The specialization map $sp:\mathfrak{X}_\eta\to \mathfrak{X}_{red}$ has an important role in rigid analytic geometry. I tried looking in Huber's papers ("Continuous Valuations", "A generalization of formal schemes and rigid analytic varieties", and "Etale cohomology of Rigid Analytic Varieties and Adic Spaces") browsing for a treatment of this map and I didn't find much.

Ideally, I would like to find a statement like the following:

Proposition: Given a Tate Huber-pair $(A,A^+)$ consider the topological space $Spa(A,A^+)$. Then there is a map of topological spaces $sp: Spa(A,A^+)\to Spf(A^+)$, this map is a spectral map of spectral spaces and it is a closed map.

I know that in a traditional Noetherian formal scheme context (or probably even topologically finite type over a valuation ring context) this proposition is a consequence of interpreting the $Spa(A,A^+)$ as the limit of admissible blow-ups, but I was wondering if this was known and written in a non-Noetherian context like perfectoid spaces or even in a non-sheafy context.

Any reference to the specialization map in a Huber-like approach will be much appreciated.

  • 1
    $\begingroup$ What do you mean by $\operatorname{Spf}(A^+)$? If $A^{+}$ isn't bounded (for example, $A^+=A^0$ and $A$ is not uniform), then the topological ring $A^+$ is not an adic ring (and, moreover, I don't see any reason for this to be a pre-admissible ring). So there is no much sense to speak about $\operatorname{Spf}(A^+)$. If you assume that $A^+$ is adic, then there is always a map $\operatorname{Spa}(A^+,A^+) \to \operatorname{Spf}(A^+)$. And the composition map $\operatorname{Spa}(A, A^+) \to \operatorname{Spa}(A^+, A^+) \to \operatorname{Spf}(A^+)$ is the desired specialization map. $\endgroup$
    – gdb
    Jul 26, 2018 at 23:27

1 Answer 1


In case you're still interested: Bhatt recently proved that for any Tate-Huber pair $(A,A^+)$, the topological space $\mathrm{Spa}(A,A^+)$ is homeomorphic to an inverse limit of admissible blowups in the expected manner (Theorem 8.1.2 here). For fun I'll summarize the construction (which probably won't seem too surprising). Fix a topologically nilpotent unit $\varpi \in A^+$, and let $I$ denote the category of proper birational maps of schemes $f_i:X_i \to \mathrm{Spec}(A^+)$ which restrict to isomorphisms over the open subset $\mathrm{Spec}(A) \subset \mathrm{Spec}(A^+)$. This is cofiltered, because any $X_i \times_{\mathrm{Spec}(A^+)}X_{i'}\to \mathrm{Spec}(A^+)$ is still in $I$. Let $\overline{X_i} \subset X_i$ denote the vanishing loci of $\varpi$, so these are also cofiltered, and they all map compatibly to $\mathrm{Spec}(A^+/\varpi)$.

Now, there is a canonical map $\Phi: \mathrm{Spa}(A,A^+) \to \lim_{\leftarrow} \overline{X_i}$ given by the following recipe: for any point $x \in \mathrm{Spa}(A,A^+)$, let $K_x$ and $K_x^+$ be the associated residue field and valuation subring, respectively, so there are canonical maps $A \to K_x$ and $A^+ \to K_x^+$. We then get canonical compatible maps $\mathrm{Spec}(K_x) \to X_i$ for all $i$, simply because the loci in the $X_i$'s where $\varpi \neq 0$ all identify with $\mathrm{Spec}(A)$ by definition, so we can lift the given map $\mathrm{Spec}(K_x)\to \mathrm{Spec}(A)$ uniquely. But we also have a map $\mathrm{Spec}(K_x^+) \to \mathrm{Spec}(A^+)$, so by the valuative criterion of properness this lifts uniquely along each $f_i$, giving a compatible system of maps $r_i : \mathrm{Spec}(K_x^+) \to X_i$. Applying the $r_i$'s to the unique closed point of $\mathrm{Spec}(K_x^+)$ then gives the desired point in $\lim \overline{X_i}$.

Anyway, Bhatt proves that $\Phi$ is always a homeomorphism. The specialization map $\mathrm{sp}$ you want is just the composition of $\Phi$ with the natural map $\lim \overline{X_i} \to \mathrm{Spec}(A^+/\varpi)$. Of course, you don't need to go through all of this to define the specialization map, but it does seem to be the easiest way to prove the following:

Fact: The map $\mathrm{sp}$ is continuous, spectral, and closed.

Proof sketch: Continuity and spectrality follow from Bhatt's result together with some basic nonsense about limits of spectral spaces. Closedness can be deduced from the following useful criterion: any quasicompact, spectral, specializing map of locally spectral spaces is closed.


Your Answer

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