A First Course in Differential Equations, 3rd ed.

Springer-Verlag, NY (2015)
J. David Logan, University of Nebraska

SOLUTIONS TO EVEN-NUMBERED EXERCISES

This supplement contains solutions, partial solutions, or hints to most of the

even-numbered exercises in the text. Many of the plots required in the Exercises
are not displayed, but rather left to the reader. Solutions to the odd-numbered
exercises are posted on the author’s web site:
http://www.math.unl.edu/ jlogan1/diff-eqs3.htm

Chapter 1 Exercises

 Sec. 1.1.2, page 9

2. x = tan t, so x′ = sec2 t = 1 + tan2 t = 1 + x2 . Clearly x(0) = 0.
6. Let x = at2 + bt + c, so x′ = 2at + b. Substitute into the DE to get
2at + b + 2(at2 + bt + c)t2 + 4t + 7. Equating like coefficients gives 2a =
1, a + b = 2, b + 2c = 7, giving a = 1/2, b = 3/2, c = 11/4.
8. Substitute x = tm to get t2 m(m − 1)tm−2 − 6tm = 0, or m2 − m − 6 = 0.
So m = 3, −2.
10. Note x′ = C. So tx′ − x + f (x′ ) = tC − (Ct + f ′ (C)) + f ′ (C) = 0.
12. I(x) = I0 e−1.4x . Setting I(x) = 0.01I0 and solving for x gives x = 3.29
m.
14. C(t) = 6.68e−0.000121t . Setting C(t) = 0.09 and solving for t gives t =
35596 yrs.
16. In 2050 the population will be approximately 11.5 billion.
 Section 1.1.3, page 15
√
2. Set x2 + t2 = C. So the isolines are circles of radius C.
4. The nullcline is t = x2 , a parabola opening to the right. The slope field is
positive when t > x2 (region between the two branches of the parabola)
and negative when t < x2 (outside the branches).

1
 Section 1.2, page 19
2. x = (t1/2 + t−1/2 )dt = (2/3)t3/2 + 2t1/2 + C; x(1) = 4 implies C = 4/3.
R

4. (a) Use integration by parts: x(t) = C − (2t + 1)/(4 exp(2t)). (b) Use the
substitution
√ y = ln t to get x(t) = ln(ln(t)) + C. (c) Use the substitution
y = t.
Rt √
6. x(t) = 1 e−s / s ds.
8. Integrate the equation x′′ = −g twice, and evaluate the two constants
using the initial conditions. We get x = − 21 gt2 + vt + h.
√ 2
10. (a) Note that the derivative of the erf function is (2/ √π)e−t . √ Thus
Rt
f ′ (t) = erf ′ (sin t) cos t. (b) By the product rule f ′ (t) = te− t + 1 e− s ds.

12. Let the initial velocity have x and y components V cos θ and V sin θ. In-
tegrating the x and y equations twice and applying the initial conditions
gives x(t) = (V cos θ)t + L and y(t) = −(1/2)gt2 + (V sin θ)t + H. Maxi-
mize the trajectory at the wall, y(L) = −(1/2)gL2 + (V sin θ)L + H with
respect to θ. You get θ = π/2. So, construct the wall higher than y(L).
14. Pull the e−at factor out of the integral and differentiate using the funda-
mental theorem of calculus. We obtain a differential equation x′ + (1 +
a)x = ab, with x(0) = b. Thus x(t) = Ce−(1+a)t − ab/(1 + a). Use the
initial condition to find C.
 Section 1.3.1, page 26
2. y = Ce−rt + a.
4. (a) x(t) = C(t + 1)2 . (b) θ(t) = arcsin 23 (t + 1)3/2 + C. (c) Integrate

to get u2 + u = 21 t2 + t + C. Us the quadratic formula to obtain explicit

solutions. (d) R(t) = tan(C + (t + 1)2 /2). (d) Write the equation as
ydy/(1 + y 2 ) = −dt and integrate to get arctan(y 2 + 1) = −2t + C. (f )
x(t) = −1/(C − ln(t + 1)).
6. Use the hint to obtain the differential equation y ′ = 4 − y 2 . Separate
variables and integrate, using partial fractions, to get

|2 − y|
ln = 4t + C, y(0) = −1.
|2 + y|

So, C = ln 3. Solve for y to get y(t) = 2 − 12/(exp(4t) + 3). So x(t) =

4t − y(t), which is valid on (−∞, ∞).
8. Separate variables to get ex dx = t2 dt, giving ex = (1/3)t3 + C. x(0) = ln 2
gives C = 2. Then x(t) = ln((1/3)t3 + 2)). The interval of existence is
when t3 > −6, or t > −31/3 .

2
 10. (a) Separate variables to get e−x dx = et dt. (b) Use partial fractions on
the T integral. (c) Use the fact tan y = sin y/ cos y.
12. y(t) = −1/(ln(t2 + 1) − 1/y0 ). If y0 < 0 the solution exists for all t. If
y0 > 0 the solution exists for ln(t2 + 1) < 1/y0 .
14. (a) The integral curves are ellipses. (b) Use implicit differentiation to get
2xx′ + 4t = 0, or x′ = −2t/x. (c) When t = 1, x = 4 we get C = 18.
16. See the formula in the text.
′
18. By the definitions, xx = r − ci x = r − D
H x, which reduces to x′ = rx(1 −
x/K), where K = rH/D.
20. Integrate both sides with respect to t to get ln x = a ln y, where the con-
stant of integration is set to zero. Then x = y a by properties of logarithms.
22. Let y = at + bx + c, then y ′ = a + bx′ = a + bF (y). (a) Let y = t + x; then
the equation becomes y ′ = 1 + y 2 , giving y = tan t. Then x = tan t − t.
24. Write (tx′ )′ = −2t and integrate to get tx′ = −t2 + c1 . Divide by t and
integrate again to get x = −t2 /2 + c1 ln t + c2 .
26. The equation is homogeneous. The solution is y(t) = ct2 /(1 − Ct).
2 Rt 2
28. Write
√ −1/y 2dy = e−t dt and integrate to get 1/y = 0 e−s ds + C =
( π/2)erf(t) + C. The initial condition gives C = 2. Then solve for y.
 Section 1.3.2, page 32
2. T ′ = −h(T − Te ), where Te is the unknown refrigerator temperature, and
T (0) = 46. So T (t) = (46 − Te )e−ht + Te . Now T (10) = (46 − Te )e−10h +
Te = 39, and T (20) = (46 − Te)e−20h + Te = 33. Now use software to solve
the last two equations for h and Te .
4. T (t) = 58e−ht + 10. Then T (9) = 58e−9h + 10 = 57, giving h = 0.023.
Then T (17) ≈ 49 degrees.
 Section 1.3.3, page 34
2. 100C ′ = 0.5(0.0002 − C), C(0) = 0. Separate variables and get C(t) =
(1/5000)(1 − exp(−t/200)). The equilibrium occurs when C = 0.0002.
4. 1000C ′ = −0.5C, C(0) = 5/1000. Then C(t) = (1/200) exp(−t/2000).
6. The constant solutions occur when C ′ = 0, or when qCin −qC −kV C 2 = 0.
Solve the quadratic for C to find physically meaningful constant solutions.
8. The equation is V C ′ = −kV C, C(0) = C0 . So C(t) = C0 exp(−kt). The
residence time T satisfies T = −(1/k) ln(0.9).

3
 Section 1.4.1, page 41
2. (a) x(t) = C/t2 + t2 /4. (c) x(t) = C exp(−t2 ) + t exp(−t2 ). (f ) x(t) =
C(t2 + 1)−3/2 + 2.
4. The integrating factor is exp(at); write the integral with variable upper
limit.
Rt
6. Write the integrating factor as exp( 1 e−s /s ds).
8. Write the equation as y ′ = 1 + x′ = 1 + y 2 , so y(t) = tan(t + C).

