2
$\begingroup$

many books and lecture notes say that it is difficult to prove that the composite of two proper morphisms of rigid analytic varieties are also proper. That is, if two morphisms of rigid analytic varieties $f:X\to Y$ and $g:Y\to Z$ are proper, then their composite $g\circ f$ is also proper. However I could solve this by a straightforward calculation. I guess I've made some mistake somewhere but I don't know. Are there any mistakes? Here's the proof:

We can assume that $Z$ is an affinoid variety $Z = SpA $. Since $g:Y\to SpA$ is proper, there are two admissible affinoid coverings $\{ U_i\}$ and $\{ U'_i \} $ of $Y$ such that $ U_i \Subset_{SpA} U'_i$ for all i. Since properness is stable under base changes, $f^{-1}(U'_i)\to U'_i$ is also proper for each i. Hence we can also take two admissible affinoid coverings $\{V_{ij}\}$ and $\{V'_{ij}\}$ of $f^{-1}(U'_{i})$ such that $ V_{ij} \Subset_{U'_i} V'_{ij} $. Let $U_i=SpB_i$, $U'_i= SpB'_i$, $V_{ij} =SpC_{ij}$ and $V'_{ij} =SpC'_{ij}$ and fix i and j. To prove that $ g\circ f$ is proper, it suffices to show that $f^{-1}(U_i) \cap V_{ij} \Subset_{SpA} V'_{ij}$. Since $f^{-1}(U_i) \cap V_{ij}$ is an inverse image of the morphism $X\to X\times_{SpA}Y$, which is the base change of the diagonal morphism $Y\to Y\times_{SpA} Y$ (note that this is a closed immersion since $g$ is proper) by $X\times_{SpA}Y \to Y\times _{SpA} Y$. Since base changes of closed immersions are also closed immersions, hence separated, it follows that $f^{-1}(U_i) \cap V_{ij}$ is an affinoid space. Now, let $g_1,...,g_r\in B'_i$ be the affinoid generating system of $B'_i$ over $A$ such that $U_i \subset \{x\in U'_i ; \forall k, |g_k(x)|<1 \}$ and $h_1,...h_s\in C'_ij$ be that of $C'_ij$ over $B_i$ such that $ V_{ij}\subset\{x\in V'_{ij};\forall l, |h_l(x)|<1\}$.then,$ \{f^*(g_k),h_l\}$ is an affinoid generating system of $C'_{ij}$ over$A$ and \begin{equation} f^{-1}(U_i) \cap V_{ij}\subset \{ x \in V'_{ij}; \forall k\ and\ \forall l,|f^*(g_k)(x)|<1,\ |h_l(x)|<1\} \end{equation} which proves that $f^{-1}(U_i) \cap V_{ij}\Subset_{SpA} V'_{ij}$.

$\endgroup$
5
  • $\begingroup$ You write that $f^{-1}(U'_i) \rightarrow U'_i$ is proper but don't say what the $V_{ij}$ or the $V'_{ij}$ are covering ($U'_i$ or $U_i$?) and then write that $V_{ij}$ is relatively compact in $V'_{ij}$ relative to $U_i$. Probably this $U_i$ is meant to be $U'_i$ or else the $V_{ij}$ and $V'_{ij}$ cover $U_i$, but anyway this is where your error lies: in Kiehl's definition one has no control over the exact affinoids in the base over which the nested pair of covers is found, so there is no way to link the relatively compact covers made for the two maps. That blocks all "easy" attempts. $\endgroup$
    – nfdc23
    Commented Nov 14, 2016 at 10:36
  • $\begingroup$ thank you for your comment. I revised some of what you indicated. But I thought that $\{f^{-1}(U'_i)\cap V_{ij}\}_{i,j}$ and $\{ V'_{ij} \}_{i,j}$ were the two affinoid admissible coverings of $X$ which enjoyed the property. why there's no way to link the relatively compact covers made for two morphisms? $\endgroup$
    – Panna
    Commented Nov 14, 2016 at 10:50
  • $\begingroup$ In Kiehl's definition one cannot control the exact affinoids in the base relative to which the two covers are found -- one cannot first pass to random affinoid $Z$ and expect to find $\{U_i\}$ and $\{U'_i\}$ adapted to that choice of $Z$, and most fatally upon passing to some affinoid $Z$'s which admit such a pair of covers of $g^{-1}(Z)$ (as can be done) one definitely cannot expect that the $V_{ij}$ and $V'_{ij}$ can be found as asserted relative to the already-chosen $U'_i$. Kiehl's definition is "local on the base" in only a weak sense. There are only so many ways for me to say this. $\endgroup$
    – nfdc23
    Commented Nov 14, 2016 at 14:29
  • $\begingroup$ I understood that we had to take an admissible affinoid covering of $U_i$ again and this is the error. Thank you. $\endgroup$
    – Panna
    Commented Nov 14, 2016 at 15:24
  • $\begingroup$ @nfdc23: Why not writing an answer? $\endgroup$
    – HeinrichD
    Commented Nov 14, 2016 at 15:34

0

You must log in to answer this question.

Browse other questions tagged .