This revision chapter extends the Cauchy formula lesson with worked examples.
Roadmap
- State the formulas precisely and explain each hypothesis.
- Derive the derivative formulas from a uniform geometric expansion of the kernel.
- Give a decision procedure for "pole inside, outside, or on the contour".
- Work exact examples, including one where the naive method fails.
- Finish with a checklist.
Convention. Every contour is a simple circle traversed once, positively (counterclockwise), unless it is expressly marked as reversed. Every claim in this chapter is a circle-only claim. Versions for general contours, which use winding numbers or homotopy, are not covered here.
The statements
Theorem 9.1.1
(Cauchy's integral formula on a disk)
Assume the following:
- is open.
- is holomorphic on .
- The closed disk is contained in .
Then for every with and every integer ,
The case is the formula .
What each hypothesis means:
- open, . The function must be holomorphic on a neighborhood of the closed disk, not just on the open disk. Being holomorphic on the open disk alone is not enough: could misbehave on the circle itself.
- holomorphic. This applies to the numerator only. The single point is the only allowed singularity, and it comes from the kernel .
- . The point must lie strictly inside the circle. Points on the circle and points outside it are treated separately in the decision-procedure section.
- Positive orientation. If the circle is traversed clockwise, the integral changes sign.
Support formula. Parametrize the circle by for . For any integer ,
Deriving the derivative formulas
Proof
(Derivative formulas from the geometric series)
We take the case of Cauchy’s formula as known.
Setting up. Fix a point with . Let . This is the distance from to , so for every .
Neighborhood hypothesis. Choose . We work with in the smaller closed disk , which lies inside .
Expanding the kernel. For , write
Because , the geometric series gives
Uniform convergence. Let , which is finite because is compact and is continuous. The -th term of is then bounded by . This bound does not depend on . By the Weierstrass M-test, the series converges uniformly on , so we may integrate term by term:
Conclusion. This holds for every , so equals a convergent power series on . A power series may be differentiated term by term, which gives . That is exactly the derivative formula at .
Proof X-ray
How the series argument uses each hypothesis
- Where the closed disk enters: gives on , and compactness of makes finite. If were holomorphic only on the open disk, could be infinite and the series bound would say nothing.
- The key move: the ratio satisfies uniformly in . Uniformity in is what lets the Weierstrass M-test justify integrating term by term; a bound that degenerated as varies would not.
- How it closes: the coefficients do not depend on , so a single convergent power series represents on , and differentiating it term by term gives , which is the derivative formula.
- What the argument does not need: need not be the center . Any interior point works, but shrinks as approaches , so the series is guaranteed to converge only up to the circle.
Bonus. The argument also shows that a holomorphic is infinitely differentiable.
Decision procedure
Suppose you are asked for . The procedure below assumes that the whole integrand is a single-valued meromorphic function on an open neighborhood of the closed disk: it is holomorphic there except at isolated poles. This includes every worked example below. A branch point or a branch cut requires a separate check of the chosen single-valued domain; merely locating denominator zeros does not establish holomorphy. Work through these steps in order.
- Orientation. If is clockwise, compute the positive version and negate it.
- List all poles of the whole integrand. Check zeros of the denominator and poles of the numerator, such as those of or . Simplify cancellations so that removable singularities are not counted as poles.
- Check the circle itself. If any pole lies on , the integrand is unbounded near that point along . In that case the usual contour integral is undefined. Say so, and do not substitute a principal value. A removable singularity is different: replace the integrand by its holomorphic extension and continue.
- No poles in the closed disk. Under the domain hypothesis above, the integrand is then holomorphic on an open neighborhood of . Cauchy's theorem for a disk gives an integral of . This conclusion uses holomorphy, not just the absence of visible denominator zeros.
- Exactly one, at inside. If is the only singularity inside, and is holomorphic on a neighborhood of , then
- More than one inside. A single application of the direct formula to the whole integrand is invalid, because a kernel with one center cannot account for poles at other points while leaving a holomorphic numerator. For a rational integrand, split it by partial fractions and apply Cauchy's formulas to the separate terms. For other meromorphic integrands, an appropriate decomposition or the residue theorem is needed, if your course has made it available.
Worked examples
Worked example 9.1.2
(A simple pole with an entire numerator)
Compute .
