AN ABSTRACT OF THE THESIS OF George Earl Barr for the Ph. D. in Physics (Major) Name) Date thesis is presented Title ON THE DIFFRACTION OF A CYLINDRICAL PULSE BY A HALF PLANE Abstract approved Redacted for privacy (Major professor) The effect of a pulse produced by an acoustic or electro- magnetic line source oriented parallel to the edge of a perfectly conducting half plane is considered. The Green's functions for the modified Helmholtz equation are found by solving for the Green's functions for a cylinder of sectorial cross-section, then allowing the radius of the sector to go to infinity to form a wedge, and finally opening the wedge to form a half plane. The Green's function for the sectorial domain itself, is determined by first solving the eigenvalue problem for the sectorial domain. The solution for the half plane is repre- sented in the form of a Laplace transform integral. Then the solu- tion for a "Dirac" pulse, which is the inverse Laplace transform of the Green's function for the modified Helmholtz equation, is directly available and the result of an arbitrary pulse can be synthesized from it by integration. The results for several pulses are produced by this method and explicitly show the "transient" terms. ON THE DIFFRACTION OF A CYLINDRICAL PULSE BY A HALF PLANE by GEORGE EARL BARR A THESIS submitted to OREGON STATE UNIVERSITY in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY June 1 965 APPROVED: Redacted for privacy Professor of Mathematics In Charge of Major Redacted for privacy Ww Chairman ofgepartment of Physics Redacted for privacy Dean of Graduate School Date thesis is presented Typed by Carol Baker TABLE OF CONTENTS Page Chapter 1 INTRODUCTION 2 GREEN' S FUNCTIONS 3 REPRESENTATION OF GREEN'S FUNCTIONS 16 AS LAPLACE TRANSFORM INTEGRALS Time Harmonic Case 5 26 4 TRANSIENT SOLUTIONS 28 5 SOME SPECIAL CASES 32 BIBLIOGRAPHY 38 APPENDIX I 40 APPENDIX II 44 LIST OF FIGURES Figure Page 1 6 2 7 3 20 4 21 23 6 30 ON THE DIFFRACTION OF A CYLINDRICAL PULSE BY A HALF PLANE CHAPTER 1 INTRODUCTION The diffraction of electromagnetic or acoustic pulses by a wedge (and by its special case, a half plane) has been treated in a number of recent contributions. The majority of these are based upon a direct attack on the wave equation of arbitrary (not necessarily time harmonic) time dependency. By this method the following cases have been treated: Normal incidence of a plane pulse on a half plane [3] , [41; two dimensional pulse incident on a half plane [4]; plane unit step function pulse incident on a wedge [9], [11], [12]; arbitrary pulse incident on a wedge [8]. Another approach centers around integral transform methods. For instance, the case of a spherical pulse (unit step function) incident on a half plane [18] uses the Fourier transform method and the case of a cylindrical "Dirac" pulse incident on a half plane [17] uses the Kontorovich-Lebedev transform method [10] . The solution of the latter problem is given in the form of an infinite series involving Legendre functions of such extremely slow convergence as to make the result practically useless. It will be shown here that the solution of this problem can not only be given in elementary form but that it involves functions of algebraic character 2 only. In addition the method employed here is much more straight- forward and not just restricted to this particular case. The procedure is as follows: First we determine Green's functions -a-1 for the so-called "modified" Helmholtz equation 2 - y2u = 0 (exponential decay case), which is obtained from Helmholtz's equation there k = -iy. Au + k u = 0 (time harmonic case) putting Then the "Dirac" pulse solution is simply given by the inverse Laplace transform of Gi with respect to y (14). 