ANSWERS TO SELECTED

EXERCISES

Chapter 2
2.1.1 sin x 0 at x * n , for integer n
2.1.3 ( a) x d
dt
(x) d
dt
(sin x ) (cos x )x cos x sin x 1
2
sin 2x
2.2.1 x* 2, unstable; x * 2, stable
2.2.10 (a) x 0 (b) x sin x (c) impossible: between any two stable xed
points, there must be an unstable one (assuming that the vector eld is smooth).
(d) x 1

rm  e rt − e−rt 
2.2.13 (a) v =   , where r = gk / m . ( b) mg / k
k  e rt + e−rt 
( d) Vavg 29,300 / 116 253 ft / s 172 mph (e) V 265 ft / s
2.3.2 x* 0, unstable; x * k1a / k 1, stable
2.4.5 x* 0, f ( x * ) 0, half-stable by graphical analysis
2.4.6 x* 1, f ( x * ) 1, unstable
2.5.1 (1 c) 1

2.5.6 (a) Conservation of mass—the volume of water owing through the hole
equals the volume of water lost from the bucket. Equating the time derivatives of
these two volumes yields av(t ) Ah(t ). (b) Change in
P.E. = [ ∆m] gh = [ A( ∆h )] gh = change in K.E. = 12 (∆m)v 2 = 12 ( A∆h)v 2 .
Hence v2 2gh.
t +T x ( t +T )
2.6.2 On the one hand, ∫ t
f ( x ) dx
dt
dt = ∫
x (t )
f ( x ) dx = 0. The rst

equality follows from the chain rule, and the second from the assumption that

460 ANSWERS TO SELECTED QUESTIONS

 t +T t +T
x( t) x( t T ) . On the other hand, ∫ t
f ( x ) dx
dt
dt = ∫
t
( dx
dt
)2 dt > 0 by

assumption that T 0 and dx

dt
does not vanish identically.
2.7.5 V( x) cosh x; equilibrium x * 0, stable
2.8.1 The equation is time-independent so the slope is determined solely by x.
2.8.6 (b) From Taylor’s series, we nd x + e−x = l + 12 x 2 − 61 x3 + O ( x 4 ).
Graphical analysis shows that 1 ≤ x = x + e−x ≤ 1 + 12 x 2 for all x. Integration
then yields t x(t ) 2 tan(t / 2 ). Hence 1 ≤ x(1) ≤ 2 tan( l / 2 ) ≈ 1.208.
(c) A step size of order 10 4 is needed, and yields xEuler(l ) 1.15361. (d) A step size
t 1 gives three-decimal accuracy: t 1 xRK ( 1 ) 1.1536059; t 0.1
xRK ( 1 ) 1.1536389; t 0.01 xRK ( l ) 1.1536390.
2.8.7 (a) x ( t1 ) x ( t0 t ) x ( t0 ) t x ( t0 ) 1
2
( t ) 2 x ( t0 )
O( t) 3
x0 t f ( x0 ) 1
2
( t ) f ( x0 ) f ( x0 ) O ( t ) 3, where we’ve
2

made use of x f ( x ) and x f ( x) x f ( x) f ( x) .

(b)  x(t1 ) − x1 = 12 ( ∆t ) 2 f ′( x0 ) f ( x0 ) + O (∆t )3 . Hence C = 12 f ′( x0 ) f ( x0 ).

Chapter 3
3.1.1 rc 2
3.2.3 rc 1
Gnp
3.3.1 (a) n = − kn (c) transcritical
f + Gn
3.4.4 rc 1, subcritical pitchfork
3.4.11 (b) x * 0 , unstable (c) rc 1, subcritical pitchfork; in nitely many
saddle-node bifurcations occur as r decreases from 1 to 0 (use graphical analysis).
_
1
(d) rc ≈ (4n +1) 2  for integer n 1.

