1
$\begingroup$

Let $(X,\omega)$ be a compact K"ahler manifold of dimension $n$, and let $D=\sum_{i=1}^r D_i$ be a simple normal crossing divisor on it, i.e., a divisor with smooth components $D_i$ intersecting each other transversally in $X$. Let $\mathcal{V}$ be a locally free coherent sheaf on $X$ and let $$\nabla:\mathcal{V}\to \Omega^1_X(\log D)\otimes\mathcal{V}$$ be a $\mathbb{C}$-linear map satisfying \begin{align}\nabla(f\cdot e)=f\cdot\nabla e+df\otimes e.\end{align} One defines $$\nabla_a:\Omega^a_X(\log D)\otimes \mathcal{V}\to \Omega^{a+1}_X(\log D)\otimes \mathcal{V}$$ by the rule $$\nabla_a (\omega\otimes e)=d\omega\otimes e+(-1)^a \omega\wedge \nabla e.$$ We assume that $\nabla_{a+1}\circ\nabla_a=0$ for all $a$. Such $\nabla$ will be called an integrable logarithmic connection along $D$, or just a connection.

My Question is: Given a connection $\nabla$ in $\mathcal V$, can it induce a connection in its dual bundle $\mathcal V^*$?

Here is my thought: Locally, $\nabla$ can be written as the form $\nabla=d+\omega$, where $\omega$ is a holomorphic section of $\Omega_{X}^1(\log D)\otimes\text{End}(\mathcal{V})$. For $\mathcal{V}$ and its dual bundle $\mathcal{V}^*$, the dual pairing $$\langle\,,\,\rangle:\mathcal{V}_x^*\times\mathcal{V}_x\longrightarrow\mathbb{C}$$ induces a dual pairing $$\langle\,,\,\rangle:A^0(\mathcal{V}^*)\times A^0(\mathcal{V})\rightarrow A^0.$$ Given a connection $\nabla$ in $\mathcal V$, we define a connection, also denoted by $\nabla$, in $\mathcal V^{*}$ by the following formula: $$ d(\xi, \eta)=\langle \nabla \xi, \eta\rangle+\langle\xi, \nabla \eta\rangle \text { for } \xi \in H^{0}(\mathcal V), \eta \in H^{0}\left(\mathcal V^{*}\right). $$ Given a local frame ficld $s=\left(s_{1}, \ldots, s_{r}\right)$ of $\mathcal V$ over an open set $U,$ let $t=$ $\left(t^{1}, \cdots, t^{r}\right)$ be the local frame field of $\mathcal V^{*}$ dual to $s$ so that $$ \left\langle t^{i}, s_ j\right\rangle=\delta^{i}_j, \text { or }\left\langle t, s\right\rangle=I_{r}, $$ where $s$ is considered as a row vector and $t$ as a column vector. If $\omega=(\omega_{j}^{i})$ denotes the connection form of $\nabla \text { with respect to } s \text { so that }$ $$ \nabla s_{i}=\sum s_{j} \omega_{i}^{j} \qquad \text{or} \,\,\nabla s=s \omega{,}$$ then $$ \nabla t_{j}=-\sum \omega_{i}^{j}t^j \qquad \text{or} \,\,\nabla t= -\omega t{.}$$This follows from $$ 0=d \delta_{j}^{i}=\left\langle \nabla t^{i}, s_{j}\right\rangle+\left(t^{i}, \nabla s_{j}\right)=\left\langle \nabla t^{i}, s_{j}\right\rangle+\omega_{j}^{i}. $$


In short, $\omega_{\mathcal V^*}=-\omega_{\mathcal V}$.

Am I right? Any advice and suggestion will be appreciated. Thanks a lot.

$\endgroup$
6
  • 1
    $\begingroup$ yes, the dual meromorphic connection is indeed logarithmic, which is readily seen from the formula for the dual connection in the local coordinates $\endgroup$ Commented Jun 15, 2020 at 16:34
  • $\begingroup$ Dear @LevSoukhanov, thanks for your comment, so my method is right? $\endgroup$
    – Invariance
    Commented Jun 15, 2020 at 17:02
  • $\begingroup$ Your method looks right to me. You may want to check that your "product rule" formula holds for every pair of sections, and not just your frame fields, so that you know the dual does not depend on the choice of basis. On the other hand if you already know a connection satisfying that formula exists then you don't need to do this. $\endgroup$
    – Will Sawin
    Commented Jun 15, 2020 at 17:22
  • $\begingroup$ @WillSawin, thanks for your comment! $\endgroup$
    – Invariance
    Commented Jun 15, 2020 at 17:48
  • $\begingroup$ Yes your method is correct. $\endgroup$ Commented Jun 16, 2020 at 17:22

0

You must log in to answer this question.