A coefficient that survives integration
Definition 5.3.1
(Residue)
If has a Laurent expansion around an isolated singularity ,
its residue at is .
On a small positively oriented circle lying in the Laurent annulus, the series converges uniformly and may be integrated term by term. Every other Laurent term integrates to zero, while integrates to . This is why one coefficient can summarize a contour integral.
Proposition 5.3.2
(Residue at a pole of order )
Let have a pole of order at , and write with holomorphic near and . Then
Proof
(Read off one Taylor coefficient)
Near , . Dividing by gives the Laurent series of , in which the power comes only from . Its coefficient is .
A simple pole is the case : the residue is . Formula (1) assumes a pole, not an arbitrary essential singularity.
A residue theorem on a disk
Theorem 5.3.3
(Residue theorem on a disk)
Suppose is holomorphic on a neighborhood of the closed disk except at finitely many poles strictly inside it. Then
Proof
(Subtract the principal parts)
At each subtract its finite principal part. The resulting function has a removable singularity at every , so it extends holomorphically to a neighborhood of the closed disk. Cauchy's theorem makes its integral around the boundary zero. For each principal part, terms with have the single-valued primitive away from , so their integrals vanish. For the simple term, Cauchy’s integral formula gives because . Summing proves (2).
Proof X-ray
Why higher pole terms do not directly contribute
A higher-order pole may have a nonzero residue, but only its coefficient survives the contour integral. The other negative powers have single-valued primitives. Formula (1) finds the surviving coefficient by differentiating the regular factor .
Worked example: a double pole
Worked example 5.3.4
( on a half-disk)
Justifying the half-disk computation. The residue theorem was proved above only for disks, so the contour needs its own argument. For let run along the real axis from to and then counterclockwise along , . Call this arc . Write with . Since and , the principal part of at is
The regular part. Expand in its Taylor series about , which is valid for . This shows that agrees near with a power series, so the singularity is removable. Hence extends holomorphically to . This set is convex, excludes , and contains .
Fix and set . For the triangle lies in . Goursat's theorem from the preceding note then gives , and continuity of yields . Because is closed, .
The principal part. On the double-pole term has primitive , so its integral over is . The simple-pole term is computed directly. On the segment,
On use . The first term gives . For the second term, and has length , so its integral has modulus at most . Hence , and
The arc and the real integral. On we have , so . Subtracting gives
Only these limits have been established; no identity at finite was assumed. Since , the improper integral converges absolutely. The symmetric limit therefore equals .
A real-variable check sets , giving .
Quick check
(Why is the primitive of built on rather than on the whole set where is holomorphic?)
Solution · Answer
After the removable singularity at is filled in, is holomorphic on , because does not cancel the double pole of at . That set is not convex, so the primitive construction does not apply directly. is convex, omits , and still contains .
What can go wrong
Counterexample mode
A pole on the contour requires a different argument
The function has poles at , and the circle passes through both. On near , and , so , which is not integrable near . Hence neither the contour integral as an ordinary integral nor formula (2) is available as stated. One must change the contour or specify and justify a suitable limiting procedure.
Common mistake
(An arc integral need not vanish)
The vanishing of an arc is also a separate estimate, not a consequence of the residue theorem alone. For instance, the upper-semicircle arc integral of is for every radius: length times a bound stays constant. Here is only an arc illustration and also has a pole on the real axis.
Practice with solutions
Quick check
(What is the residue of at ?)
Use (1) with .
Solution · Answer
Here and , so (1) gives .
Quick check
(What is the residue of at ?)
The pole is simple.
Solution · Answer
The simple-pole limit is .
Quick check
(Why is the pole at absent from the upper-semicircle calculation?)
Where is relative to ?
Solution · Answer
It lies below the real axis, outside the upper half-disk bounded by . In the argument above it is also excluded from the convex set .
Quick check
(For , does tend to zero as ?)
Compare with a power of .
Solution · Answer
Yes: , which is asymptotic to .
From a disk to a general simple contour
The disk formula is enough for the calculation above, but the argument principle uses a more flexible form.
Theorem 5.3.5
(Residue theorem for a simple closed contour)
Let be a positively oriented, piecewise continuously differentiable simple closed curve, and let be its bounded interior. Suppose is holomorphic on a neighborhood of except at finitely many isolated singularities in . None lies on the curve. Then
The singularities in this statement may be essential: the residue is still the coefficient of in the local Laurent series. The finiteness and the absence of singularities on the curve matter; neither a boundary pole nor an accumulation of singularities inside the region is covered by this statement.
Proof
(Remove small disks and apply Green's theorem)
Choose pairwise disjoint closed disks around the , small enough to lie in , with each punctured disk inside the corresponding Laurent annulus. Remove their interiors from . On a neighborhood of the closure of the remaining region, is holomorphic; local power series imply that and have continuous first partial derivatives. Green's theorem, applied to the real and imaginary parts of , gives zero for the integral around the positively oriented boundary: the two area integrands vanish by and .
The outer boundary is traversed along , while each new circular boundary is clockwise. If denotes that same circle counterclockwise, the preceding equality therefore reads
Each Laurent series converges uniformly on its circle, so termwise integration makes the th integral times its residue. This proves (3), including for essential singularities, without assuming the residue theorem on the punctured region. The argument uses Green's theorem for a piecewise smooth boundary and the already established local Laurent expansion.
Worked example 5.3.6
(Residues inside a square)
Let be the counterclockwise boundary of the square , . For
both and are inside, while is outside. The expansion gives residue at the essential singularity . At , the double-pole term has no term and the other terms are holomorphic, so its residue is . Hence . The outside pole contributes nothing; its location is not an extra residue to add.
