UNIVERSITY OF LJUBLJANA

FACULTY OF COMPUTER AND INFORMATION SCIENCE

LABORATORY OF MATHEMATICAL METHODS IN COMPUTER
AND INFORMATION SCIENCE

Aljaž Zalar

SOLVED EXERCISES FROM

EXAMS IN MATHEMATICAL
MODELLING
Study material

Ljubljana, 2022
 Introduction

The exercises on the following pages appeared on theoretical exams in the study years
2018/19–2021/22 for the course Mathematical Modelling, which is an elective course for
students enrolled in the undergraduate programme of Computer and Information Science
at the University of Ljubljana.
The exercises are divided in chapters, which are covered in the course, so that you can
try to solve them immediately after studying the appropriate chapter.
This study material is not peer-reviewed and so there can be some mistakes in the
solutions. If you notice a mistake, please send me an e-mail to [email protected].
For easier navigation within the document there are some shortcuts. If you click on the
symbol after the exercise, you will move to the corresponding solution. If you click on
the symbol at the solution, you will return to the text of the exercise. For the TeX
template of the document I am grateful to dr. Aleksandra Franc.
 Notation

N ... natural numbers

R ... real numbers
C ... complex numbers
√
i ... imaginary unit i = −1
f 0 (x) ... a derivative of a function of one variable
f (x, y, z) ... a function of three variables
fx (x, y, z) ... a partial derivative of f with respect to x
grad f (x, y, z) ... a gradient of a function of three variables
Rm×n ... the set of m × n real matrices
diag(d1 , d2 , . . . , dr ) ... a diagonal r × r matrix with d1 , d2 , . . . , dr on the diagonal
0m,n ... a m × n matrix with only zero entries
In ... a n × n identity matrix
Ai,j ... the entry in the i-th row and j-th column of the matrix A
A† ... the Moore-Penrose inverse of the matrix A
ker A ... the kernel of the matrix A
(Jf )(x) ... the Jacobian matrix of the vector function f evaluated in the point x
hu, vi ... the inner product between the vectors u and v
kuk ... the Euclidean norm of the vector u
 Contents

Introduction 3
Notation 5

Part 1. Exercises 9
Chapter 1. Linear systems 11
Chapter 2. Nonlinear systems 15
Chapter 3. Curves and surfaces 17
Chapter 4. Differential equations 21

Part 2. Solutions 25
Linear systems 26
Nonlinear systems 36
Curves and surfaces 42
Differential equations 54
 Part 1

Exercises
 CHAPTER 1

Linear systems

Task. 1.
Compute the singular value decomposition (SVD) of the matrix
" #
3 1 1
B= .
−1 3 1

Task. 2.
We are given the matrix A and the vector b:
   
1 1 2 0
   
 1 0 1   1 
A= , b =  .
   
 1 1 2   0 
   
2 0 2 2

a. What is the rank of A?

b. Compute one generalized inverse of A.
c. Determine all solutions of the system Ax = b.

Task. 3.
Let A, B be m × n matrices, m, n ∈ N, such that AT B = 0 and BAT = 0. Verify the
following statements:
a. Every column of A is perpendicular to every column of B.
Hint: What is the meaning of the entry in the i-th row and j-th column of AT B?

b. A† B = B † A = 0.
Hint: Remember the geometric meaning of A† b (resp. B † a), where b (resp. a) is a column in Rm ,
and use this for every column of the matrix B (resp. A).

c. Every column of AT is perpendicular to every column of B T .

Hint: What is the meaning of the entry in the i-th row and j-th column of (B T )T AT = BAT ?

d. BA† = AB † = 0.
Hint: Assuming (b) is true, this statement can be proved by plugging AT and B T into the
appropriate variables in (b).

e. (A + B)† = A† + B † .

Task. 4.
12 Linear systems

a. Prove that if AAT is an invertible matrix, then AT (AAT )−1 is the Moore–Penrose
inverse A† of the matrix A.

b. Find the point on the intersection of the planes x + y + z = 0 and x − y = 1 that

is closest to the origin following the next two steps:
(a) Write down the matrix of the system for the intersection and find its Moore-
Penrose inverse.
(b) Among all solutions of the system find the one closest to the origin.

Task. 5.
Let    
−1 1 1
   
A= 1 1 and −1 .
b=
 
 
−1 1 1

a. Find the Moore–Penrose inverse A† of the matrix A.

b. Does the system Ax = b have a solution?
c. If the system is solvable, find the solution closest to the origin. If the system
does not have a solution, find the vector x+ such that the error kAx+ − bk2 is the
smallest possible.
d. Find the Moore–Penrose inverse (A† )† of A† .

Task. 6.
We are given the following four points:
(0, 1), (−1, 0), (1, 2), (2, 3).
We would like to fit a function of the form ax2 + bx to these points.
a. Write down the matrix A of the corresponding system of linear equations.
b. Find the Moore-Penrose inverse A† .
c. Find the function of the above form that fits the points best according to the least
squares criterion.
d. Find one more generalized inverse of A.

Task. 7.
The system of equations
2x − y + z = 3,
−x + 2y − z = 1
can be expressed in the form Ax = b.
a. Find the Moore-Penrose inverse of A.
b. Describe the property uniquely characterizing the point A† b with respect to the
system.
c. Construct a matrix, which has the following matrices as their generalized inverses:
" # " #
3 2 0 0 0 0 3 2
, .
1 −1 0 0 0 0 1 −1
 13

Task. 8.

a. Construct any non-diagonal 3 × 2 matrix A whose singular values are 2 and 1.

b. Find the Moore-Penrose inverse A† of A.
c. Let b ∈ R3 . Describe the property uniquely characterizing point A† · b with respect
to the system Ax = b.

Task. 9.
Let  
−2 2
 
A=
 −3 −2 .

−2 −3

a. Find the matrix B ∈ R3×2 of rank 1, which is the closest to A in the Frobenius
norm.
b. Calculate kA − BkF .

Task. 10.
Let  
0 2
 
A=
 −2 −1 

−2 0
be a matrix.
a. Compute the truncated singular value decomposition of A.
b. Does there exist a matrix B ∈ R3×2 of rank 1 such that kA − BkF = 1? If yes,
compute it, otherwise justify, why it does not exist.

Task. 11.
Let A ∈ Rm×n , B ∈ Rp×r and C ∈ Rm×r be matrices. Consider the solutions of the matrix
equations:
(1) AXB = C.
Let G1 ∈ Rn×m and G2 ∈ Rr×m be generalized inverses of A and B, respectively.
a. Assume that C = AG1 CG2 B. Check that G1 CG2 solves (1).
b. Prove that if (1) is solvable, then C = AG1 CG2 B holds.
Hint: Multiply (1) from left and from right by appropriate matrices and use the definitions of
G1 , G2 .
c. Assume that (1) is solvable. Check that
X = G1 CG2 + Z − G1 AZBG2
solves (1) for any Z ∈ Rn×p .
 CHAPTER 2

Nonlinear systems

Task. 12.
Perform one step of Gauss-Newton method to approximate the least squares solution of
the system
f (x, y) = (2, 3, 1),
where
1
f : R2 → R3 , f (x, y) = (x2 + y 3 + 2, x + ey−1 , sin x + y 2 − 3).
2
For the initial approximation take (x0 , y0 ) = (0, 1).

Task. 13.
Using one step of Newton’s method approximate the solution of the system
sin x + cos y + exy = arctan(x + y) − xy = 0,
with the initial approximation (x0 , y0 ) = (0, 0).

Task. 14.
Let " # " #
f1 (x, y) x2 + y 2 − 10x + y
F (x, y) := =
f2 (x, y) x2 − y 2 − x + 10y
be a vector function and a = (2, 4) ∈ R2 a point.
a. Calculate the Jacobian matrix of the function F in the point a.
b. Calculate the linear approximation of F in the point a.
c. Perform one step of Newton’s method to find the approximate solution of the
system " #
1
F (x, y) =
25
with the initial approximation a.

Task. 15.
Let
x2 + y = 37,
x − y 2 = 5,
x+y+z =3
h iT
be a nonlinear system and v (0) = 0 0 0 a vector.

a. Compute the approximation v (1) of the solution of the system using one step of
Newton’s method.
16 Nonlinear systems

b. Compute the tangent plane to the surface given by the equation z = f (x, y), where
f (x, y) = 8xy + 4,
in the point (1, 1).

Task. 16.
Let f (x, y, z) = x2 + 3xy + yz 3 be a function of three variables.
a. Compute the gradient ∇f .
b. Perform one step of Newton’s method to approximate
the stationary point of f
1
using the initial approximation (x0 , y0 , z0 ) = 1, 0, 3 .
√

Task. 17.
Let
√ 1
π ln(x21 + x22 ) − √ sin(x1 x2 ) = ln(2π),
π
1
ex1 −x2 + √ cos(x1 x2 ) = 0,
π
h √ √ iT
be a nonlinear system and v (0) = π π a vector. Compute the approximation
v (1) of the solution of the system using one step of Newton’s method.
 CHAPTER 3

Curves and surfaces

Task. 18.
Let S be a surface given by z = g(x, y), where
g(x, y) = x3 − x2 y + y 2 − 2x + 3y − 2
is a differentiable function. Determine the tangent plane to S in the point (−1, 3) in the
parametric and implicit form.

Hint: Note that the parametric equation f : R2 → R3 of the surface S is

f (x, y) = (x, y, g(x, y)).

Task. 19.
Sketch the closed curves given in polar coordinates by
r1 (ϕ) = 1 + cos ϕ and r2 (ϕ) = 1 + sin ϕ.
Compute the area of the intersection of the bounded regions determined by the curves.

Hint: You will need the formulas sin2 x = 21 (1 − cos 2x) and cos2 x = 21 (1 + cos 2x) to
compute the area.

Task. 20.
Let
r(t) = (2 sin(2t), 2 cos(2t), 3t)
be the curve in parametric coordinates with t ∈ [0, 2π].
a. Sketch the curve in R3 .
b. Sketch all three projections of the curve in the xy-, xz- and yz-coordinate planes.
c. Compute the arc length of the curve.

Task. 21.
For the parametric curve
" # " #
x(t) 2t − t2
f (t) = = ,
y(t) 3t − t3
where t ∈ R, solve the following:
a. Find intersections with coordinate axes.
b. Find points at which the tangent is horizontal or vertical.
c. Find points where x0 (t) = y 0 (t) = 0.
d. Determine the asymptotic behaviour (limits as t → ±∞).
18 Curves and surfaces

e. Show that there are no self-intersections.

Hint: To notice that the curve does not have any self-intersections verify that the equality
1 − x(t) = 1 − x(s) implies that s = 2 − t and plug this into the equation y(t) = y(s).

f. Plot the curve.

Task. 22.
Let
r1 (ϕ) = 2 sin ϕ and r2 (ϕ) = 2 cos ϕ
be curves in polar coordinates.
a. Prove that both curves are circles and provide a sketch.
Hint. Try multiplying each equation by ri and expressing ri2 , ri sin ϕ and ri cos ϕ by x and y.

b. Compute the area of the region that lies inside both circles.

