Calculus Revisited: Complex Variables and Differential Equations

1. The Complex Numbers

1. Why is complex number useful?

$$y^{\prime \prime }(t)+y(t)=0$$

can mean the acceleration of an object plus the displacement of an object equal zero.
but then if we set $y(t)=e^{rt}$, we then have

$$r^2+1=0$$

which means introduction of complex number is necessary.
We then have solution
$$y=e^{\pm it}\to \{\begin{array}{l}y=e^{it}\\ y^{\prime }=it{e}^{it}\\ y^{\prime \prime }=-t^2{e}^{it}\end{array}$$

2. Terminologies of complex number $x+iy$?

• Argand Diagram: the plane with real axis and imaginary axis
• magnitude: $\sqrt{x^2+y^2}$
• argument: $\theta =\arctan \frac{y}{x}$

3. Prove $(r_1,{\theta }_1)(r_2,{\theta }_2)=(r_1{r}_2,{\theta }_1+{\theta }_2)$
$$\begin{array}{rl}(r_1,{\theta }_1)(r_2,{\theta }_2)& =(r_1\cos {\theta }_1+i{r}_1\sin {\theta }_1)(r_2\cos {\theta }_2+i{r}_2\sin {\theta }_2)\\ & =r_1{r}_2\cos {\theta }_1\cos {\theta }_2-r_1{r}_2\sin {\theta }_1\sin {\theta }_2+i{r}_1{r}_2(\cos {\theta }_1\sin {\theta }_2+\cos {\theta }_2\sin {\theta }_1)\\ & =r_1{r}_2\cos ({\theta }_1+{\theta }_2)+i{r}_1{r}_2\sin ({\theta }_1+{\theta }_2)\\ & =(r_1{r}_2,{\theta }_1+{\theta }_2)\end{array}$$

In essence, it’s equivalent to $r_1{e}^{i{\theta }_1}{r}_2{e}^{i{\theta }_2}=r_1{r}_2{e}^{i({\theta }_1+{\theta }_2)}$
So any number $n$ times $e^{i\theta }$ means rotate the point of that number with angle $\theta$.

4. What is De Moivre’s Theorem?
$$(\cos \theta +i\sin \theta {)}^n=\cos (n\theta )+i\sin (n\theta )$$

5. (ex) How to compute $\sqrt[6]{i}$?
$$i=e^{i(\frac{\pi }{2}+2k\pi )}$$
$$\sqrt[6]{i}=\sqrt[6]{1}{e}^{i(\frac{\pi }{2}+2k\pi )/6}=e^{i(\frac{\pi }{12}+\frac{k\pi }{3})}$$

Remark: Complex numbers are closed with respect to extracting roots.
While the real numbers don’t, like $\sqrt{-1}$.
Which means any polynomials with complex coefficients have complex roots.

2. Functions of complex Variable

1. Definition of function of complex variables?
Use a machine $f$ to map an input $x+iy$ represented by $z=(x,y)$ to output $w=(u,v)$
For example, $f(z)=z^2=(x+iy{)}^2$, so $f(x,y)=(x^2-y^2,2xy)$.

2. If we see $z$ as a vector, would $\frac{dw}{dz}$ is a constant value?
Yes, which means ${\lim }_{\Delta z\to 0}\frac{\Delta u+i\Delta v}{\Delta x+i\Delta y}=\text{Constant}$

case 1: $\Delta y=0$
which means $\Delta x\to 0$ horizontally
$$\begin{array}{rl}{f}^{\prime }(z_0)& =\underset{\Delta z\to 0}{\lim }[\frac{\Delta u}{\Delta x}+i\frac{\Delta v}{\Delta x}]\\ & =\frac{\partial u}{\partial x}+i\frac{\partial v}{\partial x}{|}_{z_0=(x_0,y_0)}\end{array}$$

case 2: $\Delta x=0$

$$\begin{array}{rl}{f}^{\prime }(z_0)& =\underset{\Delta z\to 0}{\lim }[\frac{\Delta u}{i\Delta y}+\frac{\Delta v}{\Delta y}]\\ & =-i\frac{\partial u}{\partial y}+\frac{\partial v}{\partial y}{|}_{z_0=(x_0,y_0)}\end{array}$$