10. The integrating factor is e−pt . So, (xe−pt )′ = e−pt q(t). Integrate from t0
Rt
to t to get x(t) = x0 ept + ept t0 e−ps q(s)ds.

12. (x1 /x2 )′ = (x2 x′1 − x2 x′1 )/x22 . The numerator is zero because x1 and x2
satisfy the equation, and so x1 /x2 = C, a constant.
14. In changing variables y = x1−n , note that y ′ = (1 − n)x−n x′ , where the
chain rule is used.
15. (e) Let y = x−2 , so y ′ = −2x−3 x′ . Writing the equation in terms of y
gives y ′ = −2ay
p− 2b, which is linear. Clearly y = Ce
−2at
− b/a. Thus
√
x = 1/ y = 1/ C exp(−2at) − b/a.
16. If Ht = f and Hx = g, then Htx = fx = Hxt = gt . (ii) fx = 1/t and
gt = 1/t, so the equation is exact. Now find H. Note Ht = t3 + x/t and
Hx = x2 + ln t. Integrating the first gives H = (1/4)t4 + x ln t + φ(x). Now
Hx = x2 + ln t = φ′ (x) + ln t. So φ′ (x) = x2 , or φ(x) = (1/3)x3 . Then the
integral curves are H(t, x) = (1/4)t4 + x ln t + (1/3)x3 = C.
 Section 1.4.2, page 46
2. Constant solutions, or equilibria, are found by setting S ′ = 0, or (a +
rA/M )S = rA. So S = rA/(a + rA/M ).
4. T ′ = −3(T − (9 + 10 cos(2πt))). The solution is
1
T (t) = 3e−3t + 9 + (90 cos(2πt) − 90e−3t + 60π sin(2πt)).
4π 2 + 9

6. For t < T the equation is S ′ = −(a + rA/M )S + rA, and for t > T the
equation is S ′ = −aS. The solution to the first is S1 (t) = C1 e−(a+rA/M)t +
rA/(a + rA/M ), t < T. The solution to the second is S2 (t) = C2 e−at ,
t > T. Use the conditions S1 (0) = S0 and S1 (T ) = S2 (T ) to obtain C1
and C2 .
8. Note that S ′ = I(1 − S/P ) − ES/P . (a) Set S ′ = 0 and solve for S to
get the longtime number of species. (b) The equation is linear, so use an
integrating factor.

4
 (a) (d) (f)

0 2 x 0 8 x 0 3/4 x

Figure 1: Phase line diagrams for Sec. 1.5.1, Exercises 2adf.

10. The equation becomes V C ′ = −qC −kV C 2 , which is a Bernoulli equation.

To solve, make the transformation y = C −1 , giving y ′ = −C −2 C ′ . Write
the linear equation for y as V y ′ = 2qy +2kV ; the solution is y = c1 e2qt/V −
kV /q. Then C = 1/y.
12. Both equations are Bernoulli equations and can be solved by using the
transformation y = X −1 , as in Exercise 10.
 Section 1.4.3, page 53
2. The equation is 20Q′ + 100Q = 200e−5t , Q(0) = 0. Solving, Q(t) =
10t exp(−5t). To find the maximum charge set Q′ (t) = 0 and solve for t
to get t = 1/5.
4. The circuit equation is LI ′′ + RI ′ + (1/C)I = E ′ (t). By Kirchhoff’s law
I ′ (0) = (1/L)(E(0) − RI0 − Q0 /C).
6. The equation is LI ′′ + RI + I/C = −Aω sin ωt. The initial conditions are
I(0) = I0 , and I ′ (0) given by Exercise 4.
8. I ′′ + 21 13 I 3 − I + I = 0.

 Section 1.5.1, page 63

2. (a) x′ = x2 (2 − x). Equilibria are x = 0, 2. For stability check fx (x) =
2x(2 − x) − x2 . Now fx (2) = −4 < 0, so x = 2 is stable. f( 0) = 0 so at
x = 0 there is no information. The phase line plot in Figure 1 shows it is
unstable. (d) x′ = x(x − 8)3 . The equilibria are x = 0, 8. For stability,
check fx (x) = 3x(x − 8)2 + (x − 8)3 . Now, fx (0) = −83 < 0, so x = 0
is stable. fx (8) = 0, so at x = 8 there is no information. But the phase
line plot in Figure 1 shows it is unstable. (f ) x′ = 2x(1 − x) − 21 x. The
equilibria are x = 0, x = 3/4, which are unstable and stable, respectively.

4. The equilibria are P = 0, a, K which are stable, unstable, and stable,

respectively. If P < a, i.e., P is small, then the population becomes extinct
for lack of mating partners; if a < P < K or P > K, the population
approaches the carrying capacity K. See Figure 2.

5
 P L
a/b

t
0 a K t

Figure 2: Phase line diagrams for Sec. 1.5.1, Exercises 4 (left) and 8 (right).

6. The function f (T ) at the equilibrium has negative slope at the equilib-

rium T = S which is stable. The approximation leads to the differential
equation T ′ = −kT 4 , which has solution T (t) = [3kt + 2000−3]−1/3 .
8. If L is the length and V = L3 is the volume, then V ′ = 3L2 L′ and the
growth equation is V ′ = α(6L3 ) − βx3 , or L′ = a − bL for constants a and
b. If L(0) = 0 the solution is L(t) = ab (1 − exp(−bt). As t → ∞, L → a/b,
consistent with the fact that organisms reach a maximum length. See
Figure 2.
10. I ′ = aSI = aI(N − I) = aN I(1 − I/N ), which is the same as the logistic
equation. As t gets large the solution approaches N , so everyone eventually
gets the disease.
 2

12. P ′ = α Pa − bP = α a−bP

p
P . The equilibrium price is P = a/b
and it is stable. Note the graphs of a/P and bP must intersect at the
equilibrium. To solve the differential equation separate variables and write
P dP/(a − bP 2 ) = αdt; use substitution to calculate the integral.
14. For part (a), we use the chain rule to get

d
V (x(t)) = V ′ (x(t))x′ (t) = −f (x(t))f (x(t)) < 0.
dt
For (b), if V has a local minimum at x∗ , then in a small neighborhood of
x∗ , V ′ (x) ≤ 0 for x < x∗ and V ′ (x) ≥ 0 for x > x∗ . Thus f (x) > 0 for
x < x∗ and f (x) < 0 for x > x∗ . So, f ′ (x∗ ) < 0, and x∗ is stable.
 Section 1.5.2, page 70
√
1. (d) The equilibria are given by x = 1 and x = ± h > 0. See Figure 3.
2. x = 0 is the only equilibrium and fx (0) = 1 > 0, so x = 0 is unstable.
4. The equilibria are x = 0 and x = h − 1. Note h > 0 and x ≥ 0. Using
the derivative criterion, x = 0 is unstable for h < 1 and stable for h > 0;
x = h − 1 is unstable.

6
 x s
s us

us

h
s

Figure 3: Bifurcation diagram for Exercise 1d, Sec. 1.5.2, showing stable (s)
and unstable (us) branches.

