Maths Olympiad Prep

Library / /6 of 6

Algebra Difficulty 6.9 National Olympiad Prove it Thailand

Let kk be a positive integer. Show that if there exists a sequence a0,a1,a_0, a_1, \dots of integers satisfying the condition
an=an1+nkn for all n1, a_n = \frac{a_{n-1} + n^k}{n} \text{ for all } n \ge 1,
then k2k-2 is divisible by 3.

Solution

Denote f:Z02Zf: \mathbb{Z}_{\ge 0}^2 \to \mathbb{Z} such that
xn=f(0,n)+f(1,n)[(x+1)1]+f(2,n)[(x+1)(x+2)(x+1)]+ x^n = f(0, n) + f(1, n)[(x+1)-1] + f(2, n)[(x+1)(x+2) - (x+1)] + \dots
for every non-negative integers x,nx, n.
The above function always exists because the expression of xnx^n can be uniquely represented in that particular way. Moreover, f(x,n)f(x, n) is an integer for every x,nZ0x, n \in \mathbb{Z}_{\ge 0} and f(x,n)=0f(x, n) = 0 for all x>nx > n. Next, we will prove the following lemmas.

Lemma 1. f(0,n)f(0, n) is even if and only if 3n13 \mid n - 1
Proof. Observe that f(0,0)=1f(0, 0) = 1, f(0,1)=1f(0, 1) = 1, and f(0,2)=1f(0, 2) = -1, so it is sufficient to show that 2f(0,n)+f(0,n+1)+f(0,n+2)2 \mid f(0, n) + f(0, n + 1) + f(0, n + 2) for all nZ0n \in \mathbb{Z}_{\ge 0}.

Rewrite the definition of ff as
xn=g(0,n)+i=1ng(i,n)(x+1)(x+2)(x+i) x^n = g(0, n) + \sum_{i=1}^{n} g(i, n)(x+1)(x+2)\cdots(x+i)
where g(k,n)=f(k,n)f(k+1,n)g(k, n) = f(k, n) - f(k+1, n) for all k,nZ0k, n \in \mathbb{Z}_{\ge 0}. Therefore,
g(0,n)+g(1,n)++g(n,n)=i=0ng(i,n) for all nZ0 g(0, n) + g(1, n) + \cdots + g(n, n) = \sum_{i=0}^{n} g(i, n) \text{ for all } n \in \mathbb{Z}_{\ge 0}

xn+1=g(0,n)+i=1ng(i,n)x(x+1)(x+2)(x+i)=g(0,n)+i=1ng(i,n)[(x+1)(x+2)(x+i)(x+i+1)(i+1)(x+1)(x+2)(x+i)]=g(n,n)(x+1)(x+2)(x+n+1)+i=1n1[g(i1,n)(i+1)g(i,n)](x+1)(x+2)(x+i)g(0,n) \begin{align*} x^{n+1} &= g(0, n) + \sum_{i=1}^{n} g(i, n) \cdot x(x+1)(x+2) \cdots (x+i) \\ &= g(0, n) + \sum_{i=1}^{n} g(i, n) \cdot \left[ (x+1)(x+2) \cdots (x+i)(x+i+1) \right. \\ & \qquad \left. -(i+1)(x+1)(x+2) \cdots (x+i) \right] \\ &= g(n, n)(x+1)(x+2) \cdots (x+n+1) + \sum_{i=1}^{n-1} [g(i-1, n) - (i+1)g(i, n)] \\ & \qquad (x+1)(x+2) \cdots (x+i) - g(0, n) \end{align*}

Therefore, g(n+1,n+1)=g(n,n)g(n+1, n+1) = g(n, n), g(0,n+1)=g(0,n)g(0, n+1) = -g(0, n), and g(i,n+1)=g(i1,n)(i+1)g(i,n)g(i, n+1) = g(i-1, n) - (i+1)g(i, n) for each 1in1 \le i \le n. Wrapping them up, we get, for each nZ0n \in \mathbb{Z}_{\ge 0},
f(0,n+1)=i=0n+1g(i,n+1)=i=0nig(i,n). f(0, n + 1) = \sum_{i=0}^{n+1} g(i, n + 1) = - \sum_{i=0}^{n} ig(i, n).
And
f(0,n+2)=i=0n+1ig(i,n+1)=(n+1)g(n,n)i=1ni[g(i1,n)(i+1)g(i,n)](n+1)g(n,n)i=0n1(i+1)g(i,n)(mod2)i=0n(i+1)g(i,n)(mod2) \begin{align*} f(0, n + 2) &= - \sum_{i=0}^{n+1} ig(i, n + 1) \\ &= -(n + 1)g(n, n) - \sum_{i=1}^{n} i[g(i - 1, n) - (i + 1)g(i, n)] \\ &\equiv -(n + 1)g(n, n) - \sum_{i=0}^{n-1} (i + 1)g(i, n) \pmod{2} \\ &\equiv - \sum_{i=0}^{n} (i + 1)g(i, n) \pmod{2} \end{align*}
Therefore,
f(0,n)+f(0,n+1)+f(0,n+2)i=0ng(i,n)i=0nig(i,n)i=0n(i+1)g(i,n)0(mod2) \begin{align*} f(0, n) + f(0, n + 1) + f(0, n + 2) \\ &\equiv \sum_{i=0}^{n} g(i, n) - \sum_{i=0}^{n} ig(i, n) - \sum_{i=0}^{n} (i + 1)g(i, n) \\ &\equiv 0 \pmod{2} \end{align*}

