From: Rob Johnson on
In article <4c6033fd$0$10191$ba4acef3(a)reader.news.orange.fr>,
Raymond Manzoni <raymman(a)free.fr> wrote:
>Rob Johnson a �crit :
>> In article <4c5efa99$0$10185$ba4acef3(a)reader.news.orange.fr>,
>> Raymond Manzoni <raymman(a)free.fr> wrote:
>>> Shin Dongjun a �crit :
>>>> Prove (log3)^2 > (log4)^3.
>>>>
>>>> (Don't use an approximation like log3 = 0.4771)
>>>>
>>>> log is common logarithms that is, base of log is 10.
>>>
>>> You could, for example, study f(x)= (x-1) ln(ln(x)) for x > e ...
>>
>> This works well for proving the other direction for natural logs:
>>
>> ln(3)^2 < ln(4)^3 [1]
>>
>> The derivative of (x-1) ln(ln(x)) = ln(ln(x)) + (x-1) 1/(x ln(x)),
>> which is positive for x > e. Thus, (x-1) ln(ln(x)) is monotonically
>> increasing for x > e, which implies [1].
>>
>> However, since the OP is about common logs, we get
>>
>> (x-1) log(log(x))
>>
>> = (x-1) (ln(ln(x)) - ln(ln(10)))/ln(10)
>>
>> whose derivative is
>>
>> ln(ln(x)) - ln(ln(10)) + (x-1)/(x ln(x))
>> ----------------------------------------
>> ln(10)
>>
>> which is negative for x in [3,3.7517684] and positive for x in
>> [3.7517684,4]. Thus, (x-1) log(log(x)) is decreasing for about 3/4
>> of the interval in question and increasing for about 1/4. However,
>> trying to use this in a proof seems difficult.
>>
>> Rob Johnson <rob(a)trash.whim.org>
>
>
> You are right Rob I noticed my mistake too (from your other
>answer...) and the fact that the derivative became 0 in (3,4) as you
>indicated (invalidating my attempt at least directly).
>
> In fact I was searching a correct way to handle this and got some
>results 'in line' with your response :
>
>1)
> Let's prove directly that (log(3)/log(2))^2 > 8 log(2) :
>
> Tables of powers :
> 2^n : 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048
> 3^n : 3, 9, 27, 81, 243, 729, 2187
>
> so that 3^7 > 2^11 (2187>2048, 9>8 is too weak...)
>
> and (log(3)/log(2))^2 > (11/7)^2
>
> On the other side 10 log(2)= log(1024)= log(1000*(1+24/1000)) = 3
>+ log(1+24/1000) < 3 + 24/(1000 ln(10)) (the series for log(1+x) is
>alternate)
> so (using ln(10)>2 and 24<25) 10 log(2) < 3 + 1/80
>
> so that 8 log(2) < 241/100 < (11/7)^2 < (log(3)/log(2))^2
>(laborious ...)
>
>
>2)
> Starting with 3^12 > 2^19 is more direct :
> (log(3)/log(2))^2 > (19/12)^2 > (360/12^2 = 5/2)
>
> 2^16 < 10^5 so that 8 log(2) < 5/2
>
> (to find 19/12 I used continued fractions of log(3)/log(2))
>
>
>...
> Other tricks could be tried (use 3^4 > 80 to get log(3)^2 > (1 + 3
>log(2))^2/16 that has to be greater than 8 log(2)^3 and so on...)
>
> but I was overall dissatisfied by all these answers... (kind of
>cheating since we replace 'real values approximations' by large integers
>computations). Perhaps too that I didn't get the point of the problem
>since... I still prefer my initial answer even if of no value here! :-)
>
> Thanks for the correction anyway,
> Raymond

I, too, was left with an unsatisfied feeling with this problem after
my solution. I computed log(3)/log(4) and sqrt(log(4)/log(10)) and
used continued fractions to find the rational number between them
with the smallest denominator (7/9). However, I find computing 4^81
somewhat distasteful in a problem that essentially asks for hand
computation.

However, I played around a bit more and found a smaller set of
rational approximants with denominators no greater than 14 that can
be used. These can be more easily verified:

3^14 > 4^11
(11/14)^2 > 8/13
10^8 > 4^13

These inequalities prove that

(log(3)/log(4))^2 > (11/14)^2 > 8/13 > log(4)/log(10)

Since the exponents are more reasonable, this makes me feel a bit
better. However, I still feel there must be a better solution.

Rob Johnson <rob(a)trash.whim.org>
take out the trash before replying
to view any ASCII art, display article in a monospaced font
From: Gc on
On 8 elo, 20:39, Shin Dongjun <acam...(a)gmail.com> wrote:
> Prove (log3)^2  > (log4)^3.
>
> (Don't use an approximation like log3 = 0.4771)
>
> log is common logarithms that is, base of log is 10.

log(3)^2 = log(12/4)^2 = (log(12) - log(4))^2 = log(12)^2 -
2log(12)log(4) + log(4)^2 > log(4)^2, because log(12)log(12) >
2log(12)log(4) = log(12)log(8).
From: Gc on
On 10 elo, 11:51, Gc <gcut...(a)hotmail.com> wrote:
> On 8 elo, 20:39, Shin Dongjun <acam...(a)gmail.com> wrote:
>
> > Prove (log3)^2  > (log4)^3.
>
> > (Don't use an approximation like log3 = 0.4771)
>
> > log is common logarithms that is, base of log is 10.
>
> log(3)^2 = log(12/4)^2 = (log(12) - log(4))^2 = log(12)^2 -
> 2log(12)log(4) + log(4)^2 > log(4)^2, because log(12)log(12) >
> 2log(12)log(4) = log(12)log(8).

From: Gc on
On 10 elo, 11:51, Gc <gcut...(a)hotmail.com> wrote:
> On 8 elo, 20:39, Shin Dongjun <acam...(a)gmail.com> wrote:
>
> > Prove (log3)^2  > (log4)^3.
>
> > (Don't use an approximation like log3 = 0.4771)
>
> > log is common logarithms that is, base of log is 10.
>
> log(3)^2 = log(12/4)^2 = (log(12) - log(4))^2 = log(12)^2 -
> 2log(12)log(4) + log(4)^2 > log(4)^2, because log(12)log(12) >
> 2log(12)log(4) = log(12)log(8).

No,OMG, forget his.

From: Rob Johnson on
In article <dd0464f6-0c3f-4b6c-b3cf-8895e792b4f5(a)f42g2000yqn.googlegroups.com>,
Gc <gcut667(a)hotmail.com> wrote:
>On 8 elo, 20:39, Shin Dongjun <acam...(a)gmail.com> wrote:
>> Prove (log3)^2 > (log4)^3.
>>
>> (Don't use an approximation like log3 = 0.4771)
>>
>> log is common logarithms that is, base of log is 10.
>
>log(3)^2 = log(12/4)^2 = (log(12) - log(4))^2 = log(12)^2 -
>2log(12)log(4) + log(4)^2 > log(4)^2, because log(12)log(12) >
>2log(12)log(4) = log(12)log(8).

2 log(4) = log(16) not log(8). A big tip-off that something is wrong
is that you've just proven that log(3)^2 > log(4)^2, which is false.

Rob Johnson <rob(a)trash.whim.org>
take out the trash before replying
to view any ASCII art, display article in a monospaced font