Unofficial archive — problems, solutions & results © IMC, reproduced with permission.

IMC / 2016 / Problems / Day 2, P9

IMC 2016 · Day 2 · P9

killer

Let kk be a positive integer. For each nonnegative integer nn, let f(n)f(n) be the number of solutions (x1,,xk)Zk(x_1, \dots, x_k) \in \mathbb{Z}^k of the inequality x1++xkn|x_1| + \dots + |x_k| \le n. Prove that for every n1n \ge 1, we have f(n1)f(n+1)f(n)2f(n-1) f(n+1) \le f(n)^2.

(Proposed by Esteban Arreaga, Renan Finder and José Madrid, IMPA, Rio de Janeiro)

Solution 1 of 2 (official)

We prove by induction on kk. If k=1k = 1 then we have f(n)=2n+1f(n) = 2n + 1 and the statement immediately follows from the AM-GM inequality.

Assume that k2k \ge 2 and the statement is true for k1k - 1. Let g(m)g(m) be the number of integer solutions of x1++xk1m|x_1| + \dots + |x_{k-1}| \le m; by the induction hypothesis g(m1)g(m+1)g(m)2g(m-1) g(m+1) \le g(m)^2 holds; this can be transformed to g(0)g(1)g(1)g(2)g(2)g(3).\frac{g(0)}{g(1)} \le \frac{g(1)}{g(2)} \le \frac{g(2)}{g(3)} \le \dots. For any integer constant cc, the inequality x1++xk1+cn|x_1| + \dots + |x_{k-1}| + |c| \le n has g(nc)g\bigl( n - |c| \bigr) integer solutions. Therefore, we have the recurrence relation f(n)=c=nng(nc)=g(n)+2g(n1)++2g(0).f(n) = \sum_{c=-n}^{n} g\bigl( n - |c| \bigr) = g(n) + 2 g(n-1) + \dots + 2 g(0). It follows that f(n1)f(n)=g(n1)+2g(n2)++2g(0)g(n)+2g(n1)++2g(1)+2g(0)g(n)+g(n1)+(g(n1)++2g(0)+20)g(n+1)+g(n)+(g(n)++2g(1)+2g(0))=f(n)f(n+1)\begin{align*} \frac{f(n-1)}{f(n)} &= \frac{g(n-1) + 2 g(n-2) + \dots + 2 g(0)} {g(n) + 2 g(n-1) + \dots + 2 g(1) + 2 g(0)} \le \\ &\le \frac{g(n) + g(n-1) + (g(n-1) + \dots + 2 g(0) + 2 \cdot 0)} {g(n+1) + g(n) + (g(n) + \dots + 2 g(1) + 2 g(0))} = \frac{f(n)}{f(n+1)} \end{align*} as required.

Solution 2 of 2 (official)

We first compute the generating function for f(n)f(n): n=0f(n)qn=(x1,x2,,xk)Zkc=0qx1+x2++xk+c=(xZqx)k11q=(1+q)k(1q)k+1.\sum_{n=0}^{\infty} f(n) q^n = \sum_{(x_1, x_2, \dots, x_k) \in \mathbb{Z}^k} \sum_{c=0}^{\infty} q^{|x_1| + |x_2| + \dots + |x_k| + c} = \left( \sum_{x \in \mathbb{Z}} q^{|x|} \right)^k \frac{1}{1-q} = \frac{(1+q)^k}{(1-q)^{k+1}}. For each a=0,1,a = 0, 1, \dots denote by ga(n)g_a(n) (n=0,1,2,n = 0, 1, 2, \dots) the coefficients in the following expansion: (1+q)a(1q)k+1=n=0ga(n)qn.\frac{(1+q)^a}{(1-q)^{k+1}} = \sum_{n=0}^{\infty} g_a(n) q^n. So it is clear that ga+1(n)=ga(n)+ga(n1)g_{a+1}(n) = g_a(n) + g_a(n-1) (n1n \ge 1), ga(0)=1g_a(0) = 1. Call a sequence of positive numbers g(0)g(0), g(1)g(1), g(2)g(2), … good if g(n1)g(n)\frac{g(n-1)}{g(n)} (n=1,2,n = 1, 2, \dots) is an increasing sequence. It is straightforward to check that g0g_0 is good: g0(n)=(k+nk),g0(n1)g0(n)=nk+n.g_0(n) = \binom{k+n}{k}, \qquad \frac{g_0(n-1)}{g_0(n)} = \frac{n}{k+n}. If gg is a good sequence then a new sequence gg' defined by g(0)=g(0)g'(0) = g(0), g(n)=g(n)+g(n1)g'(n) = g(n) + g(n-1) (n1n \ge 1) is also good: g(n1)g(n)=g(n1)+g(n2)g(n)+g(n1)=1+g(n2)g(n1)1+g(n)g(n1),\frac{g'(n-1)}{g'(n)} = \frac{g(n-1) + g(n-2)}{g(n) + g(n-1)} = \frac{1 + \frac{g(n-2)}{g(n-1)}}{1 + \frac{g(n)}{g(n-1)}}, where define g(1)=0g(-1) = 0. Thus we see that each of the sequences g1g_1, g2g_2, …, gk=fg_k = f are good. So the desired inequality holds.

How the field did

contestants scored
314
average (of 10)
0.69
solved (≥ 80%)
3.2%
near-0 (≤ 10%)
91.7%
discrimination
0.40

Score distribution (field cohort)

Computed on contestants with a meaningful total (field cohort); discrimination is the corrected item–total correlation.

Similar problems

IMC 2023 · Day 2 · P8very hardavg 2.1/10 · solved 14% · near-0 59% · disc 0.52
IMC 2014 · Day 2 · P9killeravg 0.6/10 · solved 4% · near-0 89% · disc 0.40
IMC 2001 · Day 2 · P12killeravg 0.5/10 · solved 4% · near-0 94% · disc 0.13
IMC 2004 · Day 1 · P5killeravg 0.6/10 · solved 4% · near-0 93% · disc 0.29