nLab formal topology

This entry is about a generalized notion of topology. For the notion of formal space in the sense of rational homotopy theory, see formal dg-algebra.

---

Idea

Formal topology is a programme for doing topology in a finite, predicative, and constructive fashion.

It is a kind of pointless topology; in the context of classical mathematics, it reproduces the theory of locales rather than topological spaces (although of course one can recover topological spaces from locales).

The basic definitions can be motivated by an attempt to study locales entirely through the posites that generate them. However, in order to recover all basic topological notions (particularly those associated with closed rather than open features) predicatively, we need to add a ‘positivity’ relation to the ‘coverage’ relation of sites.

Definitions

A formal topology or formal space is a set $S$ together with

• an element $\top$ of $S$,
• a binary operation $\cap$ on $S$,
• a binary relation $\lhd$ between elements of $S$ and subsets of $S$, and
• a unary relation $\Diamond$ on $S$,

such that

1. $a = b$ whenever $a \lhd \{b\}$ and $b \lhd \{a\}$,
2. $a \lhd U$ whenever $a \in U$,
3. $a \lhd V$ whenever $a \lhd U$ and $x \lhd V$ for all $x \in U$,
4. $a \cap b \lhd U$ whenever $a \lhd U$ or $b \lhd U$,
5. $a \lhd \{ x \cap y \;|\; x \in U,\; y \in V \}$ whenever $a \lhd U$ and $a \lhd V$,
6. $a \lhd \{\top\}$,
7. $\Diamond x$ for some $x \in U$ whenever $\Diamond a$ and $a \lhd U$, and
8. $a \lhd U$ whenever $a \lhd U$ if $\Diamond a$,

for all $a$, $b$, $U$, and $V$.

We interpret the elements of $S$ as basic opens in the formal space. We call $\top$ the entire space and $a \cap b$ the intersection of $a$ and $b$. We say that $a$ is covered by $U$ or that $U$ is a cover of $a$ if $a \lhd U$. We say that $a$ is positive or inhabited if $\Diamond a$. (For a topological space equipped with a strict topological base $S$, taking these intepretations literally does in fact define a formal space; see the Examples.)

Some immediate points to notice:

• If we drop (1), then the hypothesis of (1) defines an equivalence relation on $S$ which is a congruence for $\top$, $\cap$, $\lhd$, and $\Diamond$, so that we may simply pass to the quotient set. In appropriate foundations, we can even allow $S$ to be a preset originally, then use (1) as a definition of equality.
• We can prove that $(S,\cap,\top)$ is a bounded semilattice; if (as the notation suggests) we interpret this as a meet-semilattice, then $a \leq b$ if and only if $a \lhd \{b\}$. Conversely, we could require that $(S,\cap,\top)$ be a semilattice originally, then let (1) say that $a \leq b$ whenever $a \lhd \{b\}$.
• We can prove that $\Diamond a$ holds iff every cover of $a$ is inhabited and that $\Diamond a$ fails iff $a \lhd \empty$. Accordingly, this predicate is uniquely definable (in two equivalent ways, one impredicative and one nonconstructive) in a classical treatment; only in a treatment that is both predicative and constructive do we need to include it in the axioms. See positivity predicate.

Examples

Let $X$ be a topological space, and let $S$ be the collection of open subsets of $X$. Let $\top$ be $X$ itself, and let $a \cap b$ be the literal intersection of $a$ and $b$ for $a, b \in S$. Let $a \lhd U$ if and only $U$ is literally an open cover of $a$, and let $\Diamond a$ if and only if $a$ is literally inhabited. Then $(S,\top,\cap,\lhd,\Diamond)$ is a formal topology.

The above example is impredicative (since the collection of open subsets is generally large), but now let $S$ be a base for the topology of $X$ which is strict in the sense that it is closed under finitary intersection. Let the other definitions be as before. Then $(S,\top,\cap,\lhd,\Diamond)$ is a formal topology.

