ALEX, DARIO. CLASSICAL HODGE THEORY (NOTES BY HIRO, EDITED BY DARIO) 2011 TALBOT WORKSHOP 1. Dario, Mixed Hodge Structures A mixed Hodge strcuture (MHS) is a triple pH, W, F q where (1) H is a Z-module of finite type (2) W is an increasing filtration on HQ : H bZ Q, called the weight filtration (3) F is a decreasing filtraton on HC : H b C, called the Hodge filtration such that GrFp GrFq pGrnW pHC qq 0 ifp q n. Here we are extending W to be a filtration of HC . Notice that this formula makes sense due to the following observation: Subobjects/quotients of filtered objects are again filtered. E.g., if A is filtered, then the filtration of a quotient A{B is defined to be F k pA{B q F k pAq{pB X F k pAqq. The most boring examples of MHSs are Split MHSs., obtained at follows: H `k Hk , where Hk is a pure Hodge structure of weight k, with obvious W and F . Clearly, a MHS is split if and only if W is a grading of HQ , instead of being a filtration. Some natural questions, the first of which is the more urgent: (1) Why did I make this definition? (2) Can we make this more geometric? (3) Can I assemble mixed hodge structures into a category?—in fact they form a symmetric monoidal category. And is the symm monoidal cat. of MHS a representation of a group? The answer is yes; this would take us to Tannakian duality. Let’s answer the first question: (1) Mixed Hodge Structures appear! For instance, studying cohomology of complex algebraic varieties. I’ll consider only the non-singular case. So let’s say that X is a non-singular complex variety, possibly non-compact. We know very well what happens in the compact case, namely each cohomology H n pX, Zq is a pure HS of weight n. In the non-compact case, it turns out that, for each degree n, H n pX; Zq carries a mixed Hodge structure. 1 2 2011 TALBOT WORKSHOP The idea is to compactify X, you can do it in a particularly nice way using some algebraic geometry—namely, there exists some smooth compact algebraic variety X containing X, such that if I define the complement Y : X zX, Y is a normal crossing divisor. To fix notation, let me recall that the inclusion Y ãÑ X in analytic topology locally looks like the inclusion of some coordainate axes tz1 zr 0u into Cn tpz1 , . . . zn qu. Now it turns out I can relate this to the cohomology of the compact thing. Fact: Let’s be less ambitious for now and say complex coefficients. H n pX; Cq is isomorphic to the hypercohomology of X with coefficients in a complex of sheaves H n pX, Cq H ppX q, Ω pY qX q. and this has a name—the logarithmic deRham complex. What is the definition fo this? It is a locally free sheaf generated by holomoprhic forms on X, together with forms that locally look like dzi {zi . That’s why it’s called logarithmic. Carlos: This is not just sheaves related to divisor associated to Y —for instance, dz1 {z2 is not allowed. xY y ΩX is generated by ΩX and dzi {zi . This guy comes with two filtrations. The first is the obvious filtration: F p is equal to k-forms with k The second is increasing : it is defined as Wn pΩX xY y tforms with ¤ n occurrences of ¥ p. dzi {zi u. So we have to compute hypercohomology: the coefficient here is a complex that carries two filtrations. Hence for each of these filtrations, I can write down a spectral sequence for the cohomology of a filtered complex. So we have two ways to compute the same thing and then we can compare! Mike: I thought the F p was the number of dzs showing up. Ans: We’re doing something holmoprhic, so only dz shows up. p,q F E1 H 1pX, ΩpX xY yq Ñ H p q pX, Cq After playing with the spectral sequence associated to W , we get the Leray spectral sequence: p,q p q p q pX, Cq. W E2 H pX, R j C q Ñ H where j : X Ñ X. So, H n pX, Cq comes with two induced filtrations, that I still call F (decreasing) and W (increasing). However, since the Leray spectral sequence is topological, W is actually a filtration on the rational cohomology! Theorem 1 (Deligne). H n pX, Zq, W rns , F q is a mixed Hodge Structure. Moreover this is independent of thoice of compactification, and it’s functorial in X.