Wiki
Wiki

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

Updated

../


Source. Y. Yu and K. Chen, Erdős Problem 354(i): Strong Completeness of Two Dyadic Floor Sequences, manuscript of 13 September 2026, Lemma 2.2 with its display (2.2) and the consequence stated after it, physical p. 3, in the seventeen-page PDF held by its library source card, Yu and Chen (2026).

Standing. This is an author-recorded reconstruction. It is not an independent review, changes no status and assigns no tier.

Definitions

For a finite set W⊆ZW\subseteq\mathbb Z with at least two elements, listed as w0<w1<⋯<wmw_0<w_1<\cdots<w_m,

span⁡(W)=wm−w0,gap⁡(W)=max⁡0≤j<m(wj+1−wj).\operatorname{span}(W)=w_m-w_0,\qquad \operatorname{gap}(W)=\max_{0\le j<m}(w_{j+1}-w_j).

A gap of kk means that at most k−1k-1 consecutive integers of the interval [w0,wm][w_0,w_m] are missing from WW. The hull of WW is the real interval [w0,wm][w_0,w_m]. For an integer cc, W+cW+c is the translate.

Statement

Lemma 2.2. If span⁡(W)≥c>0\operatorname{span}(W)\ge c>0 and gap⁡(W)≤k\operatorname{gap}(W)\le k, then

gap⁡(W∪(W+c))≤k,span⁡(W∪(W+c))=span⁡(W)+c.\operatorname{gap}\bigl(W\cup(W+c)\bigr)\le k,\qquad \operatorname{span}\bigl(W\cup(W+c)\bigr)=\operatorname{span}(W)+c.

Consequence. Let c1≤c2≤⋯c_1\le c_2\le\cdots be positive integers with ci+1≤2cic_{i+1}\le2c_i for every ii, and let W0W_0 have gap⁡(W0)≤k\operatorname{gap}(W_0)\le k and span⁡(W0)≥c1\operatorname{span}(W_0)\ge c_1. Put Wi=Wi−1∪(Wi−1+ci)W_i=W_{i-1}\cup(W_{i-1}+c_i). Then every WiW_i has gap at most kk, min⁡Wi=min⁡W0\min W_i=\min W_0, and span⁡(Wi)=span⁡(W0)+c1+⋯+ci\operatorname{span}(W_i)=\operatorname{span}(W_0)+c_1+\cdots+c_i.

Proof

Write w0=min⁡Ww_0=\min W and wm=max⁡Ww_m=\max W. The hull of W+cW+c is [w0+c,wm+c][w_0+c,w_m+c]. Since c≤span⁡(W)=wm−w0c\le\operatorname{span}(W)=w_m-w_0, we have w0+c≤wmw_0+c\le w_m: the two hulls intersect or touch, and the union of the hulls is the single interval [w0,wm+c][w_0,w_m+c]. The minimum of W∪(W+c)W\cup(W+c) is w0w_0 and the maximum is wm+cw_m+c, which gives the span identity. All four hull endpoints w0w_0, wmw_m, w0+cw_0+c, wm+cw_m+c belong to the union.

Let w<w′w<w' be consecutive elements of W∪(W+c)W\cup(W+c); we show w′−w≤kw'-w\le k. The open interval (w,w′)(w,w') contains no element of the union, hence no hull endpoint. There are three cases.

If w′≤wmw'\le w_m, both points lie in the hull of WW. Let w−w_- be the largest element of WW with w−≤ww_-\le w (it exists because w≥w0w\ge w_0) and w+w_+ the smallest element of WW with w+≥w′w_+\ge w' (it exists because w′≤wmw'\le w_m). No element of WW lies in (w−,w](w_-,w], by the choice of w−w_-; none lies in (w,w′)(w,w'), by consecutiveness in the union; none lies in [w′,w+)[w',w_+), by the choice of w+w_+. So w−w_- and w+w_+ are consecutive in WW, and w′−w≤w+−w−≤kw'-w\le w_+-w_-\le k.

If w≥w0+cw\ge w_0+c, both points lie in the hull of W+cW+c, and the same argument applied to W+cW+c (whose gap is also at most kk) gives w′−w≤kw'-w\le k.

Otherwise w<w0+cw<w_0+c and w′>wmw'>w_m. Since w0+c≤wm<w′w_0+c\le w_m<w', the union point w0+cw_0+c lies in (w,w′)(w,w'), contradicting consecutiveness. So this case does not occur.

For the consequence, induct on ii. Given gap⁡(Wi−1)≤k\operatorname{gap}(W_{i-1})\le k and span⁡(Wi−1)≥ci\operatorname{span}(W_{i-1})\ge c_i, the lemma gives gap⁡(Wi)≤k\operatorname{gap}(W_i)\le k, min⁡Wi=min⁡Wi−1\min W_i=\min W_{i-1} and span⁡(Wi)=span⁡(Wi−1)+ci≥2ci≥ci+1\operatorname{span}(W_i)=\operatorname{span}(W_{i-1})+c_i\ge2c_i\ge c_{i+1}, which is the hypothesis for the next step. The base case is the assumption on W0W_0.