6. x′ = x3 − x + h. Setting x′ = 0 gives the parameter h = x − x3 =

x(1 − x)(1 + x) in terms of the equilibria x. This is an easy to plot
cubic. One can rotate the graph through 90 degrees to get x vs. h, but
it is not√necessary. To √check stability, note fx (x) = 3x2 − 1 > 0 when
x > 1/ 3 or x < −1/ 3; These √ equilibria
√ are unstable. x is stable if
fx (x) = 3x2 − 1 < 0, or −1/ 3 < x < 1/ 3.
8. Note a > 0√is necessary to have an equilibrium, which are clearly given
by x = ±1/ a. These upper and lower branches of the equilibrium curves
are unstable and stable, respectively, because fx (x) = 2ax.
10. (c) See Figure 4. Equilibria are found graphically at intersection points of
p−a and the sigmoid curve ρpn /(1+pn ). If p−a > ρpn /(1+pn ) the phase
arrow it positive (to the right); if p − a < ρpn /(1 + pn ) the phase arrow it
positive (to the left). This determines the stability of the equilibrium or
intersection point.
 Section 1.5.3, page 76
2. f (t, x) and fx (t, x) are are continuous everywhere except on the vertical
lines t = ±1. Therefore the problem has a unique solution in an interval
contained in (−1, 1).
4. (a) The problem has singularities at t = 0, 5. The initial condition is given
at t = 2. The largest interval where a solution exists is (0, 5). (b) The
largest interval where a solution exists in (3, 7).
6. (a) The implicit solution to x′ = −4t/x, x(0) = x0 , is x2 + 4t2 = C,
2 2 2
which ispan ellipse in the tx plane. Now, x(0) = x0 gives x + 4t = x0 , or
2
x = ± x0 − 4t2 . If x0 > 0 then the solution is the upper branch of the
ellipse (with the + sign), and if x0 < 0 the solution is the bottom portion
of the ellipse (with the - sign). (c) The solution is x = 1/(1/x0 − t). If
x0 > 0 the solution is valid on (−∞, 1/x0 ); if x0 < 0 the solution is valid
on (1/x0 , ∞).

7
 p-a p*
p’
unstable stable
B

A
stable
-a p a
bifurcation diagram
-a

Figure 4: Exercise 1.5.2 Exercise 10. Left : Plots of p − a for two values of a
and the sigmoid curve ρpn /(1 + pn ). Equilibria are at intersection points. As a
increases from a small to large value the straight line p − a moves downward; at
first a small equilibrium, then a bifurcation occurs and two equilibria appear at
the upper tangency point; then there are three, then two at the lower tangency
point, and then one at the upper intersection. Right : Bifurcation diagram
showing how p∗ changes as a increases. The upper and lower branches are
stable and the middle branch is unstable. As the inflow rate a increases and
the equilibrium reaches the point A, it jumps quickly to the upper branch at B
(dashed) leading to eutrophication or high phosphorus content.

8. Both parts of the function clearly satisfy the differential equation and
the initial condition. But the function f (t, x), as well as fx (x, t), are not
continuous along t = 0, where the initial condition is given. Because the
problem does not satisfy the hypotheses of the theorem, nothing can be
said one way or the other. There is no contradiction.

Chapter 2 Exercises
 Sec. 2.1, page 84
1
2. Multiply the equation by Q′ to get LQ′ Q′′ + RQ′ Q′ + C QQ
′
= 0, which
by the chain rule is the same as
 
d 1 ′ 2 1 2
L(Q ) + Q = −RQ′ Q′ .
dt 2 2C
The energy in the inductor is 12 L(Q′ )2 or 21 LI 2 , and the energy on the
1
capacitor is 2C Q2 ; the power lost is −RI 2 . If there is no resistor then
energy is conserved.
 Sec. 2.2.2, page 90
2. (a) x(t) = e2t − 2te2t .
(c) x(t) = e−t (1 + t).
(d) x(t) = 23 e−t − 21 e−3t .

8
 4. Only (c) and (d), because a physical application to oscillators and circuits
require nonnegative coefficients.
6. The eigenvalues are 4 and −6 so the characteristic polynomial must be
(λ − 4)(λ + 6) = 0, or λ2 + 2λ − 24 = 0. The equation is x′′ + 2x′ − 24x = 0.
8. The general solution is x(t) = (A + Bt)e−at . Assume that x(t1 ) = x(t2 ) =
0, t1 6= t2 . Then x(t1 ) = A + Bt1 = x(t2 ) = A + Bt2 = 0. Therefore,
t1 = t2 , a contradiction.
 Sec. 2.2.3, page 94
 √ √ 
2. (a) x(t) = e−t/2 cos 152
t
+ 1
15 sin 15 t
2 .
√ √ √

(b) x(t) = exp(2t) cos 2 t − 2 sin 2 t .
(c) x(t) = cos 3t.
4. The roots of the characteristic equation must be ±5 so λ2 − 25 = 0, giving
x′′ − 25x = 0. Initial conditions give c1 = 2, c2 = 0.
6. It involves the 5 ‘most famous’ numbers 0, 1, e, π, i.
 Sec. 2.2.4, page 99
 √ √ √ 
2. x(t) = e−t/16 2 cos(( 255/16)t) + 2 255 255
sin(( 255/16)t) . The ampli-
√
tude is A = 2.004 , the frequency is 255/16, and the phase is φ = 0.0625.
1
√
4. The eigenvalues are λ = 2L (−1 ± 1 − 4L). If L > 1/4 the the roots are
complex giving a decaying oscillation. If L < 1/4 the roots are real and
negative giving decay.
6. The circuit equation is 5I ′′ + 21 I = 0. The solution is I(t) = cos √t10 +
√ √ √
10 cos √t10 . The circuit oscillates. We have A = 11, ω = 1/ 10, and
φ = 1.26, giving
√
 
t
x(t) = 11 cos √ − 1.26 .
10
 Sec. 2.3.1, page 110
1. (a) At2 + Bt + C (e) A sin 7t + B cos 7t. (e) At + B + Ce−t .
√ √ t 1
2. (a) x(t) = c1 cos( 7 t) + c2 sin( 7 t) + 16 exp(3t) − 128 exp(3t).
(b) x(t) = c1 + c2 et + 21 e2t − 6 − 6t.
(f) x(t) = c1 e−t + 4te−t .
1
(g) x(t) = c1 e2t + c2 e−2t − 8 cos 2t.
11 8t 1 −5t 1 −t
4. x(t) = 117 e − 26 e − 18 e .

9
 √
6. The differential equation is x′′ + 2x = cos 2 t. The solution with initial
conditions is
1 √ √
x(t) = (t + 2) 2 sin 2 t.
4
5
8. y(t) = 2 − e−t − 12 e−2t .
10. The circuit equation is 0.4Q′′ + 100Q = 5e−2t , Q(0) = Q′ (0) = 0. The
charge is
√
25 −2t 25 √ 10 √
Q(t) = e − cos(5 10 t) + sin(5 10 t).
508 508 508
The first term is the transient that decays away leaving the final two terms,
which represent an oscillatory steady state.
 Sec. 2.3.2, page 114
2. Using the initial conditions we get c1 = 0 and c2 = −B/ω0 .
4. The circuit equation is LQ′′ + C1 Q = V0 sin βt. Resonance occurs when the
natural frequency equals β, or
r
1 1
= β or L = 2 .
LC β C

6. The solution is

x(t) = 50 sin 2t − et/200 (50.0002 sin 1.999t) .

 Sec. 2.4.1, page 120

√
1. (b) The indicial equation is m2 − m − 4 = 0 having real roots 21 (1 + 17).
√ √
The general solution is x(t) = c1 t1/2+ 17/2 + c2 t1/2− 17/2 .
(e) The characteristic equation is m2 − 8m + 16 = 0 having roots 4 ± 2i.
The general solution is x(t) = t4 (c1 cos(2 ln t) + c2 sin(2 ln t))).
2. This is not a Cauchy-Euler equation. Hence, let y = x′ so that y ′ +t2 y = 0.
This equation is separable and has solution y(t) = exp(−t3 /3), where we
Rt
have used y(0) = 1. Therefore x(t) = 0 exp(−r3 /3)dr.
4. The equation can be written (a(t)x′ )′ = f (t), which when integrated gives
Z t
1 1
x′ (t) = f (r)dr + c1 .
a(t) α a(t)
Integrating again,
Z t Z s  Z t
1 1
x(t) = f (r)dr ds + c1 ds + c2 .
α a(s) α α a(s)

10
 Sec. 2.4.2, page 124
2. This is a Cauchy-Euler equation. The indicial equation for the homo-
geneous equation is m2 = 0, so m = 0, 0, giving independent solutions
x1 (t) = 1 and x2 (t) = ln t. The Wronskian is t−1 . Therefore the particu-
lar solution is
a ln t 1·a a
Z Z
xp (t) = − dt + ln t dt = t2 .
t−1 t−1 4
The general solution is x(t) = c1 + c2 ln t + a4 t2 .
4. (a) We have

W ′ (t) = (x1 x′2 + x′1 x2 )′ = x1 x′′2 − x2 x′′1

= x1 (−px′2 − qx2 ) − x2 (−px′1 − qx1 ) = −pW (t).

