6

This question follows a previous one

If $n$ and $r$ are coprime then $a_{n,r}=\frac{1}{n}\binom{n}{r}$ is integer but this is not a necessary condition.

Question: what is a necessary and sufficient condition on $n$ and $r$ for $a_{n,r}$ to be integer?

I know that:

$$ \binom{n}{r}\equiv 0 \mod \frac{n}{\gcd {(r,n)}} $$

or in other words $a_{n,r}.\gcd {(r,n)}$ is always an integer, therefore if $n$ and $r$ are coprime then $a_{n,r}$ is is integer.

If $r=n$ or $r=0$, that is trivial. Let us have $1\le r\le n-1$. Let $p$ be a prime and $v_p(n)$ the largest exponent of $p$ such that $p^{v_p(n)}$ divides $n$

if $v_p(n)\ge v_p(r)$ then $v_p(n-r)= v_p(r)$ since $1\le r\le n-1$. Then if $r$ and $n-r$ are written in base $p$ then when adding $r$ and $n-r$ there are $v_p(n)- v_p(r)$ carries, then from Kummer's theorem $v_p(\binom{n}{r})=v_p(n)- v_p(r) $

if $v_p(n)\lt v_p(r)$ then obiously $v_p(n)- v_p(r)\lt 0 \le v_p(\binom{n}{r})$

Then whatever $n,r$ the inequality $v_p(\binom{n}{r})+v_p(r) \ge v_p(n)$ holds and the inequality $v_p(\binom{n}{r})+v_p(r) \ge v_p(r)$ also holds obviously.

Then $v_p(\binom{n}{r})+v_p(r) \ge \max(v_p(r),v_p(n))=v_p(\ lcm(r,n))$.

That is $r.\binom{n}{r}$ is multiple of $\ lcm(r,n)$.

Since $n.r= \ lcm{(r,n)}. \gcd {(r,n)}$, that is $\binom{n}{r}.\gcd {(r,n)}$ is divisible by $n$ and since $n$ is divisible by $\gcd {(r,n)}$, that is $\binom{n}{r}$ is divisible by $\frac{n}{\gcd {(r,n)}}$.

EDIT I just realized that this question is already here. The characterization of the $n,r$ such that $\gcd (n,r) \gt1$ and $ \binom{n}{r}\equiv 0 \mod {n} $ seems quite complex.

René Gy
  • 3,716
  • 1
    Maybe ask the more complete question of what are necessary and sufficient conditions for that divisibility. (The way you wrote it there may be some answer which gives a necessary but not sufficient condition, which would not be as complete a result.) – coffeemath Nov 11 '15 at 23:16
  • Ok I will make an edit – René Gy Nov 12 '15 at 00:03
  • Another possible improvement: What you call $a_n$ might better be denoted say $a_{n,r}$ since you're asking about necessary and sufficient conditions on the pair $(n,r)$ not just on $n$ with fixed $r.$ Also you might restrict to the nontrivial part $2 \le r \le n-2.$ – coffeemath Nov 12 '15 at 01:57
  • Right. I used the notation from the previous question. – René Gy Nov 12 '15 at 07:46

0 Answers0