Synthetic topos theory

[003J] Direct images

A morphism between toposes also induces the direct image functor which is a right adjoint of the inverse image functor. The construction of the direct image functor is not so straightforward as that of the inverse image functor. We first define a special case, the global section functor, and then relativize it using [002Y].

[003K] Definition

Let \(T\) be a type theory. We work in \(T\). Let \(i:\mathord {\textnormal {\textsf {Level}}}\) and let \(X\) be a topos. We define the global section functor

\(\Gamma _{X}:\mathcal {S}(X,i)\rightarrow \mathcal {U}(i)\)
by the following composite.
\(\mathcal {S}(X,i)\)\(\simeq \) {[003G]}\(\mathord {\textnormal {\textsf {D}}}_{\mathord {\textnormal {\textsf {Sh}}}(X)}({}\vdash \mathcal {U}(i))\)\(\rightarrow \) {\(\mathord {\textnormal {\textsf {GS}}}_{\mathord {\textnormal {\textsf {Sh}}}(X)}\)}\(\mathcal {U}(i)\)
By [002R], \(\Gamma _{X}\) has a left adjoint which we call the constant sheaf functor and is denoted by \(\Delta _{X}\).

[003N] Observation

Let \(T\) be a base type theory. We work in \(T\). Let \(i:\mathord {\textnormal {\textsf {Level}}}\), let \(X_{1}\) be a topos, and let \(X_{2}(x)\) be a topos under assumption \(x:X_{1}\). Then we have the following equivalences.

\(\mathcal {S}((x:X_{1})\times X_{2}(x),i)\)\(\simeq \) {[002Y]}\(\mathord {\textnormal {\textsf {D}}}_{\mathord {\textnormal {\textsf {Glob}}}_{\mathord {\textnormal {\textsf {Sh}}}(X_{1})}(\mathord {\textnormal {\textsf {Sp}}})}(y:X_{2}(\mathord {\textnormal {\textsf {w}}}_{X_{1}})\vdash \mathcal {E}(i))\)\(\simeq \) {definition}\(\mathord {\textnormal {\textsf {D}}}_{\mathord {\textnormal {\textsf {Sh}}}(X_{1})}({}\vdash \mathcal {S}(X_{2}(\mathord {\textnormal {\textsf {w}}}_{X_{1}}),i))\)

[003L] Definition

Let \(T\) be a base type theory. We work in \(T\). Let \(i:\mathord {\textnormal {\textsf {Level}}}\), let \(X_{1}\) and \(X_{2}\) be toposes, and let \(f:X_{1}\rightarrow X_{2}\) be a morphism. We define the direct image functor

\({f}_{*}:\mathcal {S}(X_{1},i)\rightarrow \mathcal {S}(X_{2},i)\)
by the following composite.
\(\mathcal {S}(X_{1},i)\)\(\simeq \) {fibrant replacement}\(\mathcal {S}((y:X_{2})\times \mathord {\textnormal {\textsf {Fib}}}(f,y),i)\)\(\simeq \) {[003L]}\(\mathord {\textnormal {\textsf {D}}}_{\mathord {\textnormal {\textsf {Sh}}}(X_{2})}({}\vdash \mathcal {S}(\mathord {\textnormal {\textsf {Fib}}}(f,\mathord {\textnormal {\textsf {w}}}_{X_{2}}),i))\)\(\rightarrow \) {\(\Gamma _{\mathord {\textnormal {\textsf {Fib}}}(f,\mathord {\textnormal {\textsf {w}}}_{X_{2}})}\)}\(\mathord {\textnormal {\textsf {D}}}_{\mathord {\textnormal {\textsf {Sh}}}(X_{2})}({}\vdash \mathcal {U}(i))\)\(\simeq \) {[003G]}\(\mathcal {S}(X_{2},i)\)

[003M] Exercise

Let \(T\) be a base type theory. We work in \(T\). Let \(i:\mathord {\textnormal {\textsf {Level}}}\), let \(X_{1}\) and \(X_{2}\) be toposes, and let \(f:X_{1}\rightarrow X_{2}\) be a morphism. Then \({f}_{*}:\mathcal {S}(X_{1},i)\rightarrow \mathcal {S}(X_{2},i)\) is a right adjoint of \({f}^{*}:\mathcal {S}(X_{2},i)\rightarrow \mathcal {S}(X_{1},i)\).