Conclusion: To ensure the $f^{\prime }(z_0)$ has the same dirivative along $x$ and $y$ directions, we need to guarantee case 1 and case 2 has the same result. Which leads to
$$\{\begin{array}{l}{u}_x=v_y\\ u_y=-v_x\end{array}$$

Cauchy Riemann Conditions
case 3: If $f(x,y)=u(x,y)+iv(x,y)$ is differentiable (analytic function)
then $u_x=v_y,u_y=-v_x$, and vice versa.

(EX) $f(z)=(x+iy{)}^2=(x^2-y^2)+i(2xy)$

Solution:$u(x,y)=(x^2-y^2)$, $v(x,y)=2xy$.
Then we have $u_x=2x=v_y$, $u_y=-2y=-v_x.$

Also $\frac{(x_0+\Delta x{)}^2-x_0^2}{\Delta x}=2x_0$ also proves its truth.

(EX2) $f(z)=\bar{z}=x-iy$
Solution: $u_x=1,v_y=-1$. They are not equal which means function is not differentiable.

3. How to connect differentiability with Laplace’s Equation?
Let’s say we have $u+iv$ is differentiable.
Then $u_x=v_y,u_y=-v_x⟶u_{xx}=v_{yx},u_{yy}=-v_{xy}$.
So we have $u_{xx}+u_{yy}=0$
And the same can be achieved on imaginary parts.

Then we can say, if $u+iv$ is differentiable, then both real and imaginary parts satisfy Laplace’s equations.

3. Conformal Mapping

1. What is locally invertible for a complex mapping?

Solution: If $\det |\frac{\partial (u,v)}{\partial (x,y)}|=\det \left[\begin{array}{cc}{u}_x& u_y\\ v_x& v_y\end{array}\right]̸=0$ (This is Jacobian Matrix), then it’s locally invertible.

Also, remember if a complex mapping $z=(x,y)⟶w(u,v)$ is an analytic function, we have $u_x=v_y,u_y=-v_x$.

Then $\det |\frac{\partial (u,v)}{\partial (x,y)}|=u_x{v}_y-u_y{v}_x=u_x^2+v_x^2=|f^{\prime }(z){|}^2$ cannot be equal to 0.

Conclusion: If function $f=u+iv$ is analytic and $f^{\prime }(x)̸=0$, then this function is invertible.

2. Sometimes a function $f$ is invertible but not analytic, how to distinguish them?

Solution: We consider the functions $f$ embody both properties as “nicer” functions and call them conformal mapping.

3. What is a conformal mapping?

Solution: An invertible mapping is called conformal if it preserves the angles.

The “usual” linear mappings are not conformal.

Key Point: If $f=u+iv$ is analytic and $f^{\prime }(z)̸=0$, then $\{\begin{array}{l}u=u(x,y)\\ v=v(x,y)\end{array}$ is conformal.

(EX) Prove that if $z(x,y)⟶w(u,v)$ is analytic and a conformal mapping, then for $z_0\to w_0,z_1\to w_1,z_2\to w_2$, we have $\frac{z_2-z_0}{z_1-z_0}=\frac{w_2-w_0}{w_1-w_0}$, i.e., $\frac{\Delta w_2}{\Delta w_1}=\frac{\Delta z_2}{\Delta z_1}$ or $\theta =\phi$ (angle preservation).

Solution: Because $f^{\prime }(z_0)\approx \frac{\Delta w}{\Delta z}$ and derivative of complex function is independent of direction (analytic), we can have $\frac{\Delta w_2}{\Delta z_2}\approx \frac{\Delta w_1}{\Delta z_1}\approx f^{\prime }(z_0)̸=0$ (conformal), which leads to $\frac{\Delta w_2}{\Delta w_1}=\frac{\Delta z_2}{\Delta z_1}$, i.e. $\theta =\phi$.

