Contents

topos theory

# Contents

## Idea

Subsets $A$ of a set $X$ correspond precisely to maps from $X$ to the set of truth values of classical logic via their characteristic function $\chi_A:X\to \{0,1\}$. The concept of a subobject classifier generalizes this situation to toposes other than Set:

A subobject classifier in a topos is a morphism $true : * \to \Omega$ such that every monomorphism $A \hookrightarrow B$ in the topos (hence every subobject) is the pullback of this morphism along a unique morphism (the characteristic morphism of $A$) $B \to \Omega$.

In this sense $\Omega$ is the classifying object for subobjects and $true : * \to \Omega$ the generic subobject.

The existence of a subobject classifier in a category is a powerful property which induces much other structure that lies at the heart of topos theory.

By restricting the class of monomorphism appropriately, the concept can be relativized to the concept of an M-subobject classifier1: e.g. demanding only classification of strong monomorphisms leads to quasitoposes.

In type theory, a type closely related to the subobject classifier is the type of propositions, often denoted $Prop$ or (sometimes in homotopy type theory) $hProp$.

## Definition

###### Definition

In a category $C$ with finite limits, a subobject classifier is a monomorphism $true \,\colon\, * \to \Omega$ out of the terminal object, such that for every monomorphism $U \to X$ in $C$ there is a unique morphism $\chi_U \colon X \to \Omega$ such that the following square commutes and is Cartesian (a pullback):

$\array{ U &\longrightarrow& * \\ \big\downarrow &{}^{{}_{(pb)}}& \big\downarrow{{}^\mathrlap{true}} \\ X &\underset{\chi_U}{\longrightarrow}& \Omega \mathrlap{\,.} }$

See for instance (MacLane-Moerdijk, p. 32).

###### Remark

Some terminology:

If it exists, the object $\Omega$ is also called the object of truth values, a global element $K \to \Omega$ is called a truth value and the element $true : * \hookrightarrow \Omega$ is the truth value true, where all these terms allude to the internal logic of the category $C$.

Note that the subobjects classified by the truth values are subterminal objects.

The morphism $\chi_U$ is also called the characteristic map or classifying map of the subobject $U \hookrightarrow X$.

###### Proposition

If $C$ has finite limits and is in addition a locally small category, then it has a subobject classifier precisely if the subobject-assigning presheaf

$Sub : C^{op} \to Set$
$X \mapsto \{U \hookrightarrow X\}/\sim$

is representable. In this case the representing object is the subobject classifier: there is a natural isomorphism

$Sub(X) \simeq C(X, \Omega)$

in $X \in C$.

Moreover, in this case $C$ is well powered.

This appears for instance as (MacLane-Moerdijk, prop. I.3.1).

In more detail: given a morphism $f: c \to d$ in $C$, the function

$Sub(f): Sub(d) \to Sub(c)$

takes a subobject $i: t \hookrightarrow d$ to the subobject of $c$ obtained by pulling back $i$ along $f$. (Notice that monomorphisms, as discussed there, are stable under pullback.)

The representability of this functor means there is an object $\Omega$ together with a subobject $t: T \hookrightarrow \Omega$ which is universal, meaning that given any subobject $i: s \hookrightarrow c$, there is a unique morphism $f: c \to \Omega$ such that $i$ is obtained as the pullback of $t$ along $f$.

###### Proof

To see that a subobject classifier induces such a natural isomorphism, we need that the morphisms $Sub(f)$ for $f \in Mor(C)$ corresponds to the morphisms $C(f,\Omega)$. This is the pasting law for pullbacks.

Conversely, to see that a subobjects-representing object $\Omega$ is a subobject classifier, use that by naturality we have for each morphism $\phi : X \to \Omega$ a commuting diagram

$\array{ Sub(\Omega) &\stackrel{\simeq}{\to}& C(\Omega, \Omega) \\ {}^{\mathllap{Sub(\phi)}}\downarrow && \downarrow^{\mathrlap{C(\phi,\Omega)}} \\ Sub(X) &\stackrel{\simeq}{\to}& C(X, \Omega) }$

whose commutativity says that every element of $Sub(X)$ is the pullback along some $\phi : X \to \Omega$ of the subobject of $\Omega$ corresponding under the natural isomorphism to $Id : \Omega \to \Omega$.

By further playing around with this one finds that this latter subobject of $\Omega$ has to be a terminal object.

