MA614 – Applied Linear Algebra
Matrix Functions Via Jordan Canonical Form
Dr. Sairam Kaliraj
It is well known that any n x n matrix A with complex entries can be written in the Jordan
Canonical Form
Z−1 AZ = J = diag(J1 , J2 , . . . , Jp),
where
where Z is nonsingular and m1 + m2 + · · · + mp = n. The Jordan matrix J is unique up to the
ordering of the blocks Ji , but the transforming matrix Z is not unique.
Denote by λ1 , . . . , λs the distinct eigenvalues of A and let mi be the order of the largest
Jordan block in which λi appears, which is called the index of λi.
Definition 1: The function f is said to be defined on the spectrum of A, if the values
f( k) (λi ), for k = 0, 1, . . . mi − 1, and i = 1, 2, ... s
exist. These are called the values of the function f on the spectrum of A.
Here f( k) (λi ) denotes the kth derivative of f evaluated at the point λi.
Matrix Function via Jordan canonical form
Theorem 1: Let f be defined on the spectrum of A ∈ Cn×n and let A have the Jordan canonical
form
Z−1 AZ = J = diag(J1 , J2 , . . . , Jp)
as given in the previous slide. Then
Where
Remark: The lower diagonal entries in f(Jk ) are 0s only.
We consider the above formula as a definition for f(A), when f is defined on the spectrum of A.
In fact, it was derived from the way we have defined Matrix Functions for functions with
Taylor series expansion!
For the sake of motivation, I shall explain that in a casual way: Let Jk = λk I + Nk with Nk is a
mk x mk matrix.
In general, powering Nk causes the superdiagonal of 1s to move a diagonal at a time towards
the top right-hand corner, until at the mkth power it disappears. Since f is defined on the
spectrum of A, we consider the Taylor series of f with respect to the point λk.
Here we replace t by Jk
and λk by λk I, so that
(t – λk) will be replaced
by Nk .
since all powers of Nk from the mkth onwards are zero.
Remark:The above definition of matrix function coincides with the original matrices such as
An, A1/m (whenever it can be defined), when n, m are integers.
From the above discussions, it is natural to ask the following question: When f is defined on the
Spectrum of A, does there exist a polynomial p(t) such that p(A) = f(A)?
The answer to this question is affirmative and it is as follows.
Theorem 2: (Lagrange–Hermite formula): Let A be an n x n matrix with s distinct eigenvalues
λ1 , . . . , λs with the index n1, . . . , ns respectively. Suppose that f is defined on the spectrum of
A. Then the polynomial p given explicitly by the Lagrange - Hermite formula
Where has the property that p(A) = f(A).
When all the Eigenvalues of A are distinct
Corollary 3: When all the eigenvalues of A are distinct, the above Lagrange–Hermite formula
reduced to the familiar Lagrange formula
Example: Consider f (t) = √ t and
The eigenvalues are 1 and 4, so s = 2 and n1 = n2 = 1. So we can use Lagrange formula to find
The polynomial p(t) (Here, we use the positive square root of f. If we use the negative square
Root of f, we will get a different polynomial q which will give another square root).
Here p(1) = f (1) = 1 and p(4) = f (4) = 2 and the corresponding Lagrange polynomial is
given by
Verify that (f(A))2 = A.
Remark: We are not restricted to using the same branch of the square root function at
each eigenvalue. For example, with f (1) = 1 and f (4) = −2 we obtain p(t) = 2 − t and
Exercise: By using the other two possible combinations of square root of f, find the other
two possible square roots of A.
