Unofficial archive — problems, solutions & results © IMC, reproduced with permission.

IMC / 2024 / Problems / Day 2, P10

IMC 2024 · Day 2 · P10

killer

We say that a square-free positive integer nn is almost prime if n∣xd1+xd2+⋯+xdk−kxn \mid x^{d_1} + x^{d_2} + \dots + x^{d_k} - kx for all integers xx, where 1=d1<d2<⋯<dk=n1 = d_1 < d_2 < \dots < d_k = n are all the positive divisors of nn. Suppose that rr is a Fermat prime (i.e. it is a prime of the form 22m+12^{2^m} + 1 for an integer m≥0m \ge 0), pp is a prime divisor of an almost prime integer nn, and p≡1(modr)p \equiv 1 \pmod r. Show that, with the above notation, di≡1(modr)d_i \equiv 1 \pmod r for all 1≤i≤k1 \le i \le k.

(An integer nn is called square-free if it is not divisible by d2d^2 for any integer d>1d > 1.)

(proposed by Tigran Hakobyan, Yerevan State University, Vanadzor, Armenia)

Solution (official)

We first prove the following claims.

Lemma 1. If nn is almost prime then gcd⁡(n,φ(n))=1\gcd(n, \varphi(n)) = 1.

Proof. Assume to the contrary that gcd⁡(n,φ(n))>1\gcd(n, \varphi(n)) > 1 so that there are primes pp and qq dividing nn such that p≡1(modq)p \equiv 1 \pmod q. For 0≤i≤p−20 \le i \le p-2 let hih_i be the number of positive divisors of nn congruent to ii modulo p−1p - 1 and similarly for 0≤j≤q−10 \le j \le q-1 let νj\nu_j denote the number of positive divisors of nn congruent to jj modulo qq. Observe that the polynomial Fn(x)=xd1+xd2+⋯+xdk−kxF_n(x) = x^{d_1} + x^{d_2} + \dots + x^{d_k} - kx defines the zero function on Fp\mathbb{F}_p due to the condition of the problem. On the other hand, Fn(x)=(h1−k)x+∑i≠1hixiF_n(x) = (h_1 - k) x + \sum_{i \ne 1} h_i x^i in Fp[x]\mathbb{F}_p[x], so that p∣hip | h_i for all 0≤i≤p−20 \le i \le p-2, i≠1i \ne 1. It follows that 2ω(n)−1=ν0=h0+hq+h2q+⋯≡0(modp)2^{\omega(n)-1} = \nu_0 = h_0 + h_q + h_{2q} + \dots \equiv 0 \pmod p which is a contradiction (here ω(n)\omega(n) means the number of distinct prime divisors of nn). Therefore our assumption was wrong and the lemma is proved. □\square

Lemma 2. Let qq be a prime number and let hh be a positive integer coprime to q−1q - 1. If ll is the order of hh modulo q−1q - 1, then there exists a∈Fqa \in \mathbb{F}_q such that ahl=aa^{h^l} = a and a−ah+ah2−⋯+(−1)l−1ahl−1≠0a - a^h + a^{h^2} - \dots + (-1)^{l-1} a^{h^{l-1}} \ne 0 Proof. Observe that ahl=aa^{h^l} = a for any a∈Fqa \in \mathbb{F}_q since q−1∣hl−1q - 1 \mid h^l - 1. On the other hand, the numbers h0,h1,…,hl−1h^0, h^1, \dots, h^{l-1} leave different remainders upon division by q−1q - 1 and therefore the polynomial f(x)=x−xh+xh2−⋯+(−1)l−1xhl−1f(x) = x - x^h + x^{h^2} - \dots + (-1)^{l-1} x^{h^{l-1}} defines a function on Fq\mathbb{F}_q, which is not identically zero. Hence the existence of an element with the required properties is proved. □\square

Lemma 3. If nn is almost prime then for any primes pp and qq dividing nn, the order of pp modulo q−1q - 1 is an odd number.

