logo

Singularities on the Real Axis and Improper Integrals via Jordan's Lemma 📂Complex Anaylsis

Singularities on the Real Axis and Improper Integrals via Jordan's Lemma

Buildup

The overall flow is similar to Jordan’s Lemma for Improper Integrals. For two polynomials p(z),q(z)p(z) , q(z), let’s say f(z)=q(z)p(z)\displaystyle f(z) = {{q(z)} \over {p(z)}}.

If there exists a real solution aa that satisfies p(z)=0p(z) = 0, then ff has a real singularity at aa. The reason we haven’t dealt with such cases until now was to use the Residue Theorem. Of course, just because a singularity is added on the real axis, we don’t strictly have to give up on the residue theorem; we use a trick of bending the integration path.

20171227\_010018.png

Considering a simple closed path C\mathscr{C} as shown above, we can bypass the singularity with a small semicircle like γ\gamma. The semicircle does not necessarily have to be centered on 00, and it’s fine if it exists finitely and complexly on the real axis. By choosing r0r \to 0 to tighten the small semicircle and taking RR \to \infty, we can use the residue theorem as we originally did. To use this method, we need to prepare by proving the following auxiliary lemma first. Though it sounds complicated, one can easily see it resembles the Residue Theorem.

Auxiliary Lemma 1

Let’s suppose the function f:CCf : \mathbb{C} \to \mathbb{C} is analytic near the vicinity 0<za<r00 < |z - a | < r_{0} of the singularity aa. If a semicircle z(θ)a=reiθ,0θπz( \theta) - a = r e^{i \theta}, 0 \le \theta \le \pi with radius rr, which is a small positive number smaller than r0r_{0}, centered at aa limr0γf(z)dz=iπResaf(z) \lim_{r \to 0} \int_{\gamma} f(z) dz = - i \pi \text{Res}_{a} f(z)

Proof

According to the assumption, f(z)=m=0b2m+1(za)2m+1+n=0an(za)n f(z) = \sum_{m=0}^{\infty} {{b_{2m+1}} \over {(z-a)^{2m+1} }} + \sum_{n=0}^{\infty} a_{n} (z-a)^{n} (za)=reiθ(z-a) = r e^{i \theta}, so when n1n \ne -1, γ(za)ndz=irn+10πe(n+1)iθdθ=rn+1n+1(1+cosnπ) \int_{\gamma} (z-a)^{n} dz = -i r^{n+1} \int_{0}^{\pi} e^{(n+1) i \theta} d \theta = {{r^{n+1}} \over {n+1}} (1 + \cos {n \pi}) Therefore, either n0n \ge 0 or n<1n<-1, but if nn is odd, limr0γ(za)ndz=0 \lim_{r \to 0} \int_{\gamma} (z - a)^{n} dz = 0 On the other hand, if considering the case of n=1n = -1, limr0γ(za)1dz=limr00πidθ=iπ \lim_{r \to 0} \int_{\gamma} (z-a)^{-1} dz = \lim_{r \to 0} \int_{0}^{\pi} - i d \theta = - i \pi Therefore, limr0γf(z)dz=iπb1=iπResαf(z)\displaystyle \lim_{r \to 0} \int_{\gamma} f(z) dz = - i \pi b_{1} = - i \pi \text{Res}_{\alpha} f(z)

Example

As an example, let’s calculate the improper integral of the sinc function 0sinxxdx\displaystyle \int_{0}^{\infty} {{\sin x} \over {x} } dx.

Solution

Since eiz=cosz+isinze^{i z} = \cos z + i \sin z, first consider Ceizzdz\displaystyle \int_{\mathscr{C}} {{e^{iz} } \over {z}} dz, and let’s assume C\mathscr{C} as given in the figure above. Then Ceizzdz=Γeizzdz+Rreixxdx+γeizzdz+rReixxdx \int_{\mathscr{C}} {{e^{iz} } \over {z}} dz = \color{red} { \int_{ \Gamma} {{e^{iz} } \over {z}} dz } + \int_{-R}^{-r} {{e^{ix} } \over {x}} dx + \color{blue} { \int_{ \gamma} {{e^{iz} } \over {z}} dz } + \int_{r}^{R} {{e^{ix} } \over {x}} dx

Residue Theorem: Suppose an analytic function f:ACCf: A \subset \mathbb{C} \to \mathbb{C} has a finite number of singularities inside a Simple closed contour C\mathscr{C}. Then Cf(z)dz=2πik=1mReszkf(z)\displaystyle \int_{\mathscr{C}} f(z) dz = 2 \pi i \sum_{k=1}^{m} \text{Res}_{z_{k}} f(z)

