Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Department of Mathematics
Indian Institute of Technology Guwahati
1 / 11
Similarity and diagonalization
Topics:
Similarity transformation
Diagonalization of matrices and operators
Triangularization of complex matrices
2 / 11
Similar matrices
We have seen that diagonal and triangular matrices are nice in the
sense that their eigenvalues are transparently displayed.
3 / 11
Similar matrices
We have seen that diagonal and triangular matrices are nice in the
sense that their eigenvalues are transparently displayed.
It would be desirable to transform a square matrix to a diagonal or
triangular matrix in such a way that they have exactly the same
eigenvalues.
3 / 11
Similar matrices
We have seen that diagonal and triangular matrices are nice in the
sense that their eigenvalues are transparently displayed.
It would be desirable to transform a square matrix to a diagonal or
triangular matrix in such a way that they have exactly the same
eigenvalues.
3 / 11
Similar matrices
We have seen that diagonal and triangular matrices are nice in the
sense that their eigenvalues are transparently displayed.
It would be desirable to transform a square matrix to a diagonal or
triangular matrix in such a way that they have exactly the same
eigenvalues.
3 / 11
Similar matrices
We have seen that diagonal and triangular matrices are nice in the
sense that their eigenvalues are transparently displayed.
It would be desirable to transform a square matrix to a diagonal or
triangular matrix in such a way that they have exactly the same
eigenvalues.
3 / 11
Similar matrices
We have seen that diagonal and triangular matrices are nice in the
sense that their eigenvalues are transparently displayed.
It would be desirable to transform a square matrix to a diagonal or
triangular matrix in such a way that they have exactly the same
eigenvalues.
3 / 11
Similar matrices
We have seen that diagonal and triangular matrices are nice in the
sense that their eigenvalues are transparently displayed.
It would be desirable to transform a square matrix to a diagonal or
triangular matrix in such a way that they have exactly the same
eigenvalues.
3 / 11
Diagonalization
Remark: Note that A and B := P −1 AP have the same eigenvalues
and
4 / 11
Diagonalization
Remark: Note that A and B := P −1 AP have the same eigenvalues
and Bv = λv ⇔ Au = λu,
4 / 11
Diagonalization
Remark: Note that A and B := P −1 AP have the same eigenvalues
and Bv = λv ⇔ Au = λu, where u := Pv.
4 / 11
Diagonalization
Remark: Note that A and B := P −1 AP have the same eigenvalues
and Bv = λv ⇔ Au = λu, where u := Pv.
1 2 1 0
Example: Let A = and B = .
0 −1 −2 −1
4 / 11
Diagonalization
Remark: Note that A and B := P −1 AP have the same eigenvalues
and Bv = λv ⇔ Au = λu, where u := Pv.
1 2 1 0
Example: Let A = and B = . Then
0 −1 −2 −1
A ∼ B,
4 / 11
Diagonalization
Remark: Note that A and B := P −1 AP have the same eigenvalues
and Bv = λv ⇔ Au = λu, where u := Pv.
1 2 1 0
Example: Let A = and B = . Then
0 −1 −2 −1
1 −1
A ∼ B, since AP = PB, where P =
1 1
4 / 11
Diagonalization
Remark: Note that A and B := P −1 AP have the same eigenvalues
and Bv = λv ⇔ Au = λu, where u := Pv.
1 2 1 0
Example: Let A = and B = . Then
0 −1 −2 −1
1 −1
A ∼ B, since AP = PB, where P =
1 1
4 / 11
Diagonalization
Remark: Note that A and B := P −1 AP have the same eigenvalues
and Bv = λv ⇔ Au = λu, where u := Pv.
1 2 1 0
Example: Let A = and B = . Then
0 −1 −2 −1
1 −1
A ∼ B, since AP = PB, where P =
1 1
4 / 11
Diagonalization
Remark: Note that A and B := P −1 AP have the same eigenvalues
and Bv = λv ⇔ Au = λu, where u := Pv.
1 2 1 0
Example: Let A = and B = . Then
0 −1 −2 −1
1 −1
A ∼ B, since AP = PB, where P =
1 1
4 / 11
Diagonalization
Remark: Note that A and B := P −1 AP have the same eigenvalues
and Bv = λv ⇔ Au = λu, where u := Pv.