2 Having obtained the "Dirac" pulse solution a further integration yields the arbitrary pulse case [14]. Although only pulse diffraction phenomena with a half plane are considered here, we shall deal however in the first stage of our investigation with the somewhat more general configuration of a sectorial domain, as indicated in Figure 2, bounded by 0 = a, 0 = 0, p = a. In Chapter 2 we represent G1 in the form of an infinite 2 (double) series involving eigenfunctions for this sectorial domain. One summation can be carried out using a special Fourier-Bessel expansion formula in which -di is expressed in the form of a 2 single series. Upon letting a (the radius of the sector) tend to infinity, both Green's functions for a wedge of angle a are obtained and then finally the special case a = 2.7 leads to the two dimen- sional Green's functions for the half plane. In Chapter 3 we express G1 in the form of a Laplace 2 transform integral with respect to the parameter y for the special case of a half plane, i. e. G 1 2 where F(t) F (te) dt 0 does not depend on y . It should be pointed out that F(t) is found to be a very simple elementary function, while the usual integral or series expressions for Gi involve higher tran- 2 scendental functions (see for instance [1] , [61, [7], [13] [17]. , A number of summation formulae which are necessary to arrive at such a simple representation are derived in Appendix I. In Chapter 4 we derive the "Dirac" pulse solution as an inverse Laplace transform of Gi with respect to y . But this 2 property as expressed in Chapter 3, itself occurs in the form of a Laplace transform, hence the "Dirac'? pulse solution (as the Laplace inverse of a Laplace transform) is simply the elementary function F(t) itself. The solution for an arbitrary pulse excitation g(t) is then given in the form of an integral involving the pulse function g(t) and the "Dirac" pulse solution. In Chapter 5 we give a few special cases. They include pulse functions a) g(t) of the following character: Unit step function [instantaneous line source, g(t)= 1] time harmonic pulse [g(t) = exp rectangular time harmonic pulse of duration T, [g(t) = exp it, t < T, g(t) = 0. t> T. ] In all three cases g(t) = 0 for t < 0, 1. e. the pulse excitation be- gins at t = 0. Finally it is shown that the two Green's functions for the Helmholtz equation (time harmonic case) can be obtained from the case b) before, as the limiting case for t = 00. For this purpose the pulse solution corresponding to the case b) is represented as the sum of two contributions, the "transient" part and the "quasi- stationary" part. The latter part is precisely Green's function for the Helmholtz equation while the "transient" part becomes negligible when t becomes large enough. 5 CHAPTER 2 GREEN'S FUNCTIONS As mentioned in the introduction we set, in Helmholtz's equation Au + k2u (v = k= 0, (1) is the velocity of propagation, w is the frequency and X. the wave length) (2) and obtain y2u = (3) 0 which we shall call the "modified" Helmholtz equation. The transi- tion from k to y amounts to the transition from a wave problem to an exponential decay problem. At first the exponential decay problem is solved and the wave problem solution is established by returning to k. Often this simplifies questions with respect to convergence considerably, since in this case we are dealing with monotonic functions instead of with oscillatory functions. (See for instance [13] , [14] , [19]). Consequently the definition of the two Green's functions G1 of (3) for a domain D bounded by a curve 2 C, if we regard u to be independent of z, is (See Figure 1): 6 Figure 1 a) A-G-1 y-di - 2 b) G1+ 0 everywhere in D save for PQ) is regular inside 2 ,17.. x = x' and y = y' C, 2 c) When P is on C then: for the first Green s function, 0 or aG2 for the second Green's function. an and aG an is the differential quotient of G, with respect to the direction normal to C. Ko(z) is the 7 modified Hankel function (see Appendix II). The distance between P and Q is denoted by PQ. Having obtained Gi for (3) one can obtain G1 for (1) 2 2 by replacing y by ik. Since the main interest here centers around pulse problems and since the solution of those requires GI rather 2 than G1 the latter will receive only casual treatment. 