The Weierstrass factorization notes show that a zero-free entire function is for some entire . That description allows to be an arbitrary entire function. If grows with finite order, can still be transcendental? The answer is no: must be a polynomial. The proof illustrates a useful way to recover control of an analytic function from an upper bound on its real part.
This is a finite-zero case of the relationship between growth and factorization. It does not prove the general Hadamard factorization theorem for infinitely many zeros.
A real-part bound controls a smaller disk
Theorem 1.5.1
(Borel–Carathéodory estimate)
Let be holomorphic on a neighborhood of , and let . Put . Then
In particular, the nonnegative difference on the right controls the imaginary as well as real part of on the smaller disk.
Proof
(Map a half-plane into a disk)
Set and . If , the harmonic function has an interior maximum zero and is constant; the Cauchy–Riemann equations then make constant, and .
Suppose . Since , the denominator in
is nonzero, and because . Also . Schwarz's lemma on the radius- disk gives . Solving for and using ,
For , Schwarz's bound gives , so ; the last inequality uses that is increasing on . This proves the estimate.
Proof X-ray
Why a one-sided bound is enough
The only hypothesis on is one-sided: . For an arbitrary real-valued function this would say nothing about large negative values; holomorphy is what couples the real and imaginary parts. The key move is the map , which sends the half-plane into the closed unit disk and fixes ; the normalization is exactly what Schwarz's lemma needs.
The proof closes by inverting that map, and this is where the smaller disk is used: on Schwarz gives , which keeps away from zero. As the margin disappears and the constant diverges. The case is split off only because the map is undefined there.
The zero-free classification
For nonconstant entire , use the growth order from the preceding notes,
for sufficiently large , with nonzero constants assigned order zero. If , then for every , for sufficiently large when is nonconstant. A constant satisfies the weaker bound , which is all the proof needs.
Theorem 1.5.2
(A zero-free finite-order function has a polynomial logarithm)
If is zero-free, entire and of finite order , then
for a polynomial with when is nonconstant. In fact, if has positive degree , then ; if is constant, is constant and has order zero.
Proof
(From growth of an exponential to a polynomial exponent)
Because has no zeros, is entire. Integrating it on the simply connected plane yields an entire logarithm with . Consequently .
Fix . By the maximum modulus principle and the growth bound, for sufficiently large ,
where can depend on . Apply Borel–Carathéodory with outer radius and inner radius . It follows that . For any integer , Cauchy's estimate gives
For every integer , choose with . The coefficient of in vanishes. Thus is a polynomial of degree at most .
If with and , then for ,
The upper bound follows by the triangle inequality; for the lower bound choose the angle making positive real. Therefore has the displayed asymptotic and the order is exactly . A constant gives a nonzero constant .
Removing finitely many zeros
Theorem 1.5.3
(Finite zeros force an exponential polynomial)
Let be a nonzero entire function of finite order with only finitely many zeros, counted with multiplicity. Then
where is a nonzero polynomial recording precisely those zeros and is a polynomial. If is nonconstant, ; if is constant, is a nonzero polynomial and has order zero. In particular, the order of a finite-zero entire function is either zero or a positive integer.
Proof
(Dividing by the zero polynomial preserves order)
Take to be the finite product of the zeros with their multiplicities, including any zero at the origin; use when there are none. Local Taylor factorization shows extends to an entire zero-free function.
If has degree and leading coefficient , then for all sufficiently large and every ,
Thus , so has finite order no greater than . If is nonconstant, and the displayed bound gives ; taking double logarithms where is nonconstant gives the order comparison. If is constant, its order is zero by convention. The zero-free theorem gives with polynomial. If is nonconstant of degree , the same two-sided estimate shows that differs from by at most , so . If is constant, is a nonzero polynomial, whose order is zero.
Examples and limits of the theorem
Worked example 1.5.4
(Classify a specified finite zero divisor)
Suppose an entire function of order at most has exactly a double zero at and a simple zero at , and no other zeros. Then
for some . Conversely every such function has precisely those zeros and order at most one. It has order one when and order zero when . The zero data fix the polynomial factor only up to a nonzero constant; that constant can be absorbed into .
Worked example 1.5.5
(Sublinear order and an omitted value)
Let be entire of order strictly below one and suppose it omits . Then is zero-free and has order strictly below one. Indeed, the two inequalities show that the orders agree for nonconstant , since ; the constant case is immediate. The theorem makes its logarithm constant, hence is constant. Equivalently, every nonconstant entire function of order below one assumes every complex value.
The strict inequality is necessary: has order one and omits zero. This conclusion is restricted to order below one and is not a proof of Picard's theorem.
Worked example 1.5.6
(Zero-free periodic functions of finite order)
Let be zero-free, entire and of finite order, with . The theorem gives with a polynomial. Then , so the entire function takes values in ; being continuous on the connected plane, it is a constant . If with leading coefficient , then has degree and leading coefficient , so it is not constant. Hence
For the order is one; in the edge case , is constant. Without the finite-order hypothesis this fails: is zero-free and -periodic, but its logarithm is not a polynomial.
Counterexample mode
An arbitrary entire logarithm can have infinite order
The zero-free function has an entire logarithm , but its order is infinite. Along ,
As , and , which grows faster than . Thus the finite-order hypothesis cannot be omitted.
Common mistake
(Finite order does not mean finitely many zeros)
A finite order alone does not imply finitely many zeros: has order one and infinitely many zeros. The finite-zero theorem cannot replace the canonical-product part of a general factorization. Likewise has order two and infinitely many zeros, at and , so an integer order alone does not force a function of the form .
Exercises
Exercise 1.5.7
(Let be zero-free entire with . Classify its possible form.)
Core
Solution(3)
- 1
Bound the order
The bound gives , so .
- 2
Apply the zero-free classification
Key step
Then with , so for constants .
- 3
Match the explicit constant
The bound holds for every exactly when : the maximum over of is , attained at .
Because: On the maximum of is .
Exercise 1.5.8
(Let be entire, of order less than two, with exactly three simple zeros and no others. Give all possible forms and their possible orders.)
Foundation
Solution(3)
- 1
Remove the zeros
The zero polynomial is , and the finite-zero theorem gives with a polynomial.
- 2
Bound the degree of
Key step
Since the order is strictly below two, .
Because: If is nonconstant, .
- 3
List the functions
; its order is one if and zero if . All these functions meet the hypotheses.
Exercise 1.5.9
(Prove that a nonconstant entire function of order has infinitely many zeros. Can it omit a complex value?)
Core
Solution(3)
- 1
Rule out finitely many zeros
The order is neither, so there are infinitely many zeros.
Because: A finite-zero entire function has order zero or a positive integer.
- 2
Transfer an omitted value to a zero-free function
Key step
If omitted , then would be zero-free of order exactly .
Because: and .
- 3
Conclude
By the worked example on sublinear order, and hence would be constant, a contradiction. So assumes every complex value.
Exercise 1.5.10
(Show directly that has order two and no zeros. Why does being negative in some directions not reduce its maximum-modulus order?)
Foundation
Solution(3)
- 1
No zeros
The exponential is never zero.
- 2
Two-sided bound on
Key step
, while at it equals . Thus .
- 3
Read off the order
, so the order is two. Maximum modulus selects the largest direction; decay elsewhere does not remove that growth.