Re: Limit of a sequence
- From: David C. Ullrich <ullrich@xxxxxxxxxxxxxxxx>
- Date: Fri, 27 Jan 2006 07:44:18 -0600
On 27 Jan 2006 03:25:32 -0800, "Mate" <mmatica@xxxxxxxxxxx> wrote:
>Let f : [0,1] --> [0,1] be continuous and strictly decreasing such that
>f(0)=1, f(1)=0.
>Is it true that
>
>lim_(n-->oo) [ int (f(x))^(n+1) dx ] / [ int (f(x))^n dx ] = 1 (?)
Yes.
Lemma If 0 < a < 1 then
int_0^a f^n / int_a^1 f^n -> infinity.
Proof: Choose b with 0 < b < a. Then
int_0^a f^n > b f(b)^n
int_a^1 f^n < (1-a) f(a)^n,
and (f(b)/f(a))^n -> infinity. QED.
Now it's clear that all of your ratios are
less than 1, so we only need to show that the
lim sup of the ratios is >= 1, right.
But for any a as in the lemma, the lemma
shows that
lim sup int_0^1 f^{n+1} / int_0^1 f^n
= lim sup int_0^a f^(n+1) / int_0^a f^n,
and it's clear that
int_0^a f^(n+1) / int_0^a f^n > f(a),
hence its lim sup is >= f(a). So the lim sup
of the original ratio is >= f(a) for all a in (0,1).
>If the limit in (?) exists, then it must be 1 ( = ||f||_oo ),
>but I think that the sequence diverges for some f.
>I could not find such an f. Can you help?
>
>Thank you.
************************
David C. Ullrich
.
- Follow-Ups:
- Re: Limit of a sequence
- From: The World Wide Wade
- Re: Limit of a sequence
- From: Mate
- Re: Limit of a sequence
- References:
- Limit of a sequence
- From: Mate
- Limit of a sequence
- Prev by Date: Pi to 10 trillion places
- Next by Date: Re: A Basic Theorem
- Previous by thread: Limit of a sequence
- Next by thread: Re: Limit of a sequence
- Index(es):
Relevant Pages
|
Loading