Maths Olympiad Prep

Library / Iranian Mathematical Olympiad /223 of 270

Iranian Mathematical Olympiad

Algebra Difficulty 7.3 National Olympiad, round 2 Prove it Iran

Suppose that 2≤n2 \le n and a1,…,ana_1, \dots, a_n are natural numbers such that gcd(a1,…,an)=1\text{gcd}(a_1, \dots, a_n) = 1. Find all strictly increasing functions f:Z→Rf: \mathbb{Z} \to \mathbb{R} with the following property:
∀x1,…,xn∈Z:f(∑i=1nxiai)=∑i=1nf(xiai). \forall x_1, \dots, x_n \in \mathbb{Z} : f\left(\sum_{i=1}^{n} x_i a_i\right) = \sum_{i=1}^{n} f(x_i a_i).

Solution

Let Di=gcd(a1,…,ai−1,ai+1,…,ak)D_i = \text{gcd}(a_1, \dots, a_{i-1}, a_{i+1}, \dots, a_k) (i=1,…,ki = 1, \dots, k). Then f(x)=Cx+f1(x)+⋯+fk(x)f(x) = Cx + f_1(x) + \dots + f_k(x) such that f1,…,fkf_1, \dots, f_k are periodic of period DiD_i and fi(0)=0f_i(0) = 0. Let P=a1…akP = a_1 \dots a_k then if f(P)=0f(P) = 0 it follows that for all i≠ji \neq j and m≥0m \ge 0 we have f(maiaj)=0f(ma_i a_j) = 0. Indeed, letting M=f(a1a2)M = f(a_1 a_2) then f(2a1a2)=2Mf(2a_1 a_2) = 2M, f(3a1a2)=3Mf(3a_1 a_2) = 3M. By induction we find f(ma1a2)=mMf(ma_1 a_2) = mM. Choose m=a2…akm = a_2 \dots a_k we are done.
Let g(x)=f(x)−xf(P)Pg(x) = f(x) - \frac{x f(P)}{P} it follows that g(P)=0g(P) = 0 and therefore, for all i≠ji \neq j, m≥0m \ge 0 we have g(maiaj)=0g(ma_i a_j) = 0. Writing C=f(P)PC = \frac{f(P)}{P} it follows that f(x)=Cx+g(x)f(x) = Cx + g(x).
If for a given ii we have ∑j=1kujaj≡∑j=1kvjaj(modDi)\sum_{j=1}^{k} u_j a_j \equiv \sum_{j=1}^{k} v_j a_j \pmod{D_i} with ui≥0u_i \ge 0, vi≥0v_i \ge 0 then g(uiai)=g(viai)g(u_i a_i) = g(v_i a_i). It suffices to prove it for i=1i = 1. Indeed, the congruence implies that u1a1≡v1a1(modD1)u_1 a_1 \equiv v_1 a_1 \pmod{D_1}, since gcd(a1,D1)=1\text{gcd}(a_1, D_1) = 1 we conclude that u1≡v1(modD1)u_1 \equiv v_1 \pmod{D_1}. Since D1=gcd(a2,…,ak)D_1 = \text{gcd}(a_2, \dots, a_k), write
v1−u1=∑j=2kwjaj v_1 - u_1 = \sum_{j=2}^{k} w_j a_j
where wj=cj−cj′w_j = c_j - c'_j, cj,cj′≥0c_j, c'_j \ge 0. Thus,
u1+∑j=2kcjaj=v1+∑j=2kcj′aj. u_1 + \sum_{j=2}^{k} c_j a_j = v_1 + \sum_{j=2}^{k} c'_j a_j.
Multiplying both sides by a1a_1 and taking gg from both sides yields
g(u1a1)+∑j=2kg(cja1aj)=g(v1a1)+∑j=2kg(cj′a1aj). g(u_1 a_1) + \sum_{j=2}^{k} g(c_j a_1 a_j) = g(v_1 a_1) + \sum_{j=2}^{k} g(c'_j a_1 a_j).
Thus, g(u1a1)=g(v1a1)g(u_1 a_1) = g(v_1 a_1).
For every x=∑aiui∈Sx = \sum a_i u_i \in S we define a function f1(x)=g(a1u1)f_1(x) = g(a_1 u_1). If x∈Sx \in S can be written x=∑aiuix = \sum a_i u_i and also x=∑aiui′x = \sum a_i u'_i then g(a1u1)=g(a1u1′)g(a_1 u_1) = g(a_1 u'_1). Moreover, if x,zx, z both are in SS such that x≡z(modD1)x \equiv z \pmod{D_1} then we have f1(x)=f1(z)f_1(x) = f_1(z). We finally observe that every residue class mod  D1\mod D_1 contains elements in SS, in fact already elements of the subset a1u1a_1 u_1 (u1≥0u_1 \ge 0), because gcd(a1,D1)=1\text{gcd}(a_1, D_1) = 1. We may therefore extend the definition of f1(x)f_1(x) to all integers obtaining a function having the period D1D_1. Similarly, we define a function fi(x)f_i(x) (i=1,…,ki = 1, \dots, k) over all integers, having the period DiD_i with the property that fi(x)=g(aiui)f_i(x) = g(a_i u_i) if x=∑aiui∈Sx = \sum a_i u_i \in S.
And, in particular fi(0)=0f_i(0) = 0. If x=∑aiui∈Sx = \sum a_i u_i \in S then the functional equation gives
g(x)=g(∑aiui)=∑g(aiui)=∑fi(x). g(x) = g(\sum a_i u_i) = \sum g(a_i u_i) = \sum f_i(x).
Hence, g(x)=f1(x)+⋯+fk(x)g(x) = f_1(x) + \dots + f_k(x). We now use this to extend the definition of g(x)g(x) to all integers xx. Let us show that this implies g(aiui)=fi(aiui)g(a_i u_i) = f_i(a_i u_i). Indeed, if ui≥0u_i \ge 0 we have nothing to do. If ui<0u_i < 0 then we shall have g(aiui)=∑fi(aiui)=fi(aiui)g(a_i u_i) = \sum f_i(a_i u_i) = f_i(a_i u_i). It follows that g(x)g(x) satisfies the unrestricted equation
g(∑aiui)=∑g(aiui),ui∈Z. g(\sum a_i u_i) = \sum g(a_i u_i), \quad u_i \in \mathbb{Z}.
The converse is also easy. Indeed, the periodic function fi(x)f_i(x) depends on Di−1D_i - 1 arbitrary parameters and therefore the general solution depends on 1+∑(Di−1)1 + \sum (D_i - 1) arbitrary parameters. So, if all Di=1D_i = 1 the function is f(x)=Cxf(x) = Cx.
The least slope among the slopes of the sides of the polygonal graph of the function fi(x)(x∈Z)f_i(x)(x \in \mathbb{Z}) is given by −mi=min⁡x∈ZΔfi(x)-m_i = \min_{x \in \mathbb{Z}} \Delta f_i(x), where Δfi(x)=fi(x+1)−fi(x)\Delta f_i(x) = f_i(x+1) - f_i(x). Evidently, mi≥0m_i \ge 0. If we choose C≥∑miC \ge \sum m_i then f(x)f(x) is certainly non-decreasing since for all x:Δf(x)=C+∑Δfi(x)≥C−∑mi≥0x: \Delta f(x) = C + \sum \Delta f_i(x) \ge C - \sum m_i \ge 0.
We then prove that the solution of the above functional equation is non-decreasing if and only if ∑i=1kmi≤C\sum_{i=1}^k m_i \le C. Indeed, let the minimum of Δfi(x)\Delta f_i(x) is reached at x≡ci(modDi)x \equiv c_i \pmod{D_i} so that Δfi(ci)=−mi\Delta f_i(c_i) = -m_i (i=1,…,ki = 1, \dots, k). Since DiD_i are pair-wise coprime. By the CRT the system x≡ci(modD1)x \equiv c_i \pmod{D_1} has a solution tt. Then, 0≤Δf(t)=C+∑Δfi(t)=C−∑i=1kmi0 \le \Delta f(t) = C + \sum \Delta f_i(t) = C - \sum_{i=1}^k m_i. ■

Want a route through all this instead of an archive? The track puts 2,604 problems in a working order, from Junior Challenge level to the IMO shortlist.

Source: MathNet, licensed CC-BY-4.0. Statement reproduced verbatim; metadata (topic, difficulty) added by this project.