1
$\begingroup$

I am looking for a reference to the following fact: monodromy group acting on the cohomology of smooth hyperplane sections of a smooth projective variety $X$ over $\mathbb C$ is preserved under isomorphic projection.

To be precise, suppose that $X\subset \mathbb P^N$ is a smooth projective variety of dimension $n+1$ and $Y$ is its smooth hyperplane section. As $Y$ varies in the class of smooth hyperplane sections of $X$, a monodromy action on $H^n(Y,\mathbb Z)$ arises; by monodromy group of $X$ I will mean the corresponding subgroup of $\mathrm{Aut}(H^n(Y,\mathbb Z))$.

Suppose now that there exists a pont $p\in\mathbb P^N\setminus Y$ such that the projection $\pi_p\colon X\to X'\subset\mathbb P^{N-1}$ is an isomorphism. Do you know a reference to the effect that this monodromy group of $X'$ is the same as that of $X$?

Thank you in advance,
Serge

$\endgroup$
7
  • $\begingroup$ This follows from the assertion that for the dual projective space $\widehat{\mathbb{P}}^N$, for the dual variety $\widehat{X}\subset \widehat{\mathbb{P}}^N$ (typically a hypersurface unless $X$ is a linear subspace ...), for the singular locus $B\subset \widehat{X}$, for every line $\ell \subset \widehat{\mathbb{P}^N}$ that is disjoint from $B$ and that intersects $\widehat{X}$ transversally, the homomorphism $\pi_1(\ell\setminus \ell\cap \widehat{X}) \to \pi_1(\widehat{\mathbb{P}}^N\setminus \widehat{X})$ is surjective. This is straightforward to prove . . . $\endgroup$ Mar 29, 2017 at 9:33
  • $\begingroup$ . . . The fundamental group of $\pi_1(\ell\setminus \ell\cap \widehat{X})$ is a free group on the generators arising from intersection points $\widehat{X}$ with a single relation equal to the product of these generators (with appropriate orientation). Thus a homomorphism from this fundamental group to a group $G$ is trivial if and only if every generator maps to the identity. For a homomorphism from the fundamental group of $\widehat{\mathbb{P}}^N\setminus \widehat{X}$ to $G$, the restriction as above is trivial if and only if there is trivial "ramification" along $\widehat{X}$ . . . $\endgroup$ Mar 29, 2017 at 9:38
  • $\begingroup$ . . . In that case, by Seifert-van Kampen, the homomorphism is the restriction of a homomorphism from the fundamental group $\pi_1(\widehat{\mathbb{P}}^N\setminus B)$ to $G$. Since $B$ has complex codimension $2$, thus real codimension $4$, this equals $\pi(\widehat{\mathbb{P}}^N)$, which is trivial. Hmm, now I realize that this only proves that the image of $\pi_1(\ell \setminus \ell\cap \widehat{X})$ is contained in no proper normal subgroup of $\pi_1(\widehat{\mathbb{P}}^N\setminus \widehat{X})$ . . . $\endgroup$ Mar 29, 2017 at 9:44
  • $\begingroup$ The argument above can be corrected. Fix a general point $p$ of $\widehat{\mathbb{P}}^N$. Let $U\subset \widehat{\mathbb{P}}^N\setminus (\widehat{X}\cap\{p\})$ be the open subset of all points $q$ such that line $\ell$ spanned by $p$ and $q$ satisfies the transversality conditions. This is a connected open subset whose complement is a proper complex analytic subvariety of complex codimension $1$, real codimension $2$. As $q$ varies, the images in $\pi_1(\widehat{\mathbb{P}}^n\setminus\widehat{X})$ are all equal, by continuity and connectedness. . . . $\endgroup$ Mar 29, 2017 at 9:52
  • $\begingroup$ . . . Consider the covering space of $\widehat{\mathbb{P}}^n\setminus \widehat{X}$ obtained from the coset space of this common subgroup $\pi_1(\ell \setminus \ell\cap \widehat{X},p)$ inside $\pi_1(\widehat{\mathbb{P}}^N\setminus \widehat{X},p)$. Restricted over every line $\ell$, this covering has a unique cross section coming from the neutral coset. As $q$ varies, the union of these cross sections defines a cross section of the covering space when restricted over $U$. Now use that the complement of $U$ is an analytic subvariety of real codimension $2$ to see that the cross section extends. $\endgroup$ Mar 29, 2017 at 9:56

1 Answer 1

2
$\begingroup$

The comments above are too long, so I am replacing them by an answer. The reference for most such results in algebraic geometry is SGA 7_2. I am sure that the following is in there somewhere.

Your assertion follows from a more general statement. Let $X\subset \mathbb{P}^N$ be a smooth projective variety. Let $\widehat{\mathbb{P}}^N$ denote the dual projective space parameterizing hyperplanes in $\mathbb{P}^N$. Let $\widehat{X}\subset \mathbb{P}^N$ be the dual variety of $X$, i.e., the image of the natural map to $\widehat{\mathbb{P}}^N$ from the projectivized normal bundle of $X$ in $\mathbb{P}^N$, $\mathbb{P}_X \mathcal{N}_{X/\mathbb{P}^N}$. Since the projectivized normal bundle has dimension $N-1$, the image $\widehat{X}$ is a proper subvariety; typically it is a hypersurface. Denote by $B\subset \widehat{X}$ the closed locus over which the morphism $\mathbb{P}_X\mathcal{N}_{X/\mathbb{P}^N} \to \widehat{X}$ is not an isomorphism. In characteristic $0$, always $B$ has codimension $\geq 2$ in $\widehat{\mathbb{P}}^N$. In characteristic $p$, there are sufficient conditions to assure this, but usually people just re-embed $X$ in a bigger projective space by the $d$-uple Veronese map for $d\geq 2$; then $B$ will have codimension $\geq 2$.

