6.5. Dimension theory
The homotopy dimension of a topos measures how far its global sections functor can lower connectivity. For a slice topos \(\An_{/X}\), this recovers a classical finiteness invariant of the anima \(X\): it asks whether \(X\) is a retract of a finite-dimensional CW-complex. Dimension also controls the convergence problem from the preceding section. A local finite-dimensionality hypothesis forces Postnikov towers to converge, and a uniform local bound also ensures that every compatible system of truncations is realized by an object.
6.5.1. Homotopy dimension and examples
Let \(\phi_*\colon S \to T\) be a morphism of topoi. We say that \(\phi_*\) has dimension \(\leq d\) if, for every \(n \geq d-1\), the functor \(\phi_*\) sends \(n\)-connected maps to \((n-d)\)-connected maps.
We say that \(T\) has dimension \(\leq d\) if the terminal morphism \(\Gamma_* = \Hom_T(*,-)\colon T \to \An\) has dimension \(\leq d\).
A topos \(T\) has dimension \(\leq d\) if and only if every \((d-1)\)-connected object of \(T\) has a global section.
Proof
We say that a topos \(T\) is locally of dimension \(\leq d\) (resp. locally of finite dimension) if it is generated under colimits by objects \(U\) such that the slice topos \(T_{/U}\) has dimension \(\leq d\) (resp. has some finite dimension).
Here are some basic sources of dimension bounds.
Let \(X \in \An\). If \(X\) is represented by a \(d\)-dimensional CW-complex, or by a \(d\)-dimensional simplicial set, then \(\dim(\An_{/X}) \leq d\).
More precisely, \(\dim(\An_{/X}) \leq d\) if and only if \(X\) is a retract, in the homotopy category of animae, of an anima represented by a \(d\)-dimensional CW-complex. If \(d \neq 2\), this is further equivalent to \(X\) itself admitting a \(d\)-dimensional CW-model. For \(d=2\), the retract condition implies that \(X\) admits a \(3\)-dimensional CW-model, but it need not admit a \(2\)-dimensional one. See [Lurie 2009, Example 7.2.1.4].
If \(C\) admits a terminal object \(1\), then \(\PSh(C)\) has dimension \(\leq 0\). Indeed, global sections are given by evaluation at \(1\), and evaluation preserves effective epimorphisms. Equivalently, if \(\Ff\) is \((-1)\)-connected, then \(\Gamma_*(\Ff) \iso \Ff(1)\) is nonempty. See [Lurie 2009, Example 7.2.1.3].
The presheaf category \(\PSh(\N)\) on the poset of natural numbers has dimension \(\leq 1\). Indeed, a \(0\)-connected presheaf is a tower
of connected animae. Its limit is nonempty: after replacing the tower by a tower of Kan fibrations, one may choose a point of \(X_0\) and lift it successively through the tower. Thus every \(0\)-connected object has a global section, and Lemma 6.68 applies.
Let \(X\) be a paracompact Hausdorff space. We say that \(X\) has covering dimension \(\leq d\) if every open covering is refined by one with empty \((d+2)\)-fold intersections. Then
This applies, for example, to every CW-complex of dimension \(\leq d\). See [Lurie 2009, Theorem 7.2.3.6 and Corollary 7.2.3.7].
Let \(X\) be a spectral space of Krull dimension \(\leq d\). Then \(\Shv(X)\) has dimension \(\leq d\); see [Clausen and Mathew 2021, Theorem 3.12].
If \(X\) is a qcqs scheme, then the following bounds hold:
\(\dim(X_{\mathrm{Nis}}) \leq \dim_{\mathrm{Krull}}(X)\); see [Clausen and Mathew 2021, Theorem 3.18].
\(\dim(X_{\mathrm{cdh}}) \leq \dim_{\mathrm{val}}(X)\); see [Elmanto et al. 2021, Theorem 2.4.15].
6.5.2. Convergence and Postnikov-completeness
The following result explains why dimension enters the comparison among the completion conditions in Section 6.4. Its first part is local and only concerns Postnikov towers of existing objects. Essential surjectivity of \(T \to \widehat T\) requires a uniform local bound.
Proposition 6.76. ([Lurie 2009, Proposition 7.2.1.10 and Corollary 7.2.1.12])
Let \(T\) be a topos.
If \(T\) is locally of finite dimension, then for all \(X \in T\), the map \(X \iso \lim_n \tau_n X\) is an isomorphism. In particular, \(T\) is hypercomplete.
If \(T\) is locally of dimension \(\leq d\) for some \(d\), then \(T\) is Postnikov-complete.
Proof
References
- Jacob Lurie. Higher topos theory. Ann. Math. Stud. 170, Princeton, NJ: Princeton University Press. 2009.
- Dustin Clausen, Akhil Mathew. Hyperdescent and étale $K$-theory. Invent. Math., 225 (3), 981–1076. 2021.
- Elden Elmanto, Marc Hoyois, Ryomei Iwasa, Shane Kelly. Cdh descent, cdarc descent, and Milnor excision. Math. Ann., 379 (3--4), 1011–1045. 2021.