1 2 1 0
Example: Let A = and B = . Then
0 −1 −2 −1
1 −1
A ∼ B, since AP = PB, where P =
1 1
5 / 11
Diagonalization
5 / 11
Diagonalization
5 / 11
Diagonalization
5 / 11
Diagonalization
5 / 11
Diagonalization
5 / 11
Diagonalization
5 / 11
Diagonalization
5 / 11
Diagonalization
5 / 11
Characterization
6 / 11
Characterization
Exercise: Let λ1 , λ2 , . . . , λk be distinct eigenvalues of a matrix A.
Suppose Bi is a basis for the eigenspace Eλi .
6 / 11
Characterization
Exercise: Let λ1 , λ2 , . . . , λk be distinct eigenvalues of a matrix A.
Suppose Bi is a basis for the eigenspace Eλi . Then
B = B1 ∪ B2 ∪ . . . ∪ Bk is a linearly independent set.
6 / 11
Characterization
Exercise: Let λ1 , λ2 , . . . , λk be distinct eigenvalues of a matrix A.
Suppose Bi is a basis for the eigenspace Eλi . Then
B = B1 ∪ B2 ∪ . . . ∪ Bk is a linearly independent set.
6 / 11
Characterization
Exercise: Let λ1 , λ2 , . . . , λk be distinct eigenvalues of a matrix A.
Suppose Bi is a basis for the eigenspace Eλi . Then
B = B1 ∪ B2 ∪ . . . ∪ Bk is a linearly independent set.
6 / 11
Characterization
Exercise: Let λ1 , λ2 , . . . , λk be distinct eigenvalues of a matrix A.
Suppose Bi is a basis for the eigenspace Eλi . Then
B = B1 ∪ B2 ∪ . . . ∪ Bk is a linearly independent set.
6 / 11
Characterization
Exercise: Let λ1 , λ2 , . . . , λk be distinct eigenvalues of a matrix A.
Suppose Bi is a basis for the eigenspace Eλi . Then
B = B1 ∪ B2 ∪ . . . ∪ Bk is a linearly independent set.
6 / 11
Characterization
Exercise: Let λ1 , λ2 , . . . , λk be distinct eigenvalues of a matrix A.
Suppose Bi is a basis for the eigenspace Eλi . Then
B = B1 ∪ B2 ∪ . . . ∪ Bk is a linearly independent set.
6 / 11
Characterization
Exercise: Let λ1 , λ2 , . . . , λk be distinct eigenvalues of a matrix A.
Suppose Bi is a basis for the eigenspace Eλi . Then
B = B1 ∪ B2 ∪ . . . ∪ Bk is a linearly independent set.
6 / 11
Characterization
Exercise: Let λ1 , λ2 , . . . , λk be distinct eigenvalues of a matrix A.
Suppose Bi is a basis for the eigenspace Eλi . Then
B = B1 ∪ B2 ∪ . . . ∪ Bk is a linearly independent set.
7 / 11
Examples
1 2 3
" #
Example: The matrix A = 0 4 5 is diagonalizable because
0 0 6
1, 4, 6 are (distinct) eigenvalues of A.
7 / 11
Examples
1 2 3
" #
Example: The matrix A = 0 4 5 is diagonalizable because
0 0 6
1, 4, 6 are (distinct) eigenvalues of A.
1 2 16
" # " # " #
You can easily find: 1, 4, 6 have eigenvectors 0 , 3 , 25 ,
0 0 10
respectively.
7 / 11
Examples
1 2 3
" #
Example: The matrix A = 0 4 5 is diagonalizable because
0 0 6
1, 4, 6 are (distinct) eigenvalues of A.
1 2 16
" # " # " #
You can easily find: 1, 4, 6 have eigenvectors 0 , 3 , 25 ,
0 0 10
respectively.
1 1 1
" #
Example: The matrix A = 0 1 1 is not diagonalizable
0 0 1
7 / 11
Examples
1 2 3
" #
Example: The matrix A = 0 4 5 is diagonalizable because
0 0 6
1, 4, 6 are (distinct) eigenvalues of A.
1 2 16
" # " # " #
You can easily find: 1, 4, 6 have eigenvectors 0 , 3 , 25 ,
0 0 10
respectively.
1 1 1
" #
Example: The matrix A = 0 1 1 is not diagonalizable because
0 0 1
1 is the only eigenvalue with algebraic multiplicity 3 but geometric
multiplicity 1.
7 / 11
Examples
1 2 3
" #
Example: The matrix A = 0 4 5 is diagonalizable because
0 0 6
1, 4, 6 are (distinct) eigenvalues of A.
1 2 16
" # " # " #
You can easily find: 1, 4, 6 have eigenvectors 0 , 3 , 25 ,
0 0 10
respectively.
1 1 1
" #
Example: The matrix A = 0 1 1 is not diagonalizable because
0 0 1
1 is the only eigenvalue with algebraic multiplicity 3 but geometric
multiplicity 1. Indeed, the eigenspace E1 is given by
0 1 1 1
" #! " #!
E1 = null 0 0 1 = span 0 .
0 0 0 0
7 / 11
Example
Example: Let T : R2 [x] −→ R2 [x] be given by
8 / 11
Example
Example: Let T : R2 [x] −→ R2 [x] be given by
8 / 11
Example
Example: Let T : R2 [x] −→ R2 [x] be given by
8 / 11
Example
Example: Let T : R2 [x] −→ R2 [x] be given by
8 / 11
Example
Example: Let T : R2 [x] −→ R2 [x] be given by
8 / 11
Example
Example: Let T : R2 [x] −→ R2 [x] be given by
8 / 11
Example
−2 1 0
A − 3I = 0 −1 2 = null(A − I ) := span{[1, 2, 1]> }.
0 0 0
9 / 11
Example
−2 1 0
A − 3I = 0 −1 2 = null(A − I ) := span{[1, 2, 1]> }.
0 0 0
9 / 11
Example
−2 1 0
A − 3I = 0 −1 2 = null(A − I ) := span{[1, 2, 1]> }.
0 0 0
[T ]C = diag(1, 2, 3).
9 / 11
Example
−2 1 0
A − 3I = 0 −1 2 = null(A − I ) := span{[1, 2, 1]> }.
0 0 0
[T ]C = diag(1, 2, 3).
9 / 11
Example
−2 1 0
A − 3I = 0 −1 2 = null(A − I ) := span{[1, 2, 1]> }.
0 0 0
[T ]C = diag(1, 2, 3).
9 / 11
Example
−2 1 0
A − 3I = 0 −1 2 = null(A − I ) := span{[1, 2, 1]> }.
0 0 0
[T ]C = diag(1, 2, 3).
9 / 11
Example
−2 1 0
A − 3I = 0 −1 2 = null(A − I ) := span{[1, 2, 1]> }.
0 0 0
[T ]C = diag(1, 2, 3).
9 / 11
Triangularization
Theorem: Let A ∈ Mn (C). Then there is an invertible matrix
P ∈ Mn (C) such that U := P −1 AP is upper triangular.
10 / 11
Triangularization
Theorem: Let A ∈ Mn (C). Then there is an invertible matrix
P ∈ Mn (C) such that U := P −1 AP is upper triangular.
Proof: Apply induction on n. The result holds for n = 1.
10 / 11
Triangularization
Theorem: Let A ∈ Mn (C). Then there is an invertible matrix
P ∈ Mn (C) such that U := P −1 AP is upper triangular.
Proof: Apply induction on n. The result holds for n = 1. Assume
that the result holds for n − 1.
10 / 11
Triangularization
Theorem: Let A ∈ Mn (C). Then there is an invertible matrix
P ∈ Mn (C) such that U := P −1 AP is upper triangular.
Proof: Apply induction on n. The result holds for n = 1. Assume
that the result holds for n − 1.
Let λ be an eigenvalue of A and v be an eigenvector. Then
Av = λv.
10 / 11
Triangularization
Theorem: Let A ∈ Mn (C). Then there is an invertible matrix
P ∈ Mn (C) such that U := P −1 AP is upper triangular.
Proof: Apply induction on n. The result holds for n = 1. Assume
that the result holds for n − 1.
Let λ be an eigenvalue of A and v be an eigenvector. Then
Av = λv. Choose an n × (n − 1) matrix V such that P1 := [v, V ]
is invertible.
