4
$\begingroup$

Let $A$ and $B$ be matrices such that $B^2+ AB + 2I = 0$, where I denotes the identity matrix. Which of the following matrices must be nonsingular?

(A) $A + 2I$

(B) $B$

(C) $B + 2I$

(D) $A$

I tried using a few tricks assuming each option to be nonsingular and then coming to the given form but to no avail. Any hint is appreciated.

$\endgroup$
  • 1
    $\begingroup$ I edited your post a little to make the $\LaTeX$ work etc. Your can see how I did it by clicking the "edit" button. Cheers! $\endgroup$ – Robert Lewis Oct 5 '18 at 18:29
4
$\begingroup$

Since $$\det (B+A)\cdot \det (B) = \det (B^2+AB) = \det (-2I) =-2$$ we see that $\det (B)\ne 0$ so $B$ is invertibile.

$\endgroup$
3
$\begingroup$

We have

$B^2 + AB + 2I = 0, \tag 1$

so

$(B + A)B = B^2 + AB = -2I, \tag 2$

or

$\left ( -\dfrac{1}{2}(B + A) \right )B = I, \tag 3$

that is,

$B^{-1} = -\dfrac{1}{2}(B + A), \tag 4$

and the correct answer is (B).

$\endgroup$
  • $\begingroup$ @StubbornAtom: Oops! Typo; corrected. Thanks. Cheers! $\endgroup$ – Robert Lewis Oct 5 '18 at 19:10
  • 1
    $\begingroup$ @StubbornAtom: damn! another blunder! Now how does it look? That's what I get for getting on MSE before coffee! Cheers! $\endgroup$ – Robert Lewis Oct 5 '18 at 19:23
  • 1
    $\begingroup$ How do you exclude the other cases? $\endgroup$ – egreg Oct 5 '18 at 20:48
  • $\begingroup$ @egreg I didn't. I guess I will have to cite your answer! (+1) $\endgroup$ – Robert Lewis Oct 5 '18 at 22:05
3
$\begingroup$

I'm assuming you're over the real numbers.

Suppose $B$ is singular and take $v\ne0$ such that $Bv=0$; then $$ 0=(B^2+AB+2I)v=B^2v+ABv+2Iv=2v, $$ a contradiction. Therefore $B$ is nonsingular.

In order to exclude the other cases, let's see whether $A$, $A+2I$ or $B+2I$ can be the zero matrix.

The matrix $A$ can be the zero matrix, because $$ \begin{bmatrix} 0 & 2 \\ -1 & 0 \end{bmatrix}^2=-2I $$

Also $A+2I$ can be the zero matrix: the identity to satisfy would be $$ B^2-2B+2I=0 $$ and the matrix \begin{bmatrix} 2 & -1 \\ 2 & 0 \end{bmatrix} satisfies it. I looked for a matrix with trace $2$ and determinant $2$, so Hamilton-Cayley solves the problem.

Can $B+2I=0$?


Just for completeness, the same computations can be performed over any field, with the same result, provided the field doesn't have characteristic $2$. Indeed, the initial contradiction we found is definitely not such when $2=0$.

In this case the identity becomes $B^2+AB=0$ and we see that $B$ can as well be the zero matrix, with $A$ any matrix at all. If $B$ is invertible, then $A=B$.

There is no case distinction for $A+2I=A$ and $B+2I=B$. Thus in the case of characteristic $2$, none of the four conditions is forced.

$\endgroup$

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.