Notes on Exercises

This document is complementary to the answers in the book. For some of the
exercises in the study guide it contains hints, some intermediate steps, and an
answer in case of an incorrect answer in the book.

Note: this document does not contain full solutions, and should not be used as
a model for correct ways to write things down. For this, you can have a look at
the solutions of old exams.

Section 2.1
21. The general solution is y(t) = −2 − 23 (sin(t) + cos(t)) + Cet , with C ∈ R
arbitrary. If C 6= 0, the last term is unbounded, since et → ∞ as t → ∞.
If C = 0, the solution is bounded. To which y0 does C = 0 correspond?
23. The general solution is y(t) = − 23 t − 34 − 2et + Ce2t . For C 6= 0, the
last term is dominant for t → ∞. You can see this by rewriting y(t) =
e2t ( 32 te−2t − 34 e−2t − 2e−t + C). Since in the expression between brackets
all terms except C go to 0, we have y(t) ≈ Ce2t as t → ∞.
That means that for C > 0, y(t) → +∞ as t → ∞ and for C < 0,
y(t) → −∞ as t → ∞. What happens if C = 0?
24. First assume a 6= λ. Using the integrating factor µ(t) = eat , we find the
b
general solution y(t) = a−λ e−λt + Ce−at . Now use that limt→∞ e−ct = 0
for c > 0 (standard limits).
Now consider λ = a (why doesn’t the previous work in that case?). Again
we can take µ(t) = eat . We get y(t) = (bt + C)e−at , C ∈ R arbitrary. Now
use standard limits limt→∞ tr e−ct = 0 for all c > 0 and all r ∈ R.
Section 2.2
2. Don’t forget y(x) = 0. Solution in the book can be made explicit to
1
y(x) = sin(x)+C .
√
8. Solution can be made explicit: y(x) = −2 ± x3 − x + C + 4.
17. The interval of definition can be found by looking where the solution stops
to be differentiable. Since
1 + 3x2
y0 = ,
3y 2 − 12y

this happens when 3y 2 − 12y = 0 (note: numerator is nonzero), i.e. for

y = 0 or y = 4. Plugging this into the implicit solution gives cubic
equations for x. I do not expect you to be able to solve those by hand.
Section 2.4
∂f
5. Don’t forget to check the continuity of f (t, y) and ∂y (t, y).

1
 11. Typo’s in answer in the digital version
q of the book and absolute value
signs are missing. Correct: y(t) = ± 23 ln |1 + t3 | + y02 . Solution is differ-
 3 2 1
  3 2 1

entiable on two intervals: (e− 2 y0 − 1) 3 , ∞ and −∞, (−e− 2 y0 − 1) 3 .
Which is the relevant one and why?
1 1
Rt
21. a. We can pick y1 (t) = µ(t) and y2 (t) = µ(t) t0
µ(s)g(s) ds.
b. Recall that µ(t) = eP (t) , where P (t) is an anti-derivative of p(t). So
µ0 (t) = p(t)µ(t). Using this, we find:

µ0 (t) p(t)µ(t)
y10 (t) = − 2
=− = −p(t)y1 (t).
µ(t) µ(t)2

So indeed: y10 (t) + p(t)y1 (t) = 0.

c. You can check explicitly, but also argue: you get y2 (t) by setting
c = 0 in the general solution of the inhomogeneous equation. Hence
y2 (t) is a solution as well.
Section 2.5

6. a. The equation is of the form y 0 = f (y) (with f (y) = k(1 − y)2 ), hence
autonomous. f (y) = 0 ⇔ y = 1, so φ(t) = 1 is an equilibrium solu-
tion.

b.
f (y)

1
25. a. In this exercise you have to integrate (p−x)(q−x) (with p 6= q) at some
point. Note that by using partial fractions we can find
 
1 1 1 1
= − .
(p − x)(q − x) p−q q−x p−x

2
 28. c. Red means unstable, blue means stable, green means semi-stable.
y

Section 2.6
3. Note: if you use My = Nx to show exactness, don’t forget to check the
domain is simply connected (certainly the case here, since it is R2 ).
5. Answer in the book can be made explicit.:
p
– For c 6= 0, y(x) = − cb x ± 1c b2 x2 − c(ax2 − 2K);
2K−ax2
– For c = 0, y(x) = 2bx .

13. Assuming M and N are integrable (e.g. continuous), let PM and PN be

anti-derivatives of M and N respectively. Take Ψ(x, y) = PM (x) + PN (y).
Then dxd
Ψ(x, y(x)) = M (x) + N (y)y 0 , so the DE is exact.
?
17. Check the condition My = Nx before multiplying with µ and after multi-
plying with µ.
21. Hint: this equation is linear! What integrating factor works for those
equations?
Section 3.1
14. A local maximum is easy to find: solve y 0 (t) = 0, this gives t = ln( 49 ).
Check the second derivative:
9 3 1  9 9
y 00 (ln( )) = −et + e 2 t t=ln( 9 ) = − + < 0,
4 4 4 4 8
so y has a local maximum at that point, with value 94 .
Suppose that it is not a global maximum. There cannot be other lo-
cal maxima, so the only possibility would be that limt→∞ y(t) > 94 or
limt→−∞ y(t) > 49 . But limt→∞ y(t) = −∞ (do you see why?) and
limt→−∞ y(t) = 0. Hence y attains its global maximum at t = ln( 49 ).
16. The solution (in terms of β) is
1 1
y(t) = (1 + β)e 2 t + (1 − β)e− 2 t .

3
 b 1
√
20. a. Note that r± = − 2a ± 2a b2 − 4ac. Solutions negative means:

b 1p 2
− ± b − 4ac < 0.
2a 2a
Since a > 0 this is equivalent to:
p p
b > − b2 − 4ac AND b > b2 − 4ac.

