Convergence or divergence
\[\sum_{n=2}^{\infty} \frac{ 1 }{ (\ln n)^{\ln n} }\]
What should I use for the comparison test here, an and bn, I'm not sure.
Well bn
\[a_n = \frac{ 1 }{ (\ln n)^{\ln n} }\]
?
[ln(n)]^(ln(1/n)) it seems to be O(x^2) but i got no idea if thats helpful or not
I suspect it's converging so should I just try 1/n^2 or something?
1/y <= 1/n^2 n^2 <= y assuming y is positive
Looks good I think
\[\frac{ 1 }{ n^2 } > \frac{ 1 }{ (\ln n)^{\ln n} } \implies (\ln n)^{\ln n} > n^2\] \[\ln(\ln n)^{\ln n} > \ln n^2 \implies n> e^{e^{2}}\] It would be convergent?
ln(x) ^ ln(x) is not always < n^2
if p >1 actually I'll just use ratio or root test or something comparison tests as such always bother me as I don't fully understand them haha.
e^(ax) <= ln(x) ^ ln(x) ax <= ln^2(x)
does e^(-n) converge? it seems to be bigger than ln(x) ^(-ln(x)) after say x=.5
I don't think so
this is breaking the wolf ...
Haha xD, I don't think ratio test is working either, I'm so confused.
whats wrong with comparison test with 1/n^2 ?
Was it right haha? I was guessing there.
After skipping few initial terms, each term in 1/n^2 series is larger than the corresponding term in 1/(ln(n))^(ln(n)) yes ?
Yup
since the series with larger terms 1/n^2 converges by p-series test, the terms with smaller terms 1/(ln(n))^(ln(n)) also converges
that makes sense intuitively atleast right ?
Note that the initial terms dont matter for convergence
http://www.wolframalpha.com/input/?i=x%5E2+%3D+ln%28x%29%5E%28ln%28x%29%29 yeah, i was mixing the 2. the bigger value in fraction form is the smaller number
can we prove its cauchy?
i read somewhere that its easier to prove cauchy then convergence, and that every cauchy is convergent
But ln(ln n) is an increasing function right, wouldn't that be diverging, making the whole series diverge?
why are we talking about ln(ln(n)) ?
\[\ln(\ln n) = 2 \implies \ln n = e^2 \implies n = e^{e^{2}}\] and \[n = e^{e^{2}}\] is some huge number as well..wait nvm the larger series is converging.
above arguement with 1/n^2 is sufficient, we don't need to explicitly find a "n" value for proving the convergence
So is there also other ways of working this problem out without it failing?
Yeah I got that
it doesnt matter how large the n have to be for 1/n^2 terms to dominate... it is fine as long as there exists some finite n, after which all the terms in 1/n^2 are larger than the terms in the particular series in question..
i think we're trying comparison after giving up on all other tests eh ?
Lool, yeah, I tried ratio only other than this and it failed, but I think root might work, I'm just messing around with this question trying to understand what exactly is going on. But, what you said makes a lot of sense. The 1/n^2 is always dominating and the larger series converges because it's a p series with p = 2>1.
most importantly - comparison test works only when the terms in both series are positive
Right!
Thank you @amistre64 and @ganeshie8
Join our real-time social learning platform and learn together with your friends!