# Uniform convergence of real part of holomorphic functions on compact sets

The following is exercise 11.8 in Rudin's Real and Complex Analysis:

Suppose $\Omega$ is a region, $f_n \in H(\Omega)$ for $n = 1, 2, 3, \ldots$. $u_n$ is the real part of $f_n$. $\{u_n\}$ converges uniformly on compact subsets of $\Omega$, and $\{f_n(z)\}$ converges for at least one $z \in \Omega$. Prove that then $\{f_n\}$ converges uniformly on compact subsets of $\Omega$.

My thoughts: By Harnack's theorem, the limit $u$ of $\{u_n\}$ is harmonic. Thus it's the real part of a holomorphic function $f$ on $\Omega$ defined up to an imaginary constant. Using the limit of $\{f_n(z)\}$, we can find this constant. Thus $f$ is well-defined.

What's left is to show that the imaginary part of $f_n$ converges to that of $f$. I suspect I need to use the Cauchy-Riemann equations for this, but I cannot apply the familiar uniform convergence theorems with partial derivatives. What should I do?

Note: A region is a connected open subset of the complex plain.

• Are you familiar with the interior gradient estimate for harmonic functions? Using it, the proof goes like this: $u_n-u_m$ is small on compact subsets, therefore $\nabla (u_n-u_m)$ is small too, which by the Cauchy-Riemann equations translates into smallness of $f_n'-f_m'$. In other words, $f_n'$ converge uniformly on compact subsets, and from there it's easy.
– user53153
Feb 7, 2013 at 3:21
• @5PM Thanks. I wasn't familiar with this estimate but now I am. I wouldn't be surprised if Rudin expected the reader to come up with it on his/her own. It makes sense now. I can probably mimic the proof of this theorem to finish your proof. Is there an easier way? Would you like to post this as an answer for me to accept it? Feb 7, 2013 at 13:02
• I'm also interested in seeing a proof of this result, but possibly avoiding the technique for harmonic functions suggested by 5pm in the earlier comment.
– user44532
Feb 20, 2013 at 19:04
• I think the mean value property should help here, but can't solve it myself.
– Emo
Aug 3, 2018 at 21:26

The following inequality helps: if $$D$$ is a disk centered at $$z_0$$, $$f$$ is holomorphic in $$D$$ and continuous on $$\overline D$$, then $$\sup_{D'} |f|\le |\operatorname{Im}f(z_0)| + 3\max_{\partial D}|\operatorname{Re}f|, \tag1$$ where $$D'$$ is the disk concentric with $$D$$ and half the radius.

Proof of (1) is immediate from the Schwarz integral formula: $$f(z)= i\operatorname{Im}f(z_0)+\frac{1}{2\pi i}\int_{\partial D}\frac{\zeta+z-2z_0}{\zeta-z} \,\operatorname{Re}f(\zeta)\,\frac{d\zeta}{\zeta-z_0}, \tag2$$ where $$\left|\frac{\zeta+z-2z_0}{\zeta-z}\right|\le 3,\qquad \zeta\in\partial D,\quad z\in D'. \tag3$$

Returning to the problem at hand, an application of (1) to the differences $$f_n-f_m$$ yields uniform convergence of $$f_n$$ on any disk $$D$$ centered at $$z_0$$ and compactly contained in $$\Omega$$. Now (1) can be applied to disks whose centers lie in the set in which uniform convergence is already known. Repeating in this fashion, we can cover any compact subset of $$\Omega$$ in finitely many steps.

• I prefer the method of PeterM. If somebody could solve the last step of the proof (to show that the imaginary part of $f_n$ converges to that of $f$) it would help me too.
– Emo
Aug 3, 2018 at 22:06
• Wow. Thanks for the beautiful proof. Nov 29, 2019 at 5:19

Let $z_{0}=x_{0}+iy_{0}$ be the point in $\Omega$ where the sequence $(f_{n})$ converges. We write $f_{n}(z)=2u_{n}(\frac{z+\bar{z_{0}}}{2},\frac{z-\bar{z_{0}}}{2i})-u_{n}(x_{0},y_{0})+ic_{n}$ where $c_{n}\in\mathbb{R}$. Letting $z=z_{0}$, we see that $c_{n}=\mathrm{Im}f_{n}(z_{0})$. Then uniform convergence of $(u_{n})$ on compact subsets and convergence of $(f_{n}(z_{0}))$ imply uniform convergence of $(f_{n})$ on compact subsets.
• $f_n$'s are holomorphic function and their real parts are a function of $z$, i.e. a function of $x$ and $y$. When you denote this real part by $c_n$, it must be a function of $x$ and $y$. But you wrote ('wee see that') $c_n=\text{Im}f_n(z_0)$, which means $c_n$ is a constant and independent of $x$ and $y$. A fault appears here, I think. Nov 29, 2019 at 2:45