More generally, let $B$ be a subbase for the topology of $X$, and let $S$ be the free monoid on $B$, that is the set of finite lists of elements of $B$ (so this example is not strictly finitist), modulo the equivalence relation by which two lists are identified if their intersections are equal. Let $\top$ be the empty list, let $a \cap b$ be the concatenation of $a$ and $b$, let $a \lhd U$ if the intersection of $a$ is contained in the union of the intersections of the elements of $U$, and let $\Diamond a$ if the intersection of $a$ is inhabited. Then $(S,\top,\cap,\lhd,\Diamond)$ is a formal topology.

Let $X$ be an accessible locale generated by a posite whose underlying poset $S$ is a (meet)-semilattice. Let $\top$ and $\cap$ be as in the semilattice structure on $S$, and let $a \lhd U$ if $U$ contains a basic cover (in the posite structure on $S$) of $a$. Then we get a formal topology, defining $\Diamond$ in the unique way.

The last example is not predicative, and this is in part why one studies formal topologies instead of sites, if one wishes to be strictly predicative. (It still needs to be motivated that we want $\Diamond$ at all.)

 In dependent type theory

Impredicative definition

In dependent type theory, let $(\Omega, \mathrm{El}_\Omega)$ be the type of all propositions, and let the relation $a \in_S A$ for $a:S$ and $A:S \to \Omega$ as

$$a \in_S A \coloneqq \mathrm{El}_\Omega(A(a))$$

The singleton subtype function is a function $\{-\}:S \to (S \to \Omega)$, such that for all elements $a:S$, $p(a):a \in_S \{a\}$, and for all functions $R:S \to (S \to \Omega)$ such that for all elements $a:S$, $p_R(a):a \in_S R(a)$, the type of functions $(b \in_S \{a\}) \to (b \in_S R(a))$ is contractible for all $a:A$ and $b:A$.

A formal topology or formal space is a type $S$ together with

• an element $\top:S$,
• a binary operation $\cap:S \times S \to S$,
• a binary relation $a \lhd A$ between elements of $S$, $a:S$, and subtypes of $S$, $A:S \to \Omega$, and

such that

1. For all elements $a:S$ and $b:S$, the canonical function
   $$\mathrm{idToFormalTop}(a, b):(a =_S b) \to ((a \lhd \{b\}) \times (b \lhd \{a\}))$$

   is an equivalence of types.

   $$\prod_{a:S} \prod_{b:S} \mathrm{isEquiv}(\mathrm{idToFormalTop}(a, b))$$
2. For all elements $a:S$ and subsets $A:S \to \Omega$, if $a \in_S A$, then $a \lhd A$
   $$\prod_{a:S} \prod_{A:S \to \Omega} (a \in_S A) \to (a \lhd A)$$
3. For all elements $a:S$ and subsets $A:S \to \Omega$ and $B:S \to \Omega$, if $a \lhd A$ and for all elements $x:S$, $x \lhd B$ and $x \in_S A$, then $a \lhd B$
   $$\prod_{a:S} \prod_{A:S \to \Omega} \prod_{B:S \to \Omega} \left((a \lhd A) \times \prod_{x:S} (x \in_S A) \times (x \lhd B)\right) \to (a \lhd B)$$
4. For all elements $a:S$ and $b:S$ and subsets $A:S \to \Omega$, if $a \lhd A$ or $b \lhd A$, then $a \cap b \lhd A$.
   $$\prod_{a:S} \prod_{b:S} \prod_{A:S \to \Omega} ((a \lhd A) \times (b \lhd A)) \to (a \cap b \lhd A)$$
5. For all elements $a:S$ and subsets $A:S \to \Omega$ and $B:S \to \Omega$, if $a \lhd A$ and $a \lhd B$, then
   $$a \lhd \sum_{z:S} \prod_{x:S} \prod_{y:S} (x \in_S A) \times (y \in_S B) \times (z =_S x \cap u)$$
   $$\prod_{a:S} \prod_{A:S \to \Omega} \prod_{B:S \to \Omega} (a \lhd A) \times (a \lhd B) \to \left(a \lhd \sum_{z:S} \prod_{x:S} \prod_{y:S} (x \in_S A) \times (y \in_S B) \times (z =_S x \cap y)\right)$$
6. For all elements $a:S$, $a \lhd \{\top\}$,
   $$\prod_{a:S} a \lhd \{\top\}$$

