Continuity Set¶

A Borel set whose boundary has measure zero for a specified measure.

Version: v1 · 2026-09-28 · History
Domain-specific #: 8692
Domain group: Formal Sciences
Origin domain: Mathematics
Subdomains: Measure Theory, Probability Theory → Mathematics
Aliases: Measure continuity set, Μ-continuity set

Core Idea¶

A continuity set is a relation between a Borel subset, a topology, and a measure. Take the set's topological boundary—the points where membership can change under arbitrarily small neighborhood shifts—and ask whether that boundary has zero mass under the stated measure. For a signed measure, use its total variation so positive and negative contributions cannot cancel a charged boundary. The same subset may qualify under one measure and fail under another.

The condition matters in probability because weak convergence controls integrals of bounded continuous functions, while the indicator of a set jumps at its boundary. The Portmanteau theorem recovers convergence of set probabilities when the limiting measure assigns that jump set zero mass. This is not the 'continuity set' of a function, which names points where that function is continuous. Nor does a boundary-null set automatically provide any convergence without a weakly convergent measure sequence.

How would you explain it like I'm…

No Sand on the Line

Draw a chalk circle on the playground and sprinkle sand everywhere. Now look only at the chalk line itself: is any sand sitting right on the line? If no sand is on the line, the circle is a 'continuity set' for that sand. With a different sprinkle of sand, the very same circle might not be.

Clean-Edge Shapes

A continuity set is a region checked against a particular way of spreading 'weight' (like sand or chance) over a space. You look at the region's edge, the places where a tiny step could take you from inside to outside. If the edge carries no weight at all, the region counts as a continuity set for that spreading. The same region can pass for one spreading and fail for another. This matters when a series of spreadings settles toward a final one: the amount inside the region only settles properly if the final spreading puts nothing on the edge.

Measure-Zero Boundary Set

A continuity set is a set whose boundary gets zero weight under a chosen measure, the rule that says how much 'mass' each region has. The boundary is the set of points where every tiny neighborhood contains both members and non-members. Whether a set qualifies depends on the measure, not just the set: a disk is a continuity set for a smooth spread of mass, but not for a measure that puts a lump on its rim. This matters in probability because when a sequence of distributions approaches a limit in the weak sense, averages of smooth quantities converge, but the 'in or out' switch of a set jumps at its edge. If the limit puts no mass on that edge, the set's probabilities converge too. Do not confuse this with the set of points where a function is continuous.

A continuity set is defined relative to three things: a Borel set A, a topology (which fixes the boundary, closure minus interior), and a measure mu. A is a mu-continuity set when mu(boundary of A) = 0; for a signed measure one uses the total variation |mu|, so positive and negative charge on the boundary cannot cancel into a false zero. The notion is not intrinsic to A: the same set can be a continuity set for one measure and fail for another. Its main use is in weak convergence. Weak convergence of mu_n to mu only guarantees convergence of integrals of bounded continuous functions, and the indicator of A is discontinuous exactly on the boundary of A. The Portmanteau theorem says that if mu_n converges weakly to mu and A is a mu-continuity set, then mu_n(A) converges to mu(A). Being boundary-null gives nothing by itself; it only converts an existing weak convergence into convergence of set probabilities. The term is distinct from the 'set of continuity points' of a function.

Structural Signature¶

Sig role-phrases:

topological carrier — Supplies a space whose closure and interior determine a set boundary. It is constitutive. Counterfactual: A bare measurable space without topology has no specified ∂B.
Borel candidate set — Supplies the measurable subset tested against its boundary. It is constitutive. Counterfactual: An arbitrary non-Borel subset is outside this stated class.
reference measure — Fixes which sets have zero mass, with total variation used when signed. It is constitutive. Counterfactual: The same boundary may be null for one measure but not another.
boundary-null condition — Requires μ(∂B)=0 or |μ|(∂B)=0. It is constitutive. Counterfactual: A set with positive mass on its boundary fails even if its interior has large mass.
convergence use — Allows set-probability continuity under suitable weak convergence. It is boundary. Counterfactual: Portmanteau's conclusion needs weak convergence and the limiting measure, not merely any sequence.

What It Is Not¶

Borel set alone. Measurability does not force its boundary to be null.
Null set. The set's own mass need not vanish; only its boundary must.
Continuity points of a function. That usage concerns a function's local behavior, not boundary measure.
Uniformly stable event. Boundary-nullity is relative to a specified measure and topology.
Closest near-miss. An interval may look harmless but fails when an endpoint carries an atom of the reference measure; under an atomless measure the same boundary can be null and the interval qualifies.

Scope of Application¶

Weak-convergence proofs. Use limiting-measure continuity sets for event-probability convergence.
Atomic-measure checks. See whether endpoint atoms obstruct a proposed interval event.
Signed-measure reasoning. Apply total variation rather than cancellation-prone signed mass.
Measurable-set classification. Record topology and measure whenever calling a Borel set continuous.

Clarity¶

State B, the topology, the reference measure, and its boundary mass. [0,½] is positive under atomless Lebesgue probability on [0,1], but fails for a measure assigning mass to the endpoint ½. A null set can still have a large boundary and fail. For signed measures use |μ| on the boundary. This criterion concerns a set relative to a measure, not ordinary continuity of a function.

