WGWEERWG i.it..hr,i «i.*s COria ANTENNA ^ It G55-te BORATORY Technical Report No. 37 ON THE SOLUTION OF A CLASS OF WIENER-HOPF INTEGRAL EQUATIONS IN FINITE AND INFINITE RANGES CONPENCE ROOM *»v:.V .**** Raj Mittra 15 May 1959 Contract No. AF33(616)-6079 Project No. 9-113-6278) Task 40572 Sponsored by: WRIGHT AIR DEVELOPMENT CENTER ELECTRICAL ENGINEERING RESEARCH LABORATORY ENGINEERING EXPERIMENT STATION UNIVERSITY OF ILLINOIS URBANA, ILLINOIS ANTENNA LABORATORY Technical Report No, 37 ON THE SOULUTION OF A CLASS OF WIENER-HOPF INTEGRAL EQUATIONS IN FINITE AND INFINITE RANGES by Raj Mittra 15 May 1959 Contract No. AF33 l616) -6079 Project No, 9- (13 -6278) Task 40572 Sponsored by : WRIGHT AIR DEVELOPMENT CENTER Electrical Engineering Research Laboratory Engineering Experiment Station University of Illinois I rbana , Illinois A <^ CONTENTS Page L„ Introduction 1 2„ Solution of the Semi-Infinite Range Integral Equation 2 2„1 Derivation of the Simultaneous Set of Equations 2 2 2 Solution of the Set of Equations for the Unknown Coefficient A. 5 on 2„3 Extension to the Case When the Difference of Order of F (p ) and F _(p^) is more than one, 7 3 The Integral Equation with a Finite Range 10 3„1 Solution of the Simultaneous Set of Equations Obtained for the Finite Range Case 12 3 P 2 Inversion of the Matrix of the Equation (34a) 13 4, Calculation of the Determinant and the Cof actors of the Determinant of the Matrix M 16 5 The Fourier Transform Method of Formulation 20 5„1 Fourier Transform Formulation 20 5„2 Development of a Set of Simultaneous Equations 26 5„3 Iterative Method of Solution 30 6. The Eigenvalue Problem 33 6„1 The Eigenvalue Matrix Equation 33 6,2 Upper Limit Infinity 37 7„ Numerical Example 40 8„ Applications 42 9„ Conclusions 43 References 44 1. INTRODUCTION The subject of investigation in this report is the Wiener-Hopf Integral Equation g(t) o for a T either finite or infinite. It is assumed that Kit! admits of an expression of the form -k It I K Itl S C e r I I m The method of solution for the semi-infinite range (T -oo) i s discussed first. The method is then extended to T finite. Two basically different approaches are used. The first one leads to a set of simultaneous equations which can be solved by analytic means. The second one uses the technique of Fourier transforms in a complex plane and is more suitable for an iterative method of solution for the finite T case (for T ^&o f the Wiener-Hopf technique yields the answer directly) The eigenvalue problem corresponding to the above kernel is also discussed. Suitable numerical examples are included and certain applications in the fields of Information Theory and Electromagnetic Wave propagation are also pointed out. SOLUTION OF THE SEMI -INFINITE RANGE INTEGRAL EQUATION Consider the integral equation f (t) K | t-T| dT t > (1) where K|t-T| is of the type n -k ! t-T| K|t-T|= 2 C e r , Re(k )>0 (2) r r C = constant coefficients r Equation (1) is the well known Wiener-Hopf Integral Equation and a solution of the equation can be obtained for a general K(t) by the use of the method of Integral Transforms in the complex plane . For the particular type of Kernel which appears in (2), however, the equation can be solved directly by using the theory of residues of a complex variable. The first step for solving the problem is similar to the one used by 2 Laning and Battin , A set of simultaneous equations are obtained when their method is used and these equations are solved here using the technique of residues. 2.1 Deri vation of the Simultaneous Set of Equations The function g(t) can be expanded in a series of exponentials -i t g(t) = Z D e r , t>0 r Hence the problem in Equation (1) can be reduced to that of solving for -it a g(t) of the form e and superposing the contribution due to the differ- ent terms in the series. An alternative approach is the following. Operate on both sides of (1) with the operator L, where l - n ,p 2 - kr 2 ) , ,4 r=i Digitized by the Internet Archive in 2013 http://archive.org/details/onsolutionofclas37mitt Then, using 2 a "kit" 1 "! (p -k ) e = -2k 6(t-T), (6(t) = Dirac delta) T r one obtains, n 2 2 n n (n ^ 2 2 L g(t) = n (p -k ) g(t) = -2 2 k C n (p -k )f(t) (3) i S n r r i m g= 1 50= 1 m= 1 where the superscript (n) in the continued product implies that the factor corresponding to m = n is to be omitted. Equation (3) is the differential equation satisfied by f(t)„ The equation is clearly inhomogeneous and can be written in the form F. (P 2 ) -f(t) = g(t) < t < ~> (4) F 2 (P ) 2 J 2 It can be easily shown that F (-p )/F r (-p ) is the Fourier transform of the kernel function K(t) , i.e. ,°o . 2 F. <-p ) K| t | e' ipt dt^ J -^- F 2 (-p ) Returning to C4) , one knows that (4) has a particular solution say f (t) , * ' P and a general solution of the integral equation can be written in the form n-1 -a . t f(t) = f (t) + A 6ft) + S A. e 1 . t > (5) p o 1 v i=l i 2 where a. s are the roots of F (p ). From (3) one sees that the order 1 2 2 of the polynomial F m p is (n-1), so F (p ) has symmetrical roots, p = «- a , + a + .... + a -1 — 1 — 2 — n * For the case, the degree of F (p ) is one less than that of F„(p ) „ When the difference in the degrees is more than one, the method discussed in the next section is applicable Only half of these roots, however, the ones with the negative sign, will give solutions with exponential decay as t — » + oo„ The others, with a positive real part in the exponent are not admissible as solutions. Also it can be shown (see for instance Reference 2, p. 416, that the solution of the integral equation must contain the Dirac-6 included in (5). Let g (t), o < t < oo , be defined as g 1 (t) = [ f (T) K(it-Ti) dr n-1 ~ aT ] f(T) - A 6(T) - S A. e * > K(lt-T| ) dT (6) i=l X J using (1) and (2), (6) can be written as g x (t) - g(t) r -a t -k |t-T| (2 A. e )(SC e V )dT i l r r "o -k t -ASCe r t > (7) o r Upon evaluation of the integral in the r h,s. of (7), the following is obtained, -k.t g. (t) - gft) = S D. e 1 i i where D s can be calculated in terms of A , A. ' and the C s i o i s r -at The coefficients of e can be shown to be equal to zero since I . 2 a. s are the roots of F, (p ). Hence the right hand side contains terms -k t involving the exponentials e only The left hand side of (8) is known if the particular solution f (t) P and hence g it) is known From (8), remembering that the coefficients of -a , t e are zero, it follows that -k.t g.lt) -g(t)= SC, P. e X (9) x i l. l where P. s can be obtained from the knowledge of g (t). 1 -k t r Equating the coefficients of e the following set of n - equations are obtained for the n-unknowns A , A, ... A 1 n-1, n-1 A . A o 4- 2 — 4r- = P . r = 1, 2, „„.n (10) , a -k r J-l J r where P s in Equation (10) are considered as known, r The solution of (10) will now be derived using the theory of residues, 2 2 Solution of the Set of Equation for the Unknown Coefficients A ^ Consider first of all the solution of a set of equations n-1 A (m) ^ + * ^T-IT - P 6 ". m = l, 2, ...n (11) . a .