Wiki
Wiki

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

Updated


Source: arXiv v2, p. 2, Theorem 2 and its proof.

Statement

For every integer j≥5j\ge5, some minimal covering system consists of jj classes with distinct moduli, and these moduli, in increasing order, are:

21<22<⋯<2j−4<3⋅2j−5<2j−3<3⋅2j−4<3⋅2j−3.2^1<2^2<\cdots<2^{j-4}<3\cdot2^{j-5}<2^{j-3} <3\cdot2^{j-4}<3\cdot2^{j-3}.

When j=5j=5, the initial string 21,…,2j−42^1,\ldots,2^{j-4} consists of the single term 22.

Full proof

For 1≤i≤j−31\le i\le j-3, let

Di=2i−1+2iZ.D_i=2^{i-1}+2^i\mathbb Z.

For k∈{0,1,2}k\in\{0,1,2\}, let AkA_k be the simultaneous congruence class

n≡k(mod3),n≡0(mod2j−5+k).n\equiv k\pmod 3, \qquad n\equiv0\pmod {2^{j-5+k}}.

The Chinese remainder theorem makes AkA_k one progression of modulus 3⋅2j−5+k3\cdot2^{j-5+k}. Define

Cj={D1,…,Dj−3,A0,A1,A2}.\mathcal C_j=\{D_1,\ldots,D_{j-3},A_0,A_1,A_2\}.

If 2j−3∤n2^{j-3}\nmid n, there is a unique i∈{1,…,j−3}i\in\{1,\ldots,j-3\} with 2i−1∣n2^{i-1}\mid n and 2i∤n2^i\nmid n. Then n∈Din\in D_i. If 2j−3∣n2^{j-3}\mid n, choose the unique k∈{0,1,2}k\in\{0,1,2\} congruent to nn modulo 33. The divisibility condition for AkA_k is at most 2j−3∣n2^{j-3}\mid n, so n∈Akn\in A_k. Hence Cj\mathcal C_j covers every integer.

The dyadic moduli are 2,22,…,2j−32,2^2,\ldots,2^{j-3}, and the other three are 3⋅2j−53\cdot2^{j-5}, 3⋅2j−43\cdot2^{j-4}, and 3⋅2j−33\cdot2^{j-3}. They are distinct. The inequalities 2<3<42<3<4 put the first new modulus strictly between 2j−42^{j-4} and 2j−32^{j-3}; the remaining two occur after 2j−32^{j-3} in the displayed order. Thus the list has exactly jj entries in the claimed order.

It remains to prove minimality. The following give an integer covered by each class and by no other class.

  • If 1≤i≤j−41\le i\le j-4, the integer 2i−12^{i-1} lies in DiD_i and in no other DhD_h. For i≤j−5i\le j-5 its 22-adic valuation is too small for every AkA_k. For i=j−4i=j-4, it can meet only the divisibility condition of A0A_0, but the nonzero residue 2j−5(mod3)2^{j-5}\pmod3 excludes A0A_0.
  • For Dj−3D_{j-3}, choose an odd uu with 2j−4u≡2(mod3)2^{j-4}u\equiv2\pmod3. Then 2j−4u2^{j-4}u has exact 22-adic valuation j−4j-4, so it lies in Dj−3D_{j-3}, misses A2A_2, and its residue modulo 33 excludes A0A_0 and A1A_1.
  • For each k∈{0,1,2}k\in\{0,1,2\}, the Chinese remainder theorem supplies nkn_k with 2j−3∣nk2^{j-3}\mid n_k and nk≡k(mod3)n_k\equiv k\pmod3. It lies in AkA_k, in no DiD_i, and in neither of the other two AA-classes.

Every member therefore has a private witness. Deleting it leaves that witness uncovered, so Cj\mathcal C_j is minimal.

The source states minimality from its two partition statements but does not spell out the overlaps of A0,A1A_0,A_1 with the last dyadic classes. The private witnesses above are the compilation's explicit completion of that step.

Bears on

  • Problem 2: for each j≥5j\ge5 the system is a minimal distinct cover whose jj-th smallest modulus is 3⋅2j−33\cdot2^{j-3}, which the paper presents as complementing Theorem 1. Its least modulus is 22, so it says nothing new about the least-modulus question itself.
  • Problem 1188, as structural information about minimal distinct covering systems rather than an estimate for their total number.