The second implies the first, and is equivalent to:

b > 0 AND b2 > b2 − 4ac.

Given that a > 0, this is equivalent to:

b > 0 AND c > 0.

Real and distinct is equivalent to b2 > 4ac.

b. Hint: The roots are real with opposite signs precisely if r+ r− < 0
(can you see why?).
c. The argument is similar to that of a.
Section 3.2

11. Tip: show that c1 t2 + c2 t−1 solves the DE for all c1 and c2 , then of course
y1 (t) = t2 and y2 (t) = t−1 are just special cases, so you don’t have to
check those separately.
1
12. Tip: again plug in y(t) = c1 + c2 t 2 and show it is only a solution if
c1 = 0 or c2 = 0 (or both). Why is it no contradiction with the theorem
mentioned?
22. In this case we have W [y1 , y2 ](t) = e4t . How does the conclusion follow?
23. In this case we have W [y1 , y2 ](t) = x cos(x) − sin(x). Showing that this is
nonzero on (0, π) is tricky! In this case, I’m ok if you would do it visually.
Here is an argument:
d
Note that dx (x cos(x) − sin(x)) = −x sin(x) < 0 for x ∈ (0, π). Hence:
Z x
x cos(x) − sin(x) = (−s sin(s)) ds < 0
0

for x ∈ (0, π). In particular, x cos(x) − sin(x) 6= 0 on (0, π).

31. Hint: look at the Wronskian at t0 . How does its vanishing follow from the
given information?

Section 3.3
13. Intermediate step: the general solution is y(t) = et (c1 cos(2t) + c2 sin(2t)).

4
 14. Tip: write the set of equations to find c1 and c2 in matrix-vector form:
 1 1
√    
2√ 2 3 c1
=
2
.
− 21 3 1
2
c2 −2

25. Tip: Introduce the function z such that z(x) = z(ln(t)) = y(t). Then:
1
y 0 (t) = z 0 (ln(t)) ·
t
1 1
y (t) = z (ln(t)) · 2 − z 0 (ln(t)) · 2
00 00
t t
Then plug this into the differential equation, and show that z satisfies a
second order homogeneous DE with constant coefficients.
30. Tip: use 25.
Section 3.4
18. Write y2 (t) = v(t) y1 (t) = v(t) t3 .
Plugging into the DE gives: t5 v 00 (t) + 2t4 v 0 (t) = 0. Solving using your
favourite method will give v(t) = A 1t + B as general solution. To get an
independent solutions, we could take A = 1 and B = 0. Any other choice
with A 6= 0 works as well. Don’t forget to multiply with y1 to find y2
(common mistake!).
20. Write y2 (t) = v(t) y1 (t) = v(t) t.
Plugging into the DE gives: t3 v 00 (t) + t3 v 0 (t) = 0. Solving using your
favourite method will give v(t) = Aet + B as general solution. To get
an independent solutions, we could take A = 1 and B = 0. Any other
choice with A 6= 0 works as well. Don’t forget to multiply with y1 to find
y2 (common mistake!).
21. Write y2 (x) = v(x) y1 (x) = v(x) ex .
Plugging into the DE gives: (x − 1)v 00 (x) + (x − 2)v 0 (x) = 0. Solving us-
ing your favourite method (you will need some integration techniques) will
give v(t) = Axe−x + B as general solution. To get an independent solu-
tions, we could take A = 1 and B = 0. Any other choice with A 6= 0 works
as well. Don’t forget to multiply with y1 to find y2 (common mistake!).
Section 3.6
4. There are several typo’s in this exercise (in the electronic version, at least).
The closest interpretation seems y 00 + y 0 = 2 tan(t). In that case, the
homogeneous equation has {y1 (t) = 1, y2 (t) = e−t } as a fundamental set
of solutions (not unique, of course). In that case we find as particular
solution Y (t) = u1 (t)y1 (t) + u2 (t)y2 (t), with u1 and u2 satisfying:
( (
1 · u01 (t) + e−t u02 (t) = 0 u01 (t) = −2 tan(t)
⇒
0 · u01 (t) − e−t u02 (t) = 2 tan(t). u02 (t) = 2et tan(t)

5
 Hence we can take:
Z t
u1 (t) = −2 tan(s) ds = 2 ln | cos(t)|
0
Z t
u2 (t) = 2es tan(s) ds
0

The latter integral cannot be evaluated in terms of the functions you know.
So you get:
Z t
−t −t
y(t) = c1 + c2 e + 2 ln | cos(t)| + e 2es tan(s) ds.
0

HOWEVER: what the book meant to ask is: y 00 + y = 2 tan(t).

Then the hom. eq. has {y1 (t) = cos(t), y2 (t) = sin(t)} as fundamental set
of solutions. We get Y (t) = u1 (t)y1 (t) + u2 (t)y2 (t), with
( (
cos(t) · u01 (t) + sin(t)u02 (t) =0 u01 (t) = −2 sin(t) tan(t)
0 0
⇒
− sin(t) · u1 (t) + cos(t)u2 (t) = 2 tan(t). u02 (t) = 2 cos(t) tan(t) = 2 sin(t)
The second integral is easy: we can take u2 (t) = −2 cos(t). The first is
harder, and I do not expect you to be able to do it:
Z Z Z  
1  1 
−2 sin(t) tan(t) dt = −2 dt+2 cos(t) dt = −2 ln tan x +
 +2 sin(t).
cos(t) cos(x) 
1
The antiderivative of cos(x) is a standard result. I will not derive it here.
This leads to:
 
 1 
y(t) = c1 cos(t) + c2 sin(t) − 2 cos(t) ln tan x +
 .
cos(x) 
1
(standard notation, also used in the book, is sec(x) for cos(x) )

6. Note that a particular solution is not unique; you can add any solution of
the homogeneous equation to get another particular solution. If your an-
swer differs from the answer in the book by a solution of the homogeneous
equation, your answer is also correct.
3
10. Answer in the book should be 2 + 43 t2 ln(t). Also see remark at question
6.

