Wiki
Wiki

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

Updated

Erdős 699: full small-prime overlap in the remaining index-six position pattern


This is a retained earlier research note. The current reading guide gives the latest status and distinguishes completed proofs from remaining cases.

The exact localization lemma permits replacing a selected small-prime power PpP_p by

Ep=pmin⁡{vp(n−rp),vp(nj)}.E_p=p^{\min\{v_p(n-r_p),v_p\binom nj\}}.

This can strengthen a polynomial certificate when the column binomial has few carries at one of the selected primes. In the unresolved position pattern

(r2,r3,r5)=(2,1,0),(r_2,r_3,r_5)=(2,1,0),

the already proved row-zero restriction leaves column residue classes where all three overlaps equal the original prime powers.

Proposition. Suppose (n,6,j)(n,6,j) is bad and the selected positions are (2,1,0)(2,1,0). Put

A2=v2(n−2),A3=v3(n−1),A5=v5(n).A_2=v_2(n-2),\qquad A_3=v_3(n-1),\qquad A_5=v_5(n).

Then

n≡250(mod300).n\equiv250\pmod{300}.

For any integer column jj with

j≡11,23,47, or 59(mod60),j\equiv11,23,47,\text{ or }59\pmod{60},

one has

vp(nj)≥Ap(p=2,3,5),E2=2A2,E3=3A3,E5=5A5.v_p\binom nj\ge A_p\quad(p=2,3,5),\qquad E_2=2^{A_2},\quad E_3=3^{A_3},\quad E_5=5^{A_5}.

Here the column must lie in 0≤j≤n0\le j\le n for the binomial assertion. The overlap conclusion itself only needs the displayed conditions on the row and its three prime-power factors; it does not require badness.

Proof. Maximality at r2=2r_2=2 forces A2≥2A_2\ge2, hence n≡2(mod4)n\equiv2\pmod4. The selected factor at r3=1r_3=1 is divisible by three, so n≡1(mod3)n\equiv1\pmod3. Therefore neither 8∣n8\mid n nor 9∣n9\mid n. The forced valuation-gap proposition then supplies 25∣n25\mid n, giving n≡250(mod300)n\equiv250\pmod{300}.

Each listed column has j≡3(mod4)j\equiv3\pmod4, j≡2(mod3)j\equiv2\pmod3, and 5∤j5\nmid j. In Kummer's formula

vp(nj)=∑k≥11{j mod pk>n mod pk},v_p\binom nj=\sum_{k\ge1} \mathbf1_{\{j\bmod p^k>n\bmod p^k\}},

the first A2A_2 summands at two are all one: at level one the residues are 1>01>0, and at every level 2≤k≤A22\le k\le A_2 the row residue is two while the column residue is at least three. At three, every level 1≤k≤A31\le k\le A_3 has row residue one and column residue at least two. At five, every level 1≤k≤A51\le k\le A_5 has row residue zero and nonzero column residue. This proves all three valuation bounds and hence all three equalities for EpE_p. □\square

These column classes also satisfy the forced row-zero gap in its strongest simple form: v5(j)=0v_5(j)=0 and v5(n)≥2v_5(n)\ge2. They are closed under column reflection, because n≡10(mod60)n\equiv10\pmod{60} interchanges 1111 with 5959 and 2323 with 4747.

Whether such columns also avoid every common prime at least seven remains unresolved. An argument using the actual overlaps must obtain an additional restriction from those large primes; the selected-position pattern and the known row-zero valuation gap alone provide no power saving in Ep/PpE_p/P_p.