A Lefschetz pencil is a line $\ell \in \widehat{\mathbb{P}}^N$ that is disjoint from the codimension $2$ subset $B$ and that intersects $\widehat{X}$ (tangentially) transversally at every intersection point. When $B$ has codimension $2$, Lefschetz pencils exist (by the same sort of parameter count that establishes Bertini's smoothness theorem). All of the above is in SGA 7_2.

The assertion is that for every Lefschetz pencil $\ell$, the induced homomorphism $$f_{X,\ell}:\pi_1(\ell\setminus \ell\cap \widehat{X}) \to \pi_1(\widehat{\mathbb{P}}^N\setminus \widehat{X})$$ is surjective. This proves your assertion as follows. For a point $p$ such that the linear projection from $p$ restricts to an isomorphism from $X$ to its image $Y$ in $\mathbb{P}^{N-1}$, the corresponding hyperplane $H_p\subset \widehat{\mathbb{P}}^N$ equals $\widehat{\mathbb{P}}^{N-1}$ and $H_p\cap \widehat{X}$ equals $\widehat{Y}$. In characteristic $0$ at least, existence of a Lefschetz pencil implies that there exists $\ell\in H_p$ that is a Lefschetz pencil. Thus, we have a composition of group homomorphisms, $$\pi_1(\ell \setminus \ell \cap \widehat{X}) \xrightarrow{f_{Y,\ell}} \pi_1(H_p\setminus H_p\cap \widehat{X}) \xrightarrow{f_{X,Y}} \pi_1(\widehat{\mathbb{P}}^N\setminus \widehat{X}).$$ Since the composition $f_{X,Y}\circ f_{Y,\ell}$ is surjective, also $f_{X,Y}$ is surjective. Thus, for every group homomorphism $\phi:\pi_1(\widehat{\mathbb{P}}^N\setminus \widehat{X}) \to G$, the image of $\phi$ equals the image of $\phi\circ f_{X,Y}$.

Thus it suffices to prove that for every Lefschetz pencil $\ell$, $f_{X,\ell}$ is surjective. Fix a point $p$ of $\ell$. Denote by $U\subset \widehat{\mathbb{P}}^N\setminus (\widehat{X}\cup \{p\})$ the set of points $q$ such that line $\ell_q$ spanned by $p$ and $q$ is a Lefschetz pencil. The transversality conditions in the definition of a Lefschetz pencil are open conditions. Thus, $U$ is an open subset. Since for every $q\in \ell\setminus\{p\}$, $q$ is in $U$, $U$ is a nonempty open subset. Thus, $U$ is a dense open subset of $\widehat{\mathbb{P}}^N\setminus (\widehat{X}\cup \{p\})$. In particular, $U$ is connected. Thus, the images of the homomorphisms $f_{X,\ell_q}$ for $q\in U$ are all equal to a common subgroup $\Gamma$ of $\pi_1(\widehat{\mathbb{P}}^N\setminus \widehat{X},p)$.

Let $\rho:Y\to \widehat{\mathbb{P}}^N\setminus \widehat{X}$ be the covering space associated to the standard left action of $\pi_1(\widehat{\mathbb{P}}^N\setminus \widehat{X},p)$ on the set of right $\Gamma$-cosets. The restriction of this left action to the subgroup $\Gamma$ has a fixed point: the neutral coset $\Gamma$. Thus, the restriction of $\rho$ over every line $\ell_q\setminus \ell_q\cap \widehat{X}$ has a cross section. The union of all of these cross sections defines a cross section of the restriction of $\rho$ over the open dense subset $U$. For every point $r\in \widehat{\mathbb{P}}^N\setminus \widehat{X}$, there exists a neighborhood $B_\epsilon(r)$ of $r$ in this open over which $\rho$ is trivialized. Since the complement of $U$ is an analytic subvariety of complex codimension $1$, real codimension $2$, $B_\epsilon(r)\cap U$ is nonempty and connected. Thus, the cross section of $\rho$ over $U$ extends uniquely to a cross section over all of $B_\epsilon(r)$. Since this holds for every $r$, the cross section extends over all of $\widehat{\mathbb{P}}^N\setminus \widehat{X}$. It follows that the neutral coset $\Gamma$ is a fixed point for all of $\pi_1(\widehat{\mathbb{P}}^N\setminus \widehat{X},p)$. Thus, $\Gamma$ is the entire group.

$\endgroup$
1
  • $\begingroup$ Dear Jason, thank you for your time. Most certainly this assertion is absent from SGA7_2, otherwise I would not have asked for a reference:) Anyway, I guess one may safely assume that the required reference does not exist. Thank you again. $\endgroup$ Mar 29, 2017 at 16:05

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.