TFY4305 solutions extra exercises

2014

Problem 2.3.3
Tumor growth can be modelled by Gompertz’ law

Ṅ = −aN ln(bN ) , (1)

where N (t) is proportional to the number of cells in the tumor, and a, b > 0 are parameters.

a) In order to interpret b we note that N = 1/b is a stable fixed point since f (1/b) = 0
and f (N ) = −aN ln(bN ) > 0 for N < 1/b and f (N ) < 0 for N > 1/b. Thus 1/b is the lim-
iting size of the tumor. The parameter a is related to the proliferation ability (cell division)
and depends on the availability of substrate, oxygen etc.
Interestingly the Gompertz equation can be solved exactly. Separation of variables gives
dN
= −adt (2)
N ln bN
Integration yields

ln[ln(bN )] = −at + C , (3)

where C is an integration constant. Exponentiating this expression twice gives

1 eC e−at
N (t) = e . (4)
b
C
Using the initial condition N (0) = N0 , one finds N0 = 1b ee or ln(N0 b) = eC . The solution
can then be written as
1 ln(N0 b)e−at
N (t) = e . (5)
b

The solution satisfies N (0) = N0 and

1
lim = . (6)
t→∞ b

1
 2

fHN L
2

ç æ N
0.5 1.0 1.5 2.0

-1

-2

Figure 1: Vector field for the Gompertz model with b = 1.

We can also gain insight into the solution without knowing it. Taking the derivative of
the Gompertz equation, we obtain

N̈ = −aṄ (ln(bN ) + 1)
= a2 N ln(bN )(ln(bN ) + 1) . (7)

Assume first that N0 > 1/b. Then we know that N (t) decreases monotonically to N = 1/b
as t → ∞. In this case Ṅ < 0 and so N̈ > 0. This implies that there is no inflection point.
Next assume that N0 < 1/b. Then we that N (t) increases monotonically to N = 1/b as
t → ∞. In this case Ṅ < 0 but N̈ changes sign at N = 1b e−1 . Thus there is an inflection
point here.
b) The other fixed point of the differential equation is N = 0 and this fixed point is
unstable. The vector field is shown in Fig. 1 for b = 1. The graph N (t) is shown for
a = b = 1 and two initial conditions N0 = 1/20 < 1/b and N0 = 1.2 > 1/b in Fig. 2. In the
former case one can clearly see the inflection point as predicted above.
 3

N HtL
1.2

1.0

0.8

0.6

0.4

0.2

t
0 1 2 3 4 5

Figure 2: Solution N (t) for a = 1 and two initial conditions N0 = 1/20 < 1/b and N0 =
1.2 > 1/b.

Fig. 3 shows the data points for tumor growth in a laboratory experiment at NTNU. The
parameters a and b have been fitted to the data points. The agreement is very good.
 4

Figure 3: Tumor growth.

Problem 2.4.6
We have

ẋ = ln x . (8)

The fixed points are found by solving f (x∗ ) = ln x∗ = 0, which yields x∗ = 1. Moreover

1
f 0 (x) = , (9)
x
and so f 0 (1) = 1. The fixed point x∗ = 1 is therefore unstable.
 5

Problem 2.6.1
The point is that the harmonic oscillator is not a first-order system. It is a system of two
coupled differential equations. Define ẋ = y. This yields
mẏ = −kx (10)
ẋ = y , (11)
and we conclude that the system is two dimensional and so does not correspond to flow on
the line.

Problem 2.7.5
The dynamics is governed by the equation
ẋ = − sinh x . (12)
The function f (x) = − sinh x is shown in Fig. 4. The only zero of f (x), i. e. fixed point is
fHxL

> > è < < x

-2 -1 1 2

-1

-2

-3

Figure 4: Function f (x) = − sinh x.

x = 0. Since f 0 (x) = − cosh x and f 0 (0) = −1, the fixed point is stable. This can also be
inferred from inspecting the graph.
The potential is given by integration of dV /dx = sinh x, Integration yields
V (x) = cosh x + C , (13)
where C is an integration constant we set to zero.
The potential V (x) is shown in Fig 5.

