# Phragmén–Lindelöf principle

In complex analysis, the Phragmén–Lindelöf principle (or method), first formulated by Lars Edvard Phragmén (1863–1937) and Ernst Leonard Lindelöf (1870-1946) in 1908, is a technique which employs an auxiliary, parameterized function to prove the boundedness of a holomorphic function ${\displaystyle f}$ (i.e, ${\displaystyle |f(z)|) on an unbounded domain ${\displaystyle \Omega }$ when an additional (usually mild) condition constraining the growth of ${\displaystyle |f|}$ on ${\displaystyle \Omega }$ is given. It is a generalization of the maximum modulus principle, which is only applicable to bounded domains.

## Background

In the theory of complex functions, it is known that the modulus (absolute value) of a holomorphic (complex differentiable) function in the interior of a bounded region is bounded by its modulus on the boundary of the region. More precisely, if a non-constant function ${\displaystyle f:\mathbb {C} \to \mathbb {C} }$ is holomorphic in a bounded region[1] ${\displaystyle \Omega }$ and continuous on its closure ${\displaystyle {\overline {\Omega }}=\Omega \cup \partial \Omega }$, then ${\textstyle |f(z_{0})|<\sup _{z\in \partial \Omega }|f(z)|}$ for all ${\displaystyle z_{0}\in \Omega }$. This is known as the maximum modulus principle. (In fact, since ${\displaystyle {\overline {\Omega }}}$ is compact and ${\displaystyle |f|}$ is continuous, there actually exists some ${\displaystyle w_{0}\in \partial \Omega }$ such that ${\textstyle |f(w_{0})|=\sup _{z\in \partial \Omega }|f(z)|}$.) The maximum modulus principle is generally used to conclude that a holomorphic function is bounded in a region after showing that it is bounded on its boundary.

However, the maximum modulus principle cannot be applied to an unbounded region of the complex plane. As a concrete example, let us examine the behavior of the holomorphic function ${\displaystyle f(z)=\exp(\exp(z))}$ in the unbounded strip

${\displaystyle S={\Big \{}z:\Im (z)\in {\big (}-{\frac {\pi }{2}},{\frac {\pi }{2}}{\big )}{\Big \}}}$.

Although ${\displaystyle |f(x\pm \pi i/2)|=1}$, so that ${\displaystyle |f|}$ is bounded on boundary ${\displaystyle \partial S}$, ${\displaystyle |f|}$ grows rapidly without bound when ${\displaystyle |z|\to \infty }$ along the positive real axis. The difficulty here stems from the extremely fast growth of ${\displaystyle |f|}$ along the positive real axis. If the growth rate of ${\displaystyle |f|}$ is guaranteed to not be "too fast," as specified by an appropriate growth condition, the Phragmén–Lindelöf principle can be applied to show that boundedness of ${\displaystyle f}$ on the region's boundary implies that ${\displaystyle f}$ is in fact bounded in the whole region, effectively extending the maximum modulus principle to unbounded regions.

The technique: Suppose we are given a holomorphic function ${\displaystyle f}$ and an unbounded region ${\displaystyle S}$. In a typical Phragmén–Lindelöf argument, we introduce a certain multiplicative factor ${\displaystyle h_{\epsilon }}$ satisfying ${\textstyle \lim _{\epsilon \to 0}h_{\epsilon }=1}$ to "subdue" the growth of ${\displaystyle f}$, such that ${\displaystyle |fh_{\epsilon }| on the boundary of a bounded subregion ${\displaystyle S_{x_{0}}\subset S}$. This allows us to apply the maximum modulus principle to ${\displaystyle fh_{\epsilon }}$ and conclude that it is bounded on ${\displaystyle S_{x_{0}}}$. We then argue that the subregion could be expanded so as to encompass all points in ${\displaystyle S}$, establishing the boundedness of ${\displaystyle fh_{\epsilon }}$ on ${\displaystyle S}$. Finally, we let ${\displaystyle \epsilon \to 0}$ so that ${\displaystyle fh_{\epsilon }\to f}$ in order to conclude that ${\displaystyle f}$ must also be bounded on ${\displaystyle S}$.

In the literature of complex analysis, there are many examples of the Phragmén–Lindelöf principle applied to unbounded regions of differing types, and also a version of this principle may be applied in a similar fashion to subharmonic and superharmonic functions.

## Example of application

To continue the example above, we can impose a growth condition on a holomorphic function ${\displaystyle f}$ that prevents it from "blowing up" and allows the Phragmén–Lindelöf principle to be applied. To this end, we now include the condition that

${\displaystyle |f(z)|<\exp {\big (}A\exp(c\cdot |\Re (z)|){\big )}}$

for some real constants ${\displaystyle c<1}$ and ${\displaystyle A<\infty }$, for all ${\displaystyle z\in S}$. It can then be shown that ${\displaystyle |f(z)|\leq 1}$ for all ${\displaystyle z\in \partial S}$ implies that ${\displaystyle |f(z)|\leq 1}$ in fact holds for all ${\displaystyle z\in S}$. Thus, we have the following proposition:

Proposition. Let

${\displaystyle S={\Big \{}z:\Im (z)\in {\big (}-{\frac {\pi }{2}},{\frac {\pi }{2}}{\big )}{\Big \}},\quad {\overline {S}}={\Big \{}z:\Im (z)\in {\big [}-{\frac {\pi }{2}},{\frac {\pi }{2}}{\big ]}{\Big \}}}$.