13. See remark at question 6.

Section 5.1

14. a. In electronic version, part of the answer is at 13.

6
 P∞ P∞
bc y 00 = y ⇔ n=2 an n(n − 1)xn−2 = n=0 an xn . To compare we shift
indices in the first series: choose n = k + 2. In the second, we don’t
shift: choose k = n:
∞
X ∞
X
ak+2 (k + 2)(k + 1)xk = ak xk .
k=0 k=0

Now use the fact that two power series are equal if and only if the
corresponding coefficients are equal.
16. Confusing typo in the exercise: the indices under the summation sign
should be k, not n.
(in what folllows, k is the old index, m is the new one) In the first series
we choose k = m (no shift), in the second k = m − 1. We get
∞
X ∞
X
am+1 xm + am−1 xm .
m=0 m=1

Note that the summation does not start at the same index. How can you
fix this?
19. Following standard methods, you could also get
∞
X
a0 + a1 x + ((n − 1)an−1 + an ) xn .
n=2

Check that this is the same as the answer in the book.

−3
23. Intermediate result: you should get an+1 = n+1 a0 (with n = 0, 1, 2, . . .)
as recursion relation.

Section 5.2

1. a. Hint: write the x in the DE in terms of (x − 1).

b. y1 obtained by setting a0 = 1, a1 = 0, y2 obtained by setting a0 =
0, a1 = 1. Other choices are possible.

6. 1. Answer in the book treats cases n = 0 and n = 1 separately. This is

not necessary here.
9. Most important remark here is that for the solution with a0 = 1, a1 = 0,
the power series breaks off, so that we have an exact solution! Using
reduction of order, the DE can actually be solved exactly.
an
12. a. Recursion relation: an+2 = n+2 for n = 0, 1, 2, . . .. How are a0 and
0
a1 related to y(0) and y (0)?

7
 b. Graphs:

−1 1

c. “reasonably accurate” is not well defined, so any answer goes!

λ−2n
19. a. Recursion relation: an+2 = − (n+1)(n+2) an , for n = 0, 1, 2, . . ..

Section 5.3

2. Tip: to find the higher derivatives, do not rewrite the DE to y 00 = . . ., but

differentiate the DE itself. The calculations will be slightly easier then.
Intermediate results:
3
x2 y 000 + (3x + 1)y 00 + (1 + 3 ln(x))y 0 +
y=0
x
6 3
x2 y 0000 + (5x + 1)y 000 + (4 + 3 ln(x))y 00 + y 0 − 2 y = 0
x x

3. Intermediate results:

y 000 + sin(x)y 00 + 2 cos(x)y 0 − sin(x)y = 0

0000
y + sin(x)y + 3 cos(x)y 00 − 3 sin(x)y 0 − cos(x)y = 0
000

8. Intermediate results:

ex y (3) + ex y 00 + xy 0 + y = 0
ex y (4) + 2ex y (3) + (ex + x)y 00 + 2y 0 = 0
ex y (5) + 3ex y (4) + (3ex + x)y (3) + (ex + 3)y 00 = 0
ex y (6) + 4ex y (5) + (6ex + x)y (4) + (4ex + 4)y (3) + ex y (2) = 0

8
 n2 −α2
12. a. You should get the recursion relation an+2 = (n+1)(n+2) an .

Section 5.4

9. The solution blows up for t → 0.

25. General solution: y(x) = c1 x−1 + c2 x2 . Hint: first find the value(s) of c1
and c2 for which the solution is bounded near the origin, then find y.
p
26. Solution y(t) = tr implies r = − 21 (α−1)± 21 (α − 1)2 − 10. Tip: consider
the case of real roots and the case of non-real roots separately.

31. Intermediate result: from plugging in power series you find a1 = 0 and
1
an+1 = − (2n+3)(n+1) an−1 for n = 1, 2, . . ..

Section 5.5

1. Note that in combining the series you get two separate conditions:

a0 ((4r(r − 1) + r) = 0
a1 (4(r + 1)r − (r + 1)) = 0

Since in the method you assume a0 6= 0, the first condition implies 4r(r −
1) + r = 0, i.e. r = 0 or r = 43 . If you plug these into the second equation,
you see that a1 = 0 in both cases. This is important, but not mentioned
in the answer section of the book.
In the book, a general form of the nth term is given. I will not ask you to
do that, unless it is very easy.

5. Also here, a1 = 0.
6. Actually, in this case the option r = 1 does lead to a solution, namely
y2 (x) = x. In general, this might not work if the solutions of the indicial
equation differ by an integer.

7. Note: in this exercise it is more natural to shift the series such that all
terms are of the form . . . xk+r−1 , rather that . . . xk+r . This is ok: it
actually doesn’t matter what the common power is, as long as it is the
same in the series that you combine. All powers k + r is also possible;
then some of the series start at k = −1. This is also ok!

9. Answer not in the book. Some intermediate steps:

– The indicial equation becomes r2 = 0.
– Similarly to exercise 1: a1 = 0.
1 1
– For n ≥ 2 we find: an = − (n+r) 2 an−2 → for r = 0: an = − n2 ak−2 .

9
 – Odd terms vanish (since a1 = 0), for even terms: take a0 = 1, then
a2 = − 212 , a4 = 421·22 , general:
 2  2
k 1 k 1 1
a2k = (−1) = (−1) k
= (−1)k .
(2k) · (2k − 2) · · · · · 2 2 (k) · (k − 1) · · · · · 1 22k (k!)2

(I will not ask you to do this on the exam!)

– For convergence, apply the Ratio Test. Note: we do not need the full
series expression! By the recurrence relation we have aka−2
k
= − k12 ,
so:
 ak xk 
 
1 2
 ak−2 xk−2  = k 2 x → 0 for k → ∞.
 

