Complex Analysis Notes

Math 444 - Spring 2023

Jump to week: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14

Week 1

Tentative Schedule

Day Section Topic
Mon, Jan 16 1.1 Definitions and algebraic properties
Wed, Jan 18 1.2 From algebra to geometry and back
Fri, Jan 20 1.3 Geometry of complex numbers

Monday, Jan 16

We did these exercises in class.

  1. Simplify 14+3i\displaystyle\frac{1}{4+3i} by rationalizing the denominator.

The next two problems can be solved using either rectangular or polar form:

  1. Find (1+i)2(1+i)^2.

  2. Find 11+i\dfrac{1}{1+i}.

Finally, some problems are definitely easier using polar form:

  1. Find i\sqrt{-i}.

Wednesday, Jan 18

Today we derived the angle addition formulas in trigonometry from Euler’s formula. We also looked at some other trigonometry facts like the double and triple angle formulas and the law of cosines.

Friday, Jan 20

On Friday, we talked about sets of complex numbers and how multiplication by a complex number z=Reiθz = R e^{i \theta} stretches a set by a factor of RR and rotates the set by an angle of θ\theta (in the counter-clockwise direction). We also talked about the triangle inequality which says that |z±w||z|+|w||z \pm w| \le |z|+|w| for any two complex numbers zz and ww.


Week 2 Notes

Tentative Schedule

Day Section Topic
Mon, Jan 23 1.4 Elementary topology of the plane
Wed, Jan 25 1e Chapter 1 recap
Fri, Jan 27 2.1 Limits and continuity

Monday, Jan 23

Today we talked about topology which is the study of how sets are connected. We defined the following: neighborhoods, interior points, open sets, closed sets, interiors, closures, and boundaries. These are all standard definitions (which are in the book).

We also defined the following which is not in the book:

A set AA \subseteq \mathbb{C} touches a point zz \in \mathbb{C} if every neighborhood of zz contains an element of AA. Then a set is closed if it contains every element it touches.

We used the notion of touch to define continuous functions. We finished by talking about paths which are continuous functions from an interval [a,b][a,b] to \mathbb{C}.

Wednesday, Jan 25

We talked a little more about topology, including bounded sets and path connected sets. Then we looked at using Python to work with complex numbers.

Friday, Jan 27

On Friday we did this workshop about sequences and continuity. We did #1-3 in class.


Week 3 Notes

Tentative Schedule

Day Section Topic
Mon, Jan 30 2.2 Differentiability and holomorphicity
Wed, Feb 1 2.2 Holomorphic implies conformal
Fri, Feb 3 2.3 The Cauchy-Riemann equations

Monday, Jan 30

We proved that the product of two continuous functions is continuous. We also talked about how to calculate limits of complex functions and we did these examples:

  1. limziz2\displaystyle\lim_{z \rightarrow i} z^2.

  2. limz0zz¯\displaystyle\lim_{z \rightarrow 0} \dfrac{z}{\overline{z}}.

  3. Exercise 2.6.

Wednesday, Feb 1

Today we talked about derivatives of complex functions. We did the following examples:

  1. Use the definition to find the derivative of z2z^2.

  2. Show that the derivative of f(z)=z¯f(z) = \overline{z} does not exist.

We talked about properties of the derivative including the power rule, chain rule, product rule, and quotient rule. Then we talked about the meaning of the derivative in terms of linear approximations using these two apps:

Friday, Feb 3

On Friday, we talked about the Cauchy-Riemann Equations. Here is a nice video explanation:

In class we calculated the Cauchy-Riemann equations for the following examples to see whether they are (complex) differentiable:

  1. f(z)=z2f(z) = z^2.

  2. f(z)=z¯f(z) = \overline{z}.


Week 4 Notes

Tentative Schedule

Day Section Topic
Mon, Feb 6 2.3 The Cauchy-Riemann equations - con’d
Wed, Feb 8 2.4 Constant Functions
Fri, Feb 10 2e Chapter 2 recap

Monday, Feb 6

Today we talked more about the Cauchy-Riemann equations. We looked at two examples:

  1. f(z)=ezf(z) = e^z

  2. This is Exercise 2.19 in the book: Define f(z)=0f(z) = 0 if zz is on either purely real or purely imaginary, and f(z)=1f(z) = 1 otherwise. Show that ff satisfies the Cauchy–Riemann equations at z=0z = 0, yet ff is not differentiable at z=0z = 0. Why doesn’t this contradict the theorem about a complex function being complex differentiable if and only if it satisfies the Cauchy-Riemann equations?

Then we talked about smooth paths which are paths that are differentiable. If γ\gamma is a smooth path in \mathbb{C} and ff is holomorphic, then the chain rule holds and says that ddtf(γ(t))=f(γ(t))γ(t).\displaystyle\frac{d}{dt} f(\gamma(t)) = f'(\gamma(t)) \cdot \gamma'(t).

  1. Find the derivative of the path γ(t)=eit\gamma(t) = e^{it}. What does it mean?

  2. If f(z)=z2f(z) = z^2 and γ(t)=t2+it\gamma(t) = t^2 + i t, then find the derivative of the path f(γ(t))f(\gamma(t)) at t=1t=1.

Then we proved that if ff is holomorphic on a connected open set GG and f(z)=0f'(z) = 0 everywhere on GG, then ff must be constant.

Wednesday, Feb 8

Today we reviewed Homework 3. We also talked about the function f(z)=zf(z) = \sqrt{z}. This lead to a discussion of the principal branch of the square root function, and the idea of a branch cut.

Friday, Feb 10

Today we introduced Möbius transformations and talked about how they transform lines and circles into lines and circles.


Week 5 Notes

Tentative Schedule

Day Section Topic
Mon, Feb 13 3.1 Möbius transforms
Wed, Feb 15 3.4 Exponential and trigonometric functions
Fri, Feb 17 3.4 Exponential and trigonometric functions - con’d

Monday, Feb 13

Today we talked more about Möbius transformations. We proved that you can compose two Möbius transformations by multiplying their coefficient matrices. We also talked about the extended complex plane {}\mathbb{C} \cup \{ \infty \}. A Möbius transformation is a one-to-one and onto map from the extended complex plane to itself. Furthermore, a Möbius transformation is completely defined by what it does to any three elements of {}\mathbb{C} \cup \{ \infty \}.

Wednesday, Feb 15

We introduced the complex sine and cosine functions and the complex logarithm.

Friday, Feb 17

We talked about the principle branch of the complex power function zαz^\alpha.


Week 6 Notes

Tentative Schedule

Day Section Topic
Mon, Feb 20 3.5 Logarithms and complex exponentials
Wed, Feb 22 3e Chapter 3 recap
Fri, Feb 24 Midterm 1

Monday, Feb 20

Today we wrapped up our discussion of logarithms, exponentials, and power functions. We proved the following in class:

  1. If zz \in \mathbb{C}, then (ez)¯=ez¯\overline{(e^z)} = e^{\overline{z}}.

  2. If z{0}z \in \mathbb{C}\backslash \{0 \}, then Log(z)¯=Log(z¯)\overline{\operatorname{Log}(z)} = \operatorname{Log}(\overline{z}).

As a consequence, it follows that for any complex z,αz, \alpha \in \mathbb{C}, zα¯=z¯α¯\overline{z^\alpha} = \overline{z}^\overline{\alpha}. Not every complex function commutes with the conjugate function. In fact, we showed that:

  1. In order for a function f:f: \mathbb{C}\rightarrow \mathbb{C} to commute with complex conjugation, it is necessary that f()f(\mathbb{R}) \subseteq \mathbb{R}.

We also looked at Problem #7 from HW6. And we finished with this exercise:

  1. What is the image of the square with corners 0,i,1,0, i, 1, and 1+i1+i under the function f(z)=ezf(z) = e^z?

Week 7 Notes

Tentative Schedule

Day Section Topic
Mon, Feb 27 4.1 Complex integration
Wed, Mar 1 4.1 Complex integration - con’d
Fri, Mar 3 4.2 Antiderivatives

Monday, Feb 27

Today we introduced complex integrals. We did these examples:

  1. γ(z¯)2dz\displaystyle\int_\gamma (\overline{z})^2 \, dz on the path γ(t)=t(1+i)\gamma(t) = t(1+i) with t[0,1]t \in [0,1].

  2. δ(z¯)2dz\displaystyle\int_\delta (\overline{z})^2 \, dz on the path δ(t)=t+it2\delta(t) = t+it^2 with t[0,1]t \in [0,1]. (Link: Sympy code)

  3. γ1zdz\displaystyle\int_\gamma \frac{1}{z} \, dz where γ\gamma is the unit circle γ(t)=eit\gamma(t) = e^{it} with t[0,2π]t \in [0,2\pi].

  4. Use the formula length(γ)=ab|γ(t)|dt\displaystyle\operatorname{length}(\gamma) = \int_a^b |\gamma'(t)| \, dt to find the length of the unit circle.

  5. Use the formula |γf(z)dz|maxzγ|f(z)|length(γ)\displaystyle\left| \int_\gamma f(z) \, dz \right| \le \max_{z \in \gamma} |f(z)| \cdot \operatorname{length}(\gamma) to estimate an upper bound for |γ1z4+16dz|\displaystyle\left| \int_\gamma \frac{1}{z^4 + 16} \, dz \right| when γ\gamma is the unit circle.

Friday, Mar 3

Today we proved Cauchy’s Theorem using Green’s Theorem (see this video for a nice explanation of Green’s theorem).

Cauchy’s Theorem: If ff is holomorphic on a simply connected open set DD \subseteq \mathbb{C}, and CC is a (piecewise smooth) simple closed path in DD, then Cf(z)dz=0.\oint_C f(z) \, dz = 0.

