A nonconstant entire function can be periodic: repeats after . Two independent complex periods are much more restrictive. They confine the whole function to a bounded fundamental parallelogram, so an entire doubly periodic function must be constant. Allowing poles produces a rich family of meromorphic functions, whose zeros and poles are constrained by the residue theorem.
Definition 1.16.1
(Lattice and elliptic function)
Let be linearly independent over . The lattice they generate is
An elliptic function for is a meromorphic function on satisfying for every .
The lattice need not be the full period group of : a constant function, for instance, has every complex number as a period.
Geometry of a lattice
The real-linear map is invertible. By compactness of the unit circle in , there is such that
Thus is discrete, and any two distinct lattice points are separated by at least . A fundamental parallelogram can be written
Its translates cover the plane. Opposite sides differ by a period, with opposite orientations in the positively oriented boundary.
Theorem 1.16.2
(Residues and orders in a period cell)
Let be elliptic for . Choose a translate whose boundary contains no poles of . The sum of residues of in is zero. If is nonconstant, choose the boundary to avoid its zeros as well; then the number of zeros in , counted with multiplicity, equals the number of poles, counted with order.
An entire elliptic function is constant. In particular, an elliptic function cannot have exactly one simple pole modulo and no other poles modulo .
Proof
(Cancel opposite sides)
The pole set of a meromorphic function is discrete; it has finitely many representatives in a bounded union of period cells. A small generic translation of avoids these points on the boundary. For a nonconstant function, zeros are also isolated and have finitely many representatives, so the boundary can avoid them too.
Traverse the boundary counterclockwise regardless of the orientation of the chosen basis. Opposite sides are then traversed in opposite directions, and periodicity makes their integrals cancel. The residue theorem therefore gives the first assertion. For the second, is elliptic. At a zero of order it has residue , and at a pole of order it has residue , as follows by factoring a nonvanishing holomorphic function from the local power. Applying the first assertion to proves equality of the two counts. A boundary avoiding zeros and poles selects exactly one representative of each class modulo , so these counts do not depend on the chosen cell.
If is entire, it is bounded on the compact closure of one cell and hence everywhere by periodicity. Liouville's theorem makes it constant. A single simple pole has a nonzero residue, contradicting the zero sum of residues.
Proof X-ray
Orders are counted, not merely points
Periodicity is used exactly once, on the boundary: opposite sides of are translates of each other by or and are traversed in opposite directions, so without any information about inside the cell. Everything else makes the residue theorem applicable. The small translation of keeps poles, and later zeros, off the path; non-constancy is what makes zeros isolated, so that is meromorphic.
The second assertion is not proved afresh. It is the first assertion applied to the auxiliary elliptic function , whose residues are signed orders. That substitution is the step that turns “residues sum to zero” into “zeros equal poles”, and it explains why multiplicities, not locations, are counted: a double zero contributes and a triple pole . The constancy statement needs only compactness of the closed cell and Liouville's theorem.
A convergent construction
The natural sum of double poles is not suitable without correction. Define instead
The correction terms improve the decay enough to make the series absolutely and locally uniformly convergent. This is essential: we will later regroup terms by lattice translations.
Common mistake
(Summing the double poles without the correction terms)
For there are lattice points with , each of modulus at most . Their contribution to is at least , and diverges. Once we have , so is not absolutely convergent either, and its value would depend on the order of summation. Subtracting is what brings the terms down to size .
Theorem 1.16.3
(The Weierstrass ℘ function)
The series (2) defines an even elliptic function with a double pole at each lattice point and no other poles. Its derivative is
with absolute local uniform convergence away from .
Proof
(Convergence first, periodicity second)
The number of lattice points with is at most . To see this, surround each point by a disjoint disk of radius using (1); these disks lie in an annulus of area . For the shell , use a disk rather than an annulus: it contains finitely many nonzero points, each at least from zero. Summing the bounds for shells consequently gives .