Task. 23.
Let
~r(t) = (x(t), y(t)) = (t3 − 4t, t2 − 4)
be the parametric curve.
a. Find all points where it intersects the coordinate axes.
b. Find the tangent to the curve at t = 1.
c. Find the points on the curve where the tangent is horizontal or vertical.
d. If there is a self-intersection, find it and compute the area inside the loop formed
by the curve.
e. Sketch the curve.

Task. 24.
Let
γ(t) = (t3 − t + 1, t2 )
be the parametric curve.
a. Find self-intersections of γ.
b. Find the angle at which γ intersects itself in the self-intersections.
c. Find the point at which γ reaches its global minimum in the direction of y-axis.

Task. 25.
Let
γ(t) = (2 cos(t), 2 sin(t), −t)T
be the parametric curve.
a. Sketch γ.
b. Parametrize γ with a natural parameter.
c. Find the length of γ between points (2, 0, 0) and (2, 0, 2π).
 19

Task. 26.
Two surfaces in the upper halfspace z > 0 are given by the following equations:
z2
Π : x2 + y 2 = Σ : x2 + y 2 = z.
2
Curve γ is the intersection of surfaces Π and Σ. Let P = (1, 1, 2) ∈ γ.
a. Find the angle at which the surfaces intersect at P .
b. Find the line tangent to γ at P .
c. Find the plane that is tangent to Σ at (1, 2, 5).

Task. 27.
Let
f (t) = (sin t, cos(3t)), t ∈ [0, 2π]
be the parametric curve.
a. Find all points where the curve intersects the coordinate axes.
b. Find all points on the curve where the tangent is horizontal or vertical.
c. Sketch the curve.

Task. 28.
Let
f (t) = (t3 − 5t2 + 3t + 11, t2 − 2t + 3), t∈R
be the parametric curve.
a. Find all points on the curve, where the tangent is horizontal or vertical.
b. Find all self-intersections.
c. Sketch the curve.

Task. 29.
Sketch the curve given in polar coordinates by
r(ϕ) = 2 + 4 sin(ϕ)
and compute the area of the smaller bounded region determined by the curve.
 CHAPTER 4

Differential equations

Task. 30.
Let
dy
(2) 2xy − 9x2 + (2y + x2 + 1) · =0
dx
be a differential equation.

a. Rewrite (2) in the form M (x, y)dx+N (x, y)dy = 0 and prove that this DE is exact
by checking the necessary and sufficient conditions involving partial derivatives of
M and N .
b. Solve the DE (2) with an initial condition y(0) = −3.

Task. 31.
Convert the differential equation
(3) y 00 + 11y 0 + 24y = 0
into the system of first order DEs, solve this system and recover the solution of the initial
DE (3).

Task. 32.
Solve the differential equation
1
3y 0 cos x + y sin x − = 0,
y2
given the initial condition y(0) = 1.

Hint: Note that this DE can be transformed into a first order linear nonhomogeneous DE
by multiplying it with an appropriate factor and introducing aZnew variable. To compute
R 1
tan x dx use the substitution u = cos x. Also remember that dx = tan x + C.
(cos x)2

Task. 33.
Find the general solution of the system
ẋ = 2x − 3y,
ẏ = x − 2y,
and sketch the phase potrait.
22 Differential equations

Task. 34.
Solve the differential equation
y 00 − 4y 0 + 5y = 8 cos x.
Find a solution to this DE which has a local extremum in the point (0, 2).

Task. 35.
Find the solution x(t), y(t) of the nonautonomous system of first order linear differential
equations
ẋ = 2x − y,
ẏ = −2x + y + 18t,
which satisfies x(0) = 1, y(0) = 0.

Hint: One of the particular solutions of the system above is of the form
xp (t) = At2 + Bt + C, yp (t) = Dt2 + Et + F,
where A, B, C, D, E, F are constants.

Task. 36.
Let
(4) y 0 = 2xy 2
be the differential equation with an initial condition y(0) = 1.
a. Find the exact solution of the DE (4).
b. Use Euler’s method with step size 0.2 to estimate y(0.4) and compare the result
to the exact value y(0.4).

Task. 37.
Find the general solution of the nonhomogeneous second order linear equation
ẍ + ẋ − 2x = t2 .

Task. 38.
Find the general solution of the differential equation
y 0 = 2x(1 + y 2 )
and the particular solution that satisfies y(1) = 0.

Task. 39.
For the system of nonlinear differential equations
ẋ = xy + 1, ẏ = x + xy,
solve the following:
a. Find stationary points.
b. Classify each stationary point as a saddle, source, sink or center.
c. Sketch the phase portrait of the system around each stationary point.
 23

Task. 40.
Solve the differential equation
xy 0 = y + 2x3
with the initial condition y(2) = 3.

Task. 41.
Solve the differential equation
y 00 + y 0 − 6y = 36x
with the initial conditions y(0) = y 0 (0) = 1.

Task. 42.
Find the solution y of the differential equation
x2 y 0 + xy + 3 = 0
with the initial condition y(1) = 1.

Task. 43.
Solve the following system of differential equations:
x0 (t) = −2x(t) + 5y(t),
y 0 (t) = x(t) + 2y(t),
with the initial conditions x(0) = y(0) = 1.

Task. 44.
Solve the following exact differential equation
2xy + (x2 + 3y 2 )y 0 = 0.

Task. 45.
Solve the differential equation
y 00 + 9y = 2x2 − 1
with the initial condition y(0) = y 0 (0) = 1.

Task. 46.
Consider the system of nonlinear differential equations
ẋ = x(3 − x − 2y),
ẏ = y(4 − 3x − y).

a. Find the stationary points of the system.

b. Compute the linearization of the system around the nontrivial stationary point,
i.e., the one with both coordinates being nonzero.
c. Solve the linear system from the previous question and sketch its phase portrait.
24 Differential equations

Task. 47.
Let
   
ẋ1 (t) x1 (t)
   
 ẋ2 (t)  = A  x2 (t)  ,
   
ẋ3 (t) x3 (t)
where A ∈ R3×3 , be a system of differential equations with the following three solutions:
     
1 1 1
e−t 
     
−1  , et  1  and e2t  2  .
     
1 1 4
a. Write down a general solution of the system.
b. Determine the matrix A.
c. Write down a third order differential equation with constants coefficients, which is
transformed into the above system.

Task. 48.
Solve the differential equation
(5) ẍ − ẋ − 4x = 2t + et .
 Part 2

Solutions
 Linear systems

Solution of the task 1.

We have to compute the orthogonal matrices
h i h i
U = u1 u2 ∈ R2×2 , V = v1 v2 v3 ∈ R3×3
and a diagonal rectangular matrix Σ ∈ R2×3 such that
B = U ΣV T .
We have that " #
11 1
BB T = ,
1 11
which implies
det(BB T − λI2 ) = (11 − λ)2 − 1 = (11 − λ − 1)(11 − λ + 1) = (12 − λ)(10 − λ).
So the eigenvalues of BB T are λ1 = 12, λ2 = 10 and hence
" √ #
12 0 0
Σ= √ .
0 10 0

The kernel of " # " #

12 0 −1 1
BB T − =
0 12 1 −1
h iT
contains the vector 1 1 and hence
" #
1 1
u1 = √ .
2 1

The kernel of " # " #

10 0 1 1
BB T − =
0 10 1 1
h iT
contains the vector 1 −1 and hence
" #
1 1
u2 = √ .
2 −1

Now the first two columns of V are

   
1 2
1 1   1 1  
v1 = √ · B T u1 = √  2  and v2 = √ · B T u2 = √  −1 .
12 6  10 5 
1 0
Finally,  
1
v1 × v2 1  
v3 = =√  2 
.
kv1 × v2 k 30 
−5
Linear systems 27

Solution of the task 2.

a. The left upper 2×2 submatrix has determinant −1 and hence the first two columns
are linearly independent. The third column is the sum of the first two and so
rank A = 2.

b. We choose an invertible 2×2 submatrix of A, replace it with its transposed inverse,

replace the other entries with 0s and transpose the matrix obtained:
#−1 T " # T
  
"
  1 1 0   
   0 1 0 0
 1 0 0  
 =  1 −1 0 0  .

G=  " # " #   
 0 0 0 
  0 0 0 0
0 0 0

c. The candidate for the solution is

 
1
 
Gb = 
 −1 .

0

We can check that A(Gb) = b and hence Gb is really a solution of the system. All
solutions are of the form
    
1 0 0 1 z1
    
Gb + (GA − I)z = 
 −1  +  0
  0 1   z2 
 
0 0 0 −1 z3
 
1 + z3
 
= −1 + z3  ,

−z3
where z1 , z2 , z3 ∈ R.
Solution of the task 3.

a.
0 = (AT B)i,j = dot product of the i-th row of AT and j-th column of B
= dot product of the i-th column of A and j-th column of B.
b. Two possible solutions to this part:

Geometrical solution: A† b is the vector with the smallest norm among all vectors
from the set
 
n 0
S(A, b) := x ∈ R : kb − Axk = min0 n
kb − Ax k .
x ∈R
28

Since every column bj of B is perpendicular to the span of the columns of A,

min kbj − Ax0 k = kbj k
x0 ∈Rn

and hence 0 ∈ S(A, bj ). Thus 0 = A† bj .

Computational solution: Let A = U ΣV T be the singular value decomposition

of A, where U ∈ Rm×m , V ∈ Rn×n are orthogonal matrices,
" #
D 0r,(n−r)
Σ= ∈ Rm×n
0m−r,r 0m−r,n−r
and D = diag(σ1 , . . . , σr ) ∈ Rr×r is a diagonal matrix with singular values σi of A
on the diagonal. We have that
" #
B11 B12
B=U V T,
B21 B22
where B11 ∈ Rr×r , B12 ∈ Rr×(n−r) , B21 ∈ R(m−r)×r and B22 ∈ R(m−r)×(n−r) . Now
we calculate A† B:
" # !
B 11 B 12
A† B = V Σ† U T VT


U
B21 B22
" #" #
D−1 0r,(m−r) B11 B12
(6) =V VT
0n−r,r 0n−r,m−r B21 B22
" #
D−1 B11 D−1 B12
=V VT
0n−r,r 0n−r,m−r
We know that AT B = 0:
" # !
B11 B12
0 = AT B = V ΣU T
VT


U
B21 B22
" #" #
D 0r,(m−r) B11 B12
(7) =V VT
0n−r,r 0n−r,m−r B21 B22
" #
DB11 DB12
=V VT
0n−r,r 0n−r,m−r

−1
Multiplying (7) with V −1 from the left side and V T from the right side we get
" #
DB11 DB12
0= .
0n−r,r 0n−r,m−r
In particular, DB11 = 0 and DB12 = 0. Since D is invertible this implies that
B11 = 0 and B12 = 0. Using this in (6) we conclude that A† B = 0.

c.
0 = (BAT )i,j = dot product of the i-th row of B and j-th column of AT
= dot product of the i-th column of B T and j-th column of AT .
Linear systems 29

d. Plugging AT , B T for A, B in (b) we obtain

0 = (AT )† B T = (A† )T B T = (BA† )T ⇒ 0 = BA† ,

0 = (B T )† AT = (B † )T AT = (AB † )T ⇒ 0 = AB † .

e. We have to check four properties of the Moore-Penrose inverse:

(A + B)(A† + B † )(A + B) = (AA† + AB † + BA† + BB † )(A + B)

= (AA† + BB † )(A + B)
= AA† A + AA† B + BB † A + BB † B
= A† + 0 + 0 + B †
= A† + B † ,
(A† + B † )(A + B)(A† + B † ) = (A† A + A† B + B † A + B † B)(A + B)
= (A† A + B † B)(A† + B † )
= A† AA† + A† AB † + B † BA† + B † BB †
=A+0+0+B
= A + B,
((A + B)(A† + B † ))T = (A† + B † )T (A + B)T
= (A† )T AT + (A† )T B T + (B † )T AT + (B † )T B T
= (AA† )T + 0 + 0 + (BB † )T
= AA† + BB † ,
((A† + B † )(A + B))T = (A + B)T (A† + B † )T
= AT (A† )T + AT (B † )T + B T (A† )T + B T (B † )T
= (A† A)T + 0 + 0 + (B † B)T
= A† A + B † B.

