1

$-\frac12-\frac14+\frac13-\frac16-\frac18+\frac15-\frac{1}{10}...$

After rearrangement the series looks like $\sum^{\infty}_{n=2}\frac{(-1)^{n+1}}{n}$.

My way of doing this is using Taylor series of $\ln{(1+x)}=\sum^{\infty}_{n=1}\frac{(-1)^{n+1}x^n}{n}$.

Therefore let $x=1$, $\sum^{\infty}_{n=2}\frac{(-1)^{n+1}}{n}=\sum^{\infty}_{n=1}\frac{(-1)^{n+1}x^n}{n}-\sum^{1}_{n=1}\frac{(-1)^{n+1}x^n}{n}=\ln{2}-\sum^{1}_{n=1}\frac{(-1)^{n+1}x^n}{n}=\ln{2}-1$

Is my solution correct?

CoolKid
  • 2,738

3 Answers3

6

Firstly, your method is wrong as has been pointed out. You cannot rearrange an infinite series, because the definition of its value is the limit of its partial sums, and rearranging it anyhow will change the sequence of partial sums, so there is no guarantee that the limit will be the same.

Just for a concrete simple illustration:

$\frac12 - \frac13 + \frac14 - \frac15 \cdots$

$\ = ( \frac12 - \frac13 ) + ( \frac14 - \frac15 ) + \cdots$   [Note that even putting brackets here is not trivially true!]

$\ > 0$.

$- \frac13 - \frac15 + \frac12 - \frac17 - \frac19 + \frac14 - \frac1{11} - \frac1{13} + \frac16 \cdots$

$\ = ( - \frac13 - \frac15 + \frac12 ) + ( - \frac17 - \frac19 + \frac14 ) + ( - \frac1{11} - \frac1{13} + \frac16 ) + \cdots$

$\ = ( - \frac{3+5}{3 \cdot 5} + \frac12 ) + ( - \frac{7+9}{7 \cdot 9} + \frac14 ) + ( - \frac{11+13}{11 \cdot 13} + \frac16 ) + \cdots$

$\ < ( - \frac{3+5}{4 \cdot 4} + \frac12 ) + ( - \frac{7+9}{8 \cdot 8} + \frac14 ) + ( - \frac{11+13}{12 \cdot 12} + \frac16 ) + \cdots$

$\ = 0$.

Both series converge, so clearly rearrangement may not even preserve the sign of the sum.

Secondly, we can find the limit of your series quite easily.

We just need the following (loose) inequality for any $m,n \in \mathbb{N}_{>1}$ such that $m \le n$:

$\ln(n+1) - \ln(m) = \int_m^{n+1} \frac1x\ dx < \sum_{k=m}^n \frac1k < \int_{m-1}^n \frac1x\ dx = \ln(n) - \ln(m-1)$.

We first simplify every third partial sum:

$-\frac12-\frac14+\frac13 -\frac16-\frac18+\frac15 -\cdots -\frac1{4n+2}-\frac1{4n+4}+\frac1{2n+3}$

$\ = - ( \frac12+\frac14 + \cdots + \frac1{4n+2}+\frac1{4n+4} ) + ( \frac13+\frac15 + \cdots + \frac1{2n+3} )$

$\ = - ( \frac12+\frac14 + \cdots +\frac1{4n+4} ) + \left( ( \frac11+\frac12 + \cdots + \frac1{2n+3} ) - ( \frac12+\frac14 + \cdots + \frac1{2n+2} ) - 1 \right)$

$\ = -\frac12 ( \frac11+\frac12 + \cdots + \frac1{2n+2} ) + \left( ( \frac11+\frac12 + \cdots + \frac1{2n+3} ) - \frac12 ( \frac11+\frac12 + \cdots + \frac1{n+1} ) \right) - 1$

$\ = \frac12 ( \frac11+\frac12 + \cdots + \frac1{2n+2} ) - \frac12 ( \frac11+\frac12 + \cdots + \frac1{n+1} ) - 1 + \frac1{2n+3}$

$\ = \frac12 \sum_{k=n+2}^{2n+2} \frac1k - 1 + \frac1{2n+3}$

$\ \in \frac12 [ \ln(2n+3)-\ln(n+2) , \ln(2n+2)-\ln(n+1) ] - 1 + \frac1{2n+3}$

$\ = \frac12 [ \ln(2-\frac{1}{n+2}) , \ln(2) ] - 1 + \frac1{2n+3}$.

Note that we only rearranged a finite sum, not an infinite series.

Now clearly as $n \to \infty$:

$-\frac12-\frac14+\frac13 -\frac16-\frac18+\frac15 -\cdots -\frac1{4n+2}-\frac1{4n+4}+\frac1{2n+3}$

$\ \to \frac12 \ln(2) - 1$   [because $\frac1{2n+3} \to 0$ and $\frac{1}{n+2} \to 0$ and $\ln$ is continuous at $2$].

