The n-generic master reduction for the ACMAX conjecture #
For every n ≥ 12, the upper-bound clause of Kolokolnikov's Conjecture 1.5
(arXiv:1412.6147) reduces to the residual predicate ResidualCore. The residual
hypothesis is discharged downstream in LeanPool.ACMax.Band.Final; the full
entry point also covers orders 4 ≤ n ≤ 11.
The reduction #
Let G be a graph on Fin n (n ≥ 12) with 2(n-2) edges. Then algConn G ≤ 2 follows
unconditionally in each of the following cases, via the generic certificate layer (each of the
lemmas below is already stated for an arbitrary finite vertex type):
Gdisconnected →algConn_le_two_of_not_connected;- some vertex has degree
≤ 2→algConn_le_two_of_low_degree_vertex(the complement of the closed neighbourhood is nonempty sincen ≥ 12 > 3); - otherwise
δ ≥ 3, and the handshake∑ deg = 4n − 8 = 3n + (n − 8)forces at least8vertices of degree exactly3(card_deg3_ge_eight: each degree-≥4vertex absorbs at least one unit of the excessn − 8, so|D| ≥ 8); - a good triangle — three mutually adjacent vertices with
n·(∑deg − 6) ≤ 2·(3·(n−3))— is a sparse weighted cut →algConn_le_two_of_weighted_cut(the triangle sends exactly∑deg − 6edges out). Similarly a goodC₄(n·(∑deg − 8) ≤ 2·(4·(n−4))) →algConn_le_two_of_good_C4, and a goodK_{2,3}(n·(∑deg − 12) ≤ 2·(5·(n−5))) →algConn_le_two_of_good_K23. Thesen-uniform thresholds are exactly the hypotheses the generic cut lemmas consume — no slack is given away (atn = 19they specialise to the familiar∑deg ≤ 11 / ≤ 14 / ≤ 19; see the sanityexamples below); - no good triangle, and every degree-
3vertex has a degree-3neighbour → the degree-3setDis triangle-free (a triangle inDhas∑deg = 9, good forn ≥ 6), has min-D-degree≥ 1and max-D-degree≤ 3, and|D| ≥ 8, soexists_induced_2K2_of_triangleFree_smalldegyields an induced2K₂of degree sum12→algConn_le_two_of_ind_2K2; - the residual: what survives is exactly
ResidualCore n G— connected,δ ≥ 3,2(n−2)edges, an isolated degree-3vertex (one with no degree-3neighbour — the precise negation of the hypothesis step 5 consumes), no good triangle, no induced2K₂on degree-3vertices, no goodC₄, no goodK_{2,3}.
The conditional dispatcher algConn_le_two_of_card_general_cond takes the
residual case as an explicit hypothesis. residual_algConn_le_two in
LeanPool.ACMax.Band.Final proves that hypothesis from the unconditional
conjecture. The numeric examples below check the cut thresholds at order 19.
The n-uniform certificate predicates #
Each predicate carries its degree-sum threshold in the exact n-uniform form consumed by the
corresponding generic cut lemma, so that the dispatch below gives away no slack.
A good triangle: three mutually adjacent vertices whose degree sum satisfies the
n-uniform weighted-cut inequality n·(∑deg − 6) ≤ 2·(3·(n−3)). (The triangle A = {x,y,z}
sends exactly ∑deg − 6 edges to Aᶜ, so this is precisely the hypothesis of
algConn_le_two_of_weighted_cut.) At n = 19 this is ∑deg ≤ 11.
Equations
- One or more equations did not get rendered due to their size.
Instances For
An induced 2K₂ on degree-3 vertices: four distinct vertices of degree 3 spanning
exactly the two edges ab, cd. Its degree sum is 12, so algConn_le_two_of_ind_2K2
applies (for every n).
Equations
- One or more equations did not get rendered due to their size.
Instances For
A good C₄: an induced 4-cycle a-b-c-d-a with the n-uniform threshold
n·(∑deg − 8) ≤ 2·(4·(n−4)) — exactly the hypothesis of algConn_le_two_of_good_C4.
At n = 19 this is ∑deg ≤ 14.
Equations
- One or more equations did not get rendered due to their size.
Instances For
A good K_{2,3}: an induced complete bipartite K_{2,3} (parts {a,b}, {c,d,e})
with the n-uniform threshold n·(∑deg − 12) ≤ 2·(5·(n−5)) — exactly the hypothesis of
algConn_le_two_of_good_K23. At n = 19 this is ∑deg ≤ 19.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The residual predicate #
The residual core — exactly the class of graphs that survives the generic reduction steps 1–5 (see the module docstring). Every field is the precise negation of the branch condition the corresponding step consumes; nothing is weakened and nothing extraneous is added:
connected— step 1 (disconnected graphs are closed byalgConn_le_two_of_not_connected);min_degree— step 2 (a degree-≤2vertex is closed byalgConn_le_two_of_low_degree_vertex);edge_card,n_ge— the standing hypotheses (from which≥ 8degree-3vertices follow,card_deg3_ge_eight);no_good_triangle— step 4a (weighted cut on a triangle);iso_deg3— the negation of "every degree-3vertex has a degree-3neighbour" (step 5 consumes that hypothesis to extract an induced2K₂inside the degree-3set);no_deg3_ind2K2— step 5's certificate directly (an induced2K₂on degree-3vertices closes the graph regardless of how it was found);no_good_C4,no_good_K23— steps 4b, 4c.
Every such residual graph is closed by residual_algConn_le_two in
LeanPool.ACMax.Band.Final.
- connected : G.Connected
- no_good_triangle : ¬HasGoodTriangle n G
- no_deg3_ind2K2 : ¬HasDeg3Ind2K2 n G
- no_good_K23 : ¬HasGoodK23 n G
Instances For
Step 3: the handshake counting lemma #
Handshake counting. If every degree is ≥ 3 and G has 2(n−2) edges, then at least
8 vertices have degree exactly 3: with D = {deg = 3} and H = {deg ≥ 4},
3|D| + 4|H| ≤ ∑deg = 4n − 8 and |D| + |H| = n give |D| ≥ 8.
The master reduction #
The n-generic master reduction (n ≥ 12). Every simple graph on Fin n with
exactly 2(n-2) edges has algebraic connectivity at most 2, given the residual lemma
residual_algConn_le_two. Steps 1–5 of the dispatch (disconnected / low degree / good
triangle / triangle-free induced 2K₂ / induced 2K₂ on degree-3 vertices / good C₄ /
good K_{2,3}) are proved here uniformly in n. The hypothesis hres records
the residual case and is discharged by the downstream unconditional theorem.
Sanity checks at n = 19 #
The n-uniform thresholds specialize at n = 19 to these numerical bounds: