36.7. EXERCISES 747

where E (r) ,E (R) converges to 0 as r→ 0 and R→ ∞. There is no hope of taking a limitas R→∞ while keeping r > 0 fixed, but if we let both variables converge at the same time,then we could get something. Let r→ 0+ and let R = 1/r.

2πieiπ(p−1) =∫ 1/r

r

xp−1

1+ xdx(

1− e(p−1)i(2π))+E (r,1/r)

and now, as r→ 0,which equals the sum of the two integrals over the straight lines added tothe integrals over the small circles which converge to 0 as r→ 0 and R→ ∞. Top straightline converges as r→ 0 to ∫ 1

0

(Rt)p−1

1+ tRRdt

Bottom integral converges as r→ 0 to

−∫ 1

0

|tR|p−1 e(p−1)i(2π)

1+ tRRdt

Of course change variables letting x = tR and the two integrals which must be summed are∫ R

0

xp−1

1+ xdx,−e(p−1)i(2π)

∫ R

0

xp−1

1+ xdx

Thus you get the following as R→ ∞.∫∞

0

xp−1

1+ xdx(

1− e2π(p−1)i)= 2πieiπ(p−1)

Then what you get is ∫∞

0

xp−1

1+ xdx =

2πieiπ(p−1)

1− e2π(p−1)i

=−2πieiπ p

1− e2π pi =−2πi

(1− e2π pi)e−iπ p =−2πi

e−iπ p− eπip

=−2πi

(cosπ p− isin(π p))− (cosπ p+ isin(π p))=

π

sin(pπ)

I think this is quite an amazing result.Actually, people typically are a little more informal in the consideration of such inte-

grals. They regard the bottom side of the line x ≥ 0 as being associated with θ = 2π andthe top side being associated with θ = 0 and leave out the fuss with taking limits as ε → 0and so forth.

36.7 Exercises1. Find the following improper integral.

∫∞

−∞

cosx1+x4 dx Hint: Use upper semicircle con-

tour and consider instead∫

−∞

eix

1+x4 dx. This is because the integral over the semicirclewill converge to 0 as R→∞ if you have eiz but this won’t happen if you use cosz be-cause cosz will be unbounded. Just write down and check and you will see why thishappens. Thus you should use eiz

1+z4 and take real part. I think the standard calculustechniques will not work for this horrible integral.

36.7. EXERCISES 747where E (r) , E (R) converges to 0 as r + 0 and R — ©. There is no hope of taking a limitas R — ~ while keeping r > 0 fixed, but if we let both variables converge at the same time,then we could get something. Let r — 0+ and let R = I /r.Qniei™(P—!) = /r 1+xand now, as r — 0,which equals the sum of the two integrals over the straight lines added tothe integrals over the small circles which converge to 0 as r — 0 and R > . Top straightline converges as r > 0 to1 Rt p-l[ (Rr) Rdt01+tR1/r yp-1dx (1 - elr Hien) ) +E (r,1/r)Bottom integral converges as r + 0 to-[ [ERP Te DICH)01+1tROf course change variables letting x = ¢tR and the two integrals which must be summed areR yp-l R yp-li a dx, —e'?!)i(2) i a dxJo +x” Jo 1+xThus you get the following as R > .co yp—li < dx(1 — 2a 1i) = 2niei™(P—!)Jo XxThen what you get isco yPol 2nie'™(P—!)[ dx = :o 1+x 1 e2n(p-1)i—2nie!™? —2ni —2ni| e2tpi a _ e27i) e- itp e—inp _ etip—27i 1(cosap—isin(zp))—(cosxp+isin(zp)) — sin(p7)I think this is quite an amazing result.Actually, people typically are a little more informal in the consideration of such inte-grals. They regard the bottom side of the line x > 0 as being associated with 0 = 27 andthe top side being associated with @ = 0 and leave out the fuss with taking limits as € > 0and so forth.36.7 Exercises1. Find the following improper integral. [™, Ta7 dx. This is because the integral over the semicircledx Hint: Use upper semicircle con-tour and consider instead |, 7will converge to 0 as R —> if you have e but this won’t happen if you use cos z be-cause cosz will be unbounded. Just write down and check and you will see why thishappens. Thus you should use & z and take real part. I think the standard calculustechniques will not work for this horrible integral.