Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am having difficulty proving the following:

Let $V$ be a finite dimensional vector space over the complex numbers, and $T\colon V\to V$ a linear transformation. Prove that $T$ has a nontrivial (not equal to $V$ or to the zero subspace) hyperinvariant subspace iff $T$ is not a scalar transformation (that is, $T$ is not of the form $aI$).

I haven't encountered the term "hyperinvariant" before this question, so I'll append the definition given for that as well:

A subspace $W$ is called $T$-hyperinvariant if for every linear transformation $S\colon V\to V$ that satisfies $ST=TS$, $W$ is $S$-invariant.

This is a question from my course, but it's not homework (it's from a former exam), so feel free to give complete answers. Hints are also welcome, of course.

Thanks in advance!

share|cite|improve this question
Note on terminology: in English, we say "over the complex number", rather than "above the complex numbers". – Arturo Magidin Sep 19 '11 at 16:44
Ah, thanks. I'm not very familiar with English terminology, as you can see (here we sometimes call linear transformations "linear copies"!). – iroiroaru Sep 19 '11 at 16:48
@iroioaru: I fixed the title; rather than "invariant hypersubspaces", you are talking about "hyperinvariant subspaces". – Arturo Magidin Sep 19 '11 at 18:19
up vote 1 down vote accepted

If $\dim(V)\leq 1$, then every linear transformation is a scalar multiple of the identity, and $T$ has no nontrivial hyperinvariant subspace; so we may assume $\dim(V)\gt 1$.

If $T$ is a scalar transformation, then it commutes with every linear transformation; given any proper subspace $W$ of $V$, there is a linear transformation such that $S(W)\neq W$: indeed, let $\{w_1,\ldots,w_k\}$ be a basis for $W$, and let $v_{k+1},\ldots,v_n$ be vectors in $V$ such that $\{w_1,\ldots,w_k,v_{k+1},\ldots,v_n\}$ is a basis for $V$. Because we are assuming that $W$ is not the zero vector space, it follows that $k\geq 1$; because we are assuming that $W\neq V$, it follows that $k\neq n$. In particular, $w_1$ is a nonzero vector in $W$, and $v_n$ is a nonzero vector not in $W$. Now define $S\colon V\to V$ by defining it on the basis $w_1,\ldots,w_k,v_{k+1},\ldots,v_n$ as follows: $$\begin{align*} S(w_i) &= w_i&&\text{if }2\leq i\leq k,\\ S(v_j) &= v_j&&\text{if }k+1\leq j\lt n,\\ S(w_1)&=v_n,\\ S(v_n)&=w_1, \end{align*}$$ and extending linearly. Then $S(W)$ is not contained in $W$, since $w_1\in W$, but $S(w_1)=v_n\notin W$. Since $T=aI$, then $ST=TS$; so $W$ is not hyperinvariant. This holds for any nontrivial subspace of $V$, so we conclude that $T$ has no nontrivial hyperinvariant subspaces. This proves that if $T$ has a nontrivial invariant subspace, then it is not a scalar transformation.

Conversely, suppose that $T$ is not a scalar transformation; let $\lambda$ be an eigenvector of $T$, and let $W$ be the eigenspace of $T$ associated to $\lambda$. Then $W$ is a nontrivial subspace (it's not $0$ because $\lambda$ is an eigenvalue, and it's not $V$ because $T\neq \lambda I$). I claim that $W$ is hyperinvariant.

Indeed, let $S\colon V\to V$ be a linear transformation such that $TS=ST$, and let $w\in W$. Then $$\lambda S(w) = S(\lambda w) = S(T(w)) = T(S(w)),$$ so $S(w)$ lies in the eigenspace of $T$ corresponding to $\lambda$, which is none other than $W$. So $S(W)\subseteq W$, proving that $W$ is $S$-invariant, as claimed. $\Box$

share|cite|improve this answer
Thanks! Most of this is a lot simpler than I thought. – iroiroaru Sep 19 '11 at 16:58
@iroiroaru: Added. – Arturo Magidin Sep 19 '11 at 17:05
@iroiroaru: Note: you should not feel pressured to accept an answer before being sure about it. You had asked a question that I was in the process of explaining when you accepted. In any case, it's often a good idea to wait some time before accepting an answer, as that invites the possibility of other people posting answers that are different from the ones already posted. More proofs = better. Having an answer accepted diminishes the number of replies, in general. – Arturo Magidin Sep 19 '11 at 17:07

Your Answer


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.