# The Universe of Discourse

Wed, 02 Jan 2019

Observing three small examples where the digital sum of !!n^k!! was equal to !!n!!, I said:

I supposed that there were few other examples, probably none … Still I hoped there might be one or two…

but in fact there are a great many; for example the digital sum of !!217^{22}!! is !!217!!.

Dave McKee asked what my intuition was. It was something like this. Fix !!n!! and consider the sequence !!n^2, n^3, …!!. The elements become increasingly sparse, and for their digital sum to equal !!n!! requires increasingly improbable coincidences. For example, taking !!n=3!!, we would need to have !!k!! such that $$3^k = 2·10^p + 1.$$ While I can't immediately rule out this possibility, it seems extremely unlikely. This also resembles the Catalan conjecture, which is that the only nontrivial solution of $$\left\lvert2^i - 3^j\right\rvert = 1$$ is !!i=3, j=2!!. The Catalan conjecture is indeed true, but it was an open problem for 158 years.

[ Addendum 20190205: I have since learned that the Catalan conjecture actually concerns the more general claim about the equation !! \left\lvert a^i - b^j\right\rvert = 1!!. The special case of !!a=2, b=3!! turns out to be elementary. ]

I was quite mistaken, so what went wrong? First, there is one nontrivial solution to the Catalan conjecture, and here we have not one sequence of !!n^k!! but an infinite family, each of which might have an exceptional solution or two. And second, the case of !!n=3!! is atypically bleak. For !!3^k!! to have a digital sum of !!3!! is a tough order because there are very few sequences of digits whose sum is 3. But the number of sequences of !!d!! digits whose sum is !!n!! grows quite rapidly with !!n!!, and for small !!n!! is very misleading.