2 We now consider the sectorial domain D as indicated in Figure 2 i. e. our configuration is a cylinder whose cross-section is a circular sector of radius a and angle Figure 2 Since we shall give an expansion of Gi in terms of the normalized 2 eigenfunctions of Helmholtz's equation (1), we observe that a solution of (1) in cylindrical coordinates can be given in the form (see for instance [16], Chapter V) 8 U(p, 0) = (4) (P)f2(0) where f ( )= J (kp) + A1 v Y (kp) (5) B1 v f2(0) = A2cos(4) + B2sin(vsb) (6) [Jv(z) is the Bessel function and Yv(z) is the Neumann function, both of order v and argument z, (see Appendix II)] In order to establish the eigenfunctions for D we have to find solutions of (1) which are finite in D and which satisfy certain boundary We establish here the results for the first boundary value conditions. problem, only ( = 0 on the boundary of second boundary value problem ( 8u The analysis for the D). on the boundary of D) On follows the same line of reasoning. The condition u = 0 for 0 = 0 and 0 = a gives, by (6): A The function f f = 0, (p) sin(va) = 0 or v= nr a y n = 1, 2, 3, in (5) then reduces to 1(p) = Al Jnr (kp) . a The corresponding Neumann function has to be rejected since to be finite for have the form: p= U is 0; hence B1 = 0. Therefore the eigenfunctions 9 U= nil a JnTr 95 ) a They vanish for çb = 0 and tp = a. In order that U vanishes for p = a also, we have J nTr (ka) = 0 a This equation determines the eigenvalues. We denote by x = T v, m the mth positive root of J(x) = 0. Then v ka = T nTr a n, m = 1, 2, 3, ... , m and the eigenvalues are = kn, m =TriTr n, m = 1, 2, 3, . (7) a The eigenfunctions U U = Un, m for D are then = J n-rr a T nit a m pia sin (3.1r±)f a In order to obtain the normalized system U1 of eigenfunctions we form a ,c p= 0 Hence 0= 0 U2 do- , do- = pdpd0 . 10 a 7 sin2(_nArtp ) j 2 /Tn-rr do 00= 0 p/a\ pdp . 0 nn a a\ P- The integration over m yields a.12 95 . In order to obtain the integral over p we apply a well-known formula from the theory of Bessel functions [2, vol. 2 p. 901: 2 xJ2 (a x)dx = 2 v (ax) - J v+.1 (a x) J v-1 (a x)] with this formula we obtain c a J2 r, v ) pdp T mr a a. = 12-a2 m r2 / JniT (iTnTr a\a _nn- a Now, \ (T \ a / (Tmr, a. a by definition and the first term within the =0 J Tin /Tmi. ,...-,.. m CL rn a bracket on the right-hand side vanishes. And, we use the recurrence relation [ 2, vol. 2, p. 121 2v Jv -1(x) = X V (X) Jv+1 (X) hence /yr a -1 Tmr a m. = +1k a, -j nrr a nrr 11 [since again Jnir 7TnTr a = 01 . M Therefore ra 0 7 nTr nir a. a. a) pdp 1 = a 2 2 J nir a (7nil 1- M.) and hence, finally N2 = 12 aa 4 a a The system of normalized eigenfunctions of (1) for a domain as sketched in Figure 2 is then 1 1 U =U n, m =2 aa7 J (.1" +1 a a rn) (8) fir sin ( a 3nTr nTr a a , n, m = 1, 2, 3, ... )p/a We now make use of the well-known expansion for Green's function G(PQ) of Helmholtz's equation (1) into eigenfunctions of (1) [1 6, p. 183] G(PQ) = tp)ui (Q)/,..2..,_ 2, :IC 3 ... 1 ' (9) where ki are the eigenvalues and U(P) and U(Q) are the normalized eigenfunctions of (1) for the domain under consideration, 12 with reference to the point of observation P = P(p,sb) and the source point Q = Q(p', 0'1 respectively. Upon putting k = -iy according to (2) and using (7) and (8) for the eigenvalues and the normalized eigenfunctions respectively, we obtain finally, for the first Green's function of for the sector- (3) ial domain: nircp sin 2 n-rr . a -2 (Tnir m/a) iTnTr a (9) ) a (TnTr +1 a a nir Oa) 3n Tr a , . a We will now show that the summation over m in the above double series can be carried out and thus reduce (9) to a single series. For this purpose we use the following Fourier -Bessel series expansion [2, vol. 2, p. 1043 00 Y-J ( T 2 )v, m Iv(x v+ 1 (T v, m )] v v, m 93 v(Tv, m X) (10) ()K ( x) [1 v v - K (0I v v ,x)]Ivq) 13 valid for [Again, 0_ x< X < 1 and arbitrary complex. t is the mth positive root of J (x) = 0, while Iv and Tv Kv are the modified Bessel and Hankel functions, respectively]. This summation formula is applied to the case v= ,ax = p/a, X = ya , and then G1 becomes G1 = [sin -2a t n1 (nin1)) a sin (n")') a (11) InTr. (yp) 7i- a_ In'rr (ya) under the condition that p< Inrr -ya)KnTr(yp p' a a ) _ - Imr(yp')Kmr(ya) a a . Since the right-hand side of (10) is symmetric in x and X it follows that then p and p' have to be interchanged in (11) when p> p'. The same analysis applies to the second Green's function G2 . In this case, for the system of normalized eigenfunctions, instead of (8) we have 14 U H4 = Un, 1 1 T a cos +1 (nTr95 ) a JnTr (Tnir m a a \ nTr a m; p/ a) (12) n= 0,1, 2, m = 1, 2, 3, ... In this case the eigenvalues are k= kn, m , Tnir n= 0, 1, 2, . (13) a m = 1, 2, 3, If we observe that sin sin [ 22Z (p] cos cos [ "95']=4 {cos{(-)]11-1-1a a 00, T cosr-LIT(0+990 a we can represent both Green' s functions of the two dimensional modified Helmholtz equation for the cylindrical sector of cross section D in the form ; cos 11(0-95')]/ Gl= -(2a )-11 E { cos[122-1.(95-091 a 2 n=0 (14) I nTr (yp) a I nTr a (ya) I nTr a ya)KnTr (yp' ) -InTr (yp' )K nTr (ya.) a a a 15 inter- valid when p < p' and the same expression with p and changed when p> p' are defined as The Neumann numbers E . E0 = 1, E The transition to a angle a. p' 2 for n=1, 2,3, ... co leads to the domain of a wedge of opening Since I (x)-4- co as x-.00 and Kv(x) 0 as x -.00 we obtain, from (14), for the two Green' s functions for a wedge-shaped domain oo G1 = 2 2a )1:En { cos[B-Lra (04' )] ; cos[(+ (0+ 0' )] } Inff(yp)K _a 0 (yp' ) n,n. a P < 13' which are the well-known expressions for the open wedge [13] , [14]. Since our main results will be concerned with half plane diffraction problems we specialize (15) for the case a 2 Tr to obtain 00 G1= 2 En cos[ i-1(4)-01)] m=0 cos[ (yp)Kn (ypi ), (0-0. )] } 2 7 P< and if p> p' , the same formula with p and p' interchanged. This last expression will be the starting point of our investigation. 16 CHAPTER 3 REPRESENTATION OF GREEN'S FUNCTIONS AS LAPLACE TRANSFORM INTEGRALS Our next goal is to express (15) as a Laplace transform integral with y as transformation parameter, i.e. G1 should be of 2 the form op F )e dt (17) 0 2 where F(t) contains, besides the integration variable t, the proper- ties a, , p, p' as parameters but independent of y. Of , course F(t) can vanish over one part of the interval of integration (0,00). For this purpose it is desirable to represent the term tnrr (yp) a Kmr (yp' ) a occurring in (16) in such a form. We use the representation IV(bs)Kv(as) = 1 2 + a+b (ab) 2 , 21a-b Pv-1( a 2ab° cos (Trv) TT for a > b, Re s > 0 g °3 a+b . Q v- 3. 7 ( 2_ 2 -st at + t2-a2-b2 -stdt} 2ab 17 P and Q are the Legendre functions of the first and second kinds, respectively ( see Appendix II). The formula (18) seems to be new and is derived in Appendix I. now replace I tyrr (yp)KnTr We (yp' ) in (16) by (18) while putting there a v= a s= y, b= p, a = p'. , Upon interchanging the order of summation and integration [ which is easily seen to be permissible because of the absolute convergence of both sum (16) and integral (18)] we encounter terms of the form: oo EP n=0 2 n a. a2 1 2 2 -t 2pp (P , ) v) cos (n" a (1 9) and Ec n0 s(.12-:-") Q a 222 nT 1 (t -P2pp-'13 2 cos (n"v) a (20) a The series (19) can be summed for an arbitrary opening angle a of the wedge. However, it does not seem possible to do the same for (20) if a is arbitrary. But for the case a = ZiT, i. e. if the wedge becomes a half plane, (20) can also be summed. Then Green' s functions --G-1 2 for the two dimensional modified Helmholtz equation 18 for the half plane can be given in the desired form of a Laplace transform integral of the form (16). In order not to overburden the text we will present this summation in Appendix I. From now on we will consider only the half plane problem. Then by (16), (18) oo EnIn ( yp) K (yp' )cos = oo 1 1 2 171 2 P Encos(111) Pn 2 (PP' )2 { +---2 2 P -P n=0 r + (P 1 co , 2 2 -u )1e-Ytdt + -' 2pp' 2 2 P -1 flE co s (nv )Q nt -2pp 2 )1e-Ytdt} 2 JP1+ The sums under the integral signs are those given in (2), (3), Appendix I. We consider