Consider the following (easy) lemma.

Lemma. There is a subset $Q$ of the positive integers and a fixed constant $N > 0$ such that
1)$Q$ has positive asymptotic density and

2)for each rational numbers $\alpha,\beta$ it results $\alpha \beta^n \in Q$ for at most $N$ positive integers $n$.

Proof (sketch). Take $Q$ as the set of squarefree positive integers and $N=2$.

My question is: For a fixed rational number $r$, can we replace $\alpha \beta^n$ with $\alpha \beta^n + r$ in the lemma above?

Thank you in advance for any help.


Let us write $r=\frac{s}{t}$ with $s$ and $t$ coprime. The set $S$ of positive integers which are square free and congruent to $-s\pmod t$ has positive density (essentially by the same argument which is used to show that the square free integers have positive density). Let $Q=\{\frac{m+s}{t}:m \in S\}$, which also has positive density. Now assume that $\alpha \beta^n+r\in Q$. Then $t\alpha\beta^n\in S$. However, $S$ contains only square free integers, so $t\alpha \beta^n\in S$ can hold for at most two values of $n$ (unless $\beta=\pm 1$, which should be excluded from the question), so $N=2$ works.

| cite | improve this answer | |

Carl Pomerance sketched a solution for me, which I didn't write down in detail, but I think it goes like this:

If $r$ is an integer, we can take $S$ to be the squarefree integers, shifted by $r$, and $N=2$.

If $r$ is not an integer, then if $\beta$ is not an integer there is at most one value of $n$ such that $\alpha\beta^n+r$ is an integer, and we can take $S$ to be the set of all integers, with $N=1$.

Finally, if $r$ is not an integer, and $\beta$ is an integer, then we can write $\alpha\beta^n+r$ as $(st^n+p)/q$ for some integers $p$, $q$, $s$, $t$, with $\gcd(p,q)=1$. Let $U$ be the set of numbers that are squarefree and congruent to $-p$ modulo $q$; this is a set of positive asymptotic density. Let $S$ be the set of numbers of the form $(u+p)/q$, with $u$ in $U$. Then $S$ has positive asymptotic density, and if $(st^n+p)/q$ is in $S$, then $st^n$ is squarefree, so we can use this set $S$, with $N=2$.

| cite | improve this answer | |
  • $\begingroup$ ---If $r$ is not an integer, then if $\beta$ is not an integer there is at most one value of $n$ such that $\alpha\beta^n+r$ is an integer--- Really? Cannot I always solve a system $\alpha\beta^n+r=a,\alpha\beta^{n+1}+r=b$ of two linear equations with two unknowns $\alpha,r$? $\endgroup$ – fedja May 22 '14 at 11:37
  • $\begingroup$ Hmmm. Clearly, I have to rethink this. $\endgroup$ – Gerry Myerson May 22 '14 at 12:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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