Remark: If $f^{\prime }(z_0)=0$, we can’t get $\frac{\Delta w_2}{\Delta w_1}=\frac{\Delta z_2}{\Delta z_1}$ from $\frac{\Delta w_2}{\Delta z_2}\approx \frac{\Delta w_1}{\Delta z_1}\approx f^{\prime }(z_0)$.

Conformal maps and Laplace’s Equation

1. Then introduction of Conformal Maps.

Known: Region $R$ enclosed by some curve $C$ and we have a temperature distribution on $C$ and $R$.
We already know temperature satisfy Laplace’s equation in region $R$. That’s called steady-sate condition. I.e. $T_{xx}+T_{yy}=0$. Also $T=T_0(x,y)$ on $C$.
Then we can get an unique function defined in region $R$.

We need a transform from $(x,y)$ to $(u,v)$ to get a better integrand. We do the integration and then map back to $(x,y)$.
Note that, only analytic can not keep the statements like Laplace’s equation $T_{xx}+T_{yy}=0$ still true after transformation.
That’s why we need conformal maps.

(Theorem)
If $f:R\to S$ where $f=u+iv$ is analytic and $f^{\prime }(z)̸=0$. Then Laplace’s equation $T_{xx}+T_{yy}=0⟷T_{uu}+T_{vv}=0$. I.e. a conformal map preserves Laplace’s equation.

Real Case:
Suppose Region $R$ + curve $C$ conformally maps to Region $S$ and curve $C^{\prime }$. So we can trans form our problem $T=T_0(x,y)$ in $R$ to $T=T_1(u,v)$ in $S$ and solve it. Then plug $\{\begin{array}{l}u=u(x,y)\\ v=v(x,y)\end{array}$ back to $T_1(u,v)$, so we can get $T=T_0(x,y)$.

Proof:
Given $\{\begin{array}{l}u=u(x,y)\\ v=v(x,y)\end{array}$, assume $u,v$ are continuously differentiable functions.

By chain rule:
$$T_x=T_u{u}_x+T_v{v}_x$$
So
$$\begin{array}{rl}{T}_{xx}& =(T_u{u}_x+T_v{v}_x{)}_x\\ & =T_u{u}_{xx}+(T_u{)}_x{u}_x+T_v{v}_{xx}+(T_v{)}_x{v}_x\\ & =T_u{u}_{xx}+(T_{uu}{u}_x+T_{uv}{v}_x)u_x+T_v{v}_{xx}+(T_{vu}{u}_x+T_{vv}{v}_x)v_x\\ & =T_u{u}_{xx}+T_{uu}{u}_x^2+T_{uv}{u}_x{v}_x+T_v{v}_{xx}+T_{uv}{u}_x{v}_x+T_{vv}{v}_x^2\end{array}$$

By symmetric:

$$\begin{array}{rl}{T}_{yy}& =(T_u{u}_y+T_v{v}_y{)}_y\\ & =T_u{u}_{yy}+(T_u{)}_y{u}_y+T_v{v}_{yy}+(T_v{)}_y{v}_y\\ & =T_u{u}_{yy}+(T_{uu}{u}_y+T_{uv}{v}_y)u_y+T_v{v}_{yy}+(T_{vu}{u}_y+T_{vv}{v}_y)v_y\\ & =T_u{u}_{yy}+T_{uu}{u}_y^2+T_{uv}{u}_y{v}_y+T_v{v}_{yy}+T_{uv}{u}_y{v}_y+T_{vv}{v}_y^2\end{array}$$

Add them up:
$$\begin{array}{rl}{T}_{xx}+T_{yy}& =T_u(u_{xx}+u_{yy})+T_v(v_{xx}+v_{yy})+2T_{uv}(u_x{v}_x+u_y{v}_y)+T_{uu}(u_x^2+u_y^2)+T_{vv}(v_x^2+v_y^2)\end{array}$$

If Cauchy-Riemann Conditions are satisfied (analytic), we have $u_{xx}+u_{yy}=0,v_{xx}+v_{yy}=0$.
Also, $|f^{\prime }{|}^2=u_x^2+v_x^2=u_x^2+u_y^2=v_x^2+v_y^2$
Meanwhile, $u_x{v}_x+u_y{v}_y=v_y{v}_x-v_x{v}_y=0$
$$T_{xx}+T_{yy}=[T_{uu}+T_{vv}]|f^{\prime }(z){|}^2$$

If $f^{\prime }(z)̸=0$,
we then have
$$T_{xx}+T_{yy}=0⟷T_{uu}+T_{vv}=0$$

4. Sequences and Series

$$e^x=\sum _{n-0}^{\infty }\frac{x^n}{n!}$$

(Definition)
${\lim }_{n\to \infty }{a}_n=L$ means given $ϵ>0$, there exists $N$, then $n>N⟶|a_n-L|<ϵ$.
In similar way, we may define ${\sum }_{n=0}^{\infty }{c}_n={\lim }_{n\to \infty }(c_1+...+c_n)$

In particular, if S = { z : ∑ a n z n  converges } S = \{z:\sum a_nz^n \text{ converges}\} S={z:∑an​zn converges}
then either

1. S = { 0 } S= \{0\} S={0}

2. $S=C$ (all complex numbers).
3. There exists $R>0$ such that S = { z : ∣ z ∣ &lt; R } S =\{z: |z|&lt; R\} S={z:∣z∣<R}, and convergence is abosulte and uniform for $|z|\le r<R$.

Definition of lim of complex numbers

even if $a_n$ and $L$ are complex numbers, the definition still holds.

$a_n$ is any point in the disc with the origin of $L$ with radius < $ϵ$

What we mean lim here is that after a certain numbers any $n>N$, we have $a_n$ falls in the disc around $L$.

$$\begin{array}{rl}z& =r{e}^{i\theta }=r{e}^{i(\theta +2\pi k)}\\ \log z& =\log r+i(\theta +2\pi k)\end{array}$$

$\log z$ is multi-valued, so we can say principal value is $-\pi <\theta <\pi$

$\cosh ix=\frac{e^{ix}+e^{-ix}}{2}=\cos x$

$$\begin{array}{rl}{e}^z& =e^{x+iy}\\ & =e^x{e}^{iy}\\ & =e^x(\cos y+i\sin y)\\ & =e^x\cos y+i{e}^x\sin y\end{array}$$

Then $\{\begin{array}{l}u=e^x\cos y\\ v=e^{x}siny\end{array}$ is a conformal mapping.

Solution: Because the power series of $e^z$ exists, the fact the power series exists means its derivative exists.

If we square both sides, we can get
$$u^2+v^2=e^{2y}$$

$$\frac{v}{u}=\tan y$$

Application:

$\frac{1}{1-u}=1+u+u^2+...={\sum }_0^{\infty }{u}^n$ converges for $|u|<1$.

So $\frac{1}{1+x^2}=1-x^2+x^4-x^6+..+{\sum }_{n=0}^{\infty }(-1{)}^n{x}^{2n}$ converges when $|x|<1$

Question: Why we need $|x|<1$? Nothing goes wrong with $\frac{1}{1+x^2}$ apparently.

Solution: If we convert real-valued series to complex-valued series, look at ${\sum }_{n=0}^{\infty }(-1{)}^n{z}^{2n}=\frac{1}{1+z^2}=\frac{1}{(z+i)(z-i)}$

If $z=\pm i$ then obviously the equation doesn’t hold. Then go to Argand Diagram, if we have trouble on $(0,\pm i)$ all the points on the circle can’t be achieved. This include $(\pm 1,0)$.

5. Integrating Complex Functions

Review:

$$\begin{array}{rl}{\int }_{x_0}^{x_1}f(x)dx& =\underset{\max \Delta x_k\to 0}{\lim }\sum _{k=1}^{n}f(c_k^{\ast })\Delta x_k\\ & =F(x_1)-F(x_0)\end{array}$$

$$\begin{array}{rl}{\int }_{z_0}^{z_1}f(z)dz& =\underset{\max \Delta z_k\to 0}{\lim }\sum _{k=1}^{n}f(c_k^{\ast })\Delta z_k\end{array}$$

Now, we can find that unlike real case, where we do integration along x-axis. In xy-plane, we do integration along the curve $C$.

We constraint the curve $C$ to be smooth.

$$z=x(t)+iy(t),t_0\le t\le t_1$$

$${\int }_{z_0}^{z_1}f(z)dz=\underset{\max \Delta z_k\to 0}{\lim }\sum _{k=1}^n[f(c_k(t_k^{\ast }))\frac{\Delta z_k}{\Delta t_k}]\Delta t_k$$

$${\int }_{z_0}^{z_1}f(z)dz={\int }_{t_0}^{t_1}f(z(t))z^{\prime }(t)dt$$

In terms of $u$ and $v$, $f(z)=u+iv$, $\Delta z=\Delta x+i\Delta y$.

This leads to
$$\begin{array}{rl}{\int }_{z_0}^{z_1}f(z)dz& ={\int }_{(x_0.y_0)}^{(x_1,y_1)}(u+iv)(dx+idy)\\ & ={\int }_{(x_0,y_0)}^{(x_1,y_1)}(udx-vdy)+i{\int }_{(x_0,y_0)}^{(x_1,y_1)}(vdx+udy)\end{array}$$

If $u+iv$ is analytic $⟶u_x=v_y,u_y=-v_x⟶$ both the real part and imaginary part are exact.

When the diffenrential is exact, the line integral was dependent only on the end points but not the path connected the end points. So if the integrand is exact, the line integral around the closed curve is zero.

So if $f(z)$ is analytic, then ${\int }_{z_0}^{z_1}f(z)dz=F(z_1)-F(z_0)$ is independent of $C$ and ${\oint }_{C}f(z)dz=0$ when $f(z)$ is exact on and inside the curve.

EX: Compute ${\oint }_C\frac{dz}{z}$ and $C$ is a circle of radius $r$ around the origin, to show that it may be zero when $f(z)$ is not analytic.

Solution: Obviously, $f(z)=\frac{1}{z}$ is not analytic becuase it’s not differentiable at the origin $(0,0)$.

1. Apprach 01

$$\begin{array}{rl}{\oint }_C\frac{dz}{z}& ={\oint }_C\frac{dx+idy}{x+iy}\\ & ={\oint }_C\frac{(x-iy)(dx+idy)}{x^2+y^2}\\ & ={\oint }_C\frac{xdx+ydy}{x^2+y^2}+i{\oint }_C\frac{xdy-ydx}{x^2+y^2}\end{array}$$

Let $x=r\cos \theta ,y=r\sin \theta$

$$\begin{array}{rl}{\oint }_C\frac{dz}{z}& =0+i{\oint }_0^{2\pi }\frac{R^2({\sin }^2\theta +{\cos }^2\theta )}{R^2}d\theta \\ & =2\pi i̸=0\end{array}$$

2. Approach 02

$$C:Z=R{e}^{i\theta }0\le \theta \le 2\pi$$

$$\begin{array}{rl}\frac{dz}{d\theta }& =iR{e}^{i\theta }\\ {\oint }_C\frac{dz}{z}& ={\int }_0^{2\pi }\frac{1}{z(\theta )}\frac{dz}{d\theta }d\theta \\ & ={\oint }_0^{2\pi }\frac{iR{e}^{i\theta }}{R{e}^{i\theta }}d\theta \\ & ={\oint }_0^{2\pi }id\theta \\ & =2\pi i\end{array}$$

Theorem: Let’s imagine two curves $C_1$ and $C_2$ and $C_1$ is inside the region enclosed by $C_2$. If $f(z)$ is analytic on and between $C_1$ and $C_2$ then ${\oint }_{C_1}f(z)dz={\oint }_{C_2}f(z)dz$.

Which means any closed curve $C$ enclose origin has ${\oint }_C\frac{dz}{z}=2\pi i$.