Scribd
Upload
11 views15 pages

Lec 03 M

Uploaded by

taiyamxx
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
11 views15 pages

Lec 03 M

Uploaded by

taiyamxx
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

L03-S00

Eigendecompositions of Hermitian matrices

MATH 6610 Lecture 03

September 4, 2020

MATH 6610-001 – U. Utah Hermitian matrices


L03-S01
Diagonalizability
Recall:
All non-defective square matrices are diagonalizable (eigenvalue
decomposition)
All square matrices are bidiagonalizable (Jordan normal form)
All square matrices are unitarily triangularizable (Schur decomposition)

MATH 6610-001 – U. Utah Hermitian matrices


L03-S01
Diagonalizability
Recall:
All non-defective square matrices are diagonalizable (eigenvalue
decomposition)
All square matrices are bidiagonalizable (Jordan normal form)
All square matrices are unitarily triangularizable (Schur decomposition)

When are matrices unitarily diagonalizable?

MATH 6610-001 – U. Utah Hermitian matrices


L03-S02
A spectral theorem

Theorem
If A P nˆn is Hermitian, then it is unitarily diagonalizable with real
eigenvalues.
Hermitian matrices are also called self-adjoint.

MATH 6610-001 – U. Utah Hermitian matrices


L03-S02
A spectral theorem

Theorem
If A P nˆn is Hermitian, then it is unitarily diagonalizable with real
eigenvalues.
Hermitian matrices are also called self-adjoint. If A P nˆn
is unitarily
diagonalizable, then it can be written as
ÿ
n
˚ ˚
A “ U ⇤U “ j uj uj ,
j“1

where tuj unj“1 are the columns of U .

MATH 6610-001 – U. Utah Hermitian matrices


L03-S02
A spectral theorem

Theorem
If A P nˆn is Hermitian, then it is unitarily diagonalizable with real
eigenvalues.
Hermitian matrices are also called self-adjoint. If A P nˆn
is unitarily
diagonalizable, then it can be written as
ÿ
n
˚ ˚
A “ U ⇤U “ j uj uj ,
j“1

where tuj unj“1 are the columns of U .

For example, the spectral radius of a matrix A is

⇢pAq :“ max | j pAq|


j“1,...,n

If A is Hermitian, then }A}2 “ ⇢pAq.


MATH 6610-001 – U. Utah Hermitian matrices
L03-S03
(Hermitian) Positive-definite matrices
A matrix A P nˆn is Hermitian positive definite (sometimes symmetric
positive-definite or “spd”) if it’s Hermitian and its spectrum is strictly positive.
(Respectively, positive semi-definite if the spectrum is non-negative.)

MATH 6610-001 – U. Utah Hermitian matrices


L03-S03
(Hermitian) Positive-definite matrices
A matrix A P nˆn is Hermitian positive definite (sometimes symmetric
positive-definite or “spd”) if it’s Hermitian and its spectrum is strictly positive.
(Respectively, positive semi-definite if the spectrum is non-negative.)
Such matrices actually define a norm: }x}2A :“ x˚ Ax is a norm.

MATH 6610-001 – U. Utah Hermitian matrices


L03-S04
Matrix square roots
There is also a functional calculus on spd matrices.
For example, a matrix S is the square root of a matrix A if A “ S 2 .

MATH 6610-001 – U. Utah Hermitian matrices


L03-S04
Matrix square roots
There is also a functional calculus on spd matrices.
For example, a matrix S is the square root of a matrix A if A “ S 2 .

Example
If A is spd, compute a matrix square root of A.

MATH 6610-001 – U. Utah Hermitian matrices


L03-S04
Matrix square roots
There is also a functional calculus on spd matrices.
For example, a matrix S is the square root of a matrix A if A “ S 2 .

Example
If A is spd, compute a matrix square root of A.

Theorem
If A is spd, then there is a(n essentially) unique spd square root of A.

MATH 6610-001 – U. Utah Hermitian matrices

You might also like