4

For a regular 2017-gon $A_1A_2...A_{2017}$ in the plane, show that there exists a point P in the plane such that the following is true: $$ \sum_{i=1}^{2017}i\frac{\mathbf{PA_i}}{|\mathbf{PA_i}|^5}=\mathbf{0}. $$

Here is my thought: I think that the number 5 has no special meanings and that if I can answer the problem $$ \sum_{i=1}^{2017}i\frac{\mathbf{PA_i}}{|\mathbf{PA_i}|^2}=\mathbf{0}, $$ then the original can be proved by the same method. I did not do it for number 1 because I think that plugging 1 there makes the vector a unit vector and in general it shouldn't be. Then I find out that the number 2017 is not special so I tried 3 for simplicity. But I find that solving the problem by brute force even if it is only a triangle is not that easy because of the i before the term.

I saw this problem in The Simon Marais Mathematics Competition (held on 7th, October, 2017) but honestly I don't know what is the correct approach to this problem? Should I try to come up with a system of equations and try to show that there has to be a solution? Or should I just use induction (I don't think this is the proper way since I cannot find a direct relation between $P_{n+1}$ and $P_n$).

Please can someone give me some insights?

Jason
  • 126
  • @Joffan I know it is a regular polygon so the origin is equidistant from all $A_i$'s, but how to proceed from then? Since clearly P is not the origin. – Jason Oct 08 '17 at 03:30
  • @Blue I edited and put the name in the problem, is that okay? – Jason Oct 08 '17 at 03:32
  • 3
    @Jason: Looks okay to me! :) (I removed my previous comments, and will remove this one shortly.) Looking at the SMMC website, I see that the contest was held across multiple time zones, but that all sessions appear to be passed. Some contests span days or weeks. That's why it's important to identify the contest: It might have been over for you but not others. (BTW: Good for you for saying that you saw the problem in a contest to begin with.) – Blue Oct 08 '17 at 03:39
  • in addition to achille hui's soution involving writing the function as a gradient, there is another approach which is thematically similar to what I believe is Gauss' original approach to the fundamental theorem of algebra. The Simon Marais competition webpage as of today promises solutions to the problems shortly, and I expect this alternative solution will appear there shortly. If it doesn't, I can write it up here. – Peter McNamara Oct 08 '17 at 20:02

2 Answers2

3

Let $A = \{ A_1, A_2, \ldots, A_{2017} \}$ be the set of vertices. Let $D = \bar{B}(0,1)$ be the closed unit disc centered at origin. WOLOG, we will assume the vertices lie on the unit circle $S^1 = \partial D$.

Consider the function $f : D \to \mathbb{R} \cup \{ +\infty \}$ defined by

$$D \ni p\quad \mapsto\quad f(p) = \begin{cases} \sum\limits_{k=1}^{2017} \frac{k}{|pA_k|^3}, & p \in D \setminus A\\ \\ +\infty, & p \in A \end{cases}$$ The function $f$ is non-negative and continuous on $D$ and smooth on $D \setminus A$.

Notice we can cover $S^1 = \partial D$ by a bunch of closed discs centered at $A_k$ with radius $r = \sin\frac{\pi}{2017}$.
There means for any $p \in \partial D$, there is at lease one $A_{k_p}$ with $|pA_{k_p}| \le r$. This leads to

$$f(p) = \sum_{k=1}^{2017} \frac{k}{|pA_k|^3} \ge \frac{k_p}{|pA_{k_p}|^3} \ge \frac{1}{r^3} = \frac{1}{\sin^3\frac{\pi}{2017}}$$

Notice at the origin $O$, all $|OA_k| = 1$, we have

$$f(O) = \sum_{k=1}^{2017} k = \frac{2017(2018)}{2} \approx \frac12 (2017)^2$$

Since $\frac{1}{r^3} \approx \left(\frac{2017}{\pi}\right)^3$ is much larger than $f(O)$, we can conclude $f$ achieve its absolute minimum over $D$ at some point $P \in {\rm int} D \subset D \setminus A$. Since $f(p)$ is smooth at $P$, the gradient vanishes there.

$$\left.\nabla f(p)\right|_{P} = 0 \quad\iff\quad \sum_{k=1}^{2017} k \frac{PA_k}{|PA_k|^5} = 0$$

In the above proof, neither the number of sides $2017$, the exponent $5$ nor the weights $k$ attached to line segment $PA_k$ is that special. What we need is the the number of sides and exponent are large enough to force the smallest value of $f(p)$ on $\partial D$ larger than the value at some interior point. This will "push" the absolute minimum of $f(p)$ into the interior of $D$ and allow us to have at lease one point where the gradient of $f$ vanishes.

achille hui
  • 122,701
1

Here is an alternative solution (pdf 1.6Mb)