RESEARCH / Focused target
All-direction maximal clique partitions of mixed-alphabet Hamming graphs
RELEASED FOUNDATIONS / PROPOSED CONNECTION
Research connections
All source anchors (1)
- D5/S3/Combinatorics/Hamming/HammingMultipartitePartition.resultAbsent from this graph
Problem
The question
Ahmed and Kunjwal, Characterizing unitaries via quasi-process functions, arXiv:2610.00579v1, Conjecture V.1 in Section V.2, ask for genuine multipartiteness of maximal-clique partitions whenever the number of coordinates exceeds three. The surrounding definitions mean that every coordinate direction must occur. Alphabet sizes are independent and at least two. Definition V.1 defines adjacency by disagreement at exactly one coordinate; maximal means inclusion-maximal, not maximum cardinality.
The exact formal statement is: for every natural n with 4 ≤ n, every
d : Fin n → ℕ with ∀ i, 2 ≤ d i, and
V = (i : Fin n) → Fin (d i), there is a set P : Set (Set V) such that
- Every
S ∈ Pis nonempty, is a full coordinate line, and satisfiesMaximal (graph d).IsClique S. - For every
x : V, there is exactly oneSsuch thatS ∈ P ∧ x ∈ S. - For every
i : Fin n, somex : Vhasline i x ∈ P.
There is no equal-alphabet assumption, upper dimension bound, finite search bound, or assumed binary matching supplier.
Motivation
Our foothold
The graph partition is the source's combinatorial formulation of using each party in mutually exclusive controlled-circuit constructions. The Lean result supplies the entire graph statement. It does not formalize the circuit interpretation, unitary operators, or quasi-process functions.
Gap
Missing bridges
Issue #13221 preregisters the complete universal assertion before any Lean probe. The accessible-source check is recorded in the matching literature note and in the issue. No exact supplier was found in the relevant project Hamming declarations or pinned Mathlib SimpleGraph APIs. Mathlib's clique, maximality, dependent update, and finite tuple APIs are reused directly. The source's binary case is attributed to Erde, arXiv:2404.03950; the binary construction is an internal ingredient, not an independent novelty claim.
Route
Proposed approach
A binary selector chooses a coordinate at each vertex and keeps that choice
when the chosen coordinate is flipped. The four-coordinate seed selects the
edges 0110–1110, 0111–1111, 1000–1100, 1001–1101, 0000–0010,
0001–0011, 0100–0101, and 1010–1011. Kernel-reduced decide checks its
stability, every-direction witnesses, and two distinct reserve edges in the
first direction.
Induction appends one Boolean coordinate. Over one reserve edge's two tails, it chooses the new direction; elsewhere it keeps the old selector. Flip involution proves the complement of that removed orbit is invariant. Nonzero-direction witnesses avoid the orbit, the other distinguished edge survives, and its two layer copies become the next reserves. This proves the uniform construction for every dimension at least four.
For mixed alphabets, project zero to false and every positive symbol to true. At a mixed vertex use its binary image's chosen direction. Along the entire full line the binary image either stays unchanged or becomes its binary partner; therefore the selector and selected full line are constant. Fixing actual off-direction coordinates splits the inverse image of a binary edge into several full lines. The range of selected lines gives existence and unique membership. To exclude an outsider from a larger clique, retain its off-direction disagreement and select a line point that also disagrees in the direction. Each alphabet has two symbols, so such a point exists. The coordinatewise zero/one section gives a witness in every direction.
Falsifier
What would falsify this route
A vertex omitted from the family, two distinct family members containing the same vertex, a member omitting a symbol from its varying coordinate, an extendible member clique, or a missing coordinate direction would violate the statement. The theorem proves each of these clauses for all permitted parameters. A different source meaning of genuine multipartiteness or maximality would invalidate the source transfer, even with a correct Lean proof; that transfer remains subject to independent review.
Evidence
Evidence to collect
The scoped canonical build of
D5.S3.Combinatorics.Hamming.HammingMultipartitePartition succeeds at
project base e4365d8215799efe75a9e5bf74091c39d0b5d20f, Lean 4.33.0, and
Mathlib db584cd6d46c92f209a44c0f1c829460d327499d. A current exact
application with the dependent product written out also compiles. The
source-bound scoped inspector and stock axiom check agree: the result's
recursive proof-and-type closure is exactly propext, Classical.choice,
and Quot.sound. The public graph and line definitions have no axioms.
There is no sorry, new axiom, native_decide, or Lean.ofReduceBool.
The only public declarations are Vertex, graph, line, and result.
The result's statement identity is
sha256:c59baac6ce3f9997e6571f28a464a453ef48319f008315115e8290dc4854cd1c.
All seed, selector, induction, projection, and clique helpers are private
and consumed by the result. A finite seed alone is not the delivered result.
Triage
Scope assessment
theorem, with the complete universal graph assertion kernel proved.
The public definitions have proof_shape: definition; the result has
proof_shape: content. Its live construction uses the all-dimension selector
induction and full-line lifting. The private finite seed is normalization
and carries no separate contribution claim. admission_basis: open-problem-resolution refers to issue #13221. utility: kind=none applies
to the universal theorem; it is not an independently delivered positive
finite instance. Registration is paused; no Reg sources or registration
issues are part of this contribution.
The mechanism shows that binary selector stability is sufficient to lift
partitions to independently varying alphabets. It supplies existence, not
optimal line counts or prescribed numbers of lines in each direction.
Lower-dimensional impossibility and the source's quantum interpretation
are separate questions. The module is frozen in the current worktree through
the canonical uncovered deposit route; generated Blueprint validation,
independent review, required CI, and the remote PR lifecycle remain separate
gates. The current whole-repository report is
sha256:1139946c59d808af62edf7cd3fddcf2f2eec97fee78884a0f92aa0942a8d966f.
No LeanEval score mapping, acceptance, priority, or campaign completion is
claimed.
What the settlement shows
The proved mechanism is the binary selector induction together with its
stability under coordinate flips. Projecting each positive alphabet symbol to
one lifts that selector to every mixed-alphabet full line, and fixing the
off-direction coordinates gives unique membership and maximality. The result
therefore covers every dimension n ≥ 4 and every independent alphabet-size
function with d(i) ≥ 2; it does not establish optimal line counts or a
prescribed number of lines in any direction. The binary case remains credited
to Erde as recorded above, while the source's circuit, unitary and
quasi-process interpretations remain outside this formal theorem.
ASSUMED-UNVERIFIED
Unverified assumptions
Source transfer to Conjecture V.1 awaits the controller's independent review. The accessible literature check is not exhaustive; unindexed or inaccessible resolutions remain unexcluded. The source's identification with MECC circuits and quasi-process functions is not formalized. The result is a development contribution at Lean 4.33.0, not an official Lean 4.35 benchmark certificate. Repository admission and publication require their own evidence.