37.5. THE BROMWICH INTEGRAL 763

If Re p < γ for all p a pole of F (s) and if F (s) is meromorphic and satisfies the growthcondition 37.8, and if f (t) is defined by that Bromwich integral, is it true that F (s) is theLaplace transform of f (t) for large s? Thus

f (t)≡ limR→∞

12π

∫ R

−Re(γ+iy)tF (γ + iy)dy =

12πi

∫γ+i∞

γ−i∞eztF (z)dz

The limit must exist because, as discussed above,

limR→∞

12πi

(∫γ+iR

γ−iReztF (z)dz+

∫CR

eztF (z)dz)

is eventually constant because the contour will have enclosed all poles of F (z), but as Rcontinues to increase, the integral over the curved part CR converges to 0. Let Res be largerthan γ . One needs to consider

L ( f )(s) =∫

0e−st 1

2πi

∫γ+i∞

γ−i∞eztF (z)dzdt

=1

2πi

∫∞

0e−st lim

R→∞

∫ηR

eztF (z)dzdt

This equals

limr→∞

12πi

∫ r

0e−st lim

R→∞

∫ηR

eztF (z)dzdt

Eventually, for all R large enough, the contour includes all of the finitely many poles ofF (z). There are only finitely many poles because of the estimate on F (z). Thus we canpick R large enough that the limit on the inside equals the contour integral. Thus

L ( f )(s) =1

2πi

∫∞

0e−st

∫ηR

eztF (z)dzdt

= limr→∞

12πi

∫ r

0e−st

∫ηR

eztF (z)dzdt

Interchanging the two contour integrals, Theorem 35.3.8,

= limr→∞

12πi

∫ηR

∫ r

0e−(s−z)tF (z)dzdt

= limr→∞

12πi

∫ηR

(1

s− z− e−(s−z)r

s− z

)F (z)dz

=1

2πi

∫ηR

F (z)s− z

dz

Now this contour integral is not zero because F (z) is not analytic on the inside of η∗R. Letthe orientation of ηR be switched and call the new contour η̂R. Then

L ( f )(s) =1

2πi

∫η̂R

F (z)z− s

dz

37.5. THE BROMWICH INTEGRAL 763If Rep < y for all p a pole of F (s) and if F (s) is meromorphic and satisfies the growthcondition 37.8, and if f(t) is defined by that Bromwich integral, is it true that F (s) is theLaplace transform of f (t) for large s? Thusicef(t) = lim & [ ; oti) E (y+ iy)dy = | eF (2)dzRoo 27 20i J y—ic0The limit must exist because, as discussed above,lim — (f, e"F (z)dzt+ | eF(z az)Ro 27 —iR Cris eventually constant because the contour will have enclosed all poles of F(z), but as Rcontinues to increase, the integral over the curved part Cr converges to 0. Let Res be largerthan y. One needs to consider[ e | e" F (z)dzdt0 2ni Y—ic01] oe)= =a elim | e*F (z)dzdtR- oo Tr£(f)(s)This equals1 s/s’ _,lim — i elim [| e*F (z)dzdtr>0 270i Jo R>0 J npEventually, for all R large enough, the contour includes all of the finitely many poles ofF (z). There are only finitely many poles because of the estimate on F (z). Thus we canpick R large enough that the limit on the inside equals the contour integral. Thus1 i of zt=> e e* F (z) dzdt=n f fF)= = tim = | ev e" F (z) dzdtre Qi NRInterchanging the two contour integrals, Theorem 35.3.8,= lim = | fe (S-2)' F (z) dzdtr+0 20i I np1 1 —(s—z)rre 20i Ing \ S—Z S—Z1 F-5. | Qa.271 Ing S—ZNow this contour integral is not zero because F (z) is not analytic on the inside of np. Letthe orientation of 1)p be switched and call the new contour f}p. ThenLNs) = 50 | 2a:2i Jip Z—S