Since the terms of the original series tend to $0$, it also converges to $\frac12 \ln(2) - 1$. Done!

user21820
  • 57,693
  • 9
  • 98
  • 256
2

As has been noted, rearranging a conditionally convergent series needs to be done with care. Collecting terms into contiguous groups whose size is uniformly bounded is allowed, as long as the terms tend to $0$.

Assuming that the series can be continued as $$ \begin{align} &\left(-\frac12-\frac14\right)+\left(\frac13-\frac16-\frac18\right)+\left(\frac15-\frac1{10}-\frac1{12}\right)+\cdots\tag{1a}\\ &=-\frac34+\sum_{k=1}^\infty\left(\frac1{2k+1}-\frac1{4k+2}-\frac1{4k+4}\right)\tag{1b}\\ &=-\frac34+\sum_{k=1}^\infty\left(\frac1{4k+2}-\frac1{4k+4}\right)\tag{1c}\\ &=-\frac34+\frac12\sum_{k=1}^\infty\left(\frac1{2k+1}-\frac1{2k+2}\right)\tag{1d}\\ &=-1+\frac12\sum_{k=0}^\infty\left(\frac1{2k+1}-\frac1{2k+2}\right)\tag{1e}\\ &=-1+\frac12\sum_{k=1}^\infty\frac{(-1)^{k-1}}k\tag{1f}\\ &=-1+\frac12\log(2)\tag{1g} \end{align} $$ Explanation:
$\text{(1a):}$ collect the terms into contiguous groups whose size is no greater than $3$
$\text{(1b):}$ write the sum using sigma notation
$\text{(1c):}$ simplify the terms
$\text{(1d):}$ pull a factor of $\frac12$ outside the sum $\left(\sum ca_k=c\sum a_k\right)$
$\text{(1e):}$ add the $k=0$ term to the sum and subtract $\frac14$ from $-\frac34$ to compensate
$\text{(1f):}$ recognize the sum as an alternating series (from uniformly sized groups)
$\text{(1g):}$ apply the Alternating Harmonic Series


The Alternating Harmonic Series $$ \begin{align} \sum_{k=0}^\infty\frac{(-1)^k}{k+1} &=\lim_{n\to\infty}\sum_{k=0}^n\frac{(-1)^k}{k+1}\tag{2a}\\ &=\lim_{n\to\infty}\sum_{k=0}^n\int_0^1(-1)^kx^k\,\mathrm{d}x\tag{2b}\\ &=\lim_{n\to\infty}\int_0^1\sum_{k=0}^n(-1)^kx^k\,\mathrm{d}x\tag{2c}\\ &=\lim_{n\to\infty}\int_0^1\frac{1+(-1)^nx^{n+1}}{1+x}\,\mathrm{d}x\tag{2d}\\ &=\int_0^1\frac1{1+x}\,\mathrm{d}x+\lim_{n\to\infty}(-1)^n\int_0^1\frac{x^{n+1}}{1+x}\,\mathrm{d}x\tag{2e}\\ &=\int_0^1\frac1{1+x}\,\mathrm{dx}\tag{2f}\\[6pt] &=\log(2)\tag{2g} \end{align} $$ Explanation:
$\text{(2a):}$ definition of an infinite sum
$\text{(2b):}$ $\int_0^1(-1)^kx^k\,\mathrm{d}x=\frac{(-1)^k}{k+1}$
$\text{(2c):}$ a finite sum of integrals is the integral of the finite sum
$\text{(2d):}$ evaluate the finite geometric sum
$\text{(2e):}$ an integral of a finite sum is the finite sum of the integrals
$\phantom{\text{(2e):}}$ a limit of a finite sum is the finite sum of the limits
$\text{(2f):}$ $0\le\int_0^1\frac{x^{n+1}}{1+x}\,\mathrm{d}x\le\int_0^1x^{n+1}\,\mathrm{d}x=\frac1{n+2}$
$\phantom{\text{(2f):}}$ thus, $\lim\limits_{n\to\infty}(-1)^n\int_0^1\frac{x^{n+1}}{1+x}\,\mathrm{d}x=0$
$\text{(2g):}$ evaluate the integral

robjohn
  • 345,667
0

You cannot rearrange an infinite series which is not absolutely convergent. We can evaluate it as follows. Your infinite series is \begin{align} \sum_{n=1}^{\infty}\left(-\dfrac1{4n-2} - \dfrac1{4n}+\dfrac1{2n+1}\right) & = \sum_{n=1}^{\infty}\left(-\dfrac1{4n-2} - \dfrac1{4n}+ 2\cdot \dfrac1{4n+2}\right)\\ & = \sum_{n=1}^{\infty}\left(-\int_0^1x^{4n-3}dx- \int_0^1x^{4n-1}dx + 2\int_0^1x^{4n+1}dx\right)\\ & = \sum_{n=1}^{\infty}\int_0^1 x^{4n}\left(2x-\dfrac1x - \dfrac1{x^3}\right)dx\\ & = \int_0^1\left(2x-\dfrac1x - \dfrac1{x^3}\right)\left(\sum_{n=1}^{\infty}x^{4n}\right)dx\\ & = \int_0^1\left(2x-\dfrac1x - \dfrac1{x^3}\right) \dfrac{x^4}{1-x^4}dx\\ & = \int_0^1 \left(\dfrac{x}{x^2+1}-2x\right)dx = \dfrac{\ln(2)}2-1 \end{align}


