Keyboard shortcuts

Press or to navigate between chapters

Press ? to show this help

Press Esc to hide this help

Finite Repair Termination

Abstract

Strict refinements of a finite equivalence partition terminate within the available class-count gap, while an infinite carrier admits an infinite refinement tower.

Theorem 1.1 (Finite repair termination and the infinite boundary).

Proof. Machine-checked in Lean as D5/S1/FixedPoints/FiniteRepairTermination.finite_repair_termination_and_infinite_tower (✓ std3). ∎

Source. Repository-derived.

Commentary.

Let X be finite and let P_n be a sequence of partitions of all of X. The order on Mathlib finpartitions is the refinement order, so P_(n+1) <= P_n says that every repair only splits equivalence classes and never merges them.

The sequence is eventually constant. Moreover, the set of indices at which P_(n+1) differs from P_n has cardinality at most |X| minus the number of classes in P_0. This is the claimed sharp budget from the initial concept-class count to the discrete partition.

Pinned Mathlib supplies Finpartition.card_mono, Finpartition.card_parts_le_card, and WellFoundedLT.antitone_chain_condition. The local bookkeeping proves that a proper refinement has strictly more parts and injects strict change indices into the natural-number interval between the initial class count and |X|.

Finiteness is essential: on the natural numbers, the kernel of x |-> min x n has singleton classes below n and one tail class. Increasing n strictly refines this relation forever, giving the source’s infinite refinement tower. Inverse-limit construction and audit of a concrete realization are explicitly outside this mathematical declaration and remain implementation obligations.

References

  • Truth anchor: D5/S1/FixedPoints/FiniteRepairTermination.finite_repair_termination_and_infinite_tower