Today we discussed this integral $$ \oint_{|z|=10} \frac{1}{(z-1)(z-2)} dz $$ where the contour is a CCW (counter-clockwise) circle of radius $10$ (or any radius $R > 2$, that encloses $1,2$)
We used three methods to show that this is zero. Denote the integrand by $f(z)$, namely $f(z) = \frac{1}{(z-1)(z-2)}$.
We apply residue theorem, and computed $$ Res_{z=1} f(z) = \frac{1}{z-2}|_{z=1} = -1. $$ $$ Res_{z=2} f(z) = \frac{1}{z-1}|_{z=2} = 1. $$ Hence the result of the contour integral, denoted by $I$ is $$ I = (2 \pi i) [Res_{z=1} f(z) + Res_{z=2} f(z)] = 0 .$$
We can deform the contour $C$, as long as the integrand $f(z)$ when restricted to the contour remains a holomorphic function during the deformation. The resulting integral is invariant under the deformation.
Thus, we may change the circle $|z|=10$ to $|z|=R$, and let $R$ tends to $\infty$. Let $I_R$ be the integral on the contour $|z|=R$. We have two claims
$$ |I_R| \leq \oint_{|z|=R} | f(z) | |dz| \leq \oint_{|z|=R} \frac{1}{|z-1| |z-2| } |dz| $$ By triangle inequality for complex numbers, $$ |a| + |b| \geq |a+b| \geq | |a| - |b| |, $$ we have for any $z$ with $|z|=R > 10$ $$ |z-1| \geq |z|-1 = R-1 > 0, \quad |z-2| \geq |z|-2 = R - 2 > 0, $$ hence $$ \frac{1}{|z-1| |z-2| } \leq \frac{1}{(R-1)(R-2)} $$ on the contour $|z|=R$. Thus $$ \oint_{|z|=R} \frac{1}{|z-1| |z-2| } |dz| \leq \frac{1}{(R-1)(R-2)} \oint |dz| = \frac{1}{(R-1)(R-2)} (2\pi R) $$ To summarize, we have $I_R$ is a constant, and for any $R>10$, we have $$ 0 \leq |I_R| \leq \frac{2\pi R}{(R-1)(R-2)}. $$ Thus $|I_R|=0$, hence $I_R=0$.
Let $w = 1/z$, then we have
$$ I = \oint_{|w|=1/10, CW} \frac{1}{(w^{-1}-1)(w^{-1}-2)} d (w^{-1}) = \oint_{|w|=1/10, CW} \frac{-1}{(1-w)(1-2w)} dw = \oint_{|w|=1/10} \frac{1}{(1-w)(1-2w)} dw $$ where in the last step, I changed the orientation of the contour from $CW$ to $CCW$(CCW is by default, hence omited) and add an extra $(-1)$ factor to the integral.
Since the integrand is only singular at $w=1,1/2$, and the contour $|w|=1/10$ contains no singularity in its interior, the integral is 0.
It is useful to think of add a point $\infty$ to the complex plane $\C$, and think of $\C \cup \{\infty\}$ as a sphere, where $\infty$ is identified with the north pole, $0$ with the south pole, the unit circle $|z|=1$ as the equator.
The natural coordinate to use near the north pole is $w=1/z$, so that $z=\infty$ corresponds to $w=0$.
Let $C$ be the contour of $|z|=10$. Consider the following integrals.
(1) $$\oint_C \frac{1}{1+z^2} dz $$
(2) (the result for this one is not zero.) $$\oint_C \frac{z}{1+z^2} dz $$
(3) $$\oint_C \frac{z^2}{1+z^4} dz $$
Apply methods 1,2,3 to the above problems (each method need to be used once)
Let $\Omega$ be a domain (open subset), and $z_0 \in \Omega$ a point. We say a holomorphic function $f: \Omega \RM \{z_0\} \to \C$ has a pole of order $m$ at $z_0$, if $h(z) = (z-z_0)^m f(z)$ can be extended to the entire $\Omega$, and $h(z_0) \neq 0$.
Suppose $f$ has an order $m$ pole at $z_0$, then near $z_0$ (more precisely, for $z \in \C$ with $0 < |z-z_0| < r$ for some small positive $r$), we can write Laurent expansion $$ f(z) = \frac{a_{-m}}{ (z-z_0)^m} + \cdots + \frac{a_{-1}}{ (z-z_0)^1} + a_0 + a_1 (z - z_0)^1 + \cdots $$
The coefficient $a_{-1}$ is called the residue, and denoted as $Res_{z_0} f := a_{-1}$. ($:=$ read 'is defined as')
1. If the pole is order $1$, then $Res_{z_0}f(z) = \lim_{z \to z_0} f(z) (z-z_0)$.
Example: residue of $f(z) = (z-2) / (z-1)$ at $z=1$.
