(Assume that there is no friction and drag, and the ball follows the law of reflection.) Suppose we have an \(n\)-gon. There is a point-like ball at the midpoint of one of the sides of the \(n\)-gon (call that side Side A). It is launched to the midpoint of another side (call that Side B). Let \(k\) be the number of sides clockwise to Side A but counterclockwise to Side B (not including Side A and Side B). Define the function \(\delta_n(k)\) to be \(k\) if the ball bounces off every side of the \(n\)-gon before returning to the launch point, and otherwise \(0\). Find the closed form for the sum \[\sum^{n-2}_{k=0}{\delta_n(k)}\]

EDIT: Here is a hint: Euler's Totient Function

And if that doesn't make sense, here is a stepping stone: GCD. (Thanks Calvin L.)

## Comments

Sort by:

TopNewestA better hint would be Greatest Common Divisor, though it might give away too much. – Calvin Lin Staff · 4 years, 1 month ago

Log in to reply

I assume you are talking about a regular n-gon? – Tim Vermeulen · 4 years, 2 months ago

Log in to reply

– Daniel Liu · 4 years, 1 month ago

yes.Log in to reply

Are you bobthesmartypants? :P – Michael Tang · 4 years, 2 months ago

Log in to reply

– Ryan Soedjak · 4 years, 2 months ago

Yeah that's what I thought.Log in to reply

– Daniel Liu · 4 years, 1 month ago

yep :PLog in to reply

Consider a square. \(\delta_{4}(2) = 2\), because the ball will hit all sides before returning to its initial position. Also, \(\delta_{4}(0) = 0\), but that doesn't contribute to the sum. So \[\sum\limits_{k=0}^{2} \delta_{4}(k) = 2\] In general, \[\sum\limits_{k=0}^{n-2} \delta_{n}(k) \geq n - 2\] Another observation I made is that \[\sum\limits_{k=0}^{p-2} \delta_{p}(k) = \sum\limits_{k=0}^{p-2} k = \frac{(p-2)(p-1)}{2}\] for every prime number p. This is because the ball will never return to the initial position before hitting each side in such a p-gon, no matter which side you point the ball at. – Tim Vermeulen · 4 years, 2 months ago

Log in to reply

HINT: Euler's totient function – Daniel Liu · 4 years, 1 month ago

Log in to reply

I know that \[\sum\limits_{k=0}^{n-2} \delta_{n}(k) = - \phi(n) + \sum_{k=1}^{n-1} \begin{cases} k &\mbox{if } \gcd(n,k) = 1 \\ 0 &\mbox{otherwise} \end{cases}\] but that's all I got so far. ;-) Can it get better than this? – Tim Vermeulen · 4 years, 1 month ago

Log in to reply

– Daniel Liu · 4 years, 1 month ago

yep, there is actually a closed form for the second sum on the RHS. Hint on finding the closed sum: What is Gauss famous for as a schoolchild?Log in to reply

– Tim Vermeulen · 4 years, 1 month ago

The second term: \[\sum_{k=1}^{n-1} \begin{cases} k &\mbox{if } \gcd(n,k) = 1 \\ 0 &\mbox{otherwise} \end{cases}\] is the sum of all values \(k\) smaller than \(n\) that are coprime to \(n\). These values \(k\) seem symmetric about \(\frac{n}{2}\). For example, if \(n=6\) then \(k=1\) or \(k=5\). And (intuitively) this works for any \(n\). So basically, the average for all values \(k\) is \(\frac{n}{2}\). So the second term in my previous solution is equal to the average \(\left(\frac{n}{2}\right)\) multiplied by the number of terms \(\left(\phi(n)\right)\): \[ \begin{align*} \sum\limits^{n-2}_{k=0} \delta_{n}(k) &= \phi(n) * \frac{n}{2} - \phi(n) \\ &= \phi(n) \left(\frac{n}{2} - 1\right) \end{align*} \] I'm guessing this is the answer? :) That was a very entertaining problem, thanks.Log in to reply

– Daniel Liu · 4 years, 1 month ago

that might be a bad hint, so think about this: pairing values upLog in to reply

Is the answer zero?? – Shubham Srivastava · 4 years, 2 months ago

Log in to reply

– Daniel Liu · 4 years, 1 month ago

no, sorry.Log in to reply