Scribd
Upload
26 views2 pages

Further Mathematical Methods (Linear Algebra) 2002 Lecture 11: Special' Real and Complex Matrices

This document summarizes key concepts related to special types of real and complex matrices: 1) It defines notation for the complex conjugate and complex conjugate transpose of matrices. It also establishes conventions for working with real and complex inner products and vectors. 2) It defines Hermitian, unitary, symmetric, and orthogonal matrices and notes relationships between these types of real and complex matrices. 3) It introduces the concept of a normal matrix and states the theorem that a complex matrix is normal if and only if it is unitarily diagonalizable. It also provides a proof that a unitarily diagonalizable matrix is normal.

Uploaded by

Gag Paf
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
26 views2 pages

Further Mathematical Methods (Linear Algebra) 2002 Lecture 11: Special' Real and Complex Matrices

This document summarizes key concepts related to special types of real and complex matrices: 1) It defines notation for the complex conjugate and complex conjugate transpose of matrices. It also establishes conventions for working with real and complex inner products and vectors. 2) It defines Hermitian, unitary, symmetric, and orthogonal matrices and notes relationships between these types of real and complex matrices. 3) It introduces the concept of a normal matrix and states the theorem that a complex matrix is normal if and only if it is unitarily diagonalizable. It also provides a proof that a unitarily diagonalizable matrix is normal.

Uploaded by

Gag Paf
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/ 2

Further Mathematical Methods (Linear Algebra) 2002

Lecture 11: Special Real and Complex Matrices


Notation
We denote the complex conjugate of a matrix A by A and the complex conjugate transpose of a
matrix A by A . Consequently, we can see that A = (A )t = (At ) , and indeed, that (A ) = A.
Notice that this is not the notation used in the book by Ostaszewski where the complex conjugate
of a matrix A is denoted by A and the complex conjugate transpose of a matrix is denoted by A .

Convention
Throughout this part of the course, we will assume that we are working with the real Euclidean inner
product when in Rn (i.e. if x = [x1 , . . . , xn ]t and y = [y1 , . . . , yn ]t are vectors in Rn , then hx, yi =
x1 y1 + + xn yn ) and the complex Euclidean inner product when in Cn (i.e. if x = [x1 , . . . , xn ]t
and y = [y1 , . . . , yn ]t are vectors in Cn , then hx, yi = x1 y1 + + xn yn ). Further, note that for such
vectors in Cn , x y = x1 y1 + + xn yn = (x1 y1 + + xn yn ) = hx, yi .

Definitions
Here is a summary of the special types of complex matrix (and their real analogues) that we shall
meet in the lecture. If we have an n n matrix A with complex entries, i.e. the vectors that form
the column space of the matrix are in Cn , we say that

A is Hermitian if A = A.

A is unitary if AA = I.

However, if we have an n n matrix A with real entries, i.e. the vectors that form the the column
space of the matrix are in Rn , we say that

A is symmetric if At = A.

A is orthogonal if AAt = I.

As R C (and, Rn Cn ), a matrix with real entries is a special case of a matrix with complex
entries (as if a matrix A has real entries, A = A). It should be clear that all symmetric matrices
are Hermitian (as for a symmetric matrix A, A = (A )t = At = A) and all orthogonal matrices
are unitary (as for an orthogonal matrix A, AA = A(A )t = AAt = I). Consequently, most of our
theorems about Hermitian and unitary matrices can be translated into theorems about symmetric
and orthogonal matrices [respectively] by using the following rules:

the complex matrix A (i.e. CS(A) Cn ) becomes the real matrix A (i.e. CS(A) Rn )
the complex conjugate transpose of A (i.e. A ) becomes the transpose of A (i.e. At )
the unitary matrix A (i.e. AA = I) becomes the orthogonal matrix A (i.e. AAt = I)
the Hermitian matrix A (i.e. A = A) becomes the symmetric matrix A (i.e. At = A)

For example: Notice that if A and B are two complex matrices, then (AB) = B A . But, if A and
B are two real matrices, then (AB)t = Bt At . We will assume this result (but, see Problem Sheet 6,
Question 5).
Lastly, as unitary matrices have the property that AA = I, it should be clear that if A is a unitary
matrix, then A = A1 . Thus, for a unitary matrix we have A A = I as well. Similarly, if A is an
orthogonal matrix, then At = A1 and At A = I too.

11-1
Unitary Diagonalisation
As we are often keen to diagonalise matrices, it is useful to note that a particularly nice form of
diagonalisation is available in some cases. To see this, observe that:

Theorem: A [square] complex matrix A has an orthonormal set of eigenvectors iff there exists a
unitary matrix P such that the matrix P AP is diagonal. If such a P exists, then A is said to be
unitarily diagonalisable.

Proof: It should be clear that a matrix P is unitary iff the column vectors of P form an orthonormal
set. To see this, observe that if the column vectors of a 3 3 matrix P are the vectors v1 , v2 and
v3 , then the product P P is given by:

v1 v1 v1 v1 v2 v1 v3 hv1 , v1 i hv1 , v2 i hv1 , v3 i

v2 v1 v2 v3 = v2 v1 v2 v2 v2 v3 = hv2 , v1 i hv2 , v2 i hv2 , v3 i

v3 v3 v1 v3 v2 v3 v3 hv3 , v1 i hv3 , v2 i hv3 , v3 i

where we have used the fact mentioned earlier when we set up our convention. So, clearly, P P = I
iff the column vectors of P form an orthonormal set. Now, to prove the theorem, we need to show
that it holds in both directions:
LTR: Assume that A is a [square] complex matrix with an orthonormal set of eigenvectors. This
means that A is diagonalisable (as orthonormal vectors are linearly independent) and so there exists
a P such that the matrix P1 AP is diagonal. This matrix P is the matrix whose columns are the
eigenvectors of A, and as these are orthonormal, we can see from the above observation that P will be
unitary. For unitary matrices, P1 = P , and therefore there exists a P such that P AP is diagonal
(as required).
RTL: Assume that there exists a unitary matrix P such that P AP is diagonal. Now, as P is unitary,
from the observation above, its column vectors must be orthonormal. Further, as P AP is diagonal
and unitary matrices are such that P1 = P , we have a matrix P1 AP that is diagonal. Now, this
is the result of our standard diagonalisation procedure, and so P is a matrix whose columns are the
eigenvectors of the [in general, complex and square] matrix A. But we have seen that these column
vectors are orthonormal, and so the eigenvectors of A must form an orthonormal set (as required).

It is now convenient to define another special kind of complex matrix, namely:

Definition: A [square] complex matrix A is normal iff A A = AA .

and this leads to the following useful theorem:

Theorem: A [square] complex matrix is normal iff A is unitarily diagonalisable.

Proof: We have to establish that this theorem holds by proving it in both directions.
RTL: We assume that A is unitarily diagonalisable, that is, there exists a unitary matrix P such that
the matrix P AP = D is diagonal. Now P is unitary, and so P = P1 , which means that A = PDP ,
and taking the complex conjugate transpose of this we get A = (PDP ) = PD P . Further, using
these, we can see that

AA = (PDP )(PD P ) = PDD P and A A = (PD P )(PDP ) = PD DP

where, again, we have used the fact that P = P1 . But, for the diagonal matrix D = diag(1 , 2 , . . . ,
n ) it should be clear that DD = D D as

DD = diag(|1 |2 , |2 |2 , . . . , |n |2 ) = D D

Consequently, AA = A A, and so the matrix A is normal (as required).


LTR: This is quite hard, and so due to considerations of space, we shall omit it here.

11-2

You might also like