Quick check
(Let be the positively oriented boundary of the square , . Evaluate .)
Solution · Answer
Only is inside. Its residue is , while is holomorphic on a neighborhood of the closed square. The integral is . Moving the pole at onto the boundary would invalidate the hypotheses, even though the local residue at would remain unchanged.
Trigonometric integrals over a full period
The semicircle method turns an integral over the real line into a contour integral and then requires a separate estimate for the arc. Integrals of a rational expression in and over one full period need no such estimate. As runs over , the point traverses the unit circle once counterclockwise, so the real integral already is a contour integral. The disk theorem (2) with radius then applies directly.
Setting. Let and be polynomials in two variables with complex coefficients. Put , and assume
Then is continuous on , and is an ordinary Riemann integral. Hypothesis (4) is sufficient, not necessary after cancellation. For example, if , the displayed quotient is undefined at , but canceling the common factor gives the continuous function on the circle, whose integral is . A formula with uncanceled zeros must be examined separately before assigning such an extension.
The substitution. If , then . Hence
The first two identities hold only on the circle, which is the only place they are used. They suggest the function
Proposition 5.3.7
(Full-period trigonometric integrals)
Assume (4). Then is a rational function of with no pole on the circle , and
where the sum runs over the poles of in the open unit disk.
Proof
(Substitute and apply the disk theorem)
If is the zero polynomial, (4) makes the integrand zero on the circle; is the zero rational function and both sides of (6) are zero. Now suppose is nonzero. Let be the total degree of . A monomial with becomes times a constant after the substitution, and this equals times a polynomial in . Therefore
is a polynomial in . The same construction with the total degree of gives a polynomial . For with ,
so is a rational function. Apart from , it can have poles only at zeros of . At the two arguments of are exactly and , so by (4). Thus no point of the unit circle is a pole of , and on the circle agrees with the unreduced expression (5).
A rational function has only finitely many poles. Those with lie at positive distance from the closed unit disk. Hence is holomorphic on a neighborhood of except at finitely many poles strictly inside it, which is exactly the hypothesis of (2) with radius . Parametrizing the positively oriented circle by , , gives
The second equality uses the identities for and on the circle, and the factor cancels the factor in (5). Combining this with (2) proves (6).
Common mistake
(The factor can create a pole at the origin)
Two features of the proof are easy to overlook. First, the factor is part of . It often creates a pole at the origin, even when itself shows no sign of a singularity there, and that pole must be included in (6). Second, exterior poles do not appear as separate terms in the residue sum, though their positions can affect residues at interior poles. Condition (4) guarantees there are no poles on the circle.
Worked example 5.3.8
()
Let be real with . Because , the denominator never vanishes and (4) holds. Here
The factor cancels, so the origin is not a pole in this example. The roots of are . They are real and distinct because , and their product is . Neither root has modulus : a real number of modulus is , while takes the nonzero values at . Hence exactly one root, call it , lies in the open unit disk, and the other is .
- For , and .
- For , and .
The pole at is simple, with residue . By (6),
Checks. For the integrand is positive and the value is positive; for both are negative. As , the integrand is uniformly in , and (7) is , which agrees. The substitution replaces by , and it shows directly that the value is odd in . A formula that gives for every has chosen the wrong root when .
Counterexample mode
The formula does not continue to parameters of modulus at most one
For the expression is imaginary, but this is a symptom, not a value of the integral. The roots of are then . Their squared modulus is , so both poles lie on the contour. Correspondingly, has a simple zero at each with , and there . Near the integrand behaves like , so the ordinary one-sided improper integrals diverge and the absolute value is not locally integrable there. For the zero is double and the improper integral also diverges. A separately defined Cauchy principal value is a different limit and is not evaluated here. In this example (4) fails exactly when the poles reach the circle.
Worked example 5.3.9
(A pole created at the origin)
Consider
Since , (4) holds. On the circle,
so
The poles are , and , and only the first two lie in the unit disk. The numerator equals at and at , so neither pole is canceled.
At (simple pole). Write . Then
At (double pole). Write with . By (1), the residue is . Put and . Then , , and . The quotient rule gives , so
By (6),
The pole at the origin is not optional. It comes from the factor and from writing in terms of , not from the original denominator. Omitting it would give . That value is visibly wrong, because the integrand is nonnegative and not identically zero, so the integral must be positive.
Practice with solutions
Exercise 5.3.10
(Evaluate directly from (6).)
Core
Solution(4)
- 1
Substitute
on the circle, so
Because: , so (4) holds.
- 2
Locate the poles
Key step
The roots are . Only , of modulus , lies in the unit disk; the other root has modulus .
- 3
Compute the residue
The pole at is simple, with residue .
- 4
Apply (6)
Exercise 5.3.11
(Evaluate by reducing it to (7).)
Foundation
Solution(3)
- 1
Match the form of (7)
- 2
Apply (7) with
Key step
Because: and .
- 3
Multiply by
The integral is , positive as the integrand is.
Quick check
(Can (6) be used for ?)
Check condition (4).
Solution · Answer
No. vanishes to first order at and , so (4) fails and the integrand is not integrable. The corresponding poles of lie on the unit circle.
Quick check
(In computing , why are poles of with discarded rather than added with a minus sign?)
Recall the hypothesis of (2).
Solution · Answer
Theorem (2) counts only singularities inside the contour. A neighborhood of the closed disk can be chosen on which is holomorphic apart from the interior poles, and the exterior poles lie outside it, so they are not separate terms of the sum. Their positions can still affect the interior residues, and therefore the value of the integral.