Liouville, Morera, maximum modulus principle
Why this is asked: Recognise Liouville in disguise: bounded, bounded real part, bounded on a growth scale, or missing two values. Maximum modulus turns interior bounds into boundary bounds.
In one sentence
Holomorphic functions are rigid: bounded on all of forces constant, a maximum of inside forces constant, and the converse of Cauchy's theorem holds under a hypothesis weak enough to be useful.
Why the exam asks it
Because each theorem has a hypothesis that sounds like a formality and is not — "entire", "bounded on ", "on a domain", "continuous" — and because Liouville is routinely misapplied to functions bounded only on the real line. is the standard trap and it is set most years in some form.
The idea
Liouville
A bounded entire function is constant.
From Cauchy's estimate with : for every , and letting gives .
Both words carry weight. Entire means holomorphic on all of — is bounded outside the unit disc and is not entire. Bounded means bounded on , not on some subset.
is entire, satisfies for every real , and is not constant. There is no contradiction: on the imaginary axis , which grows like . At the modulus already exceeds . Bounded on is not bounded on , and the option will rely on you not noticing which is claimed.
The useful strengthenings:
- for large and entire is a polynomial of degree at most . Cauchy's estimate on .
- bounded above constant. Apply Liouville to , using .
- entire and non-constant is dense in . That is the little Picard theorem's weaker cousin and follows from Liouville applied to .
Fundamental theorem of algebra: if a non-constant polynomial had no zero, would be entire and bounded — it tends to at infinity — hence constant. Contradiction.
Morera
If is continuous on a domain and for every triangle , then is holomorphic on .
The converse of Cauchy's theorem, and the hypotheses are exactly what stop it being trivially false. Continuity is required. Vanishing on every triangle is required — one contour, or one family of contours, is not enough. Its usual role is proving that a limit of holomorphic functions is holomorphic, since uniform convergence on compact sets passes through the integral.
Maximum modulus
If is holomorphic and non-constant on a domain, has no local maximum in that domain.
So on a bounded domain with continuous up to the boundary, is attained on the boundary. On , reaches on the rim and nowhere inside.
Two corollaries worth having:
- Minimum modulus: has no interior minimum unless has a zero there. The exception is the whole content — apply the maximum principle to , which needs .
- and obey the same principle, via and .
The mechanism is the open mapping theorem: a non-constant holomorphic function maps open sets to open sets, so the image of a neighbourhood of is a neighbourhood of and contains points of larger modulus. Nothing can be a local maximum.
Schwarz's lemma
holomorphic with . Then and ; equality in either, at a single point, forces with .
Proof is the maximum principle applied to , extended holomorphically across . The rigidity — one equality anywhere pins the whole function — is the examinable part.
Where intuition breaks
" is entire and on , so is constant." . Liouville needs bounded on . The single most reliable trap in this topic.
" is bounded and holomorphic, so constant." Needs entire. on is bounded and holomorphic and not constant; so is any bounded holomorphic function on the disc.
" is entire and non-constant, so ." Only for polynomials. is entire, non-constant, and tends to along the negative real axis. What Liouville gives is unboundedness, not divergence.
"A bounded harmonic function on need not be constant." It is constant — the harmonic Liouville theorem. True, and worth recognising among false options.
" attains its minimum on the boundary." Only if has no zero inside. on attains its minimum at the centre.
"Morera needs for every closed contour." Triangles suffice, which is what makes it usable. And continuity is required — options drop it.
" for one contour, so is holomorphic." Not Morera. integrates to round any contour not enclosing the origin and is nowhere holomorphic.
"Schwarz's lemma applies to any map of the disc to itself." It needs . Without that fixed point the correct statement is the Schwarz–Pick inequality.
The exam's angle
- When Liouville is invoked, check where the bound holds. On , on a half-plane, on a strip — none of these is , and or breaks each.
- Growth conditions mean Cauchy estimates. gives a polynomial of degree ; count the degree from the exponent.
- A bounded real part means consider . This trick answers a whole family of options.
- For maximum modulus, look at the boundary and nowhere else — unless the question is about a minimum, in which case first ask whether has a zero.
- If a question says a limit of holomorphic functions is holomorphic, the tool is Morera, with uniform convergence on compact sets.
The night before
- Liouville: bounded entire constant. is bounded on and unbounded on — .
- and entire polynomial of degree .
- bounded constant, via and .
- Fundamental theorem of algebra from Liouville applied to .
- Morera: continuous, and on every triangle holomorphic. Used for limits of holomorphic functions.
- Maximum modulus: no interior maximum for non-constant ; the maximum is on the boundary.
- Minimum modulus: no interior minimum unless vanishes there. on the disc.
- Open mapping: non-constant holomorphic maps open to open. That is why no interior maximum exists.
- Schwarz: , , ; equality forces a rotation.
- Entire and non-constant gives unbounded, not : .
See it move
The trap here
“A bounded holomorphic function on an unbounded domain is constant” — false
f(z) = eᶻ on {Re z < 0}
|eᶻ| < 1 there, but f is not constant. Liouville needs the whole plane.
More in Cauchy Theory
Take it further
- The mistake this prevents: standard counterexample → real wrong options, and why each one tempts
- The mistake this prevents: boundary and endpoint → real wrong options, and why each one tempts
- Previous-year questions set on this — every option and the answer, free:June 2025 · Part BJune 2025 · Part BDecember 2025 · Part CDecember 2025 · Part CJune 2025 · Part CJune 2024 · Part BDecember 2024 · Part CJune 2024 · Part C
Last revised . Found a mistake? Tell us — corrections are the fastest thing we act on.
Next: Laurent series, classification of singularities, Casorati–Weierstrass
Create a free account to keep your place and have this feed your study plan.
Open this in the full syllabus view · Unit 2