Determinant - Diagonalization

Determinant and Diagonalization
Determinant and Diagonalization 1 / 76

Goals
Study about determinant of a matrix and diagonalization a square matrix

with application to linear recurrences
1 Compute determinant of a matrix
2 Properties of determinant
3 Use determinant to find inverse of a matrix
4 Use determinant to solve system of linear equations: Cramer’s rule
5 Diagonalize a square matrix
6 Application to linear recurrences

Determinant
Table of contents
1 Determinant
Cofactor Expansion
Elementary operations
Adjugates and Inverse Matrix
Cramer’s rule
2 Diagonalization
Eigenvalues and Eigenvectors
Diagonalization
3 An Application to Linear Recurrences
4 Application in Linear Dynamical System

Determinant Cofactor Expansion
Determinant of a 2 × 2 matrix
Determinant of a square matrix is denoted by det A or det(A) or |A|.

For a 2 × 2 matrix
" #!
a b
det = ad − bc
c d
or
a b
= ad − bc
c d

Example
Let " #
1 2
A=
3 5
then det(A) = 1 ∗ 5 − 3 ∗ 5 = −10

In order to define the determinant of 3 × 3, 4 × 4 . . . matrix with inductive

method, we need some terminology
Definition (Co-factor)
Let A be a n × n matrix and
(i, j)− minor Aij : matrix obtained from A by deleting row i and column j
Then the (i, j) - cofactor cij (A) is the scale defined by
cij (A) = (−1)i+j det(Aij )
(−1)i+j is called the sign of (i, j) - position

Matrix for sign of position
 
+ − + − ...
− + − + . . .
 
+ − + − . . .
 
− −
 
+ + . . .
.. .. .. .. ..
 
. . . . .
The signs alternate along each row and column with + in the upper left
corner

Example
Find the cofactors of positions (1, 2), (3, 1), and (2, 3) in the following
matrix.  
3 −1 6
A = 5 2 7
 
8 9 4

Example
Find the cofactors of positions (1, 2), (3, 1), and (2, 3) in the following
matrix.  
3 −1 6
A = 5 2 7
 
8 9 4
Solution
In order to compute c12 (A), first we delete row 1 and column 2 to get
" #
5 7
A12 =
8 4
The sign of position (1, 2) is (−1)1+2 . So the (1, 2) - cofactor
5 7
c12 (A) = (−1)1+2 = −(5 ∗ 4 − 7 ∗ 8) = 36
8 4

Determinant by Cofactor Expansion
Cofactor expansion along row i
ai1 ci1 (A) + ai2 ci2 )(A) + · · · + ain cin (A)
Cofactor expansion along column j
a1j c1j (A) + a2j c2j )(A) + · · · + anj cnj (A)
The cofactor expansion along any row or column of a matrix A

always gives the same result. This result is called the determinant of A
and denoted by det(A) or det A or |A|

Example
Compute the determinant of
 
3 4 5
A = 1 7 2 
 
9 8 −6
Solution
The cofactor expansion along the first row

Compute det A by expanding along the first column

Example
Compute determinant of
 
3 0 0 0
 5 1 2 0
A=
 
 2 6 0 −1

−6 3 1 0
Solution
The expansion involves multiplying entries by cofactors, so the work is
minimized when the row or column contains as many zero entries as
possible. Row 1 is a best choice in this matrix. The expansion is

Solution (cont)
Use expansion along the first row to compute the determinant of 3 × 3
matrix

Solution (cont)
Use expansion along the first row to compute the determinant of 3 × 3
matrix
It’d better to choose the column 3

Determinant of triangular matrices
Determinant of a triangular matrix is the product of the entries on

the main diagonal.
determinant = a11 a22 . . . ann

Theorem
" # " #
A X A 0
Consider matrices and where A and B are square
0 B Y B
matrices. Then
" # " #
A X A 0
det = det(A) det(B) and det = det(A) det(B)
0 B Y B
Proof.
Use induction on size k of matrix A and cofactor expansion along the first
column for upper triangular case and along last column for lower triangular
case.

Example
 
2 3 1 3 " # " #
1 −2 −1 1 2 3 1 1
det   = det det = (1)(−3) = −3
 
0 0 1 1 1 2 4 1
0 0 4 1

Geometric Meaning of Determinant
Area of parallelogram defined by two vectors is
Figure 1: Parallelogram Area of parallelogram =rectangle area − area of A − · · ·

defined by the two =(x1 + x2 )(y1 + y2 ) = x2 y1 − x1 y1 /2 = x1 y2 − x2 y1
vectors
x1 x2
=
y1 y2
determinant of the matrix’s columns (vectors)

Geometric Meaning of Determinant (cont)
(Oriented) volume of the parallelepiped P defined three vector v1 , v2 , v3

h i
vol(P ) = det( v1 v2 v3 )

Orientation - swapping changes the sign of the determinant
On the left, →−
u is the first column in the matrix and we get a positive size.
On the right, →−v is first and we get a negative size. On the left, the arc
from first vector to second is counterclockwise while on the right, the arc
from first to second is clockwise. The sign returned by the determinant
reflects the orientation or sense of the box.
Scaling effect
If we scale a column by a factor k then the size of the box scales by that
factor

Determinant is Unaffected by Combining Columns
The box formed by two vectors →

−v and k →
−
v +→
−
w is slanted at a different
→
− →
−
angle than the one form by v and w but two boxes have the same base
and height, and hence the same area

Determinant Elementary operations
Cut down the work to compute determinant
If A is a 4 × 4 matrix, the cofactor expansion along any row or

column involves calculating four cofactors, each of which involves the
determinant of a 3 × 3 matrix. And if A is 5 × 5, the expansion
involves five determinants of 4 × 4 matrices


involves five determinants of 4 × 4 matrices =⇒ tedious work


calculating a determinant is simplified a great deal when a row or
column consists mostly of zeros


calculating a determinant is simplified a great deal when a row or
column consists mostly of zeros
one method of creating zeros in a matrix is to apply elementary row
operations to it

Properties of Determinant vs Elementary Operations

Let A be a square matrix
1 If A has a row or column of zeros, det A = 0.
expand along the this row/column

2 If two distinct rows (or columns) of A are interchanged, the
determinant of the resulting matrix is − det A.

Induction
3 If a row (or column) of A is multiplied by a constant u, the
determinant of the resulting matrix is u det A.

expand along this row
4 If two distinct rows (or columns) of A are identical, det A = 0.
use result in part 2

5 If a multiple of one row of A is added to a different row (or if a
multiple of a column is added to a different column), the determinant

of the resulting matrix is det A.
expand along this row (or column) and use the result in part 2

Example

Example
Evaluate the determinant of
 
1 −1 3
A = 1 0 −1
 
2 1 6
Solution
The matrix does have zero entries, so expansion along (say) the second
row would involve somewhat less work. However, a column operation can
be used to get a zero in position (2, 3) - namely, add column 1 to column
3. Because this does not change the value of the determinant, we obtain
1 −1 4
−1 4
det A = 1 0 0 = − = 12
1 8
2 1 8

Example
a b c
If p q r = 6 , evaluate det A where
x y z
 
a+x b+y c+z
A =  3x 3y 3z 
 
−p −q −r
Solution
Take common factors out of rows 2 and 3
a+x b+y c+z

det A = 3(−1) x y z
p q r

Solution (cont)
Subtract the second row from the first
a b c
det A = −3 x y z
p q r
Interchange the last two rows.
a b c
det A = 3 p q r = 3 ∗ 6 = 18
x y z

Example
Let  
1 x x
A = x 1 x
 
x x 1
Find the values of x for which det A = 0
Solution
1 x x 1 x x
r2 −xr1 r3 −xr1
det A = 0 1 − x x − x2
2 = 0 1 − x x − x2
2
x x 1 0 x − x2 1 − x2
1 − x2 x − x2
=
x − x2 1 − x2

Solution (cont)
Taking a common factor of (1 − x) from each row
(1 − x)(1 + x) x(1 − x) 1+x x

det A = = (1 − x)2
x(1 − x) (1 − x)(1 + x) x 1+x
= (1 − x)2 (2x + 1)
Hence det A = 0 means (1 − x)2 (2x + 1) = 0,that is x = 1 or x = − 12 .

Exercise
Show that
 
1 a1 a21
det 1 a2 a22  = (a1 − a2 )(a2 − a3 )(a3 − a1 )
 
1 a3 a23

Properties of Determinant
1 If A is a n × n matrix then det(uA) = un A

Take the common factor u out of each row and use property 3 in
element row operations
2 det(AB) = det(A) det(B)
3 det(A1 . . . Ak ) = det(A1 ) . . . det(Ak )
4 det(Ak ) = (det(A))k
5 If A is invertible then det(A−1 ) = det(A) 1

Inversely if det(A) ̸= 0 then A is invertible

In fact, A can be carried to reduced row-echelon matrix R by sequence of
elementary matrices E1 , . . . , Ek . That is
R = Ek . . . E1 A.
Since det(E) ̸= 0 for all elementary matrix E and det(A) ̸= 0, we have
det(R) = det(E1 ) . . . det(Ek ) det(A) ̸= 0
It implies that the square matrix R of reduced row-echelon form has no

zero row and then R must be identity matrix. Hence A is invertible

Example
 
1 0 −c
For which values of c does A = −1 3 1  have an inverse?
 
0 2c −4

Example
 
1 0 −c
For which values of c does A = −1 3 1  have an inverse?
 
0 2c −4
Solution
Compute det(A) by first adding c times column 1 to column 3 and then
expanding along row 1.
   
1 0 c 1 0 0
det(A) = det −1 3 1  = det −1 3 1 − c
   
0 2c −4 0 2c −4
" #
3 1−c
= det = 2(c + 2)(c − 3)
2c −4
A is invertible if and only if det(A) ̸= 0. Hence c ̸= −2 and c ̸= 3

Exercise
Show that if a product A1 . . . Ak is invertible then each matrix Ai is

invertible.

Determinant of transpose
det(AT ) = det(A)
We can prove by induction with respect to the size of A

If n = 2 then
" # " #
a a a a
A = 11 12 and A = 11 21
T
a21 a22 a12 a22
thus det(A) = det(AT ) = a11 a22 − a12 a21

Suppose tht it holds for any n × n matrix A
We prove that it holds for any n + 1 × n + 1 matrix A. In fact,
expand along the first row of A and first column of AT with remark
that Aij = ATji are n × n matrix thus cij (A) = cji (AT ), we obtain
det(A) = det(AT )
Example
If det(A) = 2 and det(B) = 5, calculate det(A3 B −1 AT B 2 ).
Solution
det(A3 B −1 AT B 2 ) =
|{z} det(A3 ) det(B −1 ) det(AT ) det(B 2 )
determinant of product
=
|{z} (det(A))3 det(B −1 ) det(AT )(det(B))2
determinant of power
=
|{z} (det(A))3 det(B −1 )det(A)(det(B))2
determinant of transpose
1
= (det(A))4 (det(B))2
|{z} det(B)
determinant of inverse
1
= 24 ∗ ∗ 52 = . . .
5

Example
A square matrix is called orthogonal if A−1 = AT . What are the possible

values of det(A) if A is orthogonal?

Example
A square matrix is called orthogonal if A−1 = AT . What are the possible

values of det(A) if A is orthogonal?
Solution
If A is orthogonal then AAT = I. Take determinants to obtain
det(I) = det(AAT )
So
1 = det(A) det(AT ) = det(A) det(A) = [det(A)]2
Hence det(A) = 1 or det(A) = −1.

Determinant Adjugates and Inverse Matrix
Definition (Adjugate matrix)

The adjugate of A, denoted adj(A), is the transpose of the cofactor matrix
adj(A) = [cij (A)]T
Theorem (Adjugate Formula)

If A is a square matrix then
A(adj(A)) = (adj(A))A = det(A)I
In particular if det(A) ̸= 0 then the inverse of A is given by

1
A−1 = adj(A)
det A

Example
Compute the adjugate of

 
1 3 −2
A= 0 1 5
 
−2 −6 7
and A−1

Solution
Compute the cofactor matrix

The adjugate of A is the transpose of this cofactor matrix

The determinant of A
1 3 −2
r3 +2r1
det(A) = 0 1 5 =3
0 0 3
The inverse of A
   
37 −9 17 111 −27 51
1
A−1 = adj(A) = 3 −10 3 −5 = −30 9 −15
   
det(A)
2 0 1 6 0 3

Example
 
2 1 3
Find the (2, 3) - entry of A−1 if A = 5 −7 1 
 
3 0 −6
Solution
2 1 3 2 1 7
c3 +2c1 1 7
det(A) = 5 −7 1 = 5 −7 11 = 3 = 180
−7 11
3 0 −6 3 0 0
Since A−1 = det(A)
1
adj(A) = 1801
[cij (A)]T , the (2, 3) - entry of A−1 is the
1
(3, 2)- entry of 180 [cij (A)] and equals to
!
1 1 2 3 13
c32 (A) = − =
180 180 5 1 180

Example
Let A be an n × n matrix, n ≥ 2. Show that det(adj(A)) = (det(A))n−1

Example
Let A be an n × n matrix, n ≥ 2. Show that det(adj(A)) = (det(A))n−1
Solution
Denote d = det(A). From adjugate formula, we have
det(Aadj(A)) = det(dI) = dn det(I) = dn
Moreover
det(Aadj(A)) = det(A) det(adj(A)) = d × det(adj(A))
So
d × det(adj(A)) = dn
If d ̸= 0 then we have det(adj(A)) = (det(A))n−1 .
Inversely if d = 0 then from adjugat formula, Aadj(A) = 0. If
det(adj(A)) ̸= 0 then adj(A) is invertible and A = 0(adj(A)−1 ) = 0. It
implies that adj(A) = 0 which is impossible. Hence det(adj(A)) = 0 = d
Determinant Cramer’s rule
Come back to linear system with unique solution

Suppose that
Ax = b
is a system of linear equations of n variables x1 , . . . , xn with
A is a n × n matrix 
x1 b1
 x2   b2 
   
x=  ..  and b =  .. 
  
 .  .
xn bn
If det(A) ̸= 0 then the system solution has unique solution
1
x = A−1 b = (adj(A))b
det(A)
    
x1 c11 (A) c21 (A) . . . cnn (A) b1
 x2  c12 (A) c22 (A) . . . cn2 (A) 
  b2 
    
1 
that is 
 ..  = . .. .. ..   .
 .. .   .. 
   
 .  det(A)  . .
xn c1n c2n ... cnn bn
So we have
1
x1 = [b1 c11 (A) + b2 c22 (A) + · · · + bn cn1 (A)]
det(A)
The quantity b1 c11 (A) + b2 c22 (A) + · · · + bn cn1 (A) occurring in the
formula looks like the cofactor expansion of the determinant of a matrix.
The cofactors involved are c11 (A), c21 (A), ..., cn1 (A), corresponding to the
first column of A. If A1 is obtained from A by replacing the first column
of A by b then expanding det(A1 ) by the first column gives
det(A1 ) = b1 c11 (A) + b2 c22 (A) + · · · + bn cn1 (A) = det(A)x1
Simular result for x2 , . . .

Theorem (Cramer’s rule)

If A is invertible n × n matrix then the solution of the system
Ax = b
of n equations in n variables x1 , . . . , xn is given by
det(A1 ) det(A2 ) det(An )

x1 = , x2 = , . . . , xn =
det(A) det(A) det(A)
where, for each k, Ak is the matrix obtained from A by replacing column

k by b.

Example
Find x1 given the following system of equations
5x1 + x2 − x3 = 4
9x1 + x2 − x3 = 1
x1 − x2 + 5x3 = 2
Solution
5 1 −1
det(A) = 9 1 −1 = −16
1 −1 5
4 1 −1
det(A1 ) = 1 1 −1 = 12
2 −1 5
det(A1 ) 12
Hence x1 = det(A) = −16 = − 34 by Cramer’s rule.
Diagonalization
Table of contents
1 Determinant
Cofactor Expansion
Cramer’s rule
2 Diagonalization
Diagonalization

Diagonalization
Motivation example
Consider the evolution of the population of a species of birds. Because the
number of males and females are nearly equal, we count only females. We
assume that each female remains a juvenile for one year and then becomes
an adult, and that only adults have offspring. We make three assumptions
about reproduction and survival rates:
1 The number of juvenile females hatched in any year is twice the
number of adult females alive the year before (we say the
reproduction rate is 2).
2 Half of the adult females in any year survive to the next year (the
adult survival rate is 12 ).