at first the second term on the right-hand side of (21) and we obtain, by (3), Appendix I: 2 2 -1)nEcos(nv)Q n n--z1 -- 2 2 ,t -0 t 2 271 ('t p ' 2PP jP+PI = 71p'it 1 op r P+ (3' 2 12 -Papp' -P e-Ytdt = 2 + cos v) (22) e- Ytdt 1 2 (p + p' 2 - 2pp' cosv)j yt e-dt , 19 2 Since z = 2 t -P -P 2ppt 2 > 1 along the whole interval of integration from t = p + p' to t = co the condition for the validity of (3), Appendix I, is met. As for the first term on the right hand side of (21) we have at first, by (2), Appendix I op 2 E cos(2--.v)P n 2 1 2 3 2 )= 22[cosv-P + n (P +PI 2pp' n= 0 2PPI --1 2 2 2 2 -t (23) 1 = 4(pp9 2 (t -1) , 2 + 2ppl cosy) subject to the conditions in (2), Appendix I. Put 22 2 p +p' -t cos U = 2PP` then, since in the integration of (21) cos U varies from 1 to when t varies from p' -p to p' + p, U varies from 0 to -1 Tr. But the sum (2, Appendix I) vanishes when U <I vl [Here we write Iv I since v = ± 01] . Hence, if (23) is inserted into (21) the integra1 tion does not start with p'-p but with (p2+ p' 2-2pp' cosv)2 and ends with p' + p (corresponding to U = TT). If moreover v > Tr then the sum vanishes altogether in 0 < U < Tr and henceforth the integral. Since v = figuration 0 ± 0' = Tr. ± 0' we have to investigate for which con- 20 Let the source point Q be in the upper half plane (Because of symmetry this is no restriction), i.e. 0< < Tr and consider the cases as indicated in the Figures (3) and (4). + = Tr Q(131, 4:4) II E 0.211/401/.0r.11%."....1:0"...0505/110,..40MIOrooralP1111%.411r4r4/...010/ II (1) -(1)'= Tr Figure 3 Q'(p', -40 21 We form the image Q' with respect to the half plane or its extension and draw the straight lines QE and Q' E where E is the edge of the half plane. The domain 0 < < 2Tr is now subdivided into three domains I, II, III as indicated in the above figures. Region I is bounded by the upper boundary of the obstacle (the upper part of the half plane) and the boundary of geometric reflection., Region II is bounded by the boundary of geometric re- flection and the boundary of geometric shadow. Region III is bounded by the boundary of geometric 22 shadow and the boundary of the obstacle (lower part of the half plane) (shadow region). For a point of observation P(p, 0) lying in the respective regions: Region I + 0' < Tr , Region II 0 +9' > Tr , Region III 0 + 0' > ii, I H - 0' - 0' Tr , < Hence, by (23) 2 E cos[(-12-0: 01 )] n 2 ID1 --2 22 n( P +213'pp:t )1e-Ytdt= 2 (24) 1 = 4(pp' ) P+ [t2-[p2+p' 2-2pp' cos(0-F 0' )1} ae-Ytdt 1 [p2+ pi 2-2pp' cos(0-F in the regions I and II for the upper sign and in region I for the lower sign, while (24) is zero in region III for the upper sign and in regions II and III for the lower sign. Denote by R and R' the distance of the point of observation P from the source point Q and its image point Q' then: , respectively, 23 1 R = 15.0 = [p2 p' 2 - 2pp' cos(-' )} 1 PIT= [P2+p 2 - 2pp' cos(0+95' (25) di P(P,14)) Figure 5 One then obtains, from (16), (21), (22), (24), for the two Green's functions of the two dimensional modified Helmholtz equation; for a point of observation P lying respectively in Region I G1 =(t P+ P 1 2' 1 2 -yt dt P+PI, 2 )2-e -yt dt] + R' R 2 2 (26a) 2 1 - 4,7 P+ -2- -yt e - 2 oo dt -F (t--R' 2) 2e-Ytdt] Pi 24 Region II [ C 2-dt 1 GI =- P+13t2 (t 2-71 2 - 1 -Nt co 1 +1 ) 2 2 (t P-FPI (26b) Region III 1 G1 00 = - -477-1 2 oo (t -R' dt T P+ PI 2 2 ) 2e -yt dti ,t26c) i3+ This is the desired form (17). Note that the integrand is elementary. We write 1 1 P+PI 1 co ( t 2 2e-Ytd r- (t2 -R t2-R = `JR -Ytdt p+pt and observe that 2, vol. 2, p. 82]. 