$\DeclareMathOperator\Spa{Spa}$Fargues's Theorem for $\Spa(C,O_C)$ states that the category of (mixed characteristic) shtukas with one paw at $x_C$ is equivalent to the category of Breuil-Kisin-Fargues modules over $A_{inf}=W(O_C)$. (Here, $C$ is an algebraically closed perfectoid field, and $O_C$ is the ring of integers of $C$.)

My question is concerned with extending this theorem to shtukas over $\Spa(C,C^+)$, where $C^+$ is an open and bounded valuation subring of $C$ and BKF-modules over $W(C^+)=A_{inf}(C,C^+)$ (and $C$ is still an algebraically closed perfectoid field).

First question: Where in the proof of Fargues's theorem does this slighly bigger generality fail?

A sketch of the proof of Fargues's theorem goes as follows:

Step 1: Turn a shtuka over $\mathcal{Y}_{[0,\infty)}$ to a shtuka over $\mathcal{Y}_{[0,\infty]}$. This step is done by glueing the shtuka with a "spread" vector bundle over the Fargues–Fontaine (that is the corresponding $\phi$-module over $\mathcal{Y}_{[r,\infty]}$ for some $r$). Here we use that we understand vector bundles over $\mathrm{FF}_{(C,O_C)}$ to be able to extend over $\infty$. Now theorem 8.7.7 in Kedlaya-Liu ("Relative $p$-adic Hodge Theory: Foundations" (MSN,arxiv)), tell us that vector bundles on $\mathrm{FF}_{(C,C^+)}$ coincide with those of $\mathrm{FF}_{(C,O_C)}$. That's why I think this step can be done as well.

Step 2: Turning an "analytic" shtuka over $\mathcal{Y}_{[0,\infty]}$ into an "algebraic" shtuka over $Y_{[0,\infty]}$. This is the content of Kedlaya's theorem 4.5.10 a) in his AWS (revised) notes, and that step is stated in full generality for any perfectoid Huber pair $(R,R^+)$.

Step 3: Turning the "algebraic" shtuka over $\operatorname{Spec}(A_{inf})\setminus \{\varpi=0=p\}$ into a BKF-module over $\operatorname{Spec}(A_{inf})$. This is now Kedlaya's theorem 4.5.10 b) of his AWS notes again. It works for any perfectoid field $K$ regardless of the chosen $K^+$.

All steps seem to work to me for $\Spa(C,C^+)$, but I might be missing a subtlety and I would like to know what that is. Step 1 is the one I would have most doubts about.


1 Answer 1


Step 1 indeed fails. The functor from BKF-modules over $C^+$ to shtukas over $C^+$ is not going to be full. You can still glue so you still get an essentially surjective functor but there could be more interesting $\varphi$-modules over $\mathcal{Y}_{(0,\infty]}$ by which you could attempt to glue. If $k=O_C/\mathfrak{m}$ and $k^+=C^+/\mathfrak{m}$ then the information lost is similar to the information that you lose by passing from a $\varphi$-module over $W(k^+)$ to a $\varphi$-module over $W(k)$, certain things that weren't isomorphic become isomorphic.

  • $\begingroup$ Sorry I didn't post this earlier. $\endgroup$ Commented Apr 30, 2021 at 5:32

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.