Let $V$ be a vector space of dimension $\ge 3$ over an infinite field $k$. Let $W_1,...,W_n$ be any $1$-dimensional subspaces of $V$. Then must there exist a two dimensional subspace $W$ of $V$ such that $W \cap (\cup_{i=1}^n W_i)=\{0\}$ ?



Start with any two-dimansional subspace $U$. We can pick a basis $(v_1,v_2)$ of $U$ such that $v_1$ is in none of the $W_i$: We have infinitely many choices to replace $v_1$ with $v_1+\alpha v_2$, $\alpha\in k$, and each $W_i$ prohibits at most one value of $\alpha$.

Now pick $v_3\notin U$. and consider the vector spaces $U_\beta=\langle v_1,v_2+\beta v_3\rangle$, $\beta\in k$. As $U_\beta\cap U_{\beta'}=\langle v_1\rangle$, each $W_i$ is contained in at most one $U_\beta$. Hence by a suitable choice of $\beta$, we find that $$ U_\beta\cap \bigcup_{i=1}^nW_i=\{0\}.$$

Remark: Apparently the claim also hold for finite fields $k$, as long as $n<|k|$. And vice versa, it also works for infinite $n$, as long as $k$ is "more infinite" (e.g., for $k=\Bbb R$, we can have countably many $W_i$).


A nonconstructive solution:

Convention, in the following argument, $W$ cross $W_i$ at some point means the two subspaces have a common nonzero vector.

Suppose for the sake of contradiction that, all 2-dim subspace crosses at least a point on one $W_i$.

Since there is finitely many $W_i$, there must be one $W_k$ that crosses infinitely many 2-dim subspaces, among which there are infinitely many that do not contains any $W_i$ (otherwise there are only finitely many 2-dim subspaces whose orthogonal complement is not contains any $W_i$, this can't be true since finitely many 1-dim subspaces can't cover a 2-dim space, nor to say a 3-d space).

But this can't be true that (the statement before the previous long brackets: there are infinitely many subspaces that do not contains any $W_i$(among which is $W_k$) but crosses a certain $W_k$, in fact, there are none).

A contradiction.

This proof is quite strange, I admit.


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.