32. Introduce the congruence notation: Use ab(modn) to denote ba−1(modn), where a−1 is the modular inverse of a modulo n (note that this requires (a,n)=1).
Notice that, under the above fractional congruence notation, we only need to prove: fp(x)−fp(y)≡0(modp3). Since
fp(x)−fp(y)=k=1∑p−1((px+k)21−(py+k)21)=k=1∑p−1(px+k)2(py+k)2p2(y2−x2)+p⋅2k(y−x)
Therefore, we only need to prove: □
k=1∑p−1(px+k)2(py+k)2p(y2−x2)+2k(y−x)≡0(modp2).
And
(px+k)2(py+k)2≡(2pxk+k2)(2pyk+k2)≡2pk3(x+y)+k4(modp2),
Hence, to prove (10) holds, we only need to prove:
⇔k=1∑p−12p(x+y)k3+k4p(y2−x2)+2k(y−x)≡0(modp2)k=1∑p−1(k32(y−x)−k4+2p(x+y)k33p(y2−x2))≡0(modp2)
Furthermore, we only need to prove the following congruences hold simultaneously:
{∑k=1p−1k31≡0(modp2)∑k=1p−1k4+2p(x+y)k31≡0(modp)
For (11), since
2k=1∑p−1k31=k=1∑p−1(k31+(p−k)31)=k=1∑p−1k3(p−k)3p3−3p2k+3pk2≡k=1∑ρ−1k(p−k)33p(modp2)
Hence, we only need to prove: ∑k=1p−1k(p−k)31≡0(modp). By
k(p−k)3≡−k4(modp)
We only need to prove: ∑k=1p−1k41≡0(modp). In fact, (12) also reduces to proving this.
Below, we prove: When p>5,p is a prime, ∑k=1p−1k41≡0(modp).
By Lagrange's theorem, we know that x4≡1(modp) has at most 4 solutions modulo p, so there exists s∈{1,2,⋯,p−1} such that s4=1(modp).
Notice that, {1,21,⋯,p−11} and {s,2s,⋯,p−1s} are both reduced residue systems modulo p, thus,
k=1∑p−1k41≡k=1∑p−1(ks)4=s4k=1∑p−1k41(modp)
Thus, combining p∤(s4−1), we get ∑k=1p−1k41≡0(modp).
The proposition is proved.