The dissolution of the locale associated with a frame $F$ is the locale associated with the frame $N(F)$ of nuclei of $F$ (see, e.g., Johnstone, “Stone Spaces” (1982), §2.5). Note that there is a frame monomorphism $F \to N(F)$ taking $a\in F$ to the nucleus $c(a)\colon F\to F, x \mapsto a\vee x$, which defines an epimorphism of locales in the other direction.

Now if $\mathbf{T}$ is a topos, since $\mathbf{T}$ is equivalent to the category of internal sheaves on the subobject classifier $\Omega$ of $\mathbf{T}$ (or, more precisely, on the internal locale associated to $\Omega$, which is merely the discrete topology on a singleton), I am inclined to define the “dissolution” of $\mathbf{T}$ as the category of sheaves on $N(\Omega)$ or, more precisely, on the dissolution of the discrete topology on a singleton. Note that $N(\Omega)$ can be seen as the “object of Lawvere-Tierney topologies” on $\mathbf{T}$.

  • Does this notion already exist in the literature (perhaps under a different name)?

  • Are there perhaps alternative definitions (either alternative ways of rephrasing the definition I gave, or genuinely different notions which may compete with it for the name) of “dissolution” of a topos?

  • $\begingroup$ This is interesting; I for one haven’t come across it before. An obvious “sanity check” question is whether it really generalises the existing localic definition: when $\newcommand{\T}{\mathbf{T}}\newcommand{\Sh}{\mathrm{Sh}}\T=\Sh(X)$, do we have $\Sh_{\T}(N(\T)) \simeq \Sh(N(X))$? This seems not quite obvious to me — I guess it comes down to comparing the totalisation of the internal frame $\int_{\mathcal{O}(X)}N(\Omega_\T)$ with the externally defined $N(\mathcal{O}(X))$. $\endgroup$ Feb 28, 2023 at 9:02
  • $\begingroup$ @PeterLeFanuLumsdaine Indeed, I wrote the question assuming this point was obvious, and later realized it wasn't. I need to think about it! $\endgroup$
    – Gro-Tsen
    Feb 28, 2023 at 11:06

1 Answer 1


I'm not aware of litterature on this, but this is something I have thought about several years ago and never ended-up using or publishing. What is below is me trying to remind myself how it works - unfortunately without my notes.

So,I think a good way to understand the dissolution of a topos is in terms of its universal property. First let me recall how it works from locales (This is done in C1 in the elephant, and I think also in Stone in space but I can't check now)

The frame homomorphism $X \to N(X)$ is universal for frame homomorphism $X \to Y$ that sends all elements of $X$ to complemented element of $Y$.

Note that it is also easy to construct "partial dissolution" where you start form a set $S \subset F$ and construct $N_S(F)$ which is universal for making the element of $S$ complemented. THis can be for example constructed as a subframe of N(F).

The analogous construction for a topos $T$, is:

Definition For T a topos, and I a collection of subobject inclusion $(U_i \subset V_i)_{i\in I}$ in $T$, the dissolution of $T$ at $I$ is (if it exists) a topos $N_I(T)$ endowed with an algebraic morphism $T \to N_I(T)$ which is universal for algebraic morphisms $T \to E$ sending all $U_i \subset V_i$ to complemented subobject inclusion in $E$.

By "algebraic morphism" I just mean "Geometric morphism" but in the $f^*$ direction.

It is easy to construct such $N_I(T)$ when $T$ is a Grothendieck topos and $I$ is a set as a classyfing topos: represent $T$ as a classyfing topos in a way so that all the $V_i$ are sorts and all the $U_i$ are definable subsets and then add sort $C_i$ for the complement of $U_i$ with axioms that forces $C_i$ to be the complement of $U_i$ in $V_i$.

An easy but key lemma is the following:

Lemma Let $T$ be a Grothendieck topos, $(X_i \to Y)_{i\in I}$ be a covering family, $S \subset Y$ a subobject and $S_i = X_i \times_Y S$. Let $f:T \to E$ be an algebraic morphism. each each $f(S_i) \subset f(X_i)$ is complemented then $f(S) \subset f(Y)$ is complemented.

It follows that

Proposition: Let $T$ be a Grothendieck topos, then the dissolution $N(T)$ at all subobject inclusion in $T$ exists.

Proof Take a site of a definition for $T$ and construct the dissolution of $T$ at all subobject inclusion $S \subset c$ where $c$ is representable.

Proposition: $N(Sh(X)) = Sh(N(X))$

Proof Sh(N(X)) is clearly universal for making all $V \subset 1$ complemented in $Sh(X)$. This implies that it also makes all $U \subset V$ for $U,V$ subterminal objects complemented. But as in $Sh(X)$ every object is covered by subterminal objects, this imply that it makes all subobject inclusion complemented because of the lemma.


Proposition The dissolution $N(T)$ of a Grothendieck topos can also be constructed internally as sheaves over the internal frame $N(\Omega)$ as suggested in the question.

Proof Reasoning internally in $T$, an algebraic morphism $Sh_T(N(\Omega)) \to E$ for a $T$-topos $E$ is the same as an (so... the unique) algebaic morphism $* \to E$ such that internally it sends every sub-object inclusion to a complemented one. Expressing this externally, $E$ corresponds to a topos with an algebraic morphism $T \to E$ and an extension to $Sh_T(N(\Omega))$ eaxclty means that it sends every subobject to a complemented one, which is exactly the universal property of $N(T)$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.