De Groot Self-Duality in Topology

• De Groot self-duality is the involutive operation that interchanges open sets with the complements of compact saturated sets in stably compact spaces.
• It manifests as a formal symmetry between openness and compactness, extending naturally to pointfree locales, ko-spaces, and bi-dcpos.
• Constructive presentations reverse proximity, entailment, and lattice orders, yielding invariant constructions such as patch and powerlocales under duality.

De Groot self-duality is the involutive symmetry classically associated with stably compact spaces whereby a space $X$ is sent to a dual space $X^\partial$ on the same underlying set whose open sets are precisely the complements of compact saturated subsets of $X$; equivalently, the compact saturated subsets of $X^\partial$ are the complements of the open sets of $X$ (Abbadini et al., 24 Jul 2025). In pointfree form, the de Groot dual of a stably compact locale is the locale whose frame consists of the Scott open filters on $\Omega(X)$, and constructive presentations make the duality appear as a formal reversal of proximity and entailment data (Kawai, 2018). Later work places this involution inside a wider symmetry between open-like and compact-like structure, extending beyond the classical stably compact setting to ko-spaces and, on the pointfree side, to bi-dcpos (Abbadini et al., 24 Jul 2025).

1. Classical topological formulation

In the classical setting, de Groot self-duality applies to stably compact spaces. For such a space $X$, the dual $X^\partial$ has the same underlying set as $X$, but its topology is

$$Op(X^\partial)=\{\,X\setminus K \mid K\in KSat(X)\,\},$$

where $X^\partial$0 denotes the compact saturated subsets of $X^\partial$1 (Abbadini et al., 24 Jul 2025). The construction therefore exchanges

$X^\partial$2

The use of compact saturated sets is essential outside the $X^\partial$3 setting. A saturated set is an intersection of open sets, equivalently an upset in the specialization order. With this replacement, the duality interchanges open data with compact-saturated data in a way compatible with non-Hausdorff order-topological structure (Abbadini et al., 24 Jul 2025).

The duality is involutive: $X^\partial$4 Thus de Groot self-duality is not merely a correspondence between two different kinds of topological data; it is an exact reversal operation that returns the original object after a second application (Abbadini et al., 24 Jul 2025).

Pointfree work gives the same duality a locale-theoretic form. For a stably compact locale $X^\partial$5, the de Groot dual is the locale whose frame is the frame of Scott open filters on $X^\partial$6. In that sense, the classical topology of complements of compact saturated sets and the pointfree construction by Scott-open filters are two presentations of the same duality (Kawai, 2018).

2. Symmetry between openness and compact saturatedness

A central interpretation of de Groot self-duality is that it expresses a formal symmetry between open sets and compact saturated sets. This symmetry is visible in paired compactness principles. For $X^\partial$7 and a directed family $X^\partial$8 of opens, compactness takes the form

$X^\partial$9

Its symmetric counterpart, for $X$0 and a codirected family $X$1 of compact saturated sets, is

$X$2

The second condition is exactly well-filteredness. The two statements are obtained from one another by swapping compact with open, $X$3 with $X$4, directed with codirected, and $X$5 with $X$6 (Abbadini et al., 24 Jul 2025).

Within this perspective, de Groot self-duality is the topological realization of a deeper two-sided formal symmetry. The symmetry is described as perfect for Stone spaces and spectral spaces, still perfect for stably compact spaces, and only partially visible at the ordinary topological level for sober spaces and more generally well-filtered $X$7 spaces (Abbadini et al., 24 Jul 2025).

The obstruction outside the compact setting is concrete. If $X$8 is not compact, the complements of compact saturated sets need not form a topology, because $X$9 may fail to be among them. This is why ordinary topological spaces are too rigid a container for a global involution exchanging opens with complements of compact saturated sets across larger classes. The modern extensions of de Groot duality address precisely this failure (Abbadini et al., 24 Jul 2025).

3. Pointfree and constructive presentations

Constructive work on stably compact locales recasts de Groot duality in terms of algebraic presentations. A strong proximity lattice $X^\partial$0 represents a stably compact locale $X^\partial$1, whose frame is the frame $X^\partial$2 of rounded ideals. The de Groot dual is obtained by reversing both lattice order and proximity: $X^\partial$3 At the level of frames, the key identification is

$X^\partial$4

so $X^\partial$5 is the de Groot dual of $X^\partial$6 (Kawai, 2018).

The same duality can be presented syntactically. A strong continuous entailment relation is written $X^\partial$7, and its dual is defined by reversing entailment and proximity: $X^\partial$8 If $X^\partial$9 is generated by a set of axioms $X$0, then the dual entailment is generated by the reversed axioms

$X$1

This makes de Groot duality computable directly from generators and relations (Kawai, 2018).

These two formalisms are equivalent. The constructive account establishes equivalences

$X$2

and corresponding equivalences for the morphism categories, with dualization commuting with those equivalences up to natural isomorphism (Kawai, 2018). A plausible implication is that de Groot self-duality can be recognized either semantically, through the represented locale, or syntactically, by comparing a presentation with its formal dual.

