## 1 = 0

A forum for good logic/math puzzles.

Moderators: jestingrabbit, Moderators General, Prelates

gogobera
Posts: 3
Joined: Tue Nov 21, 2006 12:08 am UTC
Contact:

### 1 = 0

I didn't make a solution thread when I should've, but now I have. It's here: http://forums.xkcd.com/viewtopic.php?t=660

I've always enjoyed false proofs, but the only one I saw on here was an algebraic one. Here's one I stubled upon way back when I was learning the Calculus (do people mind TeX notation? It shouldn't be too cryptic.)

To evaluate the integral of 1/x, we use (the much loved) integration by parts. Recall that integration by parts exploits the following:

\int u dv = uv - \int v du.

Let u = 1/x, then du = -1/x^2 dx. Let v = x; therefore, dv = dx. We write
\int 1/x dx = \int u dv
\int 1/x dx = uv - \int v du
\int 1/x dx = (1/x)x - \int x (-1/x^2) dx
\int 1/x dx = 1 + \int 1/x dx.

Subtracting the original integral from both sides leaves 0 = 1. QED Note, this solves the pesky problem of having to ever evaluate (or even define) the log base e (an irrationally based logarithim, if ever one was!).

Anyone seen that before? I haven't.
Last edited by gogobera on Tue Nov 21, 2006 1:51 pm UTC, edited 1 time in total.

svk1325
Posts: 139
Joined: Mon Jul 17, 2006 9:10 am UTC
Location: Take a guess...
Contact:
I get the feeling that this has something to do with 1/x and -1/x^2 being discontinuous at x=0.
"Insanity in a measured dose is a good thing - the difficulty lies in the measurement."

Erasmus
Posts: 26
Joined: Wed Sep 27, 2006 10:13 am UTC
When you evaluate an indefinite integral, ... there's an arbitrary constant:

\int 1/x dx
= \int u dv
= uv - \int v du + C
= x (1/x) - \int x (-1/x^2) dx + C
= 1 + \int 1/x dx + C

Therefore
0 = 1 + C

Therefore C = -1.

A "reverse" way to write the same "proof" is to start with u and v and evaluate d(uv)/dx using the product rule (which is the inverse of integration by parts):

u = 1/x => du = -1/x^2 dx
v = x => dv = dx
d(uv) = u dv + v du = dx/x - x dx/x^2 = 0

Your proof at this point involves integrating both sides to get uv = \int u dv + \int v du, but when you integrate, the arbitrary constant has to come in: it's really uv = \int u dv + \int v du + C.

One last way of looking at it: do a definite instead of indefinite integral:

\int_a^b 1/x dx
= [ x/x ]_a^b + \int_a^b 1/x dx (a, b > 0)

Doing the same subtract-from-both-sides we get
0 = [ 1 ]_a^b
which is quite obviously the case.

SpitValve
Not a mod.
Posts: 5130
Joined: Tue Sep 26, 2006 9:51 am UTC
Location: Lower pork village
yay! maths is saved!

gogobera
Posts: 3
Joined: Tue Nov 21, 2006 12:08 am UTC
Contact:
Seems I oughtta have created a "solution thread" ... hmm. oops. Ok, doing that now....

Pathway
Leon Sumbitches...?
Posts: 647
Joined: Sun Oct 15, 2006 5:59 pm UTC
Too late. It's been quite convincingly solved.

Minerva
Posts: 947
Joined: Sat Nov 11, 2006 2:58 pm UTC
Location: Australia
Contact:
Can anyone tell me why the well known result for this problem - Ln(x) + C - doesn't fall out of the solution, using the simple integration by parts as above?

Cosmologicon
Posts: 1806
Joined: Sat Nov 25, 2006 9:47 am UTC
Location: Cambridge MA USA
Contact:
Because integration by parts isn't the way to solve this one?

Generally speaking, integration by parts just produces another true statement. Sometimes the integral in that statement is simpler and you can solve the integral, but not always. Another great example is the integral of exp(x)sin(x)dx. You can solve it using integration by parts twice even though the integral you get out of it is no easier to evaluate!