596_FUNCTIONS OF A COMPLEX VARIABLE Ch. iy eeeeseedESeerrer 1 z-1 a(z — Iz — 2) = I(z— 2) "Bz — 2) 1 ACL + 2)? + 30 (1 + 2)(z — 2X3 + 2) For each of the following functions, say whether the indicated point is regular, an essential singularity, or a pole, and if a pole of what order it is. 9. (o) 4, z=0 (@) t=! 10. (b) tan? z, z= 7/2 . cos ) 2 © Bape ea 12. oy 21k E @rp 77! (d) T(z), z=0 [See Chapter 11, equation (4.1).] 5. THE RESIDUE THEOREM Let zo be an isolated singular point of f(z). We are going to find the value of §c f(2) dz around a simple closed curve C surrounding 2» but inclosing no other singularities. Let f(2) be expanded in the Laurent series (4.1) about z= 29 that con- verges near x = 29. By Cauchy’s theorem (V), the integral of the “a” series is zero since this part is analytic. To evaluate the integrals of the terms in the “4” series in (4.1), we replace the integrals around C by integrals around a circle C’ with center at 2 and radius p as in (3.6), (3.7), and Figure 3.1. Along C’, z = 25 + pe; calculating the integral of the 5, term in (4.1), we find 7 28 igi? (5.1) poe =n [ bala tO nib,. le o (s — £0) pe It is straightforward to show (Problem 1) that the integrals of all the other 4, terms are zero. Then §¢ f(z) dz = 2nib,, or since b, is called the residue of f(z) at x = %, We can say $ (2) dz = 2ni - residue of f(z) at the singular point inside C. lcSec. 5 ‘THE RESIDUE THEOREM _ 597 FIGURE 5.1 ‘The only term of the Laurent series which has survived the integration process is the b, term; you can see the reason for the term “residue.” If there are several isolated singularities inside C, say at zo, 21, 22, ‘-*, We draw small circles about each as shown in Figure 5.1 so that f(z) is analytic in the region between C and the circles. Then, introducing cuts as in the proof of Cauchy’s integral formula, we find that the integral around C counterclockwise, plus the integrals around the circles clockwise, is zero (since the integrals along the cuts cancel), or the integral along C is the sum of the integrals around the circles (all counterclockwise). But by (5.1), the integral around each circle is 2ni times the residue of f(z) at the singular point inside. Thus we have the residue theorem: $ (2) dz = 2ni + sum of the residues of f(z) inside C, where the integral around C is in the counterclockwise direction. The residue theorem is very useful in evaluating many definite integrals; we shall consider this in Section 7. But first, in Section 6, we need to develop some techniques for finding residues. PROBLEMS, SECTION 5 1. If Cis a circle of radius p about zo, show that dz ———=2n if n=1, lc (2 — 20)" but for any other integral value of 1, positive or negative, the integral is zero. Hint: Use the fact that z = zy + pe” on C. 2. Verify the formulas (4.3) for the coefficients in a Laurent series, Hint: To get a,, divide equation (4.1) by (z — z9)"*" and use the results of Problem | to evaluate the integrals of the terms of the series. Use a similar method to find 6,. 3. Obtain Cauchy’s integral formula (3.9) from the residue theorem (5.2).398 FUNCTIONS OF A COMPLEX VARIABLE Ch. 14 6. METHODS OF FINDING RESIDUES A. Laurent Series If it is easy to write down the Laurent series for f(z) about = zo that is valid near zo, then the residue is just the coefficient b, of the term Iz — zo). Caution: Be sure you have the expansion about z = zo; the series you have memorized for ¢*, sin z, etc., ate expansions about z = 0 and so can be used only for finding residues at the origin (sce Section 4, Example 3). Here is another example: Given f(z) = e(z — 1), find the residue, R(1), of f(z) at z = 1. We want to expand ¢ in powers of z — 1; we write 21 iat) [ise-ne Pe] e =mtet z-1 Then the residue is the coefficient of 1/(z — 1), that is, R(l) =e B, Simple Pole If f(z) has a simple pole at z = zo, we find the residue by multi- plying f(z) by (z — zo) and evaluating the result at z = 29 (Problem 10) Example 