-k m r J=l J r If A and A. s can be solved for all m's, then A , A , ... etc are obtained by superposition, i.e., using A Q = S A (m) , A - S A. (m) (12) m=l J m=l J Attention will therefore be devoted presently to the solution of (11) For this purpose consider a contour integral of the form, where Qtz) satisfies the following conditions, 1 lim Q(z) = Q (constant) z — >°° o 2 Q(z) has (n-1) poles atz-a . a n . . . a 1 2 n-1 3. Q(z) has (n-1) zeros at z = k , r = 1, 2, m-1 , m+l,--n, excluding n - m Such a choice of Q(z) is readily found as, n tm) Q(z) = Q H (z-k ) o r r=^l ; (Q = constant) (13) n-1 o n (z-a ), r r-1 Now let the contour in the integral Q(z) z-k dz be chosen large enough to include all the singularities of Q(z). If one lets the contour C recede to infinity, say by taking an infinitely large circle for C, one has: n-1 Q( . q ^2 dz - 2 7T jQ = 2ff j 2 ^L (z-a ) o z-k o . z-k p r p=l r 2 7T JQ(z) zaa (14) z=k Since Q(z) has zeros for z = k , r ^ m, (14) can be written as Q(z) n-1 Q(z)(z-a ) q - 2 P o . (z-k ) p=l r zsa n z=k m Using the expression of Q(z) given in (13) in the Equation (15), the following results, (15) n (m) ( k -k ) 1 n r=l m r = q < — — ->6 a -k P r v. n-l n (k-a ) , m r (16) Comparing (16) with (11), it is seen that if one puts the following relations in the equation (II), n (m) (k -k > Ifc.1 m r o A n-1 n (k -a ) V s ! Q m = A (m) (17a) (17b) -Q n (B) (a -k ) D.1 P - r n tp) (a =a ) 1*1 P r = A< m > (17c) The set of 'n 1 equations are identically satisfied. Q in (17) can be O , X calculated from the knowledge of P . The unknowns in (11), viz A , . m o A .... A etc., are then obtained using (17b) and (17c). 1 n Since the a s and k s are known, the solution of (11) is complete p r Also, as was pointed out earlier, the solution of (10) can be constructed from the solutions of (11) with the use of (12). The solution of the Integral Equation (1) is then obtained by substituting the coefficients Ai , and A in (5) „ 2 2 2.3 Extension to the Case When the Difference of Order of F (p, ) and F„ (p ) is more than one In the previous section the derivation of a simultaneous set of equaf tions for A , A n ... etc was based on the assumption, ;that the order of ° 1 2 2 the polynomial F (p ) is one less than that of F (p ). In this section this condition will be relaxed and the process for formulating a new set of equations will be outlined. The method of solution will also be presented, 2 2 Let, ' g 1 and } n' be the order of F (p ) and F (p ) respectively where J. ^j g < n. As has been pointed out by Laning and Battin (Ref 2.2 p„ 416), the solution f(t) will now contain the derivatives of the Dirac Delta and f(t) will be of the form s_1 ( . q f(t) = A 6(t) + 2 (-1) X A b{\\ + 2 A e^ 1 * t ? (18) 1=1 1=1 S - n-q > 1 , 6 = i-th derivative of 6 Following exactly the same method as was used for the derivation of (10) one now obtains the set of equations, / s-1 q A . A f 2 A.k 1 + 2 — = P (19) o . . i r . , a -k r ' 1=1 i=l i r where P ? s are the same as in (10) r Again consider solving the equation similar to (11), viz , (m) + S 2 A (m) k ± + o . , i r i=l * Cm) q a. z x a.-k 1=1 1 r P 6 m r m (20) by using the method of residues Let M(z) be a function of the complex variable z, which satisfies 1. lim < M(z) - 2 l.z 1 z-* - 3 i=l 1 (constant) 2„ M(z) has q poles at z = a , i = 1, 2, ., „, q 3. M(z) has (n-1) zeros at z = k , r = 1, 2, „ , n, r/m A choice of M(z) which staisfies the above conditions is n (m) f ,- M(z) ■ M i=l (z-k.) l o q n (z-a ) i=l i (21) One can find the constants L and L in terms of M , k and a. from i o o r i the equation (21), Now consider the integral f mi/ ^ s_1 L.Z c r i=l r where C is an infinitely large circle. Since the integrand behaves as L /z as z-^oo, one has o' (22) j J 2 Mp. (z _a . ) o i. , z-k l 1=1 r Equation (23) can be rewritten as s-1 - 2 L.k 1 + M(z) z=a i=l l r z=k Y = 1, 2, ... n s-1 f L f 2 L.k - 2 - o i r i=l l-l M(z) (z-a, ) z=a a,-k l r = M(z) Dl z=k ra (23) (24) Hence, comparing the r h.s of (24) and (20) and using the expression for M(z) in (21), one sets M(z) z=k = P , i.e. m m M =P o m n (k -a.) i=l m X n (m !k -k.) i=i m x (25) As was explained above the constants L , L , once M is known, o Again comparing (24) with (20) one has, L , ,. are determined (s-1) L =A (m) , L. =A, fm) , i = 1, 2, ..., (s-1) o o 11 -M(z)(z-a ) r s M z=a / (m) r = 1, 2, (26) The expression (26) yields all the unknowns in (20). The solution of (19) is obtained by using superposition. The method of solution has thus been extended in this section for the case when the degree of the 2 2 polynomial in F (p ) differs by more than one from the degree of F (p ) , 2 / , 2 Recall here that F (,-p )/F (-p ) is the Fourier Transform of the kernel 1 ^ K |t I . 10 3. THE INTEGRAL EQUATION WITH A FINITE RANGE The solution of the integral equation, .T g(t) = f(T) Kl t-TjdT (27) which is similar to (1) except for the upper limit, will be detailed in this section, K|tj is again assumed to be expressible as a series of -k. I t| exponentials e i as in (2). The method formulating the problem in terms of a set of simultaneous equations, is similar to the one used in the previous section. The solution of these equations, however, is considerably more involved. The Differential Equation for f (t) and the particular solution fp(t) are the same as obtained above in (24), if of course g(t) and K(t) are the same. The range however is different, i.e. F (p ) — — f (t) = g(t) < t < t (28) F 2 (P ) Hence a general solution of (28) is of the form n-1 -a.t n-1 a.(t-T) f(t) = f (t) + 2 A. e 1 + £ B. e * (29) P i=l X i=l X where a . s are the same as defined in the previous section, l 2 2 It is assumed that the order of F (p ) is one less than that of F (p ) -L £ The method can be extended for the case when this difference is more than one, as was discussed before. It can also be shown that the solution of the Integral Equation must contain Dirac - 6's i.e. the complete, expression for f(t) which satisfies (27) is, n-1 -a.t a (t-T) f(t) = f (t) + A 6(t) + B 6 (t-T) + 2 A.e 1 + 2 Be 1 (30) P o o i i i=l i=l 11 The problem in hand is to find the 2n coefficients, A , B , A s and . o o i B, s. l One again defines a g (t) as g 1 (t) = f (t) K|t-r|dT — 'o and it is possible to verify, by following the lines used in deriving (29) that n -k t n k (t-T) g x (t) - g(t) = S C r P r e r + 2 C r R r e r (31) R=l r=l Substituting (30) in (27) one obtains after carrying out the integration (for details see Laning, ibid. p. 320) r C } r fn-1 -a.T n-1 a. (t-T) if(T)-f (t) I K|t-T|dT = J 2 A.e + 2 B.e 1 ^K|t-T|dT _a .t / n-1 A. B.e v -k t = S f A + 2 ~. 2 -~ — z ) C e r (32) \ o , a .-k . a.+k / r r\ 1=1 l r l ir/ -a T / B, A.e 1 > k (t-T) + 2 B +2 ~i— - 2 n X . . C e r \ o „ Q . -k a.+k r r \ iir i ir/ Since (31) and (32) should be identical for all values of (t), one -k t gets the following set of equations by equating the coefficients of e r , k (t-T). and e r _q t n-1 A. n-1 B.e A + 2 n 1 . - S ~ — r— = P (33a) o J , a . -k a.+k r i=l l r i=l l r -a T n-1 B„ n-1 A.e 1 B + 2 — 2 -| ■ = Q (33b) i=l i r i=l i r r = 1, 2,. . . n UWVEKSITY Of nUNOli UiftMtf 12 (33a) and (33b) are the desired set ( of 2n equations for the unknowns A , A. , B , B . o 1 o i It is to be noted that (33) reduces to the form (8 0) when T— ><=*». 3.1 Solution of the Simultaneous Set of Equations Obtained for the Finite Range Case The solution of the set of 2n equations for A , A , B , and B., are o i o i discussed in this section. As a first step to solving (33), form a new set of equations by adding and subtracting (33a) and (33b) and obtaining F o + 2 F J -a T 1 l e a.+k l r )■ P r + Q r a -k l r -a T 1 e x N )■ P r - Q r a.-k a +k J + S J ; — = P - Q (34b) o i I a -k a +k / r r i \ l r i r / where F - A + B , J = A - B o o o o o o F . = A . + B . , J = A -B ill l i l The advantage of using (34a) and (34b) in preference to (33a) and (33b) becomes immediately clear if one notices that the set (34a) and (34b) are uncoupled. This is because the matrix corresponding to (34a) is the same as that of (34b). Hence, it is only necessary to invert the (nxn) matrix corresponding to one set say (34a), rather than the 2n x 2n) matrix of (33a) and (33b) „ The inversion of the matrix (34a) is not straight forward, however, since the residue method is not directly applicable. It is nevertheless possible to construct the elements of the inverted matrix for (34a) , when the cof actors and the determinant of the matrix for the set (10) is known. This is detailed in the following section. 13 3.2 Inversion of the Matrix of the Equation (34a) The matrix \ M 1 for the equation (34a) is, [ M i = 1, 1 e -c^T V k i VV' lV k i W 1 e -a T 2 -a T-, n-1 \ (— — - ' ) r"'v.-r k i °n-i-v -a 1, 1 e IT 1 e -a T r V k 2 I V k 2 «V k 2> ! a n" k 2 1, 1 e -a t -a T n-1 (a -k ) a +k In In a -k a + k n-1 n n-1 n (35) The matrix corresponding to (10), say Tm] is [„]. a -k 1 1 1. T a -k 1 r a -k r r a -k r r n-1 1 1, a -k 1 r a -k n-1 n (36) Let the cofactors of the determinant of the matrix [Ml be designated by the symbol A pq Then A = A (a , a . .a a , . . .a k n , k , . . .k . , k . , . . .k ) (37) pq pq 1 2 q-2 q n-1 1 2' p-1' p+1 n 14 i.e. the expression for A involves all a. . ' s and k 's excluding o. pq 1 r q-1 and k and this can be easily verified from the form of (36) . P The determinant A for Tm1 however involves all the a- .' s and k ' s. Suppose now A is known, then, A corresponding to TlVI'l can be pq pq L J expressed in terms A , as follows, pq n-1 -a. T / „ l. A = A + S e 1 , A (-a, ) pq pq =1 ' T>Q ij n-2 n-1 -(a. +a . ) T + S 7 s' e l l X 2 A (-a. , -a. ) h-vv 1 pq ^ l2 T n-3 n-2 n-1 -(a. +a , +a ) + Z S 7 2/ e x l x 2 x 3 A (-a -a , -a . ) H' 1 W 1 W 1 pq ij l2 l3 -(a +a +. . . a ) + e X l X 2 1 n-l A (-a -a. , -a i ) pq 1 i 2 n-1 (38) where the prime on the summation symbol implies that the term corresponding to (q-1) is to be omitted. Also, A (- a . , - a ) implies that the signs of the particular set of pq i 1 i 2 a' s appearing in the argument are to be changed in the expression for A keeping the signs of the other a's unchanged. As an example, if then A (-a ) = (k -k ) (k -k ) (-a -a ) (a -a ) pq 1 2 3 12 12 2 3' and A P q ( - Q l'-V - (k 2" 1 V (k r k 2 )( - Q r a 2 )(a 2 +a 3 ) and so on. 15 Equation (38) is a rather long and involved one, but it is a straight- forward process to show its validity. Furthermore, in a practical problem it is only necessary to compute a few of the terms in the series (38). This is because the higher order terms have an exponential decay with the order. Equation (38) provides a means for obtaining the cofactors of the matrix [M' ] when the cofactor of the matrix TmI is known. The cofactors of [m] can be obtained by a method outlined in the next section. In order to find the inverse of the matrix [M' 1 , one still needs the expression for the determinant A" of [m"] in addition to the expression for the cofactors. The expression is very similar to that of A' and is pq given in the following. -a T n-1 i A - A + 2 e A (-a . ) i 1= l n-2 n-1 -T + S S e A(-a. , -a . ) i 1=1 1 2 =1 1+ 1 1 2 -fa +a +. . .+a )T + e A (-a. -a, , , . .a ) (39) V X 2 Vl where the symbols have the same meaning as used in connection with (38). Using (38) and (39) it is possible to construct the element N pq which is p, qth element of the inverse of Tm' 1 by using the well known expression n = -H3. qp A' The problem of inverting the matrix [M* ] and solving the set of Equations (34a) and (34b) can thus formally be considered as solved. It has been tacitly assumed in the above that A and A can be pq obtained by some reasonably simple means. The actual method for calculating them is discussed in the following section. 16 CALCULATION OF THE DETERMINANT AND THE COFACTORS OF THE DETERMINANT OF THE MATRIX M In the previous chapter it was shown that the elements of the matrix Tm p can be developed from the knowledge of the determinant and the cofactors of the determinant of Tm] . The matrix I'm] is associated with the set of Equation (10). Consider first of all the task of expressing the determinant A of [Ml in a suitable form. The determinant is rewritten here for convenience A = *W~ " ( VrV i i (a -k ) " (a -k ) 12 n-1 2' 1 1 (d -k ) n-1 n (a -k ) 1 n (39) The determinant can be easily transformed to the form of a double alternant (for the definition of an alternant see Reference 3), by subtract- ing the first row from the rest. After this operation, when the common factors are taken out, there results, A = I ! 1 i ( VV (a 2 - ki ) (a -k ) (a -k ) v 1 2 v 2 2 n-1 1 (a . -k ) n-1 1 (a -k ) (a - k ) In 2 n (a -k ) n-1 n xF (40) where 17 F = n (k -k ) r=2 r X n-1 n (a -k) i r 1 r=l Now consider the (n-l)th order determinant V where V = (VV (a 2 - k 2 ) a -k 1 n 2 n n-1 Z (a -k ) n-1 n (41) The determinant V is a double alterbant and can be conveniently expressed as a ratio of products as follows V= , "- 1 t h (a a ..... a )• fc*(-k_, -k ..., -k ) (-1) 1 2 n-1 2 3 n_ n-1 , n n (a =k ) r in n=l , m=2 (42) where t, 2 is a difference product defined as -i 1-1, q I (x x . x ) = ' n ( x k ~ x «) q k=l,i=2 k X (43) From (40) A = V-F 1 i v & * ( °1- "V ' ' Vl' l * ( - k l'- k 2 •••• "V n-1 ,n n (a -k ) i , r m r=l , m=l (44) 18 Equation (44) gives the desired expression for the determinant A of the matrix [m] . The expression for the cofactors are also determined in a similar manner. Consider first of all the cofactors A , , where ml Xi = <-« mH-1 a -k 1. 1 a -k 1, m-1 I i a i " k m i 1 m+l a -k 1 n a -k n-1 1 a -k n-1 m-1 i i 1 -k a- n-1 m+l a -k n-1 n (45) The determinant in expression (45) is a double alternant. Hence, can be expressed as A _„■* l*«V V •• Vl> *<•*!• " k 2 •' - k n )(m> ml n-1 , n n ( a „-k„) r=l , p=l Pi^m (46) r p where the superscript on the second product implies that k is omitted m from the argument. Next consider A , q ^ 1 mq A mq - ("I) m«-q V k i VrV i ! 1 1 a ,-k a -k 1 m-1 q-1 m-1 a -k a -k 1 m*l q-1 m+l (a -k ) 1 n a -k ' q+1 1 a -k q+1 m-1 a -k q+1 m+l i i a - k n-1 1 a -k . n-1 m-1 a -k n-1 m+l a -k n-1 n (47) 19 The determinant appearing in (47) is of the type A, and the method used above for expressing A is applicable here. The first step is to subtract the first row from the others and take out the common factors. The resulting determinant is a double alternant and can be expressed conveniently. Omitting the details, the final expression for A , q ^ 1 is given below ^(a , a .... a i ) (q) ;*<-*, -k ..,, -k ) (m) A , ,^m+q v 1' 2 n-r 1' 2 n A mq = t" 1 ) ^l~n— ' (48) n (a r -k e ) r=l,i=l r^q,i/m Equations (44), (46) and (48) are the desired expressions for the determinant and the cofactors of the determinant A of the matrix I'm] . 20 5. THE FOURIER TRANSFORM METHOD OF FORMULATION In the previous chapter the method of solving the equation, g(t) = f(t) K|t-T|)dT < t < T Jo g(t) specified for < t < T involved as a first s^ep, the formulation in terms of a set of simultaneous equations. This was possible because the kernel of the Integral Equation could be expressed in a suitable series of exponentials. In this section there is discussed an alternative method of formulating the problem using the technique of Fourier Transformation in the complex plane. The advantage of this method is that the formulation is useful for a general KJt| , the only requirement being that the existence of the transform for the function K|t|. It is also assumed that the asymptotic behavior of g(t) is such that the transform can be defined when gCt) is extrapolated beyond the range it is specified. It is also demonstrated \^.iat when the kernel K(t-T) can be expressed in the form of a series of exponentials, as was the case assumed before, the simultaneous set of equations can also be derived from the present formulation. The exact solution is then possible by the method described in the previous section. For a general kernel K(t-T), however, the solution is obtained approximately by a series of iterations. An insight into this approach is gained by first solving the particular case of the series form of K(t-T). This is pointed out in the discussion given in connection with the iterative method of solution. 5.1 Fourier Transform Formulatio n The formulation in terms of the Fourier Transforms is discussed in the following. The Fourier Transform H(oo) of a function h(t) is usually defined as H(«") = 10Jt e h(t)dt, w real -oo 21 The inverse transform of H(oj) is then given by, h(t) 27T - icot „ v e H(co)doo For the purpose of this work, however, the transforms will be defined in terms of a complex variable s = s + is , The definitions of Fourier Transform of h(t) in the complex plane which are suitable here are given below. -a, t 1 2 Let h(t) be a function such that f (t) e , a > is in class L " a 9 t 1 2 in the interval < t < °o, and f (t) e , a < 0, is in L for -°o < t < 0, then r .OO F (s) = + „, . ist f (t) e dt, s = s +is , s > a (49) and F (s) f(t) e 1St dt, s < a 2 2 (50) The inversion formulas are and 1_ 277 1_ 27T oo+ia. -<30+ia. r 00 ^ F + (s) e 1S ds = f (t), t > ist F (s) e ds = f (t) t < J -oo+i "2 =0, t > Hence the Fourier Transform for h(t) is given by (51) '.T. [h(t)] = H (s) + H_(s) 22 Now consider the Fourier Transform of a function f (t) defined as f x (t) = f(t) o < t < T (52) = otherwise In general f(t) can be expressed as f(t) = q(t) + r(t) o < t < T where q(t) =ZA e n -c t n Re(c ) > n — (53) r(t) =Sfe eS (t " T) n Re(dn) > (54) Note that q(t) and r(t) decrease exponentially for 't positive and negative respectively. The transform for f (t) , F (s) can be expressed as F 1 (s) = F.T. of f x (t) C -X) -ist , v e f (t) dt q(t) + r(t) -co -ist e dt (55) Rewriting (53) one has F 1 (s e 1St dt q(t) e iSt dt (56) i -ist + r(t) e dt - r(t) e" 1St dt -oo 23 Using the shift theorem for transforms it is possible to write, q(t) e" iSt dt = e" lsT V (s) (57) + T where similarly Also let and V (s) = q(t+T) e~ 1St dt >T , v -ist -is* . . ,■_„* r(t) e dt = e W (s) (58) -ist W (s) = r(t+T) e dt ^30 jx> Q (s) = q(t) e 1St dt, (59) -ist R (s) = / r(t) e dt (60) Putting (57), (58), (59) and (60) in (56), there results, F,(s) = Q (s) -e" 1St V (s) 1 + + e 1S W (s) - R (s) (61) Note that the integral in (55) was broken up into (56) because of the behavior of q(t) and r(t) given in (53) and (54). The range was changed from o < t < T to o < t < and T < t < -T5, so that the transforms could be defined in terms Q (s), V (s) and W (s) , R_(s) which are analytic 24 in the complex s-plane in the ranges shown below. Q (s) and V (s) analytic for Im(s) > + + W (s) and R (s) analytic for Im(s) < Now consider the transform of a function g(t) which is known in the range o < t < T Then let g(t) - g x (t) + g 2 (t) o < t < T (62) where say -v t g, (t) = S P e n Re(v ) > (63) 1 n n — and 6 (t-T) g_(t) = S Q e n ' Re (6 ) > (64) 2. n n — Also let g(t) be extrapolated for all t in the form, g(t) = g x (t) U(t) + g 2 (t) U(T-t) + g 3 (t) U(t-T) + g 4 (t) U(-t) (65) In (65) the unknowns are g~(t) and g 4 (t) and U(t) = unit step, i.e. U(t) =0 t < = 1 t > 1 Let the F.T. of g(t) be G(s), then 00 OO e" 1St g(t) dt = - OO e~ ist g(t) dt (66) ,^x -ist + g(t) e dt -oO Using (65) and (66), the following is obtained by obvious application of the shift theorem. 25 ~, x ~ / ^ tisT , -isT G(s) = G 1+ (s) + e I x _(s) + e J 2 + (s) + G 4_ (s > (67) where .00 e 1St g 2 (t) U(T-t)dt = I e iat g 2 (t)dt -00 -00 -ist -ist ,. mXJ , -isT / e g (t+T)dt = e I, (s) 1- -00 and similarly I 2+ (s) = / e _ist g 3 (t+T)dt Also and F.T. F.T. s 1 (t) U(t) g 4 (t) U(-t) = G. (s) 1 + = G 4 _(s). (68) Also if K(t) — *- o as |t|— *°o, then the transform of K(t) say ^(s) is, (s) = . . I . 1 -ist,, K t e dt (69; -00 The equations (61), (67) and (69) give the transforms of f(t), g(t) and K(|t|) respectively under the assumptions mentioned. If now the F.T. is taken of the original Integral Equation, J f(t) k|t-T|dT O o < t < T by first writing it in the form 26 g (t) U(t) + g 2 (t) U(T-t) + g 3 (t) U(t-T) + g 4 (t) U(-t) OO = J f (t) k |t-T|dT J -OO One gets using the expressions for the appropriate transforms developed above, -isT f G (s) + G. (s) + e \l ■ (s) + I (s) 1+ 4- 1- 2 + [Q (s) - R_(s) + e" lsT |V-(s) - V (s)\] (s) (70) The unknowns in (70) are G„ (s) , I (s) , Q (s) , R (s) , W (s) and V (s) . 