As a corollary, we showed that when a function f(z)f(z) is holomorphic on a simply connected open set DD, then the integral on any two paths γ1\gamma_1 and γ2\gamma_2 that both start at a point z0z_0 and end at z1z_1 and stay inside DD must be the same:

γ1f(z)dz=γ2f(z)dz.\int_{\gamma_1} f(z) \, dz = \int_{\gamma_2} f(z) \, dz.

In other words, the value of the integral is independent of the path.


Week 8 Notes

Tentative Schedule

Day Section Topic
Mon, Mar 13 4.3 Cauchy’s theorem
Wed, Mar 15 4.4 Cauchy’s integral formula
Fri, Mar 17 4e Chapter 4 recap

Monday, Mar 13

Today we reviewed Cauchy’s theorem. We pointed out that many of the assumptions we made last time aren’t really necessary. The general version of Cauchy’s theorem says:

Cauchy’s Theorem. Let ff be holomorphic on a simply connected open set DD \subseteq \mathbb{C}. Then Cf(z)dz=0\oint_C f(z) \, dz = 0 for every piecewise smooth closed path CC in DD.

Our proof using Green’s theorem required extra assumptions that aren’t really necessary, but without them the proof gets more complicated:

  1. We assumed that ff' is continuous.
  2. We assumed that our paths were simple.

We reviewed consequences of Cauchy’s theorem like path independence of integrals. Then we defined an antiderivative of a complex function. If you know FF is an antiderivative of ff and γ:[a,b]\gamma: [a,b] \rightarrow \mathbb{C} is a piecewise smooth path, then the evaluation theorem says that γf(z)dz=F(γ(b))F(γ(a)).\int_\gamma f(z) \, dz = F(\gamma(b)) - F(\gamma(a)).

We finished by proving that any holomorphic function ff on a simply connected open set DD always has an antiderivative. If the domain DD is not simply connected, then you have to assume that ff is continuous and the integral of ff on any closed path in the domain is 0 (see Theorem 4.15 in the book).

Wednesday, Mar 15

Today we introduced the Cauchy integral formula which extends Cauchy’s theorem. The proof required two key insights:

  1. If ff is complex differentiable at ww, then f(z)=f(w)+f(w)(zw)+ϵ(z),f(z) = f(w) + f'(w) (z-w) + \epsilon(z), where the error term ϵ(z)\epsilon(z) satisfies limzwϵ(z)zw=0\lim_{z \rightarrow w} \frac{\epsilon(z)}{z-w} = 0.

  2. If C1C_1 and C2C_2 are two different positively oriented piecewise smooth closed paths in a region where ff is holomorphic, then C1f(z)zwdz=C2f(z)zwdz.\int_{C_1} \frac{f(z)}{z-w} \, dz = \int_{C_2} \frac{f(z)}{z-w} \, dz.

Then we used the Cauchy integral formula to evaluate the following integrals.

  1. Cezz1dz\displaystyle\int_C \frac{e^z}{z-1} \, dz where CC is the square with vertices at 10,10i,10,10i10, 10i, -10, -10i. (https://youtu.be/NJap6Vm5mEk)

  2. |zi|=11z2+1dz\displaystyle\oint_{|z-i|=1} \frac{1}{z^2+1} \, dz

  3. |z|=3ezz22zdz\displaystyle\oint_{|z|=3} \frac{e^z}{z^2 - 2z} \, dz

Friday, Mar 17

We started by proving the geometric sum formula: 1+z+z2+=11z1 + z + z^2 + \ldots = \frac{1}{1-z} for any zz \in \mathbb{C} with |z|<1|z| < 1. Then use that to prove this:

Theorem If ff is holomorphic in an open set containing a closed disk BR(z0)B_R(z_0), then f(z)f(z) has a power series f(z)=n=0an(zz0)nf(z) = \sum_{n = 0}^\infty a_n (z-z_0)^n which converges absolutely inside the disk BR(z0)B_R(z_0). The coefficients of the power series are an=12πi|ξz0|=Rf(ξ)(ξz0)n+1dξ.a_n = \frac{1}{2\pi i} \oint_{|\xi-z_0|=R} \frac{f(\xi)}{(\xi-z_0)^{n+1}} \, d\xi.

In the special case where z0=0z_0 = 0 you get a Maclaurin series with coefficients: an=12πi|ξ|=Rf(ξ)ξn+1dξ.a_n = \frac{1}{2\pi i} \oint_{|\xi|=R} \frac{f(\xi)}{\xi^{n+1}} \, d\xi. This follows from Cauchy’s integral formula and the geometric series formula f(z)=12πi|ξ|=Rf(ξ)ξzdξ=12πi|ξ|=Rf(ξ)ξ(n=0znξn)dξ.f(z) = \frac{1}{2\pi i} \oint_{|\xi|=R} \frac{f(\xi)}{\xi - z} \, d\xi = \frac{1}{2\pi i} \oint_{|\xi|=R} \frac{f(\xi)}{\xi} \left( \sum_{n = 0}^\infty \frac{z^n}{\xi^n} \right) \, d\xi.