2. If the pole at $z_0$ is of order $m>1$, then just Laurent expand. More precisely, do Taylor expansion of $$(z-z_0)^m f(z) = a_{-m} + \cdots + a_{-1} (z-z_0)^{m-1} + \cdots $$ and then you get $a_{-1}$.
Example: residue of $f(z) = [(z-2) / (z-1)]^2$ at $z=1$.
Thm: let $f$ has a pole (of some order) at $z_0$, and let $C$ be a small circle of radius $\epsilon$ around $z_0$ positively oriented (i.e in counter-clockwise direction) so that $f$ is holomorphic on $C$, and $C$ encloses only one pole. Then $$\oint_{C} f(z) dz = (2\pi i) Res_{z_0} f. $$
Proof sketch: write $f$ as a Laurent expansion, then integrate term by term.
Due Monday in class.
1. Taylor expand $(z+1)(z+2)$ around $z=3$.
2. Laurent expand $1/[(z-1)(z-2)]$ around $z=1$. And do it again, this time around $z=2$.
3.Compute $\int_{0}^{2\pi} 1 / (z(t)) d z(t). $ for the following three contours (a) For $t \in [0, 2\pi]$, let $z(t) = e^{it}$.
(b). For $t \in [0, 2\pi]$, let $z(t) = e^{i2t}$.
(c ). For $t \in [0, 2\pi]$, let $z(t) = e^{-it}$.
Boas, Ch 14, Section 3, #4, 6
Let's see some cool application of contour integral first.
1. $\int_\R 1/(1+x^2) dx$
didn't finish, the rest will wait till next week.
2. $\int_0^\infty 1/(1+x^3) dx$
3. $\int_0^\infty e^{-x}/(1+x^3) dx$
4. $\int_\R \frac{x}{x^4+1} e^{-x^2} dx$
Reading: Boas, Ch14, section 1-5
So, you have learned what holomorphic function looks like, and you know there are functions which are 'bad' only at a few points. What do you want to do with these functions?
Just like calculus, you can do differentiation, and you can do integration. Differentiation is easy, let's talk about integration.
Recall what we do in real analysis case: given $f(x)$ on $\R$, we can find one primitive $F(x)$ by considering $$ F(x) = C + \int_{x_0}^x f(u) du $$ where we set the initial condition that $F(x_0) = C$, and $F'(x)= f(x)$.
Can we do the same here? Say $f(z)$ is a holomorphic function, we can define $$ F(z) = C + \int_{z_0}^z f(u) du $$ Now, we immediately run into trouble: how do we go from $z_0$ to $z$? Does the integration depends on how we choose the path from $z_0$ to $z$? Thanks to the fact that $f$ is holomorphic, the integration is independent of the choice of path.
OK, $f(z) = 1/z$ is not a holomorphic function on the entire $\C$. We can say, it is a holomorphic function on the 'punctured complex plane' $\C^* = \C \RM \{0\}$, or it is a meromorphic function on $\C$ with a pole of order $1$ at $z=0$. Either way, we can ask, can we find the primitive of $1/z$ on $\C^*$? Namely, is there a hol'c function $F(z)$ such that $F'(z) = 1/z$?
You probably know that, for $x>0$, if you integrate $1/x$, you get $\log (x) + C$. (why is that? )
The same holds for complex analytic function. almost. We can say the primitive of $1/z$ is $\log(z) + C$, but $\log(z)$ is a multivalued function on $\C^*$.
Let $f: \Omega \to \C$ be a meromorphic function. Let $\gamma$ be a closed contour in $\Omega$ (“contour” just means a smooth path) that avoids the pole of $f$. Then, $$ \int_\gamma f(z) dz = (2 \pi i) \sum_{z_0 \text{poles of } f} Res_{z_0} f. $$ The $Res_{z_0} f$ is the residue of $f$, which is the Laurent expansion of $f$ at $z$, the coefficient in front of $1/(z-z_0)$.
Example: $f(z) = 1/ [(z-1)(z-2)(z-3)]$, $\gamma$ is a contour around the two poles $1$ and $2$.
1. For $t \in [0, 2\pi]$, let $z(t) = e^{it}$. Compute $\int_{0}^{2\pi} 1 / (z(t)) d z(t). $
2. For $t \in [0, 2\pi]$, let $z(t) = e^{i2t}$. Compute $\int_{0}^{2\pi} 1 / (z(t)) d z(t). $
3. For $t \in [0, 2\pi]$, let $z(t) = e^{-it}$. Compute $\int_{0}^{2\pi} 1 / (z(t)) d z(t). $