Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Let $X$ be a projective, smooth algebraic variety defined over the field of algebraic numbers. Consider algebraic de Rham cohomology $H_{dR}(X)$ and singular cohomology of $X(\mathbb{C})$ with rational coefficients $H_B(X)$. One can look at the class in $\mathbb{C}^\ast/\overline{\mathbb{Q}}^\ast$ of the determinant of the period isomorphism

$H^k_{dR}(X) \otimes_{\overline{\mathbb{Q}}} \mathbb{C} \stackrel{\sim}{\longrightarrow} H_B(X) \otimes_{\mathbb{Q}} \mathbb{C}$

In his paper "Operads and motives in deformation quantization", p. 63 Kontsevich claims that it is a power of $2\pi i$. More precisely that, for any basis, the determinant has the form $\pm \sqrt{a} (2\pi i)^n$ for $a \in \mathbb{Q}^\ast$ and $n\geq 1$. Does anyobody know how to prove this result?

share|improve this question

1 Answer 1

The argument is a series of more and more general cases.

(1) The map $H_{DR}^1(\mathbb{C}^{\ast}) \otimes \mathbb{C} \to H_B^1(\mathbb{C}^{\ast}) \otimes \mathbb{C}$ is multiplication by $(2 \pi i) \mathbb{Q}$. Proof: $H^1_{DR}$ is generated by $dz/z$. Let $\gamma$ be a closed path in $\mathbb{C}^{\ast}$ which circles the origin once, then $H^1_{B}$ is generated by $[\gamma] \mapsto 1$. Since $\int_{\gamma} dz/z= 2 \pi i$, we see that $H^1_{DR} \otimes \mathbb{C} \to H^1_B \otimes \mathbb{C}$ is multiplication by $2 \pi i$ with respect to the given generators.

(2) $H_{DR}^j(\mathbb{C}^k \times (\mathbb{C}^{\ast})^{\ell}) \otimes \mathbb{C} \to H_B^j(\mathbb{C}^k \times (\mathbb{C}^{\ast})^{\ell}) \otimes \mathbb{C}$ is multiplication by $(2 \pi i)^j \mathbb{Q}$. Proof: The Kunneth isomorphism is compatible with everything in the picture.

(3) $H^{2j}_{DR}(\mathbb{P}^n) \otimes \mathbb{C} \to H^{2j}_{B}(\mathbb{P}^n) \otimes \mathbb{C}$ is multiplication by $(2 \pi i)^j \mathbb{Q}$. Proof: Write down an open cover of $\mathbb{P}^n$ where every intersection of open sets is of the form (2). We get corresponding spectral sequence in $H_{DR}$ and $H_B$, and compatible maps between them. Keep track of degrees to get the correct power of $2 \pi i$.

(4) Let $X$ be smooth projective of complex dimension $n$. Then $H^{2n}_{DR}(X) \to H^{2n}_B(X)$ is multiplication by $(2 \pi i)^n \mathbb{Q}$. Proof: Write $X$ as a finite cover of $\mathbb{P}^n$ and draw the corresponding commuting square in $H^{2n}$.

(5) (The one that involves determinants.) Let $X$ be smooth projective of dimension $n$. Let $\zeta_B$ be the singular cohomology class of a hyperplane and let $\zeta_{DR}$ be the deRham class $(2 \pi i) \zeta_B$. Fix a cohomological degree $k \leq n$; the claim for $k$ larger than $n$ follows by Poincare duality.

Hard Lefschetz tells us that $$(\alpha, \beta) \mapsto \alpha \cup \beta \cup \zeta_T^{n-k}$$ is a perfect pairing $H^k_T \times H^k_T \to H^{2n}_T$, where $T$ is either $DR$ or $B$. We instead write this as a pairing $H^k_T \times H^k_T \to \mathbb{Q}$ given by $$\langle \alpha, \beta \rangle_T := (\alpha \cup \beta \cup \zeta_T^{n-k})/\zeta_T^n.$$ Division by $\zeta_T^n$ makes sense because $H^{2n}$ is one dimensional.

Observe that we have $$\langle \alpha, \beta \rangle_{DR} = \langle \alpha, \beta \rangle_{B} (2 \pi i)^{-k}$$ as pairings on $H^k_{DR} \otimes \mathbb{C} \cong H^k_B \otimes \mathbb{C}$, because of the distinction between $\zeta_{DR}$ and $\zeta_T$.

Choose bases for these cohomology theories and let $A_{DR}$ and $A_B$ be the matrices of the Hard-Lefschetz pairing in these two bases. Let $P$ be the matrix of the period isomorphism $H_{DR}^K \to H_B^k$. The above discussion turns into the matrix equation $$A_{DR} = P^T A_B P (2 \pi i)^{-k}.$$ So $(\det A_{DR})= (\det A_B) (\det P)^2 (2 \pi i)^{-k b_k}$ where $b_k = \dim H^k$. The two $A$-matrices have rational entries and, since the pairings were perfect, $\det A_B$ and $\det A_{DR} \neq 0$. So $(\det P)^2 \in (2 \pi i)^{k b_k} \mathbb{Q}^{\times}$ as desired.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.