Skip to main content

Math 382-0: Kursobjekt

Section 1.12 Complex paths

Subsection Complex paths

Definition 1.12.1. Complex path.

A complex path is a continuous function \(\phi \colon[a,b]\rightarrow \C\) where \([a,b]\subseteq \R\) is a real interval.
We say that a path \(\phi\colon[a,b]\rightarrow \C\) parametrizes its image \(\mathcal{C}=\{\phi(t)\colon t\in [a,b]\}\subseteq \C\text{.}\)

Remark 1.12.2. Paths.

Note that since \([a,b]\subseteq \R\subseteq \C\) can be thought of as a subset of \(\C\text{,}\) a path \(\phi\colon [a,b]\rightarrow \C\) can be treated as a complex function with a particularly restricted domain \(D=[a,b]\text{.}\) As such, the same conventions around complex functions apply to paths. In particular, we may write
\begin{equation*} \phi(t)=u(t)+iv(t) \end{equation*}
in terms of its real and imaginary parts. Furthermore, identifying \(\C\) with \(\R^2\text{,}\) we can also think of a complex path \(\phi=u+iv\) as a continuous function
\begin{align*} \phi\colon \R \amp \rightarrow \R^2 \\ t \amp \longmapsto (u(t),v(t))\text{.} \end{align*}

Example 1.12.3. Parametrized curves.

For the given curve \(\mathcal{C}\) find a path \(\phi\colon [a,b]\rightarrow \C\) that parametrizes \(\mathcal{C}\text{.}\)
  1. Fix \(z_0=a_0+ib_0\) and \(r> 0\text{.}\) Let \(\mathcal{C}=\{z\in \C\colon \abs{z-z_0}=r\}\text{.}\)
  2. Fix \(z_0=a_0+ib_0\) and \(z_1=a_1+ib_1\text{.}\) Let \(\mathcal{C}\) be the directed line segment that begins at \(z_0\) and ends at \(z_1\text{.}\)
Solution.
  1. Since \(e^{it}\text{,}\)\(t\in [0,2\pi]\text{,}\) parametrizes the unit circle centered at the origin, we see easily that
    \begin{equation*} \phi(t)=z_0+Re^{it}, t\in [0,2\pi] \end{equation*}
    is a parametrization of \(\mathcal{C}\text{.}\)
  2. For any points \(P,Q\in \R^2\text{,}\) the line segment between them is parametrized by \(P+t(P-Q)=(1-t)P+tQ\text{,}\) \(t\in [0,1]\text{.}\) The equivalent complex parametrization from \(z_0\) to \(z_1\) is thus
    \begin{equation*} \phi(t)=(1-t)z_0+tz_1, t\in [0,1]\text{.} \end{equation*}

Definition 1.12.4. Path jargon.

Let \(\phi\colon [a,b]\rightarrow \C\) be a path.
  • The points \(z_0=\phi(a)\) and \(z_1=\phi(b)\) are called the initial and terminal points of \(\phi\text{,}\) respectively.
  • \(\phi\) is closed if \(\phi(a)=\phi(b)\text{.}\)
  • \(\phi\) is simple if \(\phi(t)\ne \phi(s)\) for all \(t,s\in (a,b)\text{.}\)
  • The path \(-\phi\colon [a,b]\rightarrow \C\) defined as \(-\phi(t)=\phi(ta+(1-t)b)\) is called the reverse of \(\phi\text{.}\)
  • Given a path \(\psi\colon [b,c]\rightarrow \C\) satisfying \(\phi(b)=\psi(b)\text{,}\) the concatentation of \(\phi\) and \(\psi\text{,}\) denoted \(\phi*\psi\text{,}\) is the path \(\phi*\psi\colon [a,c]\rightarrow \C\) defined as \(\phi*\psi\vert_{[a,b]}=\phi\) and \(\phi*\psi\vert_{[b,c]}=\psi\text{.}\)
We now define a notion of real differentiability of paths. Note that though we can indeed think of a path \(\phi\colon [a,b]\rightarrow \C\) as a complex function, since no element \(t_0\) in its domain \([a,b]\) is an interior point of \([a,b]\text{,}\) considered as a subset of \(\C\text{,}\) the notion of complex differentiability does not apply to \(\phi\text{.}\)

Definition 1.12.5. Path derivative.