Problem 6.3.13
The dynamics is governed by the set of equations
ẋ = −y − x3 , (14)
ẏ = x . (15)
 6

VHxL

1.5

1.4

1.3

1.2

1.1

x
-1.0 -0.5 0.5 1.0

Figure 5: Function V (x) = cosh x.

The Jacobian matrix is

 
−3x2 −1
A = . (16)
1 0

Evaluated at the origin, we find

 
0 −1
A = . (17)
1 0

This yields the eigenvalues λ1,2 = ±i, i. e. linearization predicts a center. In order to gain
some insight into the flow, we switch to polar coordinates. One finds
xẋ + y ẏ
ṙ =
r
= −r cos4 θ ,
3
(18)

which shows that ṙ ≤ 0 and r is a nonincreasing function. Only for θ = π/2 and θ = 3π/2
is the radial flow vanishing. Moreover, one finds
xẏ − y ẋ
θ̇ =
r2
= 1 + r2 cos3 θ sin θ . (19)

θ̇ is always nonzero when ṙ = 0 and so the origin is the only fixed point. For r < 1, θ̇ > 0
and so the trajectory must be spiraling inwards. The phase portrait is shown in Fig. 6

Problem 6.3.14
The dynamics is governed by the set of equations

ẋ = ax3 − y , (20)
ẏ = x + ay 3 . (21)
 7

-1

-2

-2 -1 0 1 2

Figure 6: Phase portrait of problem 6.3.13.

a) The Jacobian matrix is

 
3ax2 −1
A = . (22)
1 3ay 2
Evaluated at the origin, we obtai
 
0 −1
A = . (23)
1 0
The eigenvalues are λ1.2 = ±i and so it is a linear center. However, note that this is a
borderline case and so we have to be careful. Going to polar coordinates, we find
ṙr = a(x4 + y 4 ) ,
Thus for r 6= 0, ṙ > 0 for a > 0 and ṙ < 0 for a < 0. Hence r is increasing for a > 0 and
decreasing for a < 0. Moreover
θ̇ = 1 − ar2 cos θ sin θ(cos2 θ − sin2 θ) , (24)
and so θ is an increasing function when r is sufficiently small. Hence the origin is an unstable
spiral for a > 0 and a stable spiral for a < 0. In both cases it spirals counterclockwise. Finally
for a = 0, the origin is a center since the equation reduces to that of a harmonic oscillator
(with appropriately scaled coordinates).

Problem 6.5.1
The second-order equation is
ẍ = x3 − x . (25)
 8

2 2

1 1

0 0

-1 -1

-2 -2

-2 -1 0 1 2 -2 -1 0 1 2

Figure 7: Phase portrait of problem 6.3.14. Left plot is for a = 1 and right plot is for a = −1.

This can be written as a set of first-order equations

ẋ = y , (26)
ẏ = x3 − x . (27)

a) The fixed points are (x, y) = (0, 0) and (x, y) = (±1, 0). The Jacobian matrix is
 
0 1
A = 2 . (28)
3x − 1 0

Evaluated at the origin, we find

 
0 1
A = , (29)
−1 0

and so λ1,2 = ±i. The origin is a center. The other fixed points are analyzed in the same
manner:
 
0 1
A = , (30)
2 0
√
and the eigenvalues are λ1,2 = ± 2. (±1, 0) are saddle points.

b) The potential energy is given by

dV
= x − x3 . (31)
dx
 9

Integration yields
1 2 1 4
V (x) = x − x , (32)
2 4
where we have set the integration constant to zero. The conserved quantity is then
1 2
E = ẋ + V (x)
2
1 2 1 2 1 4
= ẋ + x − x . (33)
2 2 4

c) The phase portrait is shown in Fig. 8.

-1

-2

-2 -1 0 1 2

Figure 8: Phase portrait of problem 6.5.1.

Problem 6.6.3
a) The function f (x, y) = sin y is odd in y and the function g(x, y) = sin x is even in y.
Hence the system is reversible.

b) The equations are

ẋ = sin y , (34)
ẏ = sin x , (35)
(36)
The fixed points are x∗ = nπ and y ∗ = mπ, where n and m are integers. These points are
the zeros of the sine function.
 10

The Jacobian matrix is

cos y ∗
 
∗ ∗ 0
A(x , y ) = , (37)
cos y ∗ 0

The eigenvalues are

√
λ1,2 = ± cos x∗ cos y ∗ . (38)

cos nπ = 1 if n even and cos nπ = −1 if n odd. This implies that the eigenvalues are
λ1,2 = ±1 if n and m are even or if n and m are odd. The fix point is then a saddle. If n
is even and m odd or vice versa, the eigenvalues are λ = ±i. The fixed point is then a center.

c) We have

dy ẏ
=
dx ẋ
sin x
= . (39)
sin y
At a point on the line y = ±x , this ratio is ±1 and so vector field points along the line
itself. Hence, we stay on it.

d) The phase portrait and the type of fixed point can be seen in Fig. 9.

-2

-4

-4 -2 0 2 4

Figure 9: Phase portrait of problem 6.6.3.

 11

Problem 6.6.10
The dynamics is given by the equations

ẋ = −y − x2 , (40)
ẏ = x . (41)

a) The only fixed point is the origin. The Jacobian matrix is given by
 
0 − 2x −1
A(x, y) = . (42)
1 0

Evaluated at the origin, we find

 
0 −1
A(0, 0) = , (43)
1 0

and so the eigenvalues are λ = ±i. The origin is therefore a linear center. Note that the
system in reversible since it is invariant under t → −t and y → −y. Theorem 6.6.1 then
applies and the origin is a nonlinear center. The phase portrait is shown in Fig. 10.

-1

-2

-2 -1 0 1 2

Figure 10: Phase portrait of problem 6.6.10

Problem 6.7.1
The equation for the damped pendulum reads

θ̈ + bθ̇ + sin θ = 0 , (44)

 12

where b > 0 is a parameter. Introducing ν = θ̇, we can write the above equation as

θ̇ = ν , (45)
ν̇ = −bν − sin θ . (46)

The fixed points are given by ν̇ = θ̇ = 0, i. e. ν = 0 and therefore sin θ = 0. The latter
yields θ = 0 or θ = π. The fixed points are therefore (0, 0) and (π, 0). The Jacobian matrix
reads
 
0 1
A(θ, ν) = (47)
− cos θ −b

This gives the characteristic equation for the eigenvalues

λ(λ + b + cos θ) = 0 . (48)

The solutions are

√
−b ± b2 − 4 cos θ
λ = . (49)
2
This yields τ = −b and ∆ = cos θ. For the fixed point (0, 0), this yields τ = −b and ∆ = 1.
If τ 2 − 4∆ = b2 − 4 > 0, i.e. if b > 2, the fixed point is a stable node, and if b < 2, the fixed
point is a stable spiral. Moreover, For the fixed point (0, π), this yields τ = −b and ∆ = −1.
Thus the fixed point is a saddle point for all values of b.

Problem 7.1.4
The dynamics is governed by the equations

ṙ = r sin r , (50)
θ̇ = 1 . (51)

In Cartesian coordinates, this becomes

p
ẋ = x sin x2 + y 2 − y , (52)
p
ẏ = y sin x2 + y 2 + x . (53)

We note that θ̇ = 1 everywhere and so the flow is always counterclockwise. The radial flow
vanishes for r = π, 2π etc and so we have circular orbits for these values of r. Between these
values of r, the radial flow always has the same sign and it alternating. Between r = 0 and
r = π, ṙ > 0 and so flow is away from the origin. Between r = π and r = 2π, ṙ < 0 and so
flow is towards the origin. The circular orbits are therefore limit cycles and their stability is
alternating.
The phase portrait is shown in Fig. 11.
 13

-2

-4

-6

-6 -4 -2 0 2 4 6

Figure 11: Phase portrait of problem 7.1.4.