1 (t2- 2 2e- yt dt = K(yR) Then G1 can be written in the form 2 Region 1 1 1- -Tfr1 2 0 (yR) T Ko(yR')] + 4Tr (c) (t2-R P+PI e - dt-F (t -R'2 ) 2e-ytdtj 'N-P' cc° 2yt (27a) 25 Region II 1 sx (t2 -R2) Gi=---K (yR)-F[ 0 2 27 p+ p, LITr 1 1 e -ytdt- (t P+Pt 2 2e-Ytdt] (27b) Region III 1 GI= 2 r°° 2 4Tr 2 (t -R ) faf 170 1 e _vt 00 7 - dt] (27c) p+pI These equations admit a physical interpretation. Region I The field consists of the result of two line sources, one located at Q(original source) and one located at its image point Q' plus an additional term which we shall call the "diffraction" term. Region II The field consists of the result of a single line source located at Q(the original source) plus a " diffraction" term. Region III The field consists of a " diffraction" term only. This " diffraction' term is equal to: 26 1 1 oo DI - 4-rr j (t2-R2) (t2-R' dt] P+Pt P+131 _1 1 oo (t2-R2)2e- I 2, 2 dt] jP+ 1 oo III = - D (t2-R2) 4Tr e P+ PI for the regions I, II, III, respectively. Time Harmonic Case In order to obtain the corresponding expressions for the time harmonic case (Helmholtz's equation) we have, according to (2) to replace y by ik. Since K0 (iz) = - 1 2 im-F1 (2) 0 (z) it follows from (27a, b, c) that Region I 1 (2) ( G1 --7.0(kR)-f.1-1 0 2 1 )(kR)] + ---[ 4Tr c (t - * OP' 2 1 2-ild ST2 2,-2 e d -t t -R1 ) e P+Pi t i dtj 27 Region II G= 1 4 H (2) 0 -1 TT )+ -4 1 1 ___ 00 (t2 -R2) 2e-iktdt 2 P+ PI (t2 - 2 e -ikt dt.1,(28b) P+ Pt Region III 1 1 C 7t2..R S12_012-e-iktdt P+P' (3+ 13' e -ikt dti (28c) 28 CHAPTER 4 TRANSIENT SOLUTIONS We will now consider the same configuration as before but under the assumption that a " Dirac" pulse is applied, starting at t = 0, to the line source. The field produced by such a " Dirac" pulse will be labeled Then D(t). . D(t) =" -1 {Gi} (29) 2 i. e.D(t) is the inverse Laplace transform of the Green's function for the modified Helmholtz equation with respect to y as inversion parameter [14]. One first has: 1 t>R -1 y , {K0(yR)} = 0 , t<R The remaining terms in (27a) to (27c) are expressed as Laplace transforms whose inversions are the function F(t) itself. Then one immediately finds for a point of observation P lying in the respective regions: Region I 1 (t) = -[(t ZIT 1 . - 2 1 -f(t2 -R' 2)2]Li-4Tr1 [(t 2 - 7. 1 ..._ 2 1 2 -T. (t -R' 2 -21 ) 1(30a) 29 Region II 1 1 (t) = --(t D 2 - 2 -2- + 1 4Tr 2 2 (t -R ) 1 (t2 -R' 2 ) ] (30b) Region III 1 (t2-R' 2) 2] (30c) The last two terms in (30a) to (30c) are equal to zero when t< p+ p' while the others are zero for t<R and t<R' , respectively. Similarly as before one can interpret (30a) to (30c) as: Region I: The field consists of the result of a " Dirac" pulse due to two line sources, one located at Q and the other at its image point plus a ' diffraction" term. Region II: The field consists of the result of a " Dirac" pulse due to line source at Q plus a "diffraction" term. Region III: The field consists of a "diffraction" term only and moreover this term is zero when t< p + p' 30 Figure 6 The formula (29) is valid under the assumption that the velocity of propogation is unity. Hence it stands to reason that t<R in the regions I, II and /.D(t) = 0 and slD(t) = 0 D(t) = 0 when when t < p- F for region III(geometric shadow). We are now in a position to deter- mine the field 1(t) due to an arbitrary time dependent excitation function g(t) of the line source at Q [14] , under the assumption that g(t) = 0 for t< 0 . (t-T)g(T)dT = (t) = 0 where (x) g(t- (x)dx (31) -oo is the "Dirac' pulse solution. Then from (30a) to (30c) and (31) we obtain: 31 for a point P lying in Region I 1t 1 2 - -R2) 2 g(t- 1 txx)dx 2-RT 2 --2 , gtt-x)dx + R,' (32a) 1 g(t-x)dx 4: - 4Tr (x2-R' 2)g(t-x)clx] P+ P' P+13' Region II 1 - = t 2 2g(t-x)dx (x2-R2) 2g(t-x)dx + (32b) (A (x2p+pI , 2 j )g(t- dx Region III 1 1,(t) = - 4.Tr [C j (x2P+ 2 2g(t- 1 (ix 5t(x2-R' 2)-70-x)dx] P+13' Integrals whose upper limit t is smaller than the lower limit R, R' , p+p' , respectively have to be set equal to zero. (32c 32 CHAPTER 5 SOME SPECIAL CASES As an example for (32a) to (32c) we choose g(t) ei(A)t , 0< t< T 0 , t< 0, t>T = This is a time harmonic cylindrical pulse starting at t duration T. The special cases 0 and of (,) = 0 (rectangular pulse) and T = 00 (time harmonic pulse of infinite duration starting at t = 0) are treated separately. The integrals occurring in (32a) to (32c) are of the form 1 g(t-x)dx Since g(t-x) = e e =0 for t-x> T two cases have to be distinguished: t-c<T t-c>T The lower limit of integration is therefore t-T in case a and c in 33 case b). We then obtain Region I _1 (t)et= C (x _ 1 2 2e -i(4 x dX 2Tr -R' 2 2 -icox e ) dx 1 (33a) 1 ryt 2 2 --2ei (x -R ) -o.)xdx 1 471. 