Recently, I encountered a problem about infinite series. So my question is how to know whether the infinite series $\sum _{n=2}^{\infty } \frac{1}{n \log (n)}$ is convergent?
-
2Compare it to the integral $\int_2^\infty \frac{dt}{t\log t}$. – Daniel Fischer Nov 20 '13 at 11:44
-
See also: http://math.stackexchange.com/questions/1419868/how-can-i-show-that-sum-limits-n-2-infty-frac1n-ln-n-is-divergent-wit – Martin Sleziak Oct 31 '15 at 20:30
6 Answers
Check the conditions fit for the Condensation Test :
$$a_n:=\frac1{n\log n}\implies 2^na_{2^n}=\frac{2^n}{2^n\log2^n}=\frac1{\log 2}\frac1n$$
and since the series of the rightmost sequence is just a multiple of the harmonic series and thus diverges, also our series diverges.
- 56,060
- 214,715
To see whether $\sum_2^\infty 1/(n \log n)$ converges, we can use the integral test. This series converges if and only if this integral does: $$ \int_2^\infty \frac{1}{x \log x} dx = \left[\log(\log x)\right]_2^\infty $$ and in fact the integral diverges.
This is part of a family of examples worth remembering. Note that $$ d/dx \log(\log(\log x)) = d/dx \log(\log x) \cdot \frac{1}{\log (\log x)} = \frac{1}{x \log x \log(\log x)} $$ and $\log (\log (\log x)) \to \infty$ as $x \to \infty$ hence $\sum \frac{1}{n \log n \log (\log n)}$ diverges as well. Similarly, by induction we can put as many iterated $\log$s in the denominator as we want (i.e. $\sum \frac{1}{n \log n \log(\log n) \ldots \log (\ldots (\log n) \ldots )}$ where the $i$th log is iterated $i$ times), and it will still diverge. However, as you should check, $\sum \frac{1}{x \log^2x}$ converges, and in fact (again by induction) if you square any of the iterated logs in $\sum \frac{1}{n \log n \log(\log n) \ldots \log (\ldots (\log n) \ldots )}$ the sum will converge.
- 4,976
-
,Thanks for your detailed solutioin.I know the series can be see as a area of trapezoid with curved edge that the delta x=1,so the sum[1/(n*Ln(n)),{n,2,Infinity}] is less than the integrate[1/(x*Ln(x),{x,2,Infinity}].However,I cannot understand why the latter integral converge can reduct the former series converge?Can you tell me why?Thanks sincerely! – abc Nov 21 '13 at 14:01
-
1Excellent answer. Two comments that might look obvious:
- It is not only squaring, but also raising to the power of any $c>1$.
- When you square (or raise to the power of $c>1$), you should do that to an outermost $\log$, or else the series will still diverge. For example, $\sum1/\left(n\log n\log\log^{2}n\right)$ diverges.
-
Julien, are you sure about your claim that $\frac{1}{n\log^2 n}$ is summable? – Jürgen Sukumaran Nov 08 '19 at 16:30
-
Yes, because the condensation test https://math.stackexchange.com/a/574528/426834 yields $\sum_n \frac1{(\log 2)^2}\frac1{n^2} = \frac1{(\log 2)^2} \sum_n \frac1{n^2}$, which converges. This is also well known. – user3810316 May 11 '22 at 09:49
We circumvent using the integral test or its companion, the Cauchy condensation test. Rather, we use creative telescoping to show that the series $\sum_{n=3}^\infty \frac{1}{n\log(n)}$ diverges. To that end, we now proceed.
We will use the well-known inequalities for the logarithm (SEE THIS ANSWER)
$$\frac{x-1}{x} \le \log(x)\le x-1 \tag1$$
Using the right-hand side inequality in $(1)$, we see that
$$\log\left(\frac{n+1}{n}\right)\le \frac1n \tag 2$$
and
$$\log\left(\frac{\log(n+1)}{\log(n)}\right)\le \frac{\log(n+1)}{\log(n)}-1 \tag3$$
Applying $(2)$ and $(3)$ yields
$$\begin{align} \sum_{n=3}^N \frac{1}{n\log(n)} &\ge \sum_{n=3}^N \frac{\log\left(\frac{n+1}{n}\right)}{\log(n)}\\\\ &=\sum_{n=3}^N \left(\frac{\log(n+1)}{\log(n)} -1\right)\\\\ &\ge \sum_{n=3}^N \log\left(\frac{\log(n+1)}{\log(n)}\right)\\\\ &=\sum_{n=3}^N \left(\log(\log(n+1)) -\log(\log(n)) \right)\\\\ &=\log(\log(N+1))-\log(\log(3)) \end{align}$$
Inasmuch as $\lim_{N\to \infty}\log(\log(N+1))=\infty$, the series of interest diverges by comparison.
