Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Greedy block charge
Proved range contraction, not a resolution. A least-budget minimal-ratio core for necessarily satisfies
This still leaves an unbounded range above the hereditary span threshold.
Throughout, , , , is -free, and
We use the hereditary caps and the independent-set expansion . Independence-criticality is unnecessary. Write
and , . Thus
1. An ordered partition with controlled holes
Repeatedly choose a minimum-degree vertex in the remaining graph and remove its closed neighborhood. Let the resulting blocks be , with , and let count their internal nonedges. If is the residual just before block , put
The edges removed at this step number exactly
Consequently . The budget gives
Thus , while the chosen pivots are independent, so .
Set , , and
Here , so nonnegativity of follows from integrality. Summing (2) gives the exact identity
2. A matching charge for an arbitrary induced set
Let have order , and let have edges. If has a matching of size , then
For a matched pair , the number of common neighbors in is . Independent-pair expansion therefore gives
Let be the matching endpoints. Then
For the second inequality, the matched edges contribute . Between two matched pairs, the number of cross edges minus their contributions to the two common-neighbor counts is at most two: for zero through four cross edges the respective maxima are . Common neighbors outside only reduce the expression. Summing and using proves (4).
3. Charging all blocks
Every block satisfies
For , this is immediate. If , -freeness supplies a missing edge; apply (4) with .
If , the cap on vertices gives . A graph with matching number at most one has all its edges in a star or a triangle. The star case would leave an independent -set in , hence a in . The triangle case has at most three edges, less than . Therefore contains a matching of size two, and (4) gives .
For , minimum residual degree implies
The hereditary cap gives ; the final inequality holds at every size at least , since and every subsequent increment is at least one. The sum exceeds the right side of (6) by
Summing (6) and using (1) and (3) yields
Thus
Since , we obtain . For the last summand is less than , and integrality gives the stated theorem.
4. Why this has not produced an iteration
Before block , write the remaining order as , with independence exactly . The original proper-induced ratio bounds give
for independent in that residual. If the next block has order and its complementary graph has a matching of size , the calculation in (5) instead gives
In particular a block of order costs at least in the exact edge decomposition (2).
Initially ; deleting a block updates to . For proper nonempty suffixes , and after the last block it is zero. Positive deviations can accumulate and weaken (8), so the range contraction cannot simply be repeated with a smaller parameter. No universal contradiction from these suffix charges has been proved.
The exact removal count, the matching charge, and all four cases of (6) were checked separately. No search or external theorem is used.
Status of the renewed closure attempt
No stronger valid descent was obtained. The complementary stability partition has at most target cross edges, but this total alone does not extend an internal nonedge to a larger independent set: its endpoints may remove a whole neighboring list. Concentrating on parts incident with many cross edges did not handle dependency chains through many lighter parts. No safe suppression or termination theorem for those chains was proved.
Several proposed scalar block schedules were discarded because they omitted the final clique condition or the global minimum-degree constraint. They are not counterexamples or certificates of feasibility. The established obstruction to naive iteration remains exactly (8): prefix deletion weakens expansion by the accumulated deficiency .