Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

In $\mathbb{R}^n$, given an unit $L_1$ sphere $\mathcal{B}_n: |x_1|+|x_2|+\ldots+|x_n|\leq 1$ and a hyperplane $\mathcal{P}: a_1x_1+a_2x_2+\ldots+a_nx_n=0$. Does there always exist a rotation such that $\mathcal{B}_n\cap\mathcal{P}$ is embedded into the $(n-1)$ dimensional unit $L_1$ sphere: $\mathcal{B}_{n-1}: |x_1|+|x_2|+\ldots+|x_{n-1}|\leq 1, x_n=0$ after the rotation?

share|improve this question
1  
Oh, I misread your question. Well, try counting extreme points of ${\mathcal B}_n\cap {\mathcal P}$ when $a_1=\dots=a_n=1$; I think that may lead towards an answer (but I haven't checked the details) –  Yemon Choi Apr 5 '11 at 20:00
    
Hint: what is the shape of the $\ell_1^3$ sphere, and what is the shape of a cross-section parallel to one of its faces? –  David Eppstein Apr 5 '11 at 23:24
    
Thanks for the hint. In 3 dimensional case, the sphere is a diamond shape and the cross-section parallel to one of this faces is a regular hexagon, which can be proved inside a $\ell_1^2$ sphere. On high dimensional case, it is same as the example given by Yemon Choi. However, I'm not sure how to embed it in $\ell_1^{n-1}$. –  Chao Li Apr 6 '11 at 0:03
add comment

1 Answer

up vote 1 down vote accepted

Thanks for the comments above. I have just proofed this problem is only true when $n\leq 4$.

When $n\leq 4$, without loss of generality, assume $|a_n|\geq |a_1|, \ldots, |a_{n-1}|$. Let $Q$ be the rotation on the plane spanned by $(a_1, \ldots, a_n)$ and $(0,\ldots, 0, 1)$ such that applying $Q$ to $(a_1, \ldots, a_n)$ rotates it to $(0,\ldots, 0, 1)$. Apply the rotation to $\mathcal{B}\cap\mathcal{P}$ will inscribe it into $\mathcal{B}_{n-1}$.

For $n=5$, the intersection between $\mathcal{B}_n$ and $\mathcal{P}: a_1+a_2+a_3+a_4+a_5=0$ can not be inscribed in to the 4-dimension $L_1$ sphere. The intuition of the proof is that the $\mathcal{B}_n\cap\mathcal{P}$ contains the intersection between the 3-dimension $L_1$ sphere and hyperplane $a_1+a_2+a_3+a_4=0$. This intersection can only be inscribed into the 4-dimensional $L_1$ sphere under two rotations (with symmetry cases ignored). One can verified that non of those two rotations can fit $\mathcal{B}_n\cap\mathcal{P}$ into a 4-dimensional $L_1$ sphere.

share|improve this answer
add comment

Your Answer

 
discard

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.