4- 2+ + + Regrouping terms in (70) there results, G. (s) + G. (s) - Q (s) - R (syf^s) 1+ 4- + " ^ = e~ isT [ |w_(s) - V + (s)j K(s) - I x _(s) - I 2+ (s)] (71) From the definition of the various transforms, it is possible to verify that the only terms dependent on T are the ones appearing on the r.h s. of (71). Since (71) is to be valid for all T, the left and right sides of (71) must separately be equal to zero. The two equations which are obtained using this condition are, JQ + (s) - R_(s)} J^(s) = G 1+ (s) + G 4 _(s) (72) |w_(s) - V + (s)jJc(s) = ^(s) + I 2+ (s) (73) \e known q are JL ( s ) and ^1 ( s ^ • and The known quantities in (72) arte J^ (s) and G, (s) and in (73) , they "1- The equations can be solved if there exists a common strip of analyticity of the transformed functions involved. 5.2 Development of a Set of Simultaneous Equations Consider the case when there exists an expansion of K(t) in the form 27 of a series M -k lt| K(|t|) = S C r e r r=i Re(k r ) > and when n -k t n k (t-T) g(t) =SCPe r + SCQe r ° r r r r It was shown in Chapter 1, that such a g(t) is obtained when the particular solution f (t) is filtered out of the integral equation by calculating f (t) - f (t) ). K t-T|dT P Then from (62) , g. (t) = S C P^ e 1 - r r Y=l -k t r and Using g 2 (t) = S C r Q r e Y =l k (t-T) r r oo J. -k t -ist . 1 re dt = e k +is one has the expressions for G 1 (s) and I-,_( s ), a s , C P r r G. (s) = S- ~ 1+ k +is (74) and -k T C Q e r I ( S ) = H-*-?—. 1- k -is (75) Also 28 •J& C k r r "- 1 , 2 2 n (s +a r ) 4 ?=1 ' „ 2 2 k +3 r n 2 2 n (s +k r= 1 F.T. (K t|) = /T (s) = 22 where d is a constant and a.r s are the zeros of /£, (s) . Equation (72) and (73) for this case becomes n ; L . 2 2 > n (S ^r > C P Q + (s) - R_( 3 )|,d ^ = 2 r ^"f- + G 4 _(s) (76) n (s +k r r=l and n-1 2 2 , n (S +a r > C Q -k T w (s) - v (s)Ld — = ^tt^tI e r + l o (s > (77) + n k tis 2+ n (s 2 +k 2 ) r=l From the nature of the differential equation of which 1 (t) is the P particular solution one knows that the complementary solution f(t) - f (t) has the form, n-1 -a T n-1 -a (t-T) A 6(t) + 2 A e r +B 6 (t-T) + 2 B e r o r o . . r r=l i=l Therefore Q (s) can be assumed to be of the form + A Q (s) = A + 2 —A- (78) + o a -is Similarly B W (s) = B + 2 ' r . (79) O CL +1S r The unknowns are A , A s , B and B s . o l o i It is possible to express the remaining quantities in the l.h s. of (76) and (77), i.e. R_(s) and V (s) , in terms of the coefficients 29 A s and B s. For this purpose refer to (57) and (58) where the definitions of R (s) and V (s) were given. It is seen that and F.T." 1 R (s) = r(t) U(-t) -1 F.T. W_(s) = r(t+T) U(-t) If R (s) is chosen to be (80) (81) R (s) = S B e r -a . T x a. +is r r (82) The Equations (80) and (81) are consistent as is verified by inverting R_(s), and using (79). Similarly -1 and Hence F.T. V (s) = q(t+T) + F.T " Q (s) = q(t) + a T i A e V (s) = A + 2 -£ r- + o a -is (83) Substituting (78} , (79), (80) and (81) in (76) and (77), one gets after dividing both sides hyjCis) -a T f A Be 1 ^ 1 CP G (s) A o + s 5^Ii ~ 2 -^T" } ^^(^ S FTi7 + ^"Ts7 and B A e r r B + S —I— - £ o a +is -a T -k T 1 ;- C rV r I 2 + (S) ,/& (84) (85) Notice that lA/Ms) has zeros for i' = + k . Hence if one sets / s — r i = k in (84), m = 1, 2, . „.„n, one obtains the set of equations, -a T A B e , C P A o + ^ a -k " ^ a + k ~ T~(k -i J M^TT r m rm I m s ^ c J 30 m=l, 2, ... n (86) s=ik m This is because all the terms in the summation on the right hand side go to zero except the one containing l/k -i in the denominator. Notice / m s that G (s) can not have a pole for s = -ik since it is regular in the 4- m lower half of the s - plane. Similarly, from (37) one obtains by setting i = -k or s = ik , s m m -a T -k T B Ae r CQe m r r mm B + S — - - S : = T-P7-: 7T r-n m = 1, 2, ... n (87) o a -k a +k [ £(s)(k +i 1 ' r m r m lj^ m s J . , s=ik m By comparison one notices that (86) and (87) have the same form as (33a) and (33b) in Chapter 3. In particular the matrix associated with these equations is exactly the same. Methods discussed in Chapter 3 can therefore be used to solve (86) and (87) . 5.3 Iterative Method of Solution It was shown in the discussion above that the transformed equations (72) and (73) can be reformulated in terms of a set of simultaneous equations similar to the ones obtained in Chapter 3. This was shown to be possible when the kernel K(t-T) is expressible in a suitable series of exponentials. However, this might not be a convenient thing to do for all kernels. It is therefore useful to find alternative methods for solving (72) and (73). An iterative method for solving the above equations is presented here. The method of solution is based on the following reasoning. Suppose one starts with an approximation of q(t), say q (t) . Let q (t) be the solution of the problem for T - °o. q (t) is therfore obtainable by using Wiener-Hopf method. The following steps are carried out for the iterative procedure. (1) 1. Predict q (t+T solution of the equation corresponding to T = °o, 1. Predict q (t+T) - approximate q(t+T) using q ' (t) which is the 31 2. Find V (s) , the transform of q (t+T) , substitute in (73) and solve (1) + for W (s) , the transform of r(t+T) (see the definitions of V and W in + - (57) and (58), by using Wiener-Hopf technique 3. Find V ' (t) by inverting W (s) and substitute in (73). Next solve (2) for q (t) by applying Wiener-Hopf method to (72) . 4. Repeat steps (1), (2) and (3) until the solution converges. No proof is included here regarding the convergence of the process. However, it is possible to make the following observations. If q(t+T) is small for t > o, i.e. if the solution q(t) decays with increasing t and if T is large enough to make q(t+T) small, then the solution r (t+T) is a good approximation of the actual r(t). Also if r(t) is fairly small for t < o, then from the definition of R_(s) it is again seen that q.. (t) obtained from (72) is a close approximation of q(t). The process is then expected to converge fast. Although it might not be practically convenient to express K 1 1— t | in a series of exponentials, it is nevertheless useful for making certain statements about the convergence of the iterative process by considering a formal series expansion of K|t-r| in terms of exponentials. From the discussion in Chapter 4 (see Equation (42)) that JO, t a. (t-T) f (t) - f (t) = A 6(t) + 2 A e 1 + B + S B, e 1 P ° i = l ± ° i=l l From the above expression it is seen that q(t+T) for t > T and r(t) for t < are small provided T is large or the a . ' s are sufficiently large. It is therefore observed that if the kernel has a reasonably fast decay with the increase of its argument, and the particular solution is filtered out from f(t) in the beginning, it is expected that the conver- gence of the iterative process will also be reasonably fast. It is not necessary that the kernel be of an exponentially decaying type. The essential condition for a rapid convergence of the process is that q(t) and r(t-T) tend asymptotically to the corresponding semi-infinite range solution at t = T. It is assumed in the above that the particular solution has been filtered out from the general one. Considerations such 32 as these apply to kernels of the type Ho (2) Mt-T which appear in the formulation of electro-magnetic wave problems, say for instance in the problem of diffraction of e„m. waves by a perfectly conducting strip. 