Skip to Content
Math416Complex Variables (Lecture 27)

Math416 Lecture 27

Continue on Application to evaluate cosx1+x4dx\int_{-\infty}^\infty \frac{\cos x}{1+x^4}dx

Consider the functionf(z)=eiz1+z4=cosz+isinz1+z4f(z)=\frac{e^{iz}}{1+z^4}=\frac{\cos z+i\sin z}{1+z^4}.

Our desired integral can be evaluated by RRf(z)dz\int_{-R}^R f(z)dz

To evaluate the singularity, z4=1z^4=-1 has four roots by the De Moivre’s theorem.

z4=1=eiπ+2kπiz^4=-1=e^{i\pi+2k\pi i} for k=0,1,2,3k=0,1,2,3.

So z=eiθz=e^{i\theta} for θ=π4+kπ2\theta=\frac{\pi}{4}+\frac{k\pi}{2} for k=0,1,2,3k=0,1,2,3.

So the singularities are z=eiπ/4,ei3π/4,ei5π/4,ei7π/4z=e^{i\pi/4},e^{i3\pi/4},e^{i5\pi/4},e^{i7\pi/4}.

Only z=eiπ/4,ei3π/4z=e^{i\pi/4},e^{i3\pi/4} are in the upper half plane.

So we can use the semi-circle contour to evaluate the integral. Name the path as γ\gamma.

γf(z)dz=2πi[Resz=eiπ/4(f)+Resz=ei3π/4(f)]\int_\gamma f(z)dz=2\pi i\left[\operatorname{Res}_{z=e^{i\pi/4}}(f)+\operatorname{Res}_{z=e^{i3\pi/4}}(f)\right].

The two poles are simple poles.

Resz0(f)=limzz0(zz0)f(z)\operatorname{Res}_{z_0}(f)=\lim_{z\to z_0}(z-z_0)f(z).

So

Resz=eiπ/4(f)=limzeiπ/4(zeiπ/4)eiz1+z4=(zeiπ/4)eiz(zeiπ/4)(zei3π/4)(zei5π/4)(zei7π/4)=eieiπ/4(eiπ/4ei3π/4)(eiπ/4ei5π/4)(eiπ/4ei7π/4)\begin{aligned} \operatorname{Res}_{z=e^{i\pi/4}}(f)&=\lim_{z\to e^{i\pi/4}}(z-e^{i\pi/4})\frac{e^{iz}}{1+z^4}\\ &=\frac{(z-e^{i\pi/4})e^{iz}}{(z-e^{i\pi/4})(z-e^{i3\pi/4})(z-e^{i5\pi/4})(z-e^{i7\pi/4})}\\ &=\frac{e^{ie^{i\pi/4}}}{(e^{i\pi/4}-e^{i3\pi/4})(e^{i\pi/4}-e^{i5\pi/4})(e^{i\pi/4}-e^{i7\pi/4})} \end{aligned}

A short cut goes as follows:

We know p(z)=1+z4p(z)=1+z^4 has four roots z1,z2,z3,z4z_1,z_2,z_3,z_4.

limzz0(zz0)p(z)=1p(z0)\lim_{z\to z_0}\frac{(z-z_0)}{p(z)}=\frac{1}{p'(z_0)}

So

Resz=eiπ/4(f)=eieiπ/44ei3π/4\operatorname{Res}_{z=e^{i\pi/4}}(f)=\frac{e^{ie^{i\pi/4}}}{4e^{i3\pi/4}}

Similarly,

Resz=ei3π/4(f)=eiei3π/44eiπ/4\operatorname{Res}_{z=e^{i3\pi/4}}(f)=\frac{e^{ie^{i3\pi/4}}}{4e^{i\pi/4}}

So the sum of the residues is

Resz=eiπ/4(f)+Resz=ei3π/4(f)=eieiπ/44ei3π/4+eiei3π/44eiπ/4=ei2e124[12+i12]+ei2e124[12+i12]=π22e12(cos12+sin12)\begin{aligned} \operatorname{Res}_{z=e^{i\pi/4}}(f)+\operatorname{Res}_{z=e^{i3\pi/4}}(f)&=\frac{e^{ie^{i\pi/4}}}{4e^{i3\pi/4}}+\frac{e^{ie^{i3\pi/4}}}{4e^{i\pi/4}}\\ &=\frac{e^{\frac{i}{\sqrt{2}}} e^{-\frac{1}{\sqrt{2}}}}{4[-\frac{1}{\sqrt{2}}+i\frac{1}{\sqrt{2}}]}+\frac{e^{-\frac{i}{\sqrt{2}}}-e^{-\frac{1}{\sqrt{2}}}}{4[\frac{1}{\sqrt{2}}+i\frac{1}{\sqrt{2}}]}\\ &=\frac{\pi\sqrt{2}}{2}e^{-\frac{1}{\sqrt{2}}}(\cos\frac{1}{\sqrt{2}}+\sin\frac{1}{\sqrt{2}}) \end{aligned}

For the semicircle part, we can bound our estimate by

CRf(z)dzπRmaxzCRf(z)π1R40\left|\int_{C_R}f(z)dz\right|\leq\pi R\max_{z\in C_R}|f(z)|\leq \pi \frac{1}{R^4}\to 0

as RR\to\infty.