(b) Separate variables

R to get dW/W = −p(t)dt and integrate to get
W (t) = C exp(− p(t)dt). If C = 0 then W (t) ≡ 0 for all t, and if C =
6 0,
then W (t) is never zero.
(c) Assume x1 (a) = x1 (b) = 0. Then x′1 (a) and x′1 (b) have opposite signs.
Now, W (t) = x1 (t)x′2 (t) − x′1 (t)x2 (t); so that W (a) = −x′1 (a)x2 (a) and
W (b) = −x′1 (b)x2 (b). These two expressions must have the same sign, so
x2 (a) and x2 (b) must have opposite signs. Therefore x2 (t) must equal zero
at some point between a and b.
Rt
6. xp (t) = emt 0 (t − s)e−ms f (s)ds.
 Sec. 2.4.3, page 125
2. Let x2 = tv and substitute into the equation to get tv ′′ + (2 − t2 )v ′ = 0.
Letting y = v ′ we get a separable equation dy/y = (t − 2/t)dt. Integrating,
we get ln y = 12 t2 − 2 ln t. Hence,

1 2 1 t2 /2
Z
y(t) = 2 et /2 , giving x2 (t) = t e dt.
t t2
Note: this solution can be written in different ways using special functions
defined by integrals.
5. Let x2 = tv. then x′2 = tv ′ + v and x′′2 = tv ′′ + 2v ′ . The equation reduces
to v ′′ − v ′ = 0, having solution v = et . Therefore, a second independent
solution is x2 (t) = tet .
7. Hint: to solve the equation for z separate variables and integrate.
8. In Exercise 7 let y(t) = et ; we know z(t)et = x2 (t) = cos t. Thus
z(t) = e−t cos t. Substitute this into the z equation to get p(t) = (sin t −
cos t)/ cos t. Return to the original equation for x and then compute
q(t) = −1 − p(t).

11
 Sec. 2.5, page 130
1. (b) The characteristic equation is λ4 + λ = λ(λ3 + 1) = 0. The second
factor has root λ = −1 and so the characteristic polynomial √
turns into
λ(λ + 1)(λ2 − λ + 1) = 0, having roots λ = 0, −1, 21 ± i 23 . The general
solution to the homogeneous equation is
√ √
xh (t) = c1 + c2 e−t + et/2 (c3 cos 3 t/2 + c4 sin 3 t/2).

We guess a particular solution as xp (t) = At which upon substitution gives

A = 1. Thus, x(t) = xh (t) + t.
2. The characteristic equation is λ3 + λ2 − 4λ − 4 = 0. Easily, λ = −1 is a
root so the polynomial factors into (λ + 1)(λ2 − 4) = 0, giving eigenvalues
−1, 2, −2. The general solution is therefore x(t) = c1 e−t + c2 e2t + c3 e−2t .
Applying the initial conditions gives three equations with three unknowns
c1 , c2 , c3 , and we find that c1 = 5/3, c2 = 1/12, c3 = −3/4.
4. The characteristic polynomial is (λ2 + 1)(λ2 + 1) = 0 giving the differential
equation x′′′′ +2x′′ +x = 0. The general solution is x(t) = A cos t+B sin t+
t(C cos t + D sin t).
 Sec. 2.6, page 135
2. u(x) = −x3 /6 + x4 /240 + 100x/3. The rate heat leaves the right end is
−Ku′ (20).
4. There are no nontrivial solutions for λ ≤ 0. When λ = n2 π 2 , n = 1, 2, 3, . . .
we get solutions un (x) = sin nπx.
√
6. The characteristic equation is p2 + 2p + λ = 0 with roots p = −1 ± 1 − λ.
If λ ≤ 1 then the solutions are exponential and cannot√satisfy the bound-
ary conditions; so λ√> 1. The roots are√then p = −1 ± λ − 1 i. Therefore
u(x) = e−x (A cos( λ − 1 x) +√ B sin( λ − 1 x). Now, u(0) = 0 implies
A = 0. Then u(x) = Be−x sin( √λ − 1 x). Apply the right √ boundary con-
dition to get u(1) = Be−1 sin( λ − 1) = 0. Therefore λ − 1 = nπ,
n = 1, 2, , 3, . . .. Thus the eigenvalues are λn = 1 + n2 π 2 with un (x) =
e−x sin nπx, n = 1, 2, 3, . . . .

Chapter 3 Exercises

 Section 3.1, page 144

R∞ R∞ 1
2. F (s) = 0 e−3t H(t − 2)e−st dt = 2 e−(3+s)t dt = 3+s e−2(3+s) .
R∞ R∞
6. L[f (t)H(t−a)] = a f (t)e−st dt = 0 f (r+a)e−(r+a)s ds = e−as Lf (r+a).
 
5
8. (a) 6s + s+2 1
+ s−3) 2. (c) cos(5t). 2
(e) 23 1s + s+1 .

12
 √ q
3√2 7
9. (c) 2t2 e5t . (d) 7H(t − 4). (g) 2 7
sin 2 t.

10. (b) Use sin t−π/2) = − cos t. (c) L[H(t−π/2) sin(t−π/2)] = e−πs/2 s21+1 .
1 −2(s+1)
12. By the shift property, X(s) = s+1 e .
2
14. x(t) = et is not of exponential order,
R ∞ 1 i.e., it is not bounded by M ert
−st
for any r. The improper integral 0 t e dt does not exist at t = 0.
X(s) = 1s es does not go to zero as s → ∞ so its transform does not exist.
16. Letting st = r2 , so that sdt = 2rdr gives
√
Z ∞ Z ∞
1 2
√ e−st dt = 2 e−r dr = π.
0 t 0

18. Let Lf = F and Lg = G. Then

L−1 [aF + bG] = L−1 [aLf + bLg]

= L−1 L[af + bg] = af + bg = aL−1 F + bL−1 G.

 Section 3.2, page 156

2. (a) Use cos t = − cos(t − π). Then −L[H(t − π) cos(t − π)] = − s2s+1 e−πs .
(b) Use the Table to get 3 · 4!/s5 .
(e) Use the Table to get 3/((s + 6)2 + 9).
R∞
(f) 0 ea+bi−s)t dt = −1/(a − s + bi) provided s > a.
4. In multiline form


 0, t < 1;
t − 1, 1 ≤ t < 3;


 t − 3, 3 ≤ t < 4;
t − 3 + e−t/2 , t ≥ 4.


1 2
6. (b) The differential equation transforms to X(s) = s+1 s2 +4 . Using partial
fractions we get x(t) = 51 (e−t − 2 cos 2t + sin 2t).
1 1 1
(e) The differential equation transforms to X(s) = (s−1) 2 +1 + s+1 (s−1)2 +1 .
1 −t 1 t
Using partial fractions x(t) = 5 e − 5 e (cos t + 7 sin t).
(i) The differential equation transforms to X(s) = s2s−2 + 1 1
s s2 −2 . Partial
√ √
fraction gives x(t) = 43 (exp( 2 t) + exp(− 2 t) − 21 .
8. L[t2 H(t − 3)] = e−3s L[(t + 3)2 ] = e−3s (2//s3 + 6/s2 + 9/s).
10. f (t) = t − (t − 4)H(t − 4).

13
 1 2 −s
12. Taking Laplace transforms of the equation gives X(s) = s−1 − s(s−1) e .
Inverting, x(t) = et − H(t − 1)(et−1 − 1).
1 1
14. Taking Laplace transforms of the equation gives X(s) = s+1 + s(s+1) e−s −
1
s(s+1) e
−2s
. Inverting gives x(t) = e−t e1−t H(t − 1) − e2−t H(t − 2).

16. Write
Z ∞ ∞ Z
X (n+1)p
F (s) = f (t)e−st dt = f (t)e−st dt
0 0 np
∞ Z
X p ∞ Z
X p
= f (r + np)e−s(r+np) dr = f (r)e−nsp e−sr dr
0 0 0 0
∞
!Z
p p
1
X Z
−nsp −sr
= e f (r)e dr = f (r)e−sr dr.
0 0 1 − e−ps 0

18. f (t) = 2H(t) − H(t − 1) + H(t − 2) − H(t − 3) + H(t − 4) + · · · .

20. Take the s-derivative of F (s) by bringing the derivative under the integral
sign to get
Z ∞ Z ∞
F ′ (s) = f (t)(−te−st )dt = (−tf (t))e−st dt = −L(tf (t)).
0 0

22. Proceed as in Exercise 20 by bringing the s-derivatives under the integral

sign to get
Z ∞ Z ∞
d ′
′′
F (s) = F (s) = −st
(−tf (t))(−te )dt = t2 f (t)e−st dt = L(t2 f (t)),
ds 0 0

and so on. Each derivative increases the power on t by one, and they
alternate in sign.
24. Taking transforms of the equations gives (s − 1)X + 2Y = 1, −3X + (s −
1)Y = 0. Solving for X and Y gives
s−1 3
X(s) = 2
, Y (s) = .
(s − 1) + 6 (s − 1)2 + 6
√ √ √
Therefore x(t) = et cos 6 t, y(t) = (3/ 6)et cos 6 t.
 Section 3.3, page 162
1 1 1
2. Take x(t) = t and y(t) = 1. Then L(t) = s2 ; yet L(t)L(1) = s2 s .
1 1
4. (a) Note that f (t) = et ⋆ t, so L(f ) = L(et )L(t) = s−1 s .

14
 6. Take the transform of the equation to get s2 X(s) − ω 2 X(s) = F (s). Thus
1
X(s) = s2 −ω 2 F (s). By convolution

t
1 1
Z
x(t) = sinh ωt ⋆ f (t) = sinh ω(t − τ )f (τ )dτ.
ω ω 0

1
8. Taking transforms, s2 X + 3sX + 2X = s+4 . Thus

1 1
X(s) = .
(s + 1)(s + 2) s + 4
Therefore,
   
−1 1 −1 1
x(t) = L ⋆L .
(s + 1)(s + 2) s+4

The two inverse transforms on the right are found in the Table.
10. Use the convolution theorem to invert the equation and get
Z t Z t
x(t) = H(t − 3) ⋆ f (t) = H(τ − 3)f (t − τ )dτ = f (t − τ )dτ.
0 3

Rt
12. L−1 s2 (s12 +1) = t ⋆ sin t = 0
τ sin(t − τ )dτ.

14. Take the transform of the equation using convolution on the integral to
get X(s) = F (s) + K(s)X(s). Thus X(s) = F (s)/(1 − K(s)).
16. The integral term in the equation is the convolution t2 ⋆ x(t). Thus, taking
the transform of the equation gives sX(s) − s13 X(s) = − s12 . Therefore

s s 1
X(s) = − =− 2 .
s4 − 1 s + 1 s2 − 1
The first factor on the right inverts to cos t and the secondR converts
t
to sinh t. Therefore by the convolution theorem, x(t) = − 0 cos(t −
τ ) sinh τ dτ . The integral can be found by writing cos and sinh in terms
of exponential functions.
 Section 3.4, page 173
2. Taking the transform of the equation and solving for X(s) gives
1 1 −s 1
X(s) = + e + e−4s .
s+3 s+3 s(s + 3)
Therefore
1 
x(t) = e−3t + e−3(t − 1)H(t − 1) − 1 − e−3(t−4) H(t − 4).
3

15
 1 −2s
4. Taking the transform and solving for X(s) gives X(s) = s2 +1 e . There-
fore x(t) = sin(t − 2)H(t − 2).
6. Taking the transform and solving for X(s) gives X(s) = s21+1 e−2s −
1 −5s
s2 +1 e . Thus x(t) = H(t − 2) sin(t − 2) − H(t − 5) sin(t − 5).

8. The differential equation is x′′ + x = ∞

P
n=0 δnπ (t) with zero initial condi-
tions. Taking the transform and solving for X(s) gives
∞ ∞
1 X −nπs 1 1 X −nπs
X(s) = e = + e .
s2 + 1 n=0 s2 + 1 s2 + 1 n=1

Thus ∞
X
x(t) = sin t + sin(t − nπ)H(t − nπ).
n=1

10. Solving for Y (s) gives

p
1 −s
p 3/4 −s
Y (s) = 2 e = 4/3 2e .
s +s+1 1 2
p
(s + 2 ) + 3/4

Therefore, from the Table,

p p 
y(t) = 4/3 sin 3/4 (t − 1) e−(t−1)/2 .

Chapter 4 Exercises

 Sec. 4.1, page 190

1. (d) The orbits are on y = −2x. As t ranges over −∞ to ∞ the orbit

goes from infinity to the origin along the straight line y = −2x in the 4th
quadrant.
2. (a) The orbits are the ellipses x2 + 3y 2 /2 = C. By eliminating y from
the differential equations we get x′′ +√6x = 0, which √has solution x(t) =
√ √ √ √
c1 cos 6 t + c2 sin 6 t. Then y(t) = 36 c1 sin 6 t − 36 c2 cos 6 t. So,
√ ! √ !
cos 6 t sin 6 t
x(t) = c1 √
6
√ + c2 √
− 6
√ .
3 sin 6 t 3 cos 6 t

3. (c) The system is equivalent to x′′ − x′ − 2x = 0 which has x(t) = c1 e−t +

c2 e2t . Then y(t) = 21 (x′ − x) = 21 (−2c1 e−t + c2 e2t . Then
 −t   2t 
e e
x(t) = c1 + c2 1 2t .
−e−t 2e

16
 4. The y equation becomes y ′ = βx − (α + γ)y + k.

 Sec. 4.2.1, page 198

 
−1 −2 3/2
2. We have A = . The solution to the systems is x = −2.5,
1 −1/2
y = 1.5. Geometrically, the solution is the intersection of the two straight
lines x + 3y = 2, 2x + 4y = 1.
4. For Exercise 3(a) λ = −1, 7. For Exercise 3(b), when λ = −1 there are
infinitely many solutions lying on the line 3x + 8y = 0; when λ = 7 there
are infinitely many solutions x = 0 (the entire y axis).
6. (a) The determinant is zero; there is no solution unless m = −5/3, in which
case there are infinitely many solutions lying on the line −6x − 9y = 5.
(b) The determinant is 2 − m. There is no solution if m = 2, and there is
a unique solution when m 6= 2.
 
2 −4
8. We have det = 4 6= 0. Therefore the vectors are linearly
−3 8
independent.
 Sec. 4.2.2, page 201
2. (a) (0, 0). (b) (0, 0). (e) The line x − 2y = 0.
3. (a) The critical point is x∗ = −14, y ∗ = 28/3. Make the transformation
z = (x + 14, y − 28/3)T to get the linear system z1′ = 2z1 + 3z2 , z2′ = −z1 .
 Sec. 4.3, page 206
2. (a) The characteristic
√ equation is λ2 − 2λ + β = 0. The eigenvalues are
λ = 1 ± 1 − β. If β < 1 the eigenvalues are real and unequal, if β = 1
there is a double real eigenvalue, and if β > 1 the eigenvalues are complex
numbers with positive real part.
p
(b) The eigenvalues are λ = 2 ± 1 − 2β 2 . If β 2 > 12 the eigenvalues are
complex; if β 2 < 21 the eigenvalues are real and unequal. If β 2 = 12 we
obtain a double real eigenvalue 2, 2.
4. Assume Ax = λx. Multiply on the left by A−1 to get x = λA−1 x, or
A−1 x = λ−1 x.
6. Yes, zero is an eigenvalue for any matrix with zero determinant.
 Sec. 4.4.1, page 217
1. (a) Both eigenvalues are real and negative so the origin is an asymptoti-
cally stable node. The general solution is
   −2t 
−e−t e
x(t) = c1 + c2 .
2e−t 2e−2t

17
 2. (b) The origin a saddle. The general solution is
   t 
0 5e
x(t) = c1 + c 2 .
e−4t 3et

4. x(t) = 14e−2t − 11e−4 , y(t) = 21e−2t − 11e−4 .

 Sec. 4.4.2, page 220
1. (a) The general solution is
   
cos t sin t
x(t) = c1 + c2 .
− sin t cos t
√
2. The eigenvalues are λ = 3/2 ± i 3/2, so the origin is an unstable spiral.
The vector field is vertical crossing the line y = x and horizontal crossing
the line y = −x/2. At the point (0, 1) we have x′ = −1, y ′ = 2, which is
NW, indicating a counterclockwise orbit.
 Sec. 4.4.3, page 224
1. (a) The general solution is

x(t) = (c1 + c2 t)e−3t , y(t) = c2 e−3t .

 Sec. 4.5, page 235

4. (a) The origin is a saddle point.
   2 4t 
−e−t 3e
x(t) = c1 + c 2 .
e−t −2e4t

(b) The origin is an asymptotically stable node.

1
x(t) = (c1 + c2 t)e−3t , y(t) = c2 e−3t .
4

(f) The origin is an unstable spiral point.

1
x(t) = − et (−c1 (cos 2t + sin 2t) + c2 (cos 2t + sin 2t))
2
y(t) = et (c1 cos 2t − c2 sin 2t).

6. There is a line of equilibria y = x/2. (b) The eigenpairs are 0, (2, 1)T , and
5, (1, −2)T . There is one linear orbit, x(t) = (1, −2)T e5t . The general
solution is
e5t
   
2
x(t) = c1 + c2 .
1 −2e5t
The phase diagram shows the line of equilibria with parallel lines coming
out of the equilibria of slope −2.

18
 8. (a) The characteristic equation is λ2 + 5α − 4 = 0. If 5α − 4 > 0 the
eigenvalues are purely imaginary and the origin is a center. For 5α − 4 < 0
the eigenvalues are real and unequal and the origin is a saddle point. Thus
there is a bifurcation at α = 4/5.
10. The trace is T = a and the determinant is D = −2 + a2 /4. Plotting D
vs T gives a√parabola in the T D plane which is concave
√ up, and passes
through (± √8, 0) and (0,
√ −2). Therefore, for a < −√ 8 we have a stable
node, for − 8 < a < 8 a saddle, √ and for a > − 8 an unstable node.
Thus, bifurcations occur at a = ± 8.

12. No. The trace of the matrix is nonzero.

14. (d) There are no equilibria. The general solution is
1 3 7 9 7
x(t) = −c1 − c2 e−2t − + t, y(t) = c1 + c2 e−2t + − t.
3 4 2 4 2

16. The origin is a stable spiral point, so the current and voltage go to zero.
The general solution is
√
2 √ √
I(t) = − (c1 cos( 2t) + c2 sin( 2t))e−t
2 √ √
V (t) = (−c1 sin( 2t) + c2 cos( 2t))e−t

 Sec. 4.6, page 243

2. The system is

(V1 C1′ ) = (q + r)c − rC1 − qC1 , (V2 C2 )′ = qC1 − qC2 .

Initially, C1 = C2 = 0 at t = 0.
4. The fundamental matrix and its inverse are
2e4t e4t
 −4t   
3e −e−11t −1 1
Φ= , Φ = .
e −4t
2e −11t
7 −e11t 3e11t

For undetermined coefficients, take xp (t) = (Ae−t , Be−t )T . Substitute into

the system to get A = 3/10, B = 1/15.
8. The only equilibrium is (0,0). The x nullcline y = (r1 + r3 )/r2 lies above
the y nullcline y = r1 /r2 . In the region under the y nullcline the vector
field is NW, between the nullclines it is SW, and above the x nullcline
it is SE. This clearly shows the origin is an asymptotically stable node.
(Analytically one can show both eigenvalues are negative.) Therefore the
solution beginning at (x0 , 0) approaches zero as t → ∞.

19
 10. The model equations are

E ′ = −hE + bP, P ′ = hE − mP.

If A is the coefficient matrix then trA = −h−m < 0 and det A = h(m−b).
If m > b the det A > 0 and the origin is asymptotically stable; thus the
population dies out. If m < b then det A < 0 and the origin is a saddle
point. The population grows without bound. If eggs are eaten at the
constant rate ρ, then the equations become

E ′ = −hE + bP − ρ, P ′ = hE − mP.

The equilibrium is now

ρm ρ
E= P = ,
h(b − m) b−m

which is viable only if b > m. In this case the population approaches the
coexistent equilibrium state. In this case the equilibrium is a saddle and,
depending on the initial conditions, orbits go to infinity or go to the line
E = 0 (no eggs). If b < m there are no equilibria and the egg population
dies out.

Chapter 5 Exercises

 Sec. 5.1, page 257

 
1 − y + 2hx −x
2. Note that the Jacobian matrix is J(x, y) = . In
y x−1
all cases, regardless of the value of h, the Jacobian matrix shows that the
origin (0, 0) is a saddle point. The two other critical points are (−1/h, 0)
and (1, h + 1). The Jacobian matrix at those two critical points in the
cases h = 1, −1, 8 give the results, which we leave to the reader. The
nullclines show that orbits are vertical along x = 0 and y = 1 + hx and
horizontal along y = 0 and x = 1.

4. The x nullcline is the circle x2 + y 2 = 4 of radius 2; the y nullcline√ is the

√
straight line
√ y = 2x.
√ These intersect at the two equilibria (2/ 5, 4/ 5
and (−2/ 5, −4/ 5. To get the direction field note x′ > 0 outside the
circle and x′ < 0 inside the circle; y ′ > 0 √
above the ′
√ straight line and y < 0
below the straight
√ line.√ This shows (−2/ 5, −4/ 5 is a saddle point. The
orbits near (2/ 5, 4/ 5 have a circular pattern, so it is not clear if that
critical point is a center or a spiral.
6. Setting sin y = 0 and x = 0 gives equilibria (0, nπ), n = 0, ±1, ±2, . . . , all
along the y axis.

20
 8. Dividing the two equations and separating variables gives (1/y)dy = ex /(ex −
1). Integrating yields the orbits y = C(ex − 1). The only critical point
is (0, 0) and examining the Jacobian matrix shows it is an unstable node
with eigenvalues 1, 1.
10. By checking the Jacobian matrix, the critical points are (0, 0) (saddle
point) and (2, 0) (asymptotically stable node).
12. The critical points are (1, 0) (saddle point) and (−1, 0) (stable spiral).
The orbits are vertical along y = 0 and horizontal along the parabola
y = x2 − 1.
 
0 1
14. The equilibrium is (0,0) and the Jacobian matrix is J(0, 0) = .
−1 −1
The trace is −1 (negative) and the determinant is +1 (positive), so by the
theorem (0, 0) is asymptotically stable.

 Sec. 5.2, page 269

2. V (x) = − F (x)dx = 12 x2 − 14 x4 . Conservation of energy is
R

1 2 1 2 1 4
y + x − x = E, x′ = y.
2 2 4
At x = 2, x′ = 1 when t = 0 we get E = −3/2. The orbit is given by
√ p
y = ± 2 −3/2 − x2 /2 + x4 /4.

4. V (x) = − F (x)dx = x3 /3. Conservation of energy is 12 y 2 + 31 x3 = 31 ,

R
where we used the
p initial
√ conditions to obtain E = 1/3. Solving for y gives
the orbit y = ± 2/3 1 − x3 . Replacing y by dx/dt, separating variables,
and then integrating gives the implicit solution
1
Z p
± 3
dx = 2/3 t + C.
1−x

6. The force is F (x) = −V ′ (x) = −2(x+1)(x−2)(2x−1).p Conservation of en-

ergy is y 2 +(x+1)2 (x−2)2 = E, giving orbits y = ± E − (x + 1)2 (x − 2)2 .
Not that orbits move to the right in the upper half plane and to the left
in the√lower half plane. Specifically, when x = 0, y = 3 initially, we find
E = 13.
8. The conservation law is given to be 12 ml2 (θ′ )2 +mgl(1−cosθ) = E. Taking
the derivative with respect to t, gives
1 2 ′ ′′
ml 2θ θ + mgl sin(θ)θ′ = 0.
2
This immediately reduces to θ′′ + (g/l) sin θ = 0.

21
 10. Substituting the given quantities into the conservation law gives E =
1.699(10)−3 J.
12. The critical points are (1, 0) and (−1, 0). The Jacobian matrix shows
easily that (1, 0) is a saddle point and (−1, 0) is an asymptotically stable
spiral.
14. We have E = 12 m(x′ )2 + V (x). Then, taking the time derivative,
dE
= mx′ x′′ + V ′ (x)x′
dt
= x′ (mx′′ + V ′ (x))
= x′ (−kx′ ) (by the equation of motion)
′ 2
= −k(x ) < 0.

 Sec. 5.3.2, page 282

4. The x nullclines are the straight lines x = 0, y = a/b. The y nullcline
is the curve y = M/(dx − c) which has a vertical asymptote at x = c/d;
that is, y → +∞ as x → c/d+, and y → −∞ as x → c/d−. There is
one equilibrium at ((M b + ac)/ad, a/b). The trace of the Jacobian matrix
at the equilibrium is M b/a, which is positive. Therefore, it is unstable.
A direction field shows a counterclockwise circulation, so the equilibrium
is an unstable spiral. Eventually the predator population will become
extinct as an orbit intersects the x axis.
6. The equilibria are at (0, 0) and (4, 6). The Jacobian matrix shows (0, 0) is
an unstable node and (4, 6) is an asymptotically stable node. The latter
is a coexistent state in this competition problem. Regardless of the initial
condition, the orbit approaches (4, 6).
10. The nullclines y = x and y = 5x2 /(4 + x2 ) (a sigmoid type curve) cross at
the
 critical points (0, 0), (1, 1), (4, 4). The Jacobian matrix is J(x, y) =
−1 1
40x . Evaluating J at the critical points shows directly that
(4+x2 )2 −1
(0, 0) is an asymptotically stable node, (1, 1) is a saddle point, and (4, 4)
is an asymptotically stable node.
14. (a) Because f (0) = 0, substituting (0, 0) into the equations shows x′ = 0,
y ′ = 0. So (0, 0) is an equilibrium. When x = K and y = 0 again x′ = 0,
y ′ = 0. So (K, 0) is an equilibrium. (b) The Jacobian at (0,0) is
   
r − 2rx/K −f (x) r 0
J(x, y) = = .
cf ′ (x)y −m + cf (x) (0,0) 0 −m

The eigenvalues are real and opposite sign, and thus (0, 0) is a saddle
point. Evaluating the Jacobian at (K, 0) gives
 
−r −f (K)
J(K, 0) = .
0 −m + cf (K)

22
 The eigenvalues are λ1 = −r and λ2 = −m + cf (K). If m > cf (K) then
both eigenvalues are negative and (K, 0) is an asymptotically stable node;
if m < cf (K) then eigenvalues have opposite signs and (K, 0) is a saddle
point. (c) For a positive equilibrium we must have y ′ = y(−m+cf (x)) = 0;
this means there must be a positive solution x∗ to f (x) = m/c, or f (x∗ ) =
m/c. Because f (x) → M as x → +∞, we need M > m/c. Therefore, to
get x′ = 0 we need rx∗ (1 − x∗ /K) − f (x∗ )y = 0 or y = mr ∗ ∗
c x (1 − x /K).
∗
This must be positive, so x < K.
 Sec. 5.3.3, page 292

4. Here r = 1/3, S(0) = 180, N = 200, and S ∗ = 100 did not get the flu.
1
Therefore the equation −S ∗ + N + 3a ln(S ∗ /S0 ) = 0 holds. Substituting
and solving for a gives a = 0.001959. The maximum number of infectives
occur when S = r/a = 170. Substituting values into I = −S + N +
r
a ln(S/S0 ) gives I = 20.

8. We need only the first two equations since R = N − S − I will then be

determined. In the SI plane only the equilibrium point is (0,0). The
Jacobian matrix at that point has two negative eigenvalues and thus is an
asymptotically stable node. Eventually, therefore, there are no infected
and susceptible individuals remaining and all become removed. To get
the phase plane, note that S ′ < 0, always, and I ′ > 0 if S > r/a, I ′ <
0if S < r/a. Therefore, if S(0) > r/a, and I(0) is small, the orbit rises as
infectives increase, it crosses the line S = r/a horizontally, and then goes
to (0, 0) as t → ∞.
10. (a) The reaction rate is r = kx and x′ = −kx and y ′ = 2kx. Note that
2x + y = C = 2x0 + y0 . Therefore y = 2x0 + y0 − 2x. But x = x0 e−kt , and
y is given by the last equation. (b) There are two reactions with rates r1 =
k1 ax, r2 = k−1 x2 . We have x′ = k1 ax−k−1 x2 , which is a logistic equation
with unstable equilibrium x = 0 and stable equilibrium x = k1 a/k−1 . (d)
The rate is r = kx2 y. Therefore, x′ = 3kx2 y, y ′ = −kx2 y, z ′ = kx2 y. Note
that x+3y = C, where C = x0 +3y0 . Therefore y = (1/3)(C−x), and the x
equation becomes x′ = kx2 (C −x). The phase line diagram shows x = C is
a stable equilibrium, so x → C. Thus, y → 0; and because y +z = C1 = y0
we get z = y0 − y → y0 . (f) The reactions have rates r1 = k1 x, r2 = k2 y.
Therefore the equations are x′ = −k1 x + k2 y, y ′ = k1 x − k2 , z ′ = k2 y.
Clearly x + y = C = x0 . Thus, x′ = −(k1 + k2 )x + k2 x0 . This has a stable
equilibrium at x∗ = k2 x0 /(k1 + k2 ). Therefore, y → x0 − x∗ .
12. First show that (S + I + I)′ = 0 by adding the equations; thus S + E + R =
N . Making the change of variables as indicated, we get

x′ = µ(1 − x) − xy, y ′ = (r + µ)y(R0 x − 1), R0 > 1.