Proof. Observe that due to Lemma 1 the order ll of pp modulo q−1q - 1 is well defined and assume to the contrary that ll is an even number. According to Lemma 2 there exists a∈Fqa \in \mathbb{F}_q such that apl=aa^{p^l} = a and f(a)≠0f(a) \ne 0, where f(x)=x−xp+xp2−⋯+(−1)l−1xpl−1f(x) = x - x^p + x^{p^2} - \dots + (-1)^{l-1} x^{p^{l-1}}. Let us consider the sequence (ai)i=0l⊂Fq(a_i)_{i=0}^{l} \subset \mathbb{F}_q defined by a0=aa_0 = a and ai+1=−aipa_{i+1} = -a_i^p for 0≤i≤l−10 \le i \le l-1. Notice that since ll is even by the assumption, we have al=a0pl=a0a_l = a_0^{p^l} = a_0. It follows that ∑i=0l−1∑d∣naid=∑i=0l−1(∑d∣npaid+∑d∣npaipd)=∑i=0l−1∑d∣np(ai+1d+aipd)=0,\sum_{i=0}^{l-1} \sum_{d \mid n} a_i^d = \sum_{i=0}^{l-1} \left( \sum_{d \mid \frac{n}{p}} a_i^d + \sum_{d \mid \frac{n}{p}} a_i^{pd} \right) = \sum_{i=0}^{l-1} \sum_{d \mid \frac{n}{p}} \bigl( a_{i+1}^d + a_i^{pd} \bigr) = 0, since dd is always odd being a divisor of nn (Recall that gcd⁡(n,φ(n))=1\gcd(n, \varphi(n)) = 1 due to Lemma 1, so that nn is odd, except the trivial case n=2n = 2), and ai+1=−aipa_{i+1} = -a_i^p for all 0≤i≤l−10 \le i \le l-1. On the other hand, according to the condition of the problem, ∑d∣naid=kai\sum_{d \mid n} a_i^d = k a_i in Fq\mathbb{F}_q for all ii, which shows that kf(a)=k∑i=0l−1ai=∑i=0l−1kai=∑i=0l−1∑d∣naid=0k f(a) = k \sum_{i=0}^{l-1} a_i = \sum_{i=0}^{l-1} k a_i = \sum_{i=0}^{l-1} \sum_{d \mid n} a_i^d = 0 in Fq\mathbb{F}_q which is impossible, since f(a)≠0f(a) \ne 0 by construction and k=2ω(n)−1k = 2^{\omega(n)-1}

is coprime to qq. The attained contradiction shows that our assumption was wrong and concludes the proof of the lemma. □\square

Let us get back to the problem. Suppose that p∣np | n is prime and r=22m+1r = 2^{2^m} + 1 is a Fermat's prime such that p≡1(modr)p \equiv 1 \pmod r. If qq is any prime divisor of nn, then by Lemma 3 we have that ql≡1(modp−1)q^l \equiv 1 \pmod{p-1} for some odd ll, so that ql≡1(modr)q^l \equiv 1 \pmod r and therefore q=qgcd⁡(l,r−1)≡1(modr)q = q^{\gcd(l, r-1)} \equiv 1 \pmod r. Hence d≡1(modr)d \equiv 1 \pmod r for any divisor dd of nn. □\square

How the field did

contestants scored
397
average (of 10)
0.26
solved (≥ 80%)
2.0%
near-0 (≤ 10%)
96.7%
discrimination
0.29

Score distribution (field cohort)

Computed on contestants with a meaningful total (field cohort); discrimination is the corrected item–total correlation.

Similar problems

IMC 2016 · Day 1 · P5killeravg 0.3/10 · solved 2% · near-0 96% · disc 0.29
IMC 1999 · Day 2 · P12killeravg 0.4/10 · solved 2% · near-0 91% · disc 0.30
IMC 2019 · Day 1 · P4killeravg 0.4/10 · solved 2% · near-0 92% · disc 0.33
IMC 2001 · Day 1 · P4killeravg 0.3/10 · solved 2% · near-0 95% · disc 0.32