Solution of the task 4.

a. We denote G := AT (AAT )−1 . There are four requirements to check for G to be

equal to A† :

−1 
−1
AGA = A AT AAT A = AAT AAT


A = IA = A,

−1 
−1
GAG = GA AT AAT = G AT A AT A


= G,
 
−1 T 
−1 T
(AG)T = A AT AAT = AAT AAT


= I = AG,
 
T

T

 T
T −1 T

T T
T
T −1
(GA) = A AA A = A AA A = GA,

−1
where we used that AT A is symmetric in the last equality.
30

b. (a) The matricial form of the system is the following:

 
" # x " #
1 1 1   0
y  = .
1 −1 0   1
| {z } z |{z}
A b

The Moore-Penrose inverse of A is the following:

   
1 1
1 1 "1 #
† T T −1
  3 0  13 2 1 
A = A (AA ) =  1 −1 0 1 =  3 − 2 .
  
2 1
1 0 3
0

(b) The solution of the system of the smallest norm is

   
1 1 1
 31 2 2
x+ = A† b =
  1

3 − 12 
 = − 2 .
 
1
3
0 0

Solution of the task 5.

−1
a. Since rank A = 2, the matrix AT A ∈ R2×s is invertible and hence A† = AT A AT .
So,
" #−1 " 3 1 #−1 " 1 1 #
†
3 −1 T 8 8 T 4 2
− 14
A = ·A = 1 3 ·A = .
−1 3 8 8
− 14 12 41

b. If the systema Ax = b has a solution, then one of the solutions is x+ = A† b. So,

we have to compute Ax+ and see, if the result is b.
 
" 1# 0
−
x+ = A † b = 1 2
 
⇒ Ax+ =  −1 6= b.

2
0

So the system Ax = b does not have a solution.

c. The vector x+ such that the error kAx+ − bk2 is the smallest possible, is
" #
− 12
A† b = 1
.
2

d. The Moore–Penrose inverse of A† is always A by the symmetry in A and A† in the

conditions the MP inverse must satisfy.

Solution of the task 6.

Linear systems 31

a. The matricial form of the system is the following:

   
0 0 1
 " #  
1 −1 a 0
=  .
   
 
1 1  b 2
   
4 2 3
| {z } |{z}
A c

b. Since rank A = 2, also rank(AT A) = 2 and hence A† is equal to

" #−1 " #
† T −1 T
18 8 0 1 1 4
A = (A A) A =
8 6 0 −1 1 2
" 3 2
# " #
22
− 11 0 1 1 4
= 2 9
− 11 22
0 −1 1 2
" 7 1 2
#
0 22
− 22 11
= .
0 − 13
22
5
22
1
11

c. The solution a, b such that ax2 + bx fits the data best w.r.t. the least squares error
method is
 
1
" #   " #
7 1 2 5
†
0 22 − 22 0
11   11
Ac=  = 8 .
0 − 13
22
5
22
1 2
11   11
3

d. Another generalized inverse of A is

 " # T
0 0
 
0 0
" #
0 0 −1 21
 
 
G= " #−1 T  = .

 1 1

 0 0 2 − 12
   
4 2

Solution of the task 7.

a. The matricial form of the system is the following:

 
" # x " #
2 −1 1   3
y  = .
−1 2 −1   1
| {z } z |{z}
A b
32

Since rank A = 2, also rank(AAT ) = 2 and hence A† is equal to

 
2 −1 " #−1
† T T −1
  6 −5
A = A (AA ) =  −1 2  −5 6

1 −1
 
2 −1 " 6 5 #
  11 11
=−1 2 
 5 6
11 11
1 −1
 
7 4
11 11
4 7 

= 
 11 11 
.
1 1
11
− 11
b. Since A ∈ R2×3 and rank A = 2, it follows that the system Ax = b is solvable and
the kernel of A is one-dimensional. Hence, there is a one-dimensional family of
solutions of the system Ax = b. The vector A† b is the solution of the system of
the smallest norm among all solutions.
c. The matrix A is of size 4 × 2. By construction of some generalized inverses, the
matrix " #
3 2 0 0
1 −1 0 0
is a generalized inverse of any matrix of the form
" #−1 
3 2
 
 1 −1 ,
 
X
where X ∈ R2×2 is any matrix, and the matrix
" #
0 0 3 2
0 0 1 −1
is a generalized inverse of any matrix of the form
 
Y
 3 2 −1  ,
" # 
 
1 −1
where Y ∈ R2×2 is any matrix. Hence,
" #−1   1 2 
3 2
  5 5 
 1 −1  1 −3
 5 5
A = " #−1  =  1 2  .

 3 2   
  5 5 
1
1 −1 5
− 35

Solution of the task 8.

Linear systems 33

a. One possible solution is

    
√1 √1 0 2 0 "1 3 1
 21 23 
#
 12 2
  √ √1
A= − √12 0 0 1 2 2
= 2 2 .
√  
 2   √1 − √1
2 2
0 0 1 0 0 0 0
b. The Moore-Penrose inverse of A is
 T
" #T " 1 # √1 √1 0 " 3 #
†
√1
2
√1
2 2
0 0  12 2
 4
− 14 0
A = √
 2 − 12
√
 = −1 3 0 .
0
√1 − √12 0 1 0
2 4 4
0 0 1
c. If the system Ax = b is solvable, then the vector A† b is the solution of the system
of the smallest norm among all solutions. Otherwise the vector A† b is the unique
solution of the system of the smallest norm w.r.t. the least squares method, i.e.,
A(A† b) − b = min kAx − bk : x ∈ R2 .


Solution of the task 9.

Let
A = U ΣV T
be the truncated SVD of A, where
h i
U = u1 u2 ∈ R3×2 with U T U = I2 ,
h i
V = v1 v2 ∈ R2×2 with V T V = I2
and
Σ = diag(σ1 , σ2 ), where σ1 ≥ σ2 ≥ 0 are the singular values of A.
By the Eckart-Young theorem it follows that
B = σ1 u1 v1T ,
and hence
kA − BkF = kσ2 u2 v2T kF = σ2 ku2 v2T kF .
h iT
We write v2 = v2,1 v2,2 . Then
q
ku2 v2T kF = k[u2 v2,1 , u2 v2,2 ]kF = ku2 v2,1 k2F + ku2 v2,2 k2F
q
= (v2,1 )2 + (v2,2 )2 = kv2 kF = 1,
where in the third equality we used that ku2 kF = 1. So we need to compute only
σ1 , σ2 , v1 , u1 to determine B and kA − BkF . We have that
" #
17 8
AT A = ,
8 17
which implies
det(AT A − λI2 ) = (17 − λ)2 − 82 = (17 − λ − 8)(17 − λ + 8) = (9 − λ)(25 − λ).
So the eigenvalues of AT A are λ1 = 25, λ2 = 9 and hence
" #
5 0
Σ= .
0 3
34

The kernel of
" # " #
T
25 0 −8 8
A A− =
0 25 8 −8
h iT
contains the vector 1 1 and hence
" #
1 1
v1 = √ .
2 1
Now the first column of U is
   
0 0
1 1   = √1  −1 
  
u1 = Av1 = √  −5
5 5 2  2 
−5 −1
Finally,
   
0 0 0
1   1 h i 5 
B = 5u1 v1T = 5 · √  −1 · √ 1 1 =  −1 −1 
2   2 2  
−1 −1 −1
and
kA − BkF = 3.
Solution of the task 10.

a. We have that
" #
8 2
AT A = ,
2 5
which implies
det(AT A − λI2 ) = (8 − λ)(5 − λ) − 22 = λ2 − 13λ + 36 = (λ − 9)(λ − 4).
So the eigenvalues of AT A are λ1 = 9, λ2 = 4. Hence, Σ in the SVD of A = U ΣV T
is equal to
 
3 0
 
 0 2 .
 
0 0
The kernel of
" # " #
9 0 −1 2
AT A − =
0 9 2 −4
h iT
contains the vector 2 1 and hence
" #
1 2
v1 = √ .
5 1
Linear systems 35

Further on,
 
2
1 1  
u1 = Av1 = √  −5 .
3 3 5 
−4
The kernel of
" # " #
4 0 4 2
AT A − =
0 4 2 1
h iT
contains the vector 1 −2 and hence
" #
1 1
v2 = √ .
5 −2
Further on,
 
−4
1 1  
u2 = Av2 = √  0 .
2 2 5 
−2
So, the truncated SVD of A is equal to
" #
h i 3 0 h iT
A = u1 u2 v1 v2 .
0 2
b. By the Eckart-Young theorem the matrix B of rank 1, which minimizes the norm
kA − BkF is equal to σ1 u1 v12 . The distance kA − BkF is kσ2 u2 v2T kF = σ2 = 2.
Hence, there does not exist a matrix B of rank 1 satisfying kA − BkF = 1.
Solution of the task 11.

a.
A(G1 CG2 )B = AG1 CG2 B = C.
b. We multiply (1) from left by AG1 and from right by G2 B to obtain
(8) AG1 AXBG2 B = AG1 CG2 B.
Since G1 (resp. G2 ) is a generalized inverse of A (resp. B), we have that AG1 A = A
(resp. BG2 B = B). Hence, (8) implies that
(9) C = AXB = AG1 CG2 B,
where in the first equality we used (1). This proves (b).

c.
AXB = A(G1 CG2 + Z − G1 AZBG2 )B
= AG1 CG2 B + A(Z − G1 AZBG2 )B
= C + AZB − AG1 AZBG2 B
= C + AZB − AZB = C,
36

where we used (b) in the second equality and the definitions of G1 , G2 in the third
equality.

Note: If (1) is solvable, then all solutions are of the form given in (c). Indeed,
if X solves (1), then X = G1 CG2 + X − G1 AXBG2 , which means that Z = X is
one appropriate choice.

Nonlinear systems

Solution of the task 12.

We define a vector function F : R2 → R3 by the rule
1
F (x, y) := f (x, y) − (2, 3, 1) = (x2 + y 3 , x + ey−1 − 3, sin x + y 2 − 4).
2
After one step of Gauss-Newton method the approximate of the least squares solution of
the system F (x, y) = (0, 0, 0) will be
" # " #
x1 x0
= − ((JF )(x0 , y0 ))† F (x0 , y0 ).
y1 y0
We have
   
2x 3y 2 0 3
   
(JF )(x, y) = 
 1 ey−1 
 ⇒  1 1 .
(JF )(0, 1) =  
cos x y 1 1
Since  
" # 0 3 " #
T 0 1 1   2 2
((JF )(0, 1)) · (JF )(0, 1) =  1 1 =
3 1 1   2 11
1 1
is invertible, it follows that
 −1
† T
((JF )(0, 1)) = ((JF )(0, 1)) · (JF )(0, 1) ((JF )(0, 1))T
" #−1 " #
2 2 0 1 1
=
2 11 3 1 1
" #" #
1 11 −2 0 1 1
=
18 −2 2 3 1 1
" #  1 1 1 
1 −6 9 9 −3 2 2
= = 
18 6 0 0 1
0 0
3

and  
  1  
− 13 1 1 37
" # " #
x1 0 2 2   12
= −   −2  =  .
y1 1 1
0 0

7
 2
3 −2 3
Nonlinear systems 37

Solution of the task 13.

We define a vector function f : R2 → R2 by the rule
" # " #
f1 (x, y) sin x + cos y + exy
f (x, y) = = .
f2 (x, y) arctan (x + y) − xy

We are approximating the zero f (x, y) = 0 of f using one step of Newton’s method with
the initial approximation (x0 , y0 ) = (0, 0). The Jacobian Jf (x, y) of f is
   
∂f1 ∂f1
∂x ∂y
cos x + yexy − sin y + xexy
Jf (x, y) =  ∂f2 ∂f2
=
1 1
.
∂x ∂y (x+y)2 +1
−y (x+y)2 +1
−x

So
" # " #
x1 x0
= − (Jf (x0 , y0 ))−1 f (x0 , y0 )
y1 y0
" #−1 " #
1 0 2
=−
1 1 0
" #" # " #
1 0 2 −2
=− = .
−1 1 0 2

Solution of the task 14.

a.
" # " #
2x − 10 2y + 1 −6 9
JF (a) = (a) = .
2x − 1 −2y + 10 3 2

b.
" #
x−2
LF,a (x, y) = F (a) + JF (a)
y−4
" # " #" #
4 −6 9 x−2
= +
26 3 2 y−4
" #
−6x + 9y − 20
= .
3x + 2y + 12

c. We are searcing for zeroes of the vector function

" #
1
G(x, y) = F (x, y) −
25
38

with the initial approximation (x0 , y0 ) = a. One step of Newton’s method:

" # " #
x1 x0
= − JG(a)−1 G(x0 , y0 )
y1 y0
" # " #−1 " #
2 −6 9 3
= −
4 3 2 1
" # " #" #
2 1 2 −9 3
= +
4 39 −3 −6 1
" #
1 25
= .
13 47

Solution of the task 15.

a. We define a vector function of a vector variable F : R3 → R3 by

   
F1 (x, y, z) x2 + y − 37
   2

 F2 (x, y, z)  =  x − y − 5  .
F (x, y, z) =    
F3 (x, y, z) x+y+z−3

We are searching for the solution of F (x, y, z) = 0 using Newton’s method. We

have that

−1
v (1) = v (0) − JF (v (0) ) F (v (0) ),

where

   
∂F1 (x,y,z) ∂F1 (x,y,z) ∂F1 (x,y,z)
∂x ∂y ∂z
2x 1 0
∂F2 (x,y,z) ∂F2 (x,y,z) ∂F2 (x,y,z)
   
JF (x, y, z) = 
 ∂x ∂y ∂z
= 1
  −2y 0 

∂F3 (x,y,z) ∂F3 (x,y,z) ∂F3 (x,y,z)
∂x ∂y ∂z
1 1 1

is the Jacobian matrix of F . So

 
0 1 0
 
JF (v (0) ) = 
 1 0 0 .

1 1 1
Nonlinear systems 39

−1
We compute JF (v (0) ) using Gaussian elimination:
   
0 1 0 1 0 0 1 0 0 0 1 0
   
 1
 0 0 0 1 0  ∼ 
 |{z}  0 1 0 1 0 0 

1 1 1 0 0 1 `1 ↔`2 1 1 1 0 0 1
|  {z 
}
(0)
JF (v ) I3
 
1 0 0 0 1 0
 
∼
|{z}
 0 1 0 1
 0 0 
.
`3 =`3 −`1 −`2
0 0 1 −1 −1 1
|  {z }
(0) −1


I3 JF (v )

So     
0 1 0 −37 5
    
v (1) = − 
 1 00 

 −5  =  37  .
  
−1 −1 1 −3 −39
b. The surface is parametrized by
   
" # x x
x    
7→  y = y .
y    
f (x, y) 8xy + 4

The tangent plane to the surface in the point (1, 1) is

     
1 1 0
     
L(x, y) =  1  +  0  (x − 1) +  1  (y − 1)
     
∂f (1,1) ∂f (1,1)
f (1, 1) ∂x ∂y
     
1 1 0
     
= 1 + 0  (x − 1) +  1  (y − 1)
     
(8xy + 4)(1, 1) (8y)(1, 1) (8x)(1, 1)
     
1 1 0
     
= 1  +  0  (x − 1) +  1  (y − 1)
    
12 8 8
 
x
 
= y ,

−4 + 8x + 8y
or z = −4 + 8x + 8y.
Solution of the task 16.
40

a.
   
∂f
∂x
(x, y, z) 2x + 3y
   
∂f
   
∇f (x, y, z) =  (x, y, z)  =  3x + z 3 .
 ∂y   
   
∂f
∂z
(x, y, z) 3yz 2
h iT
b. We are searching for the solution of ∇f (x, y, z) = 0 0 0 using Newton’s
method. We have that

−1
v (1) = v (0) − J(∇f )(v (0) ) (∇f )(v (0) ),

where
   
∂ 2 f (x,y,z) ∂ 2 f (x,y,z) ∂f (x,y,z)
∂x2 ∂x∂y ∂x∂z
2 3 0
   
∂ 2 f (x,y,z) ∂ 2 f (x,y,z) ∂f (x,y,z)
   
J(∇f )(x, y, z) =   =  3 0 3z 2 
 ∂x∂y ∂y 2 ∂y∂z   
   
∂ 2 f (x,y,z) ∂ 2 f (x,y,z) ∂f (x,y,z)
∂x∂x ∂y∂z ∂z 2
0 3z 2 6yz

is the Jacobian matrix of ∇f . So

 
2 3 0
 
 (0)

J(∇f )(v ) = 
 3 0 1
.

 
0 1 0

−1
We compute J(∇f )(v (0) ) using Gaussian elimination:
   
2 3 0 1 0 0 2 3 0 1 0 0
   
 3 0 1 0 1 0  ∼  0 1 0 0 0 1 
  |{z}  
`2 ↔`3
0 1 0 0 0 1 3 0 1 0 1 0
| {z 
}
J(∇f )(v (0) ) I3
 
1 3
1 0 0 2 0 −2
 
 
∼
|{z} 
 0 1 0 0 0 1 

`1 = 21 (`1 −3`2 )
 
3 0 1 0 1 0
 
1 3
1 0 0 2 0 −2
 
 
∼
|{z} 
 0 1 0 0 0 1 .

`3 =`3 −3`1  
0 0 1 − 32 1 92
| {z }
−1


I3 J(∇f )(v (0) )
Nonlinear systems 41

So
     
1
1 2
0 − 32 2
  0
     
(1)
    1
  
v  0 − 0 0
= 1 
  3 + 3 √3  =  0  .
     
0
     
√1 − 32 1 9 2
√
3 2 3 3

Solution of the task 17.

We define a vector function of a vector variable F : R2 → R2 by
   √ 
F1 (x1 , x2 ) π ln(x21 + x22 ) − √1 sin(x1 x2 ) − ln(2π)
π
F (x1 , x2 ) =  = .
F2 (x1 , x2 ) ex1 −x2 + √1
π
cos(x1 x2 )

We are searching for the solution of F (x1 , x2 ) = 0 using Newton’s method. We have that

−1
v (1) = v (0) − JF (v (0) ) F (v (0) ),

where
 ∂F1 (x1 ,x2 ) ∂F1 (x1 ,x2 )

∂x1 ∂x2
JF (x1 , x2 ) =  
∂F2 (x1 ,x2 ) ∂F2 (x1 ,x2 )
∂x1 ∂x2
 √ √ 
π x22x+x1 2 − x2
√
π
cos(x1 x2 ) π x22x+x2 2 − x1
√
π
cos(x1 x2 )
1 2 1 2
= 
x1 −x2 x2 x1 −x2 x1
e − √
π
sin(x1 x2 ) −e − √
π
sin(x1 x2 )

is the Jacobian matrix of F . So

" # " #
1 − (−1) 1 − (−1) 2 2
JF (v (0) ) = = .
1 −1 1 −1

−1
We compute JF (v (0) ) using Gaussian elimination:

 
1
1 1 0
" #
2 2 1 0 2
∼  
1 −1 0 1 |{z}
` 0 −2 − 1
1
| {z } `1 = 21 , 2
` =` − 1 `

JF (v (0) ) I2 2 2 2 1
 
1 0 14 21
∼   .
1 1
0 1 4 −2
|{z}
`2 =− 12 `2 ,
1
| {z }
`1 =`1 + 2 `2 
−1
I2 JF (v (0) )
42

So
 √    √ 
1 1
π 4 2
π ln(2π) − ln(2π)
v (1) = √ −  
1
π 4
− 12 1− √1
π
 √ √ 
π − 14 π ln(2π) + 41 ln(2π) − 21 + 1
√
2 π
= √ √

π − 14 π ln(2π) + 41 ln(2π) + 12 − 1
√
2 π
" #
1.20
≈ .
1.64

Curves and surfaces

Solution of the task 18.

The parametric form r : R2 → R3 of the tangent plane to S in the point (x, y) = (−1, 3)
is

r(u, v) = f (−1, 3) + ufx (−1, 3) + vfy (−1, 3)

= (−1, 3, g(−1, 3)) + u · (1, 0, gx (−1, 3)) + v · (0, 1, gy (−1, 3)).