3.4.15 rc 3 / 16
3.5.4 (a) mx + bx + kx(1− L0 / ( h 2 + x 2 )1/ 2 ) = 0 (d) m b2 / k
3.5.5 (a) Tfast mr / b
3.5.7 (b) x N / K, x0 N0 / K, rt
3.6.5 (b) u x / a, R L 0 / a, h mg sin / ka. (c) R l, unique xed point;
R 1, one, two, or three xed points, depending on h.
3.7.2 (b) Cusp at x 3
3.7.4 (d) transcritical (e) saddle-node
3.7.5 (b) rc 12 (d) saddle-node curves at rc 2x / ( 1 x2 ) 2 ,
sc x2 ( 1 x2 ) / ( 1 x2 ) 2
ANSWERS TO SELECTED EXERCISES 461
Chapter 4
4.1.1 a integer. For a well-de ned vector eld on the circle, we need
sin ( a ( 2 k)) sin ( a ) for all integer k . Hence 2 ka 2 n, for some integer
n . Thus ka integer, for all integer k . This is possible only if a is itself an integer.
4.1.3 Unstable xed points: * 0, . Stable xed points: * / 2.
4.2.1 12 seconds
4.2.3 12 / 11 hours later, i.e., at approximately 1:05 and 27 seconds. This
problem can be solved in many ways. One method is based on Example 4.2.1. It
takes the minute hand T1 1 hr and the hour hand T2 12 hrs to complete one
revolution around the clockface. Hence the time required for the minute hand to
lap the hour hand is T ( 1 121 ) 1 12 11
hrs.
du ∞
4.3.2 (a) d 2 du / ( 1 u2 ) (d) T 2
−∞ u − 2au + ∫ 2
(e) x u a/ ,

2 ∞ dx 2π 2π
ω ∫−∞
r 1   a2 / 2
,T= 2
= = .
r+x ω r ω 2 − a2
∞ du
4.3.10 b= 1
2n
− 1, c = ∫ = .
−∞ 1 + u 2 n n sin( / 2n )
4.4.1 b2 m2gL3, approximation valid after an initial transient
4.5.1 (b) ω − Ω ≤ π2 A
4.6.4 (a) Ib = Ia + IR (c) Vk = k
2e
4.6.5 Let R0 R / N. Then IbR0 / Icr, a ( R0 r ) / r,
= [ 2eI c r 2 / (R0 + r )]t.
d k dQ
4.6.6 Kirchhoff’s current law gives + I c sin k + = I b , k = 1, . . . ,
2er dt dt
N, and Kirchhoff’s voltage law gives
d 2Q dQ Q N d j
L 2
+R + = ∑ .
dt dt C 2e j =1 dt

Chapter 5
5.1.9 (c) x y, stable manifold; x y, unstable manifold
5.1.10 (d) Liapunov stable (e) asymptotically stable
1  1
5.2.1 (a) = 3, v1 = (1, 2 ), v 2 = (1, 1). (b) x(t ) = c1   e 2t + c2   e 3t .
= 2,
1 2 2 1
(c) unstable node (d) x e 2t
2e , y 2e
3t 2t
2e 3t

cos t − sin t

5.2.2 x( t) C1et   + C2 e t  
 sin t   cos t 

462 ANSWERS TO SELECTED EXERCISES

5.2.3 stable node
5.2.5 degenerate node
5.2.7 center
5.2.9 non-isolated xed point
5.3.1 a 0, b 0: Narcissistic Nerd, Better Latent than Never, Flirting Fink,
Likes to Tease but not to Please. a 0, b 0: Bashful Budder, Lackluster Libido
Lover. a, b 0: Hermit, Malevolent Misanthrope (Answers suggested by my stu-
dents and also by students in Peter Christopher’s class at Worcester Polytechnic
Institute.)

Chapter 6
6.1.1 saddle point at ( 0, 0 )
6.1.5 stable spiral at ( 1,1 ) , saddle point at ( 0,0 ) , y-axis is invariant.
6.3.3 ( 0,0 ) , saddle point
6.3.6 ( 1, 1 ) , stable node; ( 1,1 ) , saddle point
6.3.8 (b) unstable
6.3.9 (a) stable node at ( 0,0 ) , saddle points at ( 2,2 ) .
6.4.1 Unstable node at ( 0,0 ) , stable node at ( 3,0 ) , saddle point at ( 0,2 ) . Null-
clines are parallel diagonal lines. All trajectories end up at ( 3,0 ) , except those
starting on the y-axis.
6.4.2 All trajectories approach ( 1,1 ) , except those starting on the axes.
6.4.4 (a) Each species grows exponentially in the absence of the other.
(b) x b2N1 / r1, y b1N2 / r1, r1t, r2 / r1 . (d) saddle point at ( ,l ) . Almost all
trajectories approach the axes. Hence one or the other species dies out.
6.5.1 (a) center at ( 0,0 ) , saddles at ( 1,0 ) (b) 12 x 2 + 12 x 2 − 14 x 4 = C
6.5.2 (c) y2 x2 2
3
x3
6.5.6 (e) Epidemic occurs if x0 ℓ/k .
6.6.1 Reversible, since equations invariant under t t, y y.
6.6.10 Yes. The linearization predicts a center and the system is reversible:
t t, x x. A variant of Theorem 6.6.1 shows the system has a nonlinear
center.
6.7.2 (e) Small oscillations have angular frequency ( 1 )
2 l/4
for 1 1.
6.8.2 xed point at ( 0,0 ) , index I 0.