Let ${\displaystyle f}$ be holomorphic on ${\displaystyle S}$ and continuous on ${\displaystyle {\overline {S}}}$, and suppose there exist real constants ${\displaystyle c<1,\ A<\infty }$ such that

${\displaystyle |f(z)|<\exp {\big (}A\exp(c\cdot |\Re (z)|){\big )}}$

for all ${\displaystyle z\in S}$ and ${\displaystyle |f(z)|\leq 1}$ for all ${\displaystyle z\in {\overline {S}}\setminus S=\partial S}$. Then ${\displaystyle |f(z)|\leq 1}$ for all ${\displaystyle z\in S}$.

Note that this conclusion fails when ${\displaystyle c=1}$, precisely as the motivating counterexample in the previous section demonstrates. The proof of this statement employs a typical Phragmén–Lindelöf argument:[2]

Sketch of Proof: We fix ${\displaystyle b\in (c,1)}$, define for each ${\displaystyle \epsilon >0}$ the auxiliary function ${\displaystyle h_{\epsilon }}$ by ${\textstyle h_{\epsilon }(z)=e^{-\epsilon (e^{bz}+e^{-bz})}}$, and consider the function ${\displaystyle fh_{\epsilon }}$. The growth properties of ${\displaystyle fh_{\epsilon }}$ allow us to find an ${\displaystyle x_{0}}$ such that ${\displaystyle |fh_{\epsilon }|\leq 1}$ whenever ${\displaystyle z\in {\overline {S}}}$ and ${\displaystyle |\Re (z)|\geq x_{0}}$. In particular, ${\displaystyle |fh_{\epsilon }|\leq 1}$ holds for all ${\displaystyle z\in \partial S_{x_{0}}}$, where ${\displaystyle S_{x_{0}}}$ is the open rectangle in the complex plane defined by the vertices ${\displaystyle \{x_{0}\pm i\pi /2,-x_{0}\pm i\pi /2\}}$. Because ${\displaystyle S_{x_{0}}}$is a bounded region, the maximum modulus principle implies that ${\displaystyle |fh_{\epsilon }|\leq 1}$ for all ${\displaystyle z\in S_{x_{0}}}$. But since ${\displaystyle x_{0}}$ can be made arbitrarily large, this result extends to all of ${\displaystyle S}$, and ${\displaystyle |fh_{\epsilon }|\leq 1}$ for all ${\displaystyle z\in S}$. Finally, since ${\displaystyle h_{\epsilon }\to 1}$ as ${\displaystyle \epsilon \to 0}$, we conclude that ${\displaystyle |f|\leq 1}$ for all ${\displaystyle z\in S}$.

## Phragmén–Lindelöf principle for a sector in the complex plane

A particularly useful statement proved using the Phragmén–Lindelöf principle bounds holomorphic functions on a sector of the complex plane if it is bounded on its boundary. This statement can be used to give a complex analytic proof of the Hardy's uncertainty principle, which states that a function and its Fourier transform cannot both decay faster than exponentially.[3]

Proposition. Let ${\displaystyle F}$ be a function that is holomorphic in a sector

${\displaystyle S=\left\{z\,{\big |}\,\alpha <\arg z<\beta \right\}}$

of central angle ${\displaystyle \beta -\alpha =\pi /\lambda }$, and continuous on its boundary. If

${\displaystyle |F(z)|\leq 1}$

(1)

for ${\displaystyle z\in \partial S}$, and

${\displaystyle |F(z)|\leq e^{C|z|^{\rho }}}$

(2)

for all ${\displaystyle z\in S}$, where ${\displaystyle \rho \in [0,\lambda )}$ and ${\displaystyle C>0}$, then (1) holds also for all ${\displaystyle z\in S}$.

### Remarks

• The condition (2) can be relaxed to

${\displaystyle \liminf _{r\to \infty }\sup _{\alpha <\theta <\beta }{\frac {\log |F(re^{i\theta })|}{r^{\rho }}}=0\quad {\text{for some}}\quad 0\leq \rho <\lambda ~,}$

(3)

with the same conclusion.

## Special cases

In practice the point 0 is often transformed into the point ∞ of the Riemann sphere. This gives a version of the principle that applies to strips, for example bounded by two lines of constant real part in the complex plane. This special case is sometimes known as Lindelöf's theorem.

Carlson's theorem is an application of the principle to functions bounded on the imaginary axis.

## References

1. ^ The term region is not uniformly employed in the literature; here, a region is taken to mean a nonempty connected open subset of the complex plane.
2. ^ Rudin, Walter (1987). Real and Complex Analysis. New York: McGraw-Hill. pp. 257–259. ISBN 0070542341.
3. ^ Tao, Terence (2009-02-18). "Hardy's Uncertainty Principle". Updates on my research and expository papers, discussion of open problems, and other maths-related topics. By Terence Tao.