Maths Olympiad Prep

Library / /12 of 21

Number theory Difficulty 6.5 National olympiad Prove it North Macedonia

Find all natural numbers aa and bb greater than 11, such that ba(ab1)b^a \mid (a^b - 1).

Solution

Let pp be the least prime factor of bb, and qq be the least natural number such that p(aq1)p\mid(a^q-1) (such a number exists because p(ab1)p\mid(a^b-1)). From Little Fermat's theorem, we have that p(ap11)p\mid(a^{p-1}-1), which implies qbq\mid b and q(p1)q\mid(p-1), and from the minimality of pp, we get that q=1q=1, i.e. p(a1)p\mid(a-1). Let b=pαcb=p^\alpha c, where cc is not divisible by pp. If p2(a1)p^2\mid(a-1), then a1(modp2)a\equiv 1\pmod{p^2}, so ak1(modp2)a^k\equiv 1\pmod{p^2}, for every natural number kk from (al(p1)+al(p2)++1)p(modp2)(a^{l(p-1)} + a^{l(p-2)} + \dots + 1)\equiv p\pmod{p^2} and (apα(c1)+apα(c2)++1)c(modp2)(a^{p^\alpha(c-1)} + a^{p^\alpha(c-2)} + \dots + 1)\equiv c\pmod{p^2}, apαc1=(a1)(ap1+ap2++1)a^{p^\alpha c}-1=(a-1)(a^{p-1} + a^{p-2} + \dots + 1)\dots
(apα1(p1)+apα1(p2)++1)(apα(c1)+apα(c2)++1)\dots (a^{p^{\alpha-1}}(p-1) + a^{p^{\alpha-1}}(p-2) + \dots + 1)(a^{p^{\alpha}(c-1)} + a^{p^{\alpha}(c-2)} + \dots + 1)
so the degree of pp in ab1a1\frac{a^b-1}{a-1} is α\alpha and therefore we get that pα(a1)(a1)p^{\alpha(a-1)}\mid(a-1)
(pαa(ab1))(p^{\alpha a}\mid(a^b-1)). The last is not possible because for α1\alpha \ge 1, p2p \ge 2 and a2a \ge 2, it holds that pα(a1)>a1p^{\alpha(a-1)} > a-1.

This implies that, the degree of pp in a1a-1 must be 11. If p>2p > 2, we will prove that the degree of pp in apk1a^{p^k}-1 is k+1k+1, for every natural number kk. For k=0k=0, the statement is true. Let the statement be true for every k<nk < n. For k=nk=n from the following equalities
apn1=(apn1)p1=(apn11)(apn1(p1)+apn1(p2)++1) a^{p^n} - 1 = (a^{p^{n-1}})^p - 1 = (a^{p^{n-1}} - 1)(a^{p^{n-1}(p-1)} + a^{p^{n-1}(p-2)} + \dots + 1)
and
alpn11=(apn11)(apn1(l1)+apn1(l2)++1)pndnl(modpn+1) a^{lp^{n-1}} - 1 = (a^{p^{n-1}} - 1)(a^{p^{n-1}(l-1)} + a^{p^{n-1}(l-2)} + \dots + 1) \equiv p^n d_n l \pmod{p^{n+1}}
we get that
apn1pndn(pndn(p1+p2++1)+p)pn+1dn(pndnp12+1)(modp2n+1) \begin{aligned} a^{p^n} - 1 &\equiv p^n d_n \left( p^n d_n (p-1+p-2+\dots+1) + p \right) \equiv \\ &\equiv p^{n+1} d_n \left( p^n d_n \frac{p-1}{2} + 1 \right) \pmod{p^{2n+1}} \end{aligned}
(where dn=apn11pnd_n = \frac{a^{p^{n-1}}-1}{p^n} and dnd_n and pp coprime by the assumption) with which we've proved the statement by induction. From the equality
apαc1=(apα1)(apα(c1)+apα(c2)++1) a^{p^{\alpha}c} - 1 = (a^{p^{\alpha}} - 1)(a^{p^{\alpha}(c-1)} + a^{p^{\alpha}(c-2)} + \dots + 1)
and from apα(c1)+apα(c2)++1c(modp)a^{p^{\alpha}(c-1)} + a^{p^{\alpha}(c-2)} + \dots + 1 \equiv c \pmod{p}, we get that the degree of pp in ab1a^b - 1 is α+1\alpha + 1, and on the other hand it must be greater or equal to αa\alpha a, which is possible only for α=1\alpha = 1 and a=2a=2, but this case is not possible because of p(a1)p \nmid (a-1).

Now, the case when p=2p=2 remains. From the equality (the same as above for p>2p>2)
a2αc1=(a1)(a+1)(a2α1+1)(a2α(c1)+a2α(c2)++1). a^{2^{\alpha}c} - 1 = (a-1)(a+1)\dots(a^{2^{\alpha-1}} + 1)(a^{2^{\alpha}(c-1)} + a^{2^{\alpha}(c-2)} + \dots + 1).
The fact that cc is odd implies that a2α(c1)+a2α(c2)++1a^{2^{\alpha}(c-1)} + a^{2^{\alpha}(c-2)} + \dots + 1 is an odd number as a sum of an odd number of odd numbers (aa has to be odd, because 2(a1)2|(a-1)). From 2(a1)2|(a-1) and 2(a+1)2|(a+1), we get that 4(a2k1)4|(a^{2k}-1), for every natural number kk, but this implies that 44 is not a divisor of a2k+1a^{2k}+1. This implies that the degree of 22 in ab1a21\frac{a^b-1}{a^2-1} is α1\alpha-1, so because 2αa(ab1)2^{\alpha a}|(a^b-1), we get that 2α(a1)+1(a21)2^{\alpha(a-1)+1}|(a^2-1), and because 44 can be a divisor of only one of a+1a+1 and a1a-1, we get that 2α(a1)a+12^{\alpha(a-1)} \le a+1, which is possible if and only if α=1\alpha=1 or a=3a=3. Let rr be the least prime factor of cc, c=rβdc=r^\beta d, rr and dd are coprime and ss is the least natural number for which r(3s1)r|(3^s-1). This implies that sbs|b and s(r1)s|(r-1) (the same as before for pp). This is possible only if s=1s=1 or s=2s=2 (those are the only divisors of bb, less than rr). In both cases r(321)r|(3^2-1), which is possible only for r=2r=2, a contradiction with the choice of rr. We get that the unique solution is b=2b=2 and a=3a=3 (ba=8=ab1b^a=8=a^b-1).

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 and solution reproduced as published; topic and difficulty added by this site.