Imagine you are trying to find the largest value that a temperature distribution can achieve inside a room. If the temperature satisfies Laplace's equation — as it does in steady-state heat conduction — you discover something striking: the hottest point in the room is always on the wall, never in the interior. No matter how complicated the boundary conditions, the maximum is achieved at the boundary. This is not a coincidence or a special feature of heat; it is a deep structural property of all holomorphic functions. The Maximum Modulus Principle says that a non-constant holomorphic function cannot achieve its maximum modulus in the interior of its domain. The maximum is always on the boundary, and this single fact has consequences that ripple through all of complex analysis. The principle is surprising because it has no real analogue: the function $f(x) = \sin(x)$ on $[-\pi, \pi]$ achieves its maximum at the interior point $x = \pi/2$. What makes holomorphic functions special is that they are far more constrained than smooth real functions — the Cauchy–Riemann equations link the real and imaginary parts so tightly that a local maximum of $|f|$ propagates everywhere, forcing the function to be constant. [example: A Holomorphic Function Achieving Its Maximum Only on the Boundary] Consider $f: \overline{B}(0, 1) \to \mathbb{C}$ defined by $f(z) = z^2$, where $\overline{B}(0, 1)$ denotes the closed unit disk $\{z \in \mathbb{C} : |z| \le 1\}$. On the boundary circle $|z| = 1$, we have $|f(z)| = |z^2| = |z|^2 = 1$. At any interior point $z_0$ with $|z_0| < 1$, we have $|f(z_0)| = |z_0|^2 < 1$. The maximum of $|f|$ on $\overline{B}(0, 1)$ is $1$, and it is attained precisely on the boundary circle. No interior point achieves the value $1$. To see concretely why the maximum cannot be interior, suppose $|f(z_0)|$ were maximal at an interior point $z_0 = r_0 e^{i\theta_0}$ with $r_0 < 1$. Then $|f(z_0)| = r_0^2$. But on the circle of radius $r_0$ centered at the origin, we could write \begin{align*} f(z_0) &= z_0^2 = r_0^2 e^{2i\theta_0}, \end{align*} and for any $r > r_0$ with $r < 1$, the point $re^{i\theta_0}$ satisfies $|f(re^{i\theta_0})| = r^2 > r_0^2$. So $f(z_0)$ is not the maximum — any point radially outward from $z_0$ has strictly larger modulus. The maximum can only be achieved on the boundary. [/example] This example captures the key geometric mechanism: in a holomorphic function, you can always move outward and find a larger modulus, unless you are already on the boundary. The Maximum Modulus Principle is the theorem that makes this rigorous and universal. ## Definition The setting for the Maximum Modulus Principle is a bounded domain in $\mathbb{C}$. Recall that a domain is an open connected subset of $\mathbb{C}$. Before stating the principle, we must be precise about what we mean by holomorphic, and in particular about the distinction between holomorphicity on an open domain and continuity up to the boundary. [definition: Holomorphic Function on a Closed Domain] Let $\Omega \subset \mathbb{C}$ be a bounded domain with closure $\overline{\Omega}$. A function $f: \overline{\Omega} \to \mathbb{C}$ is said to be **holomorphic on $\overline{\Omega}$** (or: holomorphic on $\Omega$ and continuous on $\overline{\Omega}$) if: 1. $f$ is holomorphic on the open domain $\Omega$, meaning the complex derivative \begin{align*} f'(z) &= \lim_{h \to 0} \frac{f(z+h) - f(z)}{h} \end{align*} exists for every $z \in \Omega$. 2. $f$ is continuous on the closed domain $\overline{\Omega}$. We write $f \in \mathcal{O}(\Omega) \cap C(\overline{\Omega})$ to denote this class of functions. [/definition] The continuity condition at the boundary is not automatic from holomorphicity on $\Omega$ — a holomorphic function on an open disk need not extend continuously to the boundary circle. The condition $f \in \mathcal{O}(\Omega) \cap C(\overline{\Omega})$ is the natural setting for boundary-value problems and for comparing interior values with boundary values. [quotetheorem:491] The maximum is always attained because $\overline{\Omega}$ is compact and $|f|$ is continuous. What the theorem asserts is that the maximum cannot be a strictly interior point — it must occur on $\partial\Omega$. But the bounded-domain version above raises a question: what if we remove the compactness hypothesis and work on an arbitrary domain, not necessarily bounded, where no maximum is guaranteed to be attained? The stronger and more fundamental form of the principle answers this question: if $|f|$ achieves its supremum at even one interior point, the function is not merely pushed to the boundary — it is constant everywhere. This is a statement about the global rigidity of holomorphic functions, not about boundary comparisons. [quotetheorem:3345] The strong form is the heart of the matter: if the modulus achieves a local maximum at any interior point, the function is forced to be globally constant. The bounded-domain version follows by applying this to the interior of $\overline{\Omega}$. ## Interior Maximum via the Mean Value Property Why must a holomorphic function be constant if its modulus achieves an interior maximum? The key is the **mean value property** of holomorphic functions, which itself is a consequence of the Cauchy integral formula. Understanding this connection reveals why holomorphic functions are so rigid. Holomorphic functions satisfy a remarkable averaging property: the value at any point is exactly the average of its values on any circle centered at that point that fits inside the domain. This is what prevents the modulus from having an interior maximum — any local elevation in $|f|$ around a point $z_0$ would force $|f(z_0)|$ to be less than the average, contradicting a maximum. The mean value property is not merely an analogy with harmonic functions — it is a precise equality that follows from the Cauchy integral formula by parameterizing the circle $|z - z_0| = r$ as $z = z_0 + re^{i\theta}$. Every holomorphic function satisfies this exact averaging on every disk that fits inside the domain, and this is the mechanism that makes interior maxima impossible. [quotetheorem:3346] This is an exact equality, not an approximation. It says the center value must be balanced by the entire surrounding circle. If the modulus were already maximal at the center, the average could only keep the same size in a completely rigid situation: the nearby boundary values would have to line up in both size and direction. Varying the circle then forces the function to lose all local freedom, which is the mechanism behind the strong maximum principle. [remark: Constant Modulus Implies Constant Function] If $f \in \mathcal{O}(\Omega)$ and $|f|$ is constant on some open set $U \subset \Omega$, then $f$ is constant on $\Omega$. Geometrically, the image of $U$ is trapped on a circle centered at the origin. A nonconstant holomorphic function cannot have that kind of one-dimensional image on an open set: holomorphicity either opens up neighborhoods or collapses the function to a constant. Once constancy holds on a nonempty open set, the identity principle propagates it across the connected domain. [/remark] [illustration:mean-value-circle] ## The Open Mapping Theorem and a Second Perspective The Maximum Modulus Principle is also a direct corollary of the **Open Mapping Theorem**, which is one of the deepest structural facts about holomorphic functions. This perspective gives a second and illuminating proof. The Open Mapping Theorem tells us that non-constant holomorphic functions are geometrically expansive: they map open sets to open sets. This is in stark contrast to real differentiable functions, which can map open intervals to single points (via constant functions) or to closed intervals (via functions with zero derivative at the boundary of their range). [quotetheorem:358] From this viewpoint, the Maximum Modulus Principle becomes a geometric inevitability. A nonconstant holomorphic function sends neighborhoods to genuine two-dimensional neighborhoods in the target plane. Such an image cannot sit inside a closed disk while also touching its boundary from the inside at the supposed maximum value. Interior maxima are incompatible with openness; only constant functions can avoid that expansion. [explanation: Why Open Mapping Is Special to Holomorphic Functions] The Open Mapping Theorem fails for real differentiable functions: $f: \mathbb{R} \to \mathbb{R}$, $f(x) = x^2$, maps the open interval $(-1, 1)$ to $[0, 1)$, which is not open. The theorem's validity for holomorphic functions rests on the fact that a holomorphic function near a zero of order $k$ looks like $z^k$ after a coordinate change. At any point $z_0$ where $f'(z_0) \neq 0$, $f$ is locally a bijection by the inverse function theorem. At a critical point where $f'(z_0) = 0$, the function still opens up neighborhoods because the zero of $f'$ corresponds to a branch point of degree $k \ge 2$, and $z \mapsto z^k$ is still an open map — it maps a small disk around $0$ to a disk around $0$ (wrapping around $k$ times). This wrapping is possible in $\mathbb{C}^2$ dimensions but not in $\mathbb{R}$ (where $x \mapsto x^k$ for even $k$ folds the interval, failing to be open). [/explanation] ## Consequences and Applications The Maximum Modulus Principle is not an isolated theorem — it is a tool that appears throughout complex analysis, used to prove uniqueness, estimate growth, and compare functions. This section develops the most important consequences. ### The Minimum Modulus Principle Just as holomorphic functions cannot achieve an interior maximum of $|f|$, they also cannot achieve an interior minimum — provided the function has no zeros. If $f(z_0) = 0$ for some interior $z_0$, then $|f(z_0)| = 0$ is trivially a minimum. But if $f$ has no zeros on $\Omega$, then $g = 1/f$ is holomorphic on $\Omega$, and we can apply the Maximum Modulus Principle to $g$. [quotetheorem:3347] The hypothesis that $f$ has no zeros is essential. Without it, the result fails: $f(z) = z$ on $\overline{B}(0, 1)$ achieves its minimum $|f(0)| = 0$ at the interior point $z = 0$. ### Schwarz's Lemma One of the most elegant applications of the Maximum Modulus Principle is Schwarz's Lemma, which constrains holomorphic self-maps of the unit disk that fix the origin. The question is: if $f: B(0, 1) \to B(0, 1)$ is holomorphic and $f(0) = 0$, how large can $|f(z)|$ be, and how large can $|f'(0)|$ be? Since $f(0) = 0$, we can write $f(z) = z g(z)$ where $g$ is holomorphic. Schwarz's Lemma says $|g|$ is bounded by $1$ everywhere, meaning $f$ cannot be more expansive than the identity. [quotetheorem:368] Schwarz's Lemma is a rigidity statement about the unit disk. A holomorphic self-map that fixes the origin cannot push any point farther from the origin than it already was, and it cannot have derivative larger than one at the origin. The equality case is just as important as the inequality: only rotations of the disk preserve the bound sharply, so every other such map is genuinely contracting somewhere. [example: Schwarz's Lemma Cannot Be Improved] Consider $f(z) = z$ on $B(0, 1)$. This satisfies $f(0) = 0$, $|f(z)| = |z| \le 1$ for $z \in B(0, 1)$, and achieves $|f(z)| = |z|$ everywhere. So the inequality $|f(z)| \le |z|$ is sharp — it is attained by the identity map (and by all rotations $f(z) = e^{i\theta}z$). No non-rotation holomorphic self-map of the disk fixing the origin can expand distances. Now consider $f(z) = z^2$ on $B(0, 1)$. We have $f(0) = 0$ and $|f(z)| = |z|^2 \le |z|$ for $|z| \le 1$, with strict inequality for $0 < |z| < 1$. Here $|f'(0)| = 0 < 1$, confirming that the squaring map is strictly more contracting than any rotation. [/example] ### Hadamard's Three-Lines Theorem The Maximum Modulus Principle extends to unbounded strips, giving a powerful interpolation result. The **Three-Lines Theorem** is the complex-analytic engine behind the Riesz–Thorin interpolation theorem in harmonic analysis. Consider the vertical strip $S = \{z \in \mathbb{C} : 0 < \operatorname{Re}(z) < 1\}$. Suppose $f$ is holomorphic on $S$ and bounded and continuous on $\overline{S}$. Let $M(x) = \sup_{t \in \mathbb{R}} |f(x + it)|$ denote the supremum of $|f|$ on the vertical line $\operatorname{Re}(z) = x$. [quotetheorem:3348] The theorem says that the supremum of $|f|$ on any intermediate vertical line is bounded by the geometric mean of the suprema on the boundary lines, weighted by the position $x$. This is a quantitative form of the Maximum Modulus Principle for strips. The important idea is that boundary control on the two vertical sides propagates through the whole strip, but not linearly in $M(x)$ itself. The logarithm of the bound behaves convexly, so the estimate interpolates multiplicatively between the two boundary suprema. This is exactly the kind of statement the Maximum Modulus Principle is built for: reduce interior control to boundary control, even on a domain that is not compact. [illustration:three-lines-strip] [example: Applying the Three-Lines Theorem to Interpolate Bounds] Suppose $f$ is holomorphic on $S$, continuous on $\overline{S}$, bounded by $M_0 = 3$ on $\operatorname{Re}(z) = 0$ and bounded by $M_1 = 1/3$ on $\operatorname{Re}(z) = 1$. What is the bound on $|f|$ on the line $\operatorname{Re}(z) = 1/2$? By Hadamard's Three-Lines Theorem with $x = 1/2$: \begin{align*} M(1/2) &\le M(0)^{1 - 1/2} \cdot M(1)^{1/2} = 3^{1/2} \cdot (1/3)^{1/2} = \sqrt{3} \cdot \frac{1}{\sqrt{3}} = 1. \end{align*} So the supremum of $|f|$ on the middle line $\operatorname{Re}(z) = 1/2$ is at most $1$, even though the bounds on the left and right boundary lines are $3$ and $1/3$ respectively. The log-convexity interpolates between $\log 3 \approx 1.099$ and $\log(1/3) \approx -1.099$, giving $0$ at the midpoint, i.e., $M(1/2) \le e^0 = 1$. [/example] ### Phragmén–Lindelöf Principle The Maximum Modulus Principle applies to bounded domains, but what happens when the domain is unbounded? The naive extension fails: bounded domains guarantee that the supremum of $|f|$ is attained (by compactness), but on unbounded domains, the supremum may be a limit at infinity rather than an attained maximum. Nevertheless, a powerful extension holds for sectors and half-planes, provided the function does not grow too rapidly at infinity. [quotetheorem:3349] The growth condition (2) is essential. Without it, the principle fails completely. [example: Failure Without a Growth Condition] Consider the entire function $f: \mathbb{C} \to \mathbb{C}$ defined by \begin{align*} f(z) &= e^{e^z}. \end{align*} On the imaginary axis $z = iy$, we have $e^z = e^{iy} = \cos y + i \sin y$, so \begin{align*} |f(iy)| &= |e^{\cos y + i\sin y}| = e^{\cos y} \le e. \end{align*} Thus $|f|$ is bounded by $e$ on the imaginary axis (the boundary of the right half-plane). But at the interior point $z = n$ for large $n \in \mathbb{N}$: \begin{align*} |f(n)| &= |e^{e^n}| = e^{e^n}, \end{align*} which grows faster than any exponential. So $f$ is bounded by $e$ on the boundary of the right half-plane but is unbounded in the interior. The Maximum Modulus Principle fails for this function on the right half-plane because the domain is unbounded and $f$ grows super-exponentially. Under the sector convention above, a half-plane has opening angle $\pi$, so $\alpha = 1$; the Phragmén–Lindelöf hypothesis would require growth like $e^{|z|^\beta}$ with $\beta < 1$, but $e^{e^z}$ grows much faster along the positive real axis. [/example] [remark: The Role of the Growth Exponent] In the Phragmén–Lindelöf Principle for a sector of angle $\pi/\alpha$, the threshold exponent is $\alpha$. Functions growing like $e^{|z|^\alpha}$ on the sector are on the borderline: slightly slower growth ($e^{|z|^\beta}$ with $\beta < \alpha$) is controllable, while growth at the rate $e^{|z|^\alpha}$ or faster can violate the principle. The example of $f(z) = e^{e^z}$ on the right half-plane ($\alpha = 1$) shows that growth faster than every $e^{|z|^\beta}$ with $\beta < 1$ can defeat the principle entirely. [/remark] ## Uniqueness and Rigidity The bounded-domain principle does more than locate maxima: it turns boundary information into interior rigidity. If two holomorphic functions have the same boundary values, then their difference has boundary modulus zero. The natural question is whether that boundary vanishing can hide a nonzero holomorphic function in the interior, and the Maximum Modulus Principle says no. [quotetheorem:3350] This theorem is the boundary-value form of holomorphic rigidity. Once the values on $\partial\Omega$ are fixed, there is no room to alter the function inside the domain while preserving holomorphicity and continuity up to the boundary. It is the complex-analytic analogue of the uniqueness theorem for harmonic functions: a harmonic function on a bounded domain is uniquely determined by its boundary values, and the Maximum Modulus Principle is the tool that encodes the same rigidity for holomorphic functions. [example: Two Concrete Functions Forced to Coincide by Boundary Agreement] Let $\Omega = B(0, 1)$ (the open unit disk). Define \begin{align*} f(z) &= z^3 + iz, \\ g(z) &= z^3 + iz - 2i(z^2 - 1). \end{align*} Both $f$ and $g$ belong to $\mathcal{O}(B(0,1)) \cap C(\overline{B}(0,1))$ (they are polynomials). On the unit circle $|z| = 1$, write $z = e^{i\theta}$. Then $z^2 = e^{2i\theta}$, and: \begin{align*} g(e^{i\theta}) - f(e^{i\theta}) &= -2i(e^{2i\theta} - 1). \end{align*} At $\theta = 0$: $e^{2i \cdot 0} - 1 = 0$. At $\theta = \pi/2$: $e^{i\pi} - 1 = -2$, so $g - f = -2i(-2) = 4i \neq 0$. Thus $f$ and $g$ do **not** agree on the boundary, and indeed they differ in the interior too — for instance at $z = i/2$: \begin{align*} g(i/2) - f(i/2) &= -2i\left((i/2)^2 - 