Maths Olympiad Prep

Track / Stage 8 / 169 of 180 #1869 of 1964

Problem 1869

IMO Shortlist mid-range; USAMO P2/P5
Algebra Difficulty 8.8 Prove it Team selection tests · Vietnam

Given are two coprime positive integers a,ba, b with bb odd and a>2a > 2. The sequence (xn)(x_n) is defined by x0=2x_0 = 2, x1=ax_1 = a and xn+2=axn+1+bxnx_{n+2} = a x_{n+1} + b x_n for n1n \ge 1. Prove that

a) If aa is even then there do not exist positive integers m,n,pm, n, p such that xmxnxp\frac{x_m}{x_n x_p} is a positive integer.

b) If aa is odd then there do not exist positive integers m,n,pm, n, p such that mnpmnp is even and xmxnxp\frac{x_m}{x_n x_p} is a perfect square.

This one wants a proof. Work it on paper, then read the official solution and mark yourself. Be honest about it: the record is only any use to you if it is.

Official solution

The general formula for the sequence is xn=αn+βnx_n = \alpha^n + \beta^n with α\alpha and β\beta such that
α+β=a,αβ=b. \alpha + \beta = a, \quad \alpha\beta = -b.

We will now prove gcd(b,xm)=1\gcd(b, x_m) = 1 for all mm. Indeed, suppose there are mm and a prime pp such that xmx_m and bb are divisible by pp. Since gcd(a,b)=1\gcd(a, b) = 1, gcd(a,p)=1\gcd(a, p) = 1. Since
xmbxm2=axm1 x_m - b x_{m-2} = a x_{m-1}
results in axm1a x_{m-1} being divisible by pp or xm1x_{m-1} being divisible by pp. By induction, we obtain am2,,a1a_{m-2}, \dots, a_1 are divisible by pp or aa is divisible by pp, contradiction. Thus bb is coprime to all terms of the sequence.

We find the condition for the pairs (m,n)(m, n) with m>0m > 0 such that xmxnx_m \mid x_n. It is easy to check that (xn)(x_n) is a strictly increasing sequence of positive integers. Therefore, the necessary condition is mnm \le n. Next, if n>2mn > 2m, we have the following equation
xn+(b)mxn2m=αn+βn+αmβm(αn2m+βn2m)=(αm+βm)(αnm+βnm) \begin{aligned} x_n + (-b)^m x_{n-2m} &= \alpha^n + \beta^n + \alpha^m \beta^m (\alpha^{n-2m} + \beta^{n-2m}) \\ &= (\alpha^m + \beta^m)(\alpha^{n-m} + \beta^{n-m}) \end{aligned}
is divisible by xmx_m. So we have
xmxnxmbmxn2mxmxn2m. x_m \mid x_n \Leftrightarrow x_m \mid b^m x_{n-2m} \Leftrightarrow x_m \mid x_{n-2m}.
Therefore, by induction, if kk is the remainder of nn when divided by 2m2m, then xmxnx_m \mid x_n if and only if xmxkx_m \mid x_k. Here are two cases:

* If mkm \ge k, then xkxmx_k \le x_m. Hence xmxnx_m \mid x_n if and only if xm=xkx_m = x_k or m=km = k.
* If m<k<2mm < k < 2m, write k=m+dk = m + d with 0<d<k0 < d < k. We have
xdxmxm+d=(αd+βd)(αm+βm)αm+dβm+d=αdβd(αmd+βmd)=(b)dxmd. \begin{aligned} x_d x_m - x_{m+d} &= (\alpha^d + \beta^d)(\alpha^m + \beta^m) - \alpha^{m+d} - \beta^{m+d} \\ &= \alpha^d \beta^d (\alpha^{m-d} + \beta^{m-d}) \\ &= (-b)^d x_{m-d}. \end{aligned}
Therefore, xkx_k is divisible by xmx_m if and only if (b)dxmd(-b)^d x_{m-d} is divisible by xmx_m, i.e. xmdx_{m-d} is divisible by xmx_m, which contradicts the fact that (xn)(x_n) is strictly increasing.

In short, for m>0m > 0, we have xmxnx_m \mid x_n if and only if n=(2k+1)mn = (2k+1)m for some natural number kk.

a.

Since x0=2x_0 = 2 and x1=ax_1 = a are even, we can inductively prove that xnx_n is even for all nn. Considering the quotient of x(2k+1)nx_{(2k+1)n} with xnx_n, we have
x(2k+1)nxn=α(2k+1)n+β(2k+1)nαn+βn=i=0k(1)iαniβni(αn(2k2i)+βn(2k2i))=x2kn+bnx(2k2)n++b(k1)nx2n+bkn \begin{aligned} \frac{x_{(2k+1)n}}{x_n} &= \frac{\alpha^{(2k+1)n} + \beta^{(2k+1)n}}{\alpha^n + \beta^n} \\ &= \sum_{i=0}^{k} (-1)^i \alpha^{ni} \beta^{ni} (\alpha^{n(2k-2i)} + \beta^{n(2k-2i)}) \\ &= x_{2kn} + b^n x_{(2k-2)n} + \dots + b^{(k-1)n} x_{2n} + b^{kn} \end{aligned}
is odd, because bknb^{kn} is odd. So if xmx_m is divisible by xnx_n then xmxn\frac{x_m}{x_n} is odd, so this fraction is not divisible by xpx_p.

b.

Consider three numbers m,n,pm, n, p such that xmx_m is divisible by xnxpx_n x_p. Then, there exist natural numbers k,lk, l such that
m=(2k+1)n=(2l+1)p. m = (2k + 1)n = (2l + 1)p.
Since the product mnpmnp is even, all of three numbers are even. Set m=2um = 2u, n=2vn = 2v and p=2tp = 2t. For each natural number kk:
x2k=α2k+β2k=(αkβk)2+2b2k. \begin{aligned} x_{2k} &= \alpha^{2k} + \beta^{2k} \\ &= (\alpha^k - \beta^k)^2 + 2b^{2k}. \end{aligned}
On the other hand,
(αkβk)2=(a24b)(i=0k/2αiβi(αk2i+βk2i))2=(a24b)(i=0k/2(b)ixk2i)2=(a24b)Fk2 \begin{aligned} (\alpha^k - \beta^k)^2 &= (a^2 - 4b) \left( \sum_{i=0}^{\lfloor k/2 \rfloor} \alpha^i \beta^i (\alpha^{k-2i} + \beta^{k-2i}) \right)^2 \\ &= (a^2 - 4b) \left( \sum_{i=0}^{\lfloor k/2 \rfloor} (-b)^i x_{k-2i} \right)^2 \\ &= (a^2 - 4b) F_k^2 \end{aligned}
where FkF_k is the expression inside the brackets. So by assumption, if xmxnxpx_m x_n x_p is a perfect square then
k{u,v,t}((a24b)Fk2+2b2k). \prod_{k \in \{u, v, t\}} \left( (a^2 - 4b) F_k^2 + 2b^{2k} \right).
Therefore, for every prime divisor pp dividing a24ba^2 - 4b then
1=(2b2u2b2v2b2tp)=(2p) 1 = \left( \frac{2b^{2u} 2b^{2v} 2b^{2t}}{p} \right) = \left( \frac{2}{p} \right)
leads to 22 which is a quadratic residue modulo pp. This implies that pp has the form 8h+18h+1 or 8h+78h+7, for every prime divisor pp of a24ba^2 - 4b. Thus a24ba^2 - 4b divided by 88 leaves a remainder of 11 or 77. On the other hand, aa and bb are odd so
a24b145(mod8), a^2 - 4b \equiv 1 - 4 \equiv 5 \pmod{8},
this is a contradiction. So xmxnxpx_m x_n x_p is not a perfect square and the same for xmxnxp\frac{x_m}{x_n x_p}.

\square

Source: MathNet, licensed CC-BY-4.0. Statement and solution reproduced as published; topic, difficulty and ordering added by this site.