Convergence for all x follows.

10. Answer not in the book. More difficult than I would ask on the exam.
Some steps:
Using the suggestion, we find y20 = J00 v +J0 v 0 and y200 = J000 v +2J00 v 0 +J0 v 00 .
Plugging into the DE, and using that J0 itself satisfies the DE, we get:

x2 J0 v 00 + (2x2 J00 + xJ0 )v 0 = 0.

Write v 0 = u:
x2 J0 u0 + (2x2 J00 + xJ0 )u = 0.
This is a linear equation. In standard form:
 0 
0 J0 1
u + 2 + u = 0.
J0 x
J0
We have p(x) = 2 J00 + x1 , so p(x) dx = ln(J02 ) + ln(x) (check!). As
R
R
integrating factor we can take µ(x) = e p(x) dx = xJ02 . Hence:

C
(xJ02 · u)0 = 0 ⇒ u = .
xJ02

We
R can take C as we like, let’s take C = 1. We then find v(x) =
1
xJ0 (x)2 dx and y2 (x) is the given expression.
Now: since J0 (x) is analytic and nonzero (equal to 1) at x = 0, J0 (x)−2
is analytic and equal to 1 at x = 0, so we can write: J0 (x)−2 = 1 + a1 x +
a2 x2 + . . .. Hence:
Z Z  
1 1
dx = + a 1 + a2 x + . . . dx = C + ln(x) + analytic function.
xJ02 x

We find that y2 (x) = CJ0 (x) + ln(x)J0 (x) + analytic function.

Since J0 (0) 6= 0, it follows that at x = 0, the dominant behaviour of y2 is
given by the logarithm. This is what is meant in the exercise.

10
Section 6.1

7. Carfully analyze for which s the integral converges. Note that the inte-
grand can be written as a sum of exponential functions. Split into two
integrals; both have to converge!
18. Laplace transformation exists for s > 0.

Section 6.2

6. Recall that in partial fractions, in general you need the degree of the
numerator to be one less than the degree of the denominator.
s+5
9. Intermediate result: Y (s) = (s+1)(s+2) .
1
Tip: It is possible to first split (s+1)(s+2) and then rewrite Y (s), but it is
s+5
faster to split the fraction (s+1)(s+2) directly.

12. Intermediate result: Y (s) = s23s−6

−2s+4 . Hint: complete the square. After
that, you will still have to do some rewriting.
13. Typo in the electronic version: the DE should be y (4) (t) − 4y(t) = 0.
3
Intermediate result: Y (s) = ss4−2s
−4 . Hint: by factorizing numerator and
denominator the expression can be simplified.
s−1
15. Intermediate result: Y (s) = s2 −2s+2 + (s2 +1)(ss2 −2s+2) . When splitting the
second term, recall that - in general - the degree of the numerator has to
be one less than the degree of the denominator.
18. Incorrect answer in electronic version. It should be Y (s) = 1
s2 (s2 +1) − e−3s s21+3s
(s2 +1) .

Section 6.3

4.
y

6.

11
 y

8.
y

14. Hint: factorize the denominator and apply partial fractions.

16. The answer can also be written as f (t) = 12 u4 (t) e2(t−4) − e−2(t−4) .

Section 6.4

2. Incorrect answer in electronic version. The answer should be:

y(t) = 21 t + 3
2 sin(t) − 21 u6 (t)(t − 6 − sin(t − 6)).

Graph of forcing function (red) and solution (blue):

12
 y

6. Some intermediate steps:

1 −πs
– L {g(t)} (s) = 1+s2 (1 + e ).
Hint: express sin(t + π) in terms of sin(t).
s+1 1
– Y (s) = 2 5 + (1 + e−πs ).
s +s+ 4 (s + 1)(s2 + s + 54 )
2

– Remember how to split a fraction when the denominator is a product

of two quadratic polynomials:
1 As + B Cs + D
= 2 + 2 .
(s2 + 1)(s2 + s + 45 ) s +1 s + s + 45

Graph of forcing function (red) and solution (blue):

y

11. a. Graph (jumps at kπ):

y

c. Graph:

13
 y

d. As n → ∞, the amplitude of the oscillation keeps growing (undamped

resonance).
Section 6.5

1. Graph:
y

3. Graph:
y

5. Graph:
y

14
 6. Graph:
y

Section 6.6

Z t
1. Note that (f ∗ 1)(t) = f (τ ) dτ . There is actually only one function that
Z t 0

satisfies f (t) = f (τ ) dτ . Which one?

0

8. The integral can be evaluated by repeated integration by parts:

Z t
1
12
1 3
(t − τ )3 sin(2τ ) dτ = 24 1
t − 16 1
t + 32 sin(2t).
0

s+1 G(s)
13. Intermediate result: Y (s) = s2 +ω 2 + s2 +ω 2 .

Section 7.1

3. Note: you need a dependent variable for u, for u0 and for u00 .
11. Note that you need Hooke’s law: the force exerted by a spring, stretched
by an amount u w.r.t. the equilibrium position, is F = −ku, with k the
spring constant. Then you can argue as follows: the extension of spring
1 is x1 , the extension of spring 2 is x2 − x1 . Then the force on mass m1
is Fmass 1 = −k1 x1 + k2 (x2 − x1 ) + F1 (t). Looking at the picture helps in
getting the signs right! A similar argument gives the force on mass m2 .
  
c1 x1 (t) + c2 x2 (t) x1 (t) x2 (t)
13. Plug in and use that and are
c1 y1 (t) + c2 y2 (t) y1 (t) y2 (t)
solutions.
12. Mistake in my electronic version: the left hand side of the first equation
should be x01 .
The idea is to eliminate one of the dependent variables. E.g. assuming
that a12 is non-zero, the first equation can be rewritten as
1
x2 = (x0 − a11 x1 − g1 (t)).
a12 1

