6
$\begingroup$

Let $A_1, A_2\in M_n(\mathbb{R})$ two symmetric matrices s.t. $det(A_1^2+A_2^2)=0$.

Show that $det(A_1\cdot B_1 +A_2\cdot B_2)=0$ for every $B_1, B_2\in M_n(\mathbb{R})$.

My idea:

I consider the matrix C :\begin{bmatrix} A_1 & A_2 \\ B_1^t & B_2^t \end{bmatrix}

$det(C\cdot C^t)\geq 0\Rightarrow det(\begin{bmatrix} A_1^2 +A_2^2 & D \\ D^t & E\\ \end{bmatrix})\geq 0 $ where $D=A_1B_1+A_2B_2$.

I tried to expand the determinant with Laplace Rule. I am not sure if $ det(\begin{bmatrix} A_1^2 +A_2^2 & D \\ D^t & E\\ \end{bmatrix}) = -det(D\cdot D^t)$.

In this way I would get $det(D)=0$.

$\endgroup$
  • $\begingroup$ I have a question; purely for my own understanding & writing better answers: Why did you select a later and longer answer that does essentially the same thing? $\endgroup$ – copper.hat Mar 21 '18 at 19:18
  • $\begingroup$ @copper.hat Because he wrote with details. Also I vote your solution. $\endgroup$ – Problemsolving Mar 21 '18 at 19:22
  • $\begingroup$ Thanks for responding. There is exactly the same level of detail. This is why I am curious. More characters does not translate to more detail. I am not trying to change your mind, just to understand why you think there is more detail? $\endgroup$ – copper.hat Mar 21 '18 at 19:26
  • $\begingroup$ @copper.hat His solutions, which is the same as yours, looks easier for me, even if it is longer. $\endgroup$ – Problemsolving Mar 21 '18 at 19:34
  • $\begingroup$ Thanks. Please switch the accepted answer back. That was not my intention. I am not concerned about the points. I am curious about the reasoning, and you have provided some insight, thanks. $\endgroup$ – copper.hat Mar 21 '18 at 19:40
7
$\begingroup$

If $\det (A_1^2+A_2^2) = 0$, then $x^TA_1^2+x^TA_2^2=0$ for some $x\neq 0$ and so $\|x^TA_1\|^2 + \|x^TA_2\|^2 = 0$.

Since $x^TA_1=x^TA_2 = 0$, we see that $x^T (A_1 B_1+A_2 B_2) =0$ and so $\det (A_1 B_1+A_2 B_2) =0$

$\endgroup$
4
$\begingroup$

Let $\det(A^2_1+A^2_2)=0$ then there exists some vector $x\neq 0$ such that $(A^2_1+A^2_2)x=0$ this implies $$0=\langle (A_1^2+A^2_2)x,x\rangle=\langle A^2_1x,x\rangle+\langle A^2_2x,x\rangle=\langle A_1 x,A^T_1x\rangle+\langle A_2x,A^T_2x\rangle$$ Since $A_1, A_2$ are symmetric then $A_1=A_1^T, A_2=A^T_2$. Therefore $$0=\langle A_1 x,A^T_1x\rangle+\langle A_2x,A^T_2x\rangle=\langle A_1 x,A_1x\rangle+\langle A_2x,A_2x\rangle=||A_1x||^2+||A_2x||^2$$ This holds only if $A_1x=A_2x=0$ in which case we obtain that $\det A_1=\det A_2=0$. Now let $C:=A_1B_1+A_2B_2$ then $C^T=B_1^TA^T_1+B^T_2A^T_2=B_1^TA_1+B^T_2A_2$. This then gives $$C^T x=(B_1^TA_1+B^T_2A_2) x=B_1^TA_1x+B^T_2A_2x=B_1^T0+B^T_20=0$$ Therefore $\det C^T=0$ but $\det C=\det C^T$ for any matrix $C$. The result follows $$\det C=\det (A_1B_1+A_2B_2)=0$$

$\endgroup$

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.