Maths Olympiad Prep

Library / /62 of 383

Number theory Difficulty 8.0 National Olympiad, round 2 Prove it IMO

Prove that for every prime p>100p > 100 and every integer rr there exist two integers aa and bb such that pp divides a2+b5ra^{2} + b^{5} - r.

Solutions — 2

Solution 1

Fix pp, and let P={0,1,,p1}\mathcal{P} = \{0, 1, \ldots, p-1\} be the set of residue classes modulo pp. For every rPr \in \mathcal{P}, let Sr={(a,b)P×P:a2+b5r}S_{r} = \{(a, b) \in \mathcal{P} \times \mathcal{P} : a^{2} + b^{5} \equiv r\}, and let sr=Srs_{r} = |S_{r}|. Our aim is to prove sr>0s_{r} > 0 for all rPr \in \mathcal{P}.
We will use the well-known fact that for every residue class rPr \in \mathcal{P} and every positive integer kk, there are at most kk values xPx \in \mathcal{P} such that xkrx^{k} \equiv r.

Lemma. Let NN be the number of quadruples (a,b,c,d)P4(a, b, c, d) \in \mathcal{P}^{4} for which a2+b5c2+d5a^{2} + b^{5} \equiv c^{2} + d^{5}. Then
N=rPsr2 N = \sum_{r \in \mathcal{P}} s_{r}^{2}
and
Np(p2+4p4). N \leq p(p^{2} + 4p - 4).

Proof. (a) For each residue class rr there exist exactly srs_{r} pairs (a,b)(a, b) with a2+b5ra^{2} + b^{5} \equiv r and srs_{r} pairs (c,d)(c, d) with c2+d5rc^{2} + d^{5} \equiv r. So there are sr2s_{r}^{2} quadruples with a2+b5c2+d5ra^{2} + b^{5} \equiv c^{2} + d^{5} \equiv r. Taking the sum over all rPr \in \mathcal{P}, the statement follows.

(b) Choose an arbitrary pair (b,d)P(b, d) \in \mathcal{P} and look for the possible values of a,ca, c.
1. Suppose that b5d5b^{5} \equiv d^{5}, and let kk be the number of such pairs (b,d)(b, d). The value bb can be chosen in pp different ways. For b0b \equiv 0 only d=0d = 0 has this property; for the nonzero values of bb there are at most 5 possible values for dd. So we have k1+5(p1)=5p4k \leq 1 + 5(p-1) = 5p - 4.
The values aa and cc must satisfy a2c2a^{2} \equiv c^{2}, so a±ca \equiv \pm c, and there are exactly 2p12p - 1 such pairs (a,c)(a, c).
2. Now suppose b5≢d5b^{5} \not\equiv d^{5}. In this case aa and cc must be distinct. By (ac)(a+c)=d5b5(a-c)(a+c) = d^{5} - b^{5}, the value of aca-c uniquely determines a+ca+c and thus aa and cc as well. Hence, there are p1p-1 suitable pairs (a,c)(a, c).
Thus, for each of the kk pairs (b,d)(b, d) with b5d5b^{5} \equiv d^{5} there are 2p12p - 1 pairs (a,c)(a, c), and for each of the other p2kp^{2} - k pairs (b,d)(b, d) there are p1p-1 pairs (a,c)(a, c). Hence,
N=k(2p1)+(p2k)(p1)=p2(p1)+kpp2(p1)+(5p4)p=p(p2+4p4). N = k(2p - 1) + (p^{2} - k)(p - 1) = p^{2}(p - 1) + kp \leq p^{2}(p - 1) + (5p - 4)p = p(p^{2} + 4p - 4).

To prove the statement of the problem, suppose that Sr=S_{r} = \emptyset for some rPr \in \mathcal{P}; obviously r≢0r \not\equiv 0. Let T={x10:xP{0}}T = \{x^{10} : x \in \mathcal{P} \setminus \{0\}\} be the set of nonzero 10th powers modulo pp. Since each residue class is the 10th power of at most 10 elements in P\mathcal{P}, we have Tp1104|T| \geq \frac{p-1}{10} \geq 4 by p>100p > 100.
For every tTt \in T, we have Str=S_{tr} = \emptyset. Indeed, if (x,y)Str(x, y) \in S_{tr} and tz10t \equiv z^{10} then
(z5x)2+(z2y)5t1(x2+y5)r (z^{-5}x)^{2} + (z^{-2}y)^{5} \equiv t^{-1}(x^{2} + y^{5}) \equiv r
so (z5x,z2y)Sr(z^{-5}x, z^{-2}y) \in S_{r}.
So, there are at least p1104\frac{p-1}{10} \geq 4 empty sets among S1,,Sp1S_{1}, \ldots, S_{p-1}, and there are at most p4p-4 nonzero values among s0,s2,,sp1s_{0}, s_{2}, \ldots, s_{p-1}. Then by the AM-QM inequality we obtain
N=rPrTsr21p4(rPrTsr)2=P×P2p4=p4p4>p(p2+4p4), N = \sum_{r \in \mathcal{P} \setminus rT} s_{r}^{2} \geq \frac{1}{p-4}\left(\sum_{r \in \mathcal{P} \setminus rT} s_{r}\right)^{2} = \frac{|\mathcal{P} \times \mathcal{P}|^{2}}{p-4} = \frac{p^{4}}{p-4} > p(p^{2} + 4p - 4),
which is impossible by the lemma.

