Wiki
Wiki

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

Updated


Source. Fornal–Sun, Remark 1, equations (32)–(35), p. 12 of arXiv v1.

Source sketch; full quantitative proof not reconstructed. A materially different proposed partition groups integers n≤dn\le d by the numbers

ωi(n)=∣{p prime:vp(n)≥i}∣.\omega_i(n)=|\{p\text{ prime}:v_p(n)\ge i\}|.

Taking 1≤i≤⌊log⁡d/log⁡2⌋1\le i\le\lfloor\log d/\log2\rfloor includes every possible nonzero exponent. Each integer has one such vector, so its nonempty profile classes indeed form a partition. The source writes L=⌊log⁡d⌋L=\lfloor\log d\rfloor; for natural logarithms that length can omit the highest exponent of a power of 2. The base-2 length above is the appropriate complete-profile convention.

For these classes, the source asserts

Ti≪exp⁡ ⁣((2+o(1))log⁡dlog⁡log⁡d),T_i\ll\exp\!\left((2+o(1))\sqrt{\log d\log\log d}\right),

where TiT_i is the normalized gcd row sum defined in Proposition 2.1. It points to the Hardy–Ramanujan partition estimate, a divisor-function mean estimate, and an elementary enumeration, without writing that argument. This page records the proposed method and the asserted bound; it does not certify the omitted enumeration or its leading constant.

The logarithms are multiplied under the square root in this remark. That is a weaker scale than the quotient in Proposition 2.1 and Theorem 1.1. The remark is not used anywhere in the compiled main proof chain. No claim about the current status of a separately proposed improvement follows from it.

Dependencies and limits. The partition itself is elementary. The quantitative estimate remains a source sketch, with the exact partition/divisor estimates and their application not supplied here.

Bears on. Problem 202, as a possible alternative route to related structural bounds.