PDE & Complex Variables P13-1 Lesson 13 Laurent Series and Residue (EK 16) Laurent Series (EK 16.1) Motivation Singularities are routine for complex functions, for bounded entire functions must be constant (Liouville’s theorem, Lesson 10). We have to represent complex functions by Laurent series if there is singularity in the region of interest. Laurent theorem If f(z) is analytic in an open annulus D: R2<|z−z0|<R1,⇒ ∞ ∞ bn n n =1 ( z − z0 ) f(z)= ∑ a n ( z − z 0 ) n + ∑ n =0 an= f ( z ′) 1 dz ′ , ∫ C 2πi ( z ′ − z 0 ) n +1 bn= 1 ( z ′ − z0 ) n−1 f ( z ′)dz′ ∫ C 2πi (13.1) for every point z∈D, where C⊂D is a simple closed path. Edited by: Shang-Da Yang PDE & Complex Variables P13-2 Proof: Let C1, C2 ⊂D are two concentric circles of radii r1, r2. By Cauchy’s integral formula for doubly connected domain [eq. (10.10)]: f(z)= f (z* ) * 1 1 f ( z* ) * dz − dz ≡ 2πi ∫C2 z * − z 2πi ∫C1 z * − z g(z)+h(z) = (analytic part)+(principal part), for any z∈D. (1) Since z lies within C1, and f(z) is analytic on C1, by the proof of Taylor series in Lesson 12: g(z)= 1 f (z* ) f ( z* ) * ∞ 1 n dz = a ( z − z ) , where a = dz * . n ∑ n 0 ∫ C1 ( z * − z ) n +1 2πi ∫C1 z * − z 2 i π n =0 0 <Comment> The proof of Taylor series consists of two steps: (i) f(z)= is analytic everywhere inside C. (ii) ∫ C f (z* ) * dz = z* − z ∞ f (z* ) * 1 dz , which requires f(z) 2πi ∫C z * − z ∑ ∫ n =0 C f ( z ′) dz ′ ( z − z 0 ) n , which n +1 ( z′ − z0 ) only requires f(z) is analytic on C. Here we simply use (ii). 1 (2) Since z lies outside C2, * = z −z z * − z0 1 −p , q= , |q|<1. = , p≡ 1− q z − z0 z * − z0 z − z 0 ( z − z0 )1 − z − z0 −1 z * − z0 q n +1 z * − z0 1 n , ⇒ p= 1 + + ... + =(1+q+…+q )+ 1− q z − z0 z − z0 1− q ⇒ 2πi⋅h(z)= ∫ C2 p f ( z * ) z − z0 * 1 dz = z − z0 1 +…+ ( z * − z0 ) n f ( z * )dz * + n +1 ∫C 2 ( z − z0 ) ⇒ h(z)= n +1 bm ∑ (z − z ) m=1 0 m ∫ C2 f ( z * )dz * + 1 ( z − z ) n +1 0 + Rn* ( z ) , where bm= n ( z * − z0 ) n+1 . + * n ( z − z )( z − z0 ) 1 ( z − z0 ) 2 ∫ C2 ( z * − z0 ) f ( z * )dz * ( z * − z 0 ) n +1 f ( z * ) * dz . ∫C2 (z − z* ) 1 ( z * − z0 ) m−1 f ( z * )dz * , and Rn* ( z ) = 2πi ∫C2 1 ( z − z0 ) n+1 f ( z * ) * dz . (We still need to prove bm exists, and Rn* ( z ) →0.) 2πi ( z − z0 ) n+1 ∫C2 ( z − z* ) * Edited by: Shang-Da Yang PDE & Complex Variables P13-3 By the ML-inequality and the same procedure in proving Taylor theorem eq’s (12.1-2), we ∞ 1 bn , where bn= ( z * − z 0 ) n −1 f ( z * )dz * . n ∫ C 2 2πi n =1 ( z − z0 ) derive lim Rn* ( z ) =0, ⇒ h(z)= ∑ n →∞ (3) By Cauchy’s integral theorem 4 [eq. (10.8)], an= bn= 1 1 [" ]dz * = [" ]dz ′ , ∫ 2πi C1 2πi ∫C 1 1 [" ]dz * = [" ]dz ′ , for any contour C⊂ D. ∫ 2πi C2 2πi ∫C Eq. (13.1) remains valid if we continuously move C1 outward and C2 inward until they reach some singularity. ⇒ ROC of Laurent series is the open annulus D: R2<|z-z0|<R1. <Comment> 1) A function may have different power series because of different ROCs or different centers. E.g. 1 = (i) 1− z ∞ ∑z n , if |z|<1; (ii) n =0 ∞ −1 − 1 ∞ −n = ⋅ z = − z −( n +1) , if |z|>1; (iii) ∑ ∑ −1 z (1 − z ) z n =0 n =0 −1 , if 0<|z−1|<∞. z −1 ∞ 2) A power series only has one closed function form. E.g. f(z)= − ∑ z −( n +1) . The series will n =0 diverge if we substitute z= ∞ 1 1 into it to evaluate f . ⇒ Deriving the closed form of 2 2 f(z) first. f(z)= −(z-1) ∑ ( z −1 ) n = n =0 − z −1 1 1 = , ⇒ f =2. −1 1− z 1− z 2 Edited by: Shang-Da Yang PDE & Complex Variables P13-4 Singularities and Zeros (EK 16.2) Definitions 1) A function f(z) has a singularity at z=z0 if f(z) is not analytic at z=z0. 2) z=z0 is an isolated singularity if it has a neighborhood without any other singularity. E.g. tan z has isolated singularities at z= ± ( ) E.g. tan z −1 ± 2n + 1 π. 2 2 1 2n + 1 2 π , ⇒ z= ± =± =± , z 2 (2n + 1)π π is not continuous/analytic for 2 , …. For an arbitrarily small disk |z|<ε, you can always find infinitely many 3π singularities within it, ⇒ z=0 is a non-isolated singularity (other sigularities are isolated). 3) If the Laurent series of f(z) centered at z=z0 has nonzero coefficient(s) up to m-th order (bm≠0), ⇒ z0 is a pole of order m of f(z), which can be removed by multiplying (z−z0)m. It is called a simple pole if m=1. E.g. 1 1 z 7z 3 1 = + + +... has a simple , by long division, = sin z z − ( z 3 / 3!) + ( z 5 / 5!) − ... z 6 360 pole at z=0. E.g. To find the order of pole of tan z at z= π 2 , we change the variable: u=z− π 2 : f(z)= π π 1 u u3 − − ... , which has a simple pole at u=0, i.e. z= . = −tan u= tan u + − − 2 2 u 3 45 4) If the Laurent series of f(z) centered at z=z0 has infinitely many nonzero coefficients, z0 is an essential singularity of f(z). ∞ 1 has an isolated essential singularity at z=0. n n =0 n! z E.g. e1/z = ∑ ( ) E.g. tan z −1 = sin(1 / z ) 1 1 = + 3 +... has a non-isolated essential singularity at z=0. cos(1 / z ) z 3z 5) If an analytic function f(z) has a property of f(z0)= f'(z0)=…= f (n-1) (z0)=0 (i.e. the first n Taylor coefficients a0=a1=…=an-1=0), ⇒ z=z0 is a zero of order n of f(z). E.g. sin z has simple zeros (n=1) at z=0, ±π, ±2π, …; sin2 z has second-order zeros at these points. Edited by: Shang-Da Yang PDE & Complex Variables P13-5 Properties 1) If f(z) is analytic except for a pole at z=z0, ⇒ |f(z)|→∞ as z→z0 in any manner. 2) Picard’s theorem: If f(z) is analytic except for an essential singularity at z=z0, ⇒ f(z) can be equal to any complex number in an arbitrarily small neighborhood of z0. E.g. f(z)=e1/z: (i) if z=x=(+∞→0+), f(z)→∞; (ii) if z= −x=(−∞→0-), f(z)→0; (iii) For any cos θ − i sin θ given complex number c=c0⋅eiα, we can find some z=r⋅eiθ, s.t. f(z)= exp =c, r ⇒ cosr θ sin θ = c0 , − = α + 2 nπ ; e r ⇒ r = [(ln c0 ) 2 + (α + 2nπ ) 2 ] − 1 2 , tanθ = − (α + 2nπ ) . The solution z can be made arbitrarily close to 0 (r→0) by increasing n. ln c0 3) If f(z) is analytic and has an nth-order zero at z0, ⇒ 1 has an nth-order pole at z=z0. f ( z) 4) The properties of f(z) at large |z| can be investigated by: set z=1/w, investigate g(w)=f(1/w)=f(z) in the neighborhood of w=0. E.g. ez has an isolated essential singularity at ∞, for e1/w has that at w=0. Residue Integration (EK 16.3) Evaluate contour integral by residue To evaluate I= ∫ f ( z )dz for some arbitrary contour C: C 1) If f(z) is analytic for every point on and within C, ⇒ I=0 by Cauchy’s integral theorem 1. 