Given a path \(\phi\colon [a,b]\rightarrow \C\) with \(\phi=u+iv\) and element \(t_0\in [a,b]\text{,}\) we say \(\phi\) is (real) differentiable at \(t_0\) if \(u\) and \(v\) are differentiable at \(t_0\text{,}\) and define
\begin{equation*} \phi'(t_0)=u'(t_0)+iv'(t_0)\text{.} \end{equation*}
Geometrically we interpret \(\phi'(t_0)\) as a tangent vector to \(\mathcal{C}=\{\phi(t)\colon t\in [a,b]\}\) at the point \(\phi(t_0)\text{.}\) The path \(\phi\) is smooth on \([a,b]\) if \(\phi'\) exists and is continuous at all \(t\in (a,b)\text{;}\) it is piecewise smooth if we can write \(\phi=\phi_1*\phi_2\cdots *\phi_k\) as a concatenation of finitely many smooth curves.
Visualizing a path and its derivative in case of a circle
Figure 1.12.6. Visualizing \(\phi(t)=Re^{it}\) and \(\phi'(t)\)

Example 1.12.7. Tangent vectors.

Compute \(\phi'\) for your parametrizations in Example 1.12.3 and provide a sketch illustrating the relationship of \(\phi'\) to \(\phi\text{.}\)
Solution.
  1. For \(\phi(t)=z_0+Re^{it}\text{,}\) we have \(\phi'(t)=Rie^{it}\text{.}\) Since multiplication by \(i\) rotates a complex number by \(\pi/2\text{,}\) we see that the tangent vector \(\phi'(t)\) is always orthogonal to \(\phi(t)\text{.}\) See (Figure 1.12.6.) Note also that the modulus of \(\phi'(t)\) is
    \begin{equation*} \abs{\phi'(t)}=R\abs{i}\abs{e^{it}}=R\text{.} \end{equation*}
  2. For \(\phi(t)=(1-t)z_0+z_1\text{,}\) we have \(\phi'(t)=-z_0+z_1=z_1-z_0\text{.}\) Thus the derivative is a constant function, equal to the direction vector from \(z_0\) to \(z_1\text{.}\)
Of course, identifying \(\phi=u+iv\) with the planar path (or curve) \((u(t),v(t))\) , we recognize its derivative \(\phi'=u'+iv'\) as the derivative \((u'(t), v'(t))\) of this planar path (or curve). As a result, all the usual properties of derivatives of planar paths carry over to our complex setting. We will not list them here, but will make a connection with complex differentiation.

Proof.

Subsection Complex-valued function integration

Turning now to a notion of integration, we will introduce the synonymous term complex-valued function for complex paths. We will also favor lowercase Roman letters for naming these functions: e.g., \(f\colon [a,b]\rightarrow \C\text{,}\) \(f(t)=f_1(t)+if_2(t)\text{.}\) This shift in terminology and notation is meant to both draw a connection with Riemann integrals of single-variable real-valued functions, and also to make a distinction between the more general notion of line integrals of complex functions over paths we will define in Section 1.13.

Definition 1.12.9. Integration of complex-valued functions.

Let \(f\colon [a,b]\rightarrow \C\) be a complex-valued function, and write \(f=f_1+if_2\text{.}\) We say \(f\) is Riemann integrable over \([a,b]\) if the Riemann integrals
\begin{align*} \int_a^b f_1(t)\, dt \amp \amp \int_a^b f_2(t)\, dt \end{align*}
exist, in which case we write
\begin{equation*} \int_a^b f(t)\, dt=\int_a^bf_1(t)\, dt+i\int_a^b f_2(t)\, dt\text{.} \end{equation*}
Since the integral of a complex-valued function \(f\colon [a,b]\rightarrow \C\) is defined in terms of its real and imaginary parts, each of which is an integral of a real-valued function, it should come as no surprise that complex valued integrals satisfy all of the usual integration properties. For example, this integration operation is linear in the sense that
\begin{equation*} \int_a^b cf(t)+dg(t)\, dt=c\int_a^bf(t)\, dt+d\int_a^bg(t)\, dt \end{equation*}
for any complex-valued functions \(f,g\colon [a,b]\rightarrow \C\) and complex constants \(c,d\in \C\text{.}\) Furthermore, complex-valued functions satisfy slightly generalized version of the fundamental theorem of calculus.

Example 1.12.11. Complex-valued function integration.

Let \(f(t)=e^{(1+i)t}\text{.}\) Compute \(\int_0^\pi f(t)\, dt\) using two different techniques: directly, using Definition 1.12.9, and indirectly using Theorem 1.12.10.
Solution.
First we compute the integral directly:
\begin{align*} \int_0^\pif(t)\, dt \amp =\int_0^\pi e^t\cos t+ie^t\sin t\, dt\\ \amp = \frac{e^t\cos t-e^t\sin t}{2}\Bigr\vert_0^\pi+ i\, \frac{e^t\sin t-e^t\cos t}{2}\Bigr\vert_0^\pi \\ \amp =-(e^\pi+1)/2+i(e^\pi+1)/2 \text{.} \end{align*}
Observe that in the middle of that computation we needed to compute antiderivatives of \(e^t\cos t\) and \(e^t\sin t\text{,}\) which is normally done using integration by parts (twice).
Alternatively, using the fundamental theorem of calculus, we observe that \(F(t)=\frac{1}{1+i}e^{(1+i)t}\) is an antiderivative of \(f(t)\text{,}\) and hence
\begin{align*} \int_0^\pif(t)\, dt \amp = F(\pi)-F(0) \\ \amp =\frac{1}{1+i}(e^{\pi+i\pi}-e^{0})\\ \amp = \frac{1}{1+i}(-e^\pi-1)\\ \amp =-\frac{1-i}{2}(e^\pi+1) \\ \amp = -(e^\pi+1)/2+i(e^\pi+1)/2 \text{.} \end{align*}
Lastly, we record here an important inequality bounding the modulus of a complex-valued function, which is a clear generalization of a corresponding inequality about integrals of real-valued functions. Contrary to what you might expect, the proof of this fact does not proceed so directly from real-valued function properties. We will provide two proofs: one which makes use of the geometry of \(\C\text{,}\) and another that identifies \(\C\) with \(\R^2\) and uses the Cauchy-Schwarz inequality. The latter has the advantage of generalizing directly to more general vector-valued functions \(f\colon[a,b]\rightarrow \R^n\text{,}\) where \(n\) is any positive integer.

First proof: geometric.

Let \(w=\int_a^bf(t)\, dt\text{,}\) and write \(w=re^{i\alpha}\) in polar form. It follows that \(\abs{w}=r=e^{-i\alpha}w\text{,}\) and thus we have
\begin{align*} \abs{\int_a^bf(t)\, dt} \amp = e^{-i\alpha}w\\ \amp = e^{-i\alpha}\int_a^b f(t)\, dt\\ \amp = \int_a^b e^{-i\alpha}f(t)\, dt\\ \amp = \Re \int_a^b e^{-i\alpha}f(t)\, dt \amp (r=\int_a^b e^{-i\alpha}f(t)\, dt\in \R)\\ \amp = \int_a^b \Re e^{-i\alpha}f(t)\, dt \amp \text{(by def.)}\\ \amp = \abs{\int_a^b \Re e^{-i\alpha}f(t)\, dt} \amp (r\geq 0)\\ \amp \leq \int_a^b \abs{\Re e^{-i\alpha}f(t)}\, dt \amp \text{(real int. prop.)}\\ \amp \leq \int_a^b \abs{e^{-i\alpha}f(t)}\, dt \amp (\abs{\Re z}\leq \abs{z})\\ \amp = \int_a^b \abs{f(t)}\, dt \amp (\abs{e^{-i\alpha}}=1) \end{align*}

Second proof: using Cauchy-Schwarz.

