Complex analysis and residues
Introduced| English | Meaning |
|---|---|
| residue/ˈresɪdjuː/ | The coefficient of the power minus one in a Laurent series. |
| pole/pəʊl/ | An isolated singularity with finitely many nonzero negative-power terms in its Laurent series, and at least one. |
| essential singularity/ɪˈsenʃl ˌsɪŋɡjʊˈlærɪti/ | An isolated singularity with infinitely many nonzero negative-power terms in its Laurent series. |
A smooth reflection can fail the complex derivative
- Complex conjugation $f(z)=\bar z$ reflects the plane across the real axis. It is smooth as a map of real coordinates, but the difference quotient is 1 along real increments and $-1$ along imaginary increments.
- One complex derivative must agree for every direction of approach. We will connect this requirement to the Cauchy–Riemann equations, singularities and residues 留数, then use residues to evaluate contour integrals.
One point is different from a neighbourhood
- Complex differentiability at $z_0$ means the limit of $(f(z_0+h)-f(z_0))/h$ exists as the nonzero complex increment $h$ tends to zero. Agreement along two axes alone does not prove this limit.
- A function is holomorphic on an open set if it is complex-differentiable at every point there. In complex analysis this is equivalent to being analytic there. Analyticity at a point requires such a neighbourhood, not just one derivative at that point.
For complex conjugation, the difference quotient at zero is 1 along real increments and −1 along imaginary increments. What follows?
Difficulty 1/5 · 100 coins + a small bonus on first completion today
A complex derivative must be one common limit for every direction. Two different directional limits rule it out; averaging them has no role in the definition.
What Cauchy–Riemann does and does not prove
- Write $f=u+iv$ with $z=x+iy$. Complex differentiability forces $u_x=v_y$ and $u_y=-v_x$ at the point. For conjugation, $u=x$, $v=-y$, so $u_x=1$ and $v_y=-1$: it fails everywhere.
- A useful sufficient condition is that the first partial derivatives exist and are continuous near the point, and the equations hold at that point. This gives differentiability there. If the equations hold throughout an open set with these smoothness conditions, $f$ is analytic on that set.
Differentiable only at zero
- For $f(z)=|z|^2$, $u=x^2+y^2$ and $v=0$. Cauchy–Riemann requires $2x=0$ and $2y=0$, so only the origin can be a point of complex differentiability.
- At zero the quotient is $|h|^2/h=\bar h\to0$, so $f'(0)=0$. Every neighbourhood of zero contains points where the equations fail. Hence this function is differentiable at zero but analytic nowhere.
Which function is analytic on all of $\mathbb C$?
Difficulty 1/5 · 100 coins + a small bonus on first completion today
Polynomials in z satisfy Cauchy–Riemann everywhere.
For $f(z)=|z|^2$, find the complex derivative $f'(0)$.
Difficulty 2/5 · 140 coins + a small bonus on first completion today
The difference quotient at zero is $|h|^2/h=\bar h$, which tends to 0 for every complex approach. The derivative exists only at zero; this does not make the function analytic on a neighbourhood.
Equations at one point are not enough
- Define $F(0)=0$ and $F(z)=\bar z^2/z$ for $z\ne0$. At the origin its real and imaginary parts have $u_x=v_y=1$ and $u_y=v_x=0$, so the Cauchy–Riemann equations hold there.
- Yet $F(h)/h=\bar h^2/h^2$. It equals 1 for real or purely imaginary $h$, but $-1$ for $h=t(1+i)$, $t\ne0$. There is no derivative at zero. Do not omit the extra hypotheses from the sufficient-condition theorem.
The Cauchy–Riemann equations at one point, without any further regularity hypothesis, guarantee a complex derivative there.
Difficulty 1/5 · 100 coins + a small bonus on first completion today
For $F(0)=0$ and $F(z)=\bar z^2/z$ away from zero, the equations hold at zero. But $F(h)/h$ is 1 along either axis and −1 along the diagonal, so the derivative does not exist.
Define $F(0)=0$ and $F(x+iy)=xy/(x^2+y^2)$ otherwise. Show the Cauchy–Riemann equations hold at zero, yet F has no complex derivative there.
The imaginary part is zero. Along either coordinate axis the real part is zero, so all four first partial derivatives at the origin are 0 and the equations hold. Along x = y = t ≠ 0, the real part is 1/2, which does not tend to F(0) = 0. The function is not continuous at zero and hence cannot be complex-differentiable there. The pointwise equations alone are insufficient.