10 / 11
Triangularization
Theorem: Let A ∈ Mn (C). Then there is an invertible matrix
P ∈ Mn (C) such that U := P −1 AP is upper triangular.
Proof: Apply induction on n. The result holds for n = 1. Assume
that the result holds for n − 1.
Let λ be an eigenvalue of A and v be an eigenvector. Then
Av = λv. Choose an n × (n − 1) matrix V such that P1 := [v, V ]
λ h
is invertible. Then P1−1 AP1 =
0 A
b
10 / 11
Triangularization
Theorem: Let A ∈ Mn (C). Then there is an invertible matrix
P ∈ Mn (C) such that U := P −1 AP is upper triangular.
Proof: Apply induction on n. The result holds for n = 1. Assume
that the result holds for n − 1.
Let λ be an eigenvalue of A and v be an eigenvector. Then
Av = λv. Choose an n × (n − 1) matrix V such that P1 := [v, V ]
λ h
is invertible. Then P1−1 AP1 = for some row vector h
0 A
b
and (n − 1) × (n − 1) matrix A.b
10 / 11
Triangularization
Theorem: Let A ∈ Mn (C). Then there is an invertible matrix
P ∈ Mn (C) such that U := P −1 AP is upper triangular.
Proof: Apply induction on n. The result holds for n = 1. Assume
that the result holds for n − 1.
Let λ be an eigenvalue of A and v be an eigenvector. Then
Av = λv. Choose an n × (n − 1) matrix V such that P1 := [v, V ]
λ h
is invertible. Then P1−1 AP1 = for some row vector h
0 A b
and (n − 1) × (n − 1) matrix A.b
By induction hypothesis, there exists P
b such that U b −1 A
b := (P) bPb is
upper triangular.
10 / 11
Triangularization
Theorem: Let A ∈ Mn (C). Then there is an invertible matrix
P ∈ Mn (C) such that U := P −1 AP is upper triangular.
Proof: Apply induction on n. The result holds for n = 1. Assume
that the result holds for n − 1.
Let λ be an eigenvalue of A and v be an eigenvector. Then
Av = λv. Choose an n × (n − 1) matrix V such that P1 := [v, V ]
λ h
is invertible. Then P1−1 AP1 = for some row vector h
0 A b
and (n − 1) × (n − 1) matrix A.b
By induction hypothesis, thereexists P
b such that U b −1 A
b := (P) bPb is
1 0
upper triangular. Set P := P1 b .
0 P
10 / 11
Triangularization
Theorem: Let A ∈ Mn (C). Then there is an invertible matrix
P ∈ Mn (C) such that U := P −1 AP is upper triangular.
Proof: Apply induction on n. The result holds for n = 1. Assume
that the result holds for n − 1.
Let λ be an eigenvalue of A and v be an eigenvector. Then
Av = λv. Choose an n × (n − 1) matrix V such that P1 := [v, V ]
λ h
is invertible. Then P1−1 AP1 = for some row vector h
0 Ab
and (n − 1) × (n − 1) matrix A.b
By induction hypothesis, thereexists P
b such that U b −1 A
b := (P) bPb is
1 0
upper triangular. Set P := P1 b . Then
0 P
" #
λ h P
P −1 AP =
b
= upper triangular.
0 U b
10 / 11
Trace and eigenvalues
Theorem: Let A ∈ Mn (C) with eigenvalues λ1 , . . . , λn (counted
according to their algebraic multiplicities).
11 / 11
Trace and eigenvalues
Theorem: Let A ∈ Mn (C) with eigenvalues λ1 , . . . , λn (counted
according to their algebraic multiplicities). Then
n
Y
Trace(A) = λ1 + · · · + λn and det (A) = λj .
j=1
11 / 11
Trace and eigenvalues
Theorem: Let A ∈ Mn (C) with eigenvalues λ1 , . . . , λn (counted
according to their algebraic multiplicities). Then
n
Y
Trace(A) = λ1 + · · · + λn and det (A) = λj .
j=1
11 / 11
Trace and eigenvalues
Theorem: Let A ∈ Mn (C) with eigenvalues λ1 , . . . , λn (counted
according to their algebraic multiplicities). Then
n
Y
Trace(A) = λ1 + · · · + λn and det (A) = λj .
j=1
11 / 11