2
$\begingroup$

Let $E\rightarrow X$ be a vector bundle and let $\mathcal{A}$ denote the space of connections on $E$. Pulling back $E$ by the second projection we obtain a vector bundle $\mathbb{E}=p_2^*E\rightarrow X$ over $\mathcal{A}\times X$.

There exists a canonical connection $\mathbb{A}$ on $\mathbb{E}$ which is flat in the $\mathcal{A}$ direction and equal to $A$ on the slice $\{A\}\times X$. We obtain the following curvature: $$ \begin{cases} \mathbb{A}^2(v,w)= R_A(v,w) & \text{for $v,w\in T_xX$} \\ \mathbb{A}^2(\alpha,v)=\alpha(v) & \text{for $\alpha\in T_A\mathcal{A}$,$v\in T_xX$ } \\ \mathbb{A}^2(\alpha,\beta)=0 & \text{for $\alpha,\beta \in T_A\mathcal{A} $}. \end{cases} $$ These identities can be found in Donaldson's "INFINITE DETERMINANTS, STABLE BUNDLES AND CURVATURE" p236 and Itoh and Nakajima's "Yang-Mills Connections and Einstein-Hermitian Metrics" p451.

I do not understand the pairing of the middle line. How does it follow from the definition?

$\endgroup$

1 Answer 1

3
$\begingroup$

A vector field $v$ on $X$ and a vector field $\alpha$ on $\mathcal A$ give rise to two commuting vector fields on $\mathcal A\times X$, denoted by the same letters. Then we may regard $\alpha$ as (the pullback of) an element of $\Omega^1(X;\operatorname{End}E)$ as well. Write the connection as a covariant derivative $\nabla$, choose a section $s$ of $E\to X$ and compute for its pullback (still denoted $s$) that $$\underbrace{\nabla_\alpha\nabla_v}_{=\alpha(v)\in\operatorname{End}E}s-\nabla_v\underbrace{\nabla_\alpha s}_{=0} -\nabla_{\underbrace{[\alpha,v]}_{=0}}s=\alpha(v)s\;.$$ Here we have used the interpretation of $\alpha$ as a variation of $\nabla$, and we have used that $\nabla$ is trivial in the $\mathcal A$-direction.

$\endgroup$

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.