Manages Complexity¶

The two-word name compresses Borel membership and a boundary-null test into a portable event condition. That lets probability proofs replace difficult indicator limits with a single boundary check. The compression hides dependence on topology, the limiting measure, and atoms; omitting any of these can reverse the classification.

Abstract Reasoning¶

Specify the topological space and Borel σ-algebra.
Choose the candidate Borel set and compute its topological boundary.
Specify the measure, or total variation for a signed measure.
Test whether the boundary has zero mass under that measure.
Invoke Portmanteau's set-probability conclusion only with a suitable weakly converging sequence and the limiting measure.

Knowledge Transfer¶

The boundary-null test transfers across Borel measures on topological spaces, but qualification must be recalculated when topology or measure changes. One interval's Lebesgue status does not carry to an atomic law, and a Portmanteau conclusion cannot be moved to sets with charged boundaries.

Examples¶

Canonical¶

For Lebesgue probability measure on [0,1], the interval [0,½] is a continuity set: its relative boundary is the singleton {½}, which has zero measure. Under a probability measure with positive mass at ½, the same interval is no longer a continuity set of that measure.

Mapped back: topological carrier → [0,1] with its usual relative topology; Borel candidate set → [0,½]; reference measure → Lebesgue probability or one with an atom; boundary-null condition → zero for the first, positive for the second; convergence use → limiting measure must be the one tested.

Applied / In Practice¶

In the Portmanteau theorem, weak convergence of probability measures implies convergence of their assigned values on every continuity set of the limit measure. The University of Toronto notes use that boundary-null criterion to connect function-integral convergence with event probabilities; it does not promise convergence for sets whose boundary carries limiting mass.

Mapped back: topological carrier → metric probability space; Borel candidate set → event set in theorem; reference measure → limiting probability measure; boundary-null condition → zero boundary mass hypothesis; convergence use → event-probability convergence.

Structural Tensions¶

T1 — Set Membership versus Measure Dependence. The same geometric set can qualify for one measure and fail for another with an atom on its boundary.

Diagnostic: Which measure's boundary mass is being tested?

T2 — Weak Convergence versus Discontinuous Indicators. Event indicators are discontinuous at boundaries, so boundary-nullity licenses their limit behavior.

Diagnostic: Does the limiting measure charge the discontinuity locus?

Structural–Framed Character¶

The approved DAG parent is Borel Set: a continuity set is Borel in a declared topology and additionally has boundary measure zero for a specified measure. For signed measures, total variation supplies the condition.

Evaluative weight: Low; “continuity” is a measure-relative property, not general smoothness. Human-practice-bound: Low formally, though topology and measure are selected. Institutional origin: Measure theory defines the term; proofs determine qualification. Vocabulary travels: The test applies across appropriate spaces, but changing measure can change status. Import versus recognize: Recognize the set by μ(∂B)=0 or its signed variant; transferring Lebesgue status to an atomic law imports an invalid conclusion.

Its character: A measure-relative Borel-set subtype with portable boundary-null testing and explicit reference measure.

Structural Core vs. Domain Accent¶

Skeletal core. A Borel set's boundary is negligible under a specified measure.

Domain-bound accent. Topology determines the boundary, measure determines nullity, and weak-convergence use requires further premises.

Why not prime. Borel sets are broader; a charged boundary or unspecified measure defeats this child property.

This entry is a kind of Borel Set.

Strict parent — Borel set. A continuity set first belongs to the topology-generated Borel σ-algebra, then satisfies the additional zero-boundary-mass test.
Related — null set. The whole set need not have zero measure; only its boundary does.
Related — continuous function. A function's continuity points are a different use of the words.

Relationships to Other Abstractions¶

Current abstraction Continuity Set Domain-specific

Parents (1) — more general patterns this builds on

Continuity Set is a kind of Borel Set Domain-specific

Every measure-continuity set is a Borel set with the additional condition that its boundary is null for a specified measure.

Hierarchy path (1) — routes to 1 parentless root

Continuity Set → Borel Set → Set and Membership

Neighborhood in Abstraction Space¶

Continuity Set sits in a crowded region of the domain-specific corpus (33^rd percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Formal Systems & Discrete Structures (18 abstractions)

Nearest neighbors

Orthocompact Space — 0.91
Mathematical Space — 0.89
Perfect measure — 0.88
Urysohn's lemma — 0.88
Radon Measure — 0.87

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With¶

Null set. Tell: Is the boundary null, rather than the entire set?
Borel set. Tell: Has the boundary-mass condition also been checked?
Function's continuity set. Tell: Is the subject a measure-relative set or points of continuity of a function?
Weak-convergence event. Tell: Is boundary-nullity evaluated under the limiting measure?

References¶

University of Toronto, Notes on Convergence of Probability Measures, Portmanteau continuity-set theorem: https://www.math.toronto.edu/mnica/Billingsley.pdf
Frozen Wikipedia discovery revision: https://en.wikipedia.org/wiki/Continuity_set (revision 1275458824).