KC Topology: Compact Subset Separation

• KC is defined as the property in which every compact subset of a space is closed, lying between Hausdorff and US separation axioms.
• The Extended Real Line with Reentry (ERI) serves as a concrete counterexample, being compact, T1, and US while failing the KC property.
• Additional conditions such as local path connectivity or unicoherence can force a KC space to be Hausdorff, highlighting the delicate balance in separation theory.

Searching arXiv for the topology papers on the KC property and its relation to US. {"query":"(Lattenero, 3 Mar 2026) KC topology US separation Wilansky hierarchy", "max_results": 5} Searching for foundational and related KC-space work. {"query":"(Fabel, 2011) KC-spaces low dimensional topology compact subsets closed", "max_results": 5} Searching for the quantum-information use of “KC” to avoid conflating distinct meanings. {"query":"(He et al., 2022) Klein's condition KC quantum entropy cone", "max_results": 5} In general topology, KC denotes the separation property that every compact subset is closed. It occupies a precise intermediate position in separation theory: Hausdorff spaces are KC, and KC spaces in turn satisfy uniqueness of sequential limits, but neither converse holds in general. Recent work has sharpened this picture by giving an explicit compact, path-connected, uniquely sequential non-KC space—the Extended Real Line with Reentry (ERI)—while earlier work established several local, dimensional, and homotopical hypotheses under which KC spaces are forced to be Hausdorff and constructed low-dimensional compact connected non-Hausdorff KC examples (Lattenero, 3 Mar 2026, Fabel, 2011).

1. Definition and logical position

A space $X$ is KC if

$$\forall K\subseteq X\,(K\text{ compact}\Rightarrow K\text{ closed}).$$

The property is weaker than Hausdorffness but stronger than several other minimal separation conditions. In the Wilansky hierarchy recalled in recent work, one has

$$T_2 \Longrightarrow \mathrm{KC} \Longrightarrow \mathrm{US} \Longrightarrow T_1,$$

where US means uniqueness of sequential limits: $$\forall(x_n)\,\bigl[(x_n\to x)\land(x_n\to y)\Rightarrow x=y\bigr].$$ All three implications are strict (Lattenero, 3 Mar 2026).

The KC axiom is also naturally related to weak Hausdorffness. In a KC-space, the continuous image of a compact Hausdorff space is compact and therefore closed, so every KC-space is weakly Hausdorff. This places KC strictly between Hausdorff and weakly Hausdorff separation in the sense emphasized in the low-dimensional theory of KC-spaces (Fabel, 2011).

Conceptually, KC isolates one specific Hausdorff consequence—closure of compacta—without demanding full point-separation by disjoint neighborhoods. In compact Hausdorff spaces KC is automatic; in general $T_1$ spaces it is strictly stronger than US and strictly weaker than Hausdorff.

2. ERI and the explicit separation of US from KC

The most explicit modern counterexample separating US from KC is the Extended Real Line with Reentry (ERI). Start with the extended real line $\overline{\mathbb R}=[-\infty,+\infty]$ with its usual order topology, collapse the three points $\{-\infty,0,+\infty\}$ to a single point $\ast$, and write the quotient map as

$$q:\overline{\mathbb R}\to X_0,\qquad q(-\infty)=q(0)=q(+\infty)=\ast.$$

ERI equips $X_0$ with the topology

$$\forall K\subseteq X\,(K\text{ compact}\Rightarrow K\text{ closed}).$$0

This density condition at the collapsed point is the decisive mechanism: it prevents Hausdorff separation from $$\forall K\subseteq X\,(K\text{ compact}\Rightarrow K\text{ closed}).$$1, destroys first-countability at $$\forall K\subseteq X\,(K\text{ compact}\Rightarrow K\text{ closed}).$$2, and forces compact subsets with interior to fail closedness, while preserving $$\forall K\subseteq X\,(K\text{ compact}\Rightarrow K\text{ closed}).$$3 and enabling unique sequential limits (Lattenero, 3 Mar 2026).

Neighborhoods of $$\forall K\subseteq X\,(K\text{ compact}\Rightarrow K\text{ closed}).$$4 admit a canonical description. If $$\forall K\subseteq X\,(K\text{ compact}\Rightarrow K\text{ closed}).$$5 is closed and nowhere dense, then

$$\forall K\subseteq X\,(K\text{ compact}\Rightarrow K\text{ closed}).$$6

is an ERI-neighborhood of $$\forall K\subseteq X\,(K\text{ compact}\Rightarrow K\text{ closed}).$$7, and every ERI-neighborhood of $$\forall K\subseteq X\,(K\text{ compact}\Rightarrow K\text{ closed}).$$8 is of this form. From this follows the key intersection principle: if $$\forall K\subseteq X\,(K\text{ compact}\Rightarrow K\text{ closed}).$$9 is ERI-open and $$T_2 \Longrightarrow \mathrm{KC} \Longrightarrow \mathrm{US} \Longrightarrow T_1,$$0 is ERI-open, then $$T_2 \Longrightarrow \mathrm{KC} \Longrightarrow \mathrm{US} \Longrightarrow T_1,$$1. Hence every neighborhood of $$T_2 \Longrightarrow \mathrm{KC} \Longrightarrow \mathrm{US} \Longrightarrow T_1,$$2 is dense, so $$T_2 \Longrightarrow \mathrm{KC} \Longrightarrow \mathrm{US} \Longrightarrow T_1,$$3 cannot be Hausdorff-separated from any other point.

The space nonetheless has a sharp sequential theory. For a sequence $$T_2 \Longrightarrow \mathrm{KC} \Longrightarrow \mathrm{US} \Longrightarrow T_1,$$4 in $$T_2 \Longrightarrow \mathrm{KC} \Longrightarrow \mathrm{US} \Longrightarrow T_1,$$5,

$$T_2 \Longrightarrow \mathrm{KC} \Longrightarrow \mathrm{US} \Longrightarrow T_1,$$6

Equivalently, all accumulation points of $$T_2 \Longrightarrow \mathrm{KC} \Longrightarrow \mathrm{US} \Longrightarrow T_1,$$7 in $$T_2 \Longrightarrow \mathrm{KC} \Longrightarrow \mathrm{US} \Longrightarrow T_1,$$8 lie in $$T_2 \Longrightarrow \mathrm{KC} \Longrightarrow \mathrm{US} \Longrightarrow T_1,$$9, or, again equivalently, $$\forall(x_n)\,\bigl[(x_n\to x)\land(x_n\to y)\Rightarrow x=y\bigr].$$0 eventually exits every compact subset of $$\forall(x_n)\,\bigl[(x_n\to x)\land(x_n\to y)\Rightarrow x=y\bigr].$$1. This criterion yields the theorem that ERI is US: convergent sequences have unique limits even though the space is non-Hausdorff.

At the same time ERI is not KC. If $$\forall(x_n)\,\bigl[(x_n\to x)\land(x_n\to y)\Rightarrow x=y\bigr].$$2, then

$$\forall(x_n)\,\bigl[(x_n\to x)\land(x_n\to y)\Rightarrow x=y\bigr].$$3

is compact but not closed. The complement $$\forall(x_n)\,\bigl[(x_n\to x)\land(x_n\to y)\Rightarrow x=y\bigr].$$4 contains $$\forall(x_n)\,\bigl[(x_n\to x)\land(x_n\to y)\Rightarrow x=y\bigr].$$5, yet its preimage $$\forall(x_n)\,\bigl[(x_n\to x)\land(x_n\to y)\Rightarrow x=y\bigr].$$6 is open but not dense, so it is not ERI-open. More generally, $$\forall(x_n)\,\bigl[(x_n\to x)\land(x_n\to y)\Rightarrow x=y\bigr].$$7 is compact non-closed whenever $$\forall(x_n)\,\bigl[(x_n\to x)\land(x_n\to y)\Rightarrow x=y\bigr].$$8 is closed with nonempty interior. ERI is therefore compact, $$\forall(x_n)\,\bigl[(x_n\to x)\land(x_n\to y)\Rightarrow x=y\bigr].$$9, path-connected, US, and not KC. The same construction also shows that every continuous real-valued function on ERI is constant.

3. When KC collapses to Hausdorff

The older structure theory of KC-spaces identifies several strong hypotheses under which the KC axiom already forces Hausdorffness. A basic lemma states that if $T_1$0 is KC and $T_1$1 is continuous from a compact Hausdorff space $T_1$2, then $T_1$3 is Hausdorff as a subspace. In particular, each path-connected subspace of a KC-space is arcwise connected (Fabel, 2011).

Two general promotion theorems are especially important. First, if a KC-space is locally path connected and contains no simple closed curve, then it is Hausdorff. As a corollary, a locally path connected, simply connected, 1-dimensional KC-space is Hausdorff; if it is also compact and connected, then it is a dendrite. Second, if a KC-space is generalized hereditarily unicoherent and locally connected by continua, then it is Hausdorff; if it is also compact and connected, then it is a dendroid (Fabel, 2011).