1. Find R(—4) and R(5) for z (0a Multiply /(z) by (z + 4) (Caution: not by (2z + 1)] and evaluate the result at 2 = —4. We find (2+ Df) = (2 +4) eee (e+ 15 —2) 2 x) R-D= 5 =-# Similarly, 2 2 &- SSO =@- 5) FE p DG-2) 7 41? RS) = Example 2. Find R(0) for f(2) = (cos 2)/z. Since f(z) = cos 2, we have R(0) = (cos £),-9 = c05 0 = 1,Sec. 6 METHODS OF FINDING RESIDUES _ 599 To use this method, we may in some problems. have to evaluate an indeterminate form, so in general we write f° R(z) = lim(z — zo) f(z) when 2p is a simple pole. a0 Example 3. Find the residue of cot 2 at z = By (6.1), R(0) = lim = 20 SI z eH = cos 0+ lim —— 20 SINS If, as often happens, f(z) van be written as g(z)/h(z), where g(z) is analytic and not zero at Z9 and A(z) = 0, then (6.1) becomes tim AED) oy tim ETA mo R20) = Fim ey BEE) Bim Fy = 8G 0) im Heo) by L’Hépital’s rule or the definition of #(z) (Problem 11). Thus we have- ) S (2) = g(z)/h(@), and R(&o) = so) if | (20) = finite const. # 0, and (eo) h(z9) = 0, A'(zo) # 0. Often (6.2) gives the most convenient way of finding the residue at a simple pole. Example 4. Find the residue of (sin 2)/(1 — 24) at z =i By (6.2) we have sin z sini Re Tiga A = Gap THe Da A sinh. Now you may ask how you know, without finding the Laurent series, that a function has a simple pole. Perhaps the simplest answer is that if the limit obtained using (6.1) is Some constant (not 0 or <0), then f(z) does have a simple pole and the constant is the residue. [If the limit = 0, the function is analytic and the residue = 0; if the limit is infinite, the pole is of higher order.] However, you can often recognize the order of a Pole in advance. [See end of Section 4 for the simple case in which (z — 29)" is a factor of the denominator.] Suppose f(z) is written in the form g(z)/A(z), where g(z) and A(z) are analytic. Then you can think of g(z) and A(z) as power series in (z — zo). If the denominator has the factor (z — Zo) to one higher power than the numerator, then f(z) has a simple pole at 29. For example, (1 al +- el + & cos? z zcor? z= sin? z600_FUNCTIONS OF A COMPLEX VARIABLE Ch. 14 has a simple pole at z = 0. By the same method we can see whether a function has pole of any order. C. Multiple Poles When f(z) has a pole of order m, we can use the following method of finding residues. Multiply f(z) by (2 — 20)", where m is an integer greater than or equal to the order n of the pole, differentiate the result m— 1 times, divide by (m — 1)!, and evaluate the resulting expression at z = Zo. It is easy to prove that this rule is correct (Problem 12) by using the Laurent series (4.1) for f(z) and showing that the result of the outlined process is by. Example 5. Find the residue of f(z) = (z sin 2)/(z — m)° atx =n. We take m = 3 to eliminate the denominator before differentiating; this is an allowed choice for m because the order of the pole of f(z) at m is not greater than 3 since is finite at 7. (The pole is actually of order 2, but we do not need this fact.) Then following the rule stated, we get a 1 Rin) = 3 7a (@ sin 2) =4[-zsinz +2 cos z],-,=—1 (To compute the derivative quickly, use Leibniz’ rule for differentiating a product; see Chapter 12, Section 3.) PROBLEMS, SECTION 6 Find the Laurent series for the following functions about the indicated points; hence find the residue of the function at the point. (Be sure you have the Laurent series which converges near the point.) 1 1 ——,z 2+ 1) cosh z ze +z sin mz 427-1" 10. Show that rule B is correct by applying it to (4.1). 11. Derive (6.2) by using the limit definition of the derivative A(z9) instead of using L’Hépital’s rule. Remember that h(z9) = 0 because we are assuming that f(z) has a simple pole at zo. 12. Prove rule C for finding the residue at a multiple pole, by applying it to (4.1). Note that the rule is valid for » = 1 (simple pole) although we seldom use it for that case.Sec. 6 METHODS OF FINDING RESIDUES _ 601 13. Prove rule C by using (3.9). Hints: If f(z) has a pole of order m at f(2) = g(2)(z — a)" with g(2) analytic at z = a. By (3.9), , then f 82) 1. = anig(a) ic (2 — a) with Ca contour inclosing a but no other singularities. Differentiate this equation (n — 1) times with respect to a. (Or, use Problem 3.21.) Find the residues of the following functions at the indicated points. Try to select the easiest method. 