Note that the coefficients are all bounded in absolute value by max|ξ|=R|f(ξ)|Rn\displaystyle\max_{|\xi|=R} \frac{|f(\xi)|}{R^n} so the series does converge absolutely as long as |z|<R|z| < R.


Week 9 Notes

Tentative Schedule

Day Section Topic
Mon, Mar 20 5.1 Cauchy’s integral formula for derivatives
Wed, Mar 22 5.2 Antiderivatives again
Fri, Mar 24 5.3 Entire functions and Liouville’s theorem

Monday, Mar 20

Today we talked about Cauchy’s integral formula for derivatives. We did these examples in class:

  1. |z|=3eiz(z+i)2dz\displaystyle\oint_{|z| = 3} \frac{e^{iz}}{(z+i)^2} \, dz. (https://youtu.be/WJOf4PfoHow)

  2. |z|=1z2+1z(2z+1)dz\displaystyle\oint_{|z| = 1} \frac{z^2+1}{z(2z+1)} \, dz. (https://youtu.be/APoh2B5S2ok)

We finished by using a complex contour integral to find the real integral 1x4+1dx\displaystyle\int_{-\infty}^{\infty} \frac{1}{x^4+1} \, dx.

Wednesday, Mar 22

Today we talked about two theorems:

Morera’s Theorem If f:Df:D \rightarrow \mathbb{C} is continuous in an open connect set DD \subseteq \mathbb{C}, and γf(z)dz=0\int_\gamma f(z) \, dz = 0 for every piecewise smooth closed path γ\gamma in DD, then ff is holomorphic.

In class last week we proved that the functions described above have a well-defined antiderivative function FF. But then, Cauchy’s general integral formula guarantees that FF is holomorphic, and so are all of its derivatives, including ff!

Then we looked at another interesting concept:

Definition The winding number of a closed curve γ\gamma around a point z0z_0 \in \mathbb{C} 12πiγ1zz0dz\frac{1}{2\pi i} \int_\gamma \frac{1}{z-z_0} \, dz (assuming that γ\gamma does not intersect z0z_0).

Theorem If p(z)p(z) is a polynomial with no roots on the unit circle, then the winding number of p(eit)p(e^{it}) around the origin is equal to the number of roots of pp inside the unit circle.

We proved this theorem by doing the following exercises in class:

  1. If γ(t)=p(eit)\gamma(t) = p(e^{it}), show that γ1zz=|z|=1p(z)p(z)dz\displaystyle\int_\gamma \frac{1}{z} \, z = \oint_{|z|=1} \frac{p'(z)}{p(z)} \, dz.

  2. If f(z)=(zw)mg(z)f(z) = (z-w)^m g(z) where g(z)g(z) is holomorphic and g(w)0g(w) \ne 0 in a disk of radius RR around ww, then 12πi|zw|=rf(z)f(z)dz=m\displaystyle\frac{1}{2\pi i } \oint_{|z-w| = r} \frac{f'(z)}{f(z)} \, dz = m for any 0<r<R0 < r< R.

Friday, Mar 24

Today we finished the proof of the theorem from Wednesday. We observed that for a polynomial p(z)=(zr1)m1(zr2)m2(zrk)mkp(z) = (z-r_1)^{m_1} (z-r_2)^{m_2} \cdots (z-r_k)^{m_k}, the winding number of γ(t)=p(eit)\gamma(t) = p(e^{it}), 0t2π0 \le t \le 2 \pi around the origin will be 12πiγ1zdz=12πi|z|=1p(z)p(z)dz=|rj|<1mj.\frac{1}{2\pi i} \oint_\gamma \frac{1}{z} \, dz = \frac{1}{2\pi i} \oint_{|z|=1} \frac{p'(z)}{p(z)} \, dz = \sum_{|r_j|<1} m_j.

Then we talked about entire functions which are functions that are holomorphic everywhere in \mathbb{C}.

Liouville’s Theorem. Any bounded entire function is constant.

To prove Liouville’s theorem, we need a quick observation and an idea. The observation is that if f(z)=0f'(z) = 0 everywhere, then ff must be constant. The idea is that we can use Cauchy’s integral formula for derivatives to calculate f(z)f'(z) if ff is bounded and entire: f(z)=12πi|ξz|=Rf(ξ)(ξz)2dξ.f'(z) = \frac{1}{2\pi i} \oint_{|\xi - z| = R} \frac{f(\xi)}{(\xi - z)^2} \, d\xi. What happens if the radius RR gets really really big? Use the max-times-length inequality to estimate |f(z)||f'(z)|.

We finished by applying Liouville’s theorem to prove:

The Fundamental Theorem of Algebra Every nonconstant polynomial with complex coefficients has a complex root.


