4
$\begingroup$

I am currently working on some problems related to Grassmann manifolds and eventually come to the following question.

Let $S$ be a subset of $G_1(\mathbb{R}^n)$ such that any element of $G_{n-1}:=G_{n-1}(\mathbb{R}^n)$ contains a line in $S.$ Is it true that there exists a hyperplane in $G_{n-1}$ that is completely covered by $S$ (that is, every line passing through the origin contained in that hyperplane belongs to $S$)?

Has this problem ever been discussed before? So far I could not come up with any good idea for it, at least not by "geometric arguments".

$\endgroup$
3
  • $\begingroup$ In the sentence in brackets, did you mean "every line contained in that hyperplane belongs to $S$"? $\endgroup$ Commented Mar 15, 2018 at 15:00
  • $\begingroup$ I changed the statement. Thank you. $\endgroup$
    – user160393
    Commented Mar 15, 2018 at 15:06
  • $\begingroup$ If I get the question correctly, $S:=G_1(V)\subset G_1(\mathbb{R}^n)$ for a 2-dimensional linear subspace $V\subset \mathbb{R}^n$ has this property, because any hyperplane and $V$ share at least a line. $\endgroup$ Commented Apr 15, 2018 at 0:06

3 Answers 3

1
$\begingroup$

The answer is in the negative and a construction of a counterexample is provided here. I wrote this argument quickly, but I hope it is correct. Let me know if it is not. We will construct a family of lines passing through the origin in $\mathbb{R}^3$ so that:

1. Every plane passing through the origin contains at least one of the lines from the family,

2. The union of the lines does not contain any plane passing through the origin.

The shortest way to do it is to use transfinite induction. While this is not an effective construction, it shows that you cannot prove your claim and this is what is important here. The argument used here is pretty similar to the one used in this post.

Order all planes by the initial ordinal of $2^{\aleph_0}$. Denote the planes by $P_\alpha$, where $\alpha$ is an ordinal. We can construct a corresponding family of lines $\ell_\alpha$ so that

(a) $\ell_\alpha\subset P_\alpha$,

(b) $\ell_\alpha\cap P_\beta=\{ 0\}$ for $\beta<\alpha$.

Then (1) is obviously satisfied by (a). By (b), each plane $P_\beta$ can contain only lines $\ell_\alpha$, $\alpha\leq\beta$. The cardinality of such lines is less than $2^{\aleph_0}$ which is the cardinality of all lines in a plane so (2) follows.

The construction of the family of lines $\ell_\alpha$ is pretty standard. Suppose we already have lines $\ell_\beta$, $\beta<\alpha$ satisfying (1) and (2). Consider the plane $P_\alpha$. $P_\alpha\cap P_\beta$, $\beta<\alpha$ is a line and the cardinality of such lines is less than $2^{\aleph_0}$ so there is a line $\ell_\alpha\subset P_\alpha$ which is different from all lines $P_\alpha\cap P_\beta$. Hence it is not contained in any of the planes $P_\beta$, $\beta<\alpha$ which is condition (2).

$\endgroup$
0
3
$\begingroup$

There are many subsets $S$ with that property, and they need not cover any hyperplane.

For instance, let $S\subset G_1(\mathbb{R}^n)$ be the set of lines $\{\mathbb{R}a\}_{a\in A}$ generated by a nonempty, symmetric (i.e. $A=-A$), connected subset $A$ of $\mathbb{R}^n\setminus\{0\}$. For instance, a closed symmetric curve on the unit sphere (note that such a curve may meet any hyperplane in finitely many pairs of antipodal points).

The image of $A$ by any linear form on $\mathbb{R}^n$ is a non-empty symmetric interval, thus containing $0$. In other words, any hyperplane meets $A$, thus contains a line of $S$.

$\endgroup$
1
$\begingroup$

The answer to the question is "no". It is probably best to think in terms of projective geometry: you are asking for a set $S$ of points in projective $n-1$ space such that every hyperplane contains a point in $S$, and $S$ does not contain a hyperplane. Let $S_0$ be the union of two hyperplanes $H_1$ and $H_2$ with two points $p_1\in H_1$ and $p_2\in H_2$ removed. Let $S$ be the union of $S_0$ with a point $q$ lying on the line between $p_1$ and $p_2$. Any hyperplane not meeting $S_0$ must meet $H_1$ only in $p_1$ and $H_2$ only in $p_2$, and so must contain the line between $p_1$ and $p_2$, and hence must contain $q$.

$\endgroup$
5
  • $\begingroup$ I think there are some misunderstandings here. The Grassmann manifolds $G_k(\mathbb{R}^n)$ consist of all $k$-dimensional planes in $\mathbb{R}^n$ passing through the origin. What I actually meant in my statement was that whether there is a hyperplane in $G_{n-1}$ whose every line passing through the origin is contained in the given set. For example, could you give a detailed construction in $\mathbb{R}^3$ as in your counterexample? $\endgroup$
    – user160393
    Commented Mar 15, 2018 at 15:45
  • $\begingroup$ Indeed I misunderstood, perhaps my answer is clearer now. Piotr Hajlasz's answer seems relevant to the interesting question of whether there must always be a plane which contains infinitely many of the lines in $S$; it is possible that this was closer to your original intention. $\endgroup$ Commented Mar 15, 2018 at 21:14
  • $\begingroup$ @user160393 would you like any additional detail added? $\endgroup$ Commented Mar 20, 2018 at 8:44
  • $\begingroup$ I got it. Thank you. $\endgroup$
    – user160393
    Commented Mar 20, 2018 at 9:22
  • 1
    $\begingroup$ @user160393 If one of these answers resolves your question, you may want to "accept" it and mark this question as answered by clicking the checkmark to the left of the answer. $\endgroup$ Commented Mar 20, 2018 at 21:43

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .