Maths Olympiad Prep

Library / /7 of 10

, 2015

Number theory Difficulty 5.9 AIME, harder Prove it Taiwan

Find all triples (p,x,y)(p, x, y) consisting of a prime number pp and two positive integers xx and yy such that xp1+yx^{p-1} + y and x+yp1x + y^{p-1} are both powers of pp.

Find all triples of a prime number pp and a pair of positive integers (x,y)(x, y) such that xp1+yx^{p-1} + y and x+yp1x + y^{p-1} are both powers of pp.

Solution

All solutions are (p,x,y){(3,2,5),(3,5,2)}{(2,n,2kn)0<n<2k}(p, x, y) \in \{(3, 2, 5), (3, 5, 2)\} \cup \{(2, n, 2^k - n) \mid 0 < n < 2^k\}.

(1) When p=2p=2, clearly all (x,y)(x,y) whose sum is a power of 22 satisfy the problem's conditions, so we only need to consider p>2p > 2.

(2) Suppose xp1+y=pax^{p-1} + y = p^a and x+yp1=pbx + y^{p-1} = p^b. Without loss of generality, we assume xyx \le y, and hence aba \le b. We therefore have
pb=yp1+x=(paxp1)p1+x p^b = y^{p-1} + x = (p^a - x^{p-1})^{p-1} + x
Taking both sides of the above equation modulo pap^a, and noting that p1p-1 is even, we thus have
0=x(p1)2+xmod pa(1) 0 = x^{(p-1)^2} + x \quad \mathrm{mod}\ p^a \qquad (1)
If pxp \nmid x, then x(p1)21+1x^{(p-1)^2-1} + 1 is clearly not divisible by pp, but this contradicts Eq. (1) (since xxp1<pax \le x^{p-1} < p^a.)
Hence pp does not divide xx, that is,
pax(p1)21+1=xp(p2)+1. p^a \mid x^{(p-1)^2-1} + 1 = x^{p(p-2)} + 1.

(3) By Fermat's little theorem, x(p1)2=1modpx^{(p-1)^2} = 1 \mod p, so
x+1=x+x(p1)2=x(1+x(p1)21)=0modp, x + 1 = x + x^{(p-1)^2} = x(1 + x^{(p-1)^2-1}) = 0 \mod p,
that is, px+1p \nmid x + 1. Let prp^r be the highest power dividing x+1x + 1.

(4) Now let us consider the highest power dividing xp(p2)+1x^{p(p-2)} + 1. Let us expand xp(p2)=(x+11)p(p2)x^{p(p-2)} = (x + 1 - 1)^{p(p-2)} binomially as a sum of a series of terms in (x+1)k(x+1)^k. For all terms with k3k \ge 3, by the above it must be divisible by p3rp^{3r}. The k=2k=2 term is
p(p2)(p22p1)2(x+1)2, -\frac{p(p-2)(p^2-2p-1)}{2}(x+1)^2,
which is clearly divisible by p2r+1p^{2r+1}. The k=1k=1 term is
p(p2)(x+1), p(p-2)(x+1),
which is clearly divisible by pr+1p^{r+1}, but not by pr+2p^{r+2} (since rr is the highest power dividing x+1x+1). The last term is 1-1. Combining the above discussion: we know that the highest power dividing xp(p2)+1x^{p(p-2)} + 1 is pr+1p^{r+1}.

(5) But on the other hand, we assumed from the start that paxp(p2)+1p^a \mid x^{p(p-2)} + 1, so we must have ar+1a \le r+1. But at the same time,
prx+1xp1+y=pa(2) p^r \le x+1 \le x^{p-1} + y = p^a \quad (2)
so we must have a=ra=r or a=r+1a=r+1.

(6) If a=ra=r, then all equalities in Eq. (2) must hold, so x=y=1x=y=1, but this clearly contradicts p>2p>2, so a=r+1a=r+1. Furthermore, based on prx+1p^r \le x+1, we have
x=x2+xx+1xp1+yx+1=pax+1papr=p. x = \frac{x^2 + x}{x+1} \le \frac{x^{p-1} + y}{x+1} = \frac{p^a}{x+1} \le \frac{p^a}{p^r} = p.
Combining this with px+1p|x+1, we have x=p1x = p-1; in other words, r=1,a=2r=1, a=2.

(7) Now, if p5p \ge 5, we have
pa=xp1+y>(p1)4=(p22p+1)2>(3p)2>p2=pa p^a = x^{p-1} + y > (p-1)^4 = (p^2 - 2p + 1)^2 > (3p)^2 > p^2 = p^a
a contradiction! Hence p=3p=3, so x=p1=2x=p-1=2, and y=paxp1=5y = p^a - x^{p-1} = 5.

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 translated into English from en; metadata (topic, difficulty) added by this project.