Week 10 Notes

Tentative Schedule

Day Section Topic
Mon, Mar 27 5e Chapter 5 recap
Wed, Mar 29 6.1 Harmonic functions
Fri, Mar 31 6.1 Fields and flows

Monday, Mar 27

Today we gave a completely different proof of the Fundamental Theorem of Algebra by thinking comparing the contour p(Reit)p(Re^{i t}) when the radius RR is small versus when it is large. Here is a video that explains the proof we did:

Before we did that, we reviewed some facts about polynomials.

Rational Root Theorem. If p(z)=anzn++a1z+a0p(z) = a_n z^n + \ldots + a_1 z + a_0 has integer coefficients, then any rational root of pp must be expressed as a reduced fraction md\frac{m}{d} where mm divides a0a_0 and dd divides ana_n.

We used the rational root theorem to check the possible rational roots of p(z)=z33z2+z3p(z) = z^3 - 3z^2 + z - 3. We found that z=3z = 3 is a root.

Division Algorithm for Polynomials. If p(z)p(z) and d(z)d(z) are polynomials with complex coefficients and degree(d)>0\operatorname{degree}(d) > 0, then there are unique polynomials q(z)q(z) and r(z)r(z) with degree(r)<degree(d)\operatorname{degree}(r) < \operatorname{degree}(d) such that p(z)=q(z)d(z)+r(z).p(z) = q(z) d(z) + r(z).

In particular, this means that z33z2+z3=(z3)(z2+1)+ a remainderz^3 - 3z^2 + z - 3 = (z-3)(z^2 + 1) + \text{ a remainder}. But since 33 is a root, it is clear that the remainder must be zero. So (z3)(z-3) is on factor. Factoring (z2+1)(z^2+1), we see that z33z2+z3=(z3)(zi)(z+i).z^3 - 3z^2 + z - 3 = (z-3)(z - i)(z+i). In general, we have the following corollary of the Fundamental Theorem of Algebra:

Theorem. An n-th degree polynomial p(z)=anzn++a0p(z) = a_n z^n + \ldots + a_0 with complex coefficients always factors into the form an(zr1)(zr2)(zrn)a_n(z-r_1) (z-r_2) \cdots (z-r_n) where reach rkr_k is a root of p(z)p(z) (repeats are allowed).

We used these ideas to completely factor these two polynomials:

  1. p(z)=z33z2+z3p(z) = z^3 - 3z^2 + z - 3

  2. p(z)=z3+1p(z) = z^3 + 1

Wednesday, Mar 29

We started by proving this theorem which effectively says that f(w)f(w) is the average of the values of ff in a circle around ww:

Theorem. If ff is holomorphic in a region that contains a closed disk BR¯(w)\overline{B_R}(w), then f(w)=12π02πf(w+Reit)dt.f(w) = \frac{1}{2\pi} \int_0^{2\pi} f(w + R e^{i t}) \, dt.

After we proved that, we introduced harmonic functions which are functions u:2u:\mathbb{R}^2 \rightarrow \mathbb{R} such that 2ux2+2uy2=0.\frac{\partial^2 u}{\partial x^2} + \frac{\partial^2 u}{\partial y^2} = 0.

  1. Show that u(x,y)=x33xy2u(x,y) = x^3 - 3xy^2 is harmonic.

  2. Construct a real-valued function v(x,y)v(x,y) such that u+ivu+iv satisfies the Cauchy-Riemann equations.

In fact, for every harmonic function uu, it is always possible to construct another harmonic function vv such that u+ivu+iv is holomorphic (at least locally). The function vv is called a harmonic conjugate of uu. The converse is true as well, if ff is holomorphic, then both the real and imaginary parts of ff are harmonic functions.


Week 11 Notes

Tentative Schedule

Day Section Topic
Mon, Apr 3 No class
Wed, Apr 5 7.4 Regions of convergence
Fri, Apr 7 7e Chapter 7 review

Wednesday, Apr 5

Today we did a review of infinite series and power series. We started with the following two definitions: an infinite series converges if its partial sums converge. A series n=0an\sum_{n = 0}^\infty a_n converges absolutely if n=0|an|\sum_{n=0}^\infty |a_n| converges. We proved the following theorem:

Theorem If a series n=0an\sum_{n = 0}^\infty a_n converges absolutely, then it also converges.

Then we talked about power series which are series of the form n=0an(zc)n\sum_{n=0}^\infty a_n (z-c)^n where the coefficients ana_n and the center cc are complex numbers and zz is a variable.

Theorem A power series n=0an(zc)n\sum_{n=0}^\infty a_n (z-c)^n has a radius of convergence R=limn|an||an+1|.R = \lim_{n \rightarrow \infty} \frac{|a_n|}{|a_{n+1}|}. The power series converges absolutely for all zz inside the open disk with radius RR around center cc. It diverges outside the closed disk of radius RR around cc. On the boundary, the series might converge or diverge.

We looked at the following examples:

  1. Consider the following power series for Log(z)=(z1)(z1)22+(z1)33(z1)44+.\operatorname{Log}(z) = (z-1) - \frac{(z-1)^2}{2} + \frac{(z-1)^3}{3} - \frac{(z-1)^4}{4} + \ldots.

    1. Re-write this series in Σ\Sigma-notation.
    2. What are the coefficients ana_n of this power series?
    3. What is the center and radius of convergence?
  2. Find the radius of convergence for this power series n=0znn!\displaystyle\sum_{n = 0}^{\infty} \frac{z^n}{n!}.

  3. Find power series for the following functions by using the Maclaurin series for eze^z, sinz\sin z, and cosz\cos z:

    1. ez2/2e^{-z^2/2}
    2. sinzz\dfrac{\sin z}{z}

Friday, Apr 7

Today we reviewed some of the questions from HW10 and HW11. Then we worked on proving:

Theorem A power series n=0an(zc)n\displaystyle\sum_{n=0}^\infty a_n (z-c)^n converges absolutely when |zc|<R=limn|an||an+1|.|z-c| < R = \lim_{n \rightarrow \infty} \frac{|a_n|}{|a_{n+1}|}. (assuming the limit exists).

Proof. By the definition of limits, for any ϵ>0\epsilon > 0, there exists an NN big enough so that Rϵ<|an||an+1|<R+ϵR-\epsilon < \frac{|a_n|}{|a_{n+1}|} < R+ \epsilon for all nNn \ge N. This means that |an+1|<|an|(Rϵ)|a_{n+1}| < \frac{|a_n|}{(R-\epsilon)} for all n>Nn > N.

Exercise. Use mathematical induction to prove that |an||aN|(Rϵ)(nN)|a_n| \le \frac{|a_N|}{(R-\epsilon)^(n-N)} for all nNn \ge N.

Now, we can choose ϵ>0\epsilon > 0 small enough so that |zc|<(Rϵ)|z-c| < (R - \epsilon). Then we have n=0|an||zc|n=n=0N1|an||zc|n+n=N|an||zc|n.\sum_{n = 0}^{\infty} |a_n|\, |z-c|^n = \sum_{n = 0}^{N-1} |a_n|\, |z-c|^n + \sum_{n=N}^\infty |a_n| \, |z-c|^n. The first NN terms are not important, since they definitely have a finite sum. So we focus on the right side terms. Using the inequality from the induction exercise, we have:

n=N|an||zc|nn=N|aN|(Rϵ)(nN)|zc|n.\sum_{n = N}^{\infty} |a_n|\, |z-c|^n \le \sum_{n = N}^{\infty} \frac{|a_N|}{(R-\epsilon)^{(n-N)}} |z-c|^n.

That is a geometric series with common ratio (|zc|(Rϵ))<1\displaystyle\left( \frac{|z-c|}{(R-\epsilon)} \right) < 1. So it converges. □

In class I mixed up the end of the proof because I chose ϵ\epsilon so that |zc|=Rϵ|z-c| = R - \epsilon instead of |zc|<Rϵ|z-c| < R - \epsilon. That prevented the series from converging, but you can choose any ϵ>0\epsilon > 0 as small as you want, so you can definitely find one so that |zc|<Rϵ|z-c| < R- \epsilon.


Week 12 Notes

Tentative Schedule

Day Section Topic
Mon, Apr 10 8.1 Power series and holomorphic functions - con’d
Wed, Apr 12 Review
Fri, Apr 14 Midterm 2

Monday, Apr 10

Since I messed up the proof that power series converge absolutely inside their radius of convergence last Friday, I figured I’d start today be fixing the argument. But I also wanted to talk about the intuition behind the argument. To create a proof, you typically need to find an intuition for why the claim is true. Here is the intuition:

n=0|an||zc|n=|a0|+|a1||zc|+|a2||zc|2+|a3||zc|3+\sum_{n = 0}^\infty |a_n| \, |z-c|^n = |a_0| + |a_1| |z-c| + |a_2| |z-c|^2 + |a_3| |z-c|^3 + \ldots

Unlike geometric series, power series don’t have a common ratio. The ratio between consecutive terms here is: |an+1||an||zc|.\frac{|a_{n+1}|}{|a_n|} |z-c|. In the long run, this approaches |zc|R,\frac{|z-c|}{R}, which explains why the series will converge when |zc|<R|z-c| < R. In the long run, it behaves like a geometric series with that as its common ratio.

After that, we talked about the zeros of an analytic function. An analytic function is one with a converging power series in a disk with positive radius. Analytic functions are the same as holomorphic functions.

We proved the following theorem:

Theorem If ff is a non-constant analytic function in an open connected domain DD \subseteq \mathbb{C}, then the zeros of ff are isolated.

This lead to the following definition. The order of a zero z0z_0 of f(z)f(z) is the smallest NN such that the coefficient aNa_N in the power series for ff centered at z0z_0 is nonzero.

Find the orders for the following zeros:

  1. z0=0z_0 = 0 for f(z)=zsinzf(z) = z-\sin z.

  2. z0=πiz_0 = \pi i for g(z)=ez+1g(z) = e^z + 1.