The Jacobian at the equilibrium (1, 0) has two negative eigenvalues and
is therefore a saddle. The other equilibrium (R0−1 , (r + µ)(R0 − 1)) is an

23
 asymptotically stable endemic state. Check the sign of (tr J)2 − 4 det J to
determine when the critical point is a node or spiral. (Hint: Look at the
cases when R0 is slightly greater than 1 and when R0 is well above 1.)
 Sec. 5.4, page 306
2. Because x′ > 0, there can never be a critical point. So there are no
periodic orbits.
4. Note if k = 0 we can take V ′ (x) = x2 to get x′′ +2x = 0, which has periodic
orbits. If k 6= 0 then the Dulac theorem gives fx + gy = 0 + (−k) < 0,
which is of one sign. Therefore there are no periodic orbits.
6. Hint: Use the Poincare–Bendixson theorem.
8. Clearly θ(t) = −t+c1 , and the polar angle rotates clockwise. The equation
r′ = r(1 − r2 ) is a Bernoulli equation. Make the substitution w = r−2 to
get w′ + 2w = 2. So w = 1 + c2 e−2t , giving
1
r(t) = √ .
1 + c2 e−2t
Note c2 = 0 gives the circular orbit r = 1; as t → ∞ note r → 1 from
outside the circle or from inside the circle, depending on c2 > 0 or c2 < 0,
respectively.
10. We have f = Hy = 8y, g = −Hx = −2x. So the system is x′ = 8y,
y ′ = −2x. The origin (0, 0) is the only equilibrium and by dividing the
differential equations, separating variables, and integrating we get orbits
0.5x2 + 2y 2 = C, which are circles. The origin is a center.
12. To get part (a) use the law of mass action. (b) Note that (x − 2y − z)′ =
x′ − 2y ′ − z ′ = 0 (by substituting the equations) and so x − 2y − z = C.
(c) Note z(0) = 0 so z = x − 2y. Substituting into the two equations we
can eliminate z to get x′ = (2β − α)xy − βx2 and y ′ = −αxy. (d) There
is a line of critical points, x = 0 or the y axis. The x nullcline is along
the straight line y = βx/(2β − α). Above the nullcline the vector field is
SE and below the nullcline it is SW. So orbits approach the origin x = 0,
y = 0.
 Sec. 5.5, page 311
2. The equilibria are (0,0) and (1,0). The Jacobian matrices at the critical
points are
   
0 1 0 1
J(0, 0) = J(1, 0) = .
−1 c 1 c

For J(0, 0) the trace and determinant are positive. So (0, 0) is unstable; it
is an unstable spiral point if c < 2 and an unstable node if c > 2 (because

24
 the discriminant is c2 − 4.). For J(1, 0) we have the determinant negative,
so (1, 0) is always a saddle point. So the bifurcation occurs at c = 2. As
c increases from 0, the origin changes from to an unstable spiral to an
unstable node at c = 2.
4. This problem can be handled
√ exactly like Exercise 8, Section 5.4. Alter-
nately, we first note r = a is a circular, periodic solution. If a > 0 then
the only equilibrium is (0, 0). Further, θ(t) = −t + C winds clockwise.
The Jacobian matrix of the original rectangular system is
 
a 1
J(0, 0) = .
−1 a
Note that tr(J) = 2a and det J = a2 +1 > 0. The discriminant is negative.
If a < 0 the trace is negative and determinant is positive, so (0, 0) is a
stable spiral. If a > 0 the trace is positive and (0, 0) is an unstable spiral;
√
clearly, from the r-equation the orbits approach the limit cycle r = a.
6. For h < 0 there are no critical points; if h = 0 the√ only critical√point
is (0, 0); for h > 0 there are two critical points ( h, 0) and (− h, 0).
For h = 0 the origin has a nonstandard orbital structure with a saddle
structure for x > 0 and a stable node structure for x < 0. The Jacobian
matrices at the two critical points are
 √ √
√ √
  
2 h 0 −2 h 0
J( h, 0) = J(− h, 0) = .
0 −1 0 −1
These give a saddle and a stable node, respectively.
8. For a <√0 there are no √critical points. For a > 0 there are two critical
points ( a/2, 0) and (− a/2, 0). It is easily seen by calculating the Ja-
cobian matrices at these two points that they are a saddle point and an
asymptotically stable node, respectively. Therefore, as a passes through
zero two equilibria appear at a = 0.
10. The only critical point is (1, a). The Jacobian matrix is
 
a − 1 −1
J(1, a) = .
3a −1
If a < 2 the trace is negative and determinant positive, thus (1, a) is
asymptotically stable. If a > 2 the trace and determinant are positive, so
(1, a) is unstable.
12. (a) Kirchhoff’s law states VL + VR + VC = 0 or LI ′ + f (I) + C1 Q = 0.
Taking the derivative while using the chain rule gives LI ′′ + f ′ (I)I ′ =
I = 0. (b) Direct substitution into the last equation. (d) The system is
x′ = y, y ′ = −x − µ(x2 − 1)y. The Jacobian matrix at (0, 0) is
 
0 1
J(0, 0) = .
−1 µ

25
 The trace is µ and determinant is 1. The discriminant is µ2 −4. Therefore,
if µ > 2 the origin is an unstable node, and if µ < 2 it is an unstable spiral.
14. This is a linear system with critical point (0, 0). The Jacobian matrix is
 
−(ν + λN ) b
J(0, 0) = .
λN −µ

The trace is clearly negative. The determinant is det J = µ(ν+λN )−bλN .

If det J > 0 the the origin is asymptotically stable and the populations
die out. We want the other case when an outbreak occurs, namely when
the determinant is positive, so the origin is unstable. This occurs when
µ(ν + λN ) < bλN, which we now assume. In this case the L nullcline lies
below the M nullcline in the first quadrant (by our assumed determinant
condition) and so (0, 0) is a saddle point; thus orbits must veer away from
the origin and go to infinity.
18. Take the square with vertices at (0, ±1), (±1, 0) and find the vector field
along its edges.

26
 Exercise 2 Exercise 4
20 15

10
10
0

−10
5
−20

−30
0
−40

−50 −5
0 5 10 15 20 0 2 4 6

Figure 5: (Left) Euler approximation in Exercise 2 with N = 50. (Right) The

Euler approximation with h = 0.001 in Exercise 4 along with the exact solution
(dashed).

Chapter 6 Exercises

 Sec. 6.1, page 320

Rt
2. The recursion formula is xk+1 (t) = 1 + 0 (s − xk (s))ds. Then x0 (t) = 1,
Rt Rt
and x1 (t) = 1 + 0 (s − x0 (s))ds = 1 + 0 (s − 1)ds = 1 + (1/2)t2 − t, etc.

 Sec. 6.3, page 330

2. A plot of the solution using the Euler method with N = 50 and h = 0.4
is shown in Figure 5, left panel. The approximation is clearly inaccurate;
the exact solution is xex (t) = exp(sin t), which is periodic. We leave it to
the reader to check the additional cases for possible improvement.
4. A plot of the solution using the Euler method with N = 5000 and h =
0.001 is shown in Figure 5, right panel, along with the exact solution
33 t −4t/3
xex (t) = − ee .
32
The slope field is shown in Figure 6, indicating the reason for the inaccu-
racy.
6. The answer is similar to Exercise 4. The solution is relatively flat near the
exact solution. But if the initial condition is changed slightly, the solution

27
 Vector field in Exercise 4
16

14

12

10

−2
−1 0 1 2 3 4 5 6

Figure 6: The slope field in Exercise 4 showing the flatness in the region near
the solution. But there is a dramatic steepness in the slope field near that
region. A small error in the Euler method, which follows the slope field, leads
to a completely inaccurate approximation.

curves blow up. In the Euler method the approximation moves away
from the exact solution and is taken to the region where the slope field is
steep. This phenomenon can be compared to attempting to approximate
a separatrix moving into an unstable saddle point of a system; nearby
orbits veer away from the saddle near the critical point.
8 . The differential equation is x′ (t) = f (t). Integrating from tn to tn+1 and
using the fundamental theorem of calculus, we get
Z tn+1
x(tn+1 ) = f (t)dt.
tn

Approximating the right side by Simpson’s rule (see any calculus book)
gives
Z tn+1    
tn+1 − tn tn+1 + tn
f (t)dt = f (tn ) + 4f + f (tn+1 ) + O(h5 )
tn 6 2
h
= (f (tn ) + 4f (tn + h/2) + f (tn+1 )) + O(h5 ).
6
This is the Runge–Kutta formula with k1 = f (tn ), k2 = k3 = f (tn + h/2),
k4 = f (tn+1 ).

28