Wiki
Wiki

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

Updated

Lev 2004 reconstructing integer sets representation functions

../

construction_p4: Lev's single greedy perfect difference set: step n adds z_n and z_n + d_n, where d_n is the least difference not yet represented and z_n creates no non-trivial equal differences; the paper states that the nth element is O(n^3), so the counting function is at least of order x^(1/3), against the order x^(1/2) that bounds every perfect difference set in N.

theorem_1: Dombi's theorem, reproved in Lev's paper: the partition of the positive integers by the sign function T with T(1) = 1, T(2n) = -T(2n-1) and T(2n+1) = T(n+1) gives two sets A and B with the same number of representations n = a1 + a2, a1 < a2, for every positive integer n.

theorem_2: Chen and Wang's theorem, reproved in Lev's paper: the partition of the positive integers by the sign function T with T(1) = 1, T(2n) = -T(2n-1) and T(2n+1) = -T(n+1) gives two sets A and B with the same number of representations n = a1 + a2, a1 <= a2, for every integer n >= 3.

theorem_3: Lev's partition of the positive integers into infinitely many sets A_k, each a perfect difference set (every non-zero integer is uniquely a difference of two of its elements), such that every intersection of A_i with a translate A_j + z, z a positive integer, has at most two elements.


Lev, Vsevolod F., Reconstructing integer sets from their representation functions. Electron. J. Combin. 11 (2004), no. 1, Research Paper 78, 6 pp. doi:10.37236/1831. The copy read for this card is the journal's PDF; it prints no notice; the journal's article page (https://www.combinatorics.org/ojs/index.php/eljc/article/view/v11i1r78, read 2026-10-02) shows no license, and the journal's About page says it was "one of the first journals to leave copyright with authors" and names no Creative Commons license, so the authors' copyright governs with no reuse grant stated, every other right reserved.

Source: https://www.combinatorics.org/ojs/index.php/eljc/article/view/v11i1r78.

For A⊆ZA\subseteq\mathbb Z the paper compares the counts RA(1)(n)R_A^{(1)}(n), RA(2)(n)R_A^{(2)}(n) and RA(3)(n)R_A^{(3)}(n) of representations n=a1+a2n=a_1+a_2 with a1,a2∈Aa_1,a_2\in A, unrestricted, with a1<a2a_1<a_2, and with a1≤a2a_1\le a_2, and asks how far they determine AA. Theorems 1 (Dombi) and 2 (Chen and Wang) give partitions N=A∪B\mathbb N=A\cup B by a sign recursion with RA(2)=RB(2)R_A^{(2)}=R_B^{(2)} everywhere and RA(3)(n)=RB(3)(n)R_A^{(3)}(n)=R_B^{(3)}(n) for n≥3n\ge3; the paper proves both by one generating-function identity and remarks that the constructions are essentially unique (p. 3). For differences, Theorem 3 partitions N\mathbb N into infinitely many perfect difference sets AkA_k with ∣Ai∩(Aj+z)∣≤2|A_i\cap(A_j+z)|\le2 for all i,j,z∈Ni,j,z\in\mathbb N, so no three elements of a part reappear, shifted by a positive integer, in the same or another part. Section 3 simplifies that construction to a single greedy perfect difference set, whose nnth element the paper states is O(n3)O(n^3), and poses five open problems, the first on the largest possible counting function of a perfect difference set in N\mathbb N. Labels and pages here are those of the journal's PDF (pp. 1--6).

Bears on. #1194: the greedy perfect difference set (pp. 4--5) is a set in which every positive integer is uniquely a difference of two members, and the paper states that its nnth element is O(n3)O(n^3), from counts that bound the two numbers added at step nn. The problem's claim page derives from this an upper bound an≪n3a_n\ll n^3 for that set. The paper states no bound for ana_n itself; the only bound it states for every perfect difference set in N\mathbb N is A(x)≪x1/2A(x)\ll x^{1/2} on the counting function, not a bound on ana_n. So it does not settle how fast an/na_n/n must grow.

Results. Page numbers are those of the journal's PDF.

  • Theorem 1 (Dombi; p. 2, proof p. 3): the partition by T(1)=1T(1)=1, T(2n)=−T(2n−1)T(2n)=-T(2n-1), T(2n+1)=T(n+1)T(2n+1)=T(n+1) has RA(2)(n)=RB(2)(n)R_A^{(2)}(n)=R_B^{(2)}(n) for all n∈Nn\in\mathbb N.
  • Theorem 2 (Chen and Wang; p. 2, proof p. 3): the partition by T(1)=1T(1)=1, T(2n)=−T(2n−1)T(2n)=-T(2n-1), T(2n+1)=−T(n+1)T(2n+1)=-T(n+1) has RA(3)(n)=RB(3)(n)R_A^{(3)}(n)=R_B^{(3)}(n) for all integer n≥3n\ge3.
  • Theorem 3 (p. 2, proof pp. 3--4): a partition of N\mathbb N into perfect difference sets A1,A2,…A_1,A_2,\ldots with ∣Ai∩(Aj+z)∣≤2|A_i\cap(A_j+z)|\le2 for all i,j,z∈Ni,j,z\in\mathbb N.
  • Greedy perfect difference set (pp. 4--5, unnumbered): a single perfect difference set in N\mathbb N whose nnth element the paper states is O(n3)O(n^3), so A(x)≫x1/3A(x)\gg x^{1/3}, with Problem 1 (p. 5) on whether A(x)≫x1/2A(x)\gg x^{1/2} is possible.

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