0
$\begingroup$

If $A \in M_{n}$ is nonderogatory, why is rank $A \geq n-1 ?$

$\endgroup$
4
  • 1
    $\begingroup$ @coffeemath According to Wikipedia, a matrix with minimal polynomial equal to the characteristic polynomial. $\endgroup$ – Arthur Jul 16 '20 at 12:59
  • $\begingroup$ @Arthur Thanks. Just saw that... deleted my c omment. $\endgroup$ – coffeemath Jul 16 '20 at 13:00
  • $\begingroup$ If nullity is at least 2, that means in the JNF there are at least two blocks with eigenvalue 0, hence min poly and char poly are not the same. $\endgroup$ – user10354138 Jul 16 '20 at 13:14
  • $\begingroup$ For more on non-derogatory matrices, see also this post. $\endgroup$ – Ben Grossmann Jul 16 '20 at 13:29
2
$\begingroup$

Hint: If $A$ is non-derogatory, then all eigenvalues of $A$ must have geometric multiplicity $1$.

$\endgroup$
1
  • $\begingroup$ I know that. Could you give a detailed proof? $\endgroup$ – Ben Jul 16 '20 at 22:26
1
$\begingroup$

Consider the Jordan canonical form of $A$, we have $A = SJ_AS^{-1}$.

Assume that $J_A=J_{n_{1}}\left(\lambda_{1}\right) \oplus \cdots \oplus J_{n_{k}}\left(\lambda_{k}\right)$.

Since the geometric multiplicity of a given eigenvalue of a Jordan matrix is equal to the number of Jordan blocks corresponding to that eigenvalue, a matrix is nonderogatory if and only if each of its distinct eigenvalues corresponds to exactly one block in its Jordan canonical form. Thus, $\lambda_1, \cdots , \lambda_k$ are mutually distinct.

Case 1: $A$ has $0$ eigenvalue. Then by the rank-nullity theorem, we have $n - rank(A-0I) = 1$, we have $rank A = n-1\ge n-1$

Case 2: $A$ doesn't have $0$ eigenvalue, i.e., $\lambda_i \ne 0, \forall i = 1,\dots, k$. Thus, the rank of A is the number of nonzero rows of $J_A$, i.e., n

To summary, the rank of a nonderogatory matrix is at least its order $-1$.

$\endgroup$

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.