15
 This equation can the be used to eliminate x2 and x02 (how?) in the second
equation. This ultimately gives:
1 00 a11 + a22 0 a11 a22 − a12 a21 g 0 (t) − a22 g1 (t)
x1 − x1 + x1 = 1 + g2 (t).
a12 a12 a12 a12
For the boundary conditions we get:

x1 (0) = x01 (given)

x01 (0) = a11 x01 + a12 x02 + g1 (0) (from first DE)

In case we take a21 6= 0, we get in a similar way:

1 00 a11 + a22 0 a11 a22 − a12 a21 g 0 (t) − a11 g2
x2 − x2 + x2 = 2 + g1 (t),
a21 a21 a21 a21
with boundary conditions:

x2 (0) = x02 (given)

x02 (0) = a21 x01 + a22 x02 + g2 (0) (from second DE)

If the coefficients depend on t, you can follow the same procedure to

eliminate x1 or x2 . However, the expressions will not be so pretty!
17. a. Note: the picture given is not the picture corresponding to the text,
but just an example. Make a similar picture for the situation in this
exercise.
Further, we can use “change per minute = inflow per minute − out-
flow per minute” for both vessels. Note that the volume is constant
in both. For vessel 1 we get:
Q2 (t) Q1 (t)
Q01 (t) = 3q1 + −4 .
100 60
The terms on the right hand side are: external inflow, inflow from
vessel 2, outflow out of the system.
Similarly, for vessel 2 we get:
Q1 (t) Q1 (t)
Q02 (t) = q2 + 2 −3 .
60 100
b. What do we know about the derivative of equilibrium solutions?
c. Plug it in... Why is the answer you get not possible?
d. The q1 , q2 , QE E
1 and Q2 have to satisfy:
 4 1
  E  
60 − 100 Q1 3q1
2 3 E = .
− 60 100 Q 2 q2

Invert the matrix, and use the fact that q1 and q2 cannot be negative.

16
Section 7.4

8. The hardest part of this exercise is keeping track of all the notation. First
recall that for a second order homogeneous linear DE the Wronskian of
two solutions y (1) (t), y (2) (t) is defined as
 (1)
y (2) (t)

y (t)
W [y (1) , y (2) ](t) = det d (1) d (2) .
dt y (t) dt y (t)

On the other hand, for a 2 × 2 linear homogeneous system of equations,

the Wronskian of 2 solutions x(1) (t), x(2) (t) is defined as
" #
(1) (2)
x (t) x (t) (1) (2) (1) (2)
W [x(1) , x(2) ](t) = det 1(1) 1
(2) = x1 (t)x2 (t) − x2 (t)x1 (t).
x2 (t) x2 (t)

The system of DEs is constructed

 in such a way that if y(t) is a solution
y(t)
of the second order DE, the 0 is a solution to the system (Check it!).
y (t)
(1) (2)
Since y and y are
 solutions to the second order DE, it follows that
y (1) y (2)

(1)
d (1) and d (2) are solutions of the given system of DEs. Since x
dt y dt y
and x(2) form a fundamental set, we have:
 (1) 
y (1)
d (1) = c1 x + c2 x(2)
dt y
 (2) 
y (1)
d (2) = c3 x + c4 x(2)
dt y

for some constants c1 , c2 , c3 and c4 . Hence we get:

" #
(1) (2) (1) (2)
(1) (2) c1 x1 + c2 x1 c3 x1 + c4 x1
W [y , y ] = det (1) (2) (1) (2) .
c1 x2 + c2 x2 c3 x2 + c4 x2

Writing out the determinant gives:

(1) (2) (1) (2)
W [y (1) , y (2) ] = (c1 c4 − c2 c3 )(x1 x2 − x2 x1 ) = cW [x(1) , x(2) ].

Note that the calculation is simpler if you recognize that

" # " #
(1) (2) (1) (2) (1) (2) 
c1 x1 + c2 x1 c3 x1 + c4 x1 x1 (t) x1 (t) c1 c3
(1) (2) (1) (2) = (1) (2) ,
c1 x2 + c2 x2 c3 x2 + c4 x2 x2 (t) x2 (t) c2 c4

and use the rules of calculation of the determinant.

9. Typo in my electronic edition: g(t) should be g(t).
Fix a particular solution xp of the inhomogeneous equation x0 = P(t)x + g(t).
We have to show that we get precisely all solutions of the inhomogeneous

17
 equation by adding x(p) to the solutions of the homogeneous equation.
If x(c) is a solution of the homogeneous equation, so (x(c) )0 = P(t)x(c) ,
then we have:

(x(c) + x(p) )0 = P(t)x(c) + P(t)x(p) + g(t) = P(t)(x(c) + x(p) ) + g(t),

so (x(c) + x(p) ) is indeed a solution of the inhomogeneous equation.

To show that we get all solutions in this way, assume that x is any solution
of the inhomogeneous equation. Then we have:
 0    
x − x(p) = x0 −(x(p) )0 = (P(t)x + g(t))− P(t)x(c) + g(t) = P(t) x − x(p) ,


So x(c) = x − x(p) is the solution of the homogeneous equation, and we
see all solutions of the inhomogeneous equation are of the form

x = x(c) + x(p) ,

where x(p) is a fixed particular solution.