One can define the positivity predicate $\Diamond a$ on $S$ by

$$\Diamond a \coloneqq \prod_{A:S \to \Omega} (a \lhd A) \to (\exists x:S.x \in_S A)$$

Theorem

For all elements $a:S$ and subsets $A:S \to \Omega$, if $\Diamond a$ and $a \lhd A$, there exists an element $x:S$ such that $x \in_S A$ and $\Diamond x$.

$$\prod_{a:S} \prod_{A:S \to \Omega} (\Diamond a) \times (a \lhd A) \to \sum_{x:S} (x \in_S A) \times (\Diamond x)$$

Theorem

For all elements $a:S$ and subsets $A:S \to \Omega$, if $\Diamond a$ implies $a \lhd A$, then $a \lhd A$.

$$\prod_{a:S} \prod_{A:S \to \Omega} (\Diamond a \to a \lhd A) \to (a \lhd A)$$

 Predicative definition

Suppose that the dependent type theory does not have a type of all propositions, nor any type universes. This means that subtypes of a type $S$ do not form a type, which is necessary for defining the cover relation, which is inherently a relation between elements of $S$ and subtypes of $S$. However, one can resolve this problem by using a definition of formal topology which does not require a cover relation in its definition, and then inductively define the cover relation on $S$ as a higher inductive family of inductive cover relations for every single type in $A$.

In addition, the positivity predicate $S$ also needs to be separately defined.

Cover relation

Ayberk Tosun in Tosun 2020 defined a formal topology as a poset $(A, \leq)$ with families of types $x:A \vdash B(x)$, $x:A, y:B(x) \vdash C(x, y)$ and a dependent function

$$d:\prod_{x:A} \prod_{y:B(x)} C(x, y) \to A$$

such that

• for all $x:A$, $y:B(x)$, and $z:C(x, y)$, $d(x, y, z) \leq x$

$$\prod_{x:A} \prod_{y:B(x)} \prod_{z:C(x, y)} d(x, y, z) \leq x$$

• for all $x:A$ and $x':A$, $x' \leq x$ implies that for all $y:B(x)$ one can construct $y':B(x')$ such that for all $z:C(x, y)$, one can construct $z':C(x', y')$ such that $d(x', y', z') \leq d(x, y, z)$.

$$\prod_{x:A} \prod_{x:A'} (a \leq a') \to \prod_{y:B(x)}\sum_{y':B(x')} \prod_{z:C(x, y)} \sum_{z':C(x', y')} d(x', y', z') \leq d(x, y, z)$$

If one has a type universe $U$, then the locally $U$-small inductive cover relation is given by

$$\frac{\Gamma \vdash a:A \quad \Gamma \vdash I:U \quad \Gamma \vdash e:I \hookrightarrow A \quad \Gamma \vdash p:\exists x:I.e(x) =_A a}{\Gamma \vdash \mathrm{dir}_U(a, I, e, p):a \lhd_U (I, e)}$$

$$\frac{\Gamma \vdash a:A \quad \Gamma \vdash I:U \quad \Gamma \vdash e:I \hookrightarrow A \quad \Gamma \vdash b:B(a) \quad \Gamma \vdash f:\prod_{c:C(a, b)} d(a, b, c) \lhd_U (I, e)}{\Gamma \vdash \mathrm{branch}_U(a, I, e, b, f):a \lhd_U (I, e)}$$

$$\frac{\Gamma \vdash a:A \quad \Gamma \vdash I:U \quad \Gamma \vdash e:I \hookrightarrow A \quad \Gamma \vdash p:a \lhd_U (I, e) \quad \Gamma \vdash q:a \lhd_U (I, e)}{\Gamma \vdash \mathrm{squash}_U(a, I, e, p, q):p =_{a \lhd_U (I, e)} q}$$