For and ,
The -test proves local uniform convergence of (2) after the finitely many nearby terms are separated. Near any lattice point, only its own term has a pole, and its principal part is . Termwise differentiation is valid for locally uniformly convergent holomorphic series; the derivative series (3) is also absolutely locally uniformly convergent by the same lattice estimate.
Reindexing the absolutely convergent series (3) shows that is -periodic. Therefore has derivative zero on for every . The complement is connected: a line segment between two nonlattice points meets only finitely many lattice points and can be detoured around them. Hence the difference is constant there; its poles also cancel, so it extends to the same entire constant . The original series is even: pair the terms at and , justified by absolute convergence. Also . Evaluating the shift relation at and using evenness gives , so . Thus is periodic.
Half-period critical points
Theorem 1.16.4
(Three critical points modulo the lattice)
Modulo , the zeros of are exactly
and each is simple. Their -values are pairwise distinct.
Proof
(Oddness supplies three zeros; the pole count supplies no more)
The derivative is odd and periodic. At each listed half-period , , so periodicity and oddness give ; hence it vanishes. The three points are distinct modulo and none belongs to .
There is one triple pole of modulo , from the derivative of ; in particular is nonconstant. The zero/pole count therefore gives exactly three zeros counted with multiplicity. The three displayed zeros exhaust the count and are simple.
If two half-periods had the same value , then would have a zero of order at least two at each, because its derivative vanishes there. That would give at least four zeros counted with multiplicity, but the nonconstant function has only one double pole per cell and hence exactly two zeros. This contradiction proves the values distinct.
Counterexample mode
Equal counts do not mean equally many distinct points
A tempting reading of the zero/pole count is that an elliptic function has as many distinct zeros as distinct poles in a cell. The function refutes it. Modulo it has a single pole location, the lattice point , where the pole has order ; by the theorem just proved it has three distinct simple zeros at the half-periods. The counts with multiplicity agree, , while the numbers of locations are and . The correct statement counts zeros with multiplicity and poles with order.
Worked examples
Worked example 1.16.5
(Solving )
Let . The function is nonconstant and elliptic with one double pole per cell, so it has exactly two zeros modulo , counted with multiplicity. One zero is ; since is even, is another.
If , then and are distinct modulo , so they are the two zeros and both are simple. If , then is congruent to one of the three half-periods and ; there , so the zero at is double. In both cases
Worked example 1.16.6
(The square lattice and the value )
Take , , so and . For write each as ; then runs over , and absolute convergence permits the reindexing:
At we have , so and . Also . The three half-period values are pairwise distinct, so . The values are therefore with .
Worked example 1.16.7
(One double pole per cell, with residue zero)
Let , and . By the first example, the only zero of modulo is the double zero at , so has exactly one pole per cell, of order . The residue theorem for cells forces its residue to be ; we confirm it directly. Since ,
so is an even function of . Its Laurent series in has only even powers, and the coefficient of vanishes. The zero count also checks: has a double zero at each lattice point, where has its double pole, so two zeros match two poles.
Practice
Quick check
(Why does a lattice have only finitely many points in any bounded disk?)
Solution · Answer
By (1), a lattice point in has . There are only finitely many integer pairs in that disk.
Quick check
(Why must a nonconstant elliptic function have at least one pole in a period cell?)
Solution · Answer
If there were no poles in a cell, periodicity would imply no poles anywhere. The function would be entire and elliptic, hence constant by Liouville's theorem, a contradiction.
Quick check
(How many zeros does have modulo , counted with multiplicity, for any ?)
Solution · Answer
Exactly two. The nonconstant elliptic function has one double pole modulo and no other poles, so the zero/pole count gives two zeros with multiplicity.
Quick check
(At a half-period listed above, what is the order of the zero of ?)
Solution · Answer
The first derivative vanishes at , and the zero of there is simple, so . Its Taylor expansion therefore begins with a nonzero multiple of : the order is exactly two.
Quick check
(Show that has one triple pole modulo and residue zero at that pole.)
Solution · Answer
Near , a holomorphic function, so a holomorphic function. It has a triple pole and no coefficient. Periodicity transfers this local form to every lattice point, all equivalent modulo .