1\right) = -2i\left(-1/4 - 1\right) = -2i(-5/4) = 5i/2 \neq 0. \end{align*} Now modify $g$ so that it agrees with $f$ on the boundary. Set \begin{align*} g(z) &= z^3 + iz + h(z), \end{align*} where $h \in \mathcal{O}(B(0,1)) \cap C(\overline{B}(0,1))$ satisfies $h(e^{i\theta}) = 0$ for all $\theta$. By the Maximum Modulus Principle applied to $h$: \begin{align*} \max_{|z| \le 1} |h(z)| &= \max_{|z| = 1} |h(z)| = 0, \end{align*} so $h \equiv 0$ on $\overline{B}(0,1)$, and $g = f$ everywhere. The only holomorphic function on the closed disk that vanishes on the boundary circle is the zero function — every other holomorphic function leaves a non-zero trace on the boundary that distinguishes it from all others. [/example] The uniqueness theorem is qualitative — two functions that agree on the boundary are identical. But in practice one often needs a quantitative version: given that $|f|$ is bounded by some constant $M$ on $\partial\Omega$, what can we say about $|f(z_0)|$ at a specific interior point $z_0$? The answer is simply that $|f(z_0)| \le M$ — the interior bound is no worse than the boundary bound. This is the workhorse form of the Maximum Modulus Principle, and it is what analysts actually apply when they need to control holomorphic functions on domains. [quotetheorem:3351] This estimate is the form in which the Maximum Modulus Principle appears most often in practice. It reduces the task of bounding $f$ everywhere to bounding $f$ on the boundary alone — a much simpler region. ## Connection to Subharmonic Functions The Maximum Modulus Principle is not an isolated fact about holomorphic functions — it is the complex-analytic instance of a broader phenomenon governing **subharmonic functions**. Understanding this connection clarifies why the principle holds and extends it far beyond holomorphic functions. Recall that the real and imaginary parts of a holomorphic function are harmonic: they satisfy $\Delta u = 0$. A harmonic function satisfies the mean value property exactly, and the classical maximum principle for harmonic functions is the real-variable counterpart of our theorem. But the modulus $|f|$, unlike $\operatorname{Re}(f)$ or $\operatorname{Im}(f)$, is not harmonic in general — it satisfies a one-sided inequality that makes it subharmonic. What class of functions does this describe? [definition: Subharmonic Function] Let $\Omega \subset \mathbb{C}$ be a domain. A function $u: \Omega \to [-\infty, +\infty)$ that is upper semicontinuous is called **subharmonic** on $\Omega$ if for every $z_0 \in \Omega$ and every $r > 0$ with $\overline{B}(z_0, r) \subset \Omega$, \begin{align*} u(z_0) &\le \frac{1}{2\pi} \int_0^{2\pi} u(z_0 + re^{i\theta}) \, d\theta. \end{align*} That is, $u$ is dominated by the average of its values on every circle around each of its points. [/definition] A harmonic function is both subharmonic and superharmonic (the average inequality is an equality in both directions). For subharmonic functions, the same proof that establishes the Maximum Modulus Principle — using only the sub-mean-value inequality and the topology of connected sets — gives the maximum principle for subharmonic functions: a subharmonic function on a bounded domain achieves its maximum on the boundary. The connection to holomorphic functions is the following: if $f \in \mathcal{O}(\Omega)$, then $|f|^p$ is subharmonic on $\Omega$ for every $p > 0$, and $\log|f|$ is subharmonic wherever $f \neq 0$. In particular, $|f|$ itself is subharmonic, which is exactly why the Maximum Modulus Principle holds — the sub-mean-value inequality for $|f|$ is all that is needed to prevent interior maxima. The step we carried out explicitly (taking moduli of the mean value identity and applying the triangle inequality) is precisely the verification that $|f|$ is subharmonic. This perspective also explains the scope of the principle: any subharmonic function, holomorphic or not, obeys a maximum principle. The holomorphic case is special only in that holomorphicity is a clean sufficient condition for $|f|$ to be subharmonic, while for a general smooth function $|g|$ need not satisfy the sub-mean-value inequality. ## References Lars V. Ahlfors, *Complex Analysis* (3rd ed., 1979). Elias M. Stein and Rami Shakarchi, *Complex Analysis* (Princeton Lectures in Analysis, vol. II, 2003). Walter Rudin, *Real and Complex Analysis* (3rd ed., 1987). John B. Conway, *Functions of One Complex Variable I* (2nd ed., 1978).