If one doesn’t have type universes, then one could use the stack semantics instead. Given a type $I$, the cover relation for $I$ is a higher inductive type family $a \lhd_I e$ between elements $a:A$ and embeddings of types $e:I \hookrightarrow A$ generated by the constructors

$$\frac{\Gamma \vdash a:A \quad \Gamma \vdash I \; \mathrm{type} \quad \Gamma \vdash e:I \hookrightarrow A \quad \Gamma \vdash p:\exists x:I.e(x) =_A a}{\Gamma \vdash \mathrm{dir}_I(a, e, p):a \lhd_I e}$$

$$\frac{\Gamma \vdash a:A \quad \Gamma \vdash I \; \mathrm{type} \quad \Gamma \vdash e:I \hookrightarrow A \quad \Gamma \vdash b:B(a) \quad \Gamma \vdash f:\prod_{c:C(a, b)} d(a, b, c) \lhd_I e}{\Gamma \vdash \mathrm{branch}_I(a, e, b, f):a \lhd_I e}$$

$$\frac{\Gamma \vdash a:A \quad \Gamma \vdash I \; \mathrm{type} \quad \Gamma \vdash e:I \hookrightarrow A \quad \Gamma \vdash p:a \lhd_I e \quad \Gamma \vdash q:a \lhd_I e}{\Gamma \vdash \mathrm{squash}_I(a, e, p, q):p =_{a \lhd_I e} q}$$

Families of types $x:A \vdash B(x)$ are equivalently functions $B \to A$, so one can express the above definition in terms of single types. A formal topology is a poset $(A, \leq)$ with a set $B$, a function $f:B \to A$, a set $C$ with a function $g:C \to B$ and a function $d:C \to A$, such that

• for all $z:C$, $d(z) \leq g(f(z))$

$$\prod_{z:C} d(z) \leq g(f(z))$$

• for all $x:A$ and $x':A$, if $x \leq x'$, then for all $y:B$, $f(y) =_A x$ implies that one can construct a $y':B$ such that $f(y') =_A x'$ implies that for all $z:C$, $g(z) =_B y$ implies that one can construct a $z':C$ such that $g(z') =_B y'$ implies $d(z') \leq d(z)$.

$$\prod_{x:A} \prod_{x:A'} (a \leq a') \to \prod_{y:B} (f(y) =_A x) \to \sum_{y':B} (f(y') =_A x') \to \prod_{z:C} (g(z) =_B y) \to \sum_{z':C} (g(z') =_B y') \to (d(z') \leq d(z))$$

If one has a type universe $U$, then the locally $U$-small inductive cover relation is given by

$$\frac{ \begin{array}{c} \Gamma \vdash B \; \mathrm{type} \quad \Gamma \vdash C \; \mathrm{type} \quad \Gamma \vdash f:B \to A \quad \Gamma \vdash g:C \to B \quad \Gamma \vdash d:C \to A \\ \Gamma \vdash a:A \quad \Gamma \vdash I:U \quad \Gamma \vdash e:I \hookrightarrow A \quad \Gamma \vdash p:\exists x:I.e(x) =_A a \end{array} }{\Gamma \vdash \mathrm{dir}_U(a, I, e, p):a \lhd_U (I, e)}$$

$$\frac{ \begin{array}{c} \Gamma \vdash B \; \mathrm{type} \quad \Gamma \vdash C \; \mathrm{type} \quad \Gamma \vdash f:B \to A \quad \Gamma \vdash g:C \to B \quad \Gamma \vdash d:C \to A \\ \Gamma \vdash a:A \quad \Gamma \vdash I:U \quad \Gamma \vdash e:I \hookrightarrow A \quad \Gamma \vdash b:B \quad \Gamma \vdash p:f(b) =_A a \quad \Gamma \vdash h:\prod_{c:C} (g(c) =_B b) \to (d(c) \lhd_U (I, e)) \end{array} }{\Gamma \vdash \mathrm{branch}_U(a, I, e, b, p, h):a \lhd_U (I, e)}$$

$$\frac{ \begin{array}{c} \Gamma \vdash B \; \mathrm{type} \quad \Gamma \vdash C \; \mathrm{type} \quad \Gamma \vdash f:B \to A \quad \Gamma \vdash g:C \to B \quad \Gamma \vdash d:C \to A \\ \Gamma \vdash a:A \quad \Gamma \vdash I:U \quad \Gamma \vdash e:I \hookrightarrow A \quad \Gamma \vdash p:a \lhd_U (I, e) \quad \Gamma \vdash q:a \lhd_U (I, e) \end{array} }{\Gamma \vdash \mathrm{squash}_U(a, I, e, p, q):p =_{a \lhd_U (I, e)} q}$$

