What are holomorphic and anti-holomorphic components? Why don't we call them complex components and their conjugates? What is holomorphic coordinate transformation?


$\newcommand{\dd}{\partial}\newcommand{\Reals}{\mathbf{R}}\newcommand{\Cpx}{\mathbf{C}}$The $z^{\beta}$ are emphatically not holomorphic coordinates in general. "Holomorphic" refers to overlap maps being holomorphic, i.e., satisfying the Cauchy-Riemann equations, not merely to the coordinates being complex-valued.

In more detail, fix an identification of $\Cpx^{n}$ and $\Reals^{2n}$, say $z^{\beta} = x^{\beta} + iy^{\beta}$, and identify $$ (z^{1}, \dots, z^{n}) \leftrightarrow (x^{1}, \dots, x^{n}, y^{1}, \dots, y^{n}). $$ Define the complex-valued differential operators $$ \frac{\dd}{\dd z^{\beta}} = \frac{1}{2}\left(\frac{\dd}{\dd x^{\beta}} - i\frac{\dd}{\dd y^{\beta}}\right),\qquad \frac{\dd}{\dd \bar{z}^{\beta}} = \frac{1}{2}\left(\frac{\dd}{\dd x^{\beta}} + i\frac{\dd}{\dd y^{\beta}}\right), $$ and their dual one-forms $dz^{\beta} = dx^{\beta} + i\, dy^{\beta}$, $d\bar{z}^{\beta} = dx^{\beta} - i\, dy^{\beta}$.

A $C^{1}$ function $f:\Cpx^{n} \to \Cpx$ is holomorphic if $\dfrac{\dd f}{\dd \bar{z}^{\beta}} = 0$ for all $\beta$, and is anti-holomorphic if $\dfrac{\dd f}{\dd z^{\beta}} = 0$ for all $\beta$, i.e., if $\bar{f}$ is holomorphic. A smooth mapping $f:\Cpx^{n} \to \Cpx^{n}$ is holomorphic if each component function is holomorphic.

Holomorphic maps constitute a (rich but) thin subset of all smooth mappings. Among their many nice properties, a holomorphic mapping $F$ is complex analytic (representable by a convergent power series in a neighborhood of each point), and is locally invertible near a point $p$ (i.e., is a local biholomorphism at $p$) if and only if its derivative $dF(p)$ (which is a complex-linear mapping) is invertible (non-singular).

The chain rule guarantees a composition of holomorphic maps is holomorphic, so the set of local biholomorphisms forms a pseudo-group. (Loosely, forms a group, modulo fine print about composability.) It therefore makes sense to speak of a "holomorphic structure" or a "holomorphic coordinate system" by declaring two $\Cpx^{n}$-valued coordinate systems to be compatible if the change of coordinates (a.k.a. the overlap map) from one to the other is holomorphic. This notion of compatibility is an equivalence relation; particularly, compatibility is transitive.

Anti-holomorphic maps are conjugate-analytic in an obvious sense, and are locally invertible at $p$ if and only if their Jacobian is non-singular. The derivative of an anti-holomorphic map is not complex-linear, however, and a composition of anti-holomorphic maps is not anti-holomorphic. Consequently, there is no notion of an "anti-holomorphic structure" or "anti-holomorphic coordinate system": The corresponding notion of "compatibility" is not transitive.

Loosely, "holomorphic" and "anti-holomorphic" are not "dual" or "perfectly symmetric" concepts; the distinction between them is not arbitrary. That said, the definition ultimately rests on a choice of complex square root $i$ of $-1$, and arguably such a choice is arbitrary. But this choice is fixed by the time one even asks about complex differentiability.

  • $\begingroup$ thanks Andrew for the answer. Very elaborate. I still have a question about the transformation. Yesterday, as I was looking deeper into my notes, I saw this: The ‘splitting’ of an index b into $\beta$ and $\bar{\beta}$ is not preserved by general transformation of complex coordinates $z'^{\beta} = f ^{\beta}(z, \bar{z})$, but it is preserved under the special class of ‘holomorphic’ cordinate transformations $z'^{\beta}=f^{\beta}(z)$. So, accordingly, I am wondering about two things. (I will write them in a new comment) $\endgroup$ – user291321 Nov 23 '15 at 13:58
  • $\begingroup$ 1)I you have written in the beginning of your answer that this has nothing to do with being complex-values, why is there a repeated tendency to compare between transformation of complex coordinates and holomorphic coordinate transformation? 2) I still haven't really understood what do we mean by holomorphic coordinate transformation. $\endgroup$ – user291321 Nov 23 '15 at 14:00
  • $\begingroup$ For 2 (and possibly for 1), my answer to a question about $z$ and $\bar{z}$ versus $x$ and $y$ may be of interest. Formally, $z^{\beta}$ and $\bar{z}^{\beta} = z^{\bar{\beta}}$ may be treated as independent variables, alternative to the real variables $x^{\beta}$ and $y^{\beta}$. A holomorphic map depends only on the $z^{\beta}$, not on the "conjugate" variables. $\endgroup$ – Andrew D. Hwang Nov 24 '15 at 1:19
  • $\begingroup$ Oh right.. i see.. so, is there any reason behind the fact that a holomorphic map depends only on $z^{\beta}$ and not on the conjugate variables? $\endgroup$ – user291321 Nov 26 '15 at 13:34
  • $\begingroup$ That's the definition of "holomorphic": $\dd f/\dd \bar{z}^{\beta} = 0$, i.e., "$f$ is constant with respect to $\bar{z}^{\beta}$". $\endgroup$ – Andrew D. Hwang Nov 26 '15 at 14:13

You can talk about homomorphic and antiholomorphic functions, but a general complex-valued smooth function will not be the sum of a holomorphic function with an antiholomorphic one (for example, $|z|^2$ on $\mathbb{C}$). Also, the constant functions are both holomorphic and antiholomorphic.

On the other hand, complex-valued smooth $1$-forms do decompose into holomorphic and antiholomorphic components. In local coordinates $(z_i)$, the holomorphic $1$-forms are spanned by the $dz_i$, with smooth (not necessarily holomorphic) coefficient functions, and the antiholomorphic forms by their conjugates. Of course, you should check that this definition is independent of the choice of (holomorphic) coordinates.

I do not know why you would call the holomorphic components "complex," because they are no more complex than the antiholomorphic components. Usually, they would be called forms of type $(1, 0)$ or $(0, 1)$, because this generalizes to higher differential forms, and a "holomorphic $1$-form" is typically required to have holomorphic coefficients.

It is an instructive exercise to check if a holomorphic coordinate transformation is just a holomorphic map with a holomorphic inverse or not. Does the domain and/or codomain have to be an open subset of $\mathbb{C}$?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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