Generating a vector space via a T-cyclic subspace

[Bipolarity]

I've been thinking about a problem I made up. The solution may be trivial or very difficult as I have not given too much thought to it, but I can't think of an answer of the top of my head.

Let $T:V → V$ be a linear operator on a finite-dimensional vector space $V$. Does there exist a vector $v \in V$ for which the T-cyclic subspace of $V$ generated by $v$ is $V$? This is certainly not true in general, since if $T$ is the zero transformation and $V$ has dimension greater than 1 then no T-cyclic subspace will equal $V$.

But what about for an arbitrary linear map?

BiP

 

[brmath]

Let's just look at a 2-dim space. Let T = <br /> \begin{pmatrix}<br /> 0 &amp; 1 \\<br /> 1 &amp; 0\\<br /> <br /> \end{pmatrix}<br />
and v = (1,0). Then Tv = (0,1). Clearly v and Tv are a basis for the entire space.

 

[Bipolarity]

Let's just look at a 2-dim space. Let T = <br /> \begin{pmatrix}<br /> 0 &amp; 1 \\<br /> 1 &amp; 0\\<br /> <br /> \end{pmatrix}<br />
and v = (1,0). Then Tv = (0,1). Clearly v and Tv are a basis for the entire space.

How can we generalize this for an arbitrary linear map?

BiP

 

[brmath]

Assuming we are in a finite dimensional vector space, of dimension n, you need to pick a T such that T^n= T but if 1 < m < n T^m \neq T.
If you do have T^m = T for 1 < m < n then the T-cyclic subspace has only dimension m, because that is all the independent vectors you can generate with powers of T.

Now there is bound to be some specific vector v such that T^mv = Tv even though T^m \neq T. That vector is in fact any eigenvector of T^m - T. You can't generate the entire space V using that kind of vector v. However, there is bound to be a vector u which is not an eigenvector of T^m - T for any 1 < m < n. You can use u to generate the entire space.

Does every n dimensional vector space have a linear operator which is cyclic of degree n? Yes. Start with my 2 dimensional example and see if you can find something similar in 3 dimensions. Once you get that far, you'll see how to generate such a T in n dimensions.