Does This Polynomial Even Exist?

Find the number of positive integers n n such that 2 n 100 2 \leq n \leq 100 and there exists a polynomial f ( x ) f(x) with real coefficients and degree < n <n such that for all integers x , x, f ( x ) f(x) is an integer if and only if x x is not a multiple of n . n.


The answer is 35.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

1 solution

Firstly I would like to point out that there is a mistake in the problem. It should be 2 < n 100 2<n\le100 rather than the other way round. The answer would be 35 if it meant this exactly.

The problem will be split into two parts: - The case when n n has two or more distinct prime factors. - The case when n n can be written as p a p^a where p p is prime and a a is a positive integer larger than 1.

Now for the first case we shall show that there cannot exist a polynomial for n n . Assume there is such a polynomial for n n . Let n = p 1 δ 1 p 2 δ 2 p k δ k n=p_1^{\delta_1}p_2^{\delta_2}\cdots p_k^{\delta_k} . Now let f ( x ) f(x) have denominator q > 1 q>1 . Now using Newtonian interpolation which is defined for a polynomial of degree < n <n f ( ω + n ) = i = 0 n 1 ( 1 ) n 1 i ( n i ) f ( ω + i ) f(\omega+n) =\sum_{i = 0}^{n-1}(-1)^{n-1-i}\binom{n}{i}f(\omega+i) For some 1 s k 1\le s\le k we therefore have f ( p s δ s + n ) = i = 0 n 1 ( 1 ) n 1 i ( n i ) f ( p s δ s + i ) f(p_s^{\delta_s}+n) =\sum_{i = 0}^{n-1}(-1)^{n-1-i}\binom{n}{i}f(p_s^{\delta_s}+i) . Now note that the LHS is an integer. All terms on the right hand side are also integers maybe except for ( n n p s δ s ) f ( n ) \binom n{n-p^{\delta_s}_s}f(n) term. However this must mean that it is an integer. Now noting that p s ( n p s δ s ) p s q p_s\nmid \binom{n}{p^{\delta_s}_s}\implies p_s \nmid q . This holds for any s s (within limits) therefore we have gcd ( n , q ) = 1 \gcd(n, q)=1 . Finally take ω = 1 \omega=1 for the interpolation formula and we recieve f ( 1 + n ) = i = 0 n 1 ( 1 ) n 1 i ( n i ) f ( 1 + i ) f(1+n) =\sum_{i = 0}^{n-1}(-1)^{n-1-i}\binom{n}{i}f(1+i) . Once again, on the LHS and RHS, all terms are integers except possible for ( n 1 ) f ( n ) \binom n1 f(n) term. However this also must be an integer. So we can say that q n q\nmid n . But gcd ( q , n ) = 1 q = 1 \gcd(q, n)=1\implies q=1 thus a contradiction. Therefore n n cannot have two or more distinct prime factors.

Finally, for any n = p a n=p^a we have a polynomial that satisfies the condition, namely f ( x ) = p a 1 ( x 1 ) ( x 2 ) ( x p a + 1 ) ( p a ) ! f(x) =\frac{p^{a-1}(x-1)(x-2)\ldots(x-p^a+1)}{(p^a)!} . This can easily be checked to show that it satisfies the conditions.

Counting all prime powers between 2 2 and 100 100 we get 34 \cancel{\boxed{34}} . (See edit below)

Note : I do believe that this problem is one of the best on brilliant.org. Trying to solve it has opened my mind up to new skills and techniques and inspired me to learn more! The total time I spend on this problem is around 15 hours. But hey exams are over!


EDIT Now that the question's answer has been modified to suit the original wording we can ignore my note at the top. As 2 2 is prime, we add one to our count which gives us 35 \boxed{35} .

@Sreejato Bhattacharya Can you review this and see if the answer should be 35?

Calvin Lin Staff - 6 years, 12 months ago

Log in to reply

I think he got 34 34 probably because he ignored the case n = 2 n=2 (for n = 2 , n=2, f ( x ) = n 1 2 f(x)=\dfrac{n-1}{2} works). Counting prime powers including 2 , 2, I get 35. 35.

EDIT: Please ignore the above nonsense. I shouldn't comment before reading carefully. Let me check them again.

EDIT 2: Yeah, 35 35 seems right. So sorry. Sincere apologies from my part.

Sreejato Bhattacharya - 6 years, 12 months ago

A small fix: You're taking it for granted that the coefficients of f ( x ) f(x) are rational. This is almost trivial to notice (perhaps that's why you forgot to mention it), since if { a 1 , a 2 , , a k } \{a_1, a_2, \cdots , a_k\} is any set of integers for which f ( x ) Z f(x) \in \mathbb{Z} where k = deg ( f ( x ) ) + 1 , k= \deg (f(x))+1, by Lagrange Interpolation, f ( x ) = i = 1 k f ( a i ) j i x a j a i a j f(x) = \displaystyle \sum_{i=1}^{k} f(a_i) \displaystyle \prod_{j \neq i} \dfrac{x-a_j}{a_i-a_j} whose coefficients are clearly rational.

Otherwise, really nice solution! How I did it was by considering L P ( x ) , L P(x), where L L is the lcm of the denominators of the coefficients, and then for some prime p , p, looking at v p ( L P ( 0 ) ) . v_p (LP(0)). I'll be posting my solution shortly.

Sreejato Bhattacharya - 6 years, 12 months ago

i just included 1 and got the answer 36.I forgot I have to find all n≥2 and got the whole sum wrong :(

Souryajit Roy - 6 years, 11 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...