Borel reducibility compares classification problems by asking whether one can assign invariants for one problem using the equivalence classes of another. The comparison is sensitive to the definability of the assignment: the coding map must be Borel measurable and must preserve and reflect equivalence. It is a standard organizing principle in invariant descriptive set theory, where the objects being classified often live in Polish spaces. Marker's course notes introduce this perspective and the underlying descriptive set theory.
An upper bound for a family of equivalence relations is a single relation receiving a Borel reduction from every member of that family. The distinction between countable unions of compact sets and countable unions of closed sets is essential here. Rosendal's work established universal behavior for the more restrictive compact-union setting. The present target asks how far the closed-union class can be bounded when the target may have arbitrary Borel complexity. Rosendal, 2005.
The relevant chronology is:
A Polish space is a topological space admitting a complete compatible metric and a countable dense subset. An set is a countable union of closed sets. A binary relation has an graph when its set of related pairs is in the product topology. A Borel equivalence relation is an equivalence relation whose graph belongs to the Borel sigma algebra generated by that product topology.
For equivalence relations on and on , write if there is a Borel measurable map satisfying
The domain and target sigma algebras are the Borel sigma algebras of the specified topologies. All Polish source spaces are admitted, including spaces that are neither compact nor sigma-compact.
For a subgroup of a group , the coset relation is exactly when , with classes . The subgroup need not be normal. For a real linear subspace, the additive convention is exactly when .
The main target is precisely Theorem 1 of the supplied note:
In Lean the assertion is stated for every Polish target . The target graph may belong to any Borel level.
Two new supporting milestones retain the source numbering. Lemma 2 is the category-transfer statement: for standard-Borel parameters and countably many jointly Borel families of continuous group homomorphisms, a common Borel upper bound for the coset relations outside a generated-subgroup membership set forces that membership set to be Borel. Corollary 4 states that the counterexamples can be coset relations of real linear subspaces in one fixed real . The earlier -target obstruction is a separate, already-published reference milestone.
The main statement rules out a single Borel classifier for the whole class, even when no fixed bound is imposed on the complexity of that classifier's graph. This is stronger than requiring the proposed upper bound itself to have an graph. The supplied note also draws consequences for higher levels of the Borel hierarchy; those consequences are outside the present goal.
The formalization would connect topological groups, descriptive set theory and real Banach-space coset relations in a reusable Lean development. The earlier reference result already has an accepted machine-checked proof. The new theorem and its new milestones are open formal statements: successful compilation of their interfaces checks their well-formedness and does not supply their proofs.
Increasing the permitted complexity of the target changes the problem materially. The earlier theorem's hypothesis on the target graph no longer applies. For the category-transfer milestone, the individual reductions are only assumed to exist; a Borel selection of them is not part of the hypotheses. For the Banach-space milestone, an algebraic linear image need not be closed, and the source imposes injectivity without assuming a closed range or a bounded inverse. These distinctions must survive the formalization.
The mission reuses the public definitions of graphs, Borel maps and relation reductions from the earlier mission. New definitions specify Borel graphs explicitly using borel (X × X), upper bounds over all Polish sources represented in Lean's ordinary type universe, algebraically generated subgroups, and native real lp with exponent . Targets are universe-polymorphic. Empty targets and empty Polish source spaces are included.
The category-transfer parameter space carries its given standard Borel structure. Group factors carry their own Polish group topologies; joint measurability uses the product of the parameter sigma algebra and each group's Borel sigma algebra. Generated subgroups are algebraic closures under group operations. No normality, compactness, sigma-compactness or measurable selection hypothesis is added.
The definitions contain no proof placeholders. The three new theorem interfaces deliberately use the platform's open-target placeholder. Contributions proving these targets or developing their measurable-category and Banach-space prerequisites are welcome. Baire-category coding, analytic separation, countable closed covers and the tensor-image analysis are reusable infrastructure needed by a complete development.
UniversalBorelEquivalence-20261101.tex; no public manuscript URL was supplied.theorem NoBorelUpperBoundFSigma.theorem1
{Y : Type u} [TopologicalSpace Y] [PolishSpace Y] :
¬ ∃ F : Setoid Y,
NoBorelUpperBoundFSigma.IsBorelRelation F.r ∧
NoBorelUpperBoundFSigma.IsFSigmaUpperBound F.r := by sorry
For every Polish space Y, there is no equivalence relation F on Y such that its graph is Borel and every equivalence relation E with an F-sigma graph on every Polish space X Borel reduces to F. Explicitly, there is no such F for which every pair admits a Borel map satisfying for all source pairs. The target graph may have arbitrary Borel complexity; no F-sigma hypothesis is imposed on it. The source spaces are not required to be compact or sigma-compact.
Formalization note. Polish sources range over the ordinary Lean type universe, containing the standard presentations of all Polish spaces; the target universe is arbitrary. Empty spaces are permitted. Borel measurability and graph Borelness use the specified topologies.
No open leaves. Every sub-goal is proved or awaiting decomposition.