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

Say we have three infinite sequences $\{a_i\},\{b_i\},\{c_i\}$ of natural numbers, satisfying the equations $$a_1+b_1=c_1,\dots, a_n+b_n=c_n,\dots $$ Assume further that $gcd(a_i,b_i,c_i)=1$ for each $i$ and that $(a_i,b_i,c_i)\neq (a_j,b_j,c_j)$ for all $i,j$. Now let's define $S$ as the set of primes $p$ which divide $a_ib_ic_i$ for at least one $i$. From the S-unit theorem we know that $S$ has to be infinite.

Now the question is: Can $S$ be sparse? This can be taken to mean Dirichlet density zero, for example.

I haven't thought much about this but there are reasons to believe the answer is yes, indeed if there are infinitely many Mersenne primes $q=2^p-1$ then the equations $1+q=2^p$ give such a sparse $S$. However, I am looking for an unconditional result.

share|cite|improve this question
Instead of asking about density $0$, it would make more sense to ask whether for any $f:\mathbb N\to\mathbb N$ we may have $S=\{p_j\}$ with $p_{j+1}>f(p_j)$. All the counterexamples below fail dramatically even for $f(x)=x^2$. – fedja Apr 29 '10 at 3:53
Fedja, that is a very interesting question. I cannot say anything meaningful except that the Mersenne primes again suggest that the answer should be yes.. – Gjergji Zaimi Apr 29 '10 at 5:47
up vote 10 down vote accepted

Yes, in fact, you can make $S$ grow as slowly as you like. This follows, for example, from the fact that there exist 3-term arithmetic progressions of primes $(p,q,r)$ with $\min(p,q,r)$ arbitrarily large. For each such arithmetic progression, you can take $(a_i,b_i,c_i)=(p,r,2q)$. Now just choose these arithmetic progressions one at a time, making sure that the primes involved each time are much larger than any primes appearing earlier.

share|cite|improve this answer
Very nice, I knew I should have thought about it some more :) – Gjergji Zaimi Mar 27 '10 at 23:08

Alternatively, every prime dividing $2^{2^k}+1$ must be 1 mod $2^k$, hence must be at least $2^k$. So if you take your triples of the form $(1,2^{2^k},2^{2^k}+1)$ with the values of $k$ sufficiently sparse, you get what you want.

share|cite|improve this answer
That's a lot easier to justify than what I was proposing! – Bjorn Poonen Mar 28 '10 at 5:57

An even simpler answer: take triples of the form $(2,n!-1,n!+1)$ for a sufficiently sparse sequence of values of $n$.

share|cite|improve this answer

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.