The numerator is entire, and the only singularity is , which is inside. With , the answer is .
Worked example 9.1.3
(A pole of order three)
Compute .
The point is inside, and . Since , we have , so . The integral is .
Worked example 9.1.4
(A pole outside the circle)
Compute .
The only singularity is at , which lies outside the closed unit disk. The integrand is holomorphic on , so the integral is by step 4.
Worked example 9.1.5
(Clockwise orientation)
Compute around , traversed clockwise.
First compute the positive version. Here , so , and the positive integral is . Reversing the orientation gives .
Worked example 9.1.6
(A pole on the contour)
Consider .
The point lies on , and the singularity is not removable. The usual contour integral is therefore undefined.
Worked example 9.1.7
(A non-entire numerator that is still safe)
Compute .
First check the numerator. The function has its nearest poles at , and . So is holomorphic on , which is a neighborhood of the closed unit disk. Since , the answer is .
Note the warning: on , the same integrand would have poles of inside the circle, and step 6 would apply instead.
Counterexample mode
A numerator with a second pole inside the circle
Claim: for , take as the numerator and apply the formula at . This gives . The claim is false: has a pole at , inside , so is not holomorphic on a neighborhood of the closed disk. The residues of the integrand are at and at , so . The hypothesis cannot be dropped: the second pole contributes a term of its own.
The correct value comes from splitting the integrand by partial fractions, as in the next example.
Worked example 9.1.8
(Two interior poles by partial fractions)
Compute
Correct method: partial fractions. Write
Find the coefficients one at a time:
- Multiply by and set . This gives .
- Multiply by and set . This gives .
- Clear denominators to get . Comparing the coefficients gives , so .
- Check: the -coefficient is , and the constant term is . Both agree with the left-hand side.
Keep the contour fixed as . Since both and lie inside it, Cauchy's formula with gives and . The derivative formula with and gives . The pole need not be at the center of the circle. Therefore
Independent check with residues (if available). The residue at the double pole is . The residue at the simple pole is . The sum of the residues is , so . ✓
Interpretation. The naive value is exactly the contribution of the pole at alone. You can see this on the circle . There, is holomorphic on , which is a neighborhood of the closed disk, so Cauchy's formula correctly gives . On the larger circle , the pole at contributes and cancels it.
Concept lensAlgebraic
Only the simple-fraction coefficients survive
For a rational integrand with no pole on , partial fractions turn the circle integral into bookkeeping. Each term with has the primitive , so it integrates to over any closed curve, and a polynomial part does too. Only the coefficient of remains:
In the example above, at and at , so the sum is .
Checklist
- The contour is a simple circle, and its orientation is checked. If clockwise, negate the answer.
- The whole integrand is single-valued and meromorphic on an open neighborhood of the closed disk, as required by the decision procedure; branch-domain issues need a separate check.
- Every pole is listed, including those arising from the numerator, after removable cancellations are extended.
- Any pole on means the usual integral is undefined. No principal value is used.
- If there are no poles in the closed disk, the meromorphy hypothesis makes the integrand holomorphic on a neighborhood of that disk, so the answer is .
- Exactly one interior point , with the numerator holomorphic on a neighborhood of the closed disk, means the answer is . Remember the factor and the exponent .
- Several interior singularities means using partial fractions or residues, never the direct formula.
- The answer has been cross-checked by a second method where possible.
Common mistakes
Common mistake
(Replacing an on-contour pole by half a residue)
For it is tempting to answer , half of . On the modulus of the integrand is , which is not integrable near . So the usual contour integral does not exist. The number is the principal value, a different construction that this chapter does not use.
Common mistake
(Dropping the factor )
For the formula has exponent , so and the factor is . Writing gives , twice the correct value .
Quick checks
Quick check
(Evaluate .)
Identify , and .
Solution · Answer
The numerator is entire, lies inside , and the exponent gives . The integral is .
Quick check
(The point lies on the circle . Evaluate .)
Decide whether the singularity at is genuine.
Solution · Answer
Since as , the singularity at is removable. The integrand extends to an entire function, so the integral over the closed curve is . A removable singularity on the contour does not make the integral undefined.
Quick check
(Evaluate .)
Both lie inside the circle, so split the integrand.
Solution · Answer
By partial fractions . Each term has integral over , so the integral is .