Maths Olympiad Prep

Library / Harvard-MIT Mathematics Tournament /351 of 352

Harvard-MIT Mathematics Tournament

Geometry Difficulty 6.0 National Olympiad Prove it United States

Problem:

Let SS be the set of the points (x1,x2,…,x2012)(x_{1}, x_{2}, \ldots, x_{2012}) in 2012-dimensional space such that ∣x1∣+∣x2∣+⋯+∣x2012∣≤1|x_{1}| + |x_{2}| + \cdots + |x_{2012}| \leq 1. Let TT be the set of points in 2012-dimensional space such that max⁡i=12012∣xi∣=2\max_{i=1}^{2012} |x_{i}| = 2. Let pp be a randomly chosen point on TT. What is the probability that the closest point in SS to pp is a vertex of SS?

Solution

Solution:

Answer: 122011\frac{1}{2^{2011}}

Note that TT is a hypercube in 2012-dimensional space, containing the rotated hyperoctahedron SS. Let vv be a particular vertex of SS, and we will consider the set of points xx on TT such that vv is the closest point to xx in SS. Let ww be another point of SS and let ℓ\ell be the line between vv and ww. Then in order for vv to be the closest point to xx in SS, it must also be so on the region of ℓ\ell contained in SS. This condition is then equivalent to the projection of xx lying past vv on the line ℓ\ell, or alternatively that vv lies in the opposite halfspace of ww defined by the hyperplane perpendicular to ℓ\ell and passing through vv. This can be written algebraically as (x−v)⋅(w−v)≤0(x-v) \cdot (w-v) \leq 0. Therefore, vv is the closest point to xx if and only if (x−v)⋅(w−v)≤0(x-v) \cdot (w-v) \leq 0 for all ww in SS.

Note that these conditions do not depend on where ww is on the line, so for each line intersecting SS nontrivially, let us choose ww such that ww lies on the hyperplane HH containing all the vertices of SS except for vv and −v-v. We can further see that the conditions are linear in ww, so them holding for all ww in HH is equivalent to them holding on the vertices of the region S∩HS \cap H, which are simply the vertices of SS except for vv and −v-v. Let us now compute what these conditions look like.

Without loss of generality, let v=(1,0,…,0)v = (1, 0, \ldots, 0) and w=(0,1,0,…,0)w = (0, 1, 0, \ldots, 0). Then the equation is of the form (x−v)⋅(−1,1,0,…,0)≤0(x-v) \cdot (-1, 1, 0, \ldots, 0) \leq 0, which we can rewrite as (x1−1,x2,x3,…,x2012)⋅(−1,1,0,…,0)=1−x1+x2≤0(x_{1} - 1, x_{2}, x_{3}, \ldots, x_{2012}) \cdot (-1, 1, 0, \ldots, 0) = 1 - x_{1} + x_{2} \leq 0. For the other choices of ww, we get the similar conditions that 1−x1+xi≤01 - x_{1} + x_{i} \leq 0 and also 1−x1−xi≤01 - x_{1} - x_{i} \leq 0 for each i∈{2,…,2012}i \in \{2, \ldots, 2012\}.

Note that if any xi∈{2,−2}x_{i} \in \{2, -2\} for i≠1i \neq 1, then one of these conditions trivially fails, as it would require 3−x1≤03 - x_{1} \leq 0. Therefore, the only face of TT where xx can lie is the face defined by x1=2x_{1} = 2, which gives us the conditions that −1+xi≤0-1 + x_{i} \leq 0 and −1−xi≤0-1 - x_{i} \leq 0, so xi∈[−1,1]x_{i} \in [-1, 1] for all i∈{2,…,2012}i \in \{2, \ldots, 2012\}. This defines a 2011-dimensional hypercube of side length 2 on the face of TT defined by x1=2x_{1} = 2, and we obtain similar regions on each of the other faces corresponding to the other vertices of SS.

Therefore, the volume of the set of xx for which xx is closest to a vertex is 2⋅2012⋅220112 \cdot 2012 \cdot 2^{2011} and the volume of all the choices of xx is 2⋅2012⋅420112 \cdot 2012 \cdot 4^{2011}, so the desired probability is 122011\frac{1}{2^{2011}}.

Want a route through all this instead of an archive? The track puts 2,604 problems in a working order, from Junior Challenge level to the IMO shortlist.

Source: MathNet, licensed CC-BY-4.0. Statement reproduced verbatim; metadata (topic, difficulty) added by this project.