5

I know that we can just define the differential operators $$\frac{\partial}{\partial z} = \frac{1}{2}(\frac{\partial}{\partial x} - i \frac{\partial}{\partial y})$$ $$\frac{\partial}{\partial \bar{z}} = \frac{1}{2}(\frac{\partial}{\partial x} + i \frac{\partial}{\partial y})$$ and that $f$ is holomorphic means that $\frac{\partial f}{\partial \bar{z}}=0$, that's all clear. However, what I don't get it is how we can write any arbitrary function $f(z)=f(z,\bar{z})$, and then calculate 'formally' with the above differential operators and expect the outcomes to be as expected (or even well defined).

So this question is really not about the derivation/definition of the Wirtinger Derivatives, that's all document very well. It's about why we can write for example

\begin{align} \frac{\partial (g\circ f)}{\partial z}&=\frac{\partial (g(f(z,\bar z),\bar f(z,\bar z))}{\partial z}\\\\ &=\left.\frac{\partial g(w,\bar w)}{\partial w}\right|_{w=f(z,\bar z)}\times \frac{\partial f(z,\bar z)}{\partial z}+\left.\frac{\partial g(w,\bar w)}{\partial \bar w}\right|_{\bar w=\bar f(z,\bar z)}\times \frac{\partial \bar f(z,\bar z)}{\partial z}\\\\ &=\left(\frac{\partial g}{\partial z}\circ f\right)\frac{\partial f}{\partial z}+\left(\frac{\partial g}{\partial \bar z}\circ f\right)\frac{\partial \bar f}{\partial z} \end{align}

(copied from here). I just don't see how the formal definition of the Wirtinger derivatives makes it so that all of this goes through.

An answer to this question would explain, rigorously, why the steps in the above calculation are justified, starting from the fact we can write $f(z)=f(z,\bar{z})$ in a well-defined way s.t. the operators $\frac{\partial}{\partial z}$ and $\frac{\partial}{\partial \bar{z}}$ behave as expected.

Thank you

user2520938
  • 7,235
  • 1
    I don't understand your notation. If $f(z)=f(z,\overline z)$ then $f$ is both a function of one variable and a function of two variables? How does that work? – Gregory Grant Feb 19 '16 at 22:38
  • 1

    However, what I don't get it is how we can write any arbitrary function $f(z)=f(z,\bar z)$

    $$ $$

    You can't. $f(\bar z)$ means you replace your variable $z$ with $\bar z$.

    – Kaster Feb 19 '16 at 22:41
  • 1
    Yes, that's exactly my confusion, I don't get this, yet people seem to do it. – user2520938 Feb 19 '16 at 22:41
  • Dr. MV has a great write up already, but I want to add something. ANY function of a variable (say $x$) is also a function of $x$ and $y$, $y$ might just not show up. There are some technical considerations here, but morally this is true. –  Feb 20 '16 at 02:22

2 Answers2

3

Note that the referenced post does not suggest that $f$ is a function of $z$ only and nowhere do we see written "$f(z)=f(z,\bar z)$." Rather, a complex-valued function, $f$, is in general, a function of both $z$ and $\bar z$. To see this, let's take a closer look at things.

Let $\hat f$ be a complex function. Then we can write $\hat f$ in terms of its real and imaginary parts

$$\hat f(x,y)=u(x,y)+iv(x,y) \tag 1$$

where $ u(x,y)$ and $ v(x,y)$ are real=valued functions of $x$ and $y$ with

$$u(x,y)=\text{Re}(\hat f(x,y))$$

and

$$v(x,y)=\text{Im}(\hat f(x,y))$$

Next, note that we can write $x$ and $y$ in terms of $z=x+iy$ and $\bar z=x-iy$ as

$$x=\frac12(z+\bar z) \tag 2$$

and

$$y=\frac{1}{2i}(z-\bar z) \tag 3$$

Substituting $(2)$ and $(3)$ into $(1)$ reveals

$$\begin{align} \hat f(x,y)&= u\left(\frac12(z+\bar z),\frac{1}{2i}(z-\bar z)\right)+i v\left(\frac12(z+\bar z),\frac{1}{2i}(z-\bar z)\right)\\\\ &=f(z,\bar z)\\\\ \end{align}$$

for some function $f$ of $z$ and $\bar z$. So, any complex-valued function that can be expressed as in $(1)$ can be expressed as a function of $z$ and $\bar z$.

Mark Viola
  • 179,405
  • Very good write up, +1 –  Feb 20 '16 at 02:19
  • @avid19 Thanks Zach! Much appreciative. - Mark – Mark Viola Feb 20 '16 at 04:27
  • This still doesn't make sense. we have $u:\mathbb{R}^2\to \mathbb{R}$. However, if we plug in $f(i,0)=u(\frac{1}{2}i,\frac{1}{2})+v(...)$ we see this is not defined. Hence this does not answer the question, because this is still just playing with notation. $f(z,\bar{z})$ does not behave as a 'usual' function in 2 variables. – user2520938 Feb 20 '16 at 15:22
  • 3
    How can one "plug in" $z=i$ and $\bar z=0$. That is inconsistent. – Mark Viola Feb 20 '16 at 17:36
  • 4
    @Dr.MV yes, so this functions is only defined in a subset of $\mathbb{C}^2$. I'm not looking for any kind of 'intuitive' answers. Specify the domain of $f(z,\bar{z})$, show me, if necessary by going all the way back to $\epsilon-\delta$ arguments, how we calculate $\frac{\partial f}{\partial \bar{z}}$ for this function. To start:define $f:{(a,b)\in\mathbb{C}^2\mid a=\bar{b}}\to \mathbb{C}:(a,b)\mapsto u\left(\frac12(a+b),\frac{1}{2i}(a-b)\right)+i v\left(\frac12(a+b),\frac{1}{2i}(a-b)\right)$ – user2520938 Feb 20 '16 at 20:26
  • We view $z$ and $\bar z$ as independent variables when taking partial derivatives. The rest is formal operations. – Mark Viola Feb 20 '16 at 22:09
  • Observe that $x=\frac12 z+\frac12 \bar z$ while $y=\frac1{2i}z+\frac1{2i}\bar z$. Then, just use the chain rule. For $f=u+iv$, we have $$\frac{\partial f}{\partial \bar z}=\left(\frac{\partial f}{\partial x}\frac{\partial x}{\partial \bar z}+\frac{\partial f}{\partial y}\frac{\partial y}{\partial \bar z}\right)$$ – Mark Viola Feb 20 '16 at 22:33
  • 2
    @Dr.MV sorry this answer is just not what I'm looking for. You fail to adres the issue I have with this whole thing, which is that a lot of suggestive notation is used without rigorous arguments to back them up. Note for example, that the domain of definition of your $f$ has empty interior, how are you going to take derivatives in that? – user2520938 Feb 20 '16 at 23:59
  • I don't understand the "issue." $f$ is a function of $z$ and $\bar z$. $z$ and $\bar z$ are functions of $x$ and $y$. What is the issue with the chain rule? Certainly we can extend $u$ and $v$, to be functions of complex variables, if that is the "issue." – Mark Viola Feb 21 '16 at 00:41
0

Background

Throughout, I write things like $f:X\to Y$ when I either mean a partial function $f:X\not\to Y$, or at least that I only care about and am making claims about nice properties on a contextually-appropriate subset of $X$. It wasn't worth giving names to all of the relevant subsets.

For a reference on real and complex series of multiple variables, see Chapter 1 of Notes on global analysis by Andrew D. Lewis. For a higher level approach to the Wirtinger derivatives, see Why Wirtinger derivatives behave so well in the writings of MathSE's own Bart Michels.

Two Input Variables

Suppose we have a complex function $f:\mathbb{C}\to\mathbb{C}$ that is nice enough to have real-analytic components on some open domain $D$, like $f(z)=|z|+\exp(\overline{z})$. Then via $x=\Re z,y=\Im z,u=\Re f,v=\Im f$, we can interpret this as a function $\mathbf{f}:\mathbb{R}^{2}\to\mathbb{R}^{2}$ like $\mathbf{f}(x,y)=\left(u(x,y),v(x,y)\right)$ where $u$ and $v$ are analytic on $D$. Since $D$ is open, the series for $u$ and $v$ are absolutely convergent, so that we can use the power series to extend $u$ and $v$ to complex functions $\widetilde{u},\widetilde{v}:\mathbb{C}^{2}\to\mathbb{C}$. These can be put together to form $\widetilde{\mathbf{f}}:\mathbb{C}^{2}\to\mathbb{C}^{2}$. Inspired by "$x=(z+\overline{z})/2$" and "$y=(z-\overline{z})/(2i)$", we can define a helper function $\mathbf{h}:\mathbb{C}^{2}\to\mathbb{C}^{2}$ given by $h\left(z_{1},z_{2}\right)=\left((z_{1}+z_{2})/2,(z_{1}-z_{2})/(2i)\right)$. Define $\widehat{\mathbf{f}}:\mathbb{C}^{2}\to\mathbb{C}^{2}$ to be the composition $\widetilde{\mathbf{f}}\circ\mathbf{h}$. Note that for any $z\in\mathbb{C}$, we have $\widehat{\mathbf{f}}\left(z,\overline{z}\right)=\left(\Re f(z),\Im f(z)\right)\in\mathbb{R}^{2}$.

Then, at least on a compact subset of $D$ (so that we have uniform convergence for the series), the complex partial derivative $\left.\dfrac{\partial\widehat{\mathbf{f}}(z_{1},z_{2})}{\partial z_{1}}\right|_{\left(z_{1},z_{2}\right)=\left(a,\overline{a}\right)}$ can be written in terms of the Wirtinger derivative of $f$, as $\left(\Re\left.\dfrac{\partial f(z)}{\partial z}\right|_{z=a},\Im\left.\dfrac{\partial f(z)}{\partial z}\right|_{z=a}\right)$ and similarly for $\dfrac{\partial\widehat{\mathbf{f}}(z_{1},z_{2})}{\partial z_{2}}$ and $\dfrac{\partial g(z)}{\partial\overline{z}}$.

Chain Rule

Let $a$ be a complex number with real and imaginary parts $(x,y)$. For convenience, set $H:=D\left(\mathbf{h}\right)=\dfrac{1}{2}\begin{bmatrix}1 & 1\\-i & i\end{bmatrix}$.

We have

\begin{align*} &\phantom{=}\begin{bmatrix}\Re\left.\dfrac{\partial\left(g\circ f\right)(z)}{\partial z}\right|_{a} & \Re\left.\dfrac{\partial\left(g\circ f\right)(z)}{\partial\overline{z}}\right|_{a}\\ \Im\left.\dfrac{\partial\left(g\circ f\right)(z)}{\partial z}\right|_{a} & \Im\left.\dfrac{\partial\left(g\circ f\right)(z)}{\partial\overline{z}}\right|_{a} \end{bmatrix}\\ &=D\left(\widehat{\mathbf{g\circ f}}\right)_{\left(a,\overline{a}\right)}=D\left(\widetilde{\mathbf{g\circ f}}\circ\mathbf{h}\right)_{\left(a,\overline{a}\right)}\\&=D\left(\widetilde{\mathbf{g\circ f}}\right)_{(x,y)}D\left(\mathbf{h}\right)_{\left(a,\overline{a}\right)}=D\left(\widetilde{\mathbf{g\circ f}}\right)_{(x,y)}H\\&=D\left(\widetilde{\mathbf{g}}\circ\widetilde{\mathbf{f}}\right)_{(x,y)}H=D\left(\left(\widehat{\mathbf{g}}\circ\mathbf{h}^{-1}\right)\circ\left(\widehat{\mathbf{f}}\circ\mathbf{h}^{-1}\right)\right)_{(x,y)}H\\&=D\left(\widehat{\mathbf{g}}\circ\mathbf{h}^{-1}\right)_{\widehat{\mathbf{f}}(a,\overline{a})}D\left(\widehat{\mathbf{f}}\circ\mathbf{h}^{-1}\right)_{(x,y)}H\\&=\left(D\left(\widehat{\mathbf{g}}\right)_{\mathbf{h}^{-1}\left(\widehat{\mathbf{f}}(a,\overline{a})\right)}H^{-1}\right)\left(D\left(\widehat{\mathbf{f}}\right)_{\left(a,\overline{a}\right)}H^{-1}\right)H\\&=D\left(\widehat{\mathbf{g}}\right)_{\left(f(a),\overline{f(a)}\right)}\left(H^{-1}D\left(\widehat{\mathbf{f}}\right)_{\left(a,\overline{a}\right)}\right)\\&=\begin{bmatrix}\Re\left.\dfrac{\partial g(z)}{\partial z}\right|_{f(a)} & \Re\left.\dfrac{\partial g(z)}{\partial\overline{z}}\right|_{f(a)}\\ \Im\left.\dfrac{\partial g(z)}{\partial z}\right|_{f(a)} & \Im\left.\dfrac{\partial g(z)}{\partial\overline{z}}\right|_{f(a)} \end{bmatrix}\left(\begin{bmatrix}1 & i\\ 1 & -i \end{bmatrix}\begin{bmatrix}\Re\left.\dfrac{\partial f(z)}{\partial z}\right|_{a} & \Re\left.\dfrac{\partial f(z)}{\partial\overline{z}}\right|_{a}\\ \Im\left.\dfrac{\partial f(z)}{\partial z}\right|_{a} & \Im\left.\dfrac{\partial f(z)}{\partial\overline{z}}\right|_{a} \end{bmatrix}\right)\\&=\begin{bmatrix}\Re\left.\dfrac{\partial g(z)}{\partial z}\right|_{f(a)} & \Re\left.\dfrac{\partial g(z)}{\partial\overline{z}}\right|_{f(a)}\\ \Im\left.\dfrac{\partial g(z)}{\partial z}\right|_{f(a)} & \Im\left.\dfrac{\partial g(z)}{\partial\overline{z}}\right|_{f(a)} \end{bmatrix}\begin{bmatrix}\left.\dfrac{\partial f(z)}{\partial z}\right|_{a} & \left.\dfrac{\partial f(z)}{\partial\overline{z}}\right|_{a}\\ \left.\dfrac{\partial\overline{f}(z)}{\partial z}\right|_{a} & \left.\dfrac{\partial\overline{f}(z)}{\partial\overline{z}}\right|_{a} \end{bmatrix}\\&=\begin{bmatrix}\Re\left.\dfrac{\partial g(z)}{\partial z}\right|_{f(a)}\left.\dfrac{\partial f(z)}{\partial z}\right|_{a}+\Re\left.\dfrac{\partial g(z)}{\partial\overline{z}}\right|_{f(a)}\left.\dfrac{\partial\overline{f}(z)}{\partial z}\right|_{a} & \Re\left.\dfrac{\partial g(z)}{\partial z}\right|_{f(a)}\left.\dfrac{\partial f(z)}{\partial\overline{z}}\right|_{a}+\Re\left.\dfrac{\partial g(z)}{\partial\overline{z}}\right|_{f(a)}\left.\dfrac{\partial\overline{f}(z)}{\partial\overline{z}}\right|_{a}\\ \Im\left.\dfrac{\partial g(z)}{\partial z}\right|_{f(a)}\left.\dfrac{\partial f(z)}{\partial z}\right|_{a}+\Im\left.\dfrac{\partial g(z)}{\partial\overline{z}}\right|_{f(a)}\left.\dfrac{\partial\overline{f}(z)}{\partial z}\right|_{a} & \Im\left.\dfrac{\partial g(z)}{\partial z}\right|_{f(a)}\left.\dfrac{\partial f(z)}{\partial\overline{z}}\right|_{a}+\Im\left.\dfrac{\partial g(z)}{\partial\overline{z}}\right|_{f(a)}\left.\dfrac{\partial\overline{f}(z)}{\partial\overline{z}}\right|_{a} \end{bmatrix}\end{align*}

