For every prime p consider all polynomials f ( x ) with integer coefficients from 1 to p and degree at most p − 1 , such that for all integers x the number f ( 2 x ) − f ( x ) is divisible by p .
Find the sum of all primes p < 1 0 0 0 such that the number of such polynomials is strictly greater than p ⋅ 2 p − 2 .
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.
We can rewrite the problem by considering polynomials with coefficients modulo p , and the condition that for all residues x modulo p , f ( 2 x ) = f ( x ) in the field Z / p Z . However, every such polynomial of degree k ≤ p − 1 should be counted ( p − k ) times, one for each degree from k to p − 1 . If p = 2 , then this condition is equivalent to f ( 0 ) = f ( 1 ) . There are two polynomials modulo p satisfying this condition: 0 and 1 . They correspond to four polynomials with integer coefficients: 2 , 2 x + 1 , 1 , 2 x + 1 . This is strictly greater than 2 ⋅ 2 2 − 2 = 2 , so p = 2 works.
Now we assume that p is odd. Multiplication by 2 modulo p permutes residues modulo p . Suppose k is the smallest positive integer such that 2 k ≡ 1 ( m o d p ) . Then the set of all non-zero residues modulo p consists of k p − 1 cycles, with the ratio of any two elements of each cycle being a power of 2 . The property f ( 2 x ) ≡ f ( x ) ( m o d p ) is equivalent to f having the same values modulo p on all elements in each cycle. By the Lagrange Interpolation Formula (Polynomial Interpolation Theorem) in Z / p Z , for each assignment of values of f ( x ) modulo p there is exactly one polynomial modulo p of degree at most p − 1 . So the number of modular polynomials satisfying the given conditions is p k p − 1 + 1 (the "+1" is for the choice of the value at 0 ). Therefore, the number of integer polynomials is from p k p − 1 + 1 to p k p − 1 + 2 .
Cases p = 3 and p = 5 can be checked directly, and both of them work. Here is the detailed argument for p = 5 . Modular polynomials with given properties are of the form c + d ( x 4 − 1 ) , where c and d are any residues modulo 5 . If d = 0 , we get constant polynomials, so each of them gives 5 integer polynomials. If d\=1,2,3, or 4 , then for any c we get a degree four modular polynomial, which gives one integer polynomial. Altogether, we get 2 5 + 2 0 = 4 5 integer polynomial, which is strictly greater than 5 ⋅ 2 5 − 2 = 4 0 .
We will assume now that p ≥ 7 . Suppose 2 k − 1 = m p . We are going to prove that a prime p works if and only if m = 1 .
Lemma 1. If m = 1 , then the prime p satisfies the condition of the problem.
Proof. We will prove that in this case p k p − 1 + 1 > p 2 p − 2 . Dividing by p and taking to the power k , this is equivalent to p p − 1 > 2 k ( p − 2 ) . Because m = 1 , 2 k ( p − 2 ) = ( p + 1 ) p − 2 . Taking natural logarithm, our inequality is equivalent to ( p − 1 ) ln p > ( p − 2 ) ln ( p + 1 ) . This is equivalent to ln p + ( p − 2 ) ln p > ( p − 2 ) ln ( p + 1 ) , which is equivalent to ln p > ( p − 2 ) ln ( 1 + p 1 ) . For all x > − 1 , ln ( 1 + x ) ≤ x , so ( p − 2 ) ln ( 1 + p 1 ) ≤ p p − 2 < 1 < ln p
Lemma 2. If p > 5 and m > 1 , then p k p − 1 + 2 ≤ p ⋅ 2 p − 2 .
Proof. Suppose k p − 1 = l . First we consider the case k ≤ l . Then k p − 1 ≥ k , so p − 1 − k p − 1 + k − 1 ≤ p − 2 . So 2 k k − 1 ( p − 1 ) + ( k − 1 ) ≤ 2 p − 2 . Because m ≥ 2 , 2 k = m p + 1 > 2 p , thus 2 k − 1 > p . So p k p − 1 + 1 < 2 p − 2 , as desired.
Now suppose l ≤ k . Then k p − 1 ≤ p , thus it is enough to show that p p + 1 ≤ 2 p − 2 . Taking logarithms of both sides, we get lo g 2 p ( p − 1 + 1 ) ≤ p − 2 . This is equivalent to lo g 2 p ≤ p − 1 − 1 . One can check that this is true for primes p ≥ 6 7 . So it remains to check the primes p = 1 1 , 1 3 , 1 7 , 1 9 , 2 3 , 2 9 , 3 7 , 4 1 , 4 3 , 5 3 , 5 9 , 6 1 . It is not hard, but tedious, to find k and l in each case by repeated multiplication by 2 (the job can be simplified greatly sometimes by the Fermat's theorem). We get l = 3 for p = 4 3 and l ≤ 2 in all other cases. The required inequality is then very easy to check, we skip the details for the sake of brevity.
Therefore, the primes p for which the condition is satisfied, are p = 2 , p = 3 , p = 5 , and the primes p ≥ 7 of the form 2 k − 1 . For p < 1 0 0 0 , we just need to check k ≤ 9 . So we get the primes 2 , 3 , 5 , 7 , 3 1 , 1 2 7 . Adding them up, we get 1 7 5 .
Problem Loading...
Note Loading...
Set Loading...