So

cosx1+x4dx=π22e12(cos12+sin12)\int_{-\infty}^\infty\frac{\cos x}{1+x^4}dx=\frac{\pi\sqrt{2}}{2}e^{-\frac{1}{\sqrt{2}}}(\cos\frac{1}{\sqrt{2}}+\sin\frac{1}{\sqrt{2}})

Big idea of this course

ff is holomorphic     \iff ff has complex derivative.

ff is holomorphic     \iff ff satisfies Cauchy-Riemann equations ux=vy\frac{\partial u}{\partial x}=\frac{\partial v}{\partial y} and uy=vx\frac{\partial u}{\partial y}=-\frac{\partial v}{\partial x}

ff is holomorphic     \iff ff is analytic (is locally given by power series). The power series is integrable/differentiable term by term in the radius of convergence.

Laurent series

Similar to power series both with annulus of convergence.

f(z)=n=an(zz0)nf(z)=\sum_{n=-\infty}^\infty a_n(z-z_0)^n for zA(z0,r,R)z\in A(z_0,r,R).

Identity theorem: If ff is holomorphic on a domain Ω\Omega, it is uniquely determined by its values on any sets with a limit point in Ω\Omega.

Cauchy’s Theorem

γf(z)dz=0\int_\gamma f(z)dz=0

If ff is holomorphic on Ω\Omega and γ\gamma is a closed path in Ω\Omega and γintγΩ\gamma\cup \operatorname{int}\gamma\subset \Omega, then γf(z)dz=0\int_\gamma f(z)dz=0.

Favorite estimate

γf(z)dzsupzγf(z)length(γ)\left|\int_\gamma f(z)dz\right|\leq \sup_{z\in\gamma}|f(z)|\cdot \operatorname{length}(\gamma)

Cauchy’s Integral Formula

f(z0)=12πiγf(z)zz0dzf(z_0)=\frac{1}{2\pi i}\int_\gamma \frac{f(z)}{z-z_0}dz

where z0intγz_0\in \operatorname{int}\gamma and γ\gamma is a closed path.

Extension: If ff is holomorphic on Ω\Omega and z0Ωz_0\in \Omega, then ff is infinitely differentiable and

f(n)(z0)=n!2πiγf(z)(zz0)n+1dzf^{(n)}(z_0)=\frac{n!}{2\pi i}\int_\gamma \frac{f(z)}{(z-z_0)^{n+1}}dz

Residue theorem

If ff is holomorphic on Ω\Omega except for a finite number of isolated singularities z1,z2,,zpz_1,z_2,\dots,z_p, and Γ\Gamma is a curve inside Ω\Omega that don’t pass through any of the singularities (ΓΩ{z1,z2,,zp}\Gamma\subset \Omega\setminus \{z_1,z_2,\dots,z_p\}), then

Γf(z)dz=2πiziindΓ(zi)reszi(f)\int_\Gamma f(z)dz=2\pi i\sum_{z_i}\operatorname{ind}_{\Gamma}(z_i) \operatorname{res}_{z_i}(f)

Harmonic conjugate

Locally, always have harmonic conjugates.

Globally can do this iff domain is simply connected.

Schwarz-pick’s Lemma:

If ff maps DD to DD and f(0)=0f(0)=0, then f(z)z|f(z)|\leq |z| for all zDz\in D. and f(0)1|f'(0)|\leq 1.

For mobius map, f:DDf:D\to D holds, φ(f(z),f(w))=φ(z,w)\varphi(f(z),f(w))=\varphi(z,w) for all z,wDz,w\in D.

φ(z,w)=zw1wz\varphi(z,w)=\frac{z-w}{1-\overline{w}z}

Convergence

Types of convergence

Converge pointwise (Not very strong):

xX,limnfn(x)=f(x)\forall x\in X, \lim_{n\to\infty}f_n(x)=f(x).

Or, xX,ϵ>0,N>0,nN    fn(x)f(x)<ϵ\forall x\in X, \forall \epsilon>0, \exists N>0, \forall n\geq N \implies |f_n(x)-f(x)|<\epsilon.

Converge uniformly (Much better):

ϵ>0,N>0,nN    xX,fn(x)f(x)<ϵ\forall \epsilon>0, \exists N>0, \forall n\geq N \implies \forall x\in X, |f_n(x)-f(x)|<\epsilon.

Converge locally uniformly (Strong):

xX\forall x\in X, \exists open xUx\in U, such that fnff_n\to f uniformly on UU.

Converge uniformly on compact subsets (Good enough for local properties):

\forall compact KXK\subset X, fnff_n\to f uniformly on KK.

Weierstrass’ Theorem

If fnO(Ω)f_n\in O(\Omega) and fnff_n\to f locally uniformly, then fO(Ω)f\in O(\Omega).

Cauchy-Hadamard’s Theorem

For a power series, n=0an(zz0)n\sum_{n=0}^\infty a_n(z-z_0)^n, the radius of convergence is

R=1lim supnan1/nR=\frac{1}{\limsup_{n\to\infty}|a_n|^{1/n}}

On B(z0,R)B(z_0,R), the series converges locally uniformly and absolutely.

Argument and Logarithm

argz\arg z is any θ\theta such that z=reiθz=re^{i\theta}.

Argz\operatorname{Arg} z is the principal value of the argument, π<Argzπ-\pi<\operatorname{Arg} z\leq \pi.

logz\log z is the principal value of the logarithm, logz=lnz+iargz\log z=\ln |z|+i\arg z.

Logz\operatorname{Log} z is the set of all logarithms of zz, Logz={logz+2kπi:kZ}\operatorname{Log} z=\{\log z+2k\pi i: k\in\mathbb{Z}\}.

END

Last updated on