From a prescribed divisor to a function
Suppose points in the complex plane escape every bounded set. We want an entire function whose zeros, including multiplicities, are exactly those points. A finite polynomial product solves the problem for finitely many points. For infinitely many, even factors approaching can have a product tending to zero everywhere. The construction must make the errors summable uniformly on every compact set, without demanding that the zero sequence grow at a prescribed rate.
Definition 1.2.1
(Zero divisor)
A zero divisor is a list of nonzero complex numbers, in which repetition records multiplicity, with no finite accumulation point, together with an integer recording the multiplicity at the origin. Consequently, every bounded disk contains only finitely many entries, and an infinite list satisfies .
We first construct a product from these data, then show that every nonzero entire function has this form up to an exponential factor.
Counterexample mode
Plain linear factors need not converge
The tempting claim is that converges whenever . Take and : the partial products are
and at the partial products tend to , although is not a listed zero. Escape to infinity alone does not make finite. The repair is either the extra hypothesis , or exponential correction factors, which work for every zero divisor.
Elementary factors and a uniform estimate
The correction factors are the following.
Definition 1.2.2
(Weierstrass elementary factors)
For an integer , put
Each factor is entire and has exactly one zero, at , where it is simple.
The exponential cancels the first Taylor terms of , so is very close to near .
Lemma 1.2.3
(Uniform estimate for elementary factors)
There is a constant , independent of , such that
Proof
(Bound the logarithm, then exponentiate)
For , the logarithm of near satisfies
Here ; the exponential estimate therefore gives , so works.
The uniformity in is what permits the correction order to grow along the sequence. It is not enough merely to observe that each individual tends to as .
Constructing the product
Theorem 1.2.4
(Weierstrass product theorem)
Let be a finite or infinite sequence of nonzero complex numbers with in the infinite case, repeated according to the desired finite multiplicities. For any integer , there is an entire function whose zeros are exactly with multiplicity and the with their listed multiplicities. In the infinite case one choice is
Proof
(Summable errors on every disk)
Fix a disk . Because , for all sufficiently large we have throughout that disk. The lemma gives
The product consequently converges uniformly on the disk, and since was arbitrary, locally uniformly on the plane. Its limit is entire. On any compact set containing no listed zero, a sufficiently late product tail is nonzero by the locally uniform infinite-product criterion; the finite initial product is also nonzero there. Thus no extra zeros arise. Near a listed point , only finitely many factors vanish there. Isolate those factors: each contributes one simple zero, while all the other factors form a nonvanishing holomorphic function near . The product therefore has exactly the prescribed multiplicity. The factor gives the specified order at zero. For finitely many , an ordinary finite product suffices.
Proof X-ray
Why the order n makes the errors summable
Two quantities compete in each factor: the ratio , which is eventually at most on a fixed disk but may approach arbitrarily slowly, and the exponent in the lemma. Choosing the order makes the error at most no matter how slowly grows; this is the key move, and it is possible only because does not depend on . The hypothesis is used exactly once, to put all but finitely many ratios inside .
Summability of is the criterion that gives both locally uniform convergence and the absence of new zeros: a uniformly convergent product with summable errors vanishes only where a factor vanishes. The finitely many early factors are handled separately, which is why the argument never needs an estimate near .
This construction deliberately uses even when a smaller order would work. It proves existence for any allowed sequence, rather than claiming this product is the shortest or best for a particular growth problem.
Factoring an existing entire function
Theorem 1.2.5
(Weierstrass factorization theorem)
If is an entire function not identically zero, its nonzero zeros can be listed as above, and there is an entire for which
with the finite product when there are finitely many nonzero zeros. Here is the order of the zero at , or when .
Proof
(A zero-free quotient has an entire logarithm)
The identity theorem makes every zero isolated; compactness then shows that the nonzero zeros, counted by multiplicity, have no finite accumulation. Construct from those zeros. At every zero of , the local Taylor factorizations of and have equal powers, so has a removable singularity with nonzero value. The extended quotient is entire and zero-free. Its logarithmic derivative has an entire antiderivative , since the plane is simply connected. Thus is a nonzero constant; absorbing a logarithm of that constant into gives the asserted .
Common mistake
(Zeros determine an entire function only up to )
The exponential factor cannot be recovered from the zeros. For instance, and have identical zeros and multiplicities but different values and growth. If the product is fixed, two choices of differ by a constant integer multiple of : their difference is a continuous function taking values in the discrete set on a connected plane.
Worked examples and a misconception
Worked example 1.2.6
(Positive integers with a fixed-order product)
Since , the smaller fixed-order product already converges locally uniformly: the lemma with bounds each error by once . It has a simple zero at every positive integer and no other zero. The general theorem also gives a product using ; the quotient of these two constructions is a zero-free entire function. It need not be constant.
Worked example 1.2.7
(A repeated zero and an exponential)
The function has a double zero at and a triple zero at . In this finite case the factorization is immediate. Since and ,
so , and , with unique up to adding . The multiplicity three is encoded by three copies of the zero , not by a single unmarked entry.
Worked example 1.2.8
(A forbidden specification)
The sequence cannot be the entire zero set of a nonzero entire function: its zeros would accumulate at within the domain, and the identity theorem would force the function to vanish everywhere. This is the edge of the hypothesis . By contrast, an infinite zero sequence escaping to infinity poses no obstruction, however slowly it escapes: for the theorem still applies, although diverges for every , so no fixed-order product is covered by the lemma.
Practice and solutions
Quick check
(Why does have no nonzero zeros, and what is the order at ?)
Where can an exponential vanish?
Solution · Answer
The exponential never vanishes, including at zero, so the only zero is , with order four.
Quick check
(Construct an entire function with a double zero at each , , and no other zeros.)
Here converges.
Solution · Answer
The product converges locally uniformly because . Separating each vanishing factor from the nonzero tail proves the stated double zeros and excludes others.
Quick check
(If nonzero entire functions and have exactly the same zeros with multiplicities, prove that after removable singularities are filled in.)
Repeat the last step of the factorization proof.
Solution · Answer
Local factorization cancels equal zero orders, making the quotient zero-free and entire. Integrate its logarithmic derivative on and absorb the integration constant as in the proof of the factorization theorem.