Let $A$ and $B$ be $n \times n$ matrices over some field with $A$ nilpotent. Now let $c_1,\ldots,c_{n+1}$ be $n+1$ distinct scalars such that $A+c_i B$ is nilpotent for all $i=1, \ldots,n+1$. Then how can I show that $B$ is also nilpotent?



1 Answer 1


Let us define the matrix valued polynomial $$ p(t) = (A+tB)^n = \sum_{k=0}^n C_k t^k. $$ We see that each $(i,j)$-th component $p(t)_{ij}$ of $p(t)$ is a polynomial in $t$ of degree at most $n$. By the assumption, $p(t)_{ij}=0$ has $n+1$ distinct roots $c_1,c_2,\ldots, c_{n+1}$ on the underlying field $\mathbb{F}$. (This is why: if $n\times n$ matrix $T$ is nilpotent, the it holds $T^k=O$ for some $k\le n$ and hence $T^n=O$. Thus, by the assumption, $p(c_k) = O$ for all $k=1,2,\ldots, n+1$.) This means $p(t)_{ij} \equiv 0$ and it follows that $p(t) \equiv O$. Since the coefficient $C_n$ of $t^n$ is $B^n$, it follows that $B^n=O$.

  • $\begingroup$ Can you be more explicit. I don't see why $p(t)_{ij}=0$ has $n+1$ distinct roots. PS I saw that your answer was upvoted and then downvoted. I did not downvote. I am just seeking a clarification. $\endgroup$ Jan 7, 2019 at 8:07
  • $\begingroup$ Do we even need the fact that $A$ is nilpotent tho? $\endgroup$
    – EuxhenH
    Jan 7, 2019 at 8:12
  • $\begingroup$ @EuxhenH Maybe we can assume that one of $c_k$ is 0. $\endgroup$ Jan 7, 2019 at 8:13
  • $\begingroup$ @Song That would mean that $A$ is nilpotent as a consequence, but we still don't need it in the problem statement. $\endgroup$
    – EuxhenH
    Jan 7, 2019 at 8:16
  • $\begingroup$ @KaviRamaMurthy As is always, any productive comments are welcome! and I'm totally okay with the downvote. $\endgroup$ Jan 7, 2019 at 8:16

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.