2) If f(z) has only one singularity at z=z0 inside C, its “closest” Laurent series with ROC: ∞ ∞ bn ,⇒ n n =1 ( z − z0 ) {0<|z−z0|<R} is: f(z)= ∑ a n ( z − z 0 ) n + ∑ n =0 I=2πi⋅ Res f ( z ) z = z0 (13.2) Edited by: Shang-Da Yang PDE & Complex Variables P13-6 where Res f ( z ) =b1 is called the residue of f(z) at z0. [eq. (10.9) is a special case.] z = z0 Proof: by 2πi, if m = −1 m − = , z z dz ( ) 0 ∫C 0, otherwise E.g. I= ∫ f ( z )dz , where f(z)= C ∫ C f ( z )dz = b1 ∫ C 1 dz =2πib1. z − z0 1 1 , C: |z|= , in counterclockwise sense. z (1 − z ) 2 Since only one singularity z=0 lies inside C, ⇒ find the Laurent series centered at z=0: (1) 1 1 1 f(z)= +1+z2+…., for 0<|z|<1, ⇒ b1=1. (2) f(z)= − 2 + 3 + ... , for |z|>1, ⇒ b1=0. ⇒ z z z Choose series-1: b1=1, and I=2πi. <Comment> We always choose the closest Laurent series to evaluate residue, even when part of the integral path C falls outside its ROC. E.g. Let f(z)= 1 1 , C: z= + i + re iθ , z (1 − z ) 4 θ∈[0,2π], which passes through ROCs of the two Laurent series and only encloses one singularity z=0 if 17 4 <r< 5 4 . ⇒ I=2πi for arbitrary 17 4 <r< 5 4 . 3) If f(z) has finitely many isolated singularities {zi; i=1−N} inside C, ⇒ N I= 2πi ⋅ ∑ Res[ f ( z )] k =1 z = zk (13.3) Edited by: Shang-Da Yang PDE & Complex Variables P13-7 N Proof: By Cauchy’s integral theorem 4 [eq. (10.6)], ⇒ ∫ f ( z )dz = ∑ ∫ f ( z )dz , where C k =1 Ck Ck is a circle that only encloses one singularity zk, and is separated from all the other circles. By eq. (13.2), ∫ Ck f ( z )dz =2π i⋅ Res[ f ( z )] , ⇒ …. z = zk How to evaluate the residue of a singularity 1) If z=z0 is (i) a pole of unknown order, (ii) essential singularity, (iii) non-isolated singularity of f(z), ⇒ try to derive the (partial) Laurent series. 2) If z=z0 is a simple pole of f(z), ⇒ Res f ( z ) = lim ( z − z 0 ) f ( z ) z = z0 Proof: f(z)= (13.4) z → z0 b1 + a0 + a1 ( z − z 0 ) + a 2 ( z − z 0 ) 2 + ... , (z−z0)f(z)=b1+a0(z−z0)+a1(z−z0)2 z − z0 +…, = b1+ (z−z0)[ a0+a1(z−z0)+…]. 4) If z=z0 is a simple pole of p( z ) , where p(z0)≠0, q(z) has a simple zero at z0, ⇒ q( z ) Res z = z0 p( z ) p( z 0 ) = q( z ) q ′( z 0 ) (13.5) q ′′( z 0 ) (z−z0)2 +…, by eq. (13.4): 2! ( z − z 0 ) p( z ) p( z 0 ) p( z ) p( z ) Res = lim ( z − z 0 ) = lim = . z → z0 z = z 0 ( z − z )[q ′( z ) + ( z − z ) q ′′( z ) / 2!