1 I or ot —fandatz=2 0 15. —————_ at z= $andatz=4 “ Giyne-» 5 (tye 4 z 5 -2 +2 16. A at 2=Oandatz=1 17. Bo atc=4andat z= —$ +2 18. a7 at c= 3 19. nent 2 am 20. atz= a. a6 atz= V/A1 +i) 2 a : 2 asain 23. : l+e 1 — cos 2 a wu —S= ae=0 2s, < atz=0 z z ene i. cos z 2%. atz=em m7, SOE r= nfo a 1—2sinz ce 2 mt enti 2, —“—— z=? (2? + 9)(2? + 1) 4 cosh z— 5 he-l 38 32-1 yo, SEEN az =0 3. =F az=0 z z fe i+ 2 So ated! 33. RE ean (2? +4) (x — 2) uw 2 at ceoandate=t 33. —- pS at a= and at z= a ates 2(1 — 22)? - (22 +17 14" to 35’. Use the residue theorem to evaluate the contour integrals of each of the functions in Problems 14 to 35 around a circle of radius } and center at the origin, Check carefully to see which singular points are inside the circle. You may use your results in the previous problems as far as they go, but you may have to compute some more residues. 36. For complex z, J,(z) can be defined by the series (12.9) in Chapter 12. Use this definition to find the Laurent series about z = 0 for z~*Jo(z). Find the residue of the function at 2=0.602_ FUNCTIONS OF A COMPLEX VARIABLE Ch. 14 7. EVALUATION OF DEFINITE INTEGRALS BY USE OF THE RESIDUE THEOREM We are going to use (5.2) and the techniques of Section 6 to evaluate several different types of definite integrals. The methods are best shown by examples. 22 ap Example 1. Find / = ena lo 5 +4 cos 6 If we make the change of variable z = e, then as 6 goes from 0 to 2x, = traverses the unit circle |z| = 1 (Figure 7.1) in the counterclockwise direction, and we have a contour integral. We shall evaluate this integral by the residue theorem. If z =e", we have 1 dz=ie® d0=izd0 or d0=—dz, iz open FF cos § = — =—. FIGURE 7.1 ain 2 2 Making these substitutions in /, we get 1 —dz iz ee dz +Az+ Mz) i Jo set 227 +2 dz (2z + Iz + 2)’ where C is the unit circle, The integrand has poles at z= —4 and z= —2; only = = —tis inside the contour C. The residue of 1/[(2z + I)(z + 2)] at 2 = —4is le inn (ete ay eect eee =4 OS ene) QetNEF2 WDE a” Then by the residue theorem 1 2 == 2niR(—4) =n 4 =. i 3 This method can be used to evaluate the integral of any rational function of sin @ and cos @ between 0 and 2n, provided the denominator is never zero for any value of @ You can also find an integral from 0 to 7 if the integrand is even, since the integral from 0 to 2x of an even periodic function is twice the integral from 0 to 7 of the same function, (See Chapter 7, Section 9 for discussion of even and odd functions.) * dx Eval = ; Example 2, Evaluate / [ iFSec. 7. EVALUATION OF DEFINITE INTEGRALS _ 603 Here we could easily find the indefinite integral and so evaluate / by elementary methods. However, we shall do this simple problem by contour integration to illustrate a method which is useful for more complicated problems. This time we are not going to make a change of variable in /. We are going to start with a different integral and show how to find / from it. We consider b dz » 7 a Ete On semicircle where C is the closed boundary of the semi- z= neil circle shown in Figure 7.2. For any p > 1, the semicircle incloses the singular point z= and no others; the residue of the integrand at z =i is 1 1 R() = lim(z — ) ———— = —. © zen : (2-Het+i) 2 FIGURE 7.2 ‘Then the value of the contour integral is 2zi(1/2i) = m. Let us write the integral in two parts: (1) an integral along the x axis from —p to p; for this part z = x; (2) an integral cle, where & = pe”. Then we have iE dx [ pie® 40 -pl tx?” Jo 1+ per?” We know that the value of the contour integral is x no matter how large p becomes since there are no other singular