33 6. THE EIGENVALUE PROBLEM In the preceding chapters the discussion was centered around solving the integral equation f T gl"t) = f(T) K(lt-Tj) dT < t < T Jo g(t) specified in the range < t < T and methods were presented for solving the equation when n -k |t| Kit! = S C e r r=l r In a class of problems in probability theory the eigenvalue problem is encountered. The origin of the Equation (88) and a transform method 4 o f solution has been discussed by Youla in a recent paper. The purpose here is to show that the eigenvalue problem can be formulated in terms of a matrix without resorting to the transform method. The direct method of formulation is simple and involves comparatively less amount of work. 6.1 The Eigenvalue Matrix Equation Assume (t) in the form n -y t v t (h(t) =E(Ae m +Be m ) 0(T) K |t-T|dT, O can be made equal to zero, then it will be possible to obtain the solution of the eigenvalue problem. When (89) is substituted in (88) , the evaluation of the resulting expression involves integrals of the type r T -k |t-T| 2k e pt -V e^ + ( f -^ r | e P T e r d T , -Jl— - S— — (90) 2 , 2 p-t-k p-k p -k r r •/ o r Hence n -> t rt /- T X d)(t) =H 2 A e m + 2 B e m l= (t) K|t-T| dT m=l /; - / o T T 1 I n r V t n -k t-T n / -Y t -k |t-T| = S / Ae m SCe r dT+ZI B e m 2 C e r dT / m r m r m=l ) r=l m=l / J o - / o -k t k t k A C -> t A C e r e r -(k +Y )T = -2 2 _LJL£ o e m - 2 -JL* 2 A C - v r m «, 2 , 2 -y +k m r k +y mr V -k mr m r mr r m m r Y t -k t k t-(k -Y )t kBCe m BCe r B C e r _ 2 j. r m r _ m r _ m r v, 2 , 2 k +y ~ k -y (91) mr y -k mr r m mr r m m r -Yt If (91) is to be true for all *t ', the coefficients of e and r m t e in both sides of (91) must agree for all J [ e P T e ' ' iT ^r___e^ -_ (90) 2 , 2 p+k p-k p -k r r ^ o r Hence n -Vt 7t /- T r 6(t) = r f S A e m + S B e, m 1 = mr V -k mr r y m mr r m ^m r -yt If (91) is to be true for all ft', the coefficients of e and y m t e in both sides of (91) must agree for all 'm* and the coefficients -k t k t of e and e must vanish for all i r^. k t Applying the second condition, one has from (91) Coeff of e r is 3o equal to zero for all r or, A B m r m m r m -k L Coeff of e is equal to zero for all r , or (92) -Y t y t A e m B e m 2 J? + 2 J5 = k +y k +y m r m m r m -k t r Here the common factor e has been canceled from the second equation in (92) . Equation (92) is a homogeneous set of 2n' equations for the coefficients A and B . ■%* , -^ . mm -T t f x m ni t i Also by equating the coefficients of e or c ' . for each m in both sides of (91), the following is obtained. k C i = r 2 2 ^ r r = 1, 2, . .. n (93) K r T ~k m r The routine for calculating \ and the unknowns in the expansion of the eieenfunction, viz, A , B and y is the following. ° m m m Noting that Equation (93) is an algebraic equation of nth order for 2 y ' and that the coefficients of this equation in value \, the following m steps can be outlined. 2 2 2 2 1 Find 'n' roots of r , viz Y , 'V »'. . V ' from (93), Note that the m 1 2, n expression for these roots involve X.' 2. Substitute Y , Y , ... etc. in (92). Since (92) is a homogeneous equation for A 's and B 's, the determinant must be equal to zero. Using A = where A is the determinant of (92) , solve the transcendental equations for X, yielding an infinite number of solutions X , X , .... . 3 For each X, say X calculated (m = 1, . . . . n) from the expressions P mp in step 1. 4. Using these 'Y ' s solve (92) for A ( s and B 's corresponding to X mp mp mp p Note that the coefficients can be determined only within a multiplicative 36 th factor. The expression for the p eigenfunction is then -7 t +7 t <$> = S A e mp + S B e m P T p m mP m mp (94) The difficulty in actually working through the above procedure is encountered, first, in solving for the n roots of X in step 1, since the coefficients m involve the unknown X. Even more difficult is the solution of A(X) = 0, since this equation is transcendental. Only approximate solutions can therefore be expected. Example: Consider the case K|t-T| = e Equation (93) is then, -k t-T j t,e v 2k X T 2 2 r - k or 2 2 y + 2kX - k or = y= (k 2 -2^x)^ (95) From (92; one has, k-r k+r - = o Ae -YT k+y Be 7T k-r = (96) or or A i k-y -7t e k+y k+y k-r e e"* T (k-f) 2 (k+r) 2 ft e = + k-y = = - k+y (97) 37 Substituting from (95) in (97) (K 2 -2k\)* T k - (k 2 -2k X)^ , Qjn — 9, 1 k +(k -2kX) 5 Solutions of (98) which is a transcendental equation for X. are the eigenvalues of the problem. For a particular \ say X , y = (k 2 -2kX )■ P P from (95) and from (96) Hence k+y B = __i A p fTk i -7 1 k+r y t 4> = A \ e P + Z7-T- eP l 0 * — sinh X t + coshT^ t (100) y P r ' p p P where the constant factor has been dropped. 4 The same solution was obtained by Youla although by using a different method of formulation. 6.2 Upper Limit Infinity It is of interest to consider the case when T, the upper limit of the Integral Equation is replaced by infinity. The formulation presented in this section can still be used but there results considerable simplification of the process of obtaining \ and <)>. This is discussed in the following It is to be noticed in the first instance that because the range of the integral equation is from to oo , continuous eigenvalues will result 38 due to the type of kernel considered. In order for the integral -k t -y t y t (2 C e r ) (2 A e r + Be r ) dt r r to converge it is essential that if Re y = y (y >0) then r rl rl y , < k, , k real. It has been assumed that k, < k„ < ... < k and nl 1 r 1 2 n Y, - < \„- < ... < Y , • It is to be noted that since v , > 0, all -v s 11 21 nl 'rl 'r are admissible. Since the y s involve the eigenvalue X, this sets a limit to the values that X can take. Consider a simple example in which k|t-T|= e Then the integral equation is, k t-T j.^ ^ I , n -k t-T (t) = X (t) e ' IdT, k real (101) Assuming § (t) to be of the form, ('t) is then (107) cj>(t) = A - v/k -2kX t v/k"-2k\-k V k~-2\k t e " + — e k *2kX+k (108) Note that if Re X< then Re(^y)>k hence the negative values of Re(X) are inadmissible. Equation (101) was solved by the Wiener-Hopf technique and appears as an example in the book by Morse and Feshbach . When the series representation of the kernel contains more than one term say p , then it is found that the problem reduces to finding (p+1) coefficients from 'p 1 simultaneous equations. This can be done by choosing one of the coefficients arbitrarily and solving the 'p' equations. The eigenvalue spectrum is continuous, hence the allowed values of X are arbitrary, except for the restriction on Re(X). This restriction is imposed from the convergence condition, i.e. the restriction on Re (■y) , which was discussed earlier. 40 NUMERICAL EXAMPLE In this Chapter a numerical example is worked out to demonstrate the application of the theory discussed in Chapters '2 and 3. The equation considered is of the form (1) Let the kernel K |t-T | of the form .L I - - |t-x| n -2 It-T -3|t-T K t-T = e ' ' -T- 2e ' ' -e ' ' i.e. k x =1, k 2 =2, k 3 .3 2 F 0*> ) , the Fourier transform of K(,t) is F (co ) = 2 2 2 2 co +1 co +4 co +9 i 2 ci s are obtained from the roots of F (co ) , and are a = 0,95837 a = 3.44949 2 The matrix [mJ which corresponds to the case of T = oo is then, [«]■ 1 -1.84628 1 -0 64866 1 -0 39345 0.40825 0.68990 2 22474 The cofactors and the determinant for [mJ and M calculated using the expressions in Chapters 3 and 4 and verified by direct calculation are, 41 *L1 ■ -1 17167 *21 - -3 94688 *si - -1 00893 \2 = -1 53485 ^3 = 25522 ^2 = 1 81650 ^3 = -1 45283 *32 = -0 28165 ^33 = 1 19762 \l = -i 28444 4i = 4 17186 4 = -1 05800 \2 = -1 53485 4 = 27322 4 = 1 81650 4 = -1 51622 4 = -0 28165 a.„ = 1 24221 are A = 1.76627 A = 1,82941 The inverse of Tm] and Tm 1 constructed from the above data [»] -1 -0 66336 -0.86898 -0.14450 2.23458 1 02844 -0.82254 -0.57122 -0.15946 0,67805 l"']- 1 -0.70211 -0 83899 0.14935 -2,28044 0.99294 -0.82880 -0.57833 -0,15396 0.67902 Once the inverses [ M] and [m'] are known the Integral Equations for both T = o=> and a finite T can be obtained by the methods outlined in Chapters 2 and 3 42 8 APPLICATIONS The Wiener-Hopf Equation for the finite and infinite ranges occur in the problems associated with the theory of linear prediction and the design of optimum filters in the field of Information Theory, In the field of Electromagnetic Theory, the Wiener-Hopf Equation occurs in numerous problems of wave propagation. Examples are the problems of diffraction by a slit, of finite width in an infinite plane, or the diffraction by an infinite stack of equispaced metal planes Several other examples are also scattered around in the literature on wave propagation and boundary value problems. 43 9 . CONCLUS IONS In this report, methods have been developed for the solution of finite and infinite range Integral Equations of Wiener-Hopf type and the eigenvalue problem associated with the kernel. The assumption in a part of the work is that the kernel K|t-T| can be conveniently expressed as a series of exponentials. Iterative methods are presented in another part for the case where the series expansion is inconvenient. Examples are worked out to illustrate the methods. Applications in the fields of Information Theory and E.M. wave propagation are pointed out. Actual application to a particular problem in field theory viz. the diffraction by a stack of finite-sized metal planes is left for discussion in a future report. 44 REFERENCES 1. Morse and Feshbach Methods of Theoretical Physics , Pt . I, p 978, McGraw Hill 2 Laning and Battin, Random Processes in Automatic Control , Chapt 7, McGraw Hill. 3. Muir and Metzler, A T reatise on the Theory of Determinants , Longmans Green & Co. n i* 4 Youla, The Solution of a Homogeneous Wiener-Hopf Integral Equation, IRE Trans, on Information Theory, Vol. IT-3 , No. 3, Sept, 1957 ANTENNA LABORATORY TECHNICAL REPORTS AND MEMORANDA ISSUED Contract AF33 (616>-310 "Synthesis of Aperture Antennas/' Technical Report No, 1 , C„T„A Johnk October, 1954, A Synthesis Method for Broad-band Antenna Impedance Matching Networks Technical Report No, 2, Nicholas Yaru, 1 February 1955 „ The Asymmetrically Excited Spherical Antenna/' Technical Report No. 3, Robert C. Hansen, 30 April 1955. "Analysis of an Airborne Homing System," Technical Report No. 4, Paul E„ Mayes. 1 June 1955, (CONFIDENTIAL) "" Coupling of Antenna Elements to a Circular Surface Waveguide,' 1 Technical Report No,, 5, H, E, King and R. H, DuHamel, 30 June 1955 ^nput Impedance of a Spherical Ferri te Antenna wigh a Latitudinal Current,'* Technical Repo r t No > 6 , W„ L. Weeks, 20 August 1955. "Axially Excited Surface Wave Antennas," Technical Report No c 7, D, E. Royal, 10 October 1955„ "Homing Antennas for the F-86F Aircraft (450-2500mc), " Technical Report No. 8 P. E Mayes, R F, Hyneman, and R. C Becker, 20 February 1957. (CONFIDENTIAL) Ground Screen Pattern Range," Technical Memorandum No. 1, Roger R, Trapp, 10 July 1955„ ~ ~~ Contract AF33 (616 ) -3220 Effective Permeability of Spheroidal Shells,"' Technical Report No, 9, E J, Scott and R H DuHamel, 16 April 1956 "An Analytical Study of Spaced Loop ADF Antenna Systems," Technical Report No. 10, D G Berry and J, B, Kreer, 10 May 1 956 . A Technique for Controlling the Radiation from Dielectric Rod Wavequides," Tec hnical Repo rt No, 11, J W Duncan and R, H DuHamel, 15 July 1956. "Directional Characteristics of a U-Shaped Slot Antenna," Technical Report No 12, Richard C Becker, 30 September 1956, 'Impedance of Ferrite Loop Antennas," T ech nical Report No 13, V. H Rumsey and W. L Weeks, 15 October 1956, Closely Spaced Transverse Slots in Rectangular Waveguide '" Technical Report No 14, Richard F, Hyneman, 20 December 1956, Distributed Coupling to Surface Wave Antennas, Technical Report No 15. Ralph Richard Hodges, Jr., 5 January 1957. The Characteristic Impedance of the Fin Antenna of Infinite Length, " r Technical Report No, 16 , Robert L Carrel, 15 January 1957, 'On the Estimation of Ferrite Loop Antenna Impedance," Technical Report No, 17, Walter L. Weeks, 10 April 1 957 : ~~ "A Note Concerning a Mechanical Scanning System for a Flush Mounted Line Source Antenna," Te chnical Report No. 18 , Walter L. Weeks, 20 April 1957 , Broadband Logarithmically Periodic Antenna Structures, ' Technical Report No, 19 , R H DuHamel and D t E c Isbell, 1 May 1957, Frequency Independent Antennas, f Technical Report No„ 20 , V. H. Rumsey, 25 October 1957 The Equiangular Spiral Antenna," Technical Report No. 21, J, D, Dyson, 15 September 1957 'Experimental Investigation of the Conical Spiral Antenna, ' Technical Report No, 22, R. L. Carrel, 25 May 1957, "Coupling Between a Parallel Plate Waveguide and a Surface Waveguide," Technical Report No 23, E J, Scott, 10 August 1957 launching Efficiency of Wires and Slots for a Dielectric Rod Waveguide," Technical Report No. 24, J.W. Duncan and R H, DuHamel, August 1957 The Characteristic Impedance of an Infinite Biconical Antenna of Arbitrary Cross Section," Technical Report No. 25, Robert L. Carrel, August 1957. Cavity-Backed Slot Antennas," Technical Report No, 26 , R.J. Tector, 30 October 1 957 : "Coupled Waveguide Excitation of Traveling Wave Slot Antennas," Technical Report No 27, W,L Weeks, 1 December 1957. Phase Velocities in Rectangular Waveguide Partially Filled with Dielectric," Te chnic a l Repor t No 28, W,L Weeks, 20 December 1957 e 'Measuring the Capacitance per Unit Length of Biconical Structures of Arbitrary Cross Section, Techn i cal Report No, 29 , J.D Dyson, 10 January 1958 Non-Planar Logarithmically Periodic Antenna Structure," Technical Report No. 30 , D W Isbell, 20 February 1958 Electromagnetic Fields in Rectangular Slots," Technical Report No. 31 , N„J. Kuhn and P, E, Mast, 10 March 1958, The Efficiency of Excitation of a Surface Wave on a Dielectric Cylinder," T echnical Report No. 32 , J.W, Duncan, 25 May 1958, A Unidirectional Equiangular Spiral Antenna,' 1 Technical Report No, 33 , J D. Dyson, 10 July 1958, 'Dielectric Coated Spheroidal Radiators," Technical Report No, 34, W,L„ Weeks, 12 September 1958. "A Theoretical Study of the Equiangular Spiral Antenna," Technical Report No. 35, P.E Mast, 12 September 1958. Contract AF33 (616 )-6079 "Use of Coupled Waveguides in a Traveling Wave Scanning Antenna," Technical Report No. 36 , R,H, MacPhie, 30 April 1959, DISTRIBUTION LIST One copy each unless otherwise indicated Armed Services Technical Information Agency Arlington Hall Station Arlington 12, Virginia 3 copies, 1 repro. Commander Wright Air Development Center Wright-Patterson Air Force Base, Ohio ATTN: WCLRS-6, Mr. F 3 E; Burnham 3 copies Commander Wright Air Development Center ATTN. WCLNQ-4, Mr, M, Draganjac Wright-Patterson Air Force Base, Ohio Commander Wright Air Development Center ATTN WCOSI, Library Wright-Patterson Air Force Base, Ohio Director Evans Sign&l Laboratory ATTN. Technical Document Center Bel mar, New Jersey Commander U,S„ Naval Air Test Center ATTN ET-315, Antenna Section Patuxent River, Maryland Chief Bureau of Ordnance Department of the Navy ATTN Mr, C„H Jackson, Code Re 9a Washington 25, D„ C. Commander Hq„ A u F, Cambridge Research Center Air Research and Development Command Laurence G. Hanscom Field ATTN: CRRD, R E„ Hiatt Bedford. Massachusetts Commander Air Force Missile Test Center Patrick Air Force Base, Florida ATTN: Technical Library Director Ballistics Research Laboratory ATTN: Ballistics Measurement Lab Aberdeen Proving Ground, Maryland Office of the Chief Signal Officer ATTN: SIGNET -5 Eng. & Technical Division Washington 25, D C„ Commander Rome Air Development Center ATTN: RCERA-1 D.Mather Griffiss Air Force Base Rome, New YOrk Airborne Instruments Lab , Inc, ATTN; Dr E c G„ Fubini Antenna Section 160 Old Country Road Mineola, New York M/F Contract AF33 (616) -2143 Andrew Alford Consulting Engineers ATTN Dr„ A. Alford 299 Atlantic Avenue Boston 10, Massachusetts M F Contract AF33 (038)-23700 Bell Aircraft Corporation ATTN: Mr_ J Q D Shantz Buffalo 5, New York M/F Contract W-33 (038)-14169 Chief Bureau of Ships Department of the Navy ATTN: Code 838D, L. E. Shoemaker Washington 25, D,C. DISTRIBUTION LIST < CONTrNIJED » McDonnell Aircraft Corporation ATTN Engineering Library M F Contract AF33 (600)-8743 Lambert Municipal Airport St Louis 21, Missouri Dr L. Cutrona University of Michigan Aeronautical Research Center M F Contract AF33 t 038) -21 573 Willow Run Airport Ypsilanti, Michigan Professor HI, Zimmermann Research Lab of Electronics Massachusetts Institute of Technology M F Contract AF33 (616 J-2107 Cambridge, Massachusetts Dr. J, A Marsh -North American Aviation, Inc Aerophysics Laboratory M F Contract AF33 .'038)18319 12214 Lakewood Boulevard Downey, California Mr . Dave Mason Engineering Data Section North American Aviation, Inc M F Contract AF33 1038)1 8319 Los Angeles International Airport Los Angeles 45,, California Northrop Aircraft, Incorporated ATTN: Northrop Library Dept, 2135 M F Contract AF33(4oO) -22313 Hawthorne, California Radioplane Company M E Contract AF33 600j-23893 Van Nuvs, California Lockheed Aircraft Corporation ATTN C L Johnson P O Box 55 M F NOa(s)-52-763 Burbank, California Robert Borr-s Raytheon Manufacturing Company Wayland Laboratory, Wayland, Mass Republic Aviation Corporation ATTN:: Engineering Library M ,"F Contract AF33 (038)-14810 Farmingdal e Long Island, New York Sperry Gyroscope Company ATTN: Mr„ B. Berkowitz M/F Contract AF33 (038) -14524 Great Neck Long Island, New York Mr George Cramer Temco Aircraft Corporation Contract AF33 (600)21714 Garland, Texas George Giffin Farnsworth Electronics Co„ Marked: For Cont AF33 (600) -25523 Ft Wayne, Indiana Mr James D Leonard North American Aviation, Inc, Contract NOa(g) 54-323 4300 E Fifth Avenue Columbus, Ohio Mr P D Newhouser Development Engineering Westinghouse Electric Corporation Air Arm Division Contract AF33 (600)-27852 Friendship Airport, Maryland Air Force Development Field Representative ATTN: Capt , Carl B„ Ausfahl Code 1010 Naval Research Laboratory Washington 25, DC. Chief of Naval Research Department of :he Mavy ATTN: Mr Harry Harrison Code 427 Room 2604, Bldg, T-3 Washington 25, D„C„ DISTRIBUTION LIST (CONTINUED) Svlvaaia Electric Products, Inc Electronic Defense Laboratory B/J* Contract DA 36-039-sc-75012 P.O Box 205 Mountain View California Ennis Kuhlman % McDonnell Aircraft P 0„ Box 516 Lambert Municipal Airport St„ Louis 21, Missouri Stanford Electronics Laboratory Stanford University ATTN- Applied Electronics Lab„ Document Library Stanford, California Radio Corporation of America^ R„C A Laboratories Division Princeton, New Jersey Electrical Engineering Res, Lab. University of Texas Box 8026, University Station Austin, Texas Dr Robert Hansen 8356 Chase Avenue Los Angeles 45, California Technical Library Bell Telephone Laboratories 463 West Street New York 14, New York Dr„ R„E„ Beam • Microwave Laboratory Northwestern University Evanston, Illinois Dr H„G„ Booker Department of Electrical Engineering Cornell University Ithaca, New York Applied Physics Laboratory Johns Hopkins University 8621 Georgia Avenue Silver Spring, Maryland Exchange and Gift Division The Library of Congress Washington 25, DC. Mr,, Roger Bat tie Supervisor, Technical Liaison Sylvania Electric Products,, Inc, Electronic Systems Division P.O, Box 188 Mountain View, California Physical Science Lab, ATTN: R r Dressel New Mexico College of A and MA State College, New Mexico Mrs E,L. Hufschmidt, Librarian Technical Reports Collection 303 A, Pierce Hill Harvard University Cambridge 38, Massachusetts Dr, R,H. DuHamel Collins Radio Company Cedar Rapids, Iowa Dr„ R,F, Hyneman 5116 Marburn Avenue Los Angeles 43, California Director Air University Library ATTN: AUL-8489 Maxwell AFB, Alabama Mary Lou Fields,, Acquisitions Stanford Research Institute Documents Center Menl ow Park, California Dr Harry Letaw, Jr Research Division Raytheon Manufacturing Co. Waltham 54, Massachusetts Canoga Corporation 5955 Sepulveda Boulevard M/F Contract AF08 C'603)-4327 P.O, Box 550 Van Nuys, California UNIVERSITY OF ILLINOIS-URBANA Q621.365IL655TE «« TECHNICAL REPORTSURBANA 37 1959 " 3 0112 008079326