+...] z = z0 q ( z ) q( z ) q ′( z 0 ) 0 0 0 0 Proof: If z=z0 is a simple zero of q(z), q(z)= q ′( z 0 ) (z−z0)+ 5) If z=z0 is mth order pole of f(z), ⇒ Edited by: Shang-Da Yang PDE & Complex Variables P13-8 Res f ( z ) = z = z0 1 d m −1 [( z − z 0 ) m f ( z )]z = z0 (m − 1)! dz m −1 bm b Proof: f(z)= + ... + 1 + m z − z0 ( z − z0 ) ∞ ∑a n =0 n (13.6) ( z − z 0 ) n , g(z)≡(z−z0)m f(z)= bm+ bm-1(z−z0) ∞ + … +b1(z−z0)m-1 + ∑ a n ( z − z 0 ) n + m ; ⇒ b1 is the Taylor coefficient of the power (z−z0)m-1 n =0 of g(z), ⇒ b1= ( m −1) g ( z0 ) . (m − 1)! E.g. Find the residues of f(z)= lim zf ( z ) = lim( z + 2) −3 = z→0 z →0 1 at all poles. (1) z=0 is a simple pole. Res f ( z ) = z= 0 z ( z + 2) 3 1 . (2) z=−2 is a 3rd order pole, by eq. (13.6), Res f ( z ) = z = −2 8 1 d2 [( z + 2) 3 f ( z )]z =−2 = − 1 . 2 8 2! dz Note: Partial fractions reveal all the residues: f(z)= − 1/ 2 − 1/ 8 − 1/ 4 1/ 8 + . + + 2 z z + 2 ( z + 2) ( z + 2) 3 (*) E.g. Find the residue of f(z)=(cot z) at z=π. Ans: Find the order first. Let u=z−π, cot z =cot u, ⇒ {order of pole z=π of cot z}={order of pole u=0 of cot u}. cot u= cos u 1 − (u 2 / 2!) + (u 4 / 4!) − ... 1 u = − +…, ⇒ u=0 is a simple = sin u u − (u 3 / 3!) + (u 5 / 5!) − ... u 3 pole, ⇒ z=π is a simple pole of cot z. By eq. (13.4), Res z =π cos( z ) cos(π ) = =1. sin( z ) sin ′(π ) Edited by: Shang-Da Yang PDE & Complex Variables P13-9 Appendix 13A − Residue at Infinity If the contour C of ∫ f ( z )dz encloses a non-isolated singularity or a large number of C isolated singularities of f(z), evaluating the “interior” residues becomes inefficient. Instead, we can derive ∫ C f ( z )dz by evaluating “exterior” residues of isolated singularities z1, z2, , zN outside C, and perhaps a residue at infinity: Creating an infinite contour C∞. By eq. (13.3): ∫ C f ( z )dz = ∫ ∫ N C∞ N C∞ f ( z )dz − ∫ f ( z )dz = 2πi ⋅ ∑ Res[ f ( z )] , ⇒ f ( z )dz − 2πi ⋅ ∑ Res[ f ( z )] . To evaluate k =1 z = zk C ∫ C∞ k =1 z = zk f ( z )dz , let C∞: z=reiθ, r→∞, θ =[0,2π] (counterclockwise). By change of variable w=1/z, C∞ in the z-plane becomes C'∞ in the w-plane: w≡ρeiφ=(1/r)e-iθ, ρ→0, φ=[0,-2π] (clockwise), f(z)=f(1/w), ⇒ ∫ C∞′ f (1 / w) − dw = w2 ∫ -C∞′ ∫ C∞ f ( z )dz = f (1 / w) dw , since − C ∞′ only encloses one point w=0 in a w2 f (1 / w) counterclockwise sense, =2πi⋅ Res ,⇒ 2 w=0 w ∫ C f (1 / w) [ ] f ( z )dz = 2πi Res Re s f ( z ) − ∑ 2 w= 0 w zk outside C (13A.1) <Comment> Res w= 0 f (1 / w) = Res f ( z ) . Proof: Let Res f ( z ) =b1, ⇒ the Laurent series of f(z) centered at z =∞ z =∞ w2 Edited by: Shang-Da Yang PDE & Complex Variables P13-10 b1 1 + a 0 + a1 ( z − ∞) + ... } ⇒. By change of variable: w= , f(z)= f(1/w)= z−∞ z a b a a1 f (1 / w) = ... + 1 + 02 + 13 + ... , where b1 is the residue at ... + b1 w + a 0 + + ... , ⇒ 2 w w w w w z=∞ as: f(z)={…+ w=0. E.g. Evaluate z 99e1 / z ∫ C z100 + 1 dz , C: |z|=3. Ans: There are 101 singularities within C (z=0, and 100 points fall on the unit circle), and w−99e w ew z 99e1 / z none outside C, ⇒ use residue at infinity. Res 100 = Res = Res , z =∞ z + 1 w=0 w2 ( w−100 + 1) w=0 w(1 + w100 ) ew =1. ⇒ w→0 1 + w100 since w=0 is a simple pole, = lim ∫ C f ( z )dz =2πi⋅ Res f ( z ) =2πi. z =∞ Edited by: Shang-Da Yang