MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

This innocent-looking problem came to me some years ago.

These two most basic integration formulas are, of course, disturbingly different, in the eyes of any good mathematicians [ just joking ;-) ].

The challenge is to put them together in a nice manner.

Smells like analytical continuation and the zeta(1) pole is playing a role here, so I don't expect the answer to be trivial. Bonus for anyone using representation theory or p-adic methods [yes, i am thinking about Tate's thesis], or any appearances of cohomology.

share|cite|improve this question
I'm puzzled by the question. Obviously you don't want someone to just say "$\log(x)$ is the limit as $a \to -1$ of $x^{a+1}/(a + 1)$". The word "challenge", and the last sentence of your question, suggest you know something that you're not telling us. It's not really in the spirit of this site to do that... share whatever you know! – Tom Leinster Mar 30 '10 at 7:37
I'm not sure I get the question either, but I find your remark puzzling as well, Tom. Obviously we don't want someone to say "log(x) is the limit as $a \rightarrow -1$ of $\frac{x^{a+1}}{a+1}$" because it isn't true: that limit diverges! – Pete L. Clark Mar 30 '10 at 8:36
The issue can be resolved by starting the integral from 1 to define the antiderivative. Then one takes the limit as a approaches -1 of (x^{a+1} - 1)/(a+1) and this really does equal log x. – Qiaochu Yuan Mar 30 '10 at 8:48
No, I don't know the answer at the moment... one day I shall sit down and try hard at it. The challenge is, $\log(x)$ is not the limit as $a \to -1$ of $x^{a+1}/(a+1)$, at least not in any naive way. Let me clarify the question a little bit. The simplest solution will be to embed $\log(x)$ and $x^n$ and possibly other functions in a nice space (probably Hilbert, etc., and the space shall not be very artificial), such that we can find a smooth functional "A" always sending $x^n$ to the correct anti-derivative [forget the constant factor]. – Bo Peng Mar 30 '10 at 9:04
@Qiaochu: your comment is a perfectly good answer to the question; you might even want to leave it as an answer. The OP seems to want a fancier answer, but I don't yet understand why. – Pete L. Clark Mar 30 '10 at 9:27

First of all, I'll repeat what I said in the comments: writing $\frac{x^h - 1}{h} = \frac{e^{h \log x} - 1}{h}$ makes it fairly clear that as $h \to 0$, this expression tends to $\log x$, and setting $h = a + 1$ this is precisely the desired result.

I should mention that this result is implicit in a certain fact well-known to people who do competition math, which is as follows. Given non-negative real numbers $x_1, ... x_n$, let $A_p(x_1, ... x_n) = \sqrt[p]{ \frac{x_1^p + ... + x_n^p}{n} }$ denote the $p$-power mean for $p \neq 0$. For $p = 0$, define $A_0(x_1, ... x_n) = \sqrt[n]{x_1 ... x_n}$ (the geometric mean, and also the limit as $p \to 0$ of the above).

Theorem (Power Mean Inequality): If $p \le q$, then $A_p \le A_q$.

If you like fancy keywords, then I will bring to your attention that as $p \to \infty$ the $p$-power mean approaches $\text{max}(x_1, ... x_n)$, which one can think of as the "low-temperature limit" of ordinary addition becoming tropical addition. Then $p \to 0$ can be thought of as the "high-temperature limit," in which ordinary addition becomes multiplication instead. Somebody who knows more statistical mechanics than I do (that is, any) can probably tell you the physical significance of this.

share|cite|improve this answer
Nice $e^{h \log x}$ trick, although I cannot see why the well-known power mean inequality will have anything to do with it... – Bo Peng Mar 30 '10 at 10:06
Essentially the same limit is involved in both computations. You should pretend that all those p-power terms turn into logarithms and the pth root turns into an exponential to see the connection. – Qiaochu Yuan Mar 30 '10 at 10:28
As $p \to -\infty$, the $p$-mean approaches the minimum. – Michael Lugo Mar 30 '10 at 12:53
So, rather than "tropical geometry", we should be speaking of "arctic geometry"? – Pete L. Clark Mar 30 '10 at 13:52
Yes. Allen Knutson made a push a while back to convince tropical people that "tropicalization" should refer to deforming away from the low temperature limit, not towards it. But I think most people were nervous about changing terminology in a field that already has too many definitions. – David Speyer Mar 30 '10 at 14:16

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.