4 ms·
Tell HN: 2^4 == 4^2, extended to rationals
Last week there was a discussion about 2^4 == 4^2 (my search-fu is letting me down - I can't find it again, or I'd link it). It proved that that is the only pair of integers with that relationship.
But what about rationals?
First, remember that (a^b)^c) == a^(bc).
Let a = (3/2)^2 and b = (3/2)^3. Then a^b == b^a. (a and b are rational, but a^b is not.)
More generally, for any natural number n, let a = ((n+1)/n)^n and b = ((n+1)/n)^(n+1). Then a^b == b^a.
Note that the first element of this series has a = 2 and b = 4. Also note that as n increases, the series converges to both a and b equal to e.
But when I said "for any natural number n", that was actually an unnecessary restriction. The math works for any n - it doesn't have to be an integer. We could, for instance, use n = 1/2. Then a = ((3/2)/(1/2))^(1/2) and b = ((3/2)/(1/2))^(3/2). a and b are no longer rational, but a^b still equals b^a.
- gus_massa 4y agoInteger version: https://news.ycombinator.com/item?id=35635491 https://news.ycombinator.com/item?id=35635491 (520 points | 2 days ago | 156 comments) It (implicitly) shows why the e number is important and why it appears as the limit of your example. Did you prove that this are all the case? Or this are only the cases you found and the general case is still open?
- roninghost 4y agoThere are infinete number of rationals that satisfy the condition. The integer uniquenes proof relied on the shape of the function between 1 and e non inclusive, and 2 is the only integer in the open interval.
- AnimalMuppet 4y agoNo, I did not prove that this was all the cases - not even that this was all the rational cases. I tried, years ago (I literally made this post by trying to recall what I had found over 40 years ago), but I did not find a proof that this was all the possibilities.
- AnimalMuppet 4y agoAfter some time to think, yes, I'm pretty sure that it's all solutions. The shape of the function in the article means that there is only one b for any a. The formula I gave gives a solution, so it must be the solution. If you can find an n to generate an a, the formula must give you the correct b. (As n goes from 0 to infinity, a goes from 1 to e, so the formula covers the entire range of possibilities.) In my formula, however, n may be irrational, even for some rational a. I haven't proven that.
- nmmnthrowaway 4y agoI think a more interesting problem to consider after solving n^m = m^n is solving n^m = m^n + 1 in the natural numbers.
- mahalex 4y agoIf we exclude 0, the only solutions are 2^1 = 1^2 + 1 and 3^2 = 2^3 + 1
- nmmnthrowaway 4y agoThat is correct. Why are there no others?
- mahalex 4y agoBecause Catalan’s conjecture.
- yarg 4y ago> Catalan's conjecture was proven by Preda Mihăilescu in April 2002. Cool, so the conjecture's a theorem. https://uni-göttingen.academia.edu/PredaMihailescu https://xn--uni-gttingen-8ib.academia.edu/PredaMihailescu
- hgsgm 4y agoWhich has an extremely complicated proof :-(
- mahalex 4y agoI wouldn’t call it extremely complicated; it is much simpler than the proof of FLT.
- nmmnthrowaway 4y agoYes, but there is a much simpler proof.
- dvt 4y agoI feel rationals is probably the easiest case to prove, still clever though! Interesting (though not surprising) how there's a countably infinite number of rationals that satisfy this.
- bbischof 4y agoProbably this is the article you saw https://link.medium.com/bf81IpGM8yb https://link.medium.com/bf81IpGM8yb
- gslin 4y agoFor general solution you can check https://math.stackexchange.com/questions/3072828/why-didnt-they-simplify-xy-yx-to-x-y https://math.stackexchange.com/questions/3072828/why-didnt-t..., which is related to Lambert W function.
- sdwvit 4y agoDo it for imaginary components :tada: