Worked solutions on the Cauchy formula: simple and higher-order poles, poles on the contour, and exact derivative values.
§9.2Estimated reading time: 24 min
These worked problems revise the Cauchy formula and contain worked solutions.
Conventions used throughout
D(c,r)={z:∣z−c∣<r} is the open disk and D(c,r) is the closed disk.
C(c,r) is the circle ∣z−c∣=r, traversed once, positively (counterclockwise):
z(t)=c+reit,t∈[0,2π].
−C(c,r) is the same circle traversed once clockwise, z(t)=c+re−it, t∈[0,2π]. Every circle is positively oriented unless it is explicitly written as −C(c,r).
For any piecewise C1 path γ and any F continuous on γ,
∫−γF(z)dz=−∫γF(z)dz.
Results you may quote
Theorem 9.2.1
(Cauchy's theorem on a disk (T))
Suppose F is holomorphic on an open disk D(c,R) and 0<r<R. Then ∮C(c,r)F(z)dz=0. The reason is that a disk is convex, so F has a primitive there.
Theorem 9.2.2
(Cauchy's formula and the derivative formulas (CF))
Suppose g is holomorphic on an open set U containing D(c,r). Then for every a with ∣a−c∣<r and every integer n≥0,
(A genuine pole on the contour defeats the ordinary integral)
Let C=C(c,r) with z(t)=c+reit, let a=z(t0) lie on C, and let g be defined and continuous near a (for example holomorphic on an open set containing D(c,r)) with g(a)=0. Let m≥1. Then the parametrized integrand of ∮Cg(z)(z−a)−mdz is not integrable on any interval around t0, so the ordinary contour integral is undefined.
Proof
(A lower bound near the pole)
Read t modulo 2π and put s=t−t0. Since g is continuous and g(a)=0, there is δ∈(0,π) with ∣g(z(t))∣≥∣g(a)∣/2 for ∣s∣<δ. The chord formula gives ∣z(t)−a∣=r∣eis−1∣=2r∣sin(s/2)∣≤r∣s∣, and ∣z′(t)∣=r. Hence for 0<∣s∣<δ,
Since ∫0δs−mds=∞ for every m≥1, the integrand is not integrable near t0.
Proof X-ray
Which hypothesis does the work
The condition g(a)=0 keeps the integrand large: it turns the vanishing of the denominator into genuine growth. If g(a)=0, a factor of the numerator may cancel the singularity, so this hypothesis cannot be dropped (see the counterexample below).
The key move is the chord estimate ∣z(t)−a∣≤r∣s∣, which uses ∣sinx∣≤∣x∣. It turns the geometric statement that the pole lies on the circle into the explicit bound ∣s∣−m for the parametrized integrand.
The proof closes with the divergence of ∫0s−mds. Even m=1 diverges, only logarithmically, so a simple pole on the contour is already enough.
Only the continuity of g at a is used. The conclusion concerns the ordinary integral and says nothing about principal values, which are a different construction.
The next example shows why g(a)=0 is needed.
Counterexample mode
A zero of the denominator on the contour need not be a pole
Claim: if the denominator vanishes at a point of the contour, the ordinary integral is undefined. Take C=C(0,1), a=1∈C and F(z)=z−1sin(z−1). The denominator vanishes at 1, yet sin(u)/u→1 as u→0, so F extends to an entire function with value 1 at z=1. The extension is continuous on C, and (T) gives ∮CFdz=0. Here the numerator g(z)=sin(z−1) has g(1)=0, which is exactly the hypothesis g(a)=0 of the proposition that fails.
Roadmap
Step
What you do
Tool
1
Name the circle, its center, radius and orientation.
Definitions
2
Isolate the numeratorg and check that it is holomorphic on an open set containing the closed disk. Look for numerator singularities, not just the obvious denominator.
Inspection
3
Split the parameter a into three regions: interior ∣a−c∣<r, exterior ∣a−c∣>r, and on the contour ∣a−c∣=r.
Geometry
4
Interior: apply (CF) with the correct n.
(CF)
5
Exterior: find an open disk containing the circle on which the whole integrand is holomorphic, then apply (T).
(T)
6
Clockwise: multiply by −1.
