Let p1,p2,…,p2025 be real numbers, and let {an(1)}n≥0, {an(2)}n≥0, …, {an(2025)}n≥0 be 2025 real sequences satisfying: (1) a0(i) (1≤i≤2025) are all zero; (2) a1(i) (1≤i≤2025) are not all zero; (3) For i=1,2,…,2025 and any positive integer n, an−1(i)+an(i)+an+1(i)=pi⋅an(i+1), where an(2026)=an(1). Prove that there exists a positive real number r and infinitely many positive integers n such that max{∣an(1)∣,∣an(2)∣,…,∣an(2025)∣}>r.
Solution
Proof: We first prove a lemma. Lemma: Let β be a complex number. If a sequence {an} satisfies a0=0, a1=0, and for all n≥1, an−1+βan+an+1=0, then {an} does not converge to 0. Proof of Lemma: Let α1,α2 be roots of x2+βx+1=0, so α1α2=1. If α1=α2, then α1=α2=±1 and an=pn+q or an=(−1)n(pn+q). Clearly an doesn't converge to 0. If α1=α2, then an=pα1n+qα2n with p,q∈C and pq=0 (since a0=0, a1=0). Case 1: ∣α1∣=∣α2∣. Then an cannot converge to 0 since ∣α1∣∣α2∣=1. Case 2: ∣α1∣=∣α2∣=1. Let α1=e2πiθ, α2=e−2πiθ. If θ∈Q, then {an} is periodic and non-zero. If θ∈/Q, then {nθ} is dense modulo 1, making {an} have values dense on some circle. In all cases, an doesn't converge to 0. □ Now the main proof. Denote the given equations as (∗1),⋯,(∗2025). Case 1:∏i=12025pi=0. Define transformed sequences: bn(i)=2025pi+1pi+2⋯pi+2024⋅an(i) where indices are cyclic modulo 2025. These satisfy: bn−1(i)+bn(i)+bn+1(i)=2025p1⋯p2025⋅bn(i+1). ---
Thus we may assume p1=⋯=p2025=p. Let ω be a 2025th root of unity and define: Xn:=k=0∑2024ωkan(k+1). Then: ∀n≥1,Xn−1+(1−ω−1p)Xn+Xn+1=0. Since not all a1(i)=0, some X1=0. By the lemma, {Xn} doesn't converge to 0. Case 2: If there exists some pi=0. Without loss of generality, assume p1=0. Then from (∗)1 we know that for any n≥1 we have an−1(1)+an(1)+an+1(1)=0. If a1(1)=0, applying the lemma to the sequence {an(1)} shows that {an(1)} does not converge to 0. If a1(1)=0, then an(1)≡0. Consequently, from (∗)2025 we know that for any n≥1 we have an−1(2025)+an(2025)+an+1(2025)=0. In this case, if a1(2025)=0, we can apply the lemma to the sequence {an(2025)}n≥0. If a1(2025)=0 then an(2025)≡0, and similarly we obtain an−1(2024)+an(2024)+an+1(2024)=0. Continuing this process, by induction we can prove that either all sequences in the problem are identically zero, or there exists at least one sequence that does not converge to 0. The former case contradicts the problem's assumptions, thus completing the proof. □
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 and solution reproduced as published; topic and difficulty added by this site.