Laurent series around an isolated point
- Work where $f$ is holomorphic on a punctured disc $0<|z-a|<R$. If it is not defined or not holomorphic at $a$, that point is an isolated singularity. Its Laurent series may contain both nonnegative and negative powers of $z-a$; the negative-power terms form its principal part.
- A residue is the coefficient of the power minus one in a Laurent series. Thus $\operatorname{Res}_a f=c_{-1}$, the coefficient of $(z-a)^{-1}$, not necessarily the coefficient of the most negative power. The residue can be zero even at a genuine singularity.
Classify the whole principal part
- A removable singularity has no nonzero negative-power terms, so the function can be extended holomorphically to the point. A pole 极点 is an isolated singularity with finitely many nonzero negative-power terms in its Laurent series, and at least one. Order $m$ means the lowest nonzero power is $(z-a)^{-m}$.
- An essential singularity 本性奇点 is an isolated singularity with infinitely many nonzero negative-power terms in its Laurent series. At zero, $\sin z/z$ is removable, $1/z^3$ has a pole of order 3, and $e^{1/z}$ is essential. For the last example, $e^{1/z}=1+z^{-1}+z^{-2}/2!+\cdots$.
Match each function with its singularity at $z = 0$.
Difficulty 2/5 · 140 coins + a small bonus on first completion today
Count negative powers in the Laurent series.
Compute a residue after finding the actual pole
- At a simple pole, $\operatorname{Res}_a f=\lim_{z\to a}(z-a)f(z)$. More generally, if $f(z)=g(z)/(z-a)^m$ with $g$ analytic near $a$ and $g(a)\ne0$, the actual pole order is $m$ and the residue is $g^{(m-1)}(a)/(m-1)!$.
- The derivative formula selects the Taylor coefficient of degree $m-1$ in $g$, which becomes power $-1$ after division. Check cancellation first: $(z^2-1)/(z-1)=z+1$ away from 1, so its singularity there is removable, not a simple pole.
For $f(z)=(z^2-1)/(z-1)$ away from 1, what kind of singularity is at 1?
Difficulty 1/5 · 100 coins + a small bonus on first completion today
Cancellation gives z + 1 for z ≠ 1. Defining f(1) = 2 makes the function analytic there; there is no nonzero principal part.
Pole order and residue answer different questions
- At $z=1$, $3/(z-1)$ has residue 3; $e^z/(z-1)^2$ has residue $e$; and $e^{2z}/(z-1)^3$ has residue $4e^2/2=2e^2$. The derivative order is one less than the pole order.
- To see the double-pole example directly, put $w=z-1$: $e^{1+w}=e(1+w+w^2/2+\cdots)$. Dividing by $w^2$ gives coefficient $e$ at $w^{-1}$. In contrast, $5/(z-1)^2$ has only a term of power $-2$: pole order 2, residue 0.
What is the residue at z = 1?
Read the power-minus-one coefficient at z = 1, including removable and essential cases.
The residue of $\dfrac{e^{3z}}{(z-2)^2}$ at $z = 2$ is
Difficulty 1/5 · 100 coins + a small bonus on first completion today
$g = e^{3z}$, $g'(2) = 3e^6$.
Find the residue of $5/(z-1)^2$ at 1.
Difficulty 2/5 · 140 coins + a small bonus on first completion today
The Laurent series contains the term 5(z − 1)⁻² and no power-minus-one term. Its residue is 0, although the pole order is 2.
Let $a$ be any complex constant and $f_a(z)=e^{az}/z^2$ for $z\ne0$. Find the actual pole order and residue at zero. Does choosing a = 0 remove the pole?
Taylor expansion gives $e^{az}=1+az+O(z^2)$. Dividing by $z^2$ gives $f_a(z)=z^{-2}+az^{-1}+O(1)$. The power-minus-two coefficient is always 1, so the actual pole order is 2 for every a. The residue is a. For a = 0 the function is $1/z^2$: its residue is zero, but the double pole remains.
The residue theorem has geometric conditions
- Let $C$ be a positively oriented, simple, piecewise smooth closed curve, traversed once counterclockwise. Require $f$ holomorphic on an open neighbourhood of the curve and its interior except for finitely many isolated singularities $a_j$ strictly inside. No singularity may lie on $C$.
- Under these conditions, the contour integral is the sum of the enclosed residues multiplied by $2\pi i$:
Locate the poles before adding their residues
- For $f(z)=1/[z(z-3)]$ and the counterclockwise circle $|z|=2$, the pole at 0 is inside and the pole at 3 is outside. The enclosed residue is $\lim_{z\to0}1/(z-3)=-1/3$.
- Therefore $\oint_C f(z)\,dz=-2\pi i/3$. The pole at 3 does not enter this sum. A residue calculation without the locations and the contour is not yet an integral calculation.
For the unit circle traversed once counterclockwise, $\oint_{|z|=1}\dfrac{z^2+1}{z}\,dz=2\pi i\cdot k$. Find $k$.
Difficulty 2/5 · 140 coins + a small bonus on first completion today
(z² + 1)/z = z + 1/z, so the residue at the only pole z = 0 (inside |z| = 1) is 1. The integral is 2πi · 1, giving k = 1.
A pole outside the contour adds its residue to the integral.
Difficulty 1/5 · 100 coins + a small bonus on first completion today
Only enclosed singularities count.
Orientation and boundary cases
- Reversing a curve's orientation reverses its integral. The same radius-2 circle traversed clockwise therefore gives $+2\pi i/3$ for $1/[z(z-3)]$.
- Changing the radius to 3 puts a pole on the contour. The usual residue-theorem hypotheses fail, and the ordinary contour integral through that pole is not defined. Do not count a boundary pole as inside, outside or “half a residue”.
Where is the pole?
Locate the poles 0 and 3 of 1/[z(z − 3)] relative to each circle before applying the theorem.
The circle $|z|=2$ is traversed once clockwise. If $\oint_C dz/[z(z-3)]=2\pi i\,k$, find $k$.
Difficulty 2/5 · 140 coins + a small bonus on first completion today
Only the pole at zero is enclosed, with residue −1/3. Clockwise orientation gives −2πi times that residue, so k = +1/3.
Integrate $f(z)=1/[z(z-3)]$ on the circle $|z-3|=1$, traversed once counterclockwise. Then reverse orientation. What changes if the radius becomes 3?
The pole at 3 is inside the radius-1 circle, while zero is outside at distance 3 from its centre. The residue at 3 is $\lim_{z\to3}1/z=1/3$, so the counterclockwise integral is $2\pi i/3$ and the clockwise integral is $-2\pi i/3$. With radius 3 the pole at zero lies on the path. The usual residue theorem does not apply, and the ordinary contour integral through the pole is not defined.
Zero does not prove that singularities are absent
- On the counterclockwise circle $|z|=4$, both poles of $1/[z(z-3)]$ are enclosed. Their residues are $-1/3$ and $+1/3$, so the integral is zero through cancellation, even though two poles remain inside.
- Also, $5/z^2$ has a genuine double pole at zero but residue zero, so its integral around a counterclockwise unit circle is zero. Holomorphy on a neighbourhood of the curve and its interior implies zero integral. A zero integral for one curve does not prove holomorphy inside.
The circle $|z|=4$ is traversed once counterclockwise. If $\oint_C dz/[z(z-3)]=2\pi i\,k$, find $k$.
Difficulty 2/5 · 140 coins + a small bonus on first completion today
Both poles are enclosed. Their residues are −1/3 at zero and +1/3 at three, whose sum is zero. Neither singularity is thereby removed.
Keep the local and contour questions distinct
- A derivative at one point does not give analyticity on a neighbourhood. Cauchy–Riemann at a point needs additional regularity for sufficiency. For singularities, inspect the full principal part; a pole needs at least one nonzero negative-power term.
- For a contour integral, locate every singularity, reject boundary poles and state orientation. A residue is one Laurent coefficient; zero residue or a zero integral does not remove a pole. Add only enclosed residues when the theorem's hypotheses hold.
A student says: “The contour integral is zero, so the integrand has no poles inside.” Refute this using both $1/[z(z-3)]$ on $|z|=4$ and $5/z^2$ on $|z|=1$, with each circle traversed once counterclockwise.
In the first example, the residues at the two enclosed poles are −1/3 and +1/3, so they cancel. Both poles remain. In the second example, there is a genuine double pole, but its Laurent series has no power-minus-one term, so its residue and the integral are zero. These are different mechanisms. Holomorphy on a neighbourhood of the curve and its interior gives a zero integral, but the converse from one curve is false.