points besides =i in the upper half-plane. Let p> 00; then the second integral on the right in (7.1) tends to zero since the numerator contains p and the denominator p?. Thus the first term on the right tends to 7 (the value of the contour integral) as p -» 00, and we have 7 ds © dg tx? This method can be used to evaluate any integral of the form = P(x) ~@ Q(x) if P(x) and Q(x) are polynomials with the degree of Q at least two greater than the degree of P, and if Q(z) has no real zeros (that is, zeros on the x axis). If the integrand P(x)/Q(x) is an even function, then we can also find the integral from 0 to 00. dx Example 3. Evaluate / -| ees Jo +e ef dz lol +2?" We consider the contour integral604 FUNCTIONS OF A COMPLEX VARIABLE Ch. 14 where C is the same semicircular contour as in Example 2. The singular point incloseg is again z = i, and the residue there is Ter ea cing Up asin age The value of the contour integral is 2ni(1/2ie) = x/e. As in Example 2 we write the contour integral as a sum of two integrals: 0 ek de ef de ~pl ts? L+27° atong upper half of = pe (7.2) As before, we want to show that the second integral on the right of (7.2) tends to zero as p>» oo. This integral is the same as the corresponding integral in (7.1) except for the e factor. Now fe] = fer = fe] fe = y<1 since y > 0 on the contour we are considering. Since |e*| < 1, factor does not change the proof given in Example 2 that the integral along the semicircle tends to zero as the radius p—> 00. We have then gts ee ee fa ‘ or taking the real part of both sides of this equation, (ae Since the integrand (cos. x)/(1 + x?) is an even function, the integral from 0 to x is half the integral from — 0 to 00. Hence we have cosxde T= pobsc TELE lo lta? 2 la aH Observe that the same proof would work if we replaced e? by e'"* (m > 0) in the above integrals, At the point where we said ¢~” < 1 (since y > 0) we would then want 7" <1 for y > 0, which is true if m > 0. [For m <0, we could use a semicircle in the lower half-plane (y <0); then we would have e”” <1 for y <0. This is an unne- cessary complication, however, in evaluating integrals containing sin mx or cos mx since we can then choose m to be positive.] Although we have assumed here that (as in Example 2) Q(x) is of degree at least 2 higher than P(x), a more detailed proof (see books on complex variable) shows that degree at least one higher is enough to make the integral Pe) Q(z) em deSec. 7 EVALUATION OF DEFINITE INTEGRALS 605 around the semicircle tend to zero as p> 00. Thus ema, ; ; PO) ims POR) same dy = 2ni- sum a) C Dia)" A = 2m sum of the residues of G1 in the upper half-plane if all the following requirements are met P(x) and Q(x) are polynomials, and Q(x) has no real zeros, and the degree of Q(x) is at least 1 greater than the degree of P(x), and m> 0. By taking real and imaginary parts, we then find the integrals 7 “ome es io PO) sin me di. Q(x) > sin x Example 4. Evaluate | Here we remove the restriction of Examples 2 and 3 that Q(x) has no real zeros. As in Example 3, we consider [Ee z To avoid the singular point at z = 0, we integrate around the contour shown in Figure 7.3. We then let the radius r shrink to zero so that in effect we are integrating straight through the simple pole at the origin. We are going to show (later in this section and Problem 21) that the net result of integrating in the counterclockwise direction around a closed contour which passes straight* through one or more simple poles is 2xi - (sum of the residues at interior points plus one-half the sum of the residues at the simple poles on the boundary). (Warning : this rule does not hold in general for a multiple pole on a boundary.) You might expect this result. If a pole is inside a contour, it contributes 2ni - residue, to the integral; if it is outside, it contributes nothing; if it is om the straight line boundary, its contribution is just halfway between zero and 27i - residue. Using this fact, and