I define a sequence as following:
A[0] = 0
A[1] = A[2] = 1
A[n] = A[n-1] / cos (pi/n)
My question is - does this sequence have a finite limit? If yes, what that is?
Where does this come from: I try to understand if there is a limit to how circumcircles can grow if I keep building on a regular n-gon step by step, each step increasing the count of sides of the n-gon by 1. So it goes like this:
- For 0 we just assign radius as 0
- For 1 and 2 there are no n-gons, we assign 1 (this is conventional)
- For 3 we are looking on an excircle of a triangle. This is our A[3]
- For 4 we are building a square around the circumcircles of a triangle (thus that circle will be an inscribed circle for that square). Then we build circumcircles for that square. This is our A[4]
- ...
My reasoning for the sequence goes like this:

I wrote a simple recursive function that calculates the circle radius. And it looks like it converges:
(outputting last 10 members)
9999998-th rad is: 8.7000323319286
9999999-th rad is: 8.700032331929
but I was unable to progress in proving that the sequence even converges, let alone to find the limit.
"Intuitively" I feel as there must be a finite limit because n-gon gets closer and closer to a circle eventually becoming its own circumcircles. But then again we all have this awkward situation with the harmonic sum which "feels like should converge"