Problem 8.1.9
1
The equation for the anharmonic oscillator with damping reads

ẍ + bẋ − kx + x3 = 0 . (54)
This can be written as
ẋ = y , (55)
ẏ = −by + kx − x3 . (56)
√
The fixed points are (0, 0) and (± k, 0). The Jacobian matrix reads
 
0 1
A(x, y) = . (57)
k − 3x2 −b
This yields
 
0 1
A(0, 0) = , (58)
k −b
and thus
√
−b ± b2 + 4k
λ = . (59)
2
1
Note that we have damping for b > 0. For b < 0 we are pumping energy into the system. Also note that
the linear force term is the usual if k < 0. For k > 0 the force is no longer restoring but pushes the particle
away from the origin. However, the cubic term stabilizes the system for all b and k.
 14

Similarly
√
 
0 1
A(± k, 0) = , (60)
−2k −b

and thus
√
−b ± b2 − 8k
λ = . (61)
2
All fixed points are stable for b > 0 and unstable for b < 0. Whether it is a spiral or node
depends on the sign of b2 + 4k and b2 − 8k, respectively. For b = 0, there is no damping and
the system is conservative. The potential is V (x) = − 21 kx2 + 41 x4 . In this case, the origin is
stable for k < 0 and unstable for k > 0. The other fixed points only exist for k > 0 and they
are stable. This is the standard pitchfork bifurcation from chapter three. This is shown in
Fig. 12
k

1.0 Stable spiral

0.5
Stable node
0.0 b

-0.5 Stable node

-1.0 stable spiral

-1.5

-2.0

-3 -2 -1 0 1 2 3

Figure 12: Stability diagram. Everything to the left is unstable.

Problem 8.2.8
The equations governing the dynamics are

ẋ = x[x(1 − x) − y] , (62)
ẏ = y(x − a) , (63)

where a ≥ 0 is a parameter.

a) The nullclines are given by ẋ = 0 and ẏ = 0 The former yields x = 0, and y = x(1 − x),
while the latter gives y = 0 and x = a. This is shown Fig. 13 b) ẏ = 0 gives y = 0 or x = a.
If y = 0, ẋ = 0 implies that x = 0 or x = 1. If x = a ẋ = 0 implies y = a(1 − a). Thus
the fixed points are (x, y) = (0, 0), (x, y) = (1, 0), and (x, y) = (a, a(1 − a)). Note that last
 15

1.0 dxdt=0
dydt=0

0.5

dxdt=0

dydt=0

0.2 0.4 0.6 0.8 1.0 1.2 1.4

-0.5

-1.0

Figure 13: Nullclines for problem 8.2.8.

fixed point has x < 0 if a < 0 and y < 0 if a > 1 and so does not make sense biologically.
Since a ≥ 0 is specifies, we must restrict ourselves to a ∈ [0, 1] for this fixed point.
The Jacobian matrix reads
 
2x − 3x2 − y −x
A(x, y) = . (64)
y x−a

This yields
 
0 0
A(0, 0) = , (65)
0 −a

and so the eigenvalues are λ = 0 and λ = −a. The origin is therefore marginally stable for
all a ≥ 0.
The phase portrait is shown Fig. 14 which shows that the origin is unstable. Moreover
 
−1 −1
A(1, 0) = , (66)
0 1−a

and so the eigenvalues are λ = −1 and λ = 1 − a. The fixed point is therefore a saddle for
a < 1 and a stable node for a > 1. For a = 1, the fixed point is marginally stable. The
phase portrait is shown Fig. 14 indicating that the fixed point is actually stable.
Finally
 
2 a − 2a2 −a
A(a, a − a ) = , (67)
a − a2 0

with eigenvalues
√
a − 2a2 ± a 4a2 − 3
λ = . (68)
2
√
For a > 1/2 it is a unstable and for a < 1/2 √it is stable. For 4a2 − 3 > 0, i.e. for a > 3/2
it is a node and for 4a2 − 3 < 0, i.e. for a < 3/2 it is a spiral.
 16

1.0

0.8

0.6

0.4

0.2

0.0

0.0 0.5 1.0 1.5

Figure 14: Phase portrait for a = 1 in problem 8.2.8.

