A MathsJam classic question asks:

Without using a calculator, which is bigger: eπ or πe?

It’s one of those questions that looks perfectly straightforward: you just take logs and then… oh, but is π bigger than eln(π)? The Mathematical Ninja says “ln(π) is about 1.2, because π is about 20% above e, which means eln(π) is… about 20% more than e, which is about π.” Thanks, Ninja, but that doesn’t solve it.

The trick doesn’t involve logarithms, at least not directly. Instead, you make use of the Taylor expansion of ex, which is 1+x+x22!+x33!+, which is strictly larger than 1+x whenever x is positive.

That 1 is a bit awkward, though: we’re going to want to look at a power that has a 1 in it so that it cancels out. We might naively look at:

eπ1>π, which is a start; however, it only tells us that eπ>eπ, which isn’t what we want. We’ve got an e too many on the left hand side, so let’s try an argument of πe1:

eπe1>πe - now we’re getting somewhere!

eπe>π

eπ>πe, which is what we wanted to find out!

* Thanks to [twit handle = ‘daveinstpaul’] for the reminder about this problem.