We have

g(−1, 3) = (−1)3 − (−1)2 · 3 + 32 − 2(−1) + 3 · 3 − 2 = 14,

gx (x, y) = 3x2 − 2xy − 2 ⇒ gx (−1, 3) = 3 − 2(−1)3 − 2 = 7,
2
gy (x, y) = −x + 2y + 3 ⇒ gy (−1, 3) = −1 + 2 · 3 + 3 = 8.

Hence,

r(u, v) = (−1, 3, 14) + u(1, 0, 7) + v(0, 1, 8) = (−1 + u, 3 + v, 14 + 7u + 8v).

For the implicit form of the tangent plane we need its normal
     
1 0 −7
     
~n := fx (−1, 3) × fy (−1, 3) =  0  ×  1  =  −8  .
     
7 8 1

Hence, the implicit form

(x, y, z) · ~n = f (−1, 3) · ~n
is
−7x − 8y + z = 7 − 24 + 14 = −3.

Solution of the task 19.

The sketch of the curves and the bounded region in the intersection is the following, where
r1 (ϕ) is the blue curve and r2 (ϕ) is the orange one:
Curves and surfaces 43

2.0

1.5

1.0

0.5

-1.5 -1.0 -0.5 0.5 1.0 1.5 2.0

-0.5

-1.0

-1.5

Since sin ϕ and cos ϕ are periodic function with a period 2π, it is enough to restrict
ourselves to the interval [0, 2ϕ]. First we have to determine the points, where the curves
intersect. As seen from the sketch, one of the points is the origin (0, 0), where both polar
radii are 0. This is true for ϕ = π for r1 and ϕ = 3π 2
for r2 . The other intersections can
occur for nonzero radii, in which case they have to be the same at the same angle. We
have that
 
π 5π
r1 (ϕ) = r2 (ϕ) ⇔ sin ϕ = cos ϕ ⇔ ϕ ∈ , .
4 4

So the other two intersections are points

 π  π  π   π 
A = r1 cos , r1 sin
4 4 4 4
and
        
5π 5π 5π 5π
B= r1 cos , r1 sin .
4 4 4 4

We see from the

 sketch that the intersection consists of the area enclosed byπ r1 (ϕ) 5πon the
interval π4 , 5π
 
4
and the area enclosed by r2 (ϕ) on the union of intervals 0, 4
∪ 4
, 2π .
Hence,
Z 5π Z π
1 2π
Z
1 4
2 2 1 4
area = (r1 (ϕ)) dϕ + (r2 (ϕ)) dϕ + (r2 (ϕ))2 dϕ.
2 π4 2 5π4 2 0

Further on,
Z 5π Z 5π
4 4
2
(1 + cos ϕ) dϕ = (1 + 2 cos ϕ + cos2 ϕ) dϕ
π π
4 4
 5π4
√

3 1 3
= ϕ + 2 sin ϕ + sin 2ϕ = π − 2 2,
2 4 π 2
4
44

where we used cos2 x = 21 (1 + cos 2x) in the second equality. Similarly

π
Z 2π Z
4
2
(1 + sin ϕ) dϕ + (1 + sin ϕ)2 dϕ =
5π
4
0
π
Z 0 Z
4
2
= (1 + sin ϕ) dϕ + (1 + sin ϕ)2 dϕ
−3 π4 0
Z π   π4
4 3 2 1
= (1 + sin ϕ) dϕ = ϕ − 2 cos ϕ − sin 2ϕ
− 3π
4
2 4 −3π
4

3 √
= π − 2 2,
2

√
where we used sin2 x = 12 (1 − cos 2x) in the third equality. So, area = 23 π − 2 2.

Solution of the task 20.

The sketch of the curve is the following:

The sketch of the projections to xy-, xz- and yz-planes are:

Curves and surfaces 45

15 15

10 10

-2 -1 1 2
5 5

-1

-2 -2 -1 1 2 -2 -1 1 2

The arc length is the following:

Z 2π Z 2π
0
kr (t)k dt = k(4 cos(2t), −4 sin(2t), 3)k dt
0 0
Z 2π q
= 16 cos2 (2t) + 16 sin2 (2t) + 9 dt
0
Z 2π √
= 25 dt = 10π.
0

Solution of the task 21.

a.
x(t) = 0 ⇔ 2t − t2 = 0 ⇔ t(t − 2) = 0 ⇔ t ∈ {0, 2},
√ √
y(t) = 0 ⇔ 3t − t3 = 0 ⇔ t(3 − t2 ) = 0 ⇔ t ∈ {− 3, 0, 3}.

Intersections with y-axis: (0, 0), (0, −2).

√ √
Intersections with x-axis: (−2 3 − 3, 0), (0, 0), (2 3 − 3, 0).

b.
ẋ(t) = 0 ⇔ 2 − 2t = 0 ⇔ t = 1,
2
ẏ(t) = 0 ⇔ 3 − 3t = 0 ⇔ 1 − t2 = 0 ⇔ t ∈ {−1, 1}.

Horizontal tangent: (−3, −2), vertical tangents: none.

c. From the part above (1, 2).

" # " #
−∞ −∞
d. lim f (t) = , lim f (t) = .
t→−∞ ∞ t→∞ −∞
46

e. Assume that t 6= s.

1 − x(t) = 1 − x(s) ⇔ 1 − 2t + t2 = 1 − 2s + s2
⇔ (1 − t)2 = (1 − s)2
⇔ 1 − t ∈ {1 − s, s − 1}.

Since t 6= s, 1 − t = s − 1 and hence s = 2 − t. Thus

y(t) = y(2 − t) ⇔ 3t − t3 = 3(2 − t) + (2 − t)3

⇔ t3 − 3t2 + 3t − 1 = 0
⇔ (t − 1)3 = 0 ⇔ t = 1.

But then s = 2 − 1 = 1 and s = t.

f. Using the information above, the sketch of the curve is the following:

-8 -6 -4 -2

-2

-4

-6

-8

Solution of the task 22.

a. We have

r12 = r1 · 2 sin ϕ ⇒ x2 + y 2 = 2y ⇔ x2 + (y − 1)2 = 1,

r22 = r2 · 2 cos ϕ ⇒ x2 + y 2 = 2x ⇔ (x − 1)2 + y 2 = 1,

where we used ri2 = x2 + y 2 , ri cos ϕ = x and ri sin ϕ = y, i = 1, 2. The first

equation r1 (ϕ) is the equation of the circle with center at (0, 1) and radius 1, and
the second r2 (ϕ) is the circle with center at (1, 0) and radius 1. The sketch is the
following
Curves and surfaces 47

2.0

1.5

1.0

0.5

-1.0 -0.5 0.5 1.0 1.5 2.0

-0.5

-1.0

b. The area can be computed using elementary geometry or using the formula for
integrals of regions in polar coordinates. If A is the shaded region in the plot, then

1 π/2
Z Z π/2
2
A= (2 cos ϕ) dϕ = (1 + cos(2ϕ)) dϕ
2 π/4 π/4
 π/2
sin(2ϕ) π 1
= ϕ+ = − ,
2 π/4 4 2
π
and the total area is 2A = − 1.
2
Solution of the task 23.

a. The intersections with the x-axis correspond to y(t) = 0:

y(t) = 0 ⇔ t2 − 4 = 0 ⇔ t = t1 = −2, t = t2 = 2.
Hence, there are two intersections with the x-axis:
P = ~r(−2) = (0, 0), ~r(2) = (0, 0).
The intersections with the y-axis correspond to x(t) = 0:
x(t) = 0 ⇔ t3 − 4t = 0 ⇔ t(t2 − 4) = 0 ⇔ t = t1 , t = t2 , t = t3 = 0.
Hence, there are three intersections with the y-axis:
~r(−2) = (0, 0), ~r(2) = (0, 0), Q = ~r(0) = (0, −4).
b. The tangent to the curve at the point t = t0 is
~s(λ) = ~r(t0 ) + λ (~r)0 (t0 ).
Hence,
~s(λ) = (−3, −3) + λ(3t2 − 4, 2t)(1) = (−3, −3) + λ(−1, 2) = (−3 − λ, −3 + 2λ).
c. The tangent to the curve is horizontal at the points where y 0 (t) = 0. Hence,
y 0 (t) = 0 ⇔ 2t = 0 ⇔ t = 0,
and the point is Q = (0, −4).

The tangent to the curve is vertical at the points where x0 (t) = 0. Hence,
2 2
x0 (t) = 0 ⇔ 3t2 − 4 = 0 ⇔ t4 = √ , t5 = − √ ,
3 3
48

and the points are

   
16 8 16 8
R = − √ ,− ≈ (3.079, −2.67), S= √ ,− ≈ (−3.079, −2.67).
3 3 3 3 3 3

d. Self-intersections correspond to different values of t1 , t2 , where ~r(t1 ) = ~r(t2 ):

~r(t1 ) = ~r(t2 ) ⇔ t31 − 4t1 = t32 − 4t2 and t21 − 4 = t22 − 4

⇔ t31 − 4t1 = t32 − 4t2 and t21 = t22
⇔ t31 − 4t1 = (−t1 )3 + 4t1 and t2 = −t1
⇔ 2t1 (t21 − 4) = 0 and t2 = −t1
⇔ t1 = 2, t2 = −2 or t1 = −2, t2 = 2.

Hence, the only self-intersection is the point P .

The area A inside the loop formed by the curve between t1 = −2 and t2 = 2 is

1 2  3
Z

A= (t − 4t)2t − (t2 − 4)(3t2 − 4) dt
2 −2
1 2  4
Z

= −t + 8t2 − 16 dt.
2 −2

Since
t4 − 8t2 + 16 = (t2 − 4)2 = (t − 2)2 (t + 2)2 ,
we have that
 4 
−t + 8t2 − 16 = t4 − 8t2 + 16

for every t ∈ R. Hence,

2 2
1 t5 t3
Z 
1 4 2

512
A= t − 8t + 16 dt = − 8 + 16t = .
2 −2 2 5 3 −2 30

e. The sketch of the curve is the following:

-6 -4 -2 2 4 6
-1

-2

-3

-4

Solution of the task 24.

Curves and surfaces 49

a. Self-intersections are points such that γ(t1 ) = γ(t2 ) for some t1 6= t2 :

γ(t1 ) = γ(t2 ) ⇔ t31 − t1 + 1 = t32 − t2 + 1 and t21 = t22 ,

⇔ t31 − t1 + 1 = −t31 + t1 + 1 and t2 = −t1 ,
⇔ 2t1 (t21 − 1) and t21 = t22 ,
⇔ t1 = 0, t2 = 0 or t1 = −t2 = 1 or t1 = −t2 = −1.

Hence, the self-intersection is γ(1) = γ(−1) = (1, 1).

b. The angle at which γ intersects itself in the self-intersection is the angle between
the tangents γ1 , γ2 in (1, 1):

γ1 (λ) = (1, 1) + λ · γ̇(1) = (1, 1) + λ (3t2 − 1, 2t)(1) = (1, 1) + λ(2, 2),

γ2 (λ) = (1, 1) + λ · γ̇(−1) = (1, 1) + λ (3t2 − 1, 2t)(−1) = (1, 1) + λ(−4, −2).