Lemma 2. If α\alpha and β\beta are integers such that n!α(1!+2!++(n1)!)+βn! \mid \alpha \cdot (1! + 2! + \dots + (n-1)!) + \beta for all positive integer nn, then α=β=0\alpha = \beta = 0

Proof. We know that 1!+2!++k!<(k+1)!1! + 2! + \dots + k! < (k+1)! for all positive integer kk. Taking large enough nn for which (n1)!>β(n-1)! > |\beta| and n>2α+1n > 2|\alpha| + 1, we have
α(1!+2!++(n1)!)+βα(1!+2!++(n1)!)+β<(2α+1)(n1)!<n(n1)!=n! \begin{aligned} |\alpha \cdot (1! + 2! + \dots + (n-1)!)| + |\beta| &\le |\alpha| \cdot (1! + 2! + \dots + (n-1)!) + |\beta| \\ &< (2|\alpha| + 1) \cdot (n-1)! \\ &< n \cdot (n-1)! \\ &= n! \end{aligned}
But n!α(1!+2!++n!)+βn! \mid \alpha \cdot (1! + 2! + \dots + n!)+ \beta, therefore α(1!+2!++n!)+β=0\alpha \cdot (1! + 2! + \dots + n!)+ \beta = 0 for all large nn. Thus, we can conclude that α=β=0\alpha = \beta = 0, as desired. \square

Back to the problem. Let m=k1m = k - 1. The given sequence is equivalent to
a0+j=1njmj!a_0 + \sum_{j=1}^{n} j^m \cdot j!
an=a0+j=1njmj!n!.a_n = \frac{a_0 + \sum_{j=1}^{n} j^m \cdot j!}{n!}.
We have n!a0+j=1njmj!n! \mid a_0 + \sum_{j=1}^{n} j^m \cdot j! for all positive integer nn. Now taking modulo n!n! by using the expression of xmx^m as mentioned before, we get
j=1njmj!=j=1nj!{f(0,m)+f(1,m)[(x+1)1]+f(2,m)[(x+1)(x+2)(x+1)]++f(m,m)[(x+1)(x+2)(x+m)(x+1)(x+2)(x+m1)]}=f(0,m)i=1ni!+i=1nj=1mf(j,m)[(i+j)!(i+j1)!]=f(0,m)i=1ni!+j=1mi=1nf(j,m)[(i+j)!(i+j1)!]=f(0,m)i=1ni!+j=1mf(j,m)[(n+j)!j!]f(0,m)(1!+2!++(n1)!)j=1mf(j,m)j!(modn!) \begin{aligned} \sum_{j=1}^{n} j^m \cdot j! &= \sum_{j=1}^{n} j! \cdot \{f(0, m) + f(1, m)[(x+1)-1] + f(2, m)[(x+1)(x+2)-(x+1)] \\ &\quad + \dots + f(m, m)[(x+1)(x+2)\dots(x+m) - (x+1)(x+2)\dots(x+m-1)]\} \\ &= f(0, m) \cdot \sum_{i=1}^{n} i! + \sum_{i=1}^{n} \sum_{j=1}^{m} f(j, m) \cdot [(i+j)! - (i+j-1)!] \\ &= f(0, m) \cdot \sum_{i=1}^{n} i! + \sum_{j=1}^{m} \sum_{i=1}^{n} f(j, m) \cdot [(i+j)! - (i+j-1)!] \\ &= f(0, m) \cdot \sum_{i=1}^{n} i! + \sum_{j=1}^{m} f(j, m)[(n+j)! - j!] \\ &\equiv f(0, m) \cdot (1! + 2! + \dots + (n-1)!) - \sum_{j=1}^{m} f(j, m) \cdot j! \pmod{n!} \end{aligned}

Therefore, n!f(0,m)(1!+2!++(n1)!)+βn! \mid f(0, m) \cdot (1! + 2! + \dots + (n-1)!) + \beta for every positive integer nn, when β=a0j=1mf(j,m)j!\beta = a_0 - \sum_{j=1}^{m} f(j, m) \cdot j!. Applying Lemma 2, we get f(0,m)=0f(0, m) = 0. And by Lemma 1, we get 3m13 \mid m-1. Since m=k1m = k-1, therefore k2k-2 is divisible by 3, as desired.

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.