c) The phase portrait is shown Fig. 15 which shows that you will flow to the line y = 0,
i.e. the predators will go extinct. d) We have seen that the real part of the eigenvalues go
through a zero at a = ac = 1/2. Thus we have a Hopf bifurcation. The fixed point is losing
stability and after that it is surrounded by a stable limit cycle (going towards lower values
of a). It is therefore supercritical. The phase portrait√for a = 1/2 is shown in Fig. 16. e)For
a = 1/2, the imaginary part of the eigenvalue is ω = 2/4.

f)

Problem 10.1.9
The map is
2xn
xn+1 = . (69)
1 + xn
The fixed points x are given by the equation
2x
x = . (70)
1+x
This yields
x2 + x = 2x , (71)
or x = 0 and x = 1. The stability is given by the derivative of the function f (x) = 2x/(1+x):
2
f 0 (x) = . (72)
(1 + x)2
 17

1.0

0.8

0.6

0.4

0.2

0.0

0.0 0.5 1.0 1.5

Figure 15: Phase portrait for a = 2 in problem 8.2.8.

This yields f 0 (0) = 2 and f 0 (1) = 12 . Hence the fixed point x = 0 is unstable and the fixed
point x = 1 is stable.

Problem 10.1.10
The map is
1
xn+1 = 1 + sin(xn ) . (73)
2

In Fig 17, we show the functions g(x) = x and h(x) = 1 + 21 sin x from which it is evident
that there is a unique fixed point that corresponds to the intersection of the curves.
The derivative of the function f (x) = 1 + 12 sin x is is

1
f 0 (x) = cos x . (74)
2
Since |f 0 (x)| ≤ 21 , the fixed point is stable.

Problem 10.1.12
Consider the map

g(xn )
f (xn ) = xn − . (75)
g 0 (xn )
 18

1.0

0.8

0.6

0.4

0.2

0.0

0.0 0.5 1.0 1.5

Figure 16: Phase portrait for a = 1/2 in problem 8.2.8.

x_8n +1<

x_n
-3 -2 -1 1 2 3

-1

-2

-3

Figure 17: Fixed point of the map xn+1 = 1 + 12 sin(xn ).

a) For the function g(x) = x2 − 4, this yields

x2n − 4
f (xn ) = xn − . (76)
2xn

b) The fixed points are given by

x2 − 4
x = x− . (77)
2x
This yields x2 − 4 = 0 or x = ±2.

c) The derivative is
1 2
f 0 (x) = − 2 , (78)
2 x
 19

and so

f 0 (±2) = 0 . (79)

This shows that the fixed points are supestable.

d) Try it out!

Problem 10.1.13
Differentiation of f (x) gives

g(x)g 00 (x)
f 0 (x) = . (80)
[g 0 (x)]2

Unless g 0 (x) = 0, g(x) = 0 corresponds to f 0 (x) = 0, i.e. a zero of g(x) corresponds to a

superstable fixed point for f (x).

Problem 10.3.11
The map is given by

xn+1 = −(r + 1)x − x2 − 2x3 . (81)

a) Clearly x = 0 is fixed point. It stability is determined by the derivative of f (x) =

−(r + 1)x − x2 − 2x3 :

f 0 (x) = −(r + 1) − 2x − 6x2 . (82)

This yields f 0 (0) = −1 − r and so the origin is stable for −2 < r < 0.

b) We notice that f 0 (0) = −1 for r = 0, which is the criterion for a flip bifurcation.

c) The period-2 cycles satisfy f (f (x)) − x = 0. Since f (x) − x is a factor in this poly-
nomial, we obtain by long division

[−(2 + r)x − x2 − 2x3 ][−r + rx − (1 + 2r + 2r2 )x2 − 4rx3 − 2(3 + 4r)x4 − 8x5 − 8x6 ] = 0 .
(83)

We are interested in a bifurcation for small negative values of r around x = 0. We can then
approximate the second polynomial with

−r + rx − x2 = 0 , (84)
 20

√ √
whose solutions are x = (r ± r2 − 4r)/2 ≈= ± −r. Thus the cycle exists for r < 0
and coalesces with x = 0 as r → 0− . The stability is determined by√the derivative √ of
f (f (x)) √ the points of the cycle, i. e. by the chain rule f 0 ( −r)f 0 (− −r) ≈
evaluated at √
(−1 − 2 −r)(−1 + 2 −r) = 1 − 4r > 1 and the 2-cycle is therefore unstable.

d) Use a cobweb to show that x → 0 for |x| small and r < 0 and x → ∞ for |x| small
and r > 0.

