Would you please tell me what's the limit of (V_n-U_n)(VnUn) as nn approaches positive infinity ? PS: U_(n+1)= sqrt(U_nV_n)Un+1=UnVn and V_(n+1)=(U_n+V_n)/2Vn+1=Un+Vn2

3 Answers
Nov 18, 2016

lim_(n->oo)(V_n-U_n) = 0

Explanation:

Making U_n=lambda V_n and delta_n = V_n-U_n we have

delta_(n+1)=V_(n+1)-U_(n+1)=(1-lambda)V_(n+1) = (lambda+1)/2V_n-sqrt(lambda)V_n or

V_(n+1)/(V_n)= 1/2(lambda+1-2sqrt(lambda))/(1-lambda)=1/2(1-sqrt(lambda))^2/(1-(sqrt(lambda))^2)=1/2(1-sqrt(lambda))/(1+sqrt(lambda))

then delta_(n+1) = V_0(1/2(1-sqrt(lambda))/(1+sqrt(lambda)))^n so

lim_(n->oo)delta_n = 0 because 1/2((1-sqrt(lambda))/(1+sqrt(lambda)))<1

Nov 19, 2016

This is my thoughts:

U_(n+1) =sqrt(U_nV_n)
V_(n+1) = 1/2(U_n+V_n)

Let delta(n) = V_n - U_n

Assume the limit exists, and that:
lim _(n rarr oo) (V_n - U_n) = epsilon

:. lim _(n rarr oo) delta(n) = epsilon

Then it must also be the case that:
:. lim _(n rarr oo) delta(n+1) = epsilon

:. lim _(n rarr oo) (delta(n+1))/(delta(n)) = 1

Now:
delta(n+1) = V_(n+1) - U_(n+1)

delta(n+1) = (U_n+V_n)/2 - sqrt(U_nV_n)

(delta(n+1))/(delta(n)) = { (U_n+V_n)/2 - sqrt(U_nV_n) } / ( V_n - U_n ) = 1

(U_n+V_n)/2 - sqrt(U_nV_n) = V_n - U_n
U_n+V_n - 2sqrt(U_nV_n) = 2V_n - 2U_n
3U_n-V_n = 2sqrt(U_nV_n)
(3U_n-V_n)^2 = (2sqrt(U_nV_n))^2
9U_n^2 -6U_nV_n + V_n^2 = 4U_nV_n
9U_n^2 -10U_nV_n + V_n^2 = 0
(9U_n - V_n)( U_n - V_n ) = 0

9U_n - V_n = 0 => U_n = 1/9V_n
U_n - V_n = 0 => U_n = V_n

Not sure how this helps, but I welcome comments

Nov 20, 2016

lim n to oo (V_n-U_n) is non-negative.

Explanation:

U_(n+1)=sqrt(U_nV_n) to U_(n+1) > 0 to U_n > 0 to V_n > 0

Use the property of Means, GM <= AM.

Here, sqrt(U_nV_n) <=(U_n+V_n)/2

It follows that

U_(n+1)<=V_(n+1). So,

V_n-U_n>=0, for n=2, 3, 4, ...# So,

lim n to oo (V_n-U_n) is non-negative.