Hence,
     
h(2, 2), (−4, −2)i −8 − 4 −3
arccos = arccos √ √ = arccos √ ≈ 2.82.
k(2, 2)k k(−4, −2)k 8 20 10
So, the angle between the tangents is π − 2.82 ≈ 0.32.

c. The point at which γ reaches a a global minimum in the y-directions satisfies

d 2
dt
(t ) = 0. Hence, 2t = 0 and t = 0. The point is (1, 0).

Solution of the task 25.

a. The sketch of γ is the following:

50

b. The natural parameter s(t), which measures the arc length between the points
γ(0) and γ(t) is
Z t Z t
0
s(t) = kγ (u)k du = k(−2 sin u, 2 cos u, −1k du
0 0
Z tp
= 4 sin2 u + 4 cos2 u + 1 du
0
Z t√ √
= 5 du = t 5.
0

Hence, t(s) = √s and the parametrization of the curve with the natural parameter
5
s is      
s s s
γ(s) = 2 cos √ , 2 sin √ , − √ .
5 5 5
c. The point (2, 0, 0) corresponds to t = 0, while (2, 0, 2π) to t = −2π.
√ Hence, the
arc length between this points equals by symmetry to s(2π) = 2π 5.
Solution of the task 26.

a. The angle at which the surfaces intersect at P is the angle between the normals to
their tangent planes at the point P , i.e., between their gradients at the point P :
 
h(grad Π)(P ), (grad Σ)(P )i
arccos
kgrad Π)(P )k kgrad Σ)(P )k
 
h(2x, 2y, −z)(P ), (2x, 2y, −1)(P )i
= arccos
k(2x, 2y, −z)(P )k k(2x, 2y, −1)(P )k
   
h(2, 2, −2), (2, 2, −1)i 10
= arccos = arccos √ √
k(2, 2, −2)k k(2, 2, −1)k 12 9
 
5
= arccos √ ≈ 0.28.
3 3
b. The intersection of the surfaces satisfies
(x2 + y 2 )2
x2 + y 2 = ⇒ 2 = x2 + y 2 ⇒ z = 2.
2
So this is a circle with the parametrization
√ √ 
γ(t) = 2 · cos t, 2 · sin t, 2 .

The tangent to this circle in the point (1, 1, 2), which corresponds to t = π4 , is
  π 
`(λ) = (1, 1, 2) + λ · γ 0
 √4 √   π 
= (1, 1, 2) + λ · − 2 · sin t, 2 · cos t, 0
4
= (1, 1, 2) + λ · (−1, 1, 0).
c. The tangent plane to Σ at Q := (a, b, c) = (1, 2, 5), which is given implicitly by the
equation
F (x, y, z) = 0,
where
F (x, y, z) := x2 + y 2 − z,
Curves and surfaces 51

is determined by
     
∂F ∂F ∂F
0= (Q) · (x − a) + (Q) · (y − b) + (Q) · (z − c)
∂x ∂y ∂z
= ((2x)(Q)) · (x − 1) + ((2y)(Q)) · (y − 2) + ((−1)(Q)) · (z − 5)
= 2(x − 1) + 4(y − 2) − (z − 5).
Solution of the task 27.

a. The intersections with the x-axis correspond to y(t) = 0:

n π π 5π 7π 3π 11π o
cos(3t) = 0 ⇔ t ∈ , , , , , .
6 2 6 6 2 6
The corresponding points are
π   5π   π  1 
f =f = sin , 0 = ,0 ,
6  6
π 6 2
π 
f = sin , 0 = (1, 0),
2 2
 7π   11π   7π   1 
f =f = sin , 0 = − , 0 ,
6 6 6 2
 3π   3π 
f = sin , 0 = (−1, 0).
2 2
The intersections with the y-axis correspond to x(t) = 0:
sin t = 0 ⇔ t ∈ {0, π, 2π}.
The corresponding points are
f (0) = f (2π) = (0, 1),
f (π) = (0, −1).
b. The tangent to the curve is horizontal at the points where y 0 (t) = 0. Hence,
n π 2π 4π 5π o
0
(cos(3t)) = 0 ⇔ −3 sin(3t) = 0 ⇔ t ∈ 0, , , π, , , 2π .
3 3 3 3
The corresponding points are
f (0) = f (2π) = (0, 1),
 π   √3 
f = , −1 ,
3 √2
 2π   3 
f = ,1 ,
3  2 
f (π) = 0, −1 ,
 4π   √3 
f = − ,1 ,
3 √2
 5π   3 
f = − , −1 .
3 2
The tangent to the curve is vertical at the points where x0 (t) = 0. Hence,
n π 3π o
0
(sin t) = 0 ⇔ cos t = 0 ⇔ t ∈ , .
2 2
52

The corresponding points are

π 
f = (1, 0),
2
 3π 
f = (−1, 0).
2
c. The sketch of the curve is the following:

Solution of the task 28.

a. The tangent is horizontal in the points where y 0 (t) = 0:

2t − 2 = 0 ⇔ t = 1.
The corresponding point is
f (1) = (1 − 5 + 3 + 11, 1 − 2 + 3) = (10, 2).
The tangent is horizontal in the points where x0 (t) = 0:
√  
2 10 ± 100 − 36 1
3t − 10t + 3 = 0 ⇔ t1,2 = ∈ 3, .
6 3
The corresponding points are
f (3) = (27 − 45 + 9 + 11, 9 − 6 + 3) = (2, 6),
 
−1 −3 −2 −2 −1 310 22
f (3 ) = (3 −5·3 + 1 + 11, 3 −2·3 + 3) = , ≈ (11.5, 2.4).
27 9
Curves and surfaces 53

b. The curve has self-intersections, where f (t) = f (s) for t 6= s. We have that
t3 − 5t2 + 3t + 11 = s3 − 5s2 + 3s + 11
⇔ t3 − s3 = 5(t2 − s2 ) − 3(t − s)
(10) ⇔ t2 + ts + s2 = 5(t + s) − 3,
where we divided by t − s in the last line. Further on,
t2 − 2t + 3 = s2 − 2s + 3
⇔ t2 − s2 = 2(t − s)
(11) ⇔ t + s = 2,
where we divided by t − s in the last line. We use (11) in (10):
t2 + t(2 − t) + (2 − t)2 = 10 − 3 = 7
and hence
0 = t2 − 2t − 3 = (t − 3)(t + 1).
The solutions are t1 = 3, t2 = −1 with the correspoding s1 = −1 and s2 = 3. So
the only point of self-intersection is f (−1) = f (3) = (2, 6).
c. We compute
lim f (t) = (−∞, ∞) and lim f (t) = (∞, ∞).
t→−∞ t→∞

The sketch of the curve is the following:

10

-10 5 10

Solution of the task 29.

Since r(ϕ) is periodic with a period 2π, we can restrict r(ϕ) to the interval [0, 2π]. The
sketch of r(ϕ) is the following:

3 4 5 6

-2
54

Let us write down r(ϕ) for various ϕ:

π 2π 3π 4π 5π 6π 7π 8π 9π 10π 11π
ϕ 0 6 6 6 6 6 6 6 6 6 6 6
√ √ √ √
r(ϕ) 2 4 2(1 + 3) 6 2(1 + 3) 4 2 0 2(1 − 3) −2 2(1 − 3) 0
| {z } | {z } | {z } | {z }
≈5.46 ≈5.46 ≈−1.46 ≈−1.46

Using this calculations we can sketch the curve:

6

-3 -2 -1 1 2 3

We see from the sketch that the smaller bounded region enclosed by r(ϕ) is obtained by
restricting ϕ to the interval [ 7π6
, 11π
6
]. Its area A is
Z 11π Z 11π
1 6
2 1 6
A= r(ϕ) dϕ = (2 + 4 sin ϕ)2 dϕ
2 7π6 2 7π6
Z 11π
1 6
= (4 + 16 sin ϕ + 16 sin2 ϕ) dϕ
2 7π6
Z 11π
11π 11π 1 6
= 2 [ϕ] 7π + 8 [− cos ϕ] 7π +
6 6
(8 − 8 cos(2ϕ)) dϕ
6 6 2 7π6
4π √ 11π 11π 12π √
= − 8 3 + 4 [ϕ] 7π6 − 2 [sin(2ϕ)] 7π6 = − 8 3,
3 6 6 3
2 1
where we used that sin ϕ = 2 (1 − cos(2ϕ)) in the fourth equality.

Differential equations

Solution of the task 30.

a. An equivalent form of the DE (2) is

(2xy − 9x2 ) dx + (2y + x2 + 1) dy = 0.
| {z } | {z }
M (x,y) N (x,y)

Since M (x, y) and N (x, y) are differentiable functions for every (x, y) ∈ R2 , the
DE (2) is exact if and only if
∂M ∂N
(x, y) = (x, y).
∂y ∂x
Differential equations 55

We have
∂M ∂N
(x, y) = 2x = (x, y)
∂y ∂x
and hence the DE (2) is exact.
b. Solutions of the exact DE are level curves u(x, y) = C, C ∈ R, of the potential
function u(x, y), i.e., a function which satisfies
∂u ∂u
(x, y) = M (x, y) and (x, y) = N (x, y).
∂x ∂y
Hence,
Z
u(x, y) = M (x, y)dx = x2 y − 3x3 + C(y),
Z
u(x, y) = N (x, y)dy = y 2 + x2 y + y + D(x),

where C(y) and D(x) are functions of y and x, respectively. So,

u(x, y) = x2 y − 3x3 + y 2 + y + E,
where E ∈ R is a constant. The level curves of u(x, y) are given by the equations
(12) x2 y − 3x3 + y 2 + y = C, C ∈ R.
The solution satisfying y(0) = −3 is the level curve (12)
02 · (−3) − 3 · 03 + (−3)2 + (−3) = 6 = C
with C = 6.
Solution of the task 31.
We introduce the functions x1 (t) := y(t) and x2 (t) := x01 (t). The DE (3) converts into
the system
x01 (t) = x2 (t),
(13)
x02 (t) = −24x1 (t) − 11x2 (t).
The matricial form of (13) is
" # " #" #
x01 (t) 0 1 x1 (t)
= .
x02 (t) −24 −11 x2 (t)
| {z }
A

The eigenvalues of the matrix A are the roots of the following determinant
" #!
λ 0
det A − = (−λ)(−11 − λ) + 24 = λ2 + 11λ + 24 = (λ + 8)(λ + 3).
0 λ
So, λ1 = −8 and λ2 = −3.
The kernel of " # " #
−8 0 8 1
A− =
0 −8 −24 −3
contains the vector " #
1
u1 = .
−8
56

The kernel of " # " #

−3 0 3 1
A− =
0 −3 −24 −8
contains the vector " #
1
u2 = .
−3

So, the general solution of the system (13) is

" # " # " #
x1 (t) 1 1
= C1 · e−8t · + C2 · e−3t · ,
x2 (t) −8 −3
where C1 and C2 are constants. Hence, the solution of the initial DE is
y(t) = Ce−8t + De−3t ,
where C, D are constants.
Solution of the task 32.
Multiplying the DE with y 2 we get
(14) 3y 2 y 0 cos x + y 3 sin x − 1 = 0.
Introducing a new variable z = y 3 , (14) becomes
(15) z 0 cos x + z sin x − 1 = 0.
The homogeneous part
z 0 cos x + z sin x = 0
can be solved by separation of variables. We get
dz
= − tan x dx
z
and hence
Z Z
du
log |z| = − tan x dx = = log |u| + log K = log(K cos x),
u
where we used the substitution u = cos x in the third equality and K is a constant. So
the solution of the homogeneous part of (15) is
zh (x) = K cos x.