observing that, as in Example 3, the integral along the large semicircle tends to zero as R tends to infinity, we have ate is | © dy = dni 4(resitue of £ at = 0) =2ni-$-1 sin. “a Taking the imaginary parts of both sides, we get = sin x — de =n. cee * By “straight” we mean that the contour curve has a tangent at the pole, that is, it does not turn a corner there.606 FUNCTIONS OF A COMPLEX VARIABLE. Ch. 14 To show more carefully that our result is correct, let us return to the contour of Figure 7.3. Since e/g is analytic inside this contour, the integral around the whole contour is zero. As we have said, the integral along C tends to zero as R—> 00 by the theorem at the end of Example 3. Along the small semicircle C’, we have =k FIGURE 7.3 : d zr dz=re%do, “=i 40, 2 [S4-[ew lez As r—>0, 2» 0, e+ 1, and the integral (along C’ in the direction indicated in Figure 7.3) tends to ° id@ = —in. Then we have as R-> 00, and r—> 0, or as before. Principal Value Taking real and imaginary parts of this equation (and using Euler’s formula ¢* = cos x + i sin x), we get : eae [ SSF te =, [ een a cs Since (sin x)/x is an even function, we have ? sin x 1° sine n [ 7 trad [” an de =F. However, is a divergent integral since the integrand (cos x)/x is approximately 1/x near x = 0. The value zero which we found for I = {®,, (cos x)/x dx is called the principal value (or Cauchy principal value) of 7. ‘To see what this means, consider a simpler integral, namely S de [ eeSee. 7 EVALUATION OF DEFI INITE INTEGRALS _ 607 ‘The integrand becomes infinite at x = 3, and both J} dx/(x — 3) and [$ dx/(x — 3) are divergent. Suppose we cut out a small symmetric interval about x = 3, and integrate from 0 to 3 — rand from 3 + r to 5. We find oo os { 7 =[ntr—ai] =Inr—In3, 10 x jo 5 in ne a+h X ‘The sum of these two integrals is In 2~In 3=In3; this sum is independent of r. Thus, if we let r+ 0, we get the result In 3 which is called the principal value of Sa a { * often written PV’ Sin 2 |, r—3 or The terms In r and ~In r have been allowed to cancel each other; grap! infinite area above the x axis and a corresponding infinite area below the x axis have been canceled. In computing the contour integral we integrated along the x axis from — up to —r, and from +r to +0, and then let r—+ 0; this is just the process we have described for finding principal values, so the result we found for the improper integral {®,, (cos x)/x dx, namely zero, was the principal value of this integral. Example 5. Evaluate co ppd dr, 0 0 or if r— 20. (Verify this; note that the denominator is approximately 1 for small r, and approximately re for large r.) Thus the integrals along the circular parts of the contour tend to zero as the little circle shrinks to a point and the large circle expands indefinitely. We are left with the two integrals along the positive x axis with 4B now extending from 0 to © and DE from « to 0. Along 18 °° = y, and this integral is Betpey [ Ya. Noo ltr Along DE, we have 6 = 2x, so z = re*™ and this integral is { were arm —[ aii , , we agreed to have @ = 0, so z = re! 1+ re? l+r Adding the AB and DE integrals, we get (= et) [ vt d= Ini? | VrSee. 7 EVALUATION OF DEFINITE INTEGRALS _ 609 by (7.4). Then the desired integral is erect —2nie'"? nid ru (7.5) { dr = sap = Gap ee lo He a Let us use (7.5) to obtain (5.4) of Chapter 11. Putting g = 1 — p in (6.5) and (7.1) of Chapter 11, we have = yt aoi-n~[ dy and (o, Pp | lay (7.6) Boo, 1—p)=T(p)T(1—p) since T(1) = 1. Combining (7.5) and (7.6) gives (5.4) of Chapter 11, namely © yp) TPC — p) = BG, 1 -n=[ 149” 7 Sin =a Argument Principle Since w = f(z) is a complex number for each z, we can write w = Re® (just as we write z = re”) where R=|m| and © is the angle of » [or we could call it the angle of f(z)]. As z changes, » = f(z) also changes and so R and @ vary as we go from point to point in the complex (x, y) plane. We want to show that (a) if f(z) is analytic on and inside a simple closed curve C