I am not a specialist in singularity theory but currently I have to touch resolution of singularities and I'd like to know whether I have understood Hironaka's theorem correctly.

Let $k$ be a field of characteristic zero and $X$ an $k$-variety and $Z$ a closed subscheme of $X$. Then Hironaka's theorem claims that there exists a log-resolution $h: Y \longrightarrow X$ of the pair $(X,Z)$, by this I mean:

  • $Y$ is a smooth $k$-variety.
  • $h$ is projective, proper birational morphism.
  • $h$ induces an isomorphism outside $X_{\mathrm{sing}} \cup Z$.
  • If $E$ is the exceptional locus then $h^{-1}(Z) + E$ has strict normal crossings.

and $h$ itself can be taken to be a composition of successive blowing-ups along smooth centers.

I am interested in the following situation: let's assume we still have $(X,k)$ as above and we are given a morphism $f: X \longrightarrow \mathbb{A}^1_k$. Let $U$ be any smooth open subscheme of $X$ and $F = X \setminus U$, hence $X_{\mathrm{sing}} \subset F$ so we could take a log-resolution of the pair $(X,F \cup f^{-1}(0))$ (one may assume $f^{-1}(0)$ is nowhere dense if necessary).

Let $E_i$ ($i \in I$) be the smooth irreducible components of $h^{-1}(F \cup f^{-1}(0))$ with corresponding multiplicities $N_i \neq $ ($i \in I$) , i.e. $$h^{-1}(F \cup f^{-1}(0)) = \sum_{i \in I} N_i E_i.$$ What I want to know is can we control the multiplicities in the way that there exists $C \subset I$ such that $$h^{-1}(f^{-1}(0)) = \sum_{i \in C} N_i E_i \ \ \ \text{and} \ \ \ h^{-1}(F) = \sum_{i \in I \setminus C} N_i E_i.$$ If it is the case, then does restricting to $U$ (which gives a new resolution since blowing-ups commute with flat base-changes) lose some multiplicities? $\require{AMScd}$ \begin{CD} Y_{\mid U} @>{}>> X\\ @VVV @VVhV\\ (U,f^{-1}(0) \cap U) @>{}>> (X,F \cup f^{-1}(0)) \end{CD}


1 Answer 1


This has nothing to do with your morphism to $\mathbb A^1$. You are asking if there is another closed subset $F'$ (which is secretly equal to $h^{-1}(0)$), then can you separate the exceptional divisors according to the pre-images of the two sets $F$ and $F'$. This depends on how $F$ and $F'$ intersect with respect to the resolution. If none of the centers of the blow ups in the given resolution is contained in $F\cap F'$, then this holds. Then $C=\{i\in I\vert h(E_i)\subseteq F'\}$ (and then, of course, $I\setminus C=\{i\in I\vert h(E_i)\subseteq F\}$). In this case, when you restrict to $U$, you do not "lose" any multiplicities. That could only happen if you lose one of the $E_i$, but that would only happen if $h(E_i)\cap U=\emptyset$, but that would mean that $h(E_i)\subseteq F\cap F'$, which we assumed does not happen.

The problem is that if $h(E_i)\subseteq F\cap F'$, then that particular $E_i$ appears in both $h^{-1}(F)$ and $h^{-1}(F')$. This happens for sure if for instance $F\subseteq F'$.

  • $\begingroup$ So according to what you wrote, we may lose $N_i$ if we take base-change along an arbitrary open immersion? $\endgroup$
    – Alexey Do
    Commented Jan 5, 2023 at 21:09
  • $\begingroup$ Yes, but I would rather say that you lose the $E_i$, not the $N_i$. The latter is a consequence of the former. $\endgroup$ Commented Jan 6, 2023 at 4:50

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.