If one doesn’t have type universes, then one could use the stack semantics instead. Given a type $I$, the cover relation for $I$ is a higher inductive type family $a \lhd_I e$ between elements $a:A$ and embeddings of types $e:I \hookrightarrow A$ generated by the constructors

$$\frac{ \begin{array}{c} \Gamma \vdash B \; \mathrm{type} \quad \Gamma \vdash C \; \mathrm{type} \quad \Gamma \vdash f:B \to A \quad \Gamma \vdash g:C \to B \quad \Gamma \vdash d:C \to A \\ \Gamma \vdash a:A \quad \Gamma \vdash I \; \mathrm{type} \quad \Gamma \vdash e:I \hookrightarrow A \quad \Gamma \vdash p:\exists x:I.e(x) =_A a \end{array} }{\Gamma \vdash \mathrm{dir}_I(a, e, p):a \lhd_I e}$$

$$\frac{ \begin{array}{c} \Gamma \vdash B \; \mathrm{type} \quad \Gamma \vdash C \; \mathrm{type} \quad \Gamma \vdash f:B \to A \quad \Gamma \vdash g:C \to B \quad \Gamma \vdash d:C \to A \\ \Gamma \vdash a:A \quad \Gamma \vdash I \; \mathrm{type} \quad \Gamma \vdash e:I \hookrightarrow A \quad \Gamma \vdash b:B \quad \Gamma \vdash p:f(b) =_A a \quad \Gamma \vdash h:\prod_{c:C} (g(c) =_B b) \to (d(c) \lhd_I e) \end{array} }{\Gamma \vdash \mathrm{branch}_I(a, e, b, p, h):a \lhd_I e}$$

$$\frac{ \begin{array}{c} \Gamma \vdash B \; \mathrm{type} \quad \Gamma \vdash C \; \mathrm{type} \quad \Gamma \vdash f:B \to A \quad \Gamma \vdash g:C \to B \quad \Gamma \vdash d:C \to A \\ \Gamma \vdash a:A \quad \Gamma \vdash I \; \mathrm{type} \quad \Gamma \vdash e:I \hookrightarrow A \quad \Gamma \vdash p:a \lhd_I e \quad \Gamma \vdash q:a \lhd_I e \end{array} }{\Gamma \vdash \mathrm{squash}_I(a, e, p, q):p =_{a \lhd_I e} q}$$

Positivity predicate

Since we cannot quantify over subtypes, we needs to use an inference rule to define the positivity predicate:

$$\frac{\Gamma \vdash A \; \mathrm{type}}{\Gamma, x:A \vdash \Diamond x \; \mathrm{type}}$$

$$\frac{\Gamma \vdash A \; \mathrm{type}}{\Gamma, x:A, p:\Diamond x, q:\Diamond x \vdash \mathrm{squash}_\Diamond(x, p, q):p =_{\Diamond x} q}$$

$$\frac{ \begin{array}{c} \Gamma \vdash A \; \mathrm{type} \quad \Gamma \vdash I \; \mathrm{type} \quad \Gamma \vdash e:I \hookrightarrow A \\ \Gamma \vdash a:A \quad \Gamma \vdash p:(a \lhd_I e) \to (\exists x:I.e(x) =_A a) \end{array} }{\Gamma \vdash \mathrm{pos}_{A, I}(e, a, p):\Diamond a}$$

 Related concepts

• inductive cover

References

• Mike Fourman and Grayson (1982); Formal Spaces. This is the original development, intended as an application of locale theory to logic.

• Giovanni Sambin (1987); Intuitionistic formal spaces; pdf.

  • This is the probably the main reference on the subject.
  • Warning: you can ignore the material about foundations and type theory, but if you do read any of it, the term ‘category’ means (roughly) proper class; this was common in type theory in the 1980s.

• Giovanni Sambin (2001); Some points in formal topology; pdf. This has newer results, alternative formulations, etc.

• Erik Palmgren, From Intuitionistic to Point-Free Topology: On the Foundation of Homotopy Theory, Logicism, Intuitionism, and Formalism Volume 341 of the series Synthese Library pp 237-253, 2005 (pdf)

• Ayberk Tosun, Formal Topology in Univalent Foundations, (pdf, slides)