Solution 2

If 5p15 \nmid p-1, then all modulo pp residue classes are complete fifth powers and the statement is trivial. So assume that p=10k+1p = 10k + 1 where k10k \geq 10. Let gg be a primitive root modulo pp.

We will use the following facts:
(F1) If some residue class xx is not quadratic then x(p1)/21(modp)x^{(p-1)/2} \equiv -1 \pmod{p}.
(F2) For every integer dd, as a simple corollary of the summation formula for geometric progressions,
i=02k1g5di{2kif 2kd0if 2kd(modp) \sum_{i=0}^{2k-1} g^{5di} \equiv \begin{cases} 2k & \text{if } 2k \mid d \\ 0 & \text{if } 2k \nmid d \end{cases} \pmod{p}

Suppose that, contrary to the statement, some modulo pp residue class rr cannot be expressed as a2+b5a^{2} + b^{5}. Of course r≢0(modp)r \not\equiv 0 \pmod{p}. By (F1) we have (rb5)(p1)/2=(rb5)5k1(modp)(r - b^{5})^{(p-1)/2} = (r - b^{5})^{5k} \equiv -1 \pmod{p} for all residue classes bb.
For t=1,2,,k1t = 1, 2, \ldots, k-1 consider the sums
S(t)=i=02k1(rg5i)5kg5ti S(t) = \sum_{i=0}^{2k-1} (r - g^{5i})^{5k} g^{5ti}
By the indirect assumption and (F2),
S(t)=i=02k1(r(gi)5)5kg5tii=02k1(1)g5tii=02k1g5ti0(modp) S(t) = \sum_{i=0}^{2k-1} (r - (g^{i})^{5})^{5k} g^{5ti} \equiv \sum_{i=0}^{2k-1} (-1) g^{5ti} \equiv -\sum_{i=0}^{2k-1} g^{5ti} \equiv 0 \pmod{p}
because 2k2k cannot divide tt.
On the other hand, by the binomial theorem,
S(t)=i=02k1(j=05k(5kj)r5kj(g5i)j)g5ti=j=05k(1)j(5kj)r5kj(i=02k1g5(j+t)i)j=05k(1)j(5kj)r5kj{2kif 2kj+t0if 2kj+t(modp) \begin{aligned} S(t) &= \sum_{i=0}^{2k-1} \left( \sum_{j=0}^{5k} \binom{5k}{j} r^{5k-j} (-g^{5i})^{j} \right) g^{5ti} \\ &= \sum_{j=0}^{5k} (-1)^{j} \binom{5k}{j} r^{5k-j} \left( \sum_{i=0}^{2k-1} g^{5(j+t)i} \right) \\ &\equiv \sum_{j=0}^{5k} (-1)^{j} \binom{5k}{j} r^{5k-j} \begin{cases} 2k & \text{if } 2k \mid j+t \\ 0 & \text{if } 2k \nmid j+t \end{cases} \pmod{p} \end{aligned}
Since 1j+t<6k1 \leq j + t < 6k, the number 2k2k divides j+tj + t only for j=2ktj = 2k - t and j=4ktj = 4k - t. Hence,
0S(t)(1)t((5k2kt)r3k+t+(5k4kt)rk+t)2k(modp)(5k2kt)r2k+(5k4kt)0(modp) \begin{gathered} 0 \equiv S(t) \equiv (-1)^{t} \left( \binom{5k}{2k-t} r^{3k+t} + \binom{5k}{4k-t} r^{k+t} \right) \cdot 2k \pmod{p} \\ \binom{5k}{2k-t} r^{2k} + \binom{5k}{4k-t} \equiv 0 \pmod{p} \end{gathered}
Taking this for t=1,2t = 1, 2 and eliminating rr, we get
0(5k2k2)((5k2k1)r2k+(5k4k1))(5k2k1)((5k2k2)r2k+(5k4k2))=(5k2k2)(5k4k1)(5k2k1)(5k4k2)=(5k)!2(2k1)!(3k+2)!(4k1)!(k+2)!((2k1)(k+2)(3k+2)(4k1))=(5k)!22k(5k+1)(2k1)!(3k+2)!(4k1)!(k+2)!(modp) \begin{aligned} 0 &\equiv \binom{5k}{2k-2} \left( \binom{5k}{2k-1} r^{2k} + \binom{5k}{4k-1} \right ) - \binom{5k}{2k-1} \left( \binom{5k}{2k-2} r^{2k} + \binom{5k}{4k-2} \right ) \\ &= \binom{5k}{2k-2} \binom{5k}{4k-1} - \binom{5k}{2k-1} \binom{5k}{4k-2} \\ &= \frac{(5k)!^{2}}{(2k-1)! (3k+2)! (4k-1)! (k+2)!} \left( (2k-1)(k+2) - (3k+2)(4k-1) \right ) \\ &= \frac{-(5k)!^{2} \cdot 2k(5k+1)}{(2k-1)! (3k+2)! (4k-1)! (k+2)!} \pmod{p} \end{aligned}
But in the last expression none of the numbers is divisible by p=10k+1p = 10k + 1, a contradiction.

Want a route through all this instead of an archive? The track puts 2,000 problems in a working order, from AMC 10 level to the IMO shortlist.

Source: MathNet, licensed CC-BY-4.0. Statement reproduced verbatim; metadata (topic, difficulty) added by this project.