My Math Forum Convergence/divergence: strange case

 Real Analysis Real Analysis Math Forum

 December 4th, 2017, 07:09 AM #1 Newbie   Joined: Oct 2017 From: Italy Posts: 15 Thanks: 0 Convergence/divergence: strange case Everyone knows that $\sum \frac{1}{n^a}$ is divergent if $a\leq 1$ and is convergent if $a >1$. But, I don't remember who said that thing: if I have this sum $\sum \frac{1}{n^{b_n}}$ if $b_n \longrightarrow 1$ I can't say instantly if this sum is convergent or divergent. Is it true? Can you make me an example? Last edited by Berker; December 4th, 2017 at 07:11 AM.
 December 4th, 2017, 08:54 AM #2 Math Team   Joined: Dec 2013 From: Colombia Posts: 7,502 Thanks: 2511 Math Focus: Mainly analysis and algebra Well, the trivial case $b_n=1$ for all $n$ obviously doesn't converge. Nor does any case $b_n \le 1$ for all $n$. For convergence by the Root Test, we need \begin{align*}\displaystyle \lim_{n \to \infty} \left|n^{-b_n}\right|^{\frac1n} & < 1 \\ \lim_{n \to \infty} n^{-\frac{b_n}{n}} & < 1 \\ \log \lim_{n \to \infty} n^{-\frac{b_n}{n}} & < 0 \\ \lim_{n \to \infty} \log n^{-\frac{b_n}{n}} & < 0 \\ \lim_{n \to \infty} -b_n \frac{\log n}{n} & < 0 \\ \left(\lim_{n \to \infty} -b_n\right) \left(\lim_{n \to \infty} \frac{\log n}{n}\right) & < 0 \end{align*} But the left hand limit is always equal to zero (when the $b_n$ converge). So the root test will never give convergence. That doesn't mean we can't get convergence though. My gut feeling is that divergence is guaranteed. Last edited by v8archie; December 4th, 2017 at 09:29 AM.
 December 4th, 2017, 09:26 AM #3 Newbie   Joined: Oct 2017 From: Italy Posts: 15 Thanks: 0 And for $b_n \longrightarrow 1^{+}$, (so $b_n>1$) is it possible that the sum converges?
 December 4th, 2017, 10:50 AM #4 Senior Member   Joined: Sep 2016 From: USA Posts: 502 Thanks: 280 Math Focus: Dynamical systems, analytic function theory, numerics This series can't converge for any choice of $b_n$. To see this, write $b_n = 1 + c_n$ so that now $c_n \to 0$. The sum you have written is equivalent to $\zeta(1+c_n)$ whenever $c_n >0$ is real where $\zeta$ is $\zeta(s) = \sum_{n=1}^\infty n^{-s} \quad \text{Re}(s) > 1.$ Now, at $s = 1$ we have exactly the usual harmonic series implying the sum blows up. In fact, this is independent of the sum formula since $\zeta$ admits an analytic continuation which is analytic on the entire complex plane except at a simple pole precisely when $s= 1$. Thus, $s = 1$ is a simple pole requires $\lim_{s \to 1} \left|\zeta(s) \right| = \infty$. This can be seen directly from the residue theorem applied to our sequence, $c_n$ which gives $\lim_{n\to \infty} c_n \zeta(1+c_n) \to 1$ and since $c_n \to 0$, we must have $\zeta(1+c_n) = \zeta(b_n) \to \infty$.

 Thread Tools Display Modes Linear Mode

 Similar Threads Thread Thread Starter Forum Replies Last Post ineedhelpnow Calculus 2 July 26th, 2014 03:38 PM razzatazz Real Analysis 8 May 11th, 2013 12:30 AM math221 Calculus 3 April 8th, 2013 07:39 AM FreaKariDunk Real Analysis 28 April 30th, 2012 09:22 PM MathematicallyObtuse Algebra 3 January 29th, 2011 11:35 PM

 Contact - Home - Forums - Cryptocurrency Forum - Top