Assume $E\rightarrow X$ is a holomorphic vector bundle of rank $n$ with a linear connection $\nabla=\nabla^{1, 0}+\bar\partial_E$ which is compatible with the complex (somewhere the literature says "holomorphic") structure of $E$. Let $P\rightarrow X$ be the associated principal $GL_n(\mathbb{C})$-bundle (say the frame bundle of $E$).

There is a theorem which states that $\nabla$ induces a unique connection 1-form $\omega$ on $P$.

Let further $V$ be a representation of $GL_n(\mathbb{C})$ and let $P\times_{GL_n(\mathbb{C})}V$ be the associated vector bundle which should be in my opinion holomorphic as well. The principal connection $\omega$ induces always a linear connection $\tilde{\nabla}$ on the associated vector bundle $P\times_{GL_n(\mathbb{C})}V$. The question I want to pose is the following:

Is $\tilde{\nabla}$ compatible with the complex structure on $P\times_{GL_n(\mathbb{C})}V$?

P.S.:Thanks in advance for any clarifications, corrections and responses to that for me very interesting question.

  • $\begingroup$ Yes. You can do everything in the holomorphic category: a holomorphic connection induces a holomorphic connection on $P$, and this induces a holomorphic connection on any associated vector bundle. $\endgroup$
    – abx
    Apr 1, 2014 at 17:03
  • $\begingroup$ Thanks for your fast comment. $\nabla$ is NOT a holomorphic connection. It is just compatible with the complex structure (somewhere they say "holomorphic structure") of $E$. Compatible connection means $\nabla=\nabla^{1, 0}+\bar\partial_E$, where $\bar\partial_E^2=0$. $\endgroup$ Apr 1, 2014 at 17:11
  • $\begingroup$ Yes. Write out the connection on $P \times_{GL_n(\mathbb{C})} V$, in a basis of local sections, in terms of the connection on $P$ and, you should see compatibility immediately. $\endgroup$
    – Ben McKay
    May 2, 2014 at 11:10
  • $\begingroup$ Just to minimize confusion: in the common parlance, "complex" refers to topology, namely a vector bundle whose fiber is a complex vector space. The term "holomorphic bundle" refers to a holomorphic structure on a complex bundle, namely a way of picking out which sections are locally holomorphic vector valued functions on the base. The latter is most often described by a $\overline{\partial}$-operator as you've indicated above. Once you agree that it's these operators giving the holomorphic structure, the question you asked is largely reduced to a tautology. $\endgroup$ Jul 31, 2014 at 1:58
  • $\begingroup$ This question has been answered in a satisfactory way. $\endgroup$
    – abx
    Nov 28, 2014 at 6:32

1 Answer 1


According to your last comment, the holomorphic structure is irrelevant; you just have a $GL_n(\mathbb{C})$-bundle. I presume that your representation is complex, i.e., merely a homomorphism $H:=GL_n(\mathbb{C})\to G:=GL_m(\mathbb{C})$. (Otherwise, there is nothing to speak about.) This gives you a map $P\to P\times_HG$ of principal bundles, which is called extension of the structure group. (Unlike reduction of the structure group, it is well defined and involves no choices.) In this situation, any $H$-connection extends uniquely to a $G$-connection (e.g., via translations in the fibers, see the definition of connections as $1$-forms and their properties; BTW, a $1$-form is not induced by a connection: it is the connection). Since this is a $G$-connection, it is compatible with the complex structure just by definition.

  • $\begingroup$ I don't follow here. Do you basically claim that $\tilde{\nabla}$ is compatible with the complex structure of $P\times_{GL_n(\mathbb{C})}V$? $\endgroup$ Apr 1, 2014 at 20:06
  • $\begingroup$ Sure, $V$ is a complex vector space. So the representation is indeed a complex homomorphism as you pointed out. Ah, so basically it turns out that every connection on $P\times_{H}V$ coming from a connection 1-form on $P$ has to be compatible with the complex structure of $P\times_HV$? $\endgroup$ Apr 1, 2014 at 20:16
  • $\begingroup$ Any induced connection is "compatible" with the new structure group, whatever this means. If the new group is in $GL_n(\mathbb{C})$, the connection is compatible with the complex structure (e.g., just because $J\in GL_n(\mathbb{C})$). $\endgroup$ Apr 1, 2014 at 20:37

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.