Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated

Applegate lagarias 1995 density bounds 2

../

theorem_1_1: For each integer a not divisible by 3 there is a positive constant c_a such that at least c_a x^0.81 integers n with |n| at most x reach a under the 3x+1 function, for all x at least a; the proof is computer-assisted.

theorem_2_1: If the linear program attached to a strictly retarded system derived from Krasikov's difference inequalities has a feasible solution with c_1^2 positive, then every c_j^n is positive and each counting function phi_j^n(y) is at least a constant times c_j^n lambda^y for all y > 0.


David Applegate and Jeffrey C. Lagarias, Density bounds for the 3x+1 problem. II. Krasikov inequalities, Math. Comp. 64 (1995), no. 209, 427-438 (AMS open back issues, S0025-5718-1995-1270613-2; 12 pp.).

The paper studies πa(x)\pi_a(x), the number of integers nn with ∣n∣≤x|n|\le x some iterate of which under the 3x+13x+1 function TT equals aa, for a≢0(mod3)a\not\equiv0\pmod3 (p. 427). It encodes Krasikov's difference inequalities for the counting functions of the residue classes mod 3k3^k (Proposition 2.1, p. 429) as linear programs whose coefficients depend nonlinearly on λ=2γ\lambda=2^\gamma; a feasible solution with c12>0c_1^2>0 gives exponential lower bounds for those functions (Theorem 2.1, p. 430). Section 3 compares splitting rules numerically for k≤9k\le9 (Tables 3.1--3.8, pp. 432--435), and a computer-found feasible solution of a linear program from the level-99 system, with 12(39−1)\frac12(3^9-1) variables and not printed, proves Theorem 1.1 (p. 428): for each a≢0(mod3)a\not\equiv0\pmod3 there is a positive constant cac_a with πa(x)≥cax.81\pi_a(x)\ge c_ax^{.81} for all x≥ax\ge a. Part I of the series had the exponent .65.65 (p. 427). Section 4 discusses Krasikov's conjecture that the inequalities give πa(x)≥x1−ε\pi_a(x)\ge x^{1-\varepsilon} for large kk, and states Conjecture 4.1 (p. 436) comparing the optimum of the untruncated program (LkNT)(L_k^{NT}) with those of the derived programs.

Source: PDF. The file prints "©1995 American Mathematical Society" on its first page, every other right reserved.

Read status. Claims checked: Theorems 1.1 and 2.1 and the definitions they use (pp. 427--431) were read on the print, and the proof of Theorem 2.1 was followed; the computed linear program and feasible solution behind Theorem 1.1 are not printed and were not checked. Nothing here is independently reviewed.

Bears on. #1135: the paper's TT is the problem's ff extended to Z\mathbb Z, and Theorem 1.1 with a=1a=1 gives at least c1x.81c_1x^{.81} integers mm with 1≤m≤x1\le m\le x whose orbit reaches 11, for all x≥1x\ge1; it is a lower bound on how many starting values reach 11 and does not settle the problem.

Results. Theorem 1.1 (p. 428), the bound πa(x)≥cax.81\pi_a(x)\ge c_ax^{.81}; Theorem 2.1 (p. 430), the lower bounds from feasible solutions of (Lλ)(L_\lambda).

No file of this source is held: no license on record permits its redistribution, and the card cites the edition it names above.