Problem Sheet 4 With Solutions GRA 6035 Mathematics: BI Norwegian Business School
Problem Sheet 4 With Solutions GRA 6035 Mathematics: BI Norwegian Business School
Problems
Show that vi is an eigenvector for A for i = 1, 2, 3 and find the corresponding eigen-
values. Use this to find an invertible matrix P and a diagonal matrix D such that
A = PDP−1 .
6. Find an invertible matrix P such that D = P−1 AP is diagonal when
2 −7
A=
3 −8
8. Initially, two firms A and B (numbered 1 and 2) share the market for a certain
commodity. Firm A has 20% of the marked and B has 80%. In course of the next
year, the following changes occur:
We can represent market shares of the two firms by means of a market share vector,
defined as a column vector s whose components are all nonnegative and sum to 1.
Define the matrix T and the initial share vector s by
0.85 0.45 0.2
T= , s=
0.15 0.55 0.8
If this is the case, find an invertible matrix P such that P−1 AP is diagonal, and use
this to compute A17 .
These matrix problems are quite challenging and are meant for advanced students. It
is recommended that you work through the ordinary problems and exam problems
from Problem Sheet 1-4 and make sure that you master them before you attempt
Problem 13-15 (which are optional).
x2 + x3 + . . . + xn−1 + xn = 2
x1 + x3 + . . . + xn−1 + xn = 4
x1 + x2 + . . . + xn−1 + xn = 6
.. ..
. .
x1 + x2 + x3 + . . . + xn−1 = 2n
5
Solutions
1 We compute that
12 1 3
Av = = = 3v
21 1 3
This means that v is an eigenvector with eigenvalue λ = 3.
2 a) We solve the characteristic equation to find the eigenvalues:
2 − λ −7 2
3 −8 − λ = λ + 6λ + 5 = 0 ⇒ λ = −1, −5
For each eigenvalue, we compute the eigenvectors using an echelon form of the
coefficient matrix, and express the eigenvectors in terms of the free variables. For
λ = −1, we get eigenvectors
7
3 −7 3 −7 x y 7/3
99K ⇒ 3x − 7y = 0 ⇒ = 3 =y
3 −7 0 0 y y 1
For each eigenvalue, we compute the eigenvectors using an echelon form of the
coefficient matrix, and express the eigenvectors in terms of the free variables. For
λ = 5, we get eigenvectors
−4 4 −4 4 x y 1
99K ⇒ −4x + 4y = 0 ⇒ = =y
6 −6 0 0 y y 1
For each eigenvalue, we compute the eigenvectors using an echelon form of the
coefficient matrix, and express the eigenvectors in terms of the free variables. For
λ = 2, we get eigenvectors
000 010 x x 1
0 1 0 99K 0 0 2 ⇒ y = 0, 2z = 0 ⇒ y = 0 = x 0
002 000 z 0 0
b) We solve the characteristic equation to find the eigenvalues. Since the equation
(of degree three) is a bit difficult to solve, we first change the determinant by adding
the second row to the first row:
2 − λ 1 −1 2 − λ 2 − λ 0
0 1−λ 1 = 0 1 − λ 1
2 0 −2 − λ 2 0 −2 − λ
Then we transpose the matrix, and subtract the first row from the second row:
2 − λ 0 2 2 − λ 0 2
0 = 0 1 − λ −2 = (2 − λ )(λ 2 + λ ) = 0
= 2 − λ 1 − λ
0 1 −2 − λ 0 1 −2 − λ
We see that the eigenvalues are λ = 2, 0, −1. For each eigenvalue, we compute the
eigenvectors using an echelon form of the coefficient matrix, and express the eigen-
vectors in terms of the free variables. For this operation, we must use the matrix
before the transposition, since the operation of transposing the coefficient matrix
will not preserve the solutions of the linear system (but it will preserve the determi-
nant). For λ = 2, we get eigenvectors
7
0 0 0 2 0 −4 x 2
0 −1 1 99K 0 −1 1 ⇒ y = z, x = 2z ⇒ y = z 1
2 0 −4 0 0 0 z 1
We have AP = (Av1 |Av2 |Av3 ), and looking at the columns of AP we see that
The vectors are therefore eigenvectors, with eigenvalues λ = −5, −5, 5. The matrix
P is invertible since the three vectors v1 , v2 , v3 are linearly independent, and we have
AP = PD ⇒ A = PDP−1 when D is the diagonal matrix
−5 0 0
D = 0 −5 0
0 0 5
6 We use the eigenvalues and eigenvectors we found in Problem a). Since there are
two distinct eigenvalues, the matrix A is diagonalizable, and D = P−1 AP when we
put
7/3 1 −1 0
P= , D=
1 1 0 −5
8
Hence there is only one eigenvalue (with multiplicity 2). The corresponding eigen-
vectors are found by reducing the coefficient matrix to an echelon form. For λ = 3,
we get eigenvectors
05 x x 1
⇒ 5y = 0 ⇒ = =x
00 y 0 0
In particular, there is only one free variable and therefore not more than one linearly
independent eigenvector. This means that there are too few linearly independent
eigenvectors (only one eigenvector while n = 2), hence A is not diagonalizable.
8
a) We compute the matrix product
0.53
Ts =
0.47
and
3/4 3/4 0.2 0.75
T ns → =
1/4 1/4 0.8 0.25
The equilibrium marked shares are 75% for A and 25% for B.
9 We find the eigenvalues of A by solving the characteristic equation:
4 − λ 1 2
0 3 − λ 0 = (3 − λ )(λ 2 − 9λ + 18) = 0 ⇒ λ = 3, 3, 6
1 1 5−λ
Since there is two degrees of freedom, there are two linearly independent eigenvec-
tors v1 , v2 for λ = 3, given by
x −y − 2z −1 −2
y = y = y 1 + z 0 = yv1 + zv2
z z 0 1
We use this to compute A17 , since A = PDP−1 . We do not show the computation of
P−1 , which is straight-forward:
17
−1 −2 1 3 0 0 0 3 0
1
A17 = (PDP−1 )17 = PD17 P−1 = 1 0 0 0 317 0 −1 −1 1
3
0 1 1 0 0 617 1 1 2
10
This gives
2 −1 −2 224
A17 = 316 0 3 0 + 616 0 0 0
−1 −1 1 224
10 Final Exam in GRA6035 10/12/2010, Problem 2
Therefore, we get (
3, s 6= 0
rk(A) =
2, s = 0
b) We compute the characteristic equation of A, and find that
1 − λ 7 −2
0 s − λ 0 = (s − λ )(λ 2 − 5λ + 6) = (s − λ )(λ − 2)(λ − 3)
1 1 4−λ
of rank two, so there is only one free variable. If s = 3, we check the eigenspace
corresponding to the double root λ = 3: The coefficient matrix of the system has
echelon form
11
−2 7 −2 111
0 0 0 99K 0 9 0
1 1 1 000
of rank two, so there is only one free variable. In both cases, there are too few
linearly independent eigenvectors, and A is not diagonalizable. Hence A is diag-
onalixable if s 6= 2, 3.
11 Mock Final Exam in GRA6035 12/2010, Problem 1
Therefore, we get (
3, t 6= 3
rk(A) =
2, t = 3
b) We compute the characteristic equation of A, and find that
1 − λ 1 −4
0 t + 2 − λ t − 8 = (1 − λ )(λ 2 − (t + 7)λ + 10(t − 3)) = 0
0 −5 5 − λ
of rank two, so there is only one free variable. If t = 13, we check the eigenspace
corresponding to the double root λ = 10: The coefficient matrix of the system
has echelon form
12
−9 1 −4 −9 1 −4
0 5 5 99K 0 5 5
0 −5 −5 0 0 0
of rank two, so there is only one free variable. In both cases, there are too few
linearly independent eigenvectors, and A is not diagonalizable. Hence A is diag-
onalizable if t 6= 4, 13.
12 Final Exam in GRA6035 30/05/2011, Problem 2