Problem 10.4.2
The map is given by
rx2n
xn+1 = . (85)
1 + x2n
The origin is always a fixed point. The other fixed points are given by x2 − rx + 1, or
√
r ± r2 − 4
x± = . (86)
2
These fixed points exist for r ≥ 2. Moreover
2rx
f 0 (x) = , (87)
(1 + x2 )2
which shows that the origin is always stable. At r = 2, the new fixed point is x = 1 and so
f 0 (1) = 1. This shows that bifurcation is a tangent bifurcation. At the fixed point x± , we
have
2
f 0 (x± ) = , (88)
1 + x2
which is always larger than zero. Thus there is no period-doubling. We can also prove this
explicitly for a period-2 cycle. The existence of such a cycle implies f [f (x)] = x. Since the
solutions to f (x) = x are also solutions to f [f (x)] = x, we can use long division to write
f [f (x)] = x as
[(1 + x2 )(1 + x(r + x + r2 x))]
. = 0. (89)
[1 + 2x2 + (1 + r2 )x4 )]
The first term in the numerator has no solution. The second term yields (r2 +1)x2 −rx+1 = 0,
which gives
s
r2 − 4(r2 + 1)
x = −r ± . (90)
2(1 + r2 )

which has no real solution, as anticipated. No chaos and no intermittency.

 21

Problem 10.4.3
The map is given by
xn+1 = 1 − rx2n . (91)
A superstable n-cycle has by definition [f n (x)]0 = 0. Using the chain rule, this is equivalent
to f 0 (x) = 0. In the present case, we have f (x) = 1 − rx2 and so f 0 (x) = 0 yields −2rx = 0,
i.e. x = 0. Moreover
f 3 (x) = 1 − r[1 − r(1 − rx2 )2 ]2 . (92)
The 3-cycle satisfies f 3 (x) = x, where x = 0 is in the cycle. Inserting x = 0 into f 3 (x) = x
gives
1 − r(1 − r)2 = 0 , (93)
which is the sought polynomial.

Problem 10.5.4
The Lyapunov exponent is λ = log r and so chaos exists for r > 1. This implies there can
be no periodic window 1 < r1 < r < r2 after the onset of chaos.

Problem 10.7.6
Since R2 is defined by superstability of the 4-cycle, we know that x = 0 is a point on that
cycle. Thus f 4 (0, R2 ) = 0. This yields a polynomial for R2 :
R2 − R22 + 2R23 − 5R24 + 6R25 − 6R26 + 4R27 − R28 = 0 . (94)
Since R2 − R22 is a factor of the above polynomial (since R2 − R22 = 0 defines the superstable
2-cycle), we find
1 + R22 (R2 − 2)(R2 − 1)(1 + R22 ) = 0 . (95)
There are roots R2 = 1.3107 and R2 = 1.9408. The function is plotted in Fig. 18.
Comparing αf 2 (x/α, R1 ) and αf 4 (x/α2 , R2 ) and matching coefficients through order x2
with R1 = 1 gives
R2 − R22 + 2R23 − 5R24 + 6R25 − 6R26 + 4R27 − R28 = 0 , (96)
α − 4R23 + 8R24 − 12R25 + 12R26 − 4R27 = 0 . (97)
The first equation is identical to Eq. (94). The second can be solved for α:
α = 4R23 − 8R24 + 12R25 − 12R26 + 4R27 . (98)
Inserting R2 = 1.3107 gives α = −2.44433. The other root gives α = −20.3309 and is not
relevant. Note that if we had demanded the coefficient of x4 to vanish, it would have given
α = −2.334.
 22

fHrL

1.0

0.5

r
0.5 1.0 1.5 2.0

-0.5

-1.0

Figure 18: Polynomial for R2 .

Problem 11.1.6
The decimal shift map is given by

xn+1 = 10xn (mod 1) . (99)

The map is shown is Fig. 19.

x_8n +1<
0.10

0.08

0.06

0.04

0.02

x_n
0.2 0.4 0.6 0.8 1.0

Figure 19: Decimal shift map.

a) A rational number y can be represented in decimal form

y = x0 .x1 x2 x3 .. , (100)

where either the sequence of numbers is finite or it is periodic. If yo ∈ Q, then we will

eventually end up at y = 0 if the sequence is finite, or a periodic orbit given by the period-
icity of the decimal representation. Thus a rational corresponds to a periodic orbit and Q
is countable. Since the derivative of the decimal shift map is 10 (see Fig. 19), every orbit is
 23

unstable.

b) A irrational number has a aperiodic decimal representation and so if y0 ∈ R\Q will

give rise to an aperiodic orbit. The irrational numbers are uncountable.

c) A number on the form x0 = c1 c2 ....cN cN ..., will after n = N iterations be mapped to

xn = cN cN ..... After n + 1 iterations we also have xn+1 = cN ... and so it is fixed point. Such
a number will be periodic and therefore x0 ∈ Q. The set is thus countable.

Problem 12.1.1
If either |a| > 1 or |b| > 1, the sequence {(xn , yn )} will wander off to infinity. If |a| = |b| = 1,
any (x0 , y0 ) is a fixed point. If |a| < 1 and |b| < 1, the sequence {(xn , yn )} converges to the
origin. The way the sequence approaches the origin depends on the sign of a and b and in
which quadrant you start i. e. of the point (x0 , y0 ). For example if 0 < a1 < and −1 < b < 0
and (x0 , y0 ) is in the first quadrant, the sequence will hop between the first and the third
quadrant. Likewise if 0 < a < 1 and −1 < b < 0 and (x0 , y0 ) is in the second quadrant, the
sequence hops between the second and third quadrant. The other cases, can be treated in
the same way.

Problem 12.1.5
a) We can write (x0 , y0 ) as (.a1 a2 a3 ..., .b1 b2 b3 ...), where ai and bi are either 0 or 1. If x0 < 21 ,
then a1 = 0 and x0 is multplied by two which corresponds to moving all digits one place to
the left. Thus x1 = .a2 a3 .... Then y0 is multiplied by a 12 which corresponds to moving all
digits one place to the right. This implies that B(x, y) = (.a2 a3 ...0b2 b3 ...) If x0 ≥ 12 , then
a1 = 1. Multiplying by 2 and subtracting by one corresponds to moving all digits to the
left and dropping a1 . Thus x1 = .a2 a3 .... Now y0 is multiplied by a 12 and we add 12 The
first operation is again moving all digits to the right while the second implies that the first
1
digit will be one (since 2=0.1 in the binary representation). Since a1 = 1 we can write this
as y = .1b2 b3 ... = .a1 b2 b3 ... and so in both case we can write B(x, y) = (.a2 a3 ...a1 b2 b3 ...).

b) The only possibility for a period-2 orbit is that x = 0.101010... or x = 0.01010... since B
moves the all digits to the left and B 2 (x, y) = (x, y). After two iteration of y0 , we know that
b1 b2 has been replaced by a2 a1 and so y = 0.010101 is the only possibility for x = 0.101010...
and vice versa (use the same argument to determine b3 ...). The number x = 0.101010...
equals
1 1
x = + 3 + ...
2 2
2
= . (101)
3
 24

Likewise, you can show that x = 0.010101... = 13 .

c) The only periodic orbits must be rational numbers because these numbers are either
finite or periodic in the decimal representation and the action on x is to move all the digits
to the left. y must also be rational, otherwise (x, y) cannot be periodic. In addition there
are additional requirements analogous to those derived in b); b1 = an , b2 = an−1 etc for an
n-cycle. Since the cartesian product Q × Q is countable, it follows.

d) Pick x irrational and y anything. x irrational implies that the orbit cannot be peri-
odic since the action on x is to move all the digits to the left. Since the irrationals are
uncoutable the results follows.