Wiki
Wiki

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

Updated


Statement

Setting (pp. 1--2). Nf+(x)N_f^+(x) is the set of n≤xn\le x with n=k+f(k)n=k+f(k) for some kk, and in the bounds it stands for its size; φ\varphi is Euler's totient function. For λ∈[0,1]\lambda\in[0,1] put Φx(λ)=x−1#{k≤x:φ(k)/k≤λ}\Phi_x(\lambda)=x^{-1}\#\{k\le x:\varphi(k)/k\le\lambda\}; the paper recalls that Φ(λ)=lim⁡x→∞Φx(λ)\Phi(\lambda)=\lim_{x\to\infty}\Phi_x(\lambda) exists for each λ∈[0,1]\lambda\in[0,1] and is an increasing singular function.

Theorem 1.4 (p. 2).

x≪Nφ+(x)≤(12+∫01Φ(t) dt(1+t)2+o(1))x.x\ll N_\varphi^+(x)\le\Bigl(\frac12+\int_0^1\frac{\Phi(t)\,dt}{(1+t)^2}+o(1)\Bigr)x .

The lower bound says that the integers of the form k+φ(k)k+\varphi(k) have positive lower density. The upper bound is not given a numerical value in the theorem; the paper reports that numerical values of Φx\Phi_x at x=105x=10^5 predict ∫01Φ(t) dt/(1+t)2<0.17\int_0^1\Phi(t)\,dt/(1+t)^2<0.17, which would give Nφ+(x)<0.67xN_\varphi^+(x)<0.67x (p. 2), and that numerical calculations predict Nφ+(x)≈0.37xN_\varphi^+(x)\approx0.37x (p. 1). Neither is proved. The proved numerical upper bound is 0.93x0.93x, from Theorem 1.3.

Proof pointer

Section 5, pp. 11--21. Lower bound (§5.1, pp. 11--21): following Luca and Pomerance's method for σ(k)−k\sigma(k)-k, the authors take the set AA of n=mp∈(x/2,x]n=mp\in(x/2,x] with pp prime and m≤x7/15m\le x^{7/15} drawn from a set BB of integers kqrkqr with properties that hold for almost all integers (Lemmas 5.1 to 5.4, pp. 11--13); then ∣A∣≫x|A|\gg x (p. 14, (5.4)). The number EE of pairs in A2A^2 with equal values n+φ(n)n+\varphi(n) is shown to be O(x)O(x), grouping the pairs by the common yy-smooth part dd of mm and m′m' and applying Selberg's sieve (pp. 15--21), and Cauchy--Schwarz gives ≫x\gg x values up to 2x2x (pp. 14--15). Upper bound (§5.2, p. 21): if k>λxk>\lambda x and φ(k)/k>(1−λ)/λ\varphi(k)/k>(1-\lambda)/\lambda then k+φ(k)>xk+\varphi(k)>x; summing over a fine partition of λ∈(1/2,1)\lambda\in(1/2,1) with the uniform estimate Φx(λ)=Φ(λ)+O((log⁡x)−1(log⁡log⁡x/log⁡log⁡log⁡x)2)\Phi_x(\lambda)=\Phi(\lambda)+O\bigl((\log x)^{-1}(\log\log x/\log\log\log x)^2\bigr) gives x−Nφ+(x)≥(1/2−∫01Φ(t) dt/(1+t)2+o(1))xx-N_\varphi^+(x)\ge(1/2-\int_0^1\Phi(t)\,dt/(1+t)^2+o(1))x.

Read depth

Claims checked: the statement and its setting were read on the print (arXiv v1, pp. 1--2), and the proof of the upper bound on p. 21 was followed. The lower-bound proof of §5.1 was read for structure only. Nothing here is independently reviewed.

Dependencies

None in the corpus. External inputs named by the paper: Luca and Pomerance's method for the range of σ(k)−k\sigma(k)-k, results of Hall and Tenenbaum, Erdős, Luca and Pomerance, and De Koninck and Luca behind Lemmas 5.1 to 5.3, and the uniform distribution estimate (5.16) for φ(k)/k\varphi(k)/k (Postnikov, Fainleib).

Source. M. R. Gabdullin, V. V. Iudelevich and F. Luca, Numbers of the form k+f(k)k+f(k), J. Number Theory 262 (2024), 58--85, doi:10.1016/j.jnt.2024.03.010; arXiv:2306.16035. Labels and pages are those of the arXiv v1 edition named on the source card.

Bears on

  • Problem 822: the lower bound Nφ+(x)≫xN_\varphi^+(x)\gg x says the integers of the form n+φ(n)n+\varphi(n) have positive lower density, which answers the problem's question yes.