Wiki
Wiki

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

Updated

../


Source. S. Korsky, A resolution of the de Bruijn--Erdős consecutive-gap problem, arXiv:2609.07196v2, Lemma 6.3 (p. 12) of the retained PDF, read in the canonical conversion and checked against the text layer; held by its library card, Korsky 2026, resolution.

Standing. Author-recorded reconstruction; not an independent review; changes no status and assigns no tier. The source is an unrefereed preprint.

Definitions

Notation as on the Lemma 6.2 page: Δt(x,D)=Nt((x,x+D/t])−D\Delta_t(x,D)=N_t((x,x+D/t])-D, Zt(D)=∫T(Δt)+Z_t(D)=\int_{\mathbb T}(\Delta_t)_+, hypothesis (6.1) with constant A≥1A\ge1, and the rr-spans Si(t)S_i(t) of PtP_t.

Statement (Lemma 6.3, p. 12)

Under (6.1), for every sufficiently large tt,

Zt(r) ≤ A.Z_t(r)\ \le\ A .

Proof

Let n=⌊t⌋n=\lfloor t\rfloor and let y1,…,yny_1,\ldots,y_n be the points of PtP_t in cyclic order, indices mod nn; let Si−r(t)=yi−yi−rS_{i-r}(t)=y_i-y_{i-r} be the rr-span ending at yiy_i (the clockwise distance from yi−ry_{i-r} to yiy_i), and take tt large enough that r<nr<n and every span is shorter than 11.

Two decompositions. For all xx outside the finite set of endpoints,

Nt((x,x+r/t])=∑i=1n1(yi−r/t, yi](x),N_t\bigl((x,x+r/t]\bigr)=\sum_{i=1}^n\mathbf 1_{(y_i-r/t,\,y_i]}(x),

since yi∈(x,x+r/t]y_i\in(x,x+r/t] exactly when x∈[yi−r/t,yi)x\in[y_i-r/t,y_i). On the other hand, the rr-span arcs (yi−r,yi](y_{i-r},y_i] cover every point of the circle, apart from endpoints, exactly rr times: xx lies in (yi−r,yi](y_{i-r},y_i] exactly when yiy_i is one of the rr points following xx. So

r=∑i=1n1(yi−r, yi](x).r=\sum_{i=1}^n\mathbf 1_{(y_{i-r},\,y_i]}(x).

Pairing arcs with the same right endpoint. Subtracting,

Δt(x,r)=∑i=1n(1(yi−r/t, yi](x)−1(yi−r, yi](x)).\Delta_t(x,r)=\sum_{i=1}^n\Bigl(\mathbf 1_{(y_i-r/t,\,y_i]}(x) -\mathbf 1_{(y_{i-r},\,y_i]}(x)\Bigr).

The two arcs in the ii-th term share the right endpoint yiy_i and have lengths r/tr/t and Si−r(t)S_{i-r}(t), so their indicators differ on an arc of length ∣Si−r(t)−r/t∣|S_{i-r}(t)-r/t|. By the triangle inequality and the bound ∑i∣Si(t)−r/t∣≤2A+r(t−n)/t\sum_i|S_i(t)-r/t|\le2A+r(t-n)/t from the proof of Lemma 6.1,

∫T∣Δt(x,r)∣ dx ≤ ∑i=1n∣Si−r(t)−rt∣ ≤ 2A+r(t−n)t.\int_{\mathbb T}\bigl|\Delta_t(x,r)\bigr|\,dx\ \le\ \sum_{i=1}^n \Bigl|S_{i-r}(t)-\frac rt\Bigr|\ \le\ 2A+\frac{r(t-n)}t .

The mean. Each point lies in (x,x+r/t](x,x+r/t] for xx in a set of measure r/tr/t, so

∫TΔt(x,r) dx=rnt−r=−r(t−n)t.\int_{\mathbb T}\Delta_t(x,r)\,dx=\frac{rn}t-r=-\frac{r(t-n)}t .

Conclusion. The positive part of a function is half the sum of its absolute value and the function itself, so

Zt(r)=12(∫T∣Δt(x,r)∣ dx+∫TΔt(x,r) dx) ≤ 12(2A+r(t−n)t−r(t−n)t)=A.Z_t(r)=\frac12\Bigl(\int_{\mathbb T}|\Delta_t(x,r)|\,dx +\int_{\mathbb T}\Delta_t(x,r)\,dx\Bigr) \ \le\ \frac12\Bigl(2A+\frac{r(t-n)}t-\frac{r(t-n)}t\Bigr)=A .

Role in the argument

This is the starting value zt(r)=Zt(r)/r≤A/r=θ2z_t(r)=Z_t(r)/r\le A/r=\theta^2 of the scale iteration in Proposition 6.4.