At an isolated pole, the negative powers in the Laurent expansion describe the local singularity. Can such local data be assembled into one meromorphic function? Simply adding the principal parts is tempting, but a sum of terms of size can diverge even far from every pole. The remedy is to subtract a carefully chosen polynomial from each term. The subtraction preserves each term's principal part, and suitable choices make the tail uniformly small on each compact set.
These local data are made precise as follows.
Definition 1.11.1
(Meromorphic function and principal part)
A meromorphic function on is a function holomorphic except at isolated poles; the pole set has no finite accumulation point. A principal part at is a finite nonzero sum
Its residue is .
Common mistake
(Residues do not determine principal parts)
Prescribing only residues leaves the higher negative-power coefficients undetermined. For example and both have a single pole at with residue , yet their difference still has a pole there. Uniqueness up to an entire function requires equality of every coefficient .
Why correction terms are needed
Worked example 1.11.2
(A divergent sum repaired one term at a time)
Suppose a simple pole of residue is wanted at every positive integer. The naive series diverges at : its terms sum to the negative harmonic series. Instead put
On and , the absolute value of a summand is at most . The Weierstrass -test gives uniform convergence of the tail there. Around any positive integer , choose a disk containing no other positive integer. Its th term has principal part ; the remaining terms define a holomorphic function there by local uniform convergence. Thus (1) is meromorphic with exactly the desired principal parts, and .
The correction is the negative of the constant Taylor coefficient of at zero. More complicated pole data may require more Taylor coefficients, chosen separately for each pole.
How many coefficients must be subtracted depends on how fast the principal parts grow.
Worked example 1.11.3
(Growing residues need a linear correction)
Suppose the principal part , of residue , is wanted at each positive integer . At zero, . Subtracting only the constant term leaves , whose modulus on is about for large : a divergent harmonic tail. Subtracting the linear Taylor polynomial gives
which is at most on when . The series of these terms therefore converges locally uniformly off the positive integers and has the required principal parts. The edge case of constant residues, as in (1), needs only the constant correction.
The entire-plane theorem
Theorem 1.11.4
(Mittag–Leffler for the complex plane)
Let have no finite accumulation point. For each prescribe a nonzero principal part at . There is a meromorphic function on whose poles are exactly the points of and whose principal part at each is .
Any two such functions differ by an entire function. If is empty, every entire function is a solution; for finite , the finite sum of the principal parts is one solution.
Proof
(Taylor subtraction makes the tails small)
A set with no finite accumulation point meets each closed disk in finitely many points. It is therefore finite or countable. In the infinite case enumerate it as . Every such enumeration satisfies , since each disk contains only finitely many points. Write for the prescribed principal part at .
For each , is holomorphic in and has a pole at , so its Taylor series there has radius exactly . Its Taylor polynomials at zero converge uniformly on the smaller closed disk . Choose a polynomial such that
If one point is , set its correction ; at most one index is involved. Define
Fix a compact set . For all sufficiently large , , so (2) bounds the corresponding terms uniformly by . Hence the tail converges uniformly on . Around any point outside , take a closed disk disjoint from . The finitely many remaining terms are holomorphic on a neighborhood of that disk, and the Weierstrass theorem makes the sum holomorphic on its interior.
For a fixed , choose a disk around it containing no other point of . Separate the th summand in (3). Take a smaller closed disk about avoiding the other points of , and choose such that for . The finitely many terms with , , are holomorphic near ; the remaining terms are holomorphic there and bounded by . Thus their sum is holomorphic near . Since is entire, the principal part at is exactly ; its highest coefficient is nonzero, so is indeed a pole. Therefore the function defined off extends with precisely these poles to a meromorphic function on , proving existence.
If and have the same prescribed principal parts, their difference has no negative Laurent coefficients at any . Every such singularity is removable, so extends to an entire function. Conversely adding an arbitrary entire function does not change any principal part or introduce a pole.
Proof X-ray
Where each hypothesis enters the construction
The hypothesis that has no finite accumulation point is used twice: it makes countable with , so every compact set lies in the disks for all but finitely many ; and it lets each be isolated in a disk containing no other prescribed point, where its own term can be separated. The key move is (2): because the Taylor series of at has radius exactly , it converges uniformly on the half-radius disk, so a polynomial can approximate there to within any prescribed error, and the summable choice is possible.
Two features close the argument. Subtracting a polynomial changes no principal part, because is entire. Uniform convergence on compact sets, not merely pointwise convergence, is what lets the Weierstrass theorem conclude that the infinite remainder is holomorphic.
Symmetry can provide a simpler correction
Worked example 1.11.5
(Poles at all integers)
For define
For and , a paired term has modulus at most . Thus the series is locally uniformly convergent away from the integers. Separating one singular term shows that has a simple pole of residue at every integer and no other poles.
The pairing also makes odd. It is -periodic: for the symmetric partial sum ,
Pass to the limit at and , neither an integer. In particular, oddness and periodicity give . This constructs one solution; adding any entire function produces another with the same principal parts, although the added function need not preserve periodicity or oddness.
Common mistake
(The symmetric pairing is part of the definition)
It is tempting to write as an unordered sum. That series is not absolutely convergent: for fixed , behaves like . The convergent object is the paired series (4), equivalently the limit of the symmetric partial sums . Summing from to instead adds , and gives .
The paired sum is legitimate because the pole set is discrete; without that, no correction can help.
Counterexample mode
Poles cannot accumulate at a finite point
The points cannot all be poles of a meromorphic function on . They accumulate at zero, but a meromorphic function is either holomorphic near zero or has one isolated pole there; either alternative gives a neighborhood containing no other poles. Thus the theorem's no-finite-accumulation hypothesis cannot be discarded.
Practice
Quick check
(Construct a meromorphic function with a simple pole at of residue , a double pole at with principal part , and no other poles. Describe all solutions.)
Add the two principal parts.
Solution · Answer
One solution is . Every other solution is this function plus an arbitrary entire function. The double pole has residue zero because its principal part contains no term.
Quick check
(Explain why defines a meromorphic function with a double pole at each positive integer. What is its residue at each pole?)
Use the Weierstrass -test on .
Solution · Answer
On and , . The -test controls the tail locally uniformly. Separating the term at gives principal part , whose residue is zero.
Quick check
(For the function in (1), compute .)
Telescope the partial sums.
Solution · Answer
At , the th summand equals . The partial sums telescope to , so .
Quick check
(Suppose two meromorphic functions have the same poles and residues. Must their difference be entire? Give a counterexample or prove it.)
Compare all principal-part coefficients, not only residues.
Solution · Answer
No. Take and . Both have the single pole at and residue zero, but still has a pole. Equality of all principal-part coefficients is needed.
Quick check
(If and a simple pole of residue is wanted at each point, show that the uncorrected sum already converges locally uniformly off .)
Compare with a geometric series.
Solution · Answer
On , once we have . This geometric bound gives uniform convergence of the tail. Separating the single term at a pole gives the stated simple principal part, and away from the poles the sum is holomorphic.