This yields the following two equations: $$\left.\dfrac{\partial\left(g\circ f\right)(z)}{\partial z}\right|_{a}=\left.\dfrac{\partial g(z)}{\partial z}\right|_{f(a)}\left.\dfrac{\partial f(z)}{\partial z}\right|_{a}+\left.\dfrac{\partial g(z)}{\partial\overline{z}}\right|_{f(a)}\left.\dfrac{\partial\overline{f}(z)}{\partial z}\right|_{a}$$ $$\left.\dfrac{\partial\left(g\circ f\right)(z)}{\partial\overline{z}}\right|_{a}=\left.\dfrac{\partial g(z)}{\partial z}\right|_{f(a)}\left.\dfrac{\partial f(z)}{\partial\overline{z}}\right|_{a}+\left.\dfrac{\partial g(z)}{\partial\overline{z}}\right|_{f(a)}\left.\dfrac{\partial\overline{f}(z)}{\partial\overline{z}}\right|_{a}$$ Using shorthand, these can be rewritten in the more familiar form: $$\dfrac{\partial\left(g\circ f\right)}{\partial z}=\left(\dfrac{\partial g}{\partial z}\circ f\right)\dfrac{\partial f}{\partial z}+\left(\dfrac{\partial g}{\partial\overline{z}}\circ f\right)\dfrac{\partial\overline{f}}{\partial z}$$ $$\dfrac{\partial\left(g\circ f\right)}{\partial\overline{z}}=\left(\dfrac{\partial g}{\partial z}\circ f\right)\dfrac{\partial f}{\partial\overline{z}}+\left(\dfrac{\partial g}{\partial\overline{z}}\circ f\right)\dfrac{\partial\overline{f}}{\partial\overline{z}}$$

Mark S.
  • 23,925