12
$\begingroup$

Let $A$ be a $n \times n$ real matrix such that $$ A + A^t = 2I, $$ where $I$ is the $n \times n $ identity matrix.

Prove that $\det(A) \geq 1$.

It is obvious that tr$(A) = n$. Furthermore, we also have that $$A - 2I = -A^t, $$ so we get that $$\det(A-2I) = (-1)^n \cdot \det(A).$$

Now, we let $\lambda_1, \lambda_2, \cdots, \lambda_n \in \mathbb{C}$ be the eigenvalues of $A$, so we get that $$(\lambda_1 - 2)(\lambda_2 - 2) \cdots (\lambda_n - 2) = (-1)^n\lambda_1\lambda_2 \cdots \lambda_n, $$ but I couldn't derive anything about the product $\lambda_1\lambda_2 \cdots \lambda_n = \det(A)$ (the only known thing is $\lambda_1 + \cdots + \lambda_n = n$).

Also, I tried the same approach for $2 \times 2$ and $3 \times 3 $ matrices, but it didn't lead to anything (especially since the eigenvalues can be complex numbers).

$\endgroup$
14
$\begingroup$

Note that $A - I$ is a real skew-symmetric matrix, so all of its eigenvalues are pure imaginary (or zero) and appear in complex conjugate pairs. Therefore the eigenvalues of $A$ are either $1$ or of the form $1 + c i, 1-c i$ for some $c\in\mathbb{R}$. Note that $(1+ci)(1-ci) = 1+c^2 \geq 1$. Therefore

$$\det(A) = \prod (1 + c^2) \geq 1$$

$\endgroup$
  • $\begingroup$ Thanks! That's exactly what I just though about now, once I tried to write out the matrices and compute the sum. $\endgroup$ – C_M Dec 24 '18 at 18:44
8
$\begingroup$

Let $A = I+B.$ One can immediately notice that $B$ is skew-symmetric or $B = -B^T.$ This means the eigenvalues of $A$ are precisely $1+\lambda_j$, where $\lambda_j$ are the eigenvalues of $B.$ Now, it's well-known that the eigenvalues of a skew-symmetric matrix comes in pairs $\pm i\lambda_j.$ This means that: $$\det A = \prod(1+\lambda^2_j)\geq 1.$$ You are encouraged to fill in the details separating the cases where $n$ is odd or even.

$\endgroup$

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.