Skip to content

feat(Combinatorics/SimpleGraph/Inversion): add Möbius inversion between copy counts and induced copy counts#36

Open
FordUniver wants to merge 3 commits into
diffbase/copy-mobiusfrom
feat/copy-mobius
Open

feat(Combinatorics/SimpleGraph/Inversion): add Möbius inversion between copy counts and induced copy counts#36
FordUniver wants to merge 3 commits into
diffbase/copy-mobiusfrom
feat/copy-mobius

Conversation

@FordUniver

@FordUniver FordUniver commented May 27, 2026

Copy link
Copy Markdown
Owner

Adds the bijection Copy.equivSigmaEmbedding decomposing every copy of G in H as a choice of induced shape G' ∈ Icc G ⊤ together with an embedding G' ↪g H, and derives the labelled count identities: a forward sum H.copyCount G = ∑ G' ∈ Icc G ⊤, H.embeddingCount G' and its Möbius inverse over with sign (-1) ^ #(G'.edgeFinset \ G.edgeFinset). The Möbius direction reduces via the order-equivalence iccEquivPowersetEdgeFinsetSdiff between Icc G K and the powerset of the edge difference to Finset.sum_powerset_neg_one_pow_card. Supporting material on Copy.inducedShape (the pullback H.comap f.toEmbedding) and the per-fiber bijection Copy.fiberInducedShapeEquiv lives in the same file rather than being split out, since neither has a standalone consumer outside this Möbius decomposition.

Co-authored-by: Malte Jackisch 45597826+MaltyBlanket@users.noreply.github.com


Came up while preparing density-level versions of the same identity for a downstream flag-algebras formalisation. The labelled count version, stated over ℤ for the signs, is the standalone-mergeable kernel; density transports (over ℚ, dividing by descending-factorials or binomial coefficients) and unlabelled-count variants follow in separate PRs.

Diff for the changes just in this PR over its predecessors: link

@FordUniver FordUniver changed the title feat(Combinatorics/SimpleGraph/Inversion): Möbius inversion feat(Combinatorics/SimpleGraph/Inversion): add Möbius inversion between copy counts and induced copy counts May 27, 2026
@FordUniver FordUniver changed the base branch from feat/simplegraph-aut to diffbase/copy-mobius May 27, 2026 15:55
@FordUniver FordUniver force-pushed the feat/copy-mobius branch 3 times, most recently from b664e44 to d9da9e5 Compare May 28, 2026 05:54
@FordUniver FordUniver force-pushed the diffbase/copy-mobius branch from fa6e3bb to 3ae9341 Compare May 28, 2026 06:40
@FordUniver FordUniver force-pushed the feat/copy-mobius branch 4 times, most recently from ac4cb3d to 57f3d6b Compare May 28, 2026 09:44
FordUniver and others added 2 commits July 12, 2026 04:14
…opy counts and induced copy counts

Adds the bijection `Copy.equivSigmaEmbedding`, the labelled forward count
identity `copyCount_eq_sum_embeddingCount`, and its Möbius inverse
`embeddingCount_eq_sum_signed_copyCount` over ℤ. The Möbius direction reduces
via the order-equivalence `iccEquivPowersetEdgeFinsetSdiff` between `Icc G K`
and the powerset of the edge difference to `Finset.sum_powerset_neg_one_pow_card`.

Supporting `Copy.inducedShape` (the pullback `H.comap f.toEmbedding`) and the
per-fiber bijection `Copy.fiberInducedShapeEquiv` are kept inline in this file
rather than split out, since neither has a standalone consumer outside this
Möbius decomposition.

Co-authored-by: Malte Jackisch <45597826+MaltyBlanket@users.noreply.github.com>
Clean-slate restructure now complete:

* `edgeSetSdiffFinset G K` constructed via `Set.Finite.toFinset` — no per-graph
  `Fintype` synthesis required.
* `natCard_edgeSet_eq_add_card_sdiff` (cardinality bridge): proved via
  `Set.ncard_diff_add_ncard_of_subset` + `Nat.card_coe_set_eq` /
  `Set.ncard_eq_toFinset_card`.
* `iccEquivPowersetEdgeFinsetSdiff`: `left_inv` / `right_inv` proved by
  case-splitting on `fromEdgeSet_adj` membership and using the diag-freeness
  of edges in `K`.
* Kernel split into degenerate (`Icc G G`, trivial) and strict (`G < L`,
  reindex + powerset alternating sum + non-empty extra-edge finset).
* Main theorem: substitute forward identity, distribute, swap sums, factor,
  apply `Finset.sum_eq_single G` with the strict case discharging non-`G`
  terms.

Zero `sorry`, zero `Classical.decRel`, zero `open Classical in`, zero
`attribute [local instance]` in the file. All decidability requirements
satisfied by `Set.Finite.toFinset` (noncomputable but Classical-free at the
source level) and the bridge `Nat.card_coe_set_eq`.
…er catch-up

Rename Set.{mem_diff,diff_eq_empty,ncard_diff_add_ncard_of_subset} to the
sdiff spellings; omit unused DecidableEq/DecidableLE section vars on the
edgeSetSdiffFinset helpers; weaken edgeSet_sdiff_finite's Fintype V to Finite V
per the unusedFintypeInType linter.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant