# Reasons for defining sheaves of holomorphic and meromorphic functions on complex manifolds

I am hoping this question is sensible and non-trivial. I am learning algebraic geometry at the moment, and have taken a strong liking to it. Unfortunately my complex analysis is weaker and I only know it at an undergraduate level. I am trying to transfer some of what I know in algebraic geometry to the language of complex analysis, particularly complex manifolds.

My main question is, what are the benefits and drawbacks of defining the sheaf on a complex manifold (for the time being say a Riemann surface, or even just the Riemann sphere) in terms of holomorphic functions as opposed to meromorphic functions? From what I have gathered, meromorphic functions align better with the theory of discrete valuation rings on algebraic curves, since this provides a framework for studying poles. However it seems that holomorphic functions are taken to be the standard structure sheaf. What difference does this make, and why do you choose one over the other in certain situations? Does it make any difference to the sheaf cohomology? Does it make a difference if the surface is compact or not?

Again, forgive me if this question is either trivial or not particularly meaningful, but I feel like it would massively boost the speed I can learn complex geometry if I can frame it in the language of ringed spaces and algebraic geometry.

Any help is appreciated, or even some introductory notes that you think would help someone coming from my perspective.

Thanks

• Very nice first question! – José Carlos Santos Jul 8 '17 at 6:27
• I'm not really knowledgeable in this area, but wouldn't the sheaf of meromorphic functions be flabby, and hence cohomologically trivial? So I guess it is just more interesting to use holomorphic functions where we have non-trivial glueing problems. Have you reached Riemann-Roch, yet? The study of poles and zeros really is in some sense about glueing together suitably (locally) twisted holomorphic functions. – Jyrki Lahtonen Jul 8 '17 at 6:40
• @JyrkiLahtonen The sheaf of meromorphic functions is not flabby, even over $\Bbb C$: not every meromorphic function on the unit disc extends to one over $\Bbb C$. – Lord Shark the Unknown Jul 8 '17 at 7:37
• You're right, of course, @LordSharktheUnknown. I guess I was living inside the field of rational functions or some such, and forgot about those that cannot be continued analytically :-( – Jyrki Lahtonen Jul 8 '17 at 7:56
• In case it's of interest, in two or more variables, meromorphic "functions" such as $z_{1}/z_{2}$ are generally not functions at all, even allowing $\infty$ as a value. Locally, a meromorphic function is a ratio $f/g$ of holomorphic functions ($g$ not identically $0$). The zero set of a holomorphic function is a hypersurface. Generally, one cannot avoid existence of an indeterminacy set where $f$ and $g$ both vanish. (On a Riemann surface, a "hypersurface" is a point, and if two holomorphic functions vanish simultaneously, they have a highest common factor that can be canceled.) – Andrew D. Hwang Jul 8 '17 at 15:24

If you are a complex analyst, you cannot not like holomorphic functions, in many ways they are preferable to meromorphic ones. For instance, your first preference is to solve, say, differential equations so that solutions exist everywhere (in the domain where the equation is defined and is suitably regular), rather than on an open subset. You also may want to have finite dimensionality of the space of solutions (and having estimates on solutions in terms of equations themselves or other data). Your problem, however is that compact complex manifolds lack any nonconstant holomorphic functions. As the result, you compromise and work with sheaves which may come in several different forms. One of these is the sheaf of sections of, say, a line bundle (or, more generally, a vector bundle) given by a divisor, or in form of holomorphic tensors. The latter are no longer functions and thus could exist everywhere on your compact manifold. Of course, in the process you may want to work with meromorphic functions (which are allowed to blow up at the given divisor $D$), but if you do not impose any restrictions along $D$, then you loose finite-dimensionality of the space of solutions, integrability, etc. Another thing which may happen is that solutions of your equations are multi-valued. This is not good for a variety of reasons, so you try to make them single-valued by regarding them as sections of a certain sheaf and then extending holomorphically over the branching divisor. The fact that this is (sometimes) possible is due to the fact that you are working with either holomorphic equations or, at worst, equations which are meromorphic but have controlled singularities along $D$.