To find one particular solution we use variation of constants, i.e.,

zp (x) = K(x) cos x.
Plugging yp (x) into (15) we get
(16) K 0 (x)(cos x)2 = 1.
(16) is a separable DE:
1
1 dK = dx ⇒ K = tan x.
(cos x)2
Hence,
zp (x) = tan x · cos x = sin x,
and the general solution of (14) is
1 1
y(x) = K(cos x) 3 + (sin x) 3 .
Differential equations 57

Using that y(0) = 1 we get K = 1.

Solution of the task 33.

The matrix form of the system is
" # " #" #
ẋ 2 −3 x
= .
ẏ 1 −2 y
| {z }
A

The eigenvalues of the matrix A are the roots of the following determinant
" #!
λ 0
det A − = (2 − λ)(−2 − λ) + 3 = λ2 − 4 = (λ − 1)(λ + 1).
0 λ

So, λ1 = 1 and λ2 = −1.

The kernel of
" # " #
1 0 1 −3
A− =
0 1 1 −3

contains the vector

" #
3
u1 = .
1

The kernel of
" # " #
−1 0 3 −3
A− =
0 −1 1 −1

contains the vector

" #
1
u2 = .
1

So, the general solution of the system is

" # " # " #
x 3 1
= C1 · et · + C2 · e−t · ,
y 1 1

where C1 and C2 are constants.

The phase portait is the following:

58

-1

-2

-3

-3 -2 -1 0 1 2 3

Solution of the task 34.

First we solve the homogeneous part of the DE. The characteristic polynomial is λ2 −4λ+5
with zeroes √
4 ± 16 − 20
λ1,2 = = 2 ± i.
2
So, two linearly independent solutions to the homogeneous part are
y1 (t) = e(2+i)t = e2t eit and y2 (t) = e(2−i)t = e2t e−it .
Two real linearly independent solutions are
yR,1 (t) = e2t cos t and yR,2 (t) = e2t sin t.
A general solution of the homogeneous part is
yh (t) = AyR,1 (t) + ByR,2 (t), where A, B ∈ R.
To find one particular solution of the DE we can use the form
yp (t) = C cos t + D sin t,
where C, D are some constants. Plugging this form into the initial DE we obtain
(−C cos t − D sin t) − 4 (−C sin t + D cos t) + 5(C cos t + D sin t) = 8 cos x.
Comparing the coefficients at cos t and sin t on both sides we obtain the system
4C − 4D = 8 and 4D + 4C = 0,
with the solution C = −D = 1. So a general solution to the DE is
y(t) = yh (t) + yp (t) = e2t (A cos t + B sin t) + cos t − sin t.
The solution with a local extremum in the point (0, 2) is determined by the conditions
y(0) = 2 and y 0 (0) = 0.
We get
2 = y(0) = A + 1,
0
0 = y (0) = 2e2t (A cos t + B sin t) + e2t (−A sin t + B cos t) − (sin t + cos t) (0)

= 2A + B − 1.
So A = 1 and B = −1.
Differential equations 59

Solution of the task 35.

The system in the matricial form is the following:
" # " #" # " # " #
ẋ 2 −1 x 0 x
= + =: A + f (t).
ẏ −2 1 y 18t y
First we will solve the homogeneous part of the system.
" #
2 − λ −1
det(A − λI) = det = (2 − λ)(1 − λ) − 2 = λ(λ − 3).
−2 1 − λ
Hence, det(A − λI) = 0 for λ1 = 0 and λ2 = 3. Further on,
" # " # ( " # )
2 −1 2 −1 1
ker A = ker = = α :α∈R ,
−2 1 0 0 2
" # " # ( " # )
−1 −1 −1 −1 1
ker(A − 3I) = ker = = α :α∈R .
−2 −2 0 0 −1
Thus the solution (xh (t), yh (t)) of the homogeneous part is the following:
" # " # " #
xh (t) 1 1
=α + βe3t ,
yh (t)) 2 −1
where α, β ∈ R are constants.
Now we will find one particular solution (xp (t), yp (t)) of the system using the hint.
Plugging the form of the particular solution from the hint into our system we obtain:
2At + B = 2(At2 + Bt + C) − (Dt2 + Et + F )
= (2A − D)t2 + (2B − E)t + (2C − F ),
(17)
2Dt + E = −2(At2 + Bt + C) + (Dt2 + Et + F ) + 18t
= (−2A + D)t2 + (−2B + E + 18)t + (−2C + F ).
By comparing the coefficients at 1, t, t2 in (17) we get the system:
2A − D = 0, 2A = 2B − E, B = 2C − F, 2D = −2B + E + 18, E = 2C − F,
with a one parametric family of solutions
(A, B, C, D, E, F ) = (3, 2, C, 6, −2, −2 + 2C),
where C ∈ R is a constant. Choosing C = 0 we get
(xp (t), yp (t)) = (3t2 + 2t, 6t2 − 2t − 2).
Finally, a general solution is
" # " # " # " #
2
x(t) 1 1 3t + 2t
(18) =α + βe3t + .
y(t)) 2 −1 6t2 − 2t − 2
We plug x(0) = 1 and y(0) = 0 in (18) and obtain α = 1, β = 0.
Solution of the task 36.
60

a. The exact solution can be obtained using separation of variables:

dy dy 1 1
= 2xy 2 ⇒ = 2x dx ⇒ − = x2 + C ⇒ y=− .
dx y2 y x2 +C
The solution passing through the point (0, 1) is obtained by
1 1
y(0) = 1 = − ⇒ C = −1 ⇒ y = .
C 1 − x2
b. Using Euler’s method to estimate y(0.4) we get:
y1 = y(0.2) = 1 + 0.2 · (0 · 12 ) = 1,
y2 = y(0.4) = 1 + 0.2 · (2 · 0.2 · 12 ) = 1.08.
The exact solution is
1
y(0.4) = ≈ 1.2.
1 − 0.16
Solution of the task 37.
First we solve the homogenenous part
ẍ + ẋ − 2x = 0.
The characterstic polynomial is p(λ) := λ2 + λ − 2 and hence
p(λ) = 0 ⇔ (λ − 2)(λ + 1) = 0.
So, the solution of the homogeneous part is
xh (t) = Ce2t + De−t ,
where C, D ∈ R are constants. To find one particular solution we use the form
xp = At2 + Bt + C.
Hence, ẋp = 2At + B and ẍp = 2A. Plugging this into the DE we obtain
2A + (2At + B) − 2 At2 + Bt + C = t2 .


(19)
Comparing the coefficients at 1, t, t2 on both sides of (19) we get the system
2A + B − 2C = 0, 2A − 2B = 0, −2A = 1,
with the solution
1 1 3
A=− , B=− , C=− .
2 2 4
Hence, the general solution of the DE is
1 1 3
x(t) = Ce2t + De−t − t2 − t − .
2 2 4
Solution of the task 38.
The DE can be solved by separation of variables:
dy
= 2x dx ⇒ arctan y = x2 + C.
1 + y2
The particular solution, which goes through the point (1, 0), is
arctan y(1) = arctan(0) = 0 = 12 + C ⇒ C = −1 ⇒ arctan y = x2 − 1.
Solution of the task 39.
Differential equations 61

a. Stationary points satisfy ẋ = ẏ = 0. Hence,

ẋ = xy + 1 = 0 and ẏ = x + xy = 0 ⇔ xy = −1 and x = 1
⇔ y = −1 and x = 1.

b. To classify the stationary point (1, −1) we have to linearize the system. We denote
the right side of the system by

f (x, y) = (xy + 1, x + xy).

Hence,

" # " #
ẋ x−1
≈ ((Jf )(1, −1)) ·
ẏ y+1
" # ! " #
y x x−1
= (1, −1) ·
1+y x y+1
" #" #
−1 1 x − 1
=
0 1
y+1
" #
−x + y + 2
= .
y+1

The behaviour of the system depends on the eigenvalues of (Jf )(1, −1):

" #
−1 − λ 1
det ((Jf )(1, −1) − λI2 ) = = (−1 − λ)(1 − λ).
0 1−λ

Hence, the eigenvalues are λ1 = 1 and λ2 = −1. The solution of the system is of
the form
" #
t −t
−1
C · e · v1 + D · e · v2 + ,
1

where v1 , v2 are the eigenvectors of (Jf )(1, −1) for the eigenvalues 1, −1 and C, D
are constants. The point (1, −1) is a saddle.

c. The sketch of the phase portrait is the following:

62

-1

-2

-3

-3 -2 -1 0 1 2 3

Solution of the task 40.

First we solve the homogeneous part of the DE:
dy dx
xy 0 = y ⇒ = ⇒ ln |y| = ln |x| + k ⇒ yh (x) = Kx,
y x
where k, K ∈ R are constants. Now we have to determine one particular solution. By the
form of the DE we can try with the form
yp (x) = Ax3 + Bx2 + Cx + D,
where A, B, C, D ∈ R are constants. Hence,
yp0 (x) = 3Ax2 + 2Bx + C
and plugging into the DE we get
(20) x(3Ax2 + 2Bx + C) = Ax3 + Bx2 + Cx + D + 2x3 .
Comparing the coefficients at x3 , x2 , x, 1 on both sides of (20) we obtain the system
3A = A + 2, 2B = B, C = C, 0 = D.
Hence,
A = 1, B = D = 0 and C ∈ R is arbitrary.
We choose C = 0 and get
yp (x) = x3 .
The general solution of the DE is
y(x) = yh (x) + yp (x) = Kx + x3 .
The solution which passes through the point (2, 3) is
y(2) = 3 = 2K + 27 ⇒ K = −12 ⇒ y(x) = −12x + x3 .
Solution of the task 41.
First we solve the homogeneous part of the DE:
y 00 + y 0 − 6y = 0.
The characteristic polynomial is
p(λ) = λ2 + λ − 6 = (λ + 3)(λ − 2)
Differential equations 63

with zeroes λ1 = −3, λ2 = 2. Hence, the solution of the homogeneous part is

yh (x) = Ce−3x + De2x ,
where C, D ∈ R are constants.
To obtain a particular solution we can try with the form
(21) yp (x) = ax2 + bx + c ⇒ yp0 (x) = 2ax + b ⇒ yp00 (x) = 2a.
Plugging (21) into the DE we obtain
(22) 2a + (2ax + b) − 6(ax2 + bx + c) = 36x.
Comparing the coefficients at x2 , x, 1 on both sides of (22) we obtain a system
−6a = 0, −6b + 2a = 36, 2a + b − 6c = 0,
with the solution
a = 0, b = −6, c = −1.
Hence, the general solution of the DE is
y(x) = yh (x) + yp (x) = Ce−3x + De2x − 6x − 1.
The one satisfying the initial conditions
y(0) = C + D − 1 = 1
y 0 (0) = −3C + 2D − 6 = 1,
is the one with
3 13
C=− , D= .
5 5
So the final solution is
3 13
y(x) = − e−3x + e2x − 6x − 1.
5 5
Solution of the task 42.
First we solve the homogeneous part of the DE:
dy dx K
x2 y 0 + xy = 0 ⇒ − = ⇒ − ln |y| = ln |x| + k ⇒ yh (x) = ,
y x x
where k, K ∈ R are constants. Now we have to determine one particular solution. By
variation of constants the form of the particular solution is
K(x)
yp (x) = ,
x
where K(x) is a function of x. Thus,
K 0 (x)x − K(x)
(23) yp0 (x) =
x2
and plugging (23) into the initial DE we get
K 0 (x)x − K(x) K(x)
x2 · 2
+x + 3 = 0.
x x
Equivalently,
(24) K 0 (x)x + 3 = 0.
We solve the DE (24) by separation of variables:
dK dx 1 1
− = ⇒ − K = ln |x| ⇒ K = ln .
3 x 3 |x|3
64

