10
$\begingroup$

Suppose that $A, B$ are idempotent matrices ($A^2=A$), such that $A + B$ is idempotent, prove that $AB = BA = 0$

$\endgroup$
7
  • 1
    $\begingroup$ Are you sure you don't want to prove $AB+BA=0$ ? $\endgroup$
    – Belgi
    Jan 26, 2013 at 23:49
  • 1
    $\begingroup$ No it is definitely AB=BA=0 $\endgroup$
    – Chance
    Jan 26, 2013 at 23:51
  • 2
    $\begingroup$ Probably there is a typo on your book/list/etc... $\endgroup$
    – Sigur
    Jan 26, 2013 at 23:57
  • $\begingroup$ @Belgi Thanks.. $\endgroup$
    – Git Gud
    Jan 26, 2013 at 23:58
  • $\begingroup$ @GitGud - No problem, I was doing the same thing and I thought it would work $\endgroup$
    – Belgi
    Jan 27, 2013 at 0:00

2 Answers 2

15
$\begingroup$

This is true in every ring where you can divide by $2$. So it is true in particular for matrices.

We have $(A+B)^2=A^2+AB+BA+B^2=A+AB+BA+B$. Since $A+B$ is idempotent, it follows that $$AB+BA=0.$$ We call this equation (E).

Right-multiply (E) by $B$. We find $AB+BAB=0$, hence $AB=-BAB$.

Left-multiply (E) by $B$. This yields $BAB+BA=0$, hence $BA=-BAB$.

Equating the last two equations, we find $AB=BA$.

Now $AB+BA=0$ clearly yields $2AB=2BA=0$. Dividing by $2$, we get $$ AB=BA=0. $$

$\endgroup$
2
  • $\begingroup$ Very nice! I thought in this direction when I saw I coldn't get the result, but I figured it was a typo and I tried to find a trivial example so I tooked $B=A$ and then I saw that in $F_2$ I get such an example. Did you know this before, or just proved it ? (+1) $\endgroup$
    – Belgi
    Jan 27, 2013 at 0:17
  • $\begingroup$ Could you give me such example? I'm not able of find one $\endgroup$ Apr 24, 2018 at 19:25
9
$\begingroup$

This is incorrect.

For example take $F=\mathbb{F}_{2}$ the field with two elements and $A=B=I$ over $F$.

$A,B$ are clearly idempotent and $$A+B=I+I=2I=0$$ hence $$(A+B)^{2}=0^{2}=0=A+B$$ is also idempotent.

But $$AB=BA=I^{2}=I\neq0$$

From your assumption you can only get $AB+BA=0$.

$\endgroup$
5
  • $\begingroup$ Yes I can only seem to get $AB+BA=0$ from this assumption. Could possibly be a typo. $\endgroup$
    – Chance
    Jan 27, 2013 at 0:02
  • 1
    $\begingroup$ @Chance: It is true for square matrices over a field in characteristic distinct from $2$. See my answer. $\endgroup$
    – Julien
    Jan 27, 2013 at 0:13
  • $\begingroup$ Can the downvoter please explain ? maybe I can improve my answer. $\endgroup$
    – Belgi
    Jan 27, 2013 at 0:17
  • $\begingroup$ If you combine your answer with Julien's, you can arrive at $AB=BA=0$ $\endgroup$
    – Chance
    Jan 27, 2013 at 0:21
  • $\begingroup$ @Chance - Not in general, You have to assume $1+1\neq 0$ $\endgroup$
    – Belgi
    Jan 27, 2013 at 0:22

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.