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

IMC / 2012 / Problems / Day 1, P1

IMC 2012 · Day 1 · P1

easy

For every positive integer nn, let p(n)p(n) denote the number of ways to express nn as a sum of positive integers. For instance, p(4)=5p(4) = 5 because 4=3+1=2+2=2+1+1=1+1+1+1.4 = 3 + 1 = 2 + 2 = 2 + 1 + 1 = 1 + 1 + 1 + 1. Also define p(0)=1p(0) = 1.

Prove that p(n)p(n1)p(n) - p(n-1) is the number of ways to express nn as a sum of integers each of which is strictly greater than 1.

(Proposed by Fedor Duzhin, Nanyang Technological University)

Solution 1 of 2 (official)

The statement is true for n=1n = 1, because p(0)=p(1)=1p(0) = p(1) = 1 and the only partition of 1 contains the term 1. In the rest of the solution we assume n2n \ge 2.

Let Pn={(a1,,ak):kN, a1ak, a1++ak=n}P_n = \{ (a_1, \dots, a_k) : k \in \mathbb{N},\ a_1 \ge \dots \ge a_k,\ a_1 + \dots + a_k = n \} be the set of partitions of nn, and let Qn={(a1,,ak)Pn:ak=1}Q_n = \{ (a_1, \dots, a_k) \in P_n : a_k = 1 \} the set of those partitions of nn that contain the term 1. The set of those partitions of nn that do not contain 1 as a term, is PnQnP_n \setminus Q_n. We have to prove that PnQn=PnPn1|P_n \setminus Q_n| = |P_n| - |P_{n-1}|.

Define the map φ:Pn1Qn\varphi : P_{n-1} \to Q_n as φ(a1,,ak)=(a1,,ak,1).\varphi(a_1, \dots, a_k) = (a_1, \dots, a_k, 1). This is a partition of nn containing 1 as a term (so indeed φ(a1,,ak)Qn\varphi(a_1, \dots, a_k) \in Q_n). Moreover, each partition (a1,,ak,1)Qn(a_1, \dots, a_k, 1) \in Q_n uniquely determines (a1,,ak)(a_1, \dots, a_k). Therefore the map φ\varphi is a bijection between the sets Pn1P_{n-1} and QnQ_n. Then Pn1=Qn|P_{n-1}| = |Q_n|. Since QnPnQ_n \subset P_n, PnQn=PnQn=PnPn1=p(n)p(n1).|P_n \setminus Q_n| = |P_n| - |Q_n| = |P_n| - |P_{n-1}| = p(n) - p(n-1).

Solution 2 of 2 (official)

(outline) Denote by q(n)q(n) the number of partitions of nn not containing 1 as term (q(0)=1q(0) = 1 as the only partition of 0 is the empty sum), and define the generating functions F(x)=n=0p(n)xnandG(x)=n=0q(n)xn.F(x) = \sum_{n=0}^{\infty} p(n) x^n \quad \text{and} \quad G(x) = \sum_{n=0}^{\infty} q(n) x^n. Since q(n)p(n)<2nq(n) \le p(n) < 2^n, these series converge in some interval, say for x<12|x| < \frac{1}{2}, and the values uniquely determine the coefficients.

According to Euler's argument, we have F(x)=n=0p(n)xn=k=1(1+xk+x2k+)=k=111xkF(x) = \sum_{n=0}^{\infty} p(n) x^n = \prod_{k=1}^{\infty} (1 + x^k + x^{2k} + \dots) = \prod_{k=1}^{\infty} \frac{1}{1 - x^k} and G(x)=n=0q(n)xn=k=2(1+xk+x2k+)=k=211xk.G(x) = \sum_{n=0}^{\infty} q(n) x^n = \prod_{k=2}^{\infty} (1 + x^k + x^{2k} + \dots) = \prod_{k=2}^{\infty} \frac{1}{1 - x^k}. Then G(x)=(1x)F(x)G(x) = (1 - x) F(x). Comparing the coefficient of xnx^n in this identity we get q(n)=p(n)p(n1)q(n) = p(n) - p(n-1).

How the field did

contestants scored
313
average (of 10)
9.68
solved (≥ 80%)
96.8%
near-0 (≤ 10%)
0.6%
discrimination
0.13

Score distribution (field cohort)

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

Similar problems

IMC 2001 · Day 1 · P1easyavg 9.0/10 · solved 88% · near-0 5% · disc 0.23
IMC 2006 · Day 2 · P7easyavg 9.2/10 · solved 88% · near-0 3% · disc 0.35
IMC 2014 · Day 2 · P6easyavg 8.6/10 · solved 86% · near-0 12% · disc 0.38
IMC 2024 · Day 2 · P6easyavg 8.2/10 · solved 77% · near-0 11% · disc 0.35