - fJV2 AVz= /3Vj +aV2.

Let T be the matrix whose columns are VI and V2. Hence TEj = Vj for j = 1,2. Now consider y-IAT. We have

(T-1AT)EI = T-I(aV1- /3V2) = aEI - 13Hz

and similarly

Thus the matrix y-l AT is in the canonical form

y-i AT =.( a 13.). -{3 ex

We saw that the system y' = (T-1 AT) Y has phase portrait corresponding to a spiral sink, center, or spiral source depending on whether a < 0,. ex = 0, or

3.4 Changing Coordina.tes 55

a: > O. Therefore the phase portrait of X' := AX is equivalent to one of these after changing coordinates using T. •

Example .. (Another Harmonic Oscillator) Consider the second-order equation

x" +4x:= O.

This corresponds to an undamped harmonic oscillator with mass I and spring constant 4. As a system, we have

The characteristic equation is

).2+4::::0

so that the eigenvalues are ±2i. A complex eigenvector associated to A :::: 2i is a solution of the system

-2ix+ Y = 0

-4x - 2iy:::: O.

One such solution is the vector (1, 2 i). So we have a complex solution of the form

Breaking this solution into its real and imaginary parts, we find the general solution

( caS2t) (Sin2t)

X(t):=q -2sin2t +C2 2cos2t .

Thus the position of this oscillator is given by

x(t) :::: q cos 2t + C2 sin 2t,

which is a periodic function of period .7r •

56 Chapter 3 Pha.se Portrafts for Planar Systems

T -

Figure3.8 The change of variables T in the case of a center.

Now, let T be the matrix whose columns are the real and imaginary parts of the eigenvector (1, 2i). That is

T= (6 ~).

Then, we compute easily that

which is in canonical form. The phase portraits of these systems are shown in Figure 3.8. Note that T maps the circular solutions of the system Y' = (T-1 AT) Y to elliptic solutions of X' = AX. •

Example. (Repeated Eigenvalues) Suppose A has a single real eigenvalue A. If there exist a pair of linearly independent eigenvectors, then in fact A must be in the form

(A 0)

A= 0 A '

so the system X' = AX is easily solved (see Exercise 15) ..

For the more complicated case, let's assume that V is an eigenvector and that every other eigenvector isa multiple of V. Let W be any vector for which V and W are linearly independent. Then we have

AW=p.V+vW

Exercises 57

for some constants u; v E R. Note that fL i=- 0, for otherwise we would have a second linearly independent eigenvector W with eigenvalue v. We claim that v = A. If v - ).. i=- 0, a computation shows that

A ( W + (v ~ A) V) = v (w + (v ~ A) V) .

This says that v is a second eigenvalue different from X, Hence we must have V=A.

Finally, let U = (lll1-)W. Then

A .

AU=V+-W=V+)..U.

J.I.

Thus if we define TEl = V, TE2 = U, we get

as required. Thus X' = AX is again in canonical form after this change of coordinates. •

E.XERCISES

1. In Figure 3.9 on page 58, you see six phase portraits. Match each of these phase portraits with one of the following linear systems:

(a) (_~ -~) (b) (-~ -~) (c) G -2)
-2
(d) (=~ ~) (e) (-~ -n (f) (-3 ~)
-2 58 Chapter 3 Phase Portraits for Planar Systems

1.

2.

3.

7.

Figure 3.9 Match these phase portraits with the systems in Exercise 1.

2. For each of the following systems of the form X, = AX (a) Find the eigenvalues and eigenvectors of A.

(b) Find the matrix Y that puts A in canonical form.

(c) Find the general solution of both X' = AX and yl = (y-1 AT)Y. (d) Sketch the phase portraits of both systems.

(i) A = e ~) (lii)A=(_~ ~) (v) A = (_~ _!)

(ii) A = G ~) (iV)A=(_~ !) (vi) A = G -D

3. Find the general solution of the following harmonic oscillator equations: (a) x" + x' + x = 0

(b) x" + 2x' + x = 0

4. Consider the harmonic oscillator system

X' = (0 1) X

-k -b

where b 2: 0, k > 0 and the mass m = I.

Exercises 59

(a) For which values of k, b does this system have complex eigenvalues?

Repeated eigenvalues? Real and distinct eigenvalues?

(b) Find the general solution of this system in each case.

(c) Describe the motion of the mass when the mass is released from the initial position x = 1 with zero velocity in each of the cases in part (a).

5. Sketch the phase portrait of X' = AX where

For which values of a do you find a bifurcation? Describe the phase portrait for a-values above and below the bifurcation point.

6. Consider the system

x' = e: ~)X.

Sketch the regions in the ab-plane where this system has different types of canonical forms.

7. Consider the system

Xl = (~ ~) X

with ),_ i= O. Show that all solutions tend to (respectively, away from) the origin tangentially to the eigenvector (1,0) when X < 0 (respectively, A> 0).

8. Find all 2 x 2 matrices that have pure imaginary eigenvalues. That is, determine conditions on the entries of a matrix that guarantee that the matrix has pure imaginary eigenvalues.

9. Determine a computable condition that guarantees that, if a matrix A has complex eigenvalues, then solutions of Xl = AX travel around the origin in the counterclockwise direction.

10. Consider the system

X' = (~ ~) X

where a + d i= 0 but ad - be = O. Find the general solution of this system and sketch the phase portrait.

60 Chapter 3 Phase Portraits for Planar Systems

11. Find the general solution and describe completely the phase portrait for

X' = (6 ~) x.

1.2. Prove that

is the general solution of

X' = (~ ~) x.

13. Prove that a 2 x 2 matrix A always satisfies its own characteristic equa tion.

That is, if A 2 + etA + f3 = 0 is the characteristic equation associated to A, then the matrix A 2 + etA + f31 is the 0 matrix.

14. Suppose the 2 x 2 matrix A has repeated eigenvalues A. Let V E )R2.

Using the previous problem, show that either V is an eigenvector for A or else (A ~ AI) V is an eigenvector for A.

15 .. Suppose the matrix A has repeated real eigenvalues A and there exists a pair oflinearly independent eigenvectors associated to A. Prove that

A = (~ ~).

16. Consider the (nonlinear) system

Xl = Irl l=~x.

Use the methods of this chapter to describe the phase portrait.