Can Kronecker Exponentiation be Simplified for Matrices in Different Forms?

  • Thread starter Thread starter prospero
  • Start date Start date
Join the discussion
Registration is free. Ask a follow-up in this thread, or start your own.
1 reply · 1K views
prospero
Messages
1
Reaction score
0

Homework Statement



I am trying to derive the Kronecker exponentiation relation:

[tex]e^A \otimes e^B = e^{A \oplus B}[/tex]

with A,B as n-by-n and m-by-m matrices.

Homework Equations



Kronecker sum is defined as:

[tex]A \oplus B = A \otimes I_m + B \otimes I_n[/tex]

The Attempt at a Solution



I first tackle the simpler case of A,B and I all being 2-by-2 matrices.

[tex]A \otimes I = \begin{bmatrix} A_{11} I & A_{12} I \\ A_{21} I & A_{22} I \end{bmatrix}[/tex]

and

[tex]I \otimes B = \begin{bmatrix} B & 0 \\ 0 & B \end{bmatrix}[/tex]

Then I can represent [itex]A \otimes I[/itex] as the sum of one diagonal and two nilpotent matrices. This makes their matrix exponentiation easier. [itex]I \otimes B[/itex] is a diagonal block matrix and exponentiating it, we get a block matrix with blocks of [itex]e^B[/itex] along the diagonal, i.e.:

[tex]e^{I \otimes B} = \begin{bmatrix} e^B & 0 \\ 0 & e^B \end{bmatrix}[/tex]

I tried multiplying the matrices together and it might have worked but it was taking too much time and it is clearly not feasible to prove the general case.

However I notice that [itex]e^{I \otimes B} = I \otimes e^B[/itex].

Now if we can write [itex]e^{A \otimes I} = e^A \otimes I[/itex] the problem is over since [itex]( e^A \otimes I ) ( I \otimes e^B ) = e^A \otimes e^B[/itex].

However I am not aware of a property of matrix exponentiation that justifies the last step.
 
Last edited:
Physics news on Phys.org
Here is a way to tackle your last question. The statement is true if A is diagonal. Can you show it is true if A is diagonalizable? Can you show it is true if A is not diagonalizable but is in Jordan form? I believe if A is in Jordan form with blocks ##J_1, J_2 ... ## etc then ##e^A## is just ##e^{J_1}, e^{J_2},## etc.

Since most of the properties of the underlying matrices are the same under the Kronecker product, I think this would most likely work out.