Reversal
7
On the contour: decide whether the singularity is a genuine pole. If it is, explain why the usual integral is undefined.
Estimates
Checklist before you submit
I wrote down the parametrization and orientation of each circle.
I checked every singularity of the numerator against the closed disk, not just the open disk.
For the exterior region I named a specific open disk of radius larger than r on which the integrand is holomorphic.
I used n! correctly: the exponent n+1 in the denominator corresponds to g(n).
I changed the sign for clockwise traversal.
For an on-contour pole I said the usual contour integral is undefined. I did not assign it a principal value.
I checked that the numerator does not vanish at the on-contour point, so the pole is genuine and not removable.
A simple pole with a numerator that has its own singularity
Let C1=C(1,2) be the circle ∣z−1∣=2, traversed once counterclockwise by z(t)=1+2eit, t∈[0,2π].
Define
g(z)=z+3eπz,Fa(z)=z−ag(z)=(z+3)(z−a)eπz,a∈C.
(a) Show that g is holomorphic on an open set containing D(1,2), and that g has no zeros.
(b) For each parameter region, find I(a)=∮C1Fa(z)dz wherever it is defined:
(i) ∣a−1∣<2;
(ii) ∣a−1∣>2, including the special value a=−3.
(c) Evaluate exactly:
I(i);
∮−C1Fi(z)dz;
I(4).
(d) Take a=3, which lies on C1. Explain why the usual contour integral ∮C1F3(z)dz is undefined.
Solutions: simple pole
(a) The numerator is holomorphic near the closed disk.
The only singularity of g is z=−3.
Its distance from the center is ∣−3−1∣=4.
So g is holomorphic on U=D(1,4).
Since 4>2, the open set U contains D(1,2).
Also eπz=0 for every z, and z+3 is finite and nonzero on U. Hence g(z)=0 for all z=−3.
Warning. If the circle had radius larger than 4, the point −3 would lie inside it. Then g would not be holomorphic on the closed disk and (CF) could not be applied with this g.
(b)(i) Interior, ∣a−1∣<2.
By (a) and (CF) with n=0, c=1, r=2:
I(a)=2πig(a)=a+32πieπa.
Note that a=−3 automatically here, because ∣−3−1∣=4>2.
(b)(ii) Exterior, ∣a−1∣>2.
Set R=min(∣a−1∣,4). Then R>2.
On D(1,R), neither z=a nor z=−3 occurs. Indeed, both points are at distance at least R from 1.
So Fa is holomorphic on D(1,R), and D(1,R) contains C1.
By (T), I(a)=0.
This includes a=−3. In that case F−3(z)=eπz/(z+3)2 and R=4, so again I(−3)=0.
Summary:
I(a)=⎩⎨⎧a+32πieπa,0,∣a−1∣<2,∣a−1∣>2.
Worked example 9.2.4
(A complex interior parameter, its reversal and an exterior value)
Value a=i. We have ∣i−1∣=2<2, so a=i is interior. Using eπi=−1:
Value a=4. We have ∣4−1∣=3>2, so a=4 is exterior. Here R=min(3,4)=3, and
I(4)=0.
Worked example 9.2.5
(A genuine simple pole on the contour)
Step 1: the point lies on the circle. We have ∣3−1∣=2. In the parametrization, 3=z(0), so t0=0.
Step 2: the pole is genuine. We compute g(3)=e3π/6=0. So F3=g(z)/(z−3) has a genuine simple pole at 3. It is not a removable singularity.
Step 3: the integrand blows up like 1/∣t∣.
By continuity of g, there is δ∈(0,π) such that ∣g(z(t))∣≥∣g(3)∣/2 for ∣t∣<δ.
For t near 0,
∣z(t)−3∣=2∣eit−1∣=4sin(t/2)≤2∣t∣.
Also ∣z′(t)∣=2.
Hence, for 0<∣t∣<δ (with t taken near 0, or equivalently near 2π),
F3(z(t))z′(t)≥2∣t∣∣g(3)∣/2⋅2=2∣t∣∣g(3)∣.
Step 4: conclude.
The usual contour integral is ∫02πF3(z(t))z′(t)dt.
This requires the integrand to be defined and integrable on [0,2π].
Here the integrand is undefined at t=0 (equivalently t=2π) and unbounded near it.
Since ∫0δdt/t=∞, it is not even absolutely integrable near the endpoint.
So ∮C1F3(z)dz is undefined as an ordinary contour integral. Neither (CF) nor (T) applies, because both require the integrand to be holomorphic on the contour itself.
We do not assign it a principal value. Principal values are a separate construction and are not the usual integral.
A third-order denominator and exact derivative values
Let C2=C(0,3) be the circle ∣z∣=3, traversed once counterclockwise by z(t)=3eit, t∈[0,2π].
Ga is holomorphic on D(0,R), because both a and 4 have modulus at least R.
D(0,R) contains C2.
By (T), J(a)=0.
This includes a=4. In that case G4(z)=e2z/(z−4)4 and R=4.
(c) Specific values.
Value a=0 (interior):
J(0)=πi⋅(−3241)=−3241πi.
Value a=1 (interior):
J(1)=πi⋅(−2750e2)=−2750πe2i.
Clockwise, a=0:
∮−C2G0(z)dz=−J(0)=3241πi.
Value a=7/2. We have ∣7/2∣>3. Here R=min(7/2,4)=7/2, so
J(7/2)=0.
Value a=4. Here R=4, so
J(4)=0.
Note the contrast: on a circle ∣z∣=5, the point 4 would be enclosed. The numerator h would then fail to be holomorphic on the closed disk, and the computation above would not apply.
Worked example 9.2.7
(A genuine third-order pole on the contour)
Step 1: the point lies on the circle. We have ∣−3∣=3 and −3=z(π), so t0=π.
Step 2: the pole is genuine. We compute h(−3)=e−6/(−7)=0. So G−3=h(z)/(z+3)3 has a genuine pole of order 3 at −3.
Step 3: the integrand blows up like 1/∣t−π∣3.
Choose δ∈(0,π) such that ∣h(z(t))∣≥∣h(−3)∣/2 for ∣t−π∣<δ.
The parametrized integrand is undefined at t=π and unbounded near it.
Since ∫0δs−3ds=∞, it is not absolutely integrable near t=π.
So the usual contour integral ∮C2G−3(z)dz is undefined.
(CF) does not apply, because it needs ∣a∣<3 strictly. (T) does not apply, because G−3 is not holomorphic on the contour.
No principal value is being claimed.
Common mistakes
Common mistake
(Using the formula on a larger circle)
For G1=e2z/((z−4)(z−1)3) the value πih′′(1) is correct on ∣z∣=3, but not on a circle through or beyond 4. On ∣z∣=4 the numerator h has its own pole on the contour, and since e8/(4−1)3=0 the integral is undefined by the proposition above. On ∣z∣=5 the pole of h is inside, so (CF) cannot be applied with this h.
Common mistake
(Dropping the factor 1/n!)
For the denominator (z−a)3 the formula has n=2 and the factor 2!2πi=πi. Writing 2πih′′(a) doubles every interior value of J(a).
Quick checks
Quick check
(Evaluate ∮C1F1(z)dz, where C1=C(1,2) and Fa(z)=(z+3)(z−a)eπz.)
Locate a=1 relative to C1.
Solution · Answer
Since ∣1−1∣=0<2, the point is interior, and I(1)=2πig(1)=42πieπ=2πieπ.
Quick check
(Evaluate J(−1)=∮C2G−1(z)dz, where C2=C(0,3) and Ga(z)=(z−4)(z−a)3e2z.)
Use J(a)=πih′′(a) with the formula for h′′.
Solution · Answer
Here ∣−1∣<3, so J(−1)=πih′′(−1). With z−4=−5,
h′′(−1)=e−2[−54−254+−1252]=e−2⋅125−100−20−2=−125122e−2.
Therefore J(−1)=−125e2122πi.
Quick check
(For which real values of a is ∮C1Fa(z)dz undefined?)
The integral can fail only when a pole lies on the circle ∣z−1∣=2.
Solution · Answer
The real points with ∣a−1∣=2 are a=3 and a=−1. At each, g(a)=eπa/(a+3)=0, so the pole is genuine and the integral is undefined. For every other real a, the integral is a+32πieπa if ∣a−1∣<2 and 0 if ∣a−1∣>2.