1 c(x2t3 t3 2)2 e-iwx dxi Region II 1 Vt)e-icI3t = - Ct 2_ 2 Trj 2e -icax + clx-- (33b) 1 1 2e -icoxdx C (x2-R, 2) [st(3( 3 e -icox dxj 3 Region III 1 .1)(t)e -iwt= 1 ryt 4Tr 1 t3 2 -R2 ) Ze1Xdx4-- 1 t s (x-2 _R, 3 2) 2e-1(4xdx (33c) 34 The limits are: t1 = R if t-R<T and t1 = t-T when t-R>T, t2 = R' if t-Rf<T and t2 = t-T when t-Rf>T, t3 = p+ p' when t-(p+ pf ) <T and t = t-T when t- (0-11)>T. The following terms have to be omitted: The first term in (33a), (33b) when t <R, the second term in (33a) when t<Rf and the last two throughout when t<p+pl . We specialize (33a) to (33c) now to the case co = 0. This corresponds to a rectangular pulse of strength unity and duration T. In this case the above integrals become elementary since 1 g(x2-a") --2dx = in x +4x2-a2 a and we obtain Region I (t) = - in.rt+qtz- Rz Lti+Niti 1 in -R2 t2 t+ Nit2-R2 t +Nit 2-R2 3 t+Nit2-R 2 3 +Nit2 2 - R' 35 Region II - 1 2Tr t+4t2-R in t3 +Nit 2-R2' 3 in t + 4t2-R' 2 2-R, 2 3 Region III 41,(t) t+4t2-R2 1 4Tr t +Nit 2-R2 3 with in t +4t2-RI 2 1 t3+Nit3 2-R' 2 3 t1' t'2t3' as given before. Certain terms have to be replaced by zero under the following conditions: The first term in (34a) and (34b), when t<R, the second term in (34a) when t<R' when t< p-f pl , and the last two terms throughout . We conclude our investigation with the case T = oo, the incident cylindrical pulse is a time harmonic one, starting at t = 0. Here only the case b) applies, with the limits t1, t2, t3, in (33a) to (33c) being t1 = R' t2= R'' t3 = p+ p' . 36 Of particular interest is the case t> p+p' i.e. none of the terms in (33a) to (33c) vanishes. We then write .5' (x2 -R 2 -iwx2 dx= tx -R2 -iwxc°2 dx = (x -R ) 2 2 e e 2 e dx - t1 (35) 1 - 1 2 ( S oox -R 2) 2 e-iwx dx = --rr i 2 (2 )( 0 1 t - s oo 2 e -iwxdx t using for the first integral the well-known integral expression for the second Hankel function [2, vol. 2, p. 21]. The same expression holds for R and R' interchanged. Similarly t 1 2 .11(x -R t e -iwxdx = t3 -1 2 (x P+Pt 2 2e -I wx --2 -ix e dx 1 oo ., . =S(x"-R P-t-Pi oo 56(x2-R 2e-iwx dx . P+13' Again the same expression holds when R and R' are interchanged. We insert these results into (33a) to (33c) and group time independent terms together and time dependent terms together. The field appears then in the form: 1)(t) -ickt -- G1 + Y (t). 2 .I(t) 37 Here G1 represents Green's function for the time harmonic case (28a) to (28c) (Note, that since the velocity of propogation is assumed to be unity, o.) and k are equal). Obviously YT(t) represents the transient field, i. e. the deviation of the field produced by a time harmonic excitation starting at t = 0 from the quasistationary case e. t> when the time harmonic excitation started at t = oc)) provided p' . We find for the transient field ,c-icox 1 oo Y (t) = -14Tr . oo 2 e-1()x2 x -11 )2 dx c Se (38) t for all three regions I, II, III. Obviously (38) tends to zero as t oo, i.e. the transient field becomes, of course, more and more negligible with increasing time. 38 BIBLIOGRAPHY 1 . Clemmow, P. C. A note of the diffraction of a cylindrical wave by a perfectly conducting half plane. Quarterly Journal of Mechanics and Applied Mathematics 3:377-384. 1950. Erdelyi, A. Higher transcendental functions. Vol. 2. New York, McGraw-Hill, 1953. 396 p. Friedlander, F. G. The diffraction of sound pulses. Proceedings of the Royal Society, London A186:322-351. . 1946. On the half plane diffraction problem. Quar- terly Journal of Mechanics and Applied Mathematics 4:344357. 1951. . Sound pulses. sity Press, 1958. 202 p. Cambridge, Cambridge Univer- Gast, R. C. Half plane diffraction with line source excitation. Pittsburgh, Carnegie Institute of Technology Department of Mathematics, 1956. 49 p. (Technical report No. 24) Harrington, R. F. Current element near the edge of a conducting half plane. Journal of Applied Physics 24:547-550. 