390 CHAPTER 15. ISOLATED SINGULARITIES

which clearly converges to zero as r→ 0 because r lnr→ 0. Therefore, taking the limit asr→ 0, ∫

large semicircle

L(z)1+ z4 dz+ lim

r→0+

∫ −r

−R

ln(−t)+ iπ1+ t4 dt+

limr→0+

∫ R

r

ln t1+ t4 dt = 2πi

(332

√2π +

332

i√

2π +1

32

√2π− 1

32i√

).

Observing that∫

large semicircleL(z)1+z4 dz→ 0 as R→ ∞,

e(R)+2 limr→0+

∫ R

r

ln t1+ t4 dt + iπ

∫ 0

−∞

11+ t4 dt =

(−1

8+

14

i)

π2√

2

where e(R)→ 0 as R→ ∞. From an Example 15.7.2, this becomes

e(R)+2 limr→0+

∫ R

r

ln t1+ t4 dt + iπ

(√2

)=

(−1

8+

14

i)

π2√

2.

Now letting r→ 0+ and R→ ∞,

2∫

0

ln t1+ t4 dt =

(−1

8+

14

i)

π2√

2− iπ

(√2

)=−1

8

√2π

2,

and so∫

0ln t

1+t4 dt =− 116

√2π2, which is probably not the first thing you would thing of. You

might try to imagine how this could be obtained using elementary techniques. Showing theintegral exists is routine, but I think that finding it might prove impossible. This process isnot always routine.

Example 15.7.6 Let α ∈ (0,1). Find∫

0xα

1+x2 dx.

Note that z→ zα is analytic for z ̸= 0. In fact, using the branch of the logarithm usedabove, it is eln(|z|)α+iarg1(z)α . Now consider

∫Γr,R

1+z2 dz where Γr,R is the contour of theabove problem including the large semi-circle and the small semi-circle. Then it is routineto see that the integrals over the small and large semi-circles converge to 0 as R→ ∞ andr→ 0. There is only one residue at i and it is limz→i (z− i) zα

1+z2 = 12 sin 1

2 πα− 12 icos 1

2 πα.Thus in the limit,∫ 0

−∞

|x|α eiπα

1+ x2 dx+∫

0

1+ x2 dx = 2πi(

12

sin12

πα− 12

icos12

πα

)

and so(∫

0xα

1+x2 dx)(

eiπα +1)= π

(ei 1

2 πα

).Then simplifying, you get the amazing for-

mula∫

0xα

1+x2 dx = π

2cos 12 πα

.

Sometimes one must be “creative” about which contour to use. In the next case, cos(z2)

is not bounded and so integrals which involve a contour over a large semicircle like theabove, are not likely to be helpful.

Example 15.7.7 The Fresnel integrals are∫

0 cosx2dx,∫

0 sinx2dx.

390 CHAPTER 15. ISOLATED SINGULARITIESwhich clearly converges to zero as r + 0 because rlnr — 0. Therefore, taking the limit asr—>0,L(z) —In(—t)+indz+ lim / te[.. semicircle 1 + Za r>0+ J—R 1+ 14lim [tna = 2 9 an 4 Jian + L Van Livin14+ 32 32 32 32 ‘r>04+ Jr; LObserving that fiarge semicircle ee dz— 0as Ro,1121 zat ——dt = ( -~+-i)a’Vv2R)+ tin [a + xf ae ( pt qi) ev2where e(R) — 0. as R > oe. From an Example 15.7.2, this becomesv2 11) 5e(R)+2,hip [ps arvin( Ga = (-g4gi) hveNow letting r + 0+ and R > ~,° Int 1 V2 12 d — rV/2— — —_/ 27?[Maa (-t his va-in(*P) | V2and so fy” ae dt = — 7/22", which is probably not the first thing you would thing of. Youmight try to imagine how this could be obtained using elementary techniques. Showing theintegral exists is routine, but I think that finding it might prove impossible. This process isnot always routine.Example 15.7.6 Let a € (0,1). Find Jy “dx.Ta 142Note that z > z is analytic for z 4 0. In fact, using the branch of the logarithm usedar j . > 24 .above, it is el(lzI)@+iargi(@)@_ Now consider Jp Td where I. is the contour of theabove problem including the large semi-circle and the small semi-circle. Then it is routineto see that the integrals over the small and large semi-circles converge to 0 as R — oo and: . . ae ee | 1: 1r-> 0. There is only one residue at i and it is lim,_,; (z — i) Ligk = 7 SIN 5 MA — ziCOs >A.Thus in the limit,[ per +f x d. 20i I sin la l i COS laX x= 7}, =smn-= —-xl =co 1+x? 0 1422 2 2 2 2and so ( Jo. Sax) (ce +1)=n (37) Then simplifying, you get the amazing for-us2cos 4 x HaSometimes one must be “creative” about which contour to use. In the next case, cos (2)is not bounded and so integrals which involve a contour over a large semicircle like theabove, are not likely to be helpful.mula {. poadx = =Example 15.7.7 The Fresnel integrals are |x cosx°dx, Jy’ sinx*dx.