Week 13 Notes

Tentative Schedule

Day Section Topic
Mon, Apr 17 8.2 Classification of zeros
Wed, Apr 19 8.2 The identity principle
Fri, Apr 21 8e Chapter 8 recap

Monday, Apr 17

Today we reviewed the concept of zeros and winding numbers. We updated an old proof to show that if γ:[0,2π)\gamma:[0,2\pi) \rightarrow \mathbb{C} is a piecewise smooth simple closed curve in a simply connected open set where ff is holomorphic, then the winding number of f(γ(t))f(\gamma(t)) around ww \in \mathbb{C} is equal to the sum of the orders of the zeros of fwf-w that are enclosed by γ\gamma.

This is because the winding number must be equal to 12πif(γ)1zwdz\frac{1}{2\pi i} \int_{f(\gamma)} \frac{1}{z-w} \, dz by Cauchy’s integral formula. Then by the definition of complex contour integrals, this is the same as: 12πiγf(z)f(z)wdz.\frac{1}{2\pi i} \int_\gamma \frac{f'(z)}{f(z)-w} \, dz.

  1. Show that if fwf-w has a zero z0z_0 of order mm at ww, that is, f(z)w=(zz0)mg(z)f(z)-w = (z-z_0)^m g(z) where gg is holomorphic and g(z0)0g(z_0) \ne 0, then f(z)f(z)w=g(z)g(z)+m(zz0).\frac{f'(z)}{f(z)-w} = \frac{g'(z)}{g(z)} + \frac{m}{(z-z_0)}.

  2. If z0z_0 is the only zero of fwf-w inside γ\gamma, explain why this proves that the winding number of f(γ)f(\gamma) around ww is mm.

  3. What if z0z_0 is not the only zero of fwf-w inside γ\gamma?

Once we finished the proof, we observed the following: If ff is holomorphic in a open region RR and f(z)w0f(z)-w_0 has a zero of order mm at z0Rz_0 \in R, then f(γ)f(\gamma) winds around w0w_0 exactly mm times when γ\gamma is a small circle around z0z_0. Therefore f(γ)f(\gamma) also winds around any ww close to w0w_0 exactly mm times as well. As a corollary, we have:

Open Mapping Theorem. If ff is holomorphic on an open set UU \subseteq \mathbb{C}, then f(U)f(U) is an open set.

Another immediate corollary is the following theorem:

Maximum Modulus Principle. If ff is holomorphic on an open set UU, then |f||f| cannot have a local maximum in UU.

Proof. Suppose that z0z_0 is a local maximum, that is, |f(z0)||f(z)||f(z_0)| \ge |f(z)| for all zz in a small disk around z0z_0. This is a contradiction because f(Br(z0))f(B_r(z_0)) is an open set, so it contains an open disk around f(z0)f(z_0). \square

Wednesday, Apr 19

Today we did two applications of contour integrals to calculate real integrals:

  1. 1(x21)(x29)dx\displaystyle\int_{-\infty}^{\infty} \frac{1}{(x^2-1)(x^2-9)} \, dx.

  2. 02π1106cosθdθ\displaystyle\int_0^{2 \pi} \frac{1}{10 - 6 \cos \theta} \, d\theta.

To convert integrals of functions involving sine & cosine from 0 to 2π, you can use the following substitutions, all based on letting z=eiθz = e^{i \theta} and integrating over the unit circle:

Friday, Apr 21

Today we talked about the Fourier transform of a real function f:f:\mathbb{R}\rightarrow \mathbb{R} which is defined to be:

(f)=f(x)eiαxdx=F(α).\mathcal{F}(f) = \int_{-\infty}^\infty f(x) e^{i\alpha x} \, dx = F(\alpha).

The inverse Fourier transform is

1(F)=F(α)eiαxdα=f(x).\mathcal{F}^{-1}(F) = \int_{-\infty}^\infty F(\alpha) e^{-i\alpha x} \, d\alpha = f(x).

In class we used a contour integral to find the inverse Fourier transform of

  1. F(α)=21+α2\displaystyle F(\alpha) = \frac{2}{1+\alpha^2}

We also calculated the Fourier transform of the Gaussian f(x)=ex2f(x) = e^{-x^2}. Before calculating that, we needed to know that ex2dx=π\displaystyle\int_{-\infty}^\infty e^{-x^2} \, dx = \sqrt{\pi}. Here is a video that explains this.

Then we used a rectangular contour to show that (ex2)=πeα2/4\mathcal{F}(e^{-x^2}) = \sqrt{\pi} e^{-\alpha^2/4}.


Week 14 Notes

Tentative Schedule

Day Section Topic
Mon, Apr 24 6.1 Harmonic functions
Wed, Apr 26 6.2 Mean value & maximum principle
Fri, Apr 28 6e Chapter 6 recap

Monday, Apr 24

Today we used the Fourier transform to solve the 1-dimensional heat equation: Tt=2T2x\frac{\partial T}{\partial t} = \frac{\partial^2 T}{\partial^2 x} with initial condition T(0,x)=f(x)T(0,x) = f(x).

