Can someone explain to me the paradox 1-2+3-4+... =1/4? I really don't get why and how. Thanks so much.

Results 1 to 15 of 31

- Apr 24th 2010, 06:20 AM #1

- Joined
- Jan 2009
- Posts
- 290

- Apr 24th 2010, 06:38 AM #2
The binomial series $\displaystyle 1-2x+3x^2-4x^3+\ldots$ converges to $\displaystyle (1+x)^{-2}$ when $\displaystyle |x|<1$. If you ignore that restriction on x and pretend that the series also converges to the same function when x=1 then you come up with the formula $\displaystyle 1-2+3-4+\ldots = 1/4$, which of course is not true in any conventional sense.

- Apr 24th 2010, 07:26 PM #3

- Joined
- Feb 2010
- Posts
- 422

Call $\displaystyle S:= 1-1+1-\cdots$. Cancelling off the very first term by subtracting 1 gives $\displaystyle -1+1-1+\cdots= - (1-1+1-\cdots)$ so $\displaystyle S = -(S - 1)$ or $\displaystyle S=\frac12$.

Call $\displaystyle T := 1-2+3-4+\cdots$. Add the series $\displaystyle 1-2+3-4+\cdots$ and $\displaystyle 1-1+1-1+\cdots$ together term by term to get $\displaystyle T+S = (1+1) - (2+1) + (3+1) - \cdots = 2 - 3 + 4 -\cdots = -(T-1)$. Since $\displaystyle S=\frac12$, $\displaystyle T+\frac12 = -(T-1)$ or $\displaystyle T=\frac14$.

The remarkable thing, of course, is that you get the same answer if you do it "algebraically" like this as if you do it like Opalg does it (analytically, generating funtions).

- Apr 24th 2010, 07:35 PM #4

- Apr 24th 2010, 08:05 PM #5

- Joined
- Mar 2009
- Posts
- 378

- Apr 24th 2010, 08:34 PM #6

- Joined
- Feb 2010
- Posts
- 422

Any convergent series whose terms are integers has only finitely many non-zero terms. But closure only applies to finite sums. For instance, the rationals are closed under addition but the sum $\displaystyle 1+\frac1{2!}+\frac1{3!}+\frac1{4!}+\cdots$ is irrational.

- Apr 25th 2010, 05:44 AM #7

- Joined
- Mar 2009
- Posts
- 378

- Apr 25th 2010, 05:56 AM #8

- Apr 25th 2010, 06:20 AM #9

- Joined
- Jan 2009
- Posts
- 290

Read this

1 ? 2 + 3 ? 4 + · · · - Wikipedia, the free encyclopedia

I don't know enough math to understand the "proofs" in that article. Can anyone explain the number manipulation method? It seems like the proof that requires the least amount of math knowledge to understand.

- Apr 25th 2010, 07:39 AM #10

- Joined
- Mar 2009
- Posts
- 378

Things get weird when we begin talking infinity though. But then again, the abstract algebra sense of the Integers does include an infinite collection. Even the linear algebra sense gives you closure since the Integers are a vector space.

EDIT: I can see how this would be 1/4 in the formal power sense (which is what maddas was saying), but this sub forum is not that advanced.

- Apr 25th 2010, 08:32 AM #11

- Joined
- Dec 2007
- From
- Ottawa, Canada
- Posts
- 3,184
- Thanks
- 80

- Apr 25th 2010, 08:54 AM #12

- Joined
- Oct 2009
- Posts
- 4,261
- Thanks
- 3

- Apr 25th 2010, 08:57 AM #13

- Joined
- Oct 2009
- Posts
- 4,261
- Thanks
- 3

Nop, this is not a good reason: the conterexample $\displaystyle \sum^\infty_{k=0}\frac{1}{n!}=e\notin\mathbb{Q}$ shows that a sum of only rational numbers can have a non-rational sum...of course, as already noted, the gist here is that the sum is infinite and thus closedness and other related things do not apply here.

Tonio

- Apr 25th 2010, 09:00 AM #14

- Joined
- Oct 2009
- Posts
- 4,261
- Thanks
- 3

- Apr 25th 2010, 09:18 AM #15

- Joined
- Feb 2010
- Posts
- 422

The value of the series depends on the method used to sum it. Everyone knows that the series diverges in the traditional sense (aka. the partial sums are eventually graeater in magnitude than any number). The series is Abel summable to 1/4. The wikipedia article the OP linked discusses other summability methods under which it converges. In some sense, any "reasonable" summation method which sums this series must give 1/4. Hardy has a book on divergent series for anyone interested in these sums and their applications.

(Also, Wilmer, you have wrong series; the signs alternate every term, not every two terms. The partial sums are actually 0,1,-1,2,-2,3,-3,...)

Click on a term to search for related topics.