I have gone through two proofs to show that matrices of rank $k$, where $0 \leq k \leq \min(m,n)$ form a submanifold of the set $ M(m \times n ,\mathbb{R}) $. Both proofs involve writing down an arbitrary matrix in the set considered, as :
$$ X = \begin{bmatrix} A & B \\ C & D \\ \end{bmatrix}$$
where $A$ is assumed to be of a minor of order $ k \times k$ that has non-zero determinant,(which I understand) and then they ask us to consider either the matrix :
$$ M_X= \begin{bmatrix} A^{-1} & -A^{-1}B \\ 0 & I_{n-k} \\ \end{bmatrix}$$
or alternatively:
$$M_X= \begin{bmatrix} I & -A^{-1}B\\ 0 & I \end{bmatrix}$$
This is followed by multiplication of $X$ with $M_X$. I understand what follows next. But this particular choice of matrix seemed to be picked out of the air as I viewed it.Could anyone please help motivate this idea??
Ref: The proof is found in J.Lee's Introduction to Smooth Manifolds, pg 117, ex 5.30.