And we are done!
TOOLS USED: The right-hand side inequality in $(1)$ and summing a telescoping series.
- 184,670
Here is another possible answer.
We will derive the asymptotic formula of the partial sum $\sum_{1< n\leqslant x}\frac{1}{n\log n}$ to show that this series diverges.
By Euler's summation formula, we have $$\begin{align*} \sum_{1< n\leqslant x}\frac{1}{n\log n} &=\frac{1}{2\log 2}+\sum_{2< n\leqslant x}\frac{1}{n\log n}\\ &=\frac{1}{2\log 2}+\int_{2}^x\frac{1}{t\log t}dt-\int_{2}^x(t-[t])\frac{\log t+1}{t^2\log^2t}dt-\frac{x-[x]}{x\log x}\\ &=\log\log x+\frac{1}{2\log 2}-\log\log 2-\int_{2}^xI(t)dt-O\left(\frac{1}{x\log x}\right) \end{align*}$$ where $I(t)=(t-[t])\frac{\log t+1}{t^2\log^2t}$.
Since $$\int_{2}^xI(t)dt=\int_2^\infty I(t)dt-\int_x^\infty I(t)ft,$$ and $$I(t)\leqslant \frac{\log t+1}{t^2\log^2t},$$ we obtain ($x\geqslant 2$) $$\int_x^\infty I(t)dt\leqslant\int_x^\infty \frac{\log t+1}{t^2\log^2t}=O\left(\frac{1}{x\log x}\right)$$ and $$\int_2^\infty I(t)dt\leqslant\int_2^\infty \frac{\log t+1}{t^2\log^2t}=\frac{1}{2\log 2}.$$
This implies $-\int_{2}^xI(t)dt=C+O\left(\frac{1}{x\log x}\right)$ (the first integral is not related to $x$ and it's bounded by $1/2\log 2$, thus it's a constant).
Therefore, we have $$\sum_{1< n\leqslant x}\frac{1}{n\log n}=\log\log x+B+O\left(\frac{1}{x\log x}\right)$$
This tells us, as $x\to\infty$, the sum goes to infinity too.
I understand that this asymptotic formula is not needed to prove its divergence, but it may provide you with another way of proving convergence or divergence.
We use integral test to see whether series converge or diverge.
Let $f(x)=\frac{1}{x \log x}.$
Take $u=\log x$ then $du=\frac{1}{x} dx$
Now, $\int f(x) dx=\int \frac{1}{u}du$
$\int_2^\infty f(x)=\log(\log x))|^\infty_2=\lim_{b \rightarrow \infty} \log(\log b)-\log (\log 2)=\infty.$
Hence the series diverges.
- 467
$\sum_{n≥2} \frac{1}{n(\log n)}=\sum_{n≥2} \int_{1}^{\infty} \frac{1}{n^x} dx$
$=\int_{1}^{\infty} \zeta_2(x) dx$ where [$\zeta_2(x)=\zeta(x)-1$]
So, this is the area under the curve $\zeta(x), x≥1, y≥0$.
Now, for $x>1$,
$\zeta_2(x) > {\frac{1}{2^x} + 2\frac{1}{4^x}+ 4\frac{1}{8^x} +......}$
$=\frac{1}{2}(\frac{1}{2^{x-1}}+\frac{1}{4^{x-1}}+....)$
$=\frac{1}{2^x}(\frac{1}{1-\frac{1}{2^{x-1}}})$
$=\frac{1}{2}(\frac{1}{2^{x-1}-1})$.
So, $\int_{1}^{\infty} \zeta_{2}(x) dx >\int_{1}^{\infty} \frac{1}{2}(\frac{1}{2^{x-1}-1}) dx=\int_{0}^{\infty} \frac{1}{2}(\frac{1}{2^z-1})dz $ which diverges.
This means $\int_{1}^{\infty} \zeta_{2}(x) dx$ also diverges.
And this implies $\sum_{n≥2} \frac{1}{n(\log n)}$ diverges.
- 36,640
- 946