‘naly ‘cal Methods and pecial Functions +1 In *'national Series of 4 necvaph ‘n Matl itatics Bessel Functio 's and their Applications B. G. KorenevBessel Functions and their ApplicationsANALYTICAL METHODS AND SPECIAL FUNCTIONS. An International Series of Monographs in Mathematics FOUNDING EDITOR: A.P. Pradnikov (Russia) SERIES EDITORS: CE. Duaki (USA), H.-J. Glaeske (Germany) and M. Saigo (Japan) Volume 1 Series of Faber Polynomials PK. Suan ‘Volume 2 Inverse Spectral Problems for Differential Operators and Their Applications VA. Yarko Volume 3 Orthogonal Polynomials in Two Variables PK. Suetin Volume 4 Fourier Transforms and Their Approximations AM, Seles Volume 5 Hypersingular Integrals and Their Applications SG. Sambo Volume 6 “Methods of the Theory of Generalized Functions VS. Vladimirow Volume 7 Distribution, Integral Transforms and Applications W. Kierat and U. Sztaba Volume 8 Besse! Functions and Their Applications BG. Korner Forthcoming: Volume 9 H-Transforms: Theory and Applications AvA. Kilbas and M, Saigo ‘This book is par ofa series. The publisher will accept continuation orders which may be cancelled at any time and which provide for aucomatic billing and shipping of each riele in the series upon publication, Please write for details,Bessel Functions and their Applications B.G. Korenev ‘andes London and New Yorkree published 2002 by lajlor & Francis 11 New Fetter Lane, London EC4P 4EE ‘Simleancousy published inthe USA aod Canada by Taylor & Francis Inc, 29 West 35ch Steee, New York, NY 10001 “yl & Francs san irint of th lor & Francis Gro © 2002 Taylor & Francis “Tranalted by EV. Packeaiew “Types in 1012 Garamond by [Newgen Imaging Systems () Led, Chenosi, India Primed and bound In Grest Brita by ‘T)Interastonal Led, Padstow, Corowa All sighes reserved. No pat of thie ook may be eprnted or reproduced or utilised i any ference by ay elecronie, mechanical, or other means, aow know o here invented, including photocopying tnd recording, or io any informatio storage or retrieval system, ‘without permission in writing fom the publishers, very effot has been made to ensure thas dhe advice and formation inthis bok i eve tne accurate atthe time of going ro press. However, neither che publisher no the authors «an accept any legal responsibility or lability for any errs or omissions dat may be ‘pede Inthe case of drug administration, any medical procedae ofthe te of techaica tejuipment mentioned within his book, you are srongly advised 0 consul the smanufurater’s guidelines. British Library Cataleguing in Pablcton Dat ‘Acstalogue ecord fortis book ie available fom the British Library Litrary of Congres Cataloging i Publication Data ‘Acatalogue record has been requested ISBN 0415-28130Contents Preface vii Introduction 1 Part 1. Foundation of the theory of Bessel functions 5 Chapter 1. The Bessel equation. Properties of Bessel functions 7 1. The Bessel differential equation. Application of power series. Cylindrical functions of the first. kind a 2. Cylindrical functions of the second kind (Neumann functions) 10 3. Cylindrical functions of the third kind (Hankel functions) a 4. Cylindrical functions of a pure imaginary argument n 5. Cylindrical functions of a complex argument 12 6. Formulae of differentiation, recurrence relations “4 7. Cylindrical functions with a half-integer index 16 8. Some notation for functions of half-integer and fractional index. Airy integral 7 @ Wroneli det "i 10 10. Bessel integral and Jacobi expansion 22 11. Addition theorems 26 12, Lommel expansion 31 13. Differential equations reducible to the Bessel equation 31 14. The Poiscon integral 37 15. Some indefinite integrals 40 16. Functions contiguous to Bessel functions. Particular solutions of the inhomogeneous Bessel equation 43 17. On integration of the inhomogeneous Bessel equation. Cauchy functions 50 18. Products of Bessel functions 56 19. Integral representations of Bessel functions 59 Chapter 2. Definite and improper integrals. Series in Bessel functions 65 20. Definite integrals 65 21. Improper integrals 68 22. Dual integral equations 87 23. Roots of the Bessel functions 93 24. Series of Fourier-Bessel and Dini 96 25. Schlémlich series 104 26. Neumann series 3 27. Lommel functions of two variables 114vi contents 28. On partial cylindrical functions 120 29. Asymptotic expansion of Bessel functions 125 30. On Bessel functions with a large index 128 Part 2. Applications of Bessel functions 131 Chapter 3. Probleme of the theory of plates and shella 133 31. Oscillations and stability of a circular plate 133 32. Equilibrium of a circular plate lying on an elastic foundation. Axially symmetrical deformation 155 33. Equilibrium of a circular plate lying on an elastic foundation. Non- axially symmetrical deformation 162 34, Calculation of a circular conic shell under the action of axially symmetrical loadings and non-uniform heating 170 35. The method of compensating loadings in problems on membranes and plates 184 36. The problem on equilibrium of an unbounded plate which lies on a. uniform elastic foundation whose model has a circular symmetry 208 Chapter 4. Problems of the theory of oscillations, hydrodynamics and heat transfer 25 37. On the oscillations of a thread 215 38. Stability and transverse-longitudinal bending of a rectilinear rod; stability of the plane form of the bending of a strip 219 39. Plane heat waves in a half-space and a layer; heat waves in arod 241 40. A die which lies on an elastic half-space whose modulus of elasticity is a power function of the depth uT 41. Application of integral equations to the solution of some problems on membranes and plates 250 42. On wave propagation in electric power lines 256 43. The action of an impulse on cylindrical and prismatic tanks filled with a fluid 287 Appendix A. Drief information on gamma functions 265 Bibliographical notes 267 Bibliography am Index 25Preface A large number of diverse problems concerning practically all the most important areas of mathematical physics and various technical problems is connected with applications of Bessel functions. References where these functions are present are actually immense. Different parts of Bestel function theory are widely naed when solving problems of acoustics, radio physics, hydrodynamics, atomic and nuclear physics and s0 on. Applications of Bessel functions to heat conduction theory, including dynamical and linked problems are very numerous. In elasticity theory the solutions in Bessel functions are effective for all spatial problems, which are solved in spherical or cylindrical coordinates; algo for different, problems concerning the oscillations of plates and the equilibrium of plates on an elastic foundation; for a series of questions of the theory of shells; for problems on the concentration of the stresses near cracks and others. In each of these areas there is a wide range of different applications of Bessel functions. Bessel functions are well studied and there are a large number of reference books which contain the formulae and different tables. The study of the theory of the Bessel function from the viewpoint of its appli- cations has to be of interest for a large number of engineers and researchers. Special parts of different. courses of mathematical physics are devoted to the presentation of the theory of Bessel functions (see the courses of R. Courant and D. Hilbert, P.M. Morse and H. Feshbach, A.N. Tikhonov and A.A. Samareki and others). At the same time, monographs devoted to the special functions also put an emphasis on the study of the Bessel functions, It concerns, for instance, the second part of “A course of modem analysis” by E.T. Whittaker and G.N. Watson, the book “Special functions and their applications” by N.N. Lebedev, the book “Transcen- dental functions” by A.N. Kratzer and V. Franz, the monograph by [Link]. Vilenkin “Special functions and theory of group representations” and others. ‘There are several books concerning the theoty of Bessel functions in Russian. ‘We mention two editions of the book by R.O. Kuz'min “Bessel functions”, the translation of the book by E. Grey and G. Matyuse “Bessel functions auid th applications in physics and mechanics’ as weii as the transiation of the thos! course in the theory of Besse) functions, namely, the book by G. treatize on the theory of Bessel functions” . ‘This book contains a brief presentation of the elements of the theory of Bessel functions which is accompanied by examples of the applications of the theory to the solution of various physical problems.viii PREFACE, ‘The book is intended for the reader who is, basically, interested in the appli- cations of the theory, not the theory itself. This fact determined the choice of the topics as well as the style of the presentation, The author’s intention was to give the reader the opportunities to find the most necessary information so that afterwards the reader can consciously use the reference literature and can, all by himself or herself, apply the Bessel functions to the solution of applied problems. ‘Applying the methods of the theory of Bessel functions, an engineer-revearcher must have a general notion of the foundation of the theory and clearly represent the mutual connections hetween its different areas. This is the reason why the deduction of the formulae, given in this book in a large number, are important for the reader who is interested in applications. For the same reason, when choosing the schemes of the deduction, we prefer those which use some properties of the Bessel functions in order to clarify other properties. ‘The mutual connections between different parte of the theory can be cen more clearly by this approach. ‘Some short remarks or introductory sections of an heuristic nature are also intended to clarify the connections of the theory with physical applications. This is the viewpoint that should be used when considering some conclusions related, for example, with the addition formulae, the Bessel integral, improper integrals and others, as well as the applications of the Bessel functions considered in the second part of the book. ‘Tho book consists of two parts The first part of the book contains the presentation of the foundation of the theory of Bessel functions, the second part contains the applications. For the sake of convenience of reading, the first part is divided into two chapters. In the first chapter, which contains comparatively simple material, we first consider the main properties of the solutions of the homogeneous Bessel equation, which are based on the representation of the solutions in the form of series in increasing powers of the argument. Then we give the Bessel integral, the Poisson integral and, in Section 19, some their generalizations (which can be skipped in the first reading because of their complexity). The Neumann addition theorems and the foundation of the theory of products of Bessel functions are considered. Questions concerning differential equations which can be reduced to the Bessel equation and the inhomogeneous Bessel equations are considered in detail. In this connection we consider here, in particular, the functions contiguous to Bessel functions. Moreover, We consider some integrals here whose deduction immediately follows from the properties of Bessel functions which are investigated in this chapter. ‘The second chapter can be skipped in the first reading. In this case one should return to this chapter only as the case may require. It contains the theory of definite and improper integrals as well as the elements of the theory of dual integral equations. Then the representation of the functions hy the series of Fourier—estel, Dini and Schlémilch is presented. Before this group of sections we consider the roots of Bessel functions. Then we consider problems concerning the inhomogeneous Bessel equations which are more complicated than those considered in the first chapter, which lead to Lommel functions in two variables. From the same viewpoint wwe brieily discuss some facts concerning the partial cylindrical functions. In the conclusion of the chapter, we consider very briefly the asymptotic expansions of Bessel functions. ‘The second part of the book devoted to the applications also consists of two chapters. In the first one we consider problems concerning plates and shells ofPREFACE ix rotation. Among them some problems on the oscillations of a circular plate and ‘on the equilibrium of a plate on a Winkler-type foundation are considered, as well as problems which can be reduced to the use of the method of compensating Joadings. The method of initial parameters is considered in detail in the context. of the problem of the axially symmetrical deformation of a circular conic shell. In the second chapter we collected comparatively different problems concern- ing qucations of the theory of oscillations, hydrodynamics, heat theory and others. Section 11 of this chapter stands apart. In this section we consider the boundary problems whose solution can be reduced to singular equations. Although we for- mally consider plates and membranes here, the solution has a more general nature and is less connected with concrete applications. If the reader wishes, he can read the second part of the book independently of the first one. In this case the Teader can treat the second part as a continuation of the introduction, and look in the first part only to make inquires ‘The author conceives it his pleasant duty to express his deep gratitude to Pro- fessor S.G. Mikhlin, who has attentively read the manuscript, for his very valuable notes and, not least measure, for the great attention he dave this work and advice given when discussing it. ‘The author sincerely thanks his disciples, A.I. Tseitlin for the notes produced after reading the manuscript, L.M. Reanikov and M. Ya. Volotski for help during the preparation of the manuscript for publication and verifying the numerous formulae. B. KorenevIntroduction In many problems of mathematical physics, whose solution is connected with the application of cylindrical and spherical coordinates, the process of separation of the variables results in the differential equation ad’u du eee dee eet ay which is called the Bessel equation, and its solutions are called cylindrical or Bessel functions. First, we shall show in some examples what kind of problems results in the Bessel equation, and then we shall consider different. properties of Bessel functions and their applications. Consider small free oscillations of a thin homogeneous circular membrane. The differential equation of the membrane oscillations has the form Ow _ Ow low, 1 Hw 2a oF tr ort Og where w is the membrane deflection; t is the time; r and y are polar coordinates; 1/c? = p/N where p is the membrane mass relative to the unit of the area; NV is the membrane tension. We apply the method of separation of the variables, setting w= ulr)T()8(9); (0.3) then from 0.2 we immediately obtain T(t) = sin(at +A), (py) = sin(ng + B), where @ and A are the circular frequency and initial phase of the oscillations, respectively; n is the number of nodal diameters: B is a constant which can casily bbe expressed via the angle between the first nodal diameter and the polar axis; u(r) is a function of the variable r to be determined. After substituting (0.3) into (0.2), we obtain the following equation with respect tow: (0.1) (0.2) (0.4) 2, 2 fey tees ( %) eno. (0.5)2 INTRODUCTION In the introduction we denote the function u = wp(é) and call € the argument and 1m the index. In the problem under consideration € is a real number and the index nis a natural number; non-integer indices will be denoted by v. Further, we consider the problem of the oscillations of a membrane which has the form of a cizcular sector with a central angle y under the assumption that the membrane is inflexibly fixed along the whole contour. ‘We shall seek the solution of equation (0.2) in the form w= w(r)sin(at + A)sin he), #=0,1,2,..., where k is the number of nodal radii lying inside the sector. If v non-integer, then the problem reduces to solving the Besse] equation @u, iduy ete le (k-+1)/7 isa (06) with a non-integer index v ‘As a next example of an application of the Bessel equation we consider the prob- Jem of integration of the Laplace equation in cylindrical coordinates. Such problems occur very often in mathematical physics. For instance, considering the problem of heat conduction theory in the case when the temperature field is stationary and there are no heat sources, the temperatute T(r, y, z) satisfies the equation OT eT lor 1 er Tat oa or oe ol ae (on It we seek partial solution of equation (0.7) in the form Ta, = e7* sin(np + B)un(r), (0.8) then, substituting (0.8) into (0.7), we obtain for the function tia(r) the equation Pun a Let us consider the problem of the determination of the displacement w(r, ) of a membrane which lies on an elastic Winkler-type foundation (this means that along with the loading, the membrane is subjected to some reactive forces which are proportional with a coefficient kg to its deflections). The differential equation of the elastic surface of the unloaded part of the membrane has the form Vw w=0, A= Vko/N, (0.9) Ou tO. lo wa yt Pts ( aati Boe) where NV is the tension of the membrane; ko is called the bedding constant of the foundation material. If we write equation (0.9) in the polar coordinate system and asoume that the membrane occupies a circular domain, then, seeking a partial solution in the form w,_ = tn(é)sin(ny + B), we obtain for the function up the equation Puy Te (0.10) where €INTRODUCTION 3 If we introduce a new variable z = i€, then equation (0.10) will be transformed into the Bessel equation with an imaginary argument. ‘The problem of equilibrium of a floating plate which has been considered by Hertz, can be reduced to the solution of the equation VVw+Mw=0, A= V9/D, (0.11) where 7 is the volumetric weight of the liquid, D is the flezural rigidity of the plate. If we seek a partial solution of equation (0.11) in the form tum = un (€)sin(ng + Ba), then, as can be easily verified, the function un(€) can be represented in the form tn (€) = tayn(€) + Uain(€, where one of the functions in the right-hand side satisfies the first of equations (0.12) and the second function satisfies the second equation? Pun dun n? Ses ite (Eai)uso (0.12) Thus, the solution of this problem results in Besse] functions of a complex argument Evi. ‘One can find many other different applied problems whose solution connected with integration of the Laplace equation, the wave equation, the heat equation, a system of wave equations and so on, in cylindrical, spherical, and conic coordinates results after the separation of variables in the Bessel equation. Some differential ‘equations of cecond order with variable coefficients can also be reduced to this equa- tion. In the cases when there are sources distributed over the volume, the problems can often be reduced to integration of the inhomogeneous Bessel equations. A series of such problems will be considered in the second part of this book. 1 This problem in detail will be considered in p. 185.Part 1 Foundation of the theory of Bessel functionsCHAPTER 1 The Bessel equation. Properties of Bessel functions 1s The Bessel differential equation. Application of power series. Cylindrical functions of the first kind Consider the Bessel differential equation with the index v Pp, ote G teu a0. (a) Equation (1.1) is a linear differential equ integral can be expressed in the form of second order; hence, its general ‘u(z) = Ciwi(z) + Caua(z), where u; (2) and ta(z) are linearly independent partial solutions of equation (1.1). Suppose that z and v can admit any complex values. In the cases when the index v is an integer, we shall denote it by the letter n. If the argument = is a real number, then we shall denote it by the letter z. We introduce the Hessel operator of index v a da aa tat (12) and we rewrite (1.1) in the following form: Vou = 0. ‘We shall seek a solution of equation (1.1) in the form of a generalized power series in increasing powers of the argument z u(z) = s Ome, (1.3) mao where 09 7 0. Let us determine o and the coefficients am of series (1.3) here. For this purpose we find the first and the second derivatives of (1.3) W(2) = Yo am(m+a)e™o-, c=] w"(2) = So am (m+ a)(m-+a— 1)emter? mao and substitute the series obtained into the left-hand side of equation (1.1) instead of the function u(2)5 1. THE BESSEL, EQUATION. PROPERTIES OF BESSEL FUNCTIONS Collecting the terms containing equal powers of z, we obtain the following Vu = [ala ~ 1) + @— v7 Jagz” + [(0 + Ia t (a + 1) — v7 Jayz** + {[(@ + 2)(a + 1) + (a+ 2) — v*Jaz + ag}z**? +. = (a? —?)29 4 os((a + 1)? 7} + Yfenlle-t mm)? = 27] 4 am-a}et*™, ay Since V_u = 0, it is evident that all of the coefficients of 2#*” in (1.4) should ‘be equal to zero. ‘This condition gives us an infinite system [e+ 1)? -v7Ja = 0, V)+am-2=0, m=2,3, For even and odd values of m we obtain, respectively, the following two systems of equation: (a? ~ Gm{(oe +m)? (a? —v?)a9=0, [(a +2)? — v7 Jaz + a0 = 0, [(a4+4)? ~V7Jug + a2 =0,..- [(a+1)?-v7Ja, =0, [(2+3)?- Jos +a; = (1.8) Js - (6) Since ap # 0, the first equation of system (1.5) implies the equation a? ~ which is called determining; this equation implies that a = tv. ‘The second and next equations of system (1.5) allow one to express the coefli- cients ay, with even subscripts via ag and result in the recurrence relations a peat a Hert “~~ ath “Te thera’ Equations (1.6) are satisfied, if we set a, as 0. Setting a = v, we formally obtain the first partial solution of equation (1.1) in the fora a = 2 a wate) meee ft Dwr ! Paws her ECC ] Bvt NV+ HQU+3) a for a = —v we obtain the second partial solution tg(2) = az” [i - +I +8) 6 aaa}: Usually, the constants ao and af are assigned the following values: :: sau 1 O=eTeTy OO TRF)1, THE BESSEL DIFFERENTIAL EQUATION ° Using property (3) of gamma functions, one can rewrite the series obtained in a more compact form?, ‘The first series u,(z) defines a function which is called the Bessel or cylindrical function of the first kind of the Bessel function of indes? v of the argument z @ 5 Ser (17) Tut m+) JA the second series defines the Bessel function of the negative index —v ae () ys - (:/2)" | +m i) 8) It can easily be shown that the power series obtained converge on the whole plane of the complex variable z (except, may be, the point z = 0) and admit term by term differentiation. Actually, the series for J.(2) converges absolutely and uniformly in any bounded domain if variation of the index v and in any closed domain of variation of z (if Rv <0, then at z =0 the function has a singularity of the type =®*; therefore the origin is not included in the domain). The validity of this assertion follows from the fact that for |v] < N and |2| < d the absolute value of the ratio of the next erm of dhe series Wo the previous oue is Teas than one 22/4 es Leal = sat oe if m is greater than the positive root of the equation m? — mN — d?/4 = 0, which does not depend on v and z. ‘The Weierstrass criterion proves the indicated property of the power series. Hence, the Bessel function J,(z) is an analytical function for all values of v in a neighbourhood of any value of z, except, may be, z = 0. This, in particular, implies that the series obtained can he term by term differentiated and integrated Obviously, for a non-integer index the functions J,(z) and J-,(z) are linearly independent. If v = n is an integer, then I'(n) = (n ~ 1)! and the function Jn (z) can be written in the following form: (crynztmin nz) = x Beton Fa as Let us show that in this case there exists the following dependence between the functions J,(z) and J-n(z) é va Teale) = (-1)"Ja(2)- (1.10) ‘ Indeed, if k is a positive integer or zero, then; as itis indicated in the Appendix (265), P'(—k) = 00. Therefore, for the positive integer » = n each of the first n terms of series (1.8) vanishes; rewriting (1.8) starting from the fone term, we ‘nformation on gamma functions is given in the Appendix; all references on the’ formulae of the Appendix contain a number consisting of one integer; more detailed inforttation Goncerning gemma functions can be found, for instance, in [4], [58]: 2Sometimes, one ues instead ofthe term tindex” the Equivalent term “order10 1. THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS obtain 1)" (2/2) + i(-n+n41) * (w+ipl = [(2/2)" _ (2/24? acy [et "i+ (—1)"Jn(z)- (1.11) 2. Cylindrical functions of the second kind (Neumann functions) In Section 1 we showed that for a non-integer index the general solution of the Bessel equation can be written in the form = ByJp(z) + BaJ-o(2). Obviously, in this case the function = BsJy(z) + BaZ(2), where Zz) = CrJu(2) + CoJ-v(2), is also a solution of this equation. Here By, B2, Bs, By, Ci and C2 do not depend on the argument 2; C2 #0. If we set Cy = cot and C) = —cscum, then we obtain a function which has been introduced by Weber and is denoted by ¥(2): Jez) cosy = J-v(2) Ya = eee (21) In the literature this function is often called the Neumann function and is some- times denoted by N,(z). The function ¥, (2) is also called the Bessel or cylindrical Junction of the second kind of the index v of the argument z. For an integer value » = n the right-hand side of (2.1) is an indeterminacy of the type §. In order to find ¥;z) we remove the indeterminacy by the L’Hospital rule and we define Bevel 2) cosyn — Jov(2)] Ya(z) = lim + (2.2) La eee ] onton Inte) (m3 +6) 2 F oma ot (2) ym(e/aytm (ET Ba a {= iki}. es) where C is the Buler constant; it is approximately equal to 0.5772187. ‘The Neumann function ¥,(2) and the function J,(z) form a fundamental system of solutions to the Besse! equation for any, including integer, value of the index. Let us deduce formula (2.3). It is sufficient to obtain this formula for the special case n= 0 (then, using the recurrence relations considered below in Section 6, one can obtain the result for any n).4. CYLINDRICAL FUNCTIONS OF A PURE IMAGINARY ARGUMENT n From (2.2) we have Vole) = tim HEeVoosve = Jeol _ 2 ese = a Poe aE @sinve ea pg differentiating (1.7) term by term with respect to v and using formulae (9), (11), we obtain 2 Yo(z) = FJo(2) (In +0) yk 72) 3. Cylindrical functions of the third kind (Hankel functions) | (24) Any linear combination of the solutions obtained in Sections 1, 2, is also an integral of the Besse! equation. Consider the functions (3.1) HQ)(2) = J(2) -¥(2), which are called cylindrical functions of the third kind. They are also called Hankel functions of the first and second kind, respectively. One can see from (3.1) that there exist dependencies between Hankel, Bessel, and Neumann functions which are analogous to the connection between the expo- ential function of an imaginary argument, cosine, and sine. ‘The Bessel function of a real argument in the problem of integration of the wave equation is an image of a standing wave, and the Hankel functions give an image of a spreading wave. Therefore, it is obvious, what important role is played by Hankel functions in applications, especially, when studying wave processes in unbounded domains. 4. Cylindrical functions of a pure imaginary argument In this section we consider the Bessel functions of an imaginary argument z= iz. After changing the variable in (1.1) by the formula z = iz, we obtain the equation (aay One of solutions of equation (4.1) is the Bessel function-of an imaginary argument, J.(éz) defined by expression (1.7) after replacing z by ix in this expression, The function Jq(iz) is real for any even n and takes imaginary values for odd n. In order to get rid of this inconvenience one introduces, usually, in consideration the modified Bessel function La) = eV" Giz), (42) which is real for any v. The expansion of the function [,(2) into a power series has the following form co (a /aymey daa)= ater +) (43)2 1. THE BESSEL EQUATION, PROPERTIES OF BESSEL FUNCTIONS ‘As the second integral oteaaton (4.1), usually, the following function is taken K(x) = grie’™/? HQ) (iz) (4.4) which jo real for any real value of he index. ‘This function is called the Macdonald function; for integer v = n the power series expansion of this function has the form 1 -nt2m, Aye) In Z + 3 yeu Or (oc +en! LE SRS be- where, as before, Cis the Euler constant. This a, can easily be obtained, if we introduce J,(z) and Y,(z) into (4.4) with the help of the first of formulae (3.1) and afterwards change the argument z by iz in formulae (1.9) and (2.3) and express Jniz) via In(z) using formula (4.2). ‘Note that the formulae given above for the modified functions can be applied, if wwe replace x by a complex argument z; however, in this case the modified function 1,(2) is defined in the following way: T(z) Lz) = 5, (ze 84/2), 5 | 7. Cylindrical functions with a hal integer index Setting the index v = 1/2 in the expansion for J,(z) and replacing the gammo- functions by their values according to formula (15), after elementary simplifications we obtain yet oF Ant) Vek ere 7 (ay Jil) = \Eeoee—bygeame (72) ‘The we use formula (6.4); setting v = 1/2, one can easily obtain Bale) + file) = 2d-apal2)8, NOTATION FOR FUNCTIONS OF HALF-INTEGER AND FRACTIONAL INDEX 17 (Eom: (7.3) Using the recurrence relations given in Section 6, one ean find the Bewel func- tion for any index of the form n+ 1/2, where n is an integer, and prove that for any positive integer n the following formulae hold: then we find that Japl?) (=1)r(2z)"4¥?_—d* sing ae eer): | Cuyn(azyrt¥?_ a (cos Fn-4l2) a or ( ). In the same way one can obtain formulae similar to (7.1)-(7.4) for the modified functions; in particular, hpbe)= Yeahs, Kyle) (7) With the help of recurrence relations one can also easily obtain formulae for other values of the half-integer index. 8. Some notation for functions of half-integer and fractional index. Airy integral Functions of half-integer index occur often in applications, in particular, in problems of mathematical physics solved in spherical coordinates. The method of separation of variables in these problems gives, as a result, solutions which contain as a term the product of the cylindrical function and the power function of the same argument. In this connection, different. authors gave different names to various products of this kind. A review of these results can be found in [7]. We give there the notation of Sommerfeld Dalz) = (dee) dosnt 1A cote) = (Fe) (Gy) eee ea, In many papers the functions VF. (Fran. (En are called the spherical Bessel functions of the first, second, and third kind, respec- tively. i Let us now consider the following differential equation UM(s) + e°U(s) =0. (8.1) ‘The solutions of equation (8.1) can be expressed® in terms of so-called generalized Airy functions U,(s,a) and Ua(s,a) U = Ci (s, a) + C2U2(s, 0) (8.2) 8See also p. 3218 1. THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS ‘These functions and their derivatives with respect to the argument ¢ are connected with the Bessel functions by the following relations: Usle.0) = (o+a-merrr (222) ee (2 ant) . 1yaep (243) t2)/2 Ua(s, a) = (a + 2)" +r (223) Vidicoan (i oe ). q anasaye (242) glossy2 2 Ui(s,0) = = (a +2)-Mo¥ or (thet 09 Jee xycon2) (5 ; sr) ‘ (ose (243) fo 2 ye U4(s,0) = (a + 2)" eon (248) rHAY/2 ersvtesn (age a) (8.3) For a= 1 we obtain the fonctions which are called the Airy functions: 6,1) = FCA) ig (jer th(s,1) = ara (5e), Usle,1) = var(3) ibys (3) : (8.4) U{(e1 3 (5,1) = v5r(3) 54-78 (2°") : ‘The name of the functions U; and Up is stipulated by the fact that they are closely connected with the Airy integral - Osan (207), [come #20) at, (6.9) namely [owe + xt)dt = inf [ -a(2%) = Final [owe ona = inj [uae (2%) +40(2%)] en In order to prove the validity of these formulae, we have to differentiate the integral (8.5) twice with respect to the parameter z and remark that is integral satisfies the following differential equation wo , (8.6) which is « special case of equation (8.1).9. WRONSKI DETERMINANT ao Let us give the following notation for the Airy integral 1 Pace tz41e) at=ai(e) (88) eo) and the adjoint function if [er (« - 3) tein (= + e)| dt = Bi(z). (8.9) 9, Wronski determinant If us(z) and us(z) are linearly independent solutions of the Bessel equation, then they satisfy the relation uu — uur = Ci/2, (9.1) where C} is a constant. ‘Actually, introducing u in the Bessel equation (1.1), then uz and multiplying the first equation by uz and the second one by us, then subtracting the first equation from the second one, we have after a simple transformation Betws — wu = which immediately implies (9.1). Note that the left-hand side of equation(9.1) which represents the determinant woul? (9.2) is usually denoted by 20(u1,u2) and is called the Wronski determinant or the Wronskian of the Bessel equation. ‘The value of the constant Ci can easily be determined, if we pass to the limit as : +0 in formula (9.1) and use the expansions of the Bessel functions obtained in Sections 1-3 Suppose that the equation has an non-integer index v and let us find the Wron- skian BW Julz), Javl=)) = Cale. (9.3) Note that if v is non-integer, then = Ay a= ei +o0%, 2/2" Jes Say + oe, as 2+ 0, where O(2%) denotes a quantity, whose ratio to 2? is bounded as z -+ 0.2 1. THE BESSEL EQUATION, PROPERTIES OF BESSEL FUNCTIONS Introducing these expressions into (9.1), we obtain I(z)IL (2) — JL (2) Sev (z) = ‘eeearca- D7 Fehr 1). al +0(2) = iioracy +0) = RE +012); (9.4) when deducing this formula, we used formula (3). Multiplying (9.3) by 2 and using (0.4), after passing to the limit as x 0, we obtain a= an hence, Iz) I! (2) — So(z)J-v(z) = (-28in mv) /(n2z). (9.5) Just in the same way, using relations (2.1) between the Bessel functions of the first and second kind, one can obtain Tol2)¥E (2) — JE (2)¥e(2) = 2/ (m2) (9.6) ‘The Wronski determinant for the functions J,(z) and K,(z) is equal to TA2)Ki(2) ~ 12) (2) = 1/2, (9.7) and for J,(z), HS (2) and H!"(z), H!)(2) we have, respectively, ou ne ~ Hy el) =) 9 (9.8) dHS(2) _ Ai ee da = a HY 2)— - HEN2) (9.9) Setting z = zV/i in (9.8) and separating the real and imaginary parts, for the functions uy(z), ¥(2), f(z), and gu(z) we obtain to) fe (2) ~ ole) 9b (2) = wi (2) f-l2) — 4 (2)a-(2), (9.10) vu(z)f5(2) + ty(2)ah (2) = ub (x)90(2) + vb (2) f-(z) + 2/(nz), (9.11) where the prime denotes differentiation with respect to the variable z. In many problems the Basset formulae are also very convenient. In order to obtain these formulae we need to use the Wronskian and the Bessel differential9. WRoNsKr DETERMINANT: a equation: JolaV¥il(2) ~ Yule) (2) = (012) See) ¥il(2) = ~vele)ate) = 2 (1 -4). (9.13) Jue) ¥2"() — Yale) = need (e1 Sh z)¥2"(z) - Yo(2)So"(2) = (9.15) JeUeNYoN(e) — YM) 2) = (0.16) Je \¥O(2) ~ ¥o(2) de) = (5) (0.17) Jeleo(e) — Weleda) = —B (AE BEF). 18) A formula, which establishes a relationship between cylindrical functions whose indices differ by one, has been obtained by Lommel and Hankel; this formula has the form Iu(2)¥o4s (2) Ivga(z)¥.(z) = ~2/ (x2), (9.19) Using the Wronski determinant and recurrence relations, one can immediately obtain similar relations for cylindrical functions whose indices differ by an integer; we shall call these formulae the Lommel-Hankel formulae. For example, JoleWVeral#) —Jeyals)¥(2) = AWAD (920) Jule) ¥o4al2) — Joga Wo(e) = SLADE 2, (0.21) Jte\Yealt) ~ Jopalz)¥o(2) = SUE NEF AFD) 5 842) (9.29) Denoting Jule)¥oam(2) = Joam(e)¥i(2) = Blvmi2), (9.28) meas we can immediately establish the following rewrrance relation Bley tty) = LEM Bless) — Ble m 42) (0.24) With the help uf (9.24) une can easily establish that the right-hand sides of the Lommel-Hankel formulae are polynomials in 1/2, which can be represented in then 1. THE BESSEL EQUATION. PROPERTIES OP BESSEL FUNCTIONS form met2me B(v, m; 2) A) k) i (vm; 2) = 2{- ‘em Tes )+(m— a3) Tle+ 4 } 2 oe ~ 2 sin Tat omatiena - =1de-n'TTe+4() ci Zain, (a28) where, (m-1)/2, if mis odd, ‘m/2, if m is even, Similar formulae can be obtained for the modified functions TA) Kvgile) + Loar) Kz) = 4, (9.26) T(2)Ky4a(2) = ogale) Kz) = aes, (9.27) Tle) Koga) + logal@)Kv(2) = awenesa , (0.28) Le MKvaate) ~ Ievale) Kyle) = weiter a039 , M4) (0.29) Furthermore, the functional equations of the Wronski determinant type, which contain functions with different indices are of interest; they can also be easily obtained with the help of recurrence relations. So, for instance, Tuz)Yon (2) — Jogi (2)¥u(2) = (9.30) Jee) Moaale)~ JosslVE) ay Joe ¥eeale) — Jugale)¥e(e) = -M ED 4 2, (9.32) Male) - daee)= BOS"), 252 (a3) ‘One can easily obtain similar dependencies for the modified functions, for in- stance, Tele) Kusa(2) + doan(2) Ka (2) = (9.34) 10. Bessel integral and Jacobi expansion Consider the power series expansions of the functions exp(zt/2) and exp (—3 Since ok10. BESSEL INTEGRAL AND JACOBI EXPANSION 23 the product of these expansions gives ofl nBE 730 550 (10.1) Setting s = m, r= m+n and recalling that prahsyy = 0 for n < —m, we rewrite (10.1) in the following form: 0 (-D]-ES ressenre sae [ iE¢ 1 (eymantimyn and by virtue of (1.7) we immediately obtain exp ii (- »I = slayer (10.2) Using relation (1.10), we represent (10.2) in the form eg 5 (« = | acy be 4+ (C1 dal 2) Setting t = + exp id here, we obtain exp(shis sin 8) = Jo(2) +25 Jou(#) co8 2nd a £257 Janna(e)ain n+ 9. (103) a ‘This equality implies cos(z sind) = Jos) +25 > Jan(z) cos 2nd, (10.4) sin(2 sind) = 25> Jongalz)sin(Qn 4-100. (10.5) 4 If we replace @ by 1/2 ~n in (10.4) and (10.5), then we obtain Oa Tene yen greet (10) sin(z cos) = 2) (-1)" Jengalz) c0s(2n +1) (10.7) ‘These expansions have been obtained by Jacobi and are called by his name. ‘They give a representation of a plane wave in cylindrical coordinates. Replacing 8 by ¢ in (10.4), multiplying the left- and right-hand sides by cos np. and integrating with respect to p from 0 to ™, we obtain ‘Jn(2) for even n, J costesin cos mpdy = 7 ca (10.8) 3 :™ 1, THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS Similarly, from (10.6) we obtain [aoesnsiinnede={0, . Bem™ aoa) Adding (10.8) to (10.9), we obtain that for any integer n 1 2 [ costesiny —np) dp = Jn()- (10.10) 3 “The integral in the left-hand side of (10.10) is called the Bessel integral; Bessel took this equality (more precisely, slightly different) as a definition of the function Jn(z). Note that for a non-integer index this integral does not give the Besse] function and is denoted by £ f costa —ssind) dd = Iu) (10.1) 3 where J,(2) isa function which is usualy called the Anguer function; the properties of this function will be discussed in Section 16. For n= 0 integral (10.10) is called the Parseval integral - 2 foc mats (aoa) ‘The deduction of the formulae given above illustrates in a sense the character of applications of the Bessel integral. It arises naturally in cases when one should pass from a solution of the Helmholtz equation in Cartesian coordinates to a solution in polar coordinates. Below we shall show that integrals of the Bessel integral type can also be obtained when passing from other coordinate systems to the polar system. In applications the Bessel integral also appears when passing from polar coordinates to Cartesian ones. One of typical problems associated with application of the Bessel integral is given in the second part (see page 248) Consider once again the Parseval integral. Let us take a partial solution of the Helmholtz equation Paes Bet aya tua in the form u = cos Af = cos(AReos 8); here R and 0 are the polar coordinates of the point with Cartesian coordinates € and 9. Let us form the superposition of such solutions for 0 < @ < 2m, assuming that the €-axis rotates around the origin a u'(R) = | cos(AReos )d0. { Obviously, u*(R) is a solution of the axially symmetric problem for the Helmholtz equation without a singularity at the origin; hence, u*(R) = BJo(AR). Let us set10. BESSEL INTEGRAL AND JACOBI EXPANSION Fo R=. Since Jo(0) = 1, we have B= f" dO = 2x and Jo(AR) = F | cot areoe a. (10.13) In the case when u = sin Ag, we have Pa J sin0aRecos 0)d0 = BJo(AR) a Setting here R= 0, we obtain B = 0. One can also easily obtain a modified Parseval integral which only slightly differs from integral (10.12). Setting u = cosa€ cos 3n, where a? + 7 = d?, we have w'(R) = BdJo(AR) J costatecose) cos(#Rsin 0)d8. Obviously, B= 2x. Hence, Jo(/a? + BR) = J cos(arReoe 8) con(ARsin yao. (10.14) If we replace the function u(é,n) by an arbitrary function which is a solution of the Helmholtz equation and has no singularities in the finite domain under con- sideration, then rotating this function we also obtain up to a constant factor the Bessel function of zero index. So, if we use an elliptic coordinate system and set u(E,1) = ceo(€) Ceo(n) (see [37]), then, obviously, we have BJo(R) = J co(Reoso) Ceo(Rsin 6)d0. (10.15) One can continue the generalizations connected with the fact that one can inter- change the arguments of the functions cen and Ceo as well as introduce other indices, different from zero. ‘Thus, the Parseval integral (10.13) can be considered as the superposition of plane waves and integrals (10.14) and (10.15) can be considered as the superposition of waves of a more complicated structure. It should be noted that, using the Parseval integral, one can expand the Bessel function of zero index into a power series. For this purpose it is sufficient to repre- sent the cosine in (10.13) in the form of a power series, denote AR = r and change the order of integration and summation. ‘The Parseval integral can easily be generalized to the case when A = iy is a pure imaginary number. ‘Taking a partial solution of the equation Atu-Yu=0 (10.16) in the form uy (yr) = Acosh az cosh By = Acosh{ar cos g] cosh(fr sin y),2 1. THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS where r= /27-+47, y/z = tany, a? +9? = 7?, a and f are parameters with real values, we immediately obtain a Jol") ay | cba con) coshl pr sin yap (20.17) 3 Now we go from the Parseval integral to the Besecl integral. In what follows the restriction that the index n is an integer is essential. Consider the integral ay iL Avon n(0-— y)cos(arr con y) coa( jie sin p)dip a Acos nd ff corny cos(ar cong) cos(Br sin y)dp = uq( yr} cos n8. ‘The function un(7r) cos(nd) satisfies the equation V?u + 77u = 0. Hence, up toa constant factor, un(7r) is the Bessel function of the index n and the argument ‘yr. ‘Thus, for an appropriate choice of the constant A, the following equality holds A f cos npcos(ar cos p) cos(frsin p)dy = Ja(7"). 3 ‘The constant A can be obtained as a result of simple calculations and is equal to A=1/(2ncosné); 6 = arcsin(a/>). It is convenient to mention another superposition of plane waves here which can be used in order to represent the Bessel functions of non-integer index; its distinction from the previous one consists in the assumption that the axis along which the plane wave is moving rotates around the top of a circular cone and is always situated on the surface of this cone. ‘This result is given in the monograph [so]. Another heuristic method is due to Weyrich [84], who considered a cylindrical wave and its mathematical image — the appropriate Bessel function — as a erposition of spherical waves symmetrical with respect to the center, if the center moves along a straight line. The method of Weyrich admits obvious generalizations such as, for instance, consideration of the problem on the movement of a dipole or a source with abundance changing according to a sinusoidal law. 11. Addition theorems ‘The addition theorem for the Bessel function of zero index is defied as the equality dol V2 w= Beyc0sy) = 2)~ Im(2)Jma(u) cos my, (ay mao where the symbol )~’ differs from 5> by an additional factor equal to 1/2 for m=0,11. ADDITION THEOREMS ” Equality (11.1) is called the Neumann addition the- orem. We give its geometrical interpretation, assuming that z, y, and g are real numbers. The vertices O1, O, and C of the triangle O10C (Fig. 1) are called the first pole, the second pole, and the observation point, respectively. The lengths of the sides 010, OC, and O1€ will be denoted z, y, and z, respectively. The an- gles of the vertices O; and O will be denoted # and @. respectively. Obviously, z = Jz? +9? —2zyeony. ‘The left-side of (11.1) represents a standing cylindri- cal wave with the pole Oy; each term in the right-hand side describes a cylindrical wave with the second pole. ‘Thus, formula (11.1) allows one to solve the problem of decomposition of a cylindrical wave into cylindrical waves with another pole. A. proof of the theorem can easily be obtained with the help of the Parseval integral. We shall give a proof of a more general equality cos nvJn(V2?+ y — dzycos) = Figure 2" JIman(2)Jm(y)cosmp, (12.2) ‘ino where n is an integer. Obviously, this equality can be interpreted just in the same way as (11.1). Let us represent the Bessel function with the help of the Bessel integral In(z) 5 f cote sin 9 — n8)do This formula holds for arbitrary values of a. Let us take a equal to ZCO,0, then zsin¥ = ysing, zcosy = x— ycos y, hence, zsin(@ + ¥) = zsind + ysin(y — 0). Using the Jacobi formulae (10.6) and (10.7), making elementary transforma- tions and once again using the Bessel integral, we obtain oO OF Fe SE fumes =e YS Inlimgnlze™. ‘This immediately implies (11.2). :2 1. THE BESSEL. EQUATION. PROPERTIES OF BESSEL FUNCTIONS Let us note that for a non-integer index the Bessel integral on which the de- duction is based transforms into integral representation of the Anguer function (see 16). Hence, for a non-integer index the above reasoning gives an addition formula, for the Anguer function, in the right-hand side of which the functions Jingu (2) a7¢ replaced by the Anguer functions Jm4.(z)- ‘The addition theorems for the Bessel functions with an arbitrary index can be deduced in a similar way. However, in this case we use instead of the Bessel integral other integral representations (a proof of the addition theorems for the Beaecl functions with an arbitrary index is given in {7}, [4]). Now we formulate without proof the addition theorems for Bessel functions, when the index v is an arbitrary number. ‘The result which is due to Graaf, can be written in the following way: DY dam(zbdmtviem, (13) ee where # = arceotan £52s00¢ Changing the signs of y and ¥ in this formula, we obtain VI) = SD Jeam(2)In yen? (1.4) Formulae (11.3) and (11.4) imply that TAz)sRvd = Yo Imsv(z)Im(v)sinmy. (11.5) If we change signs of v and m in (11.5) and recall the relations between the cylin- drical functions of the first, second, and third kinds, then we can obtain: BAz\VO= Y> Zesm(z)Im(y)seimy, (11.6) where Z,(z) denotes a cylindrical function of the first, second or third kind. Re- placing 2, 2, and y in (11.5) and (11.6) by fz, éz, iy, respectively, one can easily obtain the following formulae: LR = LO (-1)™ram(z)m(vlssimg, (7) KA)S b= SD Kegmlz) en w)SStmy. (18) Formulae (11.8)-(11.8) are valid under the condition that |ye***] < |z|. One can easily deduce the addition formulae for the cylindrical functions of a complex argument. Replacing z, z, and y in (11.6) by zVi, xvi, and yvi, respec- tively, setting » = 0 and Z,(z) = H{)(z) and separating the real and imaginary11, ADDITION THEOREMS » parts, we obtain 25’ Um )\tim(2) = Im(¥)¥m(z)]eosmy for x < y, a1'eV9 =) = SU ot #) = O(n) m2) cs ms y 25’ Uml2)um(a) ~4m(2}im(Jeosmp for ®> x (11.9) a) 2 [em (¥) fo (2) + tm (Y) ma (2)] 808 mp forz y. [ (11.10) At the beginning of this section we indicated that formula (11.1) gives the ‘expansion of a solution of the Helmholtz equation in polar coordinates into a series in solutions of the same equation but for another pole. The physical meaning of the addition formulae consists in rebuilding the solution in polar coordinates under the necessity of changing the location of the pole. ‘This point of view on the addition formulae can be well illustrated by a num- ber of applications in which an axially non-symmetric solution of the Helmholtz equation for a circular domain is considered (see §§1, 3 of Part II) In order that the reasoning below is clearer from the physical standpoint, let us consider as an example one such application, namely, the problem on a circular membrane on an elastic foundation. ‘Taking into account what was said above, wwe give the deduction of the addition formulae based on using the expansion of a point force into a series in trigonometric functions and on applying the Wronski determinant of the Bessel equation. Consider the test Green function of the equation V2u—u = 0, which represents 1 solution of this equation bounded at infinity and with a singularity at the origin of the point force type, i.e. this function satisfies the condition ~ [Bar| jo Obvionsly, the function considered has the form zKo(r) Let us pass to another polar system and consider the observation point (R, 0). Let = r+ RP 2Rreow(0— 9) denote the distance between this point and the point (r,) at which the force is applied. We represent the function fCo(2) in a form mote convenient for problems associated with application of the second coordinate system. Consider the circum- ference of the radius r with the center at the origin of the second polar enordinate system. The unit point force is applied at the point (r,¥p) of this circumference.20 |, THE RESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS We expand the force into a trigonometric series Lisl 10-9) = 2 > (} + emtm(e- ol) a7 =D cosim(o- 9)] Accordingly, we represent the Green function in the form 1 a ay Kole) = Dy wm(R) coslm(d — y)] Obviously, the function wm(R) is a solution of the Bessel equation of an imaginary argument with the index m. It should be analytic at the origin and tend to zero at infinity. At R = r the derivative of this function should have a discontinuity such that 1 ==. (quay) dit di aR |, 7 dR We shall seek the solution n the form _ [Antn(®, RSF; ton(R) = (are, R>r. The condition of continuity of the function tm(2) implies Armlm(t) — Brn Kmn(r) = 0. Moreover, from (11.11) we have Amy (?) = Bm Kig() = 2. Solving this system and using formula (9.7) for simplification, we obtain Am = Eim(t), Bm = Eales hence, 252 Im(PVKm(r) cost gl, RSs Ko(2) =" (11.12) 2 Zi Ir) Km) conlen(9 — v)), Re2r. Differentiating (11.12) with respect. to the variable r, one can obtain a solution of the problem on a membrane to which a point moment of force is applied. At the conclusion of this section we give a formula which is called the Gegen- bauer addition theorem, which can be obtained from (1.1) by differentiating this equality n times with respect to cosy: ale) egies S~! Jetn(2) Iman (y) d” cos(m + np LP Taeongle there are no restrictions here on the ratio 2/y. (1.13)1a DIFFERENTIAL. RQUATIONS REDUCIRLE TO THE RESSEL EQUATION a1 ‘The Gegenbauer theorem for the cylindrical function Z,(z) has the form 2) oer) x Znszle) Ime (coo) (1) for \ye**| < [el. Here O%(cong) isthe coeficient of the term which contains in the expansion of (1 — Spices + 62)" 12, Lommel expansion Let us expand the function (z-+h)-“/?J, (Vz A), which is an analytic function of the argument (2 + h) for any its value, into a series in the powers of h, using formulae (6.8) when computing the derivatives. We obtain (2+ hy YP (SEER) = > pant erty, va} mao "24 om (Vz). (12.1) ‘The function (= 4 4)*!2J,(vz 7) is analytic everywhere, except the point =z. If |Al < [2], then (2+ A)? (VEFR) = LF a 21, (vA} =U em mV). (22) ao Setting v = 1/2 and v = ~1/2 in expansion (12.1), we, respectively, obtain (2) °F on (VER) = Fo naples za) mao ) an (vi +22) = > Shyam(e); W< mo For |h| < |2| we can replace the functions of the first kind in formulae (12.1) and (12.2) by the functions of the second or third kind. Both formulae are also valid in the case when the index is integer. Some additional results concerning the Lommel, expansion are given in 14 and 16. (12.4) 13. Differential equations reducible to the Bessel equation In this section we consider the problem to which differential equations can be reduced to the Bessel equation and how the necessary transformations should be performed. This problem is treated from the practical standpoint and the presen- tation which follows the settled tradition can hardly be called systematic. ‘This representation is more likely a collection of the most successful and interesting re- sults ordered in incteasing of their complexity. The most part of these results is due to Lommel. Many differential equations can be reduced to the Bessel equation using trans- formations of dependent and independent variables. We begin with the simplest, special case.2 1. THE BESSEL EQUATION, PROPERTIES OF BESSEL FUNCTIONS Consider the equation - (13.1) i.e, we obtain the ordinary Bessel equation. Hence, » = Z,(2), where Z,(z) is the cylindrical function of index v. As a result of the inverse substitution, we obtain, uaz Zy4spalic2). Setting u = we"? in (13.1), we obtain as a result of the substitution the equation (13.2) whose solution has the form w= Pu= PtN2Z,,,jo(iez). (13.3) Setting z= ¢#/g and g = 1/(2p + 1) in (13.2), we obtain the equation tw = 0. (13.4) ‘We find the solution of this equation replacing the argument z by ¢#/q in (13.3). Thus, w= C21 /a) MOD 2, agp (5064/9) Equations of a larger class can also be reduced by a direct transformation to the Bessel equation. In particular, Lommel has shown that the general solution of the equation PES + (20 2604 De 4 We + oa 260)] =0 (13.5) is = P87, (28). (13.6) In order to obtain (13.6) we need to introduce new variables w = uz*-*” and {= 724. It can easily be shown that the function w(t) satisfies the Bessel equation with index v.13, DIPPERENTIAL EQUATIONS REDUCIBLE TO THE BESSEL EQUATION 38 Let us give some special cases of equation (13.5) a 2 : Sheen, uae *Zyy (3#"). wip (Re); (13.7) Loe a (tony ee (tay Green 0, une (1), = P25 (Felt) 5 (13.8) u Maen ft a-yit—tuno, us eP2,(iV2; (13.9) a ifs Bit E+E (Spero, u=ZV3; (13.10) Gat hgiea(2 -§)a=0, u=2,(2); (13.11) yt : Set ee tys0, yee (#2); (13.12) @y | aad a > oe + (84 © 3) y=0, y= 2*Z ppr=z(B2). (13.13) Now we shall deduce a more general result. which is also due to Lommel. For this purpose we consider equation (13.2). Replacing in this equation w by y/x(z), = by U(2) and 2p by 2v~1 and setting c = i, we obtain vt) ¥@) , 9X8)] av wey ter Dg +A | at wt) YC), 9x’2)] XE) _ x") i +{ way +O Dy ot) x xa) ey tH ta }y= 0 (18.44) “The solution of equation (13.14) is y= xl)" 2192). Introducing into the consideration a new function (2) which is defined by the equation ee) _ ve) le) ~ #e) com and eliminating the function x(2) from (13.14), we obtain ay _ ole) dy + {fey - rea) 9X) et (13.15) ape]? dz? plz) dz” Lal g(2)} ~ 2 ol) (2) +350 [we . i Fo "Yu=o. (03.16) ‘The solution of equation (13.16) has the formau |. THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS In the special case, when (2) me) _3/y"t)]* eon. a (09o-abal *e—*+a] Bal fee oun ‘The solution of equation (13.17) is Coming back to the initial equation (13.2) and considering the case x(z) [w(z)]?"", we find that the solution of the equation oe (mu 4] 4 +=’ reves =0 (13.18) has the form v= [w(2)*Z.[v(2)] A special case of (13.18) is the equation @y ge git -v)y=0, (13.19) whose solution is the function y=Z(e). (13.20) ‘The following equations can be reduced to the Bessel equation with the help of the Lommel transformations*: re) £4 + [Pexpieai ~ Py aay, [yagrer aS e-2) ata +6 aes e]y=o ots [G53 oe Pe 7: fu, (ya) (MEM oy ( +) 2+(F+ ye e(So) ele ot (ed) Bien 2 = exp[-hz/2}z"-9/?Z, amt 1, a (is ) = expl-he/2] Z9(26V2); 2, oy oo mobos de ete (i hr) ¥=% y= Zplcinh 2); aye ( # ) 7 i aot teeea tee (I meee) Y= v= SelvTEES. “These equations are taken from {80], where Appendix contains a large and useful list of similar equations.13, DIFFERENTIAL EQUATIONS REDUCIBLE TO THE BESSEL EQUATION 36 Now we are going to consideration of equations of the forth order. Note that in many problems of mathematical physics the method of separation of variables gives, as a result, ordinary differential equations of the following form: @ ia Alutadu¢iuso, o= 2414 _% ee de ade If the equation has the form Alu-2bAut Mu=0, (13.21) then, passing to the variable £ = Az and keeping the same notation for the operator Seige ; AL (a= fet hh— 4H), we obtain Ay Ayu 2od,u+U=0, by = 1/9? (13.22) i: Let us note that Ay = a solution of the form ‘Vy —1, where Vz is the Bessel operator. We shall seek = AZAEVa) and we introduce this expression into equation (13.22). As a result, we obtain the characteristic equation s? + 2bs + 1 = 0, whose roots are $1.2 = —bo + /63— 1. If bo is equal to zero, then 61,2 = +ti and the case occurs which is considered in detail on page 162. If bq is non-zero but small with respect to one, then we can approximately assume that 1,2 © —bo i. In the more general case, for 0 < by < 1, we have S142 =atib, where a = —bo, b . Setting Vaz 1 = pe and squaring this equation, we obtain (a ib) = p°(cos 2p + sin2y). Since |a| < 1 and [6] < 1, setting =p, a=cosdy, b=sin2y, where y = 4 arccotan ¢, we can write the solution in the following form w= Ape (nel#) + AZHL (pel) + AZ Le (pe'?) + AGI (pe*#),, w= Artin (y) + Ariv(p) + Ashu(p) + Aaiu(p), where A i + A3, Az = i(Aj — A3), As = A+ Ai, iA} ~ AY). If bo = 1, then the characteristic equation has multiple roots. ‘If62 > 1, then both roots of the characteristic equation are real and the solution of equation (13.21) can be writtea in the following form. for by < -1 = A (EVai) + Aa¥i (EV a1) + Aste (E02) + Aa¥o(EV a2), for by > 1 Arlo(EV Tail) + Ak (EV Toul) + Asto(EV Tal) + AaKe (EV Teal).36 1. THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS Now let us consider problems in which the transition to a system of Bessel equations is less obvious. We give a result which has been obtained by Lommel and afterwards generalized by Watson. Equation (13.5) can be written in the form (D1 a)(D 4 a 26x)u = —62?2u, (13.23) Let us write the equation T[wws e— 29040-26298) which for n = 1 coincides with (13.23). It can easily be verified that a solution of equation (13.24) has the form ua P22), (-1)"89 227" u, (13.24) where y= cexp(rmi/n), or ‘Assigning to r the values 0, 1, . form a fundamental system. Setting n = 2 in (13.24), we obtain an equation of the fourth order (D+ a)(D + a —28)(D + a — 26v)(D +a — 2pv — 2f)u = fictzu. (13.25) Kirchhoff has considered the equation £ (8) au vax which describes the transverse vibrations of a conic rod. Introducing a new inde- pendent variable # = 2/7 and denoting uf? = 7, we reduce equation (13.26) to a system of two equations Ol. a-k 1m Leuecessively, we obtain n eolutions which @y dy 4 ap * pap* which implies tne + A2¥2(2V2) + Asha(2V2) + AaKa(2v2)] . (13.27) ‘The equation cousidered by Mononobe @ @u rd (: +) =m (13.28) has a solution which can be obtained, if we set the indices in (13.27) to be equal to 1 and replace 1/7 by 1/-/. This equation can be encountered in problems on vibration of a wedge. A more general problem has been studied by A.N. Dinnik [12], who considered the differential equation of rod vibrations which has the form Cees ($3) -" u=0. (13.29) Obviously, the Mononobe and Kirchhoff problems can be obtained as special cases, of Dinnik’s problem. We write the solution of Dinnik’s equation in the form us zt? {Ay Jma2(2V2) + A2Ym—2(2V2) + Aslm=2(2V2) + AaKm—2(2V2)}14. THE POISSON INTEGRAL a We also formulate a result recently published which is due to T. Lardner (32] who has shown that a solution of the differential equation Peau (- ) ~r can be expressed via Bessel functions, if m = P/Q, n = (P + 2)/Q—4, where P and Q are natural numbers and Q # 0; moreover, at least one of the numbers |P| and {QI is odd. Lardner does not give the general solution in (32), emphasizing its complexity. Hee considers only the special case P = 9, Q= Pe o [Ds(Ds — 4)(D, + 6)(Ds + 10) — e**Ju(ey (13.80) Setting = In(4z}/*), we can rewrite equation (13.30) in the form (13.31) where Dy = d/dt. ‘The solution of equation (18.81) has the following form u(t) = e7!"Di(Dy — 2)(Dy ~ 4)?(Ds ~ 8)°(Ds ~ 12) (2), where ®(t) is the solution of the equation [R(D1 - 2)? — eM} (4) which can be expressed via the Bessel functions, (t) = Aido(et) + A2Yo(e*) + Aslo(e*) + AaKole'). ‘The equations of the sccond and fourth orders considered are homogeneous, Using the Cauchy functions given in Section 17, one can easily obtain the solutions of the corresponding inhomogeneous equations. 14. The Poisson integral Parallel with the Bessel integral in the theory of Bessel functions and its appli- cations another integral representation plays an important role. This representation is due to Poisson who in his papers on heat conduction considered integrals of the type J cootzcos0)sin?™* 040, a [otrornsnmea where a positive integer or aero. Later on, similar integrals were considered by Lorimél.’ Let us consider the Poisson inteqral " , (2/2)” Fos 126072) Ju(z) f costecose)sin¥ ade.» (41) 238 1. THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS: In order to-prove (14.1) we replace the first factor of the integrand in the right-hand by the series eaten Scene ean, ae then we change the order of summation and integration and use formulae (7), (8). ‘Asa result, we obtain (yr ; Tears | omeson soa (z/2)" = Fw iaTED 0)" rama [or asin 000 =F MTs ree ayn 1/2) ~ LP 41/200 /2)F@m+ 1) Tmt y+) ee (-1ym(z/2)3m4¥ = Liter ino se ry =O) (04.3) that prove the validity of (14.1). In contrast to the Bessel integral, the result obtained holds not only for an integer index n but alse for any real v > —1/2 a8 well as for complex v provided that Rv > —1/2. ‘Taking into account the equality sin(zcos0) sin 940 = 0, (44) we can rewrite formula (14.1) in the following form: (2/26 Tories | el 098 sin™ 90, (14.5) Jz) ‘After making the change of the variable cos = t, we can write the Poisson integral in the form (2/2 Ile) = ees Ear | eB ae (046) a In Section 19 we will consider some generalizations of this integral Asan example of using the Poisson integral we give the deduction of the Lommel expansion considered in Section 12 for v = 0. Let us note that before obtaining the general solution given in Section 12, Lommel considered a special case of the expansion when v = 1 (see, for instance, [7|). The deduction given below is based on the obvious fact that the Poisson integral (10.12) does not change when the14. THE POISSON INTEGRAL 39 origin of the polar angle is shifted, hence, it can be written in the form f costrcoe@ - eid = | cosircosteas+ rsind sin edo 1 Jol f [cos(r cos @ cos y) cos(rsin 9 sin y) — sin(r cos @ cos y) sin(r sind sin y)] dp = £ cour cordon) cos(rsin@sing) dp, 2x >0>0. Let us note that a similar formula was obtained before, in Section 10 by another method (sce (10.14). Expanding the accond factor of the integrand into a power series, changing the order of integration and summation and using the Poisson integral, we obtain Jo(r) = > § fonteontens ayn eens a ap E— [-1)"(rsin gy?" 2m ET | cos(r cos @cosy) sin” pay (2m)! [ indy?" T (m+ 2) TQ) Jm(rcosd) (reos8)™ (1) (rsin 8)?” Jm(r-cos 8) :y mniz™—— (rcos8)™ Denoting rcos@ = /2, rsin@ = Vi, r= Vz +h, we find (14.7) (14.8) From this formula, using formula (6.8) and the formula for differentiations (Pda (2) aaa ols) = . ae aor (2) = ae which can be easily deduced from formulae (6.2) and (1.10), we obtain for integer values of n after simple transformations ectmton)/2 Cyst. ys eee Tal mn V2) (14.9) Expansions of a similar type can also be constructed for the integral J costzcota ~ pilin wap, (14.10) a“0 1 THE RESSEI. EQUATION. PROPERTIES OF RESSEI. PUNCTIONS, which is an obvious generalization of the Poisson integral (14.1) and when v = 0 becomes the Parseval integral ‘The Poisson integral multiplied by 2 can be considered as a superposition of degree v moments for the displacements of the plane waves type with respect to the axis coinciding with the fixed radius. ‘The series expansion of integral (14.10) similar to the Lommel expansion can be obtained in just the same way as we act when considering (14.1). Thus, vl cos{r cos(0 — y)]sin® ydp 3 = f coxivicos) cos(Vh sin y) sin™ yp dp 11(2) CMW her (vem 4). (14.1) If @ =0, then, obviously, we obtain the Poisson integral because in this case h = 0 and only the first. term remains in the expansion. 15. Some indefinite integrals First, we consider some indefinite integrals which are consequences of the dif- ferentiation formulae and relations following from consideration of the Wronski determinant and Uhe Basset formulae. ‘Recurrence relations and the differentiation formulae allow one to obtain the following indefinite integrals immediately: fe P2tyde= 22002), (15.1) [oP alesse (2, (152) In the case when z= pe~‘®, v = 0, we have J vie) do = ofa(o)cos + (0) sing, (15.3) J vito) dp = oén(o)cose — flo) sing. (15.4) One can replace f and j in these formulae by & and @, respectively. ‘The following integrals containing two Bessel functions or squared Bessel func- tions are obtained by Lommel with the help of using the Wronskian of the Bessel15, SOME INDEFINITE INTEGRALS a equation Sate ra (18.5) J SEG 3) = Tava zh. (15.6) J sags Th (15.7) / aa" eH. (15.8) Isa = (15.9) In order to prove the first of these formulae, we consider the derivative of Une fraction a [ae = Tevl)Iel2) = Fool 2)IE2) _ Ile) i) wad3(z)" (We have used formula (9.5) here.) Integrating with respect to z, we immediately ‘obtain (15.5); in this case we assume that v is non-integer. Let us note that we can obtain similar formulae for other cylindrical functions in the same way. Let us now consider the derivative of the logarithm of the fraction [m3 Jee P= Jule\douls) = Jeyle)Jo(e) ___2sinve @ 1" T@) (2) oul) © wed @)T-vl2)" Integrating this relation with respect to 2, we obtain for non-inceger v / dz ean G8) Weve) = Fame Ty" that prove (15.6). In the same way we can also prove formulae (15.7)-(15.9), where v is non- integer. Let us also notice that dz awe (a) laces Ba oo where v can be integer as well as non-integer. In the same way, but with the help of the Basset formula (9.13) we can obtain indefinite integrals which contain derivatives of the cylindrical functions 2 f(vfz*—1)dz _ Jh(z) ae FG)" ae 2 fav /e)dz _ ¥(z) a Eey (fe -1)dz_ | Ih i) 2 avy? de ve) : a) sapeywetey =" Fey" 2 (15.12) =In (15.13)© 1. THE RESSFL FQUATION PROPERTIES OF BESSEI. FUNCTIONS ‘The product of the right-hand sides of formulae (15.8) and (15.11) can be repre- sented with the help of the Basset formula (9.13) and the Wronski determinant in the following form: a] TuZ)YN2) = (= v7/2*)¥V(z)\Jv(z) 4, _ JEz)¥(2) © AEE)? Jelz)Ye(2)” (15.15) Obviously, we can introduce the functions H!)(z), H(z), J-v(z) instead of the functions J,(z) and Y,(z) in these integrals without any efforts. One can easily obtain similar formulae for the modified functions J,(z) and K,(z); for instance, (7/27 + Ide _ KL(2) aU()F = Tle)" cna / ale) KUCAP eo Teme U1) Let us consider the integral J elo) ae = 5 fo Zenon) dz = Enfen) + 2S te alee) - [2 rolonyae (15.18) Obviously, repeatedly applying formulae (15.18) to the integral f 2"Jo(az) dz, for odd n we can obtain an expression containing only the functions Jo(az) and Jh(oz); if n is even, then the result contains one more summand containing J dolar) dz Similar results can be obtained in the case when we have the functions 2"Yo(a2) and 2"H{ (az) under the integral sign. A more general formula [orate * Zoasle) + (n= WetZole) — 2 Vf 2°22) ds can easily be verified by differentiating and using formulae (6.3) and (6.4). Let us consider very important indefinite integrals due to Lommel which can be obtained as a result of using the different transformations of the Bessel equation given in Section 13. If y and 1 are solutions of the equations Sa + Ply =0, z cad +QAz)n=0, then the following formula holds dr ds [oer ome: =v nf. (5.19)16. FUNCTIONS CONTIGUOUS To BESSEL FUNCTIONS 4s Suppose that P(z) is defined by (13.17) for Y(z) = kz and y= ; Teapectively, Q(z) is obtained, by setting ¥(z2) = lz, = pa, k HU in (18.17). Then / {= P)z = (13 = 19) /2) 2, (ke) Zp, (lz) de 22 (Zt Zils) = p, we obtain f rAaleey7yeteya =a as {28s HE zee ately Grp ls (2) Zi (le) — 12 (2) Zhga(le)}, (15.21) wel yes}, (15.20) Assuming yy = for 1 = k we have the indefiniteness of the type 2. With the help of the L’Hospital rule we can obtain f %etuer7zloe) dz= [Link]) = Zy—a(kz)Z p41 (kz) ~ Zu4s (ke) Zpa(kz)}, (35.22) If we set P(z) = e*-” and Q(z) = imply that, [ee — YZ, (08) Zu,(e8 dz = Z(e2) +v* then formulae (13.19) and (13.20) Sale ) Sele) ~ Bie). * (15.23) ‘One can easily obtain similar formulae for solutions of Inhomogeneous Bessel equations (see Section 16) 16. Functions contiguous to Bessel functions. Particular solutions of the inhomogeneous Bessel equation In applications an essential role is played by functions which are particular solutions of the inhomogeneous Bessel equation Viu = 27p(2). ‘The simplest of these functions are those which can be obtained from the Besse! ‘equation whose right-hand side is represented by a power or linear function of the argument and which are called functions contiguous to Bessel functions. ‘These functions are considered in this section; in accordance with the type of the right- hand side and relations between the index of the equation and the exponent they. are called the Lommel, Struve, Anguer and Weber functions. Below, in Section 27 ‘we consider the Lommel functions in two variables and their particular cases: the probability integral and the Fresnel integral. Section 28 is devoted to a significantly more general and large class of functions — incomplete cylindrical functions; all these functions can be considered as particular solutions of the inhomogeneous Bessel equation whose right-hand side represents the product of a power function, and an exponential one. In Section 17 we consider two simple classes of particular solutions of the Bessel equation: 1) when the right-hand side represents a step-function;“ 1, THE BESSEL EQUATION. PROPERTIES OF BESSBL FUNCTIONS 2) when the right-hand side represents a Bessel function. It is expedient, before investigating different particular solutions of the inho- ‘mogeneous Bessel equation separately, to consider some general properties of these solutions which can be established in a rather simple way. Subsection 1 of this section is devoted to this question. 1. We consider here the following questions: 1) functional equations containing the Wronskian; 2) indefinite integrals; 3) the Lommel expansion; 4) integral representations which can be obtained with the help of the Green functions of the Besse] equation, the fundamental Cauchy functiona with the properties of unit matrix or by using the method of variations of arbitrary variables. First, we consider the Wronskians of the inhomogeneous Bessel equations. Let, uu, and up, respectively, denote solutions of the homogeneous and inhomogeneous Bessel equations with index v and the argument = Bo 140 2a ee ie @u; du; v Fo + 1s (1S) amt After multiplying the first of these equations by u2 and the second one by u; and subtracting the second equation from the first one, we obtain the following equation: d PAC Ce EO Denoting ty = uj — wu, we seek ty in the form ty p(z)/z. Then J set) dz, P(2)/2=-uvilz), plz) hence, oo (uve waeillg =—2 f suseleyae $ If the function u; together with its first derivative vanish at z = 0, then we have viva wang = =1 f smpaledas 3 Using the recurrence relations for the functions and up, one can easily obtain different, modifications of this equality so that the left-hand side contains functions of distinct indices which differ by an integer as well as to write the Basset and Lommel-Hankel formulae for the ingomogeneous equation (see Section 9). Espe- cially simple results can be obtained for the special case when the right-hand side of the inhomogeneous equation represents a power function. Ifboth functions tu and 1g are particular solutions of the inhomogeneous equa- tions V,u = 2%p(z) whose right-hand sides are equal to z%p,(z) and z%y2(z),respectively, then we have gus — uguale = f stuseale)— uapalel dz In Section 15 we give some indefinite integrals whose integrand contains the Bessel functions. Now we are going to show how one can obtain integrals containing functions contiguous to the Bessel functions in a similar way. Denoting [ suple)de = Aloe), 3 we have 4 (us) _ nity — uur _ uly) a \us, road Hence, similarly, Furthermore, 4. hence, i Axlp.2)d2 _ uu Different integrals containing functions contiguous to the Bessel functions can be obtained with the help of Lommel’s method. Suppose that a function u satisfies the homogeneous equation uf + Pur and a function up satisfies the inhomogeneous equation u + Qua = (2) which can be reduced to the Bessel equation. It can easily be verified that fe- Q)urue de uf? aed - [nse dz, From this equality one can easily obtain different integrals of products of cylindrical functions and functions contiguous to Bessel functions. Just in the same way, a5 the result of an obvious generalization one can obtain the corresponding integrals of two different contiguous functions. Let us give without proof several definite integrals containing a function &y,-(2) which will be defined below by formulae (16.10)-(16.12). These integrals can bea8 |. THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS easily obtained by using Lommel’s method which was applied by Lommel for cylin- drical functions and is presented at the beginning of Section 15: (uty —1) Jt Cle wa 1(2) = Jnnn(2)8nvl2) 9 Sev?) | Ze) ue): hee / Gtr Melee 2) Selena ge 2, (182) f WHY —I)Jo(z)5p-1y-a(2) — Jor(e)suyl2) 4, Jel J Suel=)de(2) a [3 7 Iy(z) dz a) = Inf(u + v — 1)zJ, (2) 5y—1,-1(2) — zJv—1(2)5p,v(2)}, (16-4) 2dy, (2)8yu9(02) dz & AJos(2)Su.u9 (02) ~ Spa (0z)Je, (2)] = a8 a(n + v4 — 1dr (2) 5 1(2) — Ju.-1(2)8(2)]- (16.5) ‘An integral representation of the function ,(z) which is a particular solution of ‘the differential equation fe 1d, (-S)e=00, Faz according to the arbitrary variables variation method, can be sought in the form (2) = Alz)Ju(z) + Blz)¥.(2). Tt can be easily shown that, for an non-integer index 4.0) = 7a fdtesvte)~L(es--0le(6) 46, 08.) : z> Zo, for an integer index n(2) ELJa(€)¥n (2) ~ Jn (z)¥n (ENCE) 48, (16.7) 2> 20.16. FUNCTIONS CONTIGUOUS TO BESSEL FUNCTIONS a In the cases when the right-hand side is a power function, the particular solution, contains the integrals AuvolleER) = [e+ du(leFR AEM, 3 which can be expressed in the form of a series being an analogue of the Lommel expansion: Angel VERB) = 5 Teng Auemanem(va. (168) Let us consider a partial solution ‘the inhomogeneous Bessel equation in some cases In Subsection 2 a particular solution will be obtained in the case when the right-hand side represents a power function. This solution can be expressed via the Lommel functions. In Subsection 3 we consider a special case when the exponent is greater by one than the index of the Bessel operator and the solution is expressed via the Struve functions. In Subsection 4 the Anguer and Weber functions are considered. These functions allow one to find a particular solution in the case when the right-hand side represents a linear function of the argument z. 2. Consider a particular solution of the equation kth, (16.9) where k and pare constants. One can easily see that the particular solution of this equation represented as a power series in increasing powers of the argument 2, has the form Vu a mts (as P=? © TF 1? v7 Fa) =v} eee 5 ("eda = be + nt bt) 24 Tepe Gat tm 8) = kbyy(z). (16.10) The function sy,»(2) is called the Lommel function. ‘The expansion (16.10) is, obviously, ineligible, if p-tv is a negative odd number. In order to obtain another representation of the Lommel function, we apply the method of arbitrary variables variation to equation (16.9). According to formulae (16.6) and (16.7) we define the Lommel function 6,.(2) in the following way: y= wl) = se [: Hi fetinee— tata fesse (16.11) for non-integer v and $uv(2) ; pt if #J,(2) dz — Jz) 2 2Y,(2) ‘ (16.12) for integer v.s 1. THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS If wv is an odd positive number, then the particular solution of equation (16.9) for & = 1 can be represented in’ the form of a Lommel faction Sp(2), which for non-integer v has the form paige eee) des ie Sul x [rs (ju = oD) ie) 008 (ju + vn) Jus)} (16.13) If vis an integer, then we have 1 1 L 1 1 1 a) = sees) POP (yp tea Er (ips at Spao(2) = Spelt) +2! v (ju 7 +3)r (juts +5 x fun (ju - ") Jo(z) — cos (fu - vn) xa] - (16.14) From (16.11) and (16.12) we can immediately obtain the recurrence relations which are duc to Lommel {7} Suprv(2) = 244 — (et 1)? — V }oy (2), (16.15) Snwlt) + ZSyol2) = (UY = Dpaiynrl2)s (16.16) awl) — 2a, (= = Mep-ayes(2)- (16.17) ‘These formulae immediately imply that w pow (2) = (Ht YM sp-ay—alz) — (Hv = Nya (2), (16.18) Doh 2) = (+ To paawnale) +H ¥ — Uapaeral2) (16.19) Notice that the functions 5,,.(z) in (16.15)-(16.19) can be replaced by Sy,v(2) ‘The case when 1 v ie an ‘odd negative integer, requires rather complicated reasoning which can be found in papers of Watson. In the literature one can meet, works where the integrals f Ai(t) dt and Bi(t) dt (see Section 8) are considered which can be easily expressed via the Lominel func- tions. From the practical viewpoint, the Lommel functions are very important special functions, because a great number of problems of mathematical physics can be reduced to the Bessel equation whose right-hand side represents a power function. ‘The Lommel functions allow one to obtain a particular solution in a convenient, form in the cases when the right-hand side of the inhomogeneous Bessel equation is a piecewise continuous function representing a polynomial on every interval of continuity. 3. The Struve function with index v is usually defined by the expression 22/2)” He) = reninram { — PY gin(t) dt en 22/2) sin(zcos 6) sin” 0 d0, (16.20) = Tes 1/9ra/a j where t = cos),16, FUNCTIONS CONTIGUOUS To BESSEL FUNCTIONS o Under the condition Rv > —1/2 the function H(z) can be represented by a power series HL) = i (21 "(2/2424 aa $ T(m +3720 (V+ m 43/2) We can easily prove this fact, by expanding sin( cos) in (16.20) into a series with respect to increasing powers of z and then making the calculations similar to those given in Section 14 where we considered the Poisson integral. Using this expansion, we can obtain the following recurrence relations: Boa) + Mogae) = Sa) + ee (16.22) Hy) — Hoga(=) = 28842) — whan (16.28) (& + A) H(2) — Hy-(e), (16.24) (€- 2) =~ a He. (16.28) which are, obviously, special cases of the corresponding recurrence relations for the Lommel function. ‘These formulae imply that the Struve function satisfies the inhomogeneous Bessel equation apy TORT F1/2)" where the right-hand side is a power function whose exponent exceeds the index of the Bessel operator by one. Thus, the Struve function can be considered as a special case of the Lommel function. If all terms in the expansions of H(z) are taken with the sign “plus”, then this series represents a function L,(z) which is studied (in case v = 0) by Nikolson, 4. As a definition of the Anguer function we can take the Bessel integral Vee) = (16.28) a= | costv9 — asin aya, (16.27) a If the value of v is an integer n, then the Anguer function coincides with the Bessel function Jn(2), and if v is non-integer, then J,(z) is different from the Bessel function of the index v. ‘The function Ez) =2 = f sin(v9 — zsin) ao (16.28) is called the Weber function. As a result of transformations of the integrals associated with the expansion of the integrand into a series in increasing powers of z we obtain the followingso 1. THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS equalities = con ton © (22/2) eee er Cee a (2yejaynt2 ena W480 mle Attra a Iv+i)P (m+ fv +l) (enn(esayinet aieabeprim pep | ON ‘The functions J,(z) and E,(z) satisfy the equations Velez) = feovieinwe (16.31) VE(2) = a Coa (16.32) Thus, it is evident that both these functions are connected with the Lommel functions because they represent the sum of solutions of the inhomogeneous equa tion for the casea when the exponent of the power function is equal to sero or one. This fact implies the formulae J(z) row() - 8-12), (16.33) v(1 — cos vm) — moor) fe) (16.34) 17. On intogration of the inhomogencous Bessel equation. Cauchy functions In solving inhomogeneous linear differential equations an important role is played along with the Green functions, by the fundamental Cauchy functions whose use facilitates the obtaining of the solutions. In Subsection 1 the Cauchy functions for the Bessel equations and the system of two Bessel equations are considered. In Subsection 2 we consider the solution of the Bessel equation with a apecial right-hand side which is based on using the fundamental Cauchy functions. 1. Consider the solution of the inhomogeneous equation Avut+u= Vin = f(2) a7.) the initial conditions du ua)=u, Hae. (17.2) ‘The particular solution which was sought in Section 16 (p. 46) has the form Jf st0astdle)Jf2) ~ Joule)ula)l de, u'(2)=17. INTEGRATION OF THE INHOMOGENEOUS BESSEL EQUATION En where v is non-integer. The general solution of equation (17.1) can be expressed in the form u(z) = AiJy(z) + AoJ-v(z) + u"(z), where A; and Ag should be determined from initial conditions (17.2). Let us introduce the Cauchy function Nilzz) 0, 2623 i ULe\dele) —Jelervla], 2>2 OT) 0, 2<3 Yale 114) ste) a ten oa) see Obviously, each of these functions satisfies the Bessel equation with index v; more- over, one can easily check that at z — 2 these functions and their derivatives have the following values: Vile, 2) [ Valz.2) y[_1 0 me 0 i Let us give other expressions for the Cauchy functions, which can also be used for integer values of the index: _)% z<8; Gea) = {Serauernca steve s2e (rs) 0, réei oe fe Unle)Yal2) ~ Yale)Jnl2)}, 222. its A solution of equation (17.1) which satisfies conditions (17.2) can now be writ- ten in the form u(z) aY (20) + vo¥9(2 20) + f Yale, 1/2) Consider a particular solution of the equation @u idu v tat ( a) H2), whose right-hand side is a step function; thus, f(z) = 0 forz <2, f(z) = 41 for <2 < 2a... f(z) = Dh 4 for ze <2 < 441, where q are constant. One can easily see that for z, <2 < 2441 the particular solution has the following form: we) = Daifsrel2) — sel) ile 2) tu [zy ts (zi) oa) Yale, 2) } In [20] a similar problem for the Bessel equation of zero index whose right-hand side is a step function is considered; in this case the particular solution obtained by integration of the Green function contain only the Bessel functions.2 1. THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS If the right-hand side of the equation considered represents a continuous fune- tion ‘ Fz) =O R(z- xn) for un <2< aes, rest then the particular solution can be written in the following form: k wz) = Safa ~ S2u()¥(2, 7) + ened + H)n4-8] cal} where fi are some constants. Let us pass to the equation of the fourth order which can be reduced to system of two Bessel equations. First, we consider the inhomogeneous equation Dw) -u= fl), az. where L is a linear differential operator of the second order. One can easily show that the general eolution of equation (17.7) hae the form usu tun, where the functions u; and up satisfy the equations Lua) £m = F5hl2), respectively. If pene le 8 babes aat ia ‘then we can obtain a particular solution to each of these two equations as has been shown on page 35. Suppose that the index of the operator is an integer, then, combining the particular colutione ui and uf, we obtain a formula of the form [rear dz, (17.8) where wz - Yale, 2) = TE { Jn(z)¥a(2) ~ Yale)da(2) + 2 Un(2) Ka (=) — Kn(2)In(2)]}; if the index is non-integer, then n should be replaced by v. ‘One can easily see that Ya(z,2) =0, ¥i(2,2) AnYa(#2)= 0, B4Ya(2,2) ‘These equalities immediately imply that we can take Y4 as one of the fundamental Cauchy functions and the derivative of this fanction with respect to 2, the result of applying the operator A,, and its derivative with respect to z can be taken as the remaining theee fundamental Cauchy fonctions. ‘Thus, we shall exptess the solution of the formulated problem as a sum of the particular solution u* and some linear combination of the fundamental Cauchy functions Yz.z), Yolz,z), Ya(z,z), Ya(z,2),17. INTEGRATION OF THE INHOMOGENEOUS BESSEL EQUATION 33 which vanish identically for z < 2 and satisfy the equation L*(u)—u =0 for z > 2; at z= 2 the following conditions hold: = m | ¥m | ¥ | Anon wooo It can easily be shown that in the case considered these functions have the form. forz > Viet) = F{ Ae al2) + ¥el@)Sal2) + 2 URCeKa(2) — Ka (2)al2)}}, Ya(=,2) = F {In(e)¥n(=) — Yn(2)Jn(2) 4 = Etre) als) Kol 2)/n]}, Yates) = FF { Ule)¥ol=) ~ Yel2)nl2) + 2m ented — Kile)tale}, Jn (2)¥a(2) + Ya (2)Jn(2) Un(2)Ken(2) — Kale )In()]}- Let us give the formulae for differentiation of the fundamental functions (zy m | Ym | ¥, | An¥on | AS Yon ee Nace gaa Ys 2) ¥2/¥s | Ys Yj; 3] Ya] vs} -% | -w 4) Ya) Ya} -¥o | -¥p ‘The solution of equation (17.7) with the initial conditions u(zo)= un, u'(z0) = 00, Antlo=C3, Aluo=Qn can be written in the form u(2) = woYi(2, 20) + vo¥a(z, 20) + O5Ya(2, 20) + QoYa(=, 20) + u"(2) In the monographs of the author {18}, [20], the fundamental Cauchy functions are used in problems on oscillations of rods with variable section, oscillations of aey 1. THE BESSEL EQUATION. PROPERTIES OF BESSEL PUNCTIONS circular plate, equilibrium of a conic shell, the theory of heat waves, as well as in Problema on the bend of a circular plate on an elastic foundation and others, ‘The Cauchy functions for a half-integer index of the Bessel equation can be expressed via elementary functions (see [24}). ‘When solving some physical problems, we have initial conditions defining the form of the Cauchy functions which contain linear differential operators of the second and third order different from the operators Ay and Af, considered before; however, this fact does not cause any essential difficulties (see, for example, [20]). At the beginning of the section we mentioned the Green functions. There exists a direct relation between the fundamental Cauchy functions and the Green functions, which follows immediately from the definition of the Green function of a differential operator. Consider the solution of the equation AoAou+u= (7.10) which has no singularities at z = 0 and is continuous with the first derivative and the operator A for z= z and such that Abticso — Abus and u —+ 0 as z + oo. This solution is called the Green function of equation (17.10). From formulae (11.9) we can immediately obtain (the notation uo, vo, fa, and go see on pages 13) Flute wots) ~ golz\vole)), => +5 K(z,2) (17.11) FE (wo(e)fo(2) ~ vol2)go(2)), = 2 2. Let us show that we can obtain the function Ya(z, 2) from K(z,z) in a simple way. We form the difference of the function K(z,2) and the expression which represents the upper row of (17.11) without the restriction z > z. Obviously, this difference is the function ¥4(z, 2) and it has the form Yale) 0, 2>z - { Fiuole) fale) — vo(2)g0(=) ~ fo(2)uo(2) + go(z)vo(z)], # > 2. ‘The formulae for the functions ¥\(z,), Y2(z,2), and Ys(z, 2), are given in {20}. Consider a particular solution of the inhomogeneous equation Alu-u= f(z). Assume that f(z) = 0 for z # ze, k = 1,2,3,... . Moreover, let ee ge eth ot atl, = Ou] + Gia ahs AL -0t Oe17. INTEGRATION OF THE INHOMOGENEOUS BESSEI. EQUATION ss ‘Then the particular solution for 24 < z < z441 has the following form ‘ DL (uo si (e,24) + v0 ¥a(z, xi) + 65,¥a(e, 21) + Qoa¥alz, 21)} 2. Consider the particular solutions of the inhomogeneous Bessel equation in the cases when the right-hand side represents the product of a cylindrical function and the squared argument. We begin with another simple problem. Let the right-hand side of the Bessel equation represent a cylindrical function whose index is different from the index of the operator: a 2 a= t igs + ( . 5) 72 (2), wey. A particular solution of this equation has the form (17.12) has the form ‘This result cannot be used when A= 1. The solution of equation (17.12) in this case will be called the resonance solution; we use the limit passage to find this solution. The solution of equation (17.12) with the initial conditions u(zo) = 0, u'(zo) = 0 can be written in the form 1 52 {- 2 (Azo) ¥i (2, 20) — AZ (Azo)¥alz, 20) + Z(Az)} where prime denotes differentiation with respect to the argument Az. For A= 1 the denominator and the expression in the braces are equal to zero; using the L’Hospital rule, we obtain a) =} sozy(e)¥ilese) + (« a 2) Zaza) ¥a( 20) + za] : Consider a simpler special case. The solution of the equation @u idu Fat rt w= Anz) with the initial conditions u(0) =0, u'(0) =0, for A # 1 has the form alvolAe) ~ Jo(2)) Applying the limit passage (A — 1), we obtain the resonance solution86 1. THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS 18. Products of Bessel functions In this section we consider some properties of products of the Bessel functions which we met before in Sections 11 and 17. We deduce the differential equations which are satisfied by these products here and give integral representations of the products. The problem concerning the expansion of the Bessel functions products into power series is considered in [7] At the beginning, following Watson [7], Orr and Appel, we consider two differ- ential equations = +Ri(z)v =0, e (184) Gat Rale=0 Denoting the product vw by y, we obtain of! = uw + 2u'w! + vw" = (Rs + Rady + 20'w!. Hence, + (Rit Ra)y = 2u'u! Differentiating this equation, we get xy" +2(Ri + Ra)y’ + (Ry + Ra)y = (Ri — Ra)(v'w — vw’). (18.2) ‘The special case Ry = Ro results in the differential equation of the third order Wi" 4 4Riy + 2Riu = 0 (18.3) HER, # Ra, then d fy" +2(K + 5 a +2(t we (+ ae =-(Ri- Ray. (18.4) Consider in detail the case R; = —Rz which results in the product containing two functions whose arguments differ by the factor equal to the imaginary unit. We have ae | £4) = —2Ry og — By 4 aRty (18.5) Suppose Ry = cz"? Then (18.5) takes the form w= 26g np actssle-y = 0 (18.6) For these representations of Ri and Rp, equations (18.1) can be reduced, as is shown in Section 13, to the Bessel equation. Hence, (see page 33), the solution of (18.6) is case et cia 9 = 2Fipag) ©) 44420) (®) Here Z* and Z can be two different cylindrical functions, for instance, Z* is the Beseel function and Z is the Hankel function. For g = 1 we have the trivial case36, PRODUCTS OF DESSEL FUNCTIONS sr vY +4cty = 0. If q = 3/2, then eae ‘ict ® v= ipa (“)2u0(™) 8.27) For z = i€ and q = 3/2 we have (18.8) Setting y = 2t(2), we obtain for q = 3/2 OY 4 3! J2— 3t"/2? 4 4c =0, t= y/2, (18.9) where y is represented by formula (18.7). If we assume Ry = Ro, then for the case Ri = e% Zav(e*), we have (see page 34) hence, y= Zs(e*)Zi,(e*) ‘The differential equation with this solution can be obtained immediately from (18.3), 40?) y 440 (£+1)9 0. (18.10) If the solution is represented by the product y= PZa(e!*)Z3,(e"), then the corresponding equation has the form Ol ae ae a By fe ee ( Bt (18.11) If Ry = kRy, k #1, then d (24420 + hay + (1+ eR a 3 = -0 om or 2 aes 4b 4 eR =o. 08.12) @ Um mj Consider the problem of construction of the fundamental Cauchy functions for ‘equation (18.5). ' alia As before, we have yovw, v4 Rw =0, w"- Rw (18.13)58 1. THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS ‘We denote the fundamental Cauchy functions® of equations (18.1) by v1, v2, wi, and wy. In this case the functions ¥, vz, w1, and wz vanish identically for = < aj for 2 =a we have v4(a) = 15 ‘ws(a) = 1. el) t; h(a) = 0; wa(a) 0; wila)=1; —wi(a)=0; wala) =0; First of all, we construct the function wemwm, 72¢ Obviously, wav + ow, Wf = vw + mul + 2Qvhuf = Qviwh, vf = 2Ry[-v.w) + wisi), BY = -4Rmw, + 2Ri[—niw, + wri). For z= a we have nl@=1, K@)=0, (a) =0, x"(@) Let us introduce the following functions ma + mu we pee, =} wo = yv2thn, =u = vi iba Ta) for 22a, Am ee for 2 =! can be obtained as a result of the representation of the Bessel functions present in the product in the form of the power series [7] For produets of Bessel functions with different arguments the following formula has been proven (7° Jylaz)Jy(bz) = fefreepy = a )(bz/2)* dee ag Asp t 1502/0). (8.17) 19. Integral representations of Bessel functions In Sections 11 and 14 above we considered two integral representations of the Bessel fonction: the Bessel integral and the Poisson integral. In this section we consider several contour integral which are generalizations of the Poisson integral, with the help of the theory of functions of a complex variable. Then, without deduction some results connected with similar generalizations of the Bessel integral will be formulated. “tee also A. Kratzer and V. Franz ‘Transcendental functions, TL, 1963.0 A. THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS Recall that the Poisson integral gives the following representation for the Bessel function: ea super reriarem [0-°) ae! We shall try to generalize this result, by considering contour integrals of the form u(z) = | T(the"* dt. (19.1) Jz) = Let us find out what restrictions we should impose on the contour C and the function T(t) in order that the function 2"v be a solution of the Bessel equation with index v. ‘The function v(2) satisfies the equation av" + Qu + Iv! +2v=0. (19.2) Alter differentiating (19.1), we substitute the expressions obtained into (19.2) and integrate the products zu” and zu by parts. As a result, we obtain [eT - TOE" |_-5 f [OF () + ATU) — (2v + 1)eT(t) - TW]e™ dt = 0. : (19.3) Here the symbol |, denotes the increase of the quantity before the vertical rule after going around the contour C in the positive direction. In order that equality (19.3) holds, it is, obviously, sufficient that each of the two terms in the left-hand side is equal to zero. By sctting the first term equal to zero, we obtain (ree — Me |, = 0, and taking into account the fact that the second term vanishes if the expression in the bracket is equal to zero, we have Aro — I) = Qv+ TE). (19.4) Integrating (19.4), we obtain that up to a constant factor TU) = (@ = 1yr¥? (19.5) We assume that the index v does not c accept the values v ~ n+ 1/2 (where n — (EAS TY 0;1,2,...), because in this case the integrand Vee ee jon at = se] and the in- Under the conditions formulated, (19.5) is a multifunction. If the contour is a closed curve inside which one of the points t = +1 lies, then after going along the contour in the positive direction, the argument increases by (v ~ 1/2). ‘Taking the contour C in the form presented in Fig. 2, we bypass both marked points in different directions; therefore, the function does not change its value after going around the whole contour. Hence, applying the integration along this closed Figure 219. INTEGRAL REPRESENTATIONS OF BESSEL FUNCTIONS a contour, we can obtain the representation of the integral of the Bessel equation in the form Y fe = 1PM eH a (19.6) & Giving one or another form to the contour, we can obtain dif- ferent integral representations of the Bessel functions. ‘The second group of the integral representations can be ob- tained, if the contour in a non-closed curve and at the infinite points the product (1? — 1)'"1/e## tends to zero. If the integration is performed along the contour presented in Fig. 3, then for the point 6 + iy of the contour, the exponential function present in the integrand has the form expliz(b+iy)], Rr >0. ‘The integrand tends to zero as y -+ co, and the integral (19.6) converges absolutely. ‘We turn our attention to the case, where the contour has the form shown in Fig. 2 ‘We agree to assign to the multifunction (1? — 1)”-¥/? for a contour of this form the sign plus for t > 1. Integral (19.6) converges for ®z > 0. Differentiation under the integral sign is legitimate, and the integral is an analytic function of v for all values of v. Let ns show that for an appropriate choice of the constant factor, the integral (19.6) represents the Bessel function J,(z). We shall prove below that the integral along the figure eight curve, which is Hankel’s generalization of the Poisson integral, can be reduced to the ordinary Poisson integral (considered in Section 14); hence, it represents the Bessel function J,(z). ‘We integrate along the symmetrical contour” shown in Fig. 4. It is sufficient to consider the integral along the half-contour which lies in the right half-plane and consists of two sloping segments OA and OE, two horizontal segments AB and ED and the arc BCD with the center at the point ¢ (1,0). Let p denote the radius of the arc, p < 1. Obviously, on the arc the following inequality holds: (eyes opin, te Figure 3 where C is a constant, This implies that the integral along the arc 1 on cannot exceed Cp!-¥/22np = 2nCp*¥/2, Thus, {Pry for v+1/2 > 0, the integral along the arc tends. “=| > to zero as p -+ 0. Obviously, as p tends to zero, the lengths of the segments OF and OA simul- taneously tend to zero, and on the segment AB the function T has the modulus (¢? — 1)’-1/? and the argument —ri(v ~ 1/2) ‘The modulus of the function T(t) on the segment ED remains equal to (¢? — 1)°-17, and ite argument on thie segment ie equal to -mi(v 1/2). Obviously, the modulus of the integrand on the sloping segments is less than one. Hence, the integral along the right half-loop is equal to the sum of the integrals along the Figure 4 Tee [28]6 1. THE BESSEL EQUATION. PROPERTIES OF BESSEL FUNCTIONS segments AB and CD, whose lengths tends to one as p -+ 0. Therefore, i [ee pyaar) faye | ; ° =a —yay freon a Comparing this result with the Poisson integral, we finally obtain u PH recs val + 1/2) a (:/2" 2 yyert/agise = ammo |" aaep aoe et aes (19.7) Je) = which is Hankel’s generalization of the Poissem integral ‘The restriction v + 1/2 > 0 is not necessary, because both parts of the equality are analytic functions of v and according to the principle of analytic continuation, the result 1s valid for any v. ‘We now consider a contour in the form of a loop (see Fig. 3) and we shall show that, by integrating along this loop, we obtain a representation of the Bessel function J_,(z). Assume that the circle |t| = 1 lies inside the contour, then (f? — 1)’-¥?? can be expanded in a series in decreasing powers of ¢ which uniformly converges on the contour: (e-yeeS Here ¢ is a complex variable with the argument y (—3n/2 < g < 1/2), and its modulus is greater than one. Then the termwise integration is admissible; therefore, (ae) 2 Py enn ’ | ee yas So ei [2-1-2 gp tis known [7], that [erro a = Cire nt TQ@m— 2 +1) where C; is the contour shown in Fig. 3; hence, (tay ees | L ee 2 an (e\ren aren Bei tnieVT(1/2) 5), (10.8) TG-¥) ‘This generalization of the Poisson integral is obtained under the following re- strictions: Rz > 0 and v + 1/2 is not a positive integer.19, INTEGRAL REPRESENTATIONS OF BESSEL FUNCTIONS 6 ay a = C= Figure 5 Figure 6 We can express the functions H!")(z) and H{2)(z) via the functions J,(z) and J-v(2). We write only the formulae without their deduction: (e/2° ee WP = ream eT) | fe-n etidt, (19.9) 10% = ecg [e- yea, (sa here v is non-integer, and the contours As and Aq are shown in Fig. 5 (the formulae are also valid for non-integer indices). From these formulae we can obtain the general- ined Meler-Sonin integrals. Changing the contour of | integration in integrals (19.9) and (19-10), we obtain the integrals along the contours shown in Fig. 6. Deforming these contours into the infinite half- segments (00, ~ 1] and [+1, 00), respectively, of the real axes, each of which is taken twice, we can obtain the formulae ga) = UB SAE _ ere fete -yrtea, (19.11) He y= en ~ atone frie aye (19.12) Generalizations of the Bessel integral are obtained in papers of Schlafli and Sonin. Without calculations, we write the formula af exp (: - a) dt, argz| Rv > 1 ‘The deduction of this formula can be performed by the same method that ‘was used in Section 14 when the Poisson integral was considered. Expanding the2, DEFINITE AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS 6 integrand (20.1) into a series in powers of z and using formula (8), we obtain 2 ff telecine) cin"? 0co™ +! 040 a e(eaymatim Be ME ae | repeat gett = x ns mil (a+ m+ Iya i Boost 0d0 1)" (2/2)aevtames mIP(u + v +m + 2) a? v) sly G _ BT +y) 2H Inavs (2). Obviously, if 4 = —1/2, then the first. Sonin integral becomes the Poisson integral After a calculation similar to the one given above, we use the expansion of the Lommel function s,,,(2) into a power series and obtain *p in!“ cos” 1 ff Yelesind) sin!" 900000 = err tntvwnenle)s (202) For y= —1/2 the equality (20.2) takes the form a .-1/2(2), (20.3) 6) sin! 0.0 = (x/(22))/7H, [cs Consider the first Sonin integral of the Struve functions i For this purpose, we expand the following integral, using (16.21), into a series in powers of 2: a Hi,(zsin6) sin'* 0 cos 0 8 se (eaym(e/ayutomen—P 2 (5 sintMt42042 9 oog20-H1 ero ter eT) { - hint S(yme/atmes T(E emg $4 PT 1) Pre ipe 7 Lr rs urms 3 (e+ y+ m+ §+$) then, setting k= 4, we obtain 2 _ r Lf Hesino) sit 0cos™*? 040 = ae (20.4) Lagu (2).20. DEFINITE INTEGRALS ie If we set k = —p, then after some simple transformations, denoting py = w+ v +1 and v1 =v — p+ 1, we have 2 f 1H, (zsind) sin!" @c06””+1 0.40 (aym(e/2yt2mar(y 41) 2, Pan = v4 2+ H/T a +n + 2m + I)/2) a T(v+1) * DP(n 4 1/2) + 3/2) Let us deduce the Sonin integral for the Lommel functions, by expanding the following integral into a series in powers of = (20.5) "2 u(zsin@) sin"*1+** Ocos+* 9 do =e JmQH“2 (2/2/2414 HP ((u ~ v + 19/2) (4 + v +1)/2) = T((a= v + 2m + 3)/2)0 (4 + v + Im + 3)/2) p xP snmenet cat oat -§ Caretta sr ecg et a RYE (A) Eimtut ds Bre $4 Timtatet 5+ 542)” — where, we obtain after some transformations Setting h "2 Su,v(z8in8) sin’** 8 cos™+* 9d0 = Mev + 1)/2)0 (r+ +1)/2) vp kpraayr BE (Gy + w+ k-+2)/2) % Sutugh/241/2.2-4h/241/2(2)- (20.6) ‘The integrals, in which the argument of the Bessel function is proportional to the squared sines, are of interest. One can easily show [73] that a2 i Jo(esin® 6) sin™*90 cos 040 = 2 Jv4a(2), (20.7) 3 a if Hi, (zsin?0) sin?”*9 6 cos 0 d0 = pH), (20.8) Rv > -1.68 2. DEFINITE AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS Now we give an important formula proven by Rutgers and afterwards by Wat- [stearate sia deat 0dn= Heya ane : Rv > -1/2. For references to publications, which contain generalizations of this formula, see ‘a Consider the second Sonin integral ? Gees) f Flo cone)lazsing) con pein pdg = £2 NAN, 3 (V2¥8) (20.10) Ru>-l, Rv>-1. In order to prove this formula, we expand the integrand into a series in powers of ‘ag, change the order of integration and summation and use the first Sonin integral. i ee es f Iulagcos g)J,(az sin g) cos" sin’! pdp vein am OP v ) artery Jaz sin g) sin’*? [Link]*#¥?"4 ip dip oo vt m —1" faq\tt™ ater Josnas (a+) ace Vary ayer - Let us briefly consider some trigonometric Captain integrals. ‘The solution of the differential equation fe su =ole) (20.11) with zero initial conditions has the form ue f eteysints—0 ae a ‘The computation of this integral in the case when the right-hand side of (20.11) ie expressed via Bessel functions, leads to the Captain integrals. Setting o(2) = Jo(2), one can easily show that J satysings - tat = 2h(2). (20.12) a21, IMPROPER INTEGRALS “ Setting y(z) = Jy(z), when computing u we have the integral f Hleisings— 1) a= sin ~ 2dn(2). (20.13) a ‘The validity of these equalities can easily be established by direct verification, after {introducing them into equation (20.11). Let us give one more Captain integral f Jolt) cos(s —t) dt = 2Jo(2) (20.14) Nee oe Sell ec oe Captain integrals containing the Struve functions J roto = t)dt = 28hy(2)- %, (20.15) uo sin(z ~t) dt = 4 sate. (20.16) ° A very important integral obtained by Bateman Hetees) — f 1 (2-a)Jpar(0) i 20.17 ptt ¢ ) is given in (7]. Using this integral, Rutgers has obtained a great number (more than four hundreds) of different definite integrals containing Bessel functions (81). 21. Improper integrals In this section we consider improper integrals, whose integrands contain the Bessel functions and contiguous functions. Many of these integrals can be encoun- tered, when obtaining the integral Bessel transforms [5, 6, 13). In Subsection 1 of this section we consider improper integrals of the Bessel functions; in Subsection 2 we calculate integrals of the Struve and Lommel func- tions. The question concerning the application of integral Bessel transforms to the calculation of some integrals is briefly discussed in Subsection 3. In Subsection 4 we consider the integrals with respect to the index. Subsection 5 is devoted to the problem of refinement of the convergence of some improper integrals. 1. Pirot, consider the Lipachits integral fees ae where a > 0 and both the numbers (a + it) are positive. ‘The value of the root is chosen such that |a + Va? + P| > [b]. In order to prove equality (21.1) we replace the Bessel function under the inte- gral sign by the Bessel integral, which in the case considered becomes the Parseval Cos: om)70 2, DEFINITE AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS integral, and change the order of integration. Thus, we obtain fe wainaen 2 few a f ent ae if pe a) an ibeosd ~ Jatae a Similarly, we can prove « more general formule e ~ 1 Vette en HJ, (bt) dt = (21.2) [071004- Fara a where n is an integer (cee [11]). In order to prove this equality, we use the Bessel integral in the form SIalz) cy if oF ¥ cosnpdy, a and then change the order of integration. Thus, we have Jn(bt) dt = of stat f eM cos np dy a E [creee [= cow) at = ci" f coanpdp =a J a eosp 3 are a (21.3) lS Lanes When deducing this formula, we asmme that nis a formula also holds for non-integer indices v, if Rv > —1. Consider a generalization of the Lipschitz integral due to Hankel. If we expand the Bessel function in a power series and change the order of integration and summation, then, as a result of term-by-term integration, we obtain > oe 1/2)" Pu tv + 2m) feger; however, this fe J (bt)te? dt J miPv+m+1) aHtetim (/2a))"T +r), (AY uty tl B = a (Be et tio) (21.4) Rv) >0. ‘The equality (21.4) holds under the condition that 181 < 1, because the term- lel by-term integration is only admissible in this case. Moreover, for convergence of‘21. IMPROPER INTEGRALS n the integral in (21.4) it is necessary that the inequality #(a) > 0 holds. With the help of analytical continuation one can prove that (21.4) holds for a larger domain of values of b, if the conditions (a + ib) hold simultaneously. Nielsen [76] has called this integral the integral representation of the hypergeometric function. Formula (21.4) can be rewritten in the form fertiecene ae (0/2)°T(u +») (4 l-n+v. B at) — So Fy vt, (21.5) Te + Dae +e or DT 2 aru) P18) ‘Similar results can also be obtained in the case when the cylindrical function under the sign of integral is the Neumann function; for this purpose one should use formula (2.1). ‘This general result contains many important special cases; some of them were considered above. Interesting results can be obtained for the cases, when the hypergeometric function can be reduced to elementary ones; in particular, for # equal to v +1 and v +2 we have the formulae which were obtained for the first time by Gegenbauer: by Tv 2) ya = TY a9) (ote at = ee. (21.7) Here Rv > —1/2 and Rv > —1, respectively. Let us write without deduction (see (7) the formula Jenana as ee ee | ei no a * [owns =r 19) [omnioa= =. (21.10) Let us go to consideration of the Weber integral Pasagerrsosta = me ay (ian 3 where ) > 0, 0-< Rp < R(v)-+3/2. Weber proved this result only for integer values of n; for arbitrary values of v this result was generalized by [Link]. Sonin. Formula (21.11) can be obtained from (21.5) as a result of the passage to the limit, setting a0.72 2. DEFINITE AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS Let us give the deduction of this formula, using the Poisson integral and change of the order of integration and assuming that 0.< Ry < Rv + 1/2. J saeco 3 a win) fo at Pont cos g)(sin y)” dp * a =i) [room de Je cotongy at 2 Toe) = area) nc Fang (cos) de Using formulae (7) and (8) from the Appendix, we obtain (21.11), Consider now the first exponential Weber integral if Jolat) exp(—p?#?)t dt spr mle" /(tr?)), (21.12) This integral converges only if | arg pl -1 ‘This can be proved just in the same manner as (21.12), namely: / . SUM 0/2)"47 Fee aogae [venencrer Hat Limosery | eo ptettkt a Sears =1)(a/24* Tv +k+1) EDU +R +1) | 2pFRHT (<1) (a/2p)* = gi te a] ore on (- we):21. IMPROPER INTEGRALS a A more general formula, which can be proved similarly, has the form i Jolat) exp(—p2e?)t"— dt 3 (21.14) = T(x + »)/2) sh (3 #) = papery! (e+ fuse LG Let us give a special case of formula (21.14) Fem (- &)» &). (21.15) Replacing » by —v and combining these two solutions, we obtai / Jau(at) exp(—p?t?) dt f Yov(at) exp(—p?t?) dt a Poxp -&) [o(G) eet in(S =) ) secon] (21.16) for [Rv] < 1/2. If we use the addition theorems (see Section 11), replacing a by Vat TP Dabcose in (21.14), and set p= 2, then we have, as it is shown in [7], that 2 a4) | (ab [oot pA) I (at) Jy (bet dt = mo 0(-S ot) a (#)- (1.17) ‘This forsmula holds for [arg] < n/4, Rv > -1. Consider now the Hankel integrals oy n Gate. (21.18) i L ‘The closed contour L is chosen so that is as shown, in Fig. 8. For z = z we have it PHO (az) = 2? Hl)(az). For z = ze", ie., on the left-hand horizontal semiaxis, we have Figure ® PHM az) = 2 ePt H\Y(aze") If R + 00, r; + 0, then Foe (a2 (aze™)\ds ele eg Naas re a) {eH (ar)}, (21.19) where the expression on the right-hand side is the residue at z = r multiplied by Qni.74 2. DEFINITE AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS HE)(2) = J,(2) + iY, (2) and setting r From (21.19), using the equali wwe obtain for Rk > 0 I- [ie - vs Je(az) + sin iF 0 — oe] lo) ea pt ene SS (Ga) (k-2K,(ak)]. (21.20) Meler has indicated a particular case of this formula ] zJo(az)de Sos = Kolak). (21.21) A great number of integrals, which are actually generalizations of (21.21), can be obtained if we assume that the numerator is a product of two Bessel functions and the denominator is an even polynomial of the fourth degree!. Here we can use the following approach: first, without changing the numerator of (21.21) we assume that the denominator contains an even polynomial of the fourth degree. Here we can use the integration along the same contour. Then with the help of the addition formulae we can obtain the solution for the case when the numerator of the integrand is a product of two Bessel functions. Furthermore, we can introduce ower functions under the integral sign ot replace one of the Bessel functions by the Lommel, Struve or Anguer function. Notice that the author has actually calculated some of these integrals by another method in [20]: all these integrals are solutions of the problem of the integration of the equation (A.A, +bA, + w= al€), (21.22) where g(¢) is either a delta-function or a finite function, for instance, a power function on a segment and identically zero everywhere outside this segment. Let us give only one of the integrals of this type, when the argument 2 and the numbers a and b are real (see 156, 214): 7 eh(az)de on : { itt: Fat Tamils (21.28) several integrals of this type are given in [20], as well as in the second part of this book. Now let us give the discontinuous integrals? of Weber-Sonin-Schafheitlin which are important in applications F Inbat) (bt) at a For some special cases these integrals were calculated by Weber; for all values of the parameters A, 4, v such that this integral converges it has been calculated by Sonin; Schafheitlin in his investigation has noted that at a = 6 this integral has (21.24) "The question concerning the integrals (which are another generalization of (21.21)), when, the condition of the evenness of the polynomial which is the denominator of the integrand, does ‘not hold, is considered in Subsection § of this section. Watson calls them the integrals of Weber and Schafheitlin.21 IMPROPER INTEGRALS: % a discontinuity. In the general case, the Weber-Sonin-Schafheitlin integral can be ‘expressed via the hypergeometric functions zene erat dv- pat {oe * BoP +10 Gat da - yt 8) +u-A+1 xan (SE, + 45) (21.25) for 0—1, A= y/2cosa(arei# + be-H), If we set z = ¢, put (21.27) into the integral (21.24), change the order of integration and afterwards use the Weber integral (21.11), then, as a result, we obtain a definite integral which can be expressed via the hypergeometric function? (ecc (4}). Consider special cases of formulae (21.25) and (21.26), setting v = 1/2 and (4) (c08 8)°*#(A2)"#" Tuae (Az) d0, (21.27) a 1/2 in them: sin{y aresin(8/a)} oa 7 me , [elon sindeat = oo or 0, baa, 3 a” coa(jer/2) i: VFaab+ vera 5% Ru) > -2; (21.28) f suo conte at = (21.29) Compare with page 70.76 2. DEFINITE AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS Setting y= 0, we obtain [sets 7 (21.30) Salt) costeat = (21:31) ‘The last two integrals are called the discontinuous factors of Weber. At the end of this section we will give a simple deduction for another special case of the Weber-Sonin-Schafheitlin integral (gee formula (21.43)). Let us proceed to consider the improper Sonin integrals. First, we consider the Sonin integrals which contain one Bessel function under the integral sign, then we consider the discontinuous Sonin integral whose integrand contains a product of two Bessel functions of different arguments. Fins consider the formula due to Sonin ff OVE TP) oon gy PED (4 ay? antl w= et (21.32) a>0, R(5-7)