We will identify \(\C\) with \(\R^2\text{.}\) Observe that under this identification, the modulus of a complex number \(z=a+ib\) is the norm of the vector \((a,b)\text{:}\) that is,
\begin{equation*} \abs{a+ib}=\norm{(a,b)}=\sqrt{a^2+b^2}\text{.} \end{equation*}
Under this identification, a complex-valued function \(f=f_1+if_2\) is identified as the vector-valued function
\begin{align*} f\colon [a,b] \amp\rightarrow \R^2\\ t \amp \longmapsto (f_1(t),f_2(t))\text{,} \end{align*}
and the integral \(\int_a^bf(t)\, dt\) is then identified as the vector-valued integral
\begin{equation*} \int_a^bf(t)\, dt=\left(\int_a^bf_1(t)\, dt, \int_a^bf_2(t)\, dt\right)\text{.} \end{equation*}
Lastly, for vectors \(\boldv=(a,b)\) and \(\boldw=(c,d)\text{,}\) we will denote their dot product as \(\angvec{\boldv, \boldw}\text{:}\) i.e.,
\begin{equation*} \angvec{\boldv, \boldw}=ac+bd\text{.} \end{equation*}
Since integration of vector-valued functions is linear, and since the inner product is linear in each variable, it follows easily that
\begin{equation} \angvec{\int_a^bf(t)\, dt, \boldv}=\int_a^b \angvec{f(t), \boldv}\, dt\tag{1.48} \end{equation}
for any integrable vector-valued function and vector \(\boldv\in \R^2\text{.}\)
In this setting we wish to show that
\begin{equation*} \norm{\int_a^bf(t)\, dt}\leq \int \norm{f(t)}\, dt\text{.} \end{equation*}
The claim is obvious if \(\int_a^b f(t)\, dt=\boldzero\text{.}\) Otherwise, since \(\norm{\boldv}^2=\angvec{\boldv,\boldv}\) for any \(\boldv\in \R^2\text{,}\) we have
\begin{align*} \norm{\int_a^b f(t)\, dt}^2 \amp = \angvec{\int_a^b f(t)\, dt, \int_a^b f(t)\, dt} \\ \amp = \int_a^b \angvec{f(t), \int_a^b f(t)\, dt}\, dt \amp \knowl{./knowl/xref/eq_dot_int.html}{\text{(1.48)}} \\ \amp = \abs{\int_a^b \angvec{f(t), \int_a^b f(t)\, dt}\, dt} \amp (\text{quantity is nonnegative})\\ \amp\leq \int_a^b\abs{\angvec{f(t), \int_a^b f(t)\, dt}} \, dt \amp (\text{real int. prop.})\\ \amp \leq \leq \int_a^b\norm{f(t)}\norm{\int_a^b f(t)\, dt} \, dt \amp (\text{Cauchy-Schwarz ineq.}) \\ \amp = \norm{\int_a^b f(t)\, dt}\int_a^b\norm{f(t)}\, dt \text{.} \end{align*}
In summary, we have
\begin{equation*} \norm{\int_a^b f(t)\, dt}^2\leq \norm{\int_a^b f(t)\, dt}\int_a^b\norm{f(t)}\, dt\text{.} \end{equation*}
Since we assume that \(\int_a^b f(t)\, dt\ne \boldzero\text{,}\) we have \(\norm{\int_a^b f(t)\, dt}\ne 0\text{,}\) allowing us to cancel and obtain
\begin{equation*} \norm{\int_a^b f(t)\, dt}\leq \int_a^b\norm{f(t)}\, dt\text{,} \end{equation*}
as desired.