and f(z) # 0 on C, then the number of zeros of f(z) inside C is equal to (1/2m) - (change in the angle of f(z) as we traverse the curve C); (b) if f(z) has a finite number of poles inside C, but otherwise meets the require- ments stated,* then the change in the angle of f(z) around C is equal to (2m) - (the number of zeros minus the number of poles). (Just as we say that a quadratic equation with equal roots has 10 equal roots, so here we mean that a zero of order n counts as n zeros and a pole of order n counts as 1 poles.) fe) d: t fe) To show this we consider By the residue theorem, the integral is equal to 2i- (sum of the residues at singu- larities inside C). It is straightforward to show (Problem 42) that the residue of F(z) = f'(2)/f (2) at a zero of f(z) of order m is n, and the residue of F(z) at a pole of f(z) of order p is —p. Then if N is the number of zeros and P the number of poles of £(2) inside C, the integral is 2xi(N — P). Now by direct integration, we have £'®) f() where R =| f(z)| and © is the angle of f(z). Recall from Chapter 2, Section 13, that Ln R means the ordinary real logarithm (to the base ¢) of the positive number R, and (7.7) dz =In f(z) =LnR| +i0 ic Ic Ic = In Re® ic * A function which is analytic except for poles is called meromorphic.610 FUNCTIONS OF A COMPLEX VARIABLE Ch. 14 is single-valued; In f(z) is multiple-valued because © is multiple-valued. Then if we integrate from a point A on C all the way around the curve and back to 4, Ln R has the same value at A both at the beginning and at the end, so the term Ln Ric is Ln R at A minus Ln R at A; this is zero. The same result may not be true for ©; that is, the angle may have changed as we go from point A all the way around C and back to A. (Think, for example, of the angle of 2 as we go from z = 1 around the unit circle and back to z = 1; the angle of z has increased from 0 to 2m.) Collecting our results, we have -1¢$f@,.1 (78) N~ P= 520 7 de = 7 ic = s + (change in the angle of f(z) around C), where N is the number of zeros and P the number of poles of f(z) inside C, with poles of order » counted as n poles and similarly for zeros of order n. Equation (7.8) is known as the argument principle (recall from Chapter 2 that argument means angle) This principle is often used to find out how many zeros (or poles) a given function has in a given region. (Locating the zeros of a function has important applications to determining the stability of linear systems such as electric circuits and servo- mechanisms. See, for example, Kuo, page 361, or Kaplan, Operational Methods, Chap- ter 7.) Example 6. Let us show that f(z) = z° + 4z + 1 = 0 at exactly one point in the first quadrant. The closed curve C in (7.8) is, for this problem, the contour OPQ in Figure 7.5, where PQ is a large quarter circle. We first observe that x? + 4x + 1 > 0 for x > 0 and (iy)> + 47y + 1 #0 for any y (since its real part, namely 1, #0); then f(z) #0 on OP or OQ. Also f(z) #0 on PQ if we 11@ choose a circle large enough to inclose all zeros. We now want to find the change in the angle © of f(z) = Re'® as we go around C. Along OP, z = x; then f(z) =/(z) is real and so © = 0. Along PQ, 2 = re"*, with r constant and very large. For very large r, the Pp z° term in f(z) far outweighs the other terms, and we have ©] ‘ f(e) = =e", As @ goes from 0 to 7/2 along PQ, © = 30 FIGURE 7.5 goes from 0 to 3n/2. On QO, x = iy, f(z) = —iy? + 4iy + 1; then 4 3 part of f(z) _ real part of f(z) For very large y (that is, at Q), we had © © 3n/2 (for y= 00, we would have tan 6 = —00, and © would be exactly 37/2). Now as y decreases along QO, the value of tan © = 4y — y® decreases in magnitude but remains negative until it becomes 0 at y= 2. This means that © changes from 31/2 to 2x. Between y=2 and y =0, the tangent becomes positive, but then decreases to zero again without becoming infinite. This means that the angle © increases beyond 2x but not as far as 2x + 7/2, and then