Since in the initial conditin x > 0, we have K = ln x13 and yp (x) = ln x13 · x1 . So, the general
solution of the DE is
 
1 1
y(x) = yh (x) + yp (x) = K + ln 3 .
x x
The solution which passes through the point (1, 1) is
 
1 1
y(1) = 1 = K + ln 1 ⇒ K=1 ⇒ y(x) = 1 + ln 3 .
x x
Solution of the task 43.
The matricial form of the system is the following:
" # " #" #
x0 (t) −2 5 x(t)
= .
y 0 (t) 1 2 y(t)
| {z }
A

We compute the eigenvalues of A:

" #
−2 − λ 5
det (A − λI2 ) = det = (−2 − λ)(2 − λ) − 5 = λ2 − 9 = (λ − 3)(λ + 3).
1 2−λ

Thus the eigenvalues are λ1 = 3 and λ2 = −3. The kernel of

" # " #
3 0 −5 5
A− =
0 3 1 −1

contains the vector

" #
1
u1 = .
1

The kernel of
" # " #
−3 0 1 5
A− =
0 −3 1 5
contains the vector
" #
−5
u2 = .
1

So, the general solution of the system is

" # " # " #
x 1 −5
= C1 · e3t · + C2 · e−3t · ,
y 1 1

where C1 and C2 are constants. The solution, which satisfies x(0) = y(0) = 1, is:
" # " # " #
1 −5 1
C1 + C2 = ⇒ C1 − 5C2 = 1, C1 + C2 = 1
1 1 1
⇒ C1 = 1, C2 = 0.
Differential equations 65

Solution of the task 44.

The DE is of the form
2xy dx + (x2 + 3y 2 ) dy = 0.
It is indeed exact, since
d(2xy) d(x2 + 3y 2 )
= = 2x.
dy dx
We have that
Z
2xy dx = x2 y + C(y),
Z
(x2 + 3y 2 ) dx = x2 y + y 3 + D(x),

where C(y) and D(x) are functions of y and x. Hence, the solution of the DE is a family
of functions
u(x, y, K) = x2 y + y 3 + K,
where K ∈ R is a constant.
Solution of the task 45.
First we solve the homogeneous part of the DE:
y 00 + 9y = 0.
The characteristic polynomial is
p(λ) = λ2 + 9 = (λ − 3i)(λ + 3i)
with zeroes λ1 = 3i, λ2 = −3i. Hence, the solution of the homogeneous part is
yh (x) = Ce3ix + De−3ix ,
where C, D ∈ C are constants. Another way of expressing all solutions of the DE is
yh (x) = C cos (3x) + D sin (3x),
where C, D ∈ C are constants.
To obtain a particular solution we can try with the form
(25) yp (x) = ax2 + bx + c ⇒ yp0 (x) = 2ax + b ⇒ yp00 (x) = 2a.
Plugging (25) into the DE we obtain
(26) 2a + 9(ax2 + bx + c) = 2x2 − 1.
Comparing the coefficients at x2 , x, 1 on both sides of (26) we obtain a system
9a = 2, 9b = 0, 2a + 9c = −1,
with the solution
2 13
a = , b = 0, c=− .
9 81
Hence, the general solution of the DE is
2 13
y(x) = yh (x) + yp (x) = C cos (3x) + D sin (3x) + x2 − .
9 81
The one satisfying the initial conditions
13
y(0) = C − = 1,
81
y 0 (0) = 3D = 1,
66

is the one with

94 1
C= , D= .
81 3
So the final solution is
94 1 2 13
y(x) = cos (3x) + sin (3x) + x2 − .
81 3 9 81
Solution of the task 46.

a. The stationary points satisfy ẋ = ẏ = 0. Hence:

x(3 − x − 2y) = 0 ⇒ x = 0 or x = 3 − 2y,
y(4 − 3x − y) = 0 ⇒ y = 0 or y = 4 − 3x.
If x 6= 0 and y 6= 0, then y = 4 − 3(3 − 2y) and thus, y = 1 and x = 1. So the
stationary points are:
(0, 0), (0, 4), (3, 0), (1, 1).
b. The right side of the system is
f1 (x, y) = 3x − x2 − 2xy,
f2 (x, y) = 4y − 3xy − y 2 .
The linearization of the system around (1, 1) is
" # " # " # " ∂f (1,1) ∂f (1,1) # " #
ẋ f1 (x, y) f1 (1, 1) 1 1
x − 1
= ≈ + ∂f2∂x(1,1)
∂y
∂f2 (1,1)
ẏ f2 (x, y) f2 (1, 1) ∂x ∂y
y−1
" #" #
(3 − 2x − 2y)(1, 1) (−2x)(1, 1) x−1
=
(−3y)(1, 1) (4 − 3x − 2y)(1, 1) y−1
" #" #
−1 −2 x−1
= .
−3 −1 y−1
c. Introducing the new variables X = x − 1 and Y = y − 1, one obtains the following
autonomous linear system:
" # " #" #
Ẋ −1 −2 X
=
Ẏ −3 −1 Y
| {z }
A
We have to compute the eigenvalues of A:
" #
−1 − λ −2
det(A − λI2 ) = det = (−1 − λ)2 − 6
−3 −1 − λ
√ √
= (λ + 1 − 6)(λ + 1 + 6).
√ √
Hence, the eigenvalues are λ1 = 6 − 1 and λ2 = − 6 − 1. The corresponding
eigenspaces are:
" √ # " #
− 6 −2 n 1 o
ker(A − λ1 I2 ) = ker √ = Lin √
−3 − 6 − 26
| {z }
v1
Differential equations 67

and " √ # " #

6 −2 n 1 o
ker(A − λ2 I2 ) = ker √ = Lin √
6
.
−3 6 2
| {z }
v2

The solutions of the system are

C1 eλ1 t v1 + C2 eλ2 t v2 ,
where C1 , C2 ∈ R. Hence, the solutions of the original system are
" #
1
+ C1 eλ1 t v1 + C2 eλ2 t v2 ,
1
where C1 , C2 ∈ R. The sketch of the phase portrait of the original system is the
following:

Solution of the task 47.

a. A general solution of the system is

     
1 1 1
C1 e−t 
  
t
 
2t 

 −1  + C2 e  1  + C3 e  2  ,
  
1 1 4
where C1 , C2 , C3 are constants.
b. The matrix A has eigenpairs (−1, v1 ), (1, v2 ), (2, v3 ), where
     
1 1 1
     
 −1  , v2 =  1  , v3 =  2  .
v1 =      
1 1 4
68

Hence, h i
A= v1 v2 v3 · diag(−1, 1, −2) · P −1 .
| {z }
P
−1
Let us compute P using Guassian elimination:
   
1 1 1 1 0 0 1 1 1 1 0 0
   
 −1 1 2 0 1 0  ∼  0 2 3 1 1 0 
  |{z}  
`2 =`2 +`1
1 1 4 0 0 1 `3 =`3 −`1 0 0 3 −1 0 1
| 
{z  }
P I3
 
1 1 0 43 0 − 13
 
∼
|{z}
 0 2 0 2 1 −1 
 
`1 =`1 − 13 `3
`2 =`2 −`3
0 0 3 −1 0 1
 
1 0 0 13 − 12 61
 
1 1 
∼ 0 1 0 1 − .

|{z}   2 2 

`1 =`1 − 12 `2
`2 = 21 `2
0 0 1 − 13 0 1
3
`3 = 13 `3
| 
{z 
}
−1
I3 P
So
   1 1 1

1 1 1 −1 0 0  3 − 2 6
   
1 1 
A =  −1 1 2   0 1 0   1
   
2
−2 
 
1 1 4 0 0 2 − 13 0 1
3
  1 1 1

−1 1 2  3 − 2 6 
  1
= 1 1 4  1 − 12 

  2

−1 1 8 − 13 0 1
3
 
0 1 0
 
=  0 0 1 .
 
 
−3 0 2
c. Since the eigenvalues of the matrix A are −1, 1, 2, the corresponding third order
polynomial is
(λ + 1)(λ − 1)(λ − 2) = (λ2 − 1)(λ − 2) = λ3 − 2λ2 − λ + 2
and hence the differential equation with constants coefficients is
x(3) = 2x(2) + x(2) − 2x.
Solution of the task 48.
First we solve the homogenenous part
ẍ − ẋ − 4x = 0.
Differential equations 69

The characterstic polynomial is p(λ) := λ2 − λ − 4 and hence

√ √
1 ± 1 + 16 1 ± 17
p(λ) = 0 ⇔ λ1,2 = = .
2 2
So, the solution of the homogeneous part is
√ √
1+ 17 1− 17
t t
xh (t) = Ae 2 + Be 2 ,
where A, B ∈ R are constants. To find one particular solution we use the form
xp = xp1 (t) + xp2 (t),
where
xp1 (t) = Ct + D, xp2 (t) = Eet .
Hence, ẋp1 = C, ẍp = 0 and ẋp2 = ẍp2 = Eet . Plugging this into the DE we obtain
(27) Eet − (C + Eet ) − 4(Ct + D + Eet ) = 2t + et .
Comparing the coefficients at 1, t, et on both sides of (27) we get the system
−C − 4D = 0, −4C = 2, −4E = 1,
with the solution
1 1 1
D= , C=− , E=− .
8 2 4
Hence, the general solution of the DE is
√ √
1+ 17
t 1− 17
t 1 1 1
x(t) = Ae 2 + Be 2 − t + − et .
2 8 4