3 One quarter of the juvenile females in any year survive into adulthood
(the juvenile survival rate is 14 ) .

If there were 100 adult females and 40 juvenile females alive initially,
compute the population of females k years late

Diagonalization
ak and jk : the number of adult and juvenile females after k years

total female population is ak + jk .
Assumption 1 shows that jk+1 = 2ak
assumptions 2 and 3 show that ak+1 = 12 ak + 14 jk .
Hence
1 1
ak+1 = ak + jk
2 4
jk+1 = 2ak
" # " #
1 1
a
Write vk = k and A = 2 4 then
jk 0 0
vk+1 = Avk matrix recurrence

So v1 = Av0 , v2 = Av1 = A2 v0 , . . . . In general
vk = Ak v0
" #
100
with v0 =
40
Diagonalization
Direct computation of the powers Ak can be time-consuming → need

an indirect method
the power Dk of a diagonal matrix D are easy to compute
   
d11 0 . . . 0 dk11 0 . . . 0
 0 d22 . . . 0  0 dk22 . . . 0
   
 =⇒ D k =  .
 
 ..
D= .. .. ..  . .. .. .. 
 . . . .

  . . . .


0 0 ... dnn 0 0 ... dknn
Idea: first diagonalize the matrix A, that is find an invertible matrix P

such that
P AP −1 = D is a diagonal matrix
then compute Ak
A2 = (P AP −1 )(P AP −1 ) = P A2 P −1 ,
A3 = AA2 = (P AP −1 )(P A2 P −1 ) = P A3 P −1

Diagonalization
Theorem
If A = P DP −1 then Ak = P Dk P −1 for k = 1, 2, . . .

Diagonalization
Theorem
If A = P DP −1 then Ak = P Dk P −1 for k = 1, 2, . . .
Question: How to find an invertible matrix P ?

Diagonalization Eigenvalues and Eigenvectors
Definition
If A is an n × n matrix, a number λ is called an eigenvalue of A if
Ax = λx for some column x ̸= 0 in Rn .
In this case, x is called an eigenvector of A corresponding to the
eigenvalue λ , or a λ-eigenvector for short.
Example
" # " #
3 5 5
If A = and x = then Ax = 4x. So λ = 4 is an eigenvalue of
1 −1 1
A corresponding to the eigenvector x.

λ is an eigenvalue of matrix A if and only if Ax = λx for some x ̸= 0. This

is equivalent to the system
(λI − A)x = 0
has non trivial solution. This happens if and only if λI − A is not
invertible.

λ is an eigenvalue of matrix A if and only if Ax = λx for some x ̸= 0. This

is equivalent to the system
(λI − A)x = 0
has non trivial solution. This happens if and only if λI − A is not
invertible.
Definition (Characteristic Polynomial of a Matrix)
If A is an n × n matrix then the characteristic polynomial of A is
cA (x) = det(xI − A)
Theorem
Let A is an n × n matrix.
1 The eigenvalues λ of A are the roots of the characteristic polynomial
cA (x) of A.
2 The λ-eigenvectors x are the nonzero solutions to the homogeneous
system
(λI − A)x = 0
Example
" #
3 5
Find the characteristic polynomial of the matrix A = and then
1 −1
find all the eigenvalues and their eigenvectors.
Solution
" # " # " #
x 0 3 5 x−3 5
xI − A = − =
0 x 1 −1 1 x+1
" #
x−3 5
cA (x) = det = x2 − 2x − 8 = (x − 4)(x + 2)
1 x+1
The roots of cA (x) are λ1 = −2 and λ2 = 4 and these are the eigenvalues

Solution (cont)
Find the eigenvectors corresponding to λ1 = −2
We have " # " #
−2 − 3 −5 −5 −5
λ1 I − A = =
−1 −2 + 1 −1 −1
" # " #
−t −1
General solution to (λ1 − A)x = 0 are x = =t . So the
t 1
" #
−t
eigenvectors corresponding to λ1 = −2 is x = for t ̸= 0.
t
" # " #
5t 5
Similar, the eigenvectors corresponding to λ2 = 4 are x = =t for
t 1
t ̸= 0

A square matrix A has many eigenvectors associated with any given

eigenvalue λ. Any set of nonzero multiples of the basic solutions of
(λI − A)x = 0 will
" be# called a set of basic eigenvectors corresponding to
−1
λ. For example, is a basic eigenvector corresponding to λ1 = −2.
1

Example
Find the characteristic polynomial, eigenvalues, and basic eigenvectors for
 
2 0 0
A = 1 2 −1
 
1 3 −2
Solution
Characteristic polynomial
 
x−2 0 0
cA (x) = det  −1 x − 2 1  = (x − 2)(x − 1)(x + 1)
 
11 −3 x + 2
so the eigenvalues are λ1 = 2, λ2 = 1 and λ3 = −1.

Diagonalization Diagonalization
Solution (cont)
To find all the eigenvectors corresponding to λ1 = 2, compute
   
λ1 − 2 0 0 0 0 0
λ1 I − A =  −1 λ1 − 2 1  = −1 0 1
   
11 −3 λ1 + 2 11 −3 4
Solve (λ1 I − A)x

 = 0 by gaussian algorithm
 to obtain the general
1 1
solution x = t  1 . So we can use x1 = 1 as the basic eigenvector
   
11 1
coresponding to λ
 1 = 2.  
0 0
Similary x2 = 1 and x3 = 1 are the basic eigenvectors corresponding
   
1 3
to λ2 = 1 and λ3 = −1

Diagonal matrix
An n × n matrix D is called a diagonal matrix if all its entries off the

main diagonal are zero
 
λ1 0 . . . 0
 0 λ2 . . . 0


 ..
D= .. . . .. 
 . . . . 

0 0 ... λn
Denote D = diag(λ1 , λ2 , . . . , λn )

Diagonalizable Matrices
An n × n matrix A is called diagonalizable if there exists an invertible

matrix P such that
P −1 AP is diagonal form
The invertible matrix P is called a diagonalizing matrix for A

h i
Denote x1 , x2 , . . . , xn the columns of P and write P = x1 x2 . . . xn
Then P −1 AP = D for some diagonal matrix D implied that
AP = P D
which is equivalent to
h i h i
Ax1 Ax2 . . . Axn = λ1 x1 λ2 x2 . . . λ n xn
So
Axk = λk xk
for all k = 1, 2, . . . , n.
In other words, P −1 AP = D holds if and only if the diagonal entries of
diagonal D are eigenvalues of A and the columns of the diagonalizing
matrix P are corresponding eigenvectors.

Diagonalization Algorithm
To diagonalize an n × n matrix A:
Step 1 Find the distinct eigenvalues λ of A.
Step 2 Compute a set of basic eigenvectors corresponding to each of these
eigenvalues λ as basic solutions of the homogeneous system
(λI − A)x = 0.
Step 3 The matrix A is diagonalizable if and only if there are n basic
eigenvectors in all.
Step 4 If A is diagonalizable, the n × n matrix P with these basic
eigenvectors as its columns is a diagonalizing matrix for A, that is, P
is invertible and P −1 AP is diagonal.

Example
Diagonalize the matrix  

2 0 0
A = 1 2 −1
 
1 3 −2

An Application to Linear Recurrences
Solution
The matrix A has three eigenvalues λ1=2, λ2 =   λ3 = −1with
1 and 
1 0 0
corresponding basis eigenvectors x1 = 1 , x2 = 1 and x3 = 1
     
1 1 3
The diagonalizing matrix
 
h i 1 0 0
P = x1 x2 x3 = 1 1 1
 
1 1 3
and    
λ1 0 0 2 0 0
−1
P AP = D =  0 λ2 0  = 0 1 0 
   
0 0 λ3 0 0 −1

Table of contents
1 Determinant
Cofactor Expansion
Cramer’s rule
2 Diagonalization
Diagonalization

Recurrece
It often happens that a problem can be solved by finding a sequence of
numbers x0 , x1 , x2 , ... where the first few are known, and subsequent
numbers are given in terms of earlier ones
Example
Fibonacci sequence
f0 = 1, f1 = 1, fn = fn−1 + fn−2 for n ≥ 2
Example
An urban planner wants to determine the number xk of ways that a row of
k parking spaces can be filled with cars and trucks if trucks take up two
spaces each. Find the first few values of xk .