To justify the swapping of the limit and integral, we have \begin{align} I_m & = \sum_{n=1}^m \int_0^1 x^{4n}\left(2x-\dfrac1x-\dfrac1{x^3}\right)dx = \int_0^1 \dfrac{x^4\left(1-x^{4m}\right)}{1-x^4}\left(2x-\dfrac1x-\dfrac1{x^3}\right)dx\\ & = \int_0^1 \left(\dfrac{x}{x^2+1}-2x\right)dx + \int_0^1 \dfrac{x(2x^2+1)}{x^2+1}\cdot x^{4m}dx \end{align}

Now $$\lim_{m\to\infty} \int_0^1 \dfrac{x(2x^2+1)}{x^2+1}\cdot x^{4m}dx = 0$$ There are multiple ways to prove this. One is to just invoke dominated convergence theorem, since $x^{4m} \leq 1$ and $\displaystyle \int_0^1 \dfrac{x(2x^2+1)}{x^2+1}dx < \infty$. Other is to simply note that $\dfrac{x(2x^2+1)}{x^2+1} \leq \dfrac32$ on $[0,1]$. Hence, $$\lim_{m\to\infty} \int_0^1 \dfrac{x(2x^2+1)}{x^2+1}\cdot x^{4m}dx \leq \lim_{m\to\infty} \int_0^1 \dfrac32\cdot x^{4m}dx = \lim_{m\to\infty} \dfrac3{2(4m+1)} = 0$$

Hence, $$\lim_{m \to \infty} I_m = \dfrac{\log(2)}2-1$$

Adhvaitha
  • 20,259
  • Not so fast! You need to justify your interchange of summation and integral, otherwise you are making the same kind of mistake as the asker, in your case rearranging the 'order of summation of infinitesimal pieces'! – user21820 Apr 21 '16 at 03:31
  • @user21820 The problem is not with swapping the integral and the summation. It is about rearranging the summation. Both are *two different things* the way I have written. An easier way to justify the above is to look at the summation till a finite number $M$ instead of $\infty$. It is then clear why the above is completely valid. Note that had I summed up term by term instead of summing them up together then it is a rearrangement error but not the way I have done now. – Adhvaitha Apr 21 '16 at 05:23
  • Your 4th equality is an explicit swap of the integral and summation. That is false in general. You claim that you can easily justify your method by using the partial sum. Show your justification. It is not as easy as what you're conveying to others. Also, your last sentence in your above comment makes no sense so I can't even respond to it. – user21820 Apr 21 '16 at 10:58
  • @user21820 As said before, swapping the limit is perfectly valid in this case and is not the same as rearranging the term in the sequence. To settle the debate, I have updated the answer. Also, I am a prof at reputed university and I know what I am talking about. Also, try to be polite atlas online. – Adhvaitha Apr 21 '16 at 13:01
  • 1
    I know it is valid, and I also know how to justify it. But frankly you have to learn how to convey what you know to people who don't. In your original answer you didn't even give any bit of justification, not to say state the use of the DCT or the necessary elementary inequalities needed to finish that step. That is simply misleading to anyone who does not know better, including many high-school students. You are active enough on this site to have seen many instances of such blind errors, and my comment was simply because your answer didn't teach correctly those who don't. Now it's fine. – user21820 Apr 21 '16 at 14:08
  • 1
    As for politeness, my comments were meant to be purely factual. This includes my comment that one of your sentences made no sense. If you take the integral as the limit of area under simple functions, then you can break down the final area into corresponding bits, and then interchanging the summation and integral does rearrange the order of the bits. In some sense it is worse than the original asker's error. Also, the situations where interchanging summation and integral is invalid are analogous to series where rearranging does not preserve the sum. – user21820 Apr 21 '16 at 14:18
  • And unfortunately I once again cannot (really) understand your last sentence. What do you mean by "polite atlas online"? Anyway if you feel any part of my comments are not factual statements, point out explicitly and I will take note and try to rephrase next time to convey more accurately what I mean. – user21820 Apr 21 '16 at 14:20
  • 1
    Anyway your updated answer still has a (small) flaw. You cannot from "$\int_0^1 \dfrac{x(2x^2+1)}{x^2+1}\cdot x^{4m}dx \leq \int_0^1 \dfrac32\cdot x^{4m}dx$" directly derive "$\lim_{m\to\infty} \int_0^1 \dfrac{x(2x^2+1)}{x^2+1}\cdot x^{4m}dx \leq \lim_{m\to\infty} \int_0^1 \dfrac32\cdot x^{4m}dx$", but need to use the squeeze theorem, or equivalently derive "$\limsup \cdots \leq \liminf \cdots$" and proceed from there. Of course, I won't doubt that you know this, but I think it's much better to make clear such steps where students commonly err. – user21820 Apr 21 '16 at 14:30