Let $A$ be a square matrix.

a) Show that there always exists a square matrix B such that Ker $B =$ Im $A$ and Ker$A =$ Im$B$.

Here is my approach:

Let $A:\ V \rightarrow P$. Denote the basis of Ker$A$ by $u_{1}, \dots , u_{r}$ where $r\leq n$, $n$ is the size of $A$. Let $B$ be a matrix with $u_{1}, u_{2}, \dots , u_{r}$ as the first $r$ columns and the rest columns from $r$ to $n$ are some linear combinations of $u_{i}$. Thus the first r columns of $B$ span a subspace of $V$, this is by definition the image of $B, \ \rightarrow$ Im$B=$Ker$A$. From here its blank. Im not even sure if my first approach is correct.

I will be grateful for any hint. I want to prove this with matrices since next question is to prove that IF A is normal then B is normal aswell.


It is probably better to think in terms of linear maps instead of matrices. So you have a linear map $a$ on a vector space $V$. Let $v_{1}, \dots , v_{n}$ be a basis of $V$ such that $v_{1}, \dots, v_{k}$ are a basis of $\ker(a)$, so that $f(v_{k+1}), \dots , f(v_{n})$ is a basis of the image of $a$. Extend the $f(v_{i})$ to a basis $w_{1}, \dots, w_{k}, f(v_{k+1}), \dots , f(v_{n})$ of $V$, and define a linear map $b$ which is zero on the $f(v_{i})$ and maps each $w_{j}$ to $v_{j}$.

  • $\begingroup$ The proof is even is way more general: Let $U$ and $W$ be any subspaces of $V$, that satisfy the dimension formula, i.e. their dimensions add up to $\dim V$. Then there exists an endomorphism with kernel and image being $U$ and $W$ respectively. $\endgroup$ – MooS Dec 8 '15 at 21:05
  • $\begingroup$ Yeah I follow. But one has to prove that this map exists aswell or is it evident? $\endgroup$ – Olba12 Dec 8 '15 at 21:08
  • $\begingroup$ The point is that one can define a linear map by assigning to it arbitrary values on a basis. $\endgroup$ – Andreas Caranti Dec 8 '15 at 21:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.