Suppose f ( x ) is of degree n ≤ p − 1 and is equal to i = 0 ∑ n a i x i . Then f ( 2 x ) − f ( x ) = i = 0 ∑ n ( 2 i − 1 ) a i x i . Consider this polynomial mod p . By Lagrange's theorem (in number theory), since this polynomial has degree at most n ≤ p − 1 , then this polynomial has at most p − 1 roots mod p if the polynomial is not 0 mod p . But we know that this polynomial is equal to 0 for all integers x , so this polynomial has p roots mod p . Therefore, this polynomial is equal to 0 mod p , i.e. all its coefficients are 0 mod p . So we have p ∣ ( 2 i − 1 ) a i , i.e. p ∣ 2 i − 1 or p ∣ a i for all 0 ≤ i ≤ n .
We shall ignore the case p = 2 for the moment. Let d be the order of 2 mod p , so d ∣ p − 1 by Fermat's little theorem ( p ∣ 2 p − 1 − 1 ) and Lagrange's theorem (in group theory). Also, we have 2 m ≡ 1 ( m o d p ) iff d ∣ m , so p ∣ 2 i − 1 iff i is a multiple of d . So, when i is a multiple of d (including i = 0 ), we can have any of the p possibilities for a i , but when i is not a multiple of d , then p ∣ a i ⇒ a i = p . So we have altogether p ⌊ d n ⌋ + 1 ways. The degree n of the polynomial f ( x ) may vary, so the total number of polynomials that we are looking for is n = 0 ∑ p − 1 p ⌊ d n ⌋ + 1 . Letting p − 1 = k d for some positive integer k , we can see that this expression simplifies to p k + 1 + d ⋅ j = 1 ∑ k p j = p k + 1 + p − 1 d p ( p k − 1 ) . We want this to be strictly greater than p ⋅ 2 p − 2 , so dividing by p , we get p k + p − 1 d ( p k − 1 ) > 2 p − 2 .
If p + 1 is not a power of 2 , then p + 1 = 2 d . But 2 d ≡ 1 ( m o d p ) , so 2 d > 2 p . Then 2 p − 1 = ( 2 d ) k > ( 2 p ) k = 2 k p k ⇒ 2 p − 2 > 2 k − 1 p k . But 2 p − 2 < p k + p − 1 d ( p k − 1 ) ≤ p k + ( p k − 1 ) < 2 p k , so 2 p k > 2 k − 1 p k ⇒ k = 1 . When k = 1 , then d = p − 1 so p k + p − 1 d ( p k − 1 ) > 2 p − 2 becomes 2 p − 1 > 2 p − 2 , i.e. 8 p > 2 p + 4 . We can easily prove by induction that p > 6 will not work since 8 × 6 < 2 6 + 4 . Therefore, p is at most 5 . We have assumed that p = 2 , and since 3 + 1 is a power of 2 , the only solution for p in this case is p = 5 . When p = 5 , we have d = 4 and k = 1 , so it clearly satisfies the inequality.
If p + 1 is a power of 2 (i.e. p is a Mersenne prime), then 2 q = p + 1 for some prime q (putting p = 2 q − 1 lets us see immediately that q is prime when p is prime). But 2 q ≡ 1 ( m o d p ) , so d ∣ q . But q is prime and d = 1 for all primes p > 2 , so d = q and we have 2 d = p + 1 . So we want p k + p − 1 d ( p k − 1 ) > 2 p − 2 . Let's consider the stricter inequality p k > 2 p − 2 and we shall prove that this is true for all Mersenne primes. Observe that p k > 2 p − 2 ⇔ ( 2 d − 1 ) k > 2 p − 2 ⇔ 2 ( 2 d − 1 ) k > 2 p − 1 = ( 2 d ) k ⇔ ( 2 d ) k ( 2 d − 1 ) k > 2 1 ⇔ ( 1 − 2 d 1 ) k > 2 1 . By Bernoulli's inequality, since k ≥ 1 and − 2 d 1 > − 1 , we have ( 1 − 2 d 1 ) k ≥ 1 + k ( − 2 d 1 ) . So it suffices to show that 1 + k ( − 2 d 1 ) > 2 1 . So we have 1 + k ( − 2 d 1 ) > 2 1 ⇔ 2 d k < 2 1 ⇔ d ⋅ 2 d p − 1 < 2 1 ⇔ d ⋅ 2 d 2 d − 2 < 2 1 ⇔ 2 d − 2 < d ⋅ 2 d − 1 . The final statement is obviously true because d ≥ 2 as established earlier. Therefore, we can conclude that all Mersenne primes satisfy the property in the question.
Finally, we would need to look at p = 2 . If f ( x ) = 1 and f ( x ) = 2 , then f ( 2 x ) − f ( x ) = 0 for all x and it is divisible by 2 . Moreover, when f ( x ) = 2 x + 1 or f ( x ) = 2 x + 2 , we must have f ( 2 x ) − f ( x ) = 2 x which is divisible by 2 for all x . So there are at least 4 such polynomials as described in the question, and since 4 > 2 ⋅ 2 2 − 2 , we count p = 2 as well.
Combining our results, we know that the primes p that satisfy the conditions in the question must be 2 , 5 or a Mersenne prime. For the Mersenne primes, we want p < 1 0 0 0 , so we just need to look at 2 s − 1 for all primes s not more than 9 . This corresponds to 2 2 − 1 = 3 , 2 3 − 1 = 7 , 2 5 − 1 = 3 1 and 2 7 − 1 = 1 2 7 which are all primes, so the answer we are looking for is 2 + 3 + 5 + 7 + 3 1 + 1 2 7 = 1 7 5 .