Wiki
Wiki

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

Updated


Source. Laczkovich (1984), printed p. 113 (PDF p. 5), equations (7)–(9) and the intervening recurrence. The source invokes standard regular continued-fraction theory. That theory is the explicit external input here; its general proof is not reconstructed on this page.

Let α\alpha be irrational and let [a0;a1,a2,…][a_0;a_1,a_2,\ldots] be its regular continued fraction, where a0∈Za_0\in\mathbb Z and aia_i is a positive integer for i≥1i\ge1. For its convergents pi/qip_i/q_i, use

p−1=1,p0=a0,q−1=0,q0=1,p_{-1}=1,\quad p_0=a_0,\qquad q_{-1}=0,\quad q_0=1,

and, for i≥0i\ge0,

pi+1=ai+1pi+pi−1,qi+1=ai+1qi+qi−1.(1)p_{i+1}=a_{i+1}p_i+p_{i-1},\qquad q_{i+1}=a_{i+1}q_i+q_{i-1}. \tag{1}

In particular q1=a1q_1=a_1. The denominators are positive and unbounded; qi+1≥qiq_{i+1}\ge q_i, with strict inequality for i≥1i\ge1. The approximation errors satisfy

0<qiα−pi<1qi+1(i even),−1qi+1<qiα−pi<0(i odd).(2)0<q_i\alpha-p_i<\frac1{q_{i+1}}\quad(i\text{ even}), \qquad -\frac1{q_{i+1}}<q_i\alpha-p_i<0\quad(i\text{ odd}). \tag{2}

If ai≤Ka_i\le K for every i≥1i\ge1, with K≥1K\ge1, then the recurrence and qi−1≤qiq_{i-1}\le q_i give

qi+1≤(K+1)qi(i≥0).(3)q_{i+1}\le(K+1)q_i\quad(i\ge0). \tag{3}

For i=0i=0, use q1=a1≤Kq0q_1=a_1\le Kq_0 directly. Thus (3) also gives qj−1≥qj/(K+1)q_{j-1}\ge q_j/(K+1) when j=1j=1; the possible equality q0=q1=1q_0=q_1=1 causes no exception.

These are exactly the facts used in Lemma 2. For the application on the whole real line, 2=[1;2,2,…]\sqrt2=[1;2,2,\ldots] supplies an irrational with bounded partial quotients, as the source notes on p. 110.

Source precision. The printed recurrence writes qi+1=aiqi+qi−1q_{i+1}=a_iq_i+q_{i-1}. With its indexing q0=1q_0=1 and q1=a1q_1=a_1, the coefficient is ai+1a_{i+1} as in (1). This is a transcription of the classical input with a corrected index, not an author-issued erratum. The separate two-denominator calculation in Lemma 2 is addressed on that result's page.

Bears on. Problem 1125, through Lemma 2 and the two monotonicity theorems.