## Examples

### In $Set$

In the category Sets of sets, the 2-element set $\mathbf{2} = \{f, t\}$ (the classical Boolean domain) plays the role of $\Omega$; the morphism $t \colon 1 \to \mathbf{2}$ just names the element $t$. Given a subset $S \subseteq X$, the characteristic function $\chi_S: X \to \mathbf{2}$ is the function defined by $\chi_S(x) = t$ if $x \in S$, and $\chi_S(x) = f$ if $x \notin S$.

###### Remark

It is not usually true in toposes that $\Omega$ is the coproduct $\mathbf{2} = 1 + 1$; toposes where that occurs are called Boolean. Thus the category $Set$ of sets is a Boolean topos, as is the presheaf topos $Set^G$ when $G$ is a groupoid.

### In a presheaf topos

###### Proposition

The subobject classifier in a presheaf topos $PSh(S)$ is the presheaf that sends each object $U \in S$ to the set $sieves(U)$ of sieves on it

$\Omega \;\colon\; U \mapsto sieves(U)$

Here $sieves(U)$ is equivalently the set of subobjects of the representable presheaf $Y(U)$.

The corresponding morphism $true : * \to \Omega$ of presheaves is the natural transformation that picks over each object the maximal sieve $true_U = maximal_{sieves(U)} : * \to sieves(U)$.

###### Remark

If one views a presheaf over a small category $S$ as a set varying (or evolving) over $S$, then the subobject classifier in $PSh(S)$ may be viewed as encapsulating the ways of an element to be in the set ranging from ‘never’ to ‘always’ being in the set.

In $Set = PSh(\ast)$ these two extremes are the only possibilities, but in the general case there are ways to become an element ‘over time’.

It might be helpful to have a look at the simple example of a subobject classifier in the presheaf topos of directed graphs worked out at Quiv to get some intuition.

#### In $G Set$

As a special case of presheaf toposes, for $G$ a discrete group and $G Set = [\mathbf{B} G, Set]$ the topos of G-sets, there are precisely two sieves on the single object of the delooping groupoid $\mathbf{B}G$: the trivial one and the empty one. Hence the subobject classifier here is the 2-element set as in Set, but now regarded as a $G$-set with trivial $G$-action.

For toposes of (left) actions of general monoids $M$ the picture changes dramatically, since then there will usually exist non-trivial left ideals and, accordingly, the structure of $\Omega$ will become richer: Its underlying set has elements all left ideals, with $m\in M$ acting on a left ideal $L$ by mapping it to the left ideal $\{n\in M| n\cdot m\in L\}$. (Of course, the same description of $\Omega$ applies to the case of groups as well, it is just, that groups lack non-trivial left ideals!)

### In a sheaf topos

The following is adapted from p. 34 of Caramello.

Let $(C, J)$ be a Grothendieck site. A sieve $S$ on an object $c \in C$ is $J$-closed when, for any arrow $f :d \to c$, if $f^*(S) \in J(d)$ then $f \in S$. It’s called closed since this is a closure property: $S$ contains all arrows it covers. If $(C, J)$ is a locale with the open covers topology, then this condition means that $S$ is a collection of subopens of $c$ which is closed under unions. Clearly, every sieve on $c$ can be saturated to a closed one.

The subobject classifier of $\mathrm{Sh}(C,J)$ is defined on objects as

$\Omega_J(c) = \{ S | \text{S is a J-closed sieve on c} \}$

and on morphisms as

$\Omega_J(f : c \to d) = f^*$

where $f^*$ denotes, as above, pullback of sieves.

The arrow $\mathsf{true} : 1 \to \Omega_J$ picks the maximal sieve on each object (the one generated by the identity).

Given a mono $m : B \to A$, the classifying arrow $\chi_m : A \to \Omega_J$ is given by

$\chi_m(c)(x) = \{f : d \to c | A(f)(x) \in B(d)\},$

for any $c \in C$ and $x \in A(c)$.

### In a non-boolean topos

An example of a non-Boolean topos is the category of sheaves over a “typical” topological space $X$ such as the real line $\mathbb{R}$ in its usual topology. In this case, $\Omega$ is the sheaf where the set of sections over an open subset $U$ is the set of open subsets of $U$, with the obvious restriction maps; the sheaf topos in this case is guaranteed to be non-Boolean provided there are some non-regular open sets in $X$ (a open set is regular if it is the interior of its closure). The “internal logic” of such a topos is intuitionistic.

### In a slice topos

###### Proposition

Let $\mathcal{E}$ be a topos and $X \in \mathcal{E}$ any object. Write $\mathcal{E}/X$ for the corresponding over-topos.

The subobject classifier of $\mathcal{E}/X$ is $p_2 : \Omega_{\mathcal{E}} \times X \to X$.

###### Proof

This follows for instance from the statement that the inverse image of any base change geometric morphism is a logical functor and hence preserves subobject classifiers: Here we are looking at the base change along $p : X \to *$ and hence $p^* \Omega_{\mathcal{E}}\simeq \Omega_{\mathcal{E}} \times X$.

But the statement is also easily directly checked.

### In a non-topos

The category $Set_\ast$ of pointed sets has a subobject classifier (specified up to unique isomorphism as the pointed set with two elements).

If one is willing to admit non-locally small categories, then the category of classes in ZFC is not a topos (it is not cartesian closed) but has a subobject classifier: any two-element set.

The opposite of the category of commutative von Neumann algebras has a subobject classifier given by $\mathbb{C}^2$ according to Simon Henry on MathOverflow, but is not a topos because it is not cartesian closed (see Dmitri Pavlov’s answer to the same MathOverflow question).

## Properties

Suppose a category $\mathbf{C}$ has a subobject classifier; this entails some striking structural consequences for $\mathbf{C}$. We list a few here:

###### Proposition

Every monomorphism in $\mathbf{C}$ is a regular monomorphism, i.e. is an equalizer of some pair of maps.

###### Proof

For $\chi_i: X \to \Omega$ the characteristic map of a mono $i: A \to X$, we find that $i$ is the equalizer of a pair of maps $X \rightrightarrows \Omega$:

$\array{ & & 1 \\ & \mathllap{!} \nearrow & \downarrow \mathrlap{t} \\ X & \underset{\chi_i}{\to} & \Omega. }$
###### Corollary

$\mathbf{C}$ is balanced, i.e., a morphism in $\mathbf{C}$ is an isomorphism iff it is both monic and epic.

###### Proof

“Only if” is trivial. The “if” comes from the fact that an epic (epimorphic) equalizer must be an isomorphism, for if $i: A \to X$ is the equalizer of $f, g: X \rightrightarrows Y$ and $i$ is epic, then $f = g$, whence $1_X$ is their equalizer, so $i: A \to X$ must have been an isomorphism.

###### Proposition

Any two epi-mono factorizations of a map in $\mathbf{C}$ are canonically isomorphic.

###### Proof

Suppose $i p = j q$ where $p, q$ are epic and $i, j$ are monic. Since $j$ is regular, it is the equalizer of some parallel pair $f, g$ as in the diagram

$\array{ A & \stackrel{p}{\to} & B & & \\ \mathllap{q} \downarrow & & \downarrow \mathrlap{i} & & \\ C & \underset{j}{\to} & D & \stackrel{\overset{f}{\to}}{\underset{g}{\to}} & E, }$

so that $f i p = f j q = g j q = g i p$, whence $f i = g i$ since $p$ is epic, whence $i$ factors through $j$ as $j$ is the equalizer: $i = j k$ for some $k: B \to C$. Then also $k p = q$ since $j k p = i p = j q$ and $j$ is monic. We have that $k$ is monic since $i$ is, and $k$ is epic since $q$ is. Thus $k$ is an isomorphism.

Already these results impose some tight restrictions on $\mathbf{C}$. We get some more by exploiting the internal structure of $\Omega$.

The subobject classifier always comes with the structure of an internal poset; that is, a relation $\subseteq\, \hookrightarrow \Omega\times\Omega$ which is internally reflexive, antisymmetric, and transitive. This can be constructed directly (see Proposition below), or obtained via the Yoneda lemma since the collection of subobjects of any object is an external poset.

Similarly, since we assume that $\mathbf{C}$ is finitely complete, each subobject poset $Sub(X)$ has intersections (gotten as pullbacks or fiber products of pairs of monics $i: A \to X,j: B \to X$), and the intersection operation

$\cap: Sub(X) \times Sub(X) \to Sub(X)$

is natural in $X$. Hence we have a family of maps

$\cap: \hom(X, \Omega \times \Omega) \to \hom(X, \Omega)$

natural in $X$; by the Yoneda lemma, we infer the presence of an internal intersection map

$\wedge: \Omega \times \Omega \to \Omega$

making $\Omega$ an internal meet-semilattice.

More significantly, $\Omega$ is an internal Heyting algebra. More accurately, it’s a Heyting algebra provided it has joins; without joins it is a cartesian closed poset:

###### Proposition

There is an internal implication operator

$\Rightarrow: \Omega \times \Omega \to \Omega$

uniquely specified by the internal condition

$w \wedge u \leq v \qquad iff \qquad w \leq u \Rightarrow v.$
###### Proof (sketch)

Construct $\subseteq \hookrightarrow \Omega \times \Omega$ as the equalizer of the pair of maps

$\Omega \times \Omega \stackrel{\overset{\pi_1}{\longrightarrow}}{\underset{\wedge}{\longrightarrow}} \Omega$

and then define $\Rightarrow: \Omega \times \Omega \to \Omega$ to be the characteristic map of $\subseteq \hookrightarrow \Omega \times \Omega$. Now if $\chi_u, \chi_v$ are two maps $X \to \Omega$, one calculates that $w \hookrightarrow X$ is contained in the subobject classified by $\chi_u \Rightarrow \chi_v$ iff $w \cap u = w \cap u \cap v$, which is just a way of saying $w \cap u \leq v$.

###### Corollary

In every subobject poset $Sub(X)$, meets distribute over any joins that exist.

###### Proof

Because $U \cap -$ is left adjoint to the external operator $U \Rightarrow -$ on $Sub(X)$, it preserves any joins that happen to exist in $Sub(X)$.

Normally these results are proved in the context of toposes, where we may say for example that $\Omega$ is an internal Boolean algebra if and only if the topos is Boolean. But as the proofs above indicate, we need only exploit the definition of subobject classifier making reference only to finite limit structure.

In a topos, the subobject classifier $\Omega$ is always injective, and, so is the power object $\Omega^X$ for every object $X$ (See at injective object for some of the details). In particular, every object $X$ embeds into an injective object by the singleton monomorphism $X\to\Omega^X$: ‘A topos has enough injective objects!’. More generally, injective objects in a topos are precisely the ones that are retracts of some $\Omega^X$ (Cf. Borceux 1994, p.315; Moerdijk-MacLane 1994, p.210).

### Johnstone’s exercise

A curiosity from Johnstone’s Topos Theory, posed as an exercise, is that any monomorphism $\Omega \to \Omega$ is an isomorphism and even an involution. Thus $\Omega$ is a Hopfian object?.

An online proof may be found here.

## Categories with a contractible subobject classifier

Suppose a topos $\mathcal{E}$ has a connected components functor $\Pi$ left adjoint to $\Delta\dashv \Gamma$ assigning to an object $X$ the set of it connected components. We call $X$ connected if $\Pi(X)$ is a singleton and contractible if $\Pi(X^Y)$ is connected for all $Y\in\mathcal{E}$. It can be shown that provided $\Pi$ preserves finite products, $\Omega$ is contractible iff $\Omega$ is connected which in turn implies the same for all other injective objects (see at sufficiently cohesive topos).

Now the subobject classifier $\Omega$ has always two disjoint points $\mathsf{true}$, $\mathsf{false}$ whence provided it is connected we can view it (together with its Heyting algebra structure) has a (highly nonlinear) generalized interval object and define a notion of homotopy relative to $\Omega$. The intuition here is that truth and falsity are continuously connected in such toposes and blend into each other, endowing (the logic in) $\mathcal{E}$ with a certain Hegelian flavor. The contractability of $\Omega$ was taken as a key property of a gros topos of spaces by William Lawvere. Further information on this particular class of cohesive toposes and discussion of properties of $\Omega$ relevant in this context is at sufficiently cohesive topos.

## Categories without subobject classifiers

As the previous section indicates, having a subobject classifier is a very strong property of a category and “most” categories with finite limits don’t have one.

For example, there is an easy condition ensuring a category2 with a terminal object can’t have a subobject classifier: if there are no nonidentity morphisms out of the terminal object. This includes the following examples.

• Any top bounded partial order.

• In $Ring$, the category of rings, there are no nonidentity morphisms out of the terminal object the zero ring.

Here’s another obstacle:

• If an abelian category had a subobject classifier, every subobject of every object would have to be the kernel of its classifying map. In particular, the subobject $0$ of every object $A$ would have to be the kernel of its classifying map, so that every object in this abelian category would embed into the subobject classifier $\Omega$ (including, say, all small products of $\Omega$ with itself) which in nontrivial cases would cause size issues.

But a real killer is the fact that all monos are regular, or its consequences of the category being balanced and uniqueness of epi-mono factorizations:

• The categories Pos, Cat, Top are not balanced (consider the map from a discrete structure on a set to an indiscrete structure on the same set, induced by the identity function). The category CMon is not balanced (consider the inclusion $\mathbb{N} \hookrightarrow \mathbb{Z}$ which is epic).

Even though all monos in Grp are regular, we can kill off $Grp$ by observing that if $t: 1 \to \Omega$ were a subobject classifier, the proof of Proposition indicates that every mono $i: A \to X$ would have to be the kernel of $\chi_i$. But not all monos in $Grp$ are kernels.

Perhaps an even more decisive killer is the observation that meets distribute over (arbitrary) joins in subobject orders. This eliminates many categories from consideration:

• Lattices of subobjects in $Grp$ or $Ab$ are rarely distributive.

• For any nontrivial category with biproducts, there are non-distributive subobject lattices. Take any object $A$, so that we have three subobjects $i_1: A \to A \oplus A$, $i_2: A \to A \oplus A$, and $\Delta: A \to A \oplus A$. Then $i_1 \vee i_2 = \top$, whereas $i_1 \wedge \Delta = \bot = i_2 \wedge \Delta$. Under distributivity we have

$\Delta = \Delta \wedge \top = \Delta \wedge (i_1 \vee i_2) = (\Delta \wedge i_1) \vee (\Delta \wedge i_2) = \bot \vee \bot = \bot$

but $\Delta = \bot$ forces $A = 0$. So the only such category that can have a subobject classifier is trivial.

Directly related to this distributivity property is the question when a variety of algebras has a subobject classifier (cf. Johnstone 1990, pp.448f).

## Generalizations

### Weak subobject classifier

There are many subclasses of monomorphisms, such as regular monomorphisms, strong monomorphisms, and extremal monomorphisms. As a result, there are weaker forms of subobject classifiers that only classify a subclass of monomorphisms instead of all monomorphisms.

Suppose there is a category $C$ with finite limits. Let $S$ be a subclass of monomorphisms in $C$, whose elements we shall refer to as $S$-monomorphisms. An $S$-subobject classifier is an $S$-monomorphism $true : * \to \Omega$ out of the terminal object, such that for every $S$-monomorphism $U \to X$ in $C$ there is a unique morphism $\chi_U : X \to \Omega$ such that there is a pullback diagram of the following form:

$\array{ U &\to& * \\ \downarrow && \downarrow^{\mathrlap{true}} \\ X &\stackrel{\chi_U}{\to}& \Omega } \,.$
• If $S$ is the class of all monomorphisms in $C$, then $\Omega$ is just a subobject classifier.
• If $S$ is the class of all extremal monomorphisms in $C$, then $\Omega$ is an extremal subobject classifier.
• If $S$ is the class of all strong monomorphisms in $C$, then $\Omega$ is a strong subobject classifier.
• If $S$ is the class of all regular monomorphisms in $C$, then $\Omega$ is a regular subobject classifier.
• If $S$ is the class of all effective monomorphisms in $C$, then $\Omega$ is an effective subobject classifier.
• If $S$ is the class of all strict monomorphisms in $C$, then $\Omega$ is a strict subobject classifier.
• If $S$ is the class of all $\kappa$-compact monomorphisms, for some regular cardinal $\kappa$, then $\Omega$ is a $\kappa$-compact subobject classifer.

Strong subobject classifiers in particular are important in the definition of a quasitopos.

Let $\kappa$ be an inaccessible cardinal. Then $\kappa$-compact subobject classifers are used in the syntactic category of Martin-Löf type theories like book HoTT.

### Object classifier

In higher topoi the subobject classifiers are the universal fibrations:

In the (n+1,1)-topos $n Grpd$ or $(n,0) Cat$ of (n,0)-categories the subobject classifier is the forgetful functor

$(n,1) true : (n-1,0)Cat_* \to (n-1,0)Cat$

from the $(n,1)$-category of pointed $(n-1,0)$-categories to that of $(n-1,0)$-categories, which forgets the point.

This is described in more detail at generalized universal bundle. See also the discussion at stuff, structure, property.

In fact, using the notion of (-1,0)-category the subobject classifier in Set does fit precisely into this pattern:

the set of truth values $\Omega$ may be regarded as the (0,1)-category of (-1,0)-categories (of which there are two) and the one-element set $*$ is the (0,1)-category of pointed (-1,0)-categories, of which there is one.

In the context of (∞,1)-topos theory subobject classifiers are discussed in section 6.1.6 of

Whereas for 1-toposes the subobject classifier is the key structural ingredient (besides the exactness properties), in higher topos theory this role is taken over by the object classifier, as pointed out in Lurie (2009).

The concept was introduced in

Discussion of the concept can be found in the usual suspects

• Francis Borceux, When is $\Omega$ a cogenerator in a topos ? , Cah. Top. Géom. Diff. Cat. XVI no.1 (1975) pp.3-15. (numdam)