1953. Kay, I. The diffraction of an arbitrary pulse by a wedge. Communications on Pure and Applied Mathematics 6: 41 9-434. 1 953. Keller, J. B. and A. Blank. Diffraction and reflection of pulses by wedges and corners. Communications on Pure and Applied Mathematics 4:75-94. 1951. Kontorovich, M. J and N. N. Lebedev. On a method of solution of some problems in diffraction theory. Journal of Physics, Moscow 1:229-241. 1939. Miles, J. W. On the diffraction of an acoustic pulse by a wedge. Proceedings of the Royal Society, London A212:543-547. 1 952. 39 . On the diffraction of an electromagnetic pulse by a wedge. Proceedings of the Royal Society, London A212:547-551. 1952. Oberhettinger, F. Diffraction of waves by a wedge. Communications on Pure and Applied Mathematics 7:551-563. 1 95 4. On the diffraction of waves and pulses by wedges and corners. Journal of Research of the National Bureau of Standards 61:343-365. 1958. . Tabellen zur Fourier Transformation. Berlin, Springer, 1957. 213 p. . Sommerfeld, A. Partial differential equations in physics. New York, Academic Press, 1964. 335 p. Turner, R. D. The diffraction of a cylindrical pulse by a half plane. Quarterly of Applied Mathematics 14:63-75. 1956. Wait, J. Diffraction of a spherical wave pulse by a half screen. Canadian Journal of Physics 35:693-696. 1957. Wells, C. P. and A. Leitner. A Lebedev transform and the "baffle" problem. Quarterly of Applied Mathematics 15: 430- 434. 1958. APPENDI C ES 40 APPENDIX I Some Integral and Sum Formulae Summation of series involving Legendre functions 00 EnP n=0 1 niT 2 a , (cos u)cos(---v) = a 1 2 1 Tt a cosv-cosu -u<v<u, u<v<2a -u 0 <u< Tr (E is Neu.mann's number, E =1 E =2 n> 1). ' n ' _ The formula (1) is a Fourier series with respect to v with period v = 2a. To prove (1) we expand the function I f(v) = J 1 (cosv-cosu)-1- , -u<v<u , otherwise into a Fourier series of period 2a. Since this is an even function of v we have, by Fourier's theorem: Ancos( nTrv a f(v) = n=0 41 with An = En a dv 2a,j -af(v)cos (117v) therefore for a > u 1 n C cosv-cosu) j 2 v cos()dv a This integral is known [2, vol. 1, p. 159. formula 271. 1 An =Ena -1 271. mr --+ 2 a 1 (cosu) This proves (1). The special case a = 27r yields oo EP -1 n n=0 1 --+ 2 n (co su) co s, nv t 2 = 2 3 1 2, lcosv - cosu) 2 (2) u --u<v v< 47r -u 0<u<7r nE2. , nQ 1 --2+ (z)cos(nv)= 2 n [2, vol.1, p. 166] . 1 , ir tz+cosv) 2 , z> 1 (3) 42 Representation of the Product of Two Modified Bessel Functions as a Laplace Transform We prove the formulae: I(bs)K(as) = V v ya+b a-b (ab) 2 1 x- ( a2+b2-t2 )e -st dt + 2ab v- -2- (4) '0° +27 -1cos(Tr a+b 0 ( 1 t2-a2-b2 ) e -St dt 2ab v---2- a> b. --2 yco (ab) 1 KV(as)K v(bs) = If (4) is established, K (X) V - (5) 1 2 p72 a+b (t 1 v--2- 2 a2-b 2 -stdt 2ab . (5) follows by the relation II 2sin(Tr -v (X) - I (x)1 (Note that K -v(x) = K (x) ) In order to prove (4) we start with the well-known formula [2, vol. 2, p. 96] I (bs)K as) = v(v2 +s2 0 a.b, Re v > -1 J V(av)J V(bv)dv 43 In the integrand above we substitute v(v2-Fs2)-1 Soo-st sin.vtdt e = 0 and obtain, upon interchanging the order of integration I V(bs)K v(as) =st - 0 0 J v(av)J v(bv)sin(vt)dv] dt. The inner integral is known [15, p. 169] icc° 0 J v(av)J v(bv)sin(vt)dv = a-b 0, 1 V-7 a2-Fb2-t2 2ab , a-b (*. t 1 2co s (Try)Q This proves (4). it 1 2-a2_ 2ab a-Fb , 2 t> a. 44 APPENDIX II List of Notations 00_on J (z) = (i)v+2n n! r(v+n+1) n= Y (z) = (sinIT v)-1[J (z)cos(Tr v) - (z)] -v H(z) = J v(z)i Y (z) v Be ssel function, Neumann function and Hankel function, respectively. iTr iTT v I( )=e 2 J ze 2 v+ 2n oo ) n! r(v+n+1) n=0 [sin ( IT 0].1- v( ) ( modified Bessel function and modified Hankel function, respectively. iTr v iTr K (ze ) = - 1 2 K (z) = K-v(z) (see[2, vol. 2, Ch. 7] ) iTr e 2 (2)( 45 P (z) crP v(z) Q (z) (see [2, vol. 1, Ch. 3] Legendre function of the first kind -1<x< 1 . Legendre function of the first kind, z not on the real axis between 1 and - oo Legendre function of the second kind, z not on the real axis between 1 and -oo ). .