Since the singularity of the function f(z):=eizz\displaystyle f(z) := {{e^{iz} } \over {z}} is only z=0z = 0 and there are no singularities inside C\mathscr{C}, by the Residue Theorem Ceizzdz=0 \int_{\mathscr{C}} {{e^{iz} } \over {z}} dz = 0 That is, Γeizzdz+Rreixxdx+γeizzdz+rReixxdx=0 \color{red} { \int_{ \Gamma} {{e^{iz} } \over {z}} dz } + \int_{-R}^{-r} {{e^{ix} } \over {x}} dx + \color{blue} { \int_{ \gamma} {{e^{iz} } \over {z}} dz } + \int_{r}^{R} {{e^{ix} } \over {x}} dx = 0

Residue at Simple Poles: Suppose function ff can be represented as f(z)=g(z)h(z)\displaystyle f(z) = {{g(z)} \over {h(z)}}. Here gg and hh are analytic at α\alpha, and if g(α)0,h(α)=0,h(α)0g(\alpha) \ne 0 , h(\alpha) = 0, h ' (\alpha) \ne 0 then α\alpha is a simple pole of ff and Resαf(z)=g(α)h(α)` \text{Res}_{\alpha} f(z) = {{g(\alpha)} \over {h ' (\alpha)}}

Therefore, since Res0eizz=ei01=1\displaystyle \text{Res}_{0} {{e^{iz} } \over {z}} = {{e^{i \cdot 0} } \over {1}} = 1, limr0γeizzdz=iπRes0eizz=iπ \lim_{r \to 0} \color{blue} { \int_{\gamma} {{e^{iz} } \over {z}} dz } = - i \pi \text{Res}_{0} {{e^{iz} } \over {z}} = - i \pi When r0r \to 0, Γeizzdz+R0eixxdxiπ+0Reixxdx=0 \color{red} { \int_{ \Gamma} {{e^{iz} } \over {z}} dz } + \int_{-R}^{0} {{e^{ix} } \over {x}} dx \color{blue} { - i \pi } + \int_{0}^{R} {{e^{ix} } \over {x}} dx = 0 Summarizing, Γeizzdz+RReixxdx=iπ \color{red} { \int_{ \Gamma} {{e^{iz} } \over {z}} dz } + \int_{-R}^{R} {{e^{ix} } \over {x}} dx = i \pi

Jordan’s Lemma: When a semicircle Γ\Gamma is represented as z(θ)=Reiθ,0θπz(\theta) = R e^{i \theta} , 0 \le \theta \le \pi, if a function ff is continuous on Γ\Gamma and limzf(z)=0\displaystyle \lim_{z \to \infty} f(z) = 0, then for a positive mR+m \in \mathbb{R}^{+}, limRΓemizf(z)dz=0\lim_{R \to \infty} \int_{\Gamma} e^{m i z } f(z) dz = 0

By Jordan’s Lemma, since limRΓeizzdz=0\displaystyle \lim_{R \to \infty} \color{red} { \int_{ \Gamma} {{e^{iz} } \over {z}} dz } = 0, eixxdx=iπ \int_{- \infty }^{ \infty } {{e^{ix} } \over {x}} dx = i \pi Once again, since eiz=cosz+isinze^{i z} = \cos z + i \sin z, cosxxdx+isinxxdx=iπ \int_{- \infty }^{ \infty } {{ \cos x } \over {x}} dx + i \int_{- \infty }^{ \infty } {{ \sin x } \over {x}} dx = i \pi Taking only the imaginary part gives sinxxdx=π \int_{- \infty}^{\infty} {{\sin x} \over {x} } dx = \pi And since the sinc function is an even function, 0sinxxdx=π2 \int_{0}^{\infty} {{\sin x} \over {x} } dx = {{\pi} \over {2} }

Corollary

It’s also easy to see that the following equation holds for aRa\in \mathbb{R} through variable substitution. 0sinxxdx=0sin(ax)xdx=π2 \int_{0}^{\infty} \frac{\sin x}{x}dx=\int_{0}^{\infty} \frac{\sin (ax)}{x}dx=\frac{\pi}{2} Although it may seem long and complicated at first glance, if you examine the solution closely, there really isn’t much to the calculation. Using complex analysis seems simpler and more straightforward.


  1. Osborne (1999). Complex variables and their applications: p170. ↩︎