These theorems clarify why non-Hausdorff KC examples must evade familiar low-dimensional regularity assumptions. The same paper constructs several compact, connected, non-Hausdorff KC-spaces of dimensions $T_1$4 and $T_1$5, including Alexandroff compactifications of first-countable Hausdorff non-locally compact spaces. Among the phenomena exhibited is the failure of a classical Hausdorff intersection theorem: a nested intersection of compact connected subsets can be disconnected. This failure is realized in multiple examples by intersections collapsing to a two-point set.

4. Sequential refinements and the role of first-countability

ERI also refines the position of KC in more delicate hierarchies. In Clontz’s hierarchy it is SC—every convergent sequence together with its limit is closed—but not weakly Hausdorff. The latter is witnessed by a continuous map $T_1$6 with non-closed image $T_1$7. Thus the implications

$T_1$8

are strict in the relevant range (Lattenero, 3 Mar 2026).

A central mechanism behind these separations is the failure of first-countability. The ERI paper isolates a precise barrier: a first-countable space is Hausdorff if and only if it is US. Hence any non-Hausdorff US space must fail first-countability somewhere; in ERI the failure occurs exactly at the collapsed point $T_1$9. The nonexistence of a countable local base is proved via a Baire-category argument using the dense-open preimages of neighborhoods of $\overline{\mathbb R}=[-\infty,+\infty]$0.

The space is also not sequential, and the contrast between sequences and nets is explicit. Although sequences have unique limits, the paper constructs nets converging simultaneously to $\overline{\mathbb R}=[-\infty,+\infty]$1 and to $\overline{\mathbb R}=[-\infty,+\infty]$2 for arbitrary $\overline{\mathbb R}=[-\infty,+\infty]$3. This is consistent with the general non-Hausdorff, non-first-countable setting: sequential uniqueness does not extend to uniqueness of limits for arbitrary nets.

5. Generalized reentry constructions

The ERI construction extends far beyond $\overline{\mathbb R}=[-\infty,+\infty]$4. Let $\overline{\mathbb R}=[-\infty,+\infty]$5 be a compact Hausdorff space without isolated points, let $\overline{\mathbb R}=[-\infty,+\infty]$6 be a finite nonempty subset, collapse $\overline{\mathbb R}=[-\infty,+\infty]$7 to a point $\overline{\mathbb R}=[-\infty,+\infty]$8, and define a topology by requiring preimages of neighborhoods of $\overline{\mathbb R}=[-\infty,+\infty]$9 to be open and dense in $\{-\infty,0,+\infty\}$0. The resulting quotient is a well-defined topology, coarser than the ordinary quotient topology, and yields a space that is compact, $\{-\infty,0,+\infty\}$1, not Hausdorff, connected, US, and not KC (Lattenero, 3 Mar 2026).

The convergence criterion persists in this setting: a sequence $\{-\infty,0,+\infty\}$2 converges to $\{-\infty,0,+\infty\}$3 exactly when $\{-\infty,0,+\infty\}$4 has no accumulation point in $\{-\infty,0,+\infty\}$5. The non-KC property follows from the existence of compact subsets $\{-\infty,0,+\infty\}$6 with nonempty interior in $\{-\infty,0,+\infty\}$7 whose complements fail the density requirement at $\{-\infty,0,+\infty\}$8. The construction therefore identifies a broad mechanism, not an isolated example.

What does not automatically generalize is path-connectedness. ERI is path-connected because $\{-\infty,0,+\infty\}$9 admits linear paths compatible with the quotient map; an arbitrary compact Hausdorff $\ast$0 without isolated points need not supply analogous paths. A plausible implication is that the reentry mechanism is fundamentally about dense-neighborhood quotients rather than about one-dimensional order structure.

6. Significance and terminological scope

Within general topology, KC now has a sharply delineated role. It is a separation axiom strictly weaker than Hausdorffness but strictly stronger than uniqueness of sequential limits. The ERI construction gives an explicit compact, path-connected witness that US does not imply KC, while the older low-dimensional theory explains why many apparently mild additional hypotheses force KC back to Hausdorffness. The open question raised for ERI’s place in the $\ast$1-Hausdorff region of Clontz’s hierarchy shows that this classification remains unfinished (Lattenero, 3 Mar 2026).

The abbreviation KC is not unique to topology. In quantum information theory, for example, KC can denote Klein’s condition, a closure property for vanishing mutual informations in the study of patterns of marginal independence. There it is formulated as downward closure in a poset of mutual-information instances and is strictly weaker than strong subadditivity for four or more parties, although equivalence is conjectured for 1-dimensional PMIs (He et al., 2022). This terminological overlap is substantial enough that precise domain specification is often required.