[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference rusure::math

Title:Mathematics at DEC
Moderator:RUSURE::EDP
Created:Mon Feb 03 1986
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:2083
Total number of notes:14613

1448.0. "Inscribed circle radius limit" by ELIS::GARSON (V+F = E+2) Fri May 24 1991 12:07

Start with a circle of unit radius.

Repeat forever
    1. Inscribe a regular n-gon within the circle
    2. Draw the inscribed circle of the n-gon

What is the limit of the radius of the circle? Obviously it depends on what
n-gon(s) we use.

A) Using the same 'n' every time clearly leads to the uninteresting case that
   the limit is 0 (for all n >= 3).

B) Using a triangle, then a square, then a pentagon etc. was discussed in 1086
   (the note not the year). The limit was shown to exist although a closed form
   wasn't found.

C) Using a square, then an octagon etc (i.e. double the number of sides each
   time) has an elegantly simple closed form for the limit which you are
   invited to find.
T.RTitleUserPersonal
Name
DateLines
1448.1SSDEVO::LARYSpukhafte FernwirkungenSun May 26 1991 00:5031
The limiting radius is prod(n=2->inf) cos(pi/(2^n)), or

(1)	r = cos(pi/4) * cos(pi/8) * cos(pi/16) * ...

From elementary trigonometry,

(2)	sin(a) = 2*cos(a/2)*sin(a/2).

Letting a = pi/4 and substituting (2) into (1) (since cos(pi/4) = sin(pi/4)),

(3)	r = 2 * cos^2(pi/8) * sin(pi/8) * cos(pi/16) * cos(pi/32) * ...

Let a = pi/8 and do it again,

(4)	r = 2^2 * cos^2(pi/8) * cos^2(pi/16) * sin(pi/16) * cos(pi/32) * ...

If you keep doing it successively with a = pi/2^n, you get:

(5)	r = lim(n->inf)(2^(n-2) * sin(pi*2^-n)) * prod(n=3->inf)(cos^2(pi/(2^n))

Now, = sin(pi*2^-n) = pi*2^-n + (higher terms), so the limit is equal to pi/4,
and the product in (5) is the square of the original product without the first
term, so 

(6)	r = pi/4 * (r / cos(pi/4))^2 = pi/2 * r^2 (since cos^2(pi/4) = 1/2).

Solving for r, 

(7)	r = 2 / pi.

As advertised, an elegant result.
1448.2SSDEVO::LARYSpukhafte FernwirkungenMon May 27 1991 02:293
A generalization of the previous result is:

	Prod(n=0 to inf)(cos(x/(2^n))) = sin(2x) / 2x.
1448.3HERON::BUCHANANbreggin no tubaWed Jun 05 1991 12:3410
>A generalization of the previous result is:
>
>	Prod(n=0 to inf)(cos(x/(2^n))) = sin(2x) / 2x.

	Can we generalize further?

	Prod(n=0 to inf)(cos(x/(K^n))) = ?,
for K = 3,4,5...

Andrew
1448.4I can'tELIS::GARSONV+F = E+2Thu Jun 20 1991 07:1561
re .3

>	Can we generalize further?
>
>	Prod(n=0 to inf)(cos(x/(K^n))) = ?,
>		for K = 3,4,5...


The obvious generalisation of the method of .1 doesn't answer your question but
here are my scratchings anyway.

Let sin Kx = f (x) K sin x
              K

    where f (x) = P (cos x)
           K       K

        where P  is a polynomial of degree K-1
               K

Then
	  oo
        +---+
        |   |
        |   | f (x/K^n) = (sin Kx)/Kx
        |   |  K
         n=0

K=2: f (x) = cos x and this is the result from .2.
      2

K=3: f (x) = (4cos�x - 1)/3 but the resulting product doesn't look
      3
particularly useful.

K=4 looks even messier but in fact gives the same product as K=2.


Today's trivia: P (1) = 1 for all K
                 K


Now some easy observations on the question asked in .3 ...

Define g (x) as the required infinite product.
        K

For any x in [0,pi/2] and K>=2, from the monotone convergence theorem it can be
seen that g (x) is defined. Considered as a sequence in K, g (x) is
           K                                                K

monotonically increasing, strictly except at x=0 and x=pi/2.

        sin 2x
g (x) = ------		from .2
 2        2x

g (x) . g (x/2) = g (x)		Is this solvable for g (x)?
 4       4         2                                  4
    
We can do similar things for other non-prime K but I don't get anywhere.