We started with some properties of the Fourier transform. Suppose f(x)f(x) is a real function with Fourier transform f̂(α)\hat{f}(\alpha). Then we have:

  1. Derivatives. (f(x))=iαf̂(α)\displaystyle\mathcal{F}(f'(x)) = - i\alpha \hat{f}(\alpha).

  2. Convolution. (f*g)=f̂(α)ĝ(α)\displaystyle\mathcal{F}(f \ast g) = \hat{f}(\alpha) \hat{g}(\alpha).

  3. Scaling. (f(cx))=1|c|f̂(αc)\displaystyle\mathcal{F}(f(cx)) = \tfrac{1}{|c|} \hat{f} \left( \tfrac{\alpha}{c} \right) and 1(f̂(cα))=1|c|f(xc)\displaystyle\mathcal{F}^{-1}(\hat{f}(c\alpha)) = \tfrac{1}{|c|} f \left( \tfrac{x}{c} \right)

Then we used these steps to solve the heat equation.

Wednesday, Apr 26

Today we started by looking at the 2-dimensional heat equation Tt=2Tx2+2Ty2.\frac{\partial T}{\partial t} = \frac{\partial^2 T}{\partial x^2} + \frac{\partial^2 T}{\partial y^2}.

We used a contour integral with Green’s theorem to show that the next flow of heat into a disk DD with boundary CC is equal to CTdν\int_C \nabla T \cdot d \nu where dν=(dy,dx)d \nu = (dy, -dx) is the normal vector (times dtdt) and T\nabla T is the gradient of the temperature function (at a fixed time). Since the gradient points in the direction of greatest increase in temperature, the heat wants to flow in the exact opposite direction.

  1. Use Green’s theorem CPdx+Qdy=DQxPydA\displaystyle\int_C P dx + Q dy = \iint_D \frac{\partial Q}{\partial x} - \frac{\partial P}{\partial y} \, dA to show that the net flow of heat into DD is D2Tx2+2Ty2dxdy\displaystyle\iint_D \frac{\partial^2 T}{\partial x^2} + \frac{\partial^2 T}{\partial y^2} \, dx dy.
Steady State Solutions

When the temperature reaches equilibrium, it stops changing so the left side of the heat equation (T/t\partial T/ \partial t) becomes zero. When that happens, T=T(x,y)T = T(x,y) is a harmonic function. We’ve talked about functions that are harmonic on the entire complex plane. But an important problem is to find functions that are harmonic on a domain DD \subset \mathbb{C} that satisfy a boundary condition. This is called the Dirichlet problem.

We looked at one solution to the Dirichlet problem on the upper-half plane: T(z)=k1+(k0k1π)Argz. T(z) = k_1 + \left(\frac{k_0 - k_1}{\pi} \right) \operatorname{Arg} z.

Level curves for T
  1. Show that the function T(z)=k1+(k0k1π)Argz\displaystyle T(z) = k_1 + \left(\frac{k_0 - k_1}{\pi} \right) \operatorname{Arg} z is harmonic. Hint: Is there a holomorphic function with T(z)T(z) as its real or imaginary part?

We finished by talking about how a harmonic function T:D1T: D_1 \rightarrow \mathbb{R} that is defined on one domain D1D_1 \subset \mathbb{C} can be turned into a harmonic function on another domain D2D_2 \subset \mathbb{C} if you have find an invertible holomorphic function f:D2D1f: D_2 \rightarrow D_1. In that case, T(f(z))T(f(z)) is a harmonic function on D2D_2.

Friday, Apr 26

If TT is a harmonic function on a domain in \mathbb{C}, then it has a harmonic conjugate function VV such that T+iVT+iV is holomorphic. If TT represents the equilibrium temperature on the domain, the what does VV represent?

  1. Show that the level curves of VV are always orthogonal to the level curves of TT. Hint: First show that the gradients of TT and VV are orthogonal.

Since the level curves of VV follow the direction indicated by the gradient of TT, the level curves of VV represent the lines of flow for the heat at equilibrium.

We took this idea and looked at three examples of level curves and flow curves for harmonic functions:

  1. We transformed the solution to the Dirichlet problem on the upper-half plane from Wednesday to a solution on the unit disk.

  2. We transformed a simple linear flow along a narrow channel (D={x+iy:π<y<π}D = \{x+iy : -\pi < y < \pi\}) to a flow escaping from the mouth of a channel using the transformation f(z)=z+ezf(z) = z+e^z.

  3. We transformed another simple linear flow (with D={x+iy:π2<x<π2}D = \{x+iy : -\frac{\pi}{2} < x < \frac{\pi}{2} \}) to a flow passing through a narrow gap in the x-axis using f(z)=sinzf(z) = \sin z. The key to writing the transformation of the flow curves c+itc + it using sinz\sin z was to using the angle addition formulas with cos(it)=cosh(t)\cos(it) = \cosh(t) and sin(it)=isinh(t)\sin(it) = i \sinh(t).