Re: How to prove this matrix identity?



On Mon, 12 May 2008 18:31:04 -0700 (PDT), Rotwang
<sg552@xxxxxxxxxxxxx> wrote:

Suppose that A is an invertible n x n matrix, D is an m x m matrix, B
is an n x m matrix and C is an m x n matrix. If the (n + m) x (n + m)
matrix M is given by

M = / A B \
\ C D /

then apparently

Det M = Det A * Det (D - C A^{-1} B).

Can anybody tell me how this is proved?

It's rather messy to write out, but one way to do it is first to
reduce to the case A = I:

/ A B \ / A^{-1} 0 \ = / I B \
\ C D / \ 0 I / \ CA^{-1} D /

therefore

(det M)*(det A)^{-1} = det(D - (CA^{-1})B)

- or else premultiply by same matrix, to get det(D - C(A^{-1}B)).

If A = I, we can reduce all the elements of B to 0 by subtracting
scalar multiples of the first n columns from the last m columns,
and this leaves I and C unchanged, and transforms D into D - CB.

(At least, that's how it looks in a few examples, but as I said,
it's messy to write out! ... and it's past my bedtime.)

--
Angus Rodgers
(twirlip@ eats spam; reply to angusrod@)
Contains mild peril
.



Relevant Pages

  • Re: SCWC 41: comments
    ... peculiar sight.) ... Angus Rodgers ... (twirlip@ eats spam; reply to angusrod@) ...
    (rec.puzzles.crosswords)
  • Re: Magpie A Grade Puzzle Isnt Easy!
    ... Angus Rodgers ... (twirlip@ eats spam; reply to angusrod@) ... care, and will revert to my usual style, which now that I think about ...
    (rec.puzzles.crosswords)
  • Re: SCWC 43
    ... Angus Rodgers ... (twirlip@ eats spam; reply to angusrod@) ...
    (rec.puzzles.crosswords)
  • Re: More with Quadratic Diophantine Theorem
    ... Angus Rodgers ... (twirlip@ eats spam; reply to angusrod@) ...
    (sci.math)
  • Re: Mini-comp
    ... engaging auto destruct sequence ... Angus Rodgers ... (twirlip@ eats spam; reply to angusrod@) ...
    (rec.puzzles.crosswords)