Within this framework, self-duality amounts to fixed-point behavior under the de Groot involution. For a represented locale $X$3, de Groot self-duality is the condition

$X$4

equivalently $X$5, or, in presentation form, $X$6 (Kawai, 2018).

4. Extension to ko-spaces and bi-dcpos

Recent work enlarges the spatial setting from stably compact spaces to ko-spaces. A ko-space is a triple

$X$7

where $X$8 is a poset, $X$9 is a family of “compact-like” upsets, and $\Omega(X)$0 is a family of “open-like” upsets, subject to three groups of axioms: closure of $\Omega(X)$1 under codirected intersections and $\Omega(X)$2 under directed unions, a double compactness principle, and principal generator conditions requiring $\Omega(X)$3 and $\Omega(X)$4 for every $\Omega(X)$5 (Abbadini et al., 24 Jul 2025).

The specialization order is recoverable from either sort: $\Omega(X)$6 The de Groot dual of a ko-space is defined by

$\Omega(X)$7

and remains involutive: $\Omega(X)$8 When $\Omega(X)$9 arises from a stably compact space, this reduces exactly to the classical de Groot duality (Abbadini et al., 24 Jul 2025).

On the pointfree side, the corresponding structures are bi-dcpos, written $X$0, where $X$1 has all codirected meets, $X$2 has all directed joins, and the relation $X$3 satisfies double compactness, weakening, and separation/extensionality axioms. Their self-duality is simply the exchange of the two sorts: $X$4 This is identified as an extension of Lawson self-duality for continuous domains (Abbadini et al., 24 Jul 2025).

The spatial and pointfree formalisms are linked by an exact classification. There is a bijective correspondence between isomorphism classes of ko-spaces, distributive bi-dcpos, and distributive embedded bi-dcpos, and at the categorical level

$X$5

The category $X$6 is self-dual by de Groot duality on objects and converse of c-relations on morphisms, and this restricts to classical de Groot self-duality on the full subcategory of stably compact spaces and closed relations. On the pointfree side, the same categorical symmetry restricts to Lawson duality under the equivalence with continuous domains (Abbadini et al., 24 Jul 2025).

5. Constructions, examples, and preserved operations

A standard concrete example is the interval $X$7: the de Groot dual of $X$8 with the upper topology is $X$9 with the lower topology. More generally, every well-filtered $X^\partial$0 space $X^\partial$1 embeds into the ko-space framework as

$X^\partial$2

and every dcpo $X^\partial$3 yields the ko-space

$X^\partial$4

Its de Groot dual is

$X^\partial$5

(Abbadini et al., 24 Jul 2025).

The constructive presentation theory makes several operations on stably compact locales transparent under dualization. The patch construction is explicitly invariant: $X^\partial$6 Upper and lower powerlocales are exchanged,

$X^\partial$7

while the double powerlocale and Vietoris powerlocale commute with de Groot duality: $X^\partial$8 Valuations and covaluations are interchanged,

$X^\partial$9

and for the probabilistic variants the constructions commute with duality after identifying probabilistic valuations with probabilistic covaluations (Kawai, 2018).

The broader ko-space/bi-dcpo framework also recovers classical space-frame dualities. In particular, there is a one-to-one correspondence between locally compact frames, bicontinuous bi-dcpos with $X$0 a bounded distributive lattice, locally compact distributive embedded bi-dcpos of the specified finite-meet/finite-join type, locally compact ko-spaces with the corresponding finite-intersection/finite-union closure, and locally compact sober spaces (Abbadini et al., 24 Jul 2025). This suggests that de Groot self-duality is structurally tied not only to stably compactness but also to a wider family of Stone-like dualities.

6. Scope, misconceptions, and distinct uses of “self-duality”

A persistent misconception is that de Groot self-duality is about complements of arbitrary compact sets. In the non-$X$1 setting, the correct objects are compact saturated sets. Another misconception is that any Hausdorff space should admit ordinary de Groot duality. This fails: Arens–Fort and Fortissimo spaces are cited as Hausdorff examples where compact sets are just the finite sets, so the double de Groot dual becomes discrete and does not recover the original topology (Abbadini et al., 24 Jul 2025).

A second source of confusion is terminological overload. “Self-duality” is used in several mathematically unrelated senses. An explicit algebraic self-duality such as

$X$2

for finite flat complete intersections is a Gorenstein or complete-intersection duality constructed from a presentation and a determinant element $X$3; it is not presented as de Groot duality, and De Groot is not mentioned there (Kuhn et al., 2021). This suggests that de Groot self-duality should be reserved for the topological and pointfree involution exchanging open structure with complements of compact saturated structure, together with its ko-space and bi-dcpo extensions.

The most precise modern interpretation is therefore twofold. In the classical setting, de Groot self-duality is the involution

$X$4

on stably compact spaces, defined by

$X$5

In the extended setting, it is the spatial side of a larger categorical symmetry between open-like and compact-like data, mirrored pointfreely by the swap

$X$6

This places de Groot self-duality among the central organizing symmetries of Stone-like duality theory rather than as an isolated operation on a special class of spaces (Abbadini et al., 24 Jul 2025).