TITLE: The topology of Fano schemes of lines
QUESTION [6 upvotes]: Is there any references concerning the computation of the fundamental groups and Hodge numbers of Fano schemes of lines in a smooth hypersurface in $\mathbb{P}^n$?

REPLY [9 votes]: As Daniel Loughran points out, my claim regarding vanishing of the fundamental group is wrong.  In all of the following, $F$ is the Fano scheme parameterizing lines on a degree $d$ hypersurface $X$ in $\mathbb{P}^n$.  I am assuming that $F$ is smooth of the expected dimension $2n-d-3$.  Here is what I can prove.
(1) When $d$ equals $2n-4$, except for $n=3$ and $d=2$, the Fano scheme of lines is a smooth, projective, geometrically connected curve, cf. Theorem V.4.3 of Kollár's "Rational Curves on Algebraic Varieties".  In this case, there is an enumerative computation for the genus $g$ of this curve: $2g-2 = a_{n-1,n-1}-a_{n-2,n}$ where
$$
(n+1-d(d+1)/2)(x+y)\cdot \prod_{e=0}^d ((d-e)x+ey) = \sum_{l=0}^{2n-2} a_{l,2n-2-l}x^ly^{2n-2-l}.
$$
(2) When $d\leq n-4$, then $F$ is simply connected.  Since the fundamental group of smooth, projective schemes is stable under specialization, we may assume that $X$ is sufficiently general.  Denote by $$\rho:\mathcal{C}\to F, \ \ \text{ev}:\mathcal{C} \to X,$$
the universal curve over $F$.  Since $\mathcal{C}$ is a $\mathbb{P}^1$-bundle over $\mathcal{C}$, the induced map of fundamental groups,$$\pi_1(\mathcal{C}) \to \pi_1(F),$$ is an isomorphism.  Thus, it suffices to prove that $\mathcal{C}$ is simply connected.
Since $X$ is sufficiently general, $\text{ev}$ is flat of relative dimension $n-d-1$, the generic fiber of $\text{ev}$ is smooth, and, over codimension $1$ points of $X$, the singular fibers of $\text{ev}$ have only ordinary double point singularities (this follows by considering the incidence correspondence of all hypersurfaces $X$ and lines contained in those hypersurfaces).  Denote by $B\subset X$ the subvariety of codimension $\geq 2$ where the fibers of $\text{ev}$ have worse than ordinary double points, and denote by $U\subset X$ the open complement.  Since codimension $2$ subvarieties do not affect the fundamental group, it suffices to prove that $\text{ev}^{-1}(U)$ is simply connected.
The fibers of $\text{ev}$ are naturally complete intersections of ample divisors in projective space.  Since the dimension is $\geq 3$, by the usual Lefschetz hyperplane theorem (for singular complete intersections), the fibers of $\text{ev}$ over $U$ are simply connected.  Probably, by analyzing more carefully the singularities, one can get the same result for $d=n-3$.  Because the fibers of $\text{ev}$ over $U$ are simply connected, and because $U$ is simply connected, it follows that $\text{ev}^{-1}(U)$ is simply connected. 
I suspect that $F$ is simply connected in a larger degree range.
Edit. By the way, $\mathcal{C}$ is an iterated intersection of $d+1$ Cartier divisors in the projectivized tangent bundle, $\mathbb{P}T_{\mathbb{P}^n}$.  This gives an approach to compute Hodge numbers via residues.  Since Hodge numbers are invariant under smooth deformations (in characteristic $0$), we can perform the computation for a general complete intersection (not just the special iterated intersections arising as universal curves over Fano schemes).  
Second Edit. To prove (algebraic) simple connectedness of $\mathcal{C}$, it suffices to prove that the higher cohomology of $\mathcal{O}_{\mathcal{C}}$ is zero.  Since the coherent sheaf $\mathcal{O}_{\mathcal{C}}$ has a Koszul resolution in the projective homogeneous space $\mathbb{P}T_{\mathbb{P}^n}$ where the locally free sheaves in the complex have filtrations with invertible subquotients, Bott vanishing will imply the vanishing of the higher cohomology of $\mathcal{O}_{\mathcal{C}}$ in an appropriate degree range.  Probably this is a broader degree range than $d\leq n-4$.  
I just checked, and the approach via Bott vanishing does not work.  The problem is that there is nonvanishing higher cohomology of the structure sheaf whenever $d(d+1)/2$ is at least $n+1$; just apply the Hodge symmetries and observe that $h^0(F,\omega_F)$ is nonzero.  This gives a worse degree range than the computation above when $d\leq n-4$.  
Third Edit.  I realized that the Lefschetz hyperplane theorem from (2) above applies whenever $d\leq n-3$, not just when $d\leq n-4$.  The ambient scheme must have depth $\geq 3$, but the ample divisor may have depth $2$.  Based on examples, I strongly suspect that also $F$ is simply connected when $d$ equals $n-2$.
When $d\geq n-1$, then there is an enumerative approach to proving non-simple connectedness.  For $S$ a smooth complete intersection in $F$, or in $\mathcal{C}$, of ample divisors, if $S$ is a surface, then $S$ is simply connected if and only if $F$ is simply connected.  Now, if $\chi(S,\mathcal{O}_S) = (c_1(T_S)^2+c_2(T_S))/12$ is negative, then $S$ is not simply connected.  Via "adjunction", we can reduce the computation of these Chern numbers to an intersection theory computation on $F$ or $\mathcal{C}$.