Recurrece
It often happens that a problem can be solved by finding a sequence of
numbers x0 , x1 , x2 , ... where the first few are known, and subsequent
numbers are given in terms of earlier ones
Example
Fibonacci sequence
f0 = 1, f1 = 1, fn = fn−1 + fn−2 for n ≥ 2
Example
An urban planner wants to determine the number xk of ways that a row of
k parking spaces can be filled with cars and trucks if trucks take up two
spaces each. Find the first few values of xk .
Solution
Some first few terms x0 = 1, x1 = 1, x2 = 2 and xk+1 = xk + xk+1 ∀k ≥ 0

Question
How to find explicit formula for terms in a sequence given by a linear

recursive relation?

Example
Suppose the number x0 , x1 . . . are given by the linear recursive relation
xk+2 = xk+1 + 6xk
with x0 = 1, x1 = 3.
Solution
The idea is to compute the sequence v0 , v1 . . . of columns instead of the
numbers x0 , x1 , . . . where
" #
xk
vk = for k ≥ 0.
xk+1
" #
1
We have v0 = and
1
" # " # " #" # " #
x xk+1 0 1 xk 0 1
vk+1 = k+1 = = = Avk with A =
xk+2 6xk + xk+1 6 1 xk+1 6 1
Solution (cont)
It is easy to see that vk = Ak v0 . So we first diagonalize A and then

compute Ak to obtain the general formula for xk .

Solution (cont)
It is easy to see that vk = Ak v0 . So we first diagonalize A and then

compute Ak to obtain the general formula for xk .
Diagonalize A √
The characteric polynomial is cA (x) = x2 − x − 1 with roots 1±2 5 so A
has two eigenvalues
√ √
1+ 5 1− 5
λ1 = , λ2 =
2 2
" # " #
1 1
The corresponding eigenvectors are x1 = , x2 = .
λ1 λ2
" #
h i 1 1
The diagonalizing matrix for A is P = x1 x2 = which is
λ1 λ2
invertible

Solution (cont)
Find general formula of "xk #
λ1 0
We have P AP = D =
−1 So
0 λ2
Dk = (P −1 AP )(P −1 AP ) . . . P −1 AP = P −1 Ak P
It implies that
Ak = P D k P − 1
and then
 
" # λk+1
1 √ −λk+1
2
vk
= vk = Ak v0 = P Dk P − 1v0 = 5
 
 λk+2 −λ k+2 
vk+1 1 √ 2
5
Comparing top entries give an exact formula for xk

λk+1
1 √ − λk+1
2
xk =
5
Exercise
Find an exact formula for the numbers in Fibonacci sequence.

Application in Linear Dynamical System
Table of contents
1 Determinant
Cofactor Expansion
Cramer’s rule
2 Diagonalization
Diagonalization

Linear Dynamical System
If A is an n × n matrix, a sequence v0 , v1 , v2 , . . . is called a linear

dynamical system if v0 is specified and v1 , v2 , . . . are given by the matrix
recurrence
vk+1 = Avk
for each k ≥ 0.
We call A the migration matrix of the system

Example - Evolution of the population

The bird population
" # was described by computing the female population
ak
profile vk = of the species where ak and jk represents the number of
jk
adults and juvenile females present k years after the initial values a0 = 100
and j0 = 40 were observed. The model assumes that these numbers are
related by the following equation
ak+1 = 0.5ak + 0.25jk

jk+1 = 2ak
We can write
" #
0.5 0.25
vk+1 = Avk where A=
2 0
We can now use our diagonalization techniques to determine the

population profile vk for all values of k in terms of the initial values.
Diagonalize A
The chacteristic polynomial of A is
cA (x) = x2 − 0.5x − 0.5 = (x − 1)(x − 0.5)
so the eigenvalues are λ1 = 1, λ2 = 0.5. " #

0.5
Gaussian elimination gives corresponding basic eigenvectors and
1
" # " # " #
−0.25 1 −1
. For convenience, we use multiplies x1 = and x2 =
1 2 4
respectively. Hence a diagonalizing matrix
" #
1 −1
P =
2 4
and we obtain " #

