# Monodromy with algebraic fundamental groups

### Topological setting

Say we have a fiber bundle: $p: X \rightarrow B$. Let $s: B \rightarrow X$ be a section. Then $\pi_1(B,b)$ acts on $\pi_1(F_b,s(b))$ (where $F_b:=p^{-1}(b)$) by: Let $\gamma \in \pi_1(B,b)$ and $\delta \in \pi_1(F_b,s(b))$. Deform $\delta$ above $\gamma$ in such a way that above $\gamma(t)$, $\delta(t)$ will be a loop in $p^{-1}(\gamma(t))$ with base point $s(\gamma(t))$. At $t=1$ we will arrive at some new loop in $\pi_1(F_b,s(b))$. This defines the action of $\gamma$ on $\delta$.

### Question

Is there a way to define this algebraically? Meaning: Let $p:X \rightarrow B$ be a map of integral schemes (varieties if you prefer?). Assume it is flat and with isomorphic fibers. Let $s:B \rightarrow X$ be a section. Can we then define, similarly, a monodromy action of $\pi_1(B,b)$ ($b$ a geo. point) on $\pi_1(X,s(b))$?

And maybe my setting is off. What is the right setting for this (algebraically), and how would you define this action?

-
I don't have time to write a full answer, but here a few comments (1) there is no way to capture the topological fundamental group in a purely algebraic fashion; the best you can hope for is some approximation. (2) For the etale fundamental group, see SGA1 exp X cor 1.4. Then you'll need to a work a bit from there to the analogue you want. – Donu Arapura Oct 23 '10 at 21:42
It is relatively rare for a bundle to have a section (then e.g. the long exact sequence of the homotopy groups splits etc). But the monodromy action exists for each locally trivial fiber bundle and more generally for a Serre bundle, since all one ever uses in the above definition is the covering homotopy property. More precisely, for a Serre bundle the action is defined up to homotopy, but for a locally trivial bundle with a connection say we get a genuine action by homeomorphisms. However, this action is seldom by algebraic maps: think e.g. of the universal level $N>2$ elliptic curve. – algori Oct 24 '10 at 15:02

Let $p:X \to B$ be a topological fiber bundle with fiber $F$, such that $F$, $X$, and $B$ are all connected and tame (say, CW complexes). Then there is a homotopy exact sequence $$\pi_2(B) \to \pi_1(F) \to \pi_1(X) \to \pi_1(B) \to 1.$$ Let $N$ be the quotient of $\pi_1(F)$ by the image of $\pi_2(X)$. Then $N$ is also a subgroup of $\pi_1(X)$, and $$\pi_1(X)/N \cong \pi_1(B).$$ So $\pi_1(B)$ acts on $N$ because a quotient group always acts by conjugation on the kernel.
If $X$ has a global section as a bundle over $B$, this implies that the term $\pi_2(B) \to \pi_1(F)$ is trivial (because a sphere in $\pi_2(B)$ then lifts to a disk mapped to $X$ whose boundary is a constant loop in $F$). If this term is trivial, whether or not $X$ has a global section, then $\pi_1(F) = N$ and so $\pi_1(B)$ acts on $\pi_1(F)$.
Now, suppose that $p:X \to B$ is a map of algebraic varieties and $F$ is some fiber over a closed point. Then the étale fundamental group $\tilde{\pi}_1$ is a functor on the relevant category. So you at least get a sequence $$\tilde{\pi}_1(F) \to \tilde{\pi}_1(X) \to \tilde{\pi}_1(B) \to 1.$$ Then maybe the composition of the first two maps is trivial because $F$ is sent to a closed point. For a general $p$ (taking simple examples from topology), this sequence is not an exact sequence. But, if there are suitable conditions on $p$ for it to be analogous to a fiber bundle, then maybe it should be exact. Note that the étale fundamental group $\tilde{\pi}_1(V)$ is the profinite completion of the analytic fundamental group $\pi_1(V)$ when $V$ is a variety over $\mathbb{C}$. On the negative side, profinite completion is not an exact functor on groups. (But what about for fundamental groups of complex algebraic varieties?)
If it is exact, then I'm not sure whether the kernel in $\tilde{\pi}_1(F)$ can be naturally related to anything like an étale $\tilde{\pi}_2(B)$, but maybe so.