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

My textbook asks this: Suppose that $K$ is a finite field with $k$ elements, and that $V$ is an $r$-dimensional vector space over $K$. Show that if $V = \bigcup_{i=1}^n U_i$, where $U_1,\dotsc,U_n$ are proper subspaces of $V$, then $n\geq (k^r - 1)/(k-1)$.

Struggling to prove this for a while, I did some googling and found this:

which claims to show $n = k+1$ is possible, a result which is independent of the dimension of $V$. Which is correct?

share|cite|improve this question
Possibly related(?): – Zev Chonoles Jun 23 '13 at 8:38
@ZevChonoles In summary: your link shows $n\geq k+1$, my link shows that this is sharp, but my textbook says $n\geq (k^r - 1)/(k-1)$. So the job now is to verify what is proved in my link is correct... – user71815 Jun 23 '13 at 8:46
@user71815 : you sure that it is not $n\leq(k^r-1)/(k-1)$ instead? If you take a set of representatives of $\Bbb P(V)$ and for each of them consider the line spanned by it, the set of these lines covers $V$ and there are exactly $(k^r-1)/(k-1)$ of them. – Andrea Mori Jun 23 '13 at 9:13
Your textbook is wrong except in the case $r=2$. The general case reduces to the case $r=2$, because we can "add the remaining stuff" to all the subspaces $U_i$, i.e. we can assume that they all share an $(r-2)$ dimensional subspace $W$, and work with $V/W$. – Jyrki Lahtonen Jun 23 '13 at 10:44
up vote 2 down vote accepted

Consider the linear equations $$\tag{eq} \begin{cases} x_{1} - \lambda_{j} x_{r} = 0\\ x_{r} = 0. \end{cases} $$ Here $K = \{\lambda_{1}, \dots, \lambda_{k} \}$, so there are $k+1$ equations in (eq), which define $k+1$ maximal subspaces $W_{j}$ of $V$.

Let $v \in V$. If the $r$-th component $v_{r}$ of $v$ is zero, then $v \in W_{k+1}$. If $v_{r} \ne 0$, then $v_{1} = \lambda_{j} v_{r}$ for $\lambda_{j} = v_{1} v_{r}^{-1}$, so $v \in W_{j}$. Thus $V = \bigcup_{j=1}^{k+1} W_{j}$.

So it can indeed be done with $k+1$ subspaces. The answer mentioned above proves that you cannot do better than this.

share|cite|improve this answer
I will check your article, but note that my argument is an explicit version of that of @JyrkiLahtonen in his comment above. – Andreas Caranti Jun 24 '13 at 7:24
@user71815, also, the main point of your article is that this decomposition is unique up to automorphisms. – Andreas Caranti Jun 24 '13 at 8:45
@user71815, and then my proof is essentially the same as in your article. – Andreas Caranti Jun 24 '13 at 9:36

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.