−1 1 0
P AP = D =
0 −0.5

Compute vk
Compute explicitly Ak for k ≥ 0
" #" # " #
k k −1 1 −1 1 0 1 4 1
A = PD P = k
2 4 0 (−0.5) 6 −2 4
" #
1 4 + 2(−0.5)k 1 − (−0.5)k
=
6 8 − 8(−0.5)k 2 + 4(−0.5)k
Hence we obtain
" # " #" #
ak 1 4 + 2(−0.5)k 1 − (−0.5)k 100
= v k = Ak v 0 =
jk 6 8 − 8(−0.5)k 2 + 4(−0.5)k 40
" #
1 440 + 160(−0.5)k
=
6 880 − 640(−0.5)k
Thus we obtain
1 1
ak = 440 + 160(−0.5)k , jk = 880 − 640(−0.5)k
6 6
Observe that
lim ak = lim 440 + 160(−0.5)k = 440

k→∞ k→∞
while
lim jk = lim 880 − 640(−0.5)k = 880
k→∞ k→∞
So in long term ak ≈ 440 while jk ≈ 880, that is the female population
stabilizes with approximately twice as many juveniles as adults.

Determinant - Diagonalization

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Determinant - Diagonalization

Uploaded by

Copyright:

Available Formats

Determinant and Diagonalization

Determinant and Diagonalization 1 / 76

Study about determinant of a matrix and diagonalization a square matrix

3 Use determinant to find inverse of a matrix

4 Use determinant to solve system of linear equations: Cramer’s rule

5 Diagonalize a square matrix

6 Application to linear recurrences

Determinant and Diagonalization 2 / 76

3 An Application to Linear Recurrences

4 Application in Linear Dynamical System

Determinant and Diagonalization 3 / 76

Determinant of a square matrix is denoted by det A or det(A) or |A|.

Determinant and Diagonalization 4 / 76

Determinant and Diagonalization 5 / 76

In order to define the determinant of 3 × 3, 4 × 4 . . . matrix with inductive

Then the (i, j) - cofactor cij (A) is the scale defined by

cij (A) = (−1)i+j det(Aij )

(−1)i+j is called the sign of (i, j) - position

Determinant and Diagonalization 6 / 76

Matrix for sign of position

Determinant and Diagonalization 7 / 76

Determinant and Diagonalization 8 / 76

The sign of position (1, 2) is (−1)1+2 . So the (1, 2) - cofactor

Determinant and Diagonalization 8 / 76

Determinant by Cofactor Expansion

Cofactor expansion along row i

ai1 ci1 (A) + ai2 ci2 )(A) + · · · + ain cin (A)

Cofactor expansion along column j

a1j c1j (A) + a2j c2j )(A) + · · · + anj cnj (A)

The cofactor expansion along any row or column of a matrix A

Determinant and Diagonalization 9 / 76

Determinant and Diagonalization 10 / 76

Compute det A by expanding along the first column

Determinant and Diagonalization 11 / 76

Determinant and Diagonalization 12 / 76

Determinant and Diagonalization 13 / 76

It’d better to choose the column 3

Determinant and Diagonalization 13 / 76

Determinant of triangular matrices

Determinant of a triangular matrix is the product of the entries on

determinant = a11 a22 . . . ann

Determinant and Diagonalization 14 / 76

Determinant and Diagonalization 15 / 76

Determinant and Diagonalization 16 / 76

Geometric Meaning of Determinant

Area of parallelogram defined by two vectors is

Figure 1: Parallelogram Area of parallelogram =rectangle area − area of A − · · ·

determinant of the matrix’s columns (vectors)

Determinant and Diagonalization 17 / 76

Geometric Meaning of Determinant (cont)

(Oriented) volume of the parallelepiped P defined three vector v1 , v2 , v3

Determinant and Diagonalization 18 / 76

Orientation - swapping changes the sign of the determinant

Determinant and Diagonalization 20 / 76

Determinant is Unaffected by Combining Columns

The box formed by two vectors →

Determinant and Diagonalization 21 / 76

Cut down the work to compute determinant

If A is a 4 × 4 matrix, the cofactor expansion along any row or

Determinant and Diagonalization 22 / 76

Cut down the work to compute determinant

If A is a 4 × 4 matrix, the cofactor expansion along any row or