Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Could you help me to solve this exercise?

Let $A$ be an $8\times8$ nilpotent matrix over $\mathbb{C}$ with $\mathrm{rank}(A)=5$ and $\mathrm{rank}(A^2)=2$. List all possible Jordan canonical forms for $A$ and show that knowledge of $\mathrm{rank}(A^3)$ would allow one to determine the Jordan canonical form.

This is what I have done: $A$ is nilpotent so the characteristic polynomial is $x^8$ and the minumum polynomial is $x^n$ with $3\leq n\leq8$. But now I don't know how to continue, any idea?

share|improve this question
1  
Start thinking about the Jordan blocks. For a Jordan block of size $m$ with eigenvalue $0$, what is the rank? What is the rank of its square? –  Dylan Moreland Jan 6 '12 at 4:16

2 Answers 2

up vote 4 down vote accepted

Since the matrix is nilpotent, as you note the characteristic polynomial is $x^8$, and the minimal polynomial is $x^i$ for some $i$, $1\leq i \leq 8$. It's not $x$, because $A$ is not zero, and it's not $x^2$, because $A^2$ is not zero either. So $3\leq i\leq 8$.

Let's go form there.

Because $A$ has rank $5$, the Rank-Nullity Theorem tells you the nullity is $3$; that means that the dimension of the eigenspace corresponding to $0$ is $3$; this tells you, by the basic theory of the Jordan Canonical Form, the number of Jordan blocks associated to zero. So the Jordan canonnical form of $A$ has exactly 3 blocks, all corresponding to zero.

The key observation to make is that if $J$ is $k\times k$ Jordan block associated to $0$, then it has rank $k-1$, and $J^s$ has rank $k-s$ ($1\leq s\leq k$); just notice that each time the $1$s "move up" one row until they disappear.

So: the original matrix consists of three blocks; say the sizes are $n_1$, $n_2$, and $n_3$, with $n_1\leq n_2\leq n_3$. We know the largest block has size at least $3$ (the degree of the minimal polynomial is the size of the largest block). And we know $n_1+n_2+n_3=8$.

We cannot have both $n_1$ and $n_2$ equal to $1$; if we did, the matrix would be a $6\times 6$ block and the rest would be zeros; after squaring, the matrix would have rank $4$ (by the observation above), which is not the case. So $n_2\geq 2$. We cannot have $n_1=1$ and $n_2=2$, because then $n_3=5$, and after squaring we would get a matrix of rank $3$, which is not the case.

In fact, if the blocks are of size $1+n_2+n_3$, with $n_3\geq n_2\geq 2$, then the rank of the square would be $n_2+n_3-4 = 3$, which is not the case. So $2\leq n_1$.

That will give the right rank, since then we will have that the rank of $A^2$ is $(n_1-2)+(n_2-2)+(n_3-2) = n_1+n_2+n_3-6 = 8-6 = 2$.

So we must have $2\leq n_1\leq n_2\leq n_3$, $n_1+n_2+n_3=8$; any of these combinations will be a possible Jordan canonical form.

After finding all possibilities, verify that in each case you get a different value for $\mathrm{rank}(A^3)$, so that will show the last clause of the problem is true.

share|improve this answer

Since Jordan blocks correspond to a direct sum decomposition of the vector space, the rank can be computed for each summand (block) and the added up. Now one easily sees a nilpotent Jordan block of size $d$ contributes $d-i$ to $\def\rk{\operatorname{rank}}\rk(A^i)$ if $d\geq i$, and contributes nothing otherwise. And therefore it contributes to $\rk(A^{i-1})-\rk(A^i)$ either $1$, when $d\geq i$, or $0$, when $d>i$.

It follows that as a function of $i$ the non-negative number $\rk(A^i)$ is (weakly) decreasing, but in such a way that the differences $\rk(A^{i-1})-\rk(A^i)$ are also (non-negative and) weakly decreasing. And that the latter numbers count the number of Jordan blocks of size at least$~i$; this will in particular allow deducing the sizes of the Jordan blocks (of a nilpotent matrix$~A$) when all values of $\rk(A^i)$ are known up to the point where they become$~0$.

In the example $(\rk(A^i))_{i=0,1,2}=(8,5,2)$, and one must have $\rk(A^3)\in\{0,1\}$ (because $\rk(A^{i+1})=\rk(A^i)$ is only possible if $\rk(A^i)=0$), and $\rk(A^4)=0$ is forced in either case. The corresponding sequences $(\rk(A^{i-1})-\rk(A^i))_{i=1,2,3,4}$ are $(3,3,2,0)$ and $(3,3,1,1)$, giving as Jordan types $(3,3,2)$ respectively $(4,2,2)$.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.