1
$\begingroup$

Can the following quantity be reduced to further

$[A \otimes B + (\mathbb{1} - A) \otimes D]^N$, for positive interger $N$? Here, $\otimes$ denotes the Kronecker (tensor) product and the matrices $A$ and $B$ may not have the same dimensions, and are non-commuting, in general.

Edit: The matrices are non-commuting, in general.

$\endgroup$
0

1 Answer 1

1
$\begingroup$

*edit: I have to correct myself. I had incorrectly permuted terms (without noticing) which of course in general for matrices doesn't work.

What you can use is $(A \otimes B)\cdot(C \otimes D) = (AC \otimes BD)$. The latter implies $(A \otimes B)^n = A^n \otimes B^n$. However, expanding the $N$-th power of a sum is in general a bit more tedious for matrices. You have $$(P+Q)^2 = P^2 + PQ + QP + Q^2,$$ which for you becomes $$A^2\otimes B^2 + A(1-A) \otimes BD + (1-A) A \otimes DB + (1-A)^2 \otimes D^2.$$ Likewise, for $N=3$ you would need $$(P+Q)^3 = P^3 + PPQ + PQP + QPP + PQQ + QPQ + QQP + Q^3,$$ which would give you $$A^3\otimes B^3 + A^2(1-A)\otimes B^2D + A(1-A)A \otimes BDB + (1-A) A^2 \otimes DB^2 + A(1-A)^2 \otimes BD^2 + (1-A) A (1-A) \otimes DBD + (1-A)^2 A \otimes D^2B + (1-A)^3 \otimes D^3,$$ i.e., just all combinations, leading to $N^3$ terms.

Now, what you can do is use the fact that $A$ and $1-A$ commute. This lets you write $$A^2\otimes B^2 + A(1-A) \otimes (BD+DB) + (1-A)^2 \otimes D^2.$$ for $N=2$ and $$A^3\otimes B^3 + A^2(1-A)\otimes (B^2D+BDB + DB^2) + A(1-A)^2 \otimes (BD^2 + DBD + D^2B)+ (1-A)^3 \otimes D^3$$ for $N=3$, respectively.

The answer I was originally posting only works if $B$ and $D$ were to commute. In this case you had $$[A \otimes B + (1-A) \otimes D)^N = \sum_{n=0}^N {N \choose n} \Big(A^n (1-A)^{N-n} \otimes B^n D^{N-n}\Big).$$ but that doesn't help you as yours don't.

You do have a sum over $A^n (1-A)^{N-n}$ but each term is Kronecker-ed with another sum that considers all ${N \choose n}$ arrangements of $n$ times the $B$ matrix and $N-n$ times the $D$ matrix.

$\endgroup$
4
  • $\begingroup$ There is no B on RHS? $\endgroup$ Commented Mar 11, 2020 at 12:58
  • $\begingroup$ @Rob: sorry, that was a typo in the equation. Fixed now. $\endgroup$ Commented Mar 12, 2020 at 9:06
  • $\begingroup$ The reason why it is not working for me is, I guess, the fact that $A$, $B$, $D$ are noncommuting matrices. Please share your opinion. $\endgroup$ Commented Jun 21, 2020 at 9:01
  • $\begingroup$ Oh my, you are indeed right. Without noticing I had implicitly assumed that $B$ and $D$ commute, which they don't. I have edited my reply. Sorry for the confusion I caused! $\endgroup$ Commented Jun 22, 2020 at 9:04

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.