# Holomorphic and Harmonic functions

I'm studying holomorphic functions in my Complex Analysis class and have encountered the following problem:

Let $U,V$ be open subsets of $\mathbb{C}$ and $f$ be a holomorphic function on $U$ with range in $V$ and let $\phi : V \longrightarrow \mathbb{R}$ be harmonic. Prove that the composite function $g(z)= \phi (f(z))$ is harmonic on $U$.

Thoughts:

I think we may need to make use of the fact that the real and imaginary parts of holomorphic functions are harmonic respectively. I've come across the term 'harmonic conjugate' whilst working through the notes and feel this may also be involved.

A proof that uses reasonably elementary properties of complex analysis would be highly appreciated. Regards as always, MM.

-

I'm writing this mostly for my own benefit, since I'm currently working on the same things.

First, we show that if $h(z)$ is harmonic in a simply connected domain $D$ (a domain being a connected open set, simply connected meaning 'no holes'), then there is $F(z)$ analytic (synonymous with holomorphic; use 'real analytic' for a real-valued funcion) with $\Re F(z) = h(z)$.

As the previous poster pointed out, the first observation is that for harmonic $h(z)$, the function $\psi = \frac{\partial h}{\partial x} - i\frac{\partial h}{\partial y}$ is also harmonic. Again, as mentioned, this follows from the C.R. equations, but one must also note that we have continuity of the partials (there are many examples where the C.R. equations hold, yet the function is not analytic).

Now, since $D$ is simply connected and the function $\psi$ is analytic, it has a primitive in $D$ (simple connectivity is crucial here, as this is where Cauchy's theorem holds). Call this primitive $G(z)$, so $$G'(z) = \frac{\partial h}{\partial x} - i\frac{\partial h}{\partial y}.\tag{1}$$ But we know that $G$ is analytic, so $G = u+iv$. Thus, \begin{gather*} G' = u_x + iv_x = v_y-iu_y \end{gather*} Comparing this last equation to (1) (using $u_x+iv_x$ for the first comparison and $v_y-iu_y$ for the second), we find \begin{gather*} \Re \frac{\partial G}{\partial x} = \frac{\partial h}{\partial x} = \frac{\partial \Re G}{\partial x} \cr \Re \frac{\partial G}{\partial y} = \frac{\partial h}{\partial y} = \frac{\partial \Re G}{\partial y} \end{gather*} Thus, \begin{gather*} \frac{\partial}{\partial x}(\Re G(z) - h(z)) = 0 \cr \frac{\partial}{\partial y}(\Re G(z) - h(z)) = 0 \end{gather*} Since $D$ is a connected domain, we must have $h(z) = \Re G(z) + c$, so setting $F(z) = G(z) + c$ gives us $h(z) = \Re F(z)$, as desired.

We now have, given a harmonic function $h$, an analytic function whose real part is $h$. This required some work and wasn't totally trivial; in particular, we needed simple connectivity of the domain. What's important for us is the total converse: given $h(z) = \Re F(z)$ for $F$ analytic, $h$ is harmonic. You can prove this using the Cauchy-Riemann equations

Now, we want to show that the composition of a harmonic function with an analytic one is harmonic. Suppose then $\phi(z)$ is harmonic in $V$, and let $f(z)$ be analytic in $U$ with $f(U) \subset V$. Then $u(f(z))$ is harmonic for $z \in U$.

To prove this, keep in mind that harmonicity is a local property. By the previous result, we have (locally) some function $\varphi(z)$, analytic, such that $\phi(z) = \Re \varphi(z)$. Let then $z_0 \in U$ with $f(z_0) = w_0 \in V$. Then $\varphi(f(z)) = \varphi(w)$ will be analytic for $w$ near $w_0$, so $$u(f(z)) = \Re \varphi(f(z)).$$ By the total converse, $u(f(z))$ is harmonic for $z$ close enough to $z_0$.

-

For a fixed $z_0\in U$, consider an open disc $D$ centered at $f(z_0)$ and contained in $V$. On $D$, the function $\phi$ is the real part of a holomorphic function $\Phi$, and $\Phi\circ f$ is holomorphic (so its real part $\phi\circ f$ is harmonic). In this way you can show that the Laplacian of $\phi\circ f$ is zero at every point in $U$.

-
Thanks for that; how can we say that $\phi$ is the real part of a holomorphic function though? –  Mathmo Jan 13 '12 at 23:56
@MiamiMaths I should have elaborated, sorry. That's where the term "harmonic conjugate" comes in: A harmonic function on a convex (or more generally, simply connected) domain always has a harmonic conjugate on that domain. One way of seeing this is to look at the function $\psi=\phi_x-i\phi_y$, which is holomorphic (it satisfies the Cauchy-Riemann equations). Now use the fact that holomorphic functions on simply connected domains have primitives. –  Nick Strehlke Jan 14 '12 at 0:12