slug: jack-2024-finite-otto-congruence
bibkey: jack2026conjugacy
doi: 10.1007/s00233-026-10614-w
url: https://arxiv.org/abs/2411.16284
triage: theorem
motivation_gids:
  - D5/S3/FiniteGroups/FiniteOttoCongruence.result

Finite Otto congruence quotients are cancellative

Problem

Trevor Jack, Answering Five Open Problems Involving Semigroup Conjugacy, arXiv:2411.16284v1, Section 9, Problem 9.4, p. 14; Semigroup Forum 112 (2026), 450–466, DOI 10.1007/s00233-026-10614-w:

Is there a finite semigroup for which ∼o is a congruence and S/∼o is not cancelative?

The answer is negative. Section 2, p. 2 defines the relation by:

∀a, b ∈ S (a ∼o b ⇔ ag = gb and bh = ha for some g, h ∈ S).

Both witnesses lie in $S$; no identity is adjoined.

Statement echo

Motivation

The source’s Problem Four constructs an infinite semigroup with an Otto congruence and a noncancellative quotient. Problem 9.4 asks whether finiteness can be retained. Finite idempotent powers prevent that failure.

Gap

The external named question is preregistered in #17635. The source explicitly asks the question. No answer was found in the inspected literature and repository scope described below; completeness of that screen is ASSUMED-UNVERIFIED.

Route

The module is D5/S3/FiniteGroups/FiniteOttoCongruence.lean, with one public settling theorem, D5.S3.FiniteGroups.FiniteOttoCongruence.result. The four public definitions are the literal conventions of #17635. Its private helpers are consumed by the cancellation proof. Their classification and that of result is proof_shape: bind-only; the module’s admission basis is open-problem-resolution (#17635; Proved). Its utility is none: the proof is structural over arbitrary finite semigroups and contains no enumeration or fixed numerical certificate.

The frozen D5.S0.Certificates.Algebra.BergmanQuestion15Refutation.positivePower supplies the positive product $a^{k+1}$ at index $k$. Mathlib’s exists_idempotent_in_compact_subsemigroup supplies an idempotent in the finite subsemigroup of positive powers, using the discrete topology. SemiconjBy.mul_left composes the witness equations in otto_trans.

Falsifier

A finite associative multiplication table with multiplication-compatible Otto conjugacy and a failure of either representative cancellation law would contradict result under its stated definitions. This criterion uses the original two-witness relation, without replacing it by a certificate or by a relation on a monoid with an adjoined identity.

Evidence

The kernel-checked settling declaration is D5.S3.FiniteGroups.FiniteOttoCongruence.result. Its axiom closure is contained in {propext, Classical.choice, Quot.sound}. The Blueprint displays the full universal implication and both cancellation laws.

The only public theorem is an escape-audit target. Its four-slot registration is unfinished: #17715 names the missing source reconstruction, exact occurrence bridge, whole-family variation/sensitivity and actual observational-dependence evidence. No successful Reg mirror or declared_validated status is claimed.

Experiment

The independent finite enumeration is in trureturing-experiments at commit 1ea0f434ab3c6f8b8803a90a4b0b73c27fe24977. The program is docs/reports/jack-2024-finite-otto-congruence/check.py; run python3 check.py in that directory. Exit code: 0; result: ALL_OK. SHA-256: 62c4082d7c5952937d8663b016643759922a7715cbd8fbb16e9542023f62c5df. The tested scope is every labelled semigroup of orders 1, 2, 3 and 4:

These counts are computed, not Lean enumeration theorems. The structural cancellation theorem applies to every finite order independently of this experiment.

Triage

What the settlement shows

• Mechanism — kernel-checked. exists_idempotent_power gives an index $k\ge 2$ with $a^{k+1}$ idempotent. Writing $n=k+1$, otto_stable_power gives $a{\sim }_o{a}^{n+1}$, and otto_idempotents relates every two idempotents. Under multiplication compatibility, idempotent_class_identity makes their common class a two-sided identity, inverse_class gives $[a^{n-1}]$ as an inverse representative, and otto_mul_swap gives $ab{\sim }_{o}ba$. These private declarations are on result’s proof path; their statements express the class laws on representatives.
• Strength — kernel-checked cancellation. quotient_cancelative and result prove both left and right cancellation for every finite semigroup, including the empty case vacuously. Finite abelian group — proved in prose for nonempty $S$. The quotient is finite as the image of $S$. Choose an element to obtain an idempotent power. The preceding identity, inverse and swap laws descend to its congruence classes, giving a finite abelian group. The module does not construct a Lean CommGroup instance. Converse — proved in prose. In an abelian group the equation $ag=gb$ becomes $ag=bg$ and cancellation gives $a=b$; equality conversely has witnesses $a,a$. Thus Otto conjugacy is equality, and every finite abelian group occurs as its own quotient. No nonemptiness is claimed for the empty carrier.
• Consequence — proved in prose from result. The noncancellative quotient of the source’s infinite Problem Four example cannot be realized by a finite semigroup. The source’s infinite construction remains valid; no other source theorem is changed by this finite obstruction. Finiteness is essential: the source’s infinite example violates the conclusion.
• Classification — open. Characterize the finite semigroups whose Otto conjugacy is multiplication compatible. The experiment computes that 3468 of the 3492 labelled order-four semigroups have this property. It is not a classification at arbitrary order.
• Experiment — computed. The pinned experiment in Evidence exhausts labelled semigroups of orders 1–4 and checks the quotient’s commutativity, identity, inverses and cancellation whenever Otto conjugacy is a congruence. The command, exit code, script SHA-256 and exact counts are stated there. No enumeration claim beyond order 4 is made.
• Literature — open completeness (ASSUMED-UNVERIFIED); source reading. The arXiv v1 source was read. The source and the inspected screen do not give an answer to Problem 9.4. The Semigroup Forum version of record, Liu–Wang, Communications in Algebra 53 (2025), 375–389, and Clifford–Cummings–Teymouri (2011) remain unread in full. The probe’s reported screen also inspected arXiv:2301.04252v3, 2605.00041, 2101.00324v2, 1503.00915v2 and 2503.02148; no answer was identified in that scope. This is not an exhaustive novelty certificate.

ASSUMED-UNVERIFIED

Literature completeness and external priority remain unverified because the journal version’s final question section and the other full texts named in Triage have not been read. The source fidelity and kernel-checked negative answer do not establish worldwide priority. The four-slot escape audit is unfinished under #17715. The general classification in Triage remains open.