ANSWERS TO SELECTED EXERCISES 463

6.8.7 ( 2,0 ) and ( 0,0 ) , saddles; ( 1,3 ) , stable spiral; ( 2,0 ) , stable node.
Coordinate axes are invariant. A closed orbit would have to encircle the node or
the spiral. But such a cycle can’t encircle the node (cycle would cross the x-axis:
forbidden). Similarly, cycle can’t encircle the spiral, since spiral is joined to saddle
at ( 2,0 ) by a branch of saddle’s unstable manifold, and cycle can’t cross this
trajectory.
6.8.9 False. Counterexample: use polar coordinates and consider
r = r( r − l)( r 2 − 9) , = r 2 − 4. This has all the required properties but there
2

are no xed points between the cycles r 1 and r 3, since r 0 in that region.
6.8.11 (c) For z z k , the origin has index k. To see this, let z rei . Then
zk rkeik . Hence k and the result follows. Similarly, the origin has index k
k
for z = ( z ) .

Chapter 7
7.1.8 (b) Period T 2 (c) stable
7.1.9 (b) R cos R, R sin k, where prime denotes differentia-
tion with respect to the central angle . (c) The dog asymptotically approaches a
circle for which R = 1− k 2 = 3
4
.
7.2.5 (b) Yes, as long as the vector eld is smooth everywhere, i.e., there are no
singularities.
2
y2
7.2.9 (c) V ex , equipotentials are circles x2 y2 C.
7.2.10 Any a, b 0 with a b suf ces.
7.2.12 a l, m 2, n 4
1
7.3.1 (a) unstable spiral (b) r = r ( l − r 2 − r 2sin 2 2 ) (c) r1 =
≈ .707
2
(d) r2 1 (e) No xed points inside the trapping region, so Poincaré-Bendixson
implies the existence of limit cycle.
7.3.7 (a) r = ar (1− r 2 − 2b cos 2 ) , = − 1+ ab sin 2 . (b) There is at least
one limit cycle in the annular trapping region 1− 2b ≤ r ≤ 1, by the Poincaré-
Bendixson theorem. Period of any such cycle is T
2π dθ
∫ dt = ∫ ( ddtθ ) d θ = ∫0 −1 + ab sin 2θ = T (a, b ).

7.3.9 (a) r( ) 1 ( 52 cos 1

5
sin ) O( 2
).
2 2
(b) rmax 1 O( ), rmin 1 5
O( ).
5

464 ANSWERS TO SELECTED EXERCISES

7.4.1 Use Liénard’s theorem.
7.5.2 In the Liénard plane, the limit cycle converges to a xed shape as
; that’s not true in the usual phase plane.
7.5.4 (d) T  ( 2 1n 3 ) .
7.5.5 T 2 2 ln 1 2

7.6.7 r ′ = 12 r (1− 18 r 4 ) , stable limit cycle at r 81 / 4 23 / 4, frequency

1 O( 2) .
7.6.8 r ′ = 12 r (1− 34 r ) , stable limit cycle at r 3
4
, ω = 1 + O(ε 2 )
7.6.9 r ′ = 161 r 3 (6 − r 2 ) , stable limit cycle at r = 6 , ω = 1 + O(ε 2 )
7.6.14 (b) x(t, ) ( a −2 + 34 t )−1 2 cos t
2
7.6.17 (b) c
1
2
(c) k = 1
4
1− 4 (d) If 1
2
, then 0 for all , and
−1
r ( T ) is periodic. In fact, r(φ ) ∝ ( γ + 12 cos 2φ ) , so if r is small initially, r ( )
remains close to 0 for all time.
7.6.19 (d) x0 = a cos ( f ) x1 = 321 a 3 (cos 3 − cos )
7.6.22 x = a cos ωt + 16 εa 2 (3 − 2 cos ωt − cos 2ωt ) + O( ε 2 ), ω = 1− 125 ε 2 a 2 + O (ε 3 )
7.6.24 ω = 1− 38 εa 2 − 256
21 2 4
ε a − 2048
81
ε3 a 6 + O(ε 4 )

Chapter 8
8.1.3 1 , 2 1
8.1.6 (b) c
1; saddle-node bifurcation
8.1.13 (a) One nondimensionalization is dx / dt x ( y 1 ) , dy / dt xy ay b,
where kt, x Gn / k, y GN / k, a f / k, b pG / k2. (d) Transcritical bifurca-
tion when a b .
8.2.3 subcritical
8.2.5 supercritical
8.2.8 (d) supercritical
8.2.12 (a) a 1
8
(b) subcritical
8.3.1 (a) x * 1, y* b / a, b ( 1 a) , a 0 . Fixed point is stable if
b 1 a , unstable if b 1 a , and linear center if b 1 a. (c ) bc 1 a
(d) b bc (e) T   2 / a
8.4.3 0.066 0.001

ANSWERS TO SELECTED EXERCISES 465

8.4.4 Cycle created by supercritical Hopf bifurcation at 1, destroyed by
homoclinic bifurcation at 3.72 0.01.

32 3 k 3
8.4.9 ( c ) bc
27 F 2
8.4.12 t O ( u 1 ln(1 )).
F 1 2α 2
8.5.4 (d) u(θ ) = + cos θ − sin θ . ( e) Fc ( ) = .
2α 1 + 4α 2 1 + 4α 2 1+ 4 2

8.6.2 (d) If | 1 | | 2a |, then lim θ1 (τ ) / θ2 ( τ ) = (1 + ω + ωφ ) / (1 + ω − ωφ ) ,

τ →∞
2 1/ 2
where ωφ = ((1− ω ) − 4a ) . On the other hand, if | l
2
| | 2a |, phase-locking
occurs and lim θ1 (τ ) / θ2 ( τ ) = 1 .
τ →∞

8.6.6 (c) Lissajous gures are planar projections of the motion. The motion in
the four-dimensional space ( x, x, y, y ) is projected onto the plane ( x, y ) . The
parameter is a winding number, since it is a ratio of two frequencies. For
rational winding numbers, the trajectories on the torus are knotted. When
projected onto the xy plane they appear as closed curves with self-crossings (like
a shadow of a knot).
8.6.7 (a) r0 ( h2 / mk ) 1 / 3, h / mr02 (c) r / 3 , which is irrational.
(e) Two masses are connected by a string of xed length. The rst mass plays the
role of the particle; it moves on a frictionless, horizontal “air table.” It is con-
nected to the second mass by a string that passes through a hole in the center of
the table. This second mass hangs below the table, bobbing up and down and
supplying the constant force of its weight. This mechanical system obeys the
equations given in the text, after some rescaling.
8.7.2 a 0 , stable; a 0, neutral; a 0 , unstable
8.7.4 A 0
8.7.9 (b) stable (c) e 2

Chapter 9
9.1.2 d
dt
( an2 + bn2 ) = 2( an an + bn bn ) = −2K ( an2 + bn2 ) . Thus

( an2 + bn2 ) ∝ e −2 Kt → 0 as t .
9.1.3 Let a1 y, b1 z q1 / K, x, and t = T , and solve for the
coef cients by matching the Lorenz and waterwheel equations. Find T 1 / K,
K. Picking K yields Kv / gr, Kv / gr. Also v / KI, Rayleigh
r  grq1 / K2v.

466 ANSWERS TO SELECTED EXERCISES

9.1.4 (a) degenerate pitchfork (b) Let [b ( r 1 ) ] 1 / 2 . Then
tlaser = (σ / κ )tLorenz , E = α x, P = α y, D = r − z, γ1 = κ / σ, γ 2 = κb / σ ,λ = r −1 .
9.2.1 (b) If 1, then C and C are stable for all r 0. (c) If r rH ,
b
then ( 1). b
x2 y2 ( z − r )2
9.2.2 Pick C so large that + 2+ > 1 everywhere on the bound-
br br r2
ary of E.
9.3.8 (a) yes (b) yes
9.4.2 (b) x * 2
3
; unstable (c) x1 2
5
, x2 4
5
; 2-cycle is unstable.
9.5.5 (a) X x, Y y, Z ( z 1 ), t/
9.5.6 Transient chaos does not occur if the trajectory starts close enough to
C or C .
9.6.1 (a) V ≤ −kV for any k min ( 2, 2b ) . Integration then yields
0 V ( t ) V0 e kt. (b)  12 e2 2 ≤V <V0 e −kt , so e2 ( t ) ( 2V0 ) 1 / 2 e kt / 2. Similarly,
e3 ( t ) O ( e kt / 2 ) . (c) Integration of e1 = (e2 − e1 ) , combined with
e2 ( t ) O( e kt / 2
) , implies e1 (t ) ≤ max {O( e − t ),O( e−kt / 2 )} . So all components of
e ( t ) decay exponentially fast.
9.6.6 According to Cuomo and Oppenheim (1992, 1993),
R      
u=
1  4 v − R3 1 + R4  u  , v = 1  R11 1 + R12 + R12 
R      
R5C1  1 R2 + R3  R1   R15C2  R10 + R11  R8 R9 
 R7      
1 +  u − R12 v − R12 uw  , w = 1  R19 uv − R18 1 + R19  w  .
 
 R 
6 R8 R9 
 R20C3  R16 R17 + R18  R16  

Chapter 10
10.1.1 xn 1 as n , for all x0 0
10.1.10 Yes
10.1.13 Differentiation yields = f ′( x*) = g ( x*) g ′′ ( x*) g ′( x*)2 . Hence
g ( x * ) 0 implies 0 (unless g ′( x*) = 0 too; this nongeneric case requires
separate treatment).
10.3.2 (b) l 5
10.3.7 (d) Any orbit starting at an irrational number x0 will be aperiodic, since
the decimal expansion of an irrational number never repeats.
10.3.12 (a) The maximum of the map occurs at x 12 . A superstable cycle of
period 2n occurs when this point is an element of a 2n -cycle, or equivalently, a
n n
xed point of f ( 2 ) ( x, r ). Hence the desired formula for Rn is f ( 2 ) ( 12 , Rn ) 12 .

ANSWERS TO SELECTED EXERCISES 467

10.3.13 (a) The curves are f k ( 12 , r ) vs. r, for k 1, 2, . . . Intuitively, points near
xm 12 get mapped to almost the same value, since the slope equals zero at xm .
So there is a high density of points near the iterates f k ( 12 ) of the maximum.
(b) The corner of the big wedge occurs when f 3 ( 12 ) f 4 ( 12 ) , as is clear from the
graphs of part (a). Hence f ( u ) u, where u f 3 ( 12 ) . So u must equal the xed
point 1 1r . The solution of f 3 ( 12 , r ) = 1− 1r can be obtained exactly as
r = 23 + 83 (19 + 297 )−1/3 + 23 (19 + 297 )1/ 3 = 3.67857 . . .
10.4.4 3.8318741. . .
10.4.7 (b) RLRR
10.4.11 For 0 < a < e −e , xn tends to a stable 2-cycle; for e−e < a < 1, xn x*
x* 1/ e
where x * is the unique root of x* a ; for 1 a e , xn tends to the smaller
root of x a x ; and for a e1/ e , xn → ∞ .
10.5.3 The origin is globally stable for r 1, by cobwebbing. There is an inter-
val of marginally stable xed points for r 1.
10.5.4 The Liapunov exponent is necessarily negative in a periodic window.
But since ln r 0 for all r 1, there can be no periodic windows after the
onset of chaos.
10.6.1 (b) r1 0.71994, r2 0.83326, r3 0.85861, r4 0.86408, r5 0.86526,
r6   0.86551.

10.7.1 (a) = −1− 3 = −2.732 . . . , c2 /2 1.366. . . (b) Solve

= ( 1 + c2 + c4 ) ,  c2 = 2
1 1
− 2 − 2, c4 = 1 + 2 − −1 simultaneously.
1 1

Relevant root is = 2.53403. . . , c2  =   1.52224. . . , c4 = 0.12761. . .

10.7.8 (e) b l/2
10.7.9 (b) The steps in the cobweb staircase for g 2 are twice as long, so 2.

Chapter 11
11.1.3 uncountable
11.1.6 (a) x0 is rational the corresponding orbit is periodic
1 2 4 1 1
11.2.1 + + + . . . =   =1
3 9 27  3  1− 2
3

11.2.4 Measure 1; uncountable.

11.2.4 (b) Hint: Write x [0, 1 ] in binary, i.e., base-2.
11.3.1 (a) d ln 2 / ln 4 1
2

11.3.4 ln 5 / ln l0

468 ANSWERS TO SELECTED EXERCISES