Section 7.5

2. Eigenvector directions in color:

4

-2

-4

-4 -2 0 2 4

3. Eigenvector directions in color:

18
 4

-2

-4

-4 -2 0 2 4

5. This one is special, because 0 is an eigenvalue. The general solution is:

   
1 1 t
x(t) = c1 + c2 e.
3 2

You see that for c2 6= 0 the trajectories in the phase

 portrait are half lines
t 1
(only half because e > 0, with direction c2 . Since limt→−∞ x(t) =
2
 
1
c1 the lines seems to emanate from the points on the line through
3
 
1
vector (the red line).
3

17. Eigenvector directions in green and red, the trajectory though (2, 3) in

19
 orange.
4

-2

-4

-4 -2 0 2 4

   1 −t 7 2t 
x (t) e + e
The orange curve is given by the solution 1 = 41 −t 47 2t . Here
x2 (t) −2e + 2e
is a plot of the component functions w.r.t. t (blue is x1 , orange is x2 ):
15

10

-2.0 -1.5 -1.0 -0.5 0.5 1.0

From the phase portrait you see that x1 (t) → +∞ both for t → ∞ as for
t → −∞, and that x2 (t) → +∞ both for t → ∞ and x2 (t) → −∞ for
t → −∞. You see the same behavior in the graphs above.

Section 7.6

Remark: the general solution of a linear system can be written in many different
ways. In particular in the case of non-real eigenvalues, equivalent solutions
might look very different. So if the answer in the book looks different from your
answer, it is not necessarily incorrect.

1. Phase diagram:

20
 1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

2. Phase diagram:
1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

3. Phase diagram:

21
 1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

6. NOTE: The answer in the electronic version corresponds to

1 −5
matrix A = .
1 −3  
1 −5
Phase diagram for A = :
1 −3
3

-1

-2

-3

-3 -2 -1 0 1 2 3
 
−1 −5
The solution of the initial value problem with A = would be:
1 −3

2 cos(2t) − 32 sin(2t)
 
−2t
x(t) = e .
cos(2t) + 12 sin(2t)

22
  
−1 −5
Phase diagram for A = :
1 −3
3

-1

-2

-3

-3 -2 -1 0 1 2 3

9. Phase diagram of the trajectory through (1, 12 ):

-1

-2

-2 -1 0 1 2

Corresponding solutions w.r.t t (blue is x1 , orange is x2 ):

23
 1.5

1.0

0.5

2 4 6 8 10

-0.5

-1.0

-1.5

x1 vs x2 vs t diagram. Projections on (x1 , x2 )-plane, (x1 , t)-plane and

(x2 , t)-plane in gray (you don’t have to be able to sketch such 3d pic-
tures!):

49
12. Asymptotically stable spiral for α > 24 , asymptotically stable node for
2 < α < 49
24 , saddle point for α < 2.
Phase portrait for α = 1.5 (saddle):

24
 1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

Phase portrait for α = 2.5 (spiral):

1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

48.5
Phase portrait for α = 24 (node, barely recognizable):

25
 1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

Section 7.7
1. a. Intermediate
  results: eigenvalues  ±1. Eigenspaces: Er=−1 =
 are
−1 −3
span{ } and Er=1 = span{ }. Recall that a fundamental
1 1
matrix is a matrix of which the columns form a fundamental set of
solutions.
b. Sign mistake in the book. It should be:
1 −e−t + 3et −3e−t + 3et
 
Φ(t) = .
2 e−t − et 3e−t − et
   
−1 −2
4. Eigenspaces: Er=−1 = span{ } and Er=2 = span{ }.
2 1
 
1
6. Eigenspaces: Er=−1+2i = span{ } and complex conjugate for other
2i
eigenvalue.
9. The matrix in this exercise should be the same matrix as in 1. In that
case you simply have:
1 −11e−t + 15et
 
x(t) = Φ(t)x(0) = .
2 11e−t − 5et
However, in my version of the book, the matrix here is the transpose of
the matrix A in ex. 1. In that case, you can also find the standard
fundamental matrix more easily:
∞ ∞
!T
AT t
X 1 T n n X 1 n n
Φ(t) = e = (A ) t = A t = Φ(t)T .
n=0
n! n=0
n!

26
 This works because of the rules of calculation for the transpose: (A +
B)T = AT + B T , (cA)T = c(AT ) and (An )T = (AT )n . Hence if you you
want to solve the system that is actually given, you can simply use the
transpose of the standard fundamental matrix of exercise 1.
12. a. Keep s fixed and write Z1 (t) = Φ(t)Φ(s) and Z2 (t) = Φ(t + s). Using
that Φ0 = AΦ, show that on the one hand:

Z10 (t) = AZ1 (t) with Z1 (0) = Φ(s),

and on the other hand:

Z20 (t) = AZ2 (t) with Z1 (0) = Φ(s),

so they satisfy the same boundary value problem. By uniqueness,

they must be equal.
b. Use the rule proved in a.
c. Use part a. and b.

13. Use the series definition of eA . Use that for a diagonal matrix we have:
 k  k 
a1 a1
 ..  =
  .. 
 . . 
an akn

Section 7.8
General remark: a generalized eigenvector is not unique. More generally, if ξ
is an eigenvector with eigenvalue r and η is a generalized eigenvector with the
same eigenvalue, then η + c ξ is also a generalized eigenvector with the same
eigenvalue, for any choice of c.
 
−2
3. Phase portrait; All trajectories are parallel to the line spanned by
1
(in red). All trajectories below that line are directed to the left, all above
to the right:

27
 1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

4. Phase portrait:
1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

 
3 + 5t
6. Answer in electronic version incorrect. Should be x(t) = . Tra-
4 + 15t
jectory and x1 -vs-t-plot show straight lines.
7. Trajectory in phase portrait:

28
 6

-2

-4

-6

-6 -4 -2 0 2 4 6

x1 vs t:
x1

t
-2 2 4 6 8 10

-2

q
4R2 C
−1 ± 1−
r
1 1 1 4 L
14. a. Eigenvalues are r1,2 =− ± − = .
2RC 2 R2 C 2 LC 2RC
20. b. Inductive argument not necessary, sufficient if the pattern is clear.
c. Use the series:
∞ ∞
1 n n X 1 λn nλn−1
X  
Jt
e = J t = .
n! n! 0 λn
n=0 n=0

The diagonal elements become:

∞
X 1 n n
λ t = eλt .
n=0
n!

29
 The non-zero off-diagonal element becomes:
∞ ∞ ∞
X n n−1 n X 1 n−1 n−1
X 1 k k
λ t =t λ t =t λ t = teλt .
n=0
n! n=1
(n − 1)! k!
k=0

In the second step we set the start index to n = 1, since the first
term is 0 anyway. In the next step we shift by setting k = n − 1. We
find:  λt
teλt

e
eJt = .
0 eλt

d. Since eJt is the standard fundamental matrix, we have

  λt
teλt x1 (0) x1 (0)eλt + x2 (0)teλt
      
x1 (t) Jt x1 (0) e
=e = = .
x2 (t) x2 (0) 0 eλt x2 (0) x2 (0)eλt

Section 7.9

2. Answer in my electronic version incorrect. In my e-book, the exercise is

to solve:    
0 2 1 − cos(t)
x (t) = x(t) + .
−5 −2 sin(t)

Some intermediate results:

 
−1
– The matrix has eigenvalues ±i. r = i has eigenspace Er=i = span .
2−i
– A fundamental matrix could be:
 
− cos(t) − sin(t)
Ψ(t) = .
2 cos(t) + sin(t) − cos(t) + 2 sin(t)

– Writing the particular solution as xp (t) = Ψ(t)c(t), we find that

    
0 −1 − cos(t) + 2 sin(t) sin(t) − cos(t) 1 − 2 sin(t) cos(t)
c (t) = Ψ(t) g(t) = =
−2 cos(t) − sin(t) − cos(t) sin(t) 2 cos2 (t)

– Integrate in your favourite way. We can take:

t + 21 cos(2t)
 
c(t) = .
t + 12 sin(2t)

– This gives:
xp (t) = Ψ(t)c(t) = . . .
– What is the general solution?
t−1
 
4. Typo in the electronic version: the inhomogeneous term should be .
2t−1 + 2

30
  
1
– Eigenvalues are r = 0 and r = −5. Eigenspaces: Er=0 = span
  2
−2
and Er=−5 = span .
1
– A fundamental matrix could be:
−2e−5t
 
1
Ψ(t) = .
2 e−5t

– We then have:  
1 1 2
Ψ(t)−1 = .
5 −2e5t e5t
– Following the standard procedure, we have:
 −1 4 
0 −1 t +5
c (t) = Ψ(t) g(t) = 2 5t .
5e

5. Some intermediate results:

 
−1
– Eigenvalues are r = ±1. Eigenspaces: Er=−1 = span and
1
 
−3
Er=1 = span .
1
– A fundamental matrix could be:
 −t
−3et

−e
Ψ(t) = .
e−t et

– We then have:  t
3et

1 e
Ψ(t) = −t .
2 −e −e−t
– Following the standard procedure, we have:
 2t 
−e
c0 (t) = Ψ(t)g(t) = .
0

cos( 12 t) sin( 12 t)
 
1
11b. You can use Ψ(t) = e− 2 t .
4 sin( 21 t) −4 cos( 12 t)

Section 9.1

31
2. Phase portrait:
1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

3. Phase portrait:
1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

4. Phase portrait:

32
 1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

5. Phase portrait:
1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

6. Phase portrait:

33
 1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

7. Phase portrait:
1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

8. Phase portrait:

34
 1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

14. a. Note that zero eigenvalue means that det(A − rI) = 0 for r = 0.
b. Since 0 is an eigenvalue, there is a non-zero vector ξ such that Aξ = 0.
It follows that A(cξ) = 0, for all c ∈ R. So all points on the line
spanned by ξ are critical.
c. The general solution looks like x(t) = c1 ξ (1) + c2 ξ (2) er2 t . If we write
s = er2 t , you see that the right hand side becomes c1 ξ (1) + c2 ξ (2) s
(with s > 0). This is a parameterization of a line through c1 ξ (1) with
direction c2 ξ (2) . If c2 = 0 we get an equilibrium solution. If c2 6= 0,
the trajectory is a half line (half, since s = er2 t > 0.

Section 9.2

 
0 2
5. Jacobi matrix: J(x, y) = . What are the eigenvalues at the
−10x 0
two critical points?
Phase portrait (critical points in blue):

35
 0.5

0.0

-0.5

-1.0

-1.5

-1.0 -0.5 0.0 0.5 1.0

Remark: The critical point ( √15 , − 12 )

turns out to be a stable center, also
for the non-linear system. This follows from the fact that the trajectories
are precisely level curves of the function H(x, y) = y+y 2 −x+ 53 x3 (you can
find this using the technique of exercises 15, 18 or 20). Since this function
is continuous, different level curves cannot intersect, hence trajectories on
different level curves cannot have a common limit point, as would be the
case the critical point were a spiral point.
√ √
6. Answer√ in (my √ version of) the book incorrect: (0, 0), (1 + 2, 1 − 2) and
(1 − 2, 1 + 2) are critical  points. 
−y 2 − 2y − x
Jacobi matrix: J(x, y) = . Eigenvalues:
−1 − 2y −1 − 2x
√
– At (0, 0): − 21 ± 12 7i.
√ √
– At other points, they are very ugly: e.g. at (1 + 2, 1 − 2) they
√ p √
are 12 (−4 − 2 ± 50 − 8 2). It is ok if you do this exercise by
qualitatively analyzing a phase portrait.
Phase portrait (critical points in blue, separatices in red):

36
 3

-1

-2

-2 -1 0 1 2 3

 
y − 3 − 2x x+3
9. Jacobi matrix: J(x, y) = . Eigenvalues:
y(1 − 2x) 2 + x − x2

– At (0, 0): r = −3 or r = 2;
– At (−3, 0): r = −10 or r = 3;
√
– At (−1, −1): r = −1 ± i 5;
√
– At (2, 2): r = − 52 ± 2i 95.
Phase portrait (critical points in blue, separatices in red):

-2

-4 -2 0 2 4

37
15. Phase portrait:

-1

-2

-2 -1 0 1 2

dy y−2xy
18. We have dx = −x+y+x 2 . This is not separable, but can be written as an

exact DE!
Phase portrait:

1.0

0.5

0.0

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

20. Phase portrait:

38
 4

-2

-4

-4 -2 0 2 4

Section 9.3
Note: on the exam I will not ask you to show that a system is locally linear.
Formally proving this requires calculating types of limits that go beyond the
scope of this course.
 0   
u −1 1 u
1. Linearized system: = .
v0 −4 −1 v
 0   
u 0 1 u
2. Linearized system: = .
v0 −4 0 v
 0   
u 1 0 u
3. Linearized system: = .
v0 1 1 v

5. Mistake in answer electronic version: Eigenvalues at ( 21 , 12 ) should be

√
r1,2 = − 12 ± 2i 7.
9. Note: there are infinitely many critical points!
18. Note that x is the angle in radians w.r.t. the equilibrium position. To
what position does x = π correspond? To what position does x = 4π
correspond? Based on the physics, what can you say about stability at
those points?
2
19. We have dxdy
= −ω ysin(x) . If you solve it, you might not immediately
recognize the form given in the exercise. Note, however, that ω 2 is a
constant. The reason to write the solution like in the exercise is that it
can be more easily related to conservation of energy.

Section 10.1

39
 3. Note: sin(2L) = 0 ⇔ L = 12 nπ, with n integer.
7. The general solution of the DE (so without the boundary conditions) is
y(t) = c1 cos(2x) + c2 sin(2x) + 31 cos(x).
For 14, 16 and 18: The eigenvalues in this context are the values for λ for
which the boundary value problem has at least one non-trivial solution.
The eigenfunctions are the non-trivial solutions for those values of λ.
14. Tip: distinguish the cases λ < 0, λ = 0 and λ > 0. The eigenvalues

2
as given in the book can also be written as λn = 21 + n , with n =
0, 1, 2, . . ..
16. Tip: distinguish the cases λ < 0, λ = 0 and λ > 0.

18. Note: this is an Euler equation! Tip: distinguish the cases λ < 1, λ = 1
and λ > 1.

Section 10.2

13. Sketch:
y

x
−L L 3L 5L

15. Sketch:

40
 y
x
−π π 3π 5π

Note: the Fourier series can also be written as, for example,
∞ 
1 − (−1)n (−1)n

3π X
− +3 cos(nx) − sin(nx) .
4 n=1
πn2 n

Instead of (−1)n you can also write cos(nπ).

16. Sketch:
y

x
−1 1 3 5
Note: the Fourier series can also be written as
∞
1 X 1 − (−1)n
+2 cos(nπx).
2 n=1
n2 π 2

Section 10.4

10. Even extension:

41
 y

Odd extension (ignoring the precise values at the jump points):

y

17. Sketch (at jump points, a Fourier series converges to the average of the
left and right limit):
y

19. Sketch:

42
 y

22. Incorrect answer in the book. The Fourier series should be:
∞  
X 2 4
− cos(nπ) + 2 2 sin( 12 nπ) sin( 21 nπx).
n=1
nπ n π

Sketch (at jump points, a Fourier series converges to the average of the
left and right limit):
y

Section 10.5

8. Note: you do not have to calculate any integral to obtain the Fourier
“series”.
12. Tip: Split the integral for the Fourier coefficients in 2 parts.
18. See table 10.5.1 for the material properties.
22. Some intermediate steps and remarks:
– The DE can be written as:
R00 1 R0 1 Θ00 1 T0
+ + 2 = 2 .
R r R r Θ α T

– Both sides have to be constant (why?). The book chooses −λ2 as

constant on both sides. You don’t have to take the square here.
It is convenient in practice, since it turns out that both sides have
to negative if you want bounded solutions, but you don’t have to
consider that here.
– The left hand side can then be rewritten as:
R00 R0 Θ00
r2 +r + λr2 = − .
R R Θ

43
 – Both sides have to be constant (why?). The book chooses −µ2 as
constant on both sides. Again, it turns out that the square is conve-
nient (has to to with the required periodicity of Θ), but you do not
have to consider it here.

Section 10.6

11. Note: the initial function u(x, 0) is a sine, but you need to write it
as
R a cosine series. For this you have to evaluate integrals of the form
sin(ax) cos(bx) dx. This can be done in various ways, e.g. by the “super-
trick” of doing 2 times partial integration, which gives the same integral,
which you can then move to the other side of the equation to evaluate it.

15.b. This situation (one insulated end, one end constant temperature) has not
been treated in class. There is a trick though: consider a rod with length
2L in which the temperature is 0 at the endpoints and symmetric w.r.t. the
mid-point. Then an differentiable temperature distrubution must satisfy
ux (L, t) = 0, by symmetry. We can apply the theory of 2 endpoints with
0 temperature to this long rod. In particular, we can find the Fourier sine
series by looking at the odd 4L-periodic extension of the temperature in
the long rod.
Since only the odd sines are symmetric on the interval [0, 2L], only they
will contribute to the Fourier series.

Section 10.7

1. See the following link: https://www.geogebra.org/m/ys9tv7t8. You

can change the time t by moving the slider.
4. See the following link: https://www.geogebra.org/m/z68twgsd. You
can change the time t by moving the slider.
10. See 10.6.15. Also here you can consider a string of length 2L with fixed
endpoints, and u(x, t) symmetric with respect to the midpoint.
15. a. The equation referred to is u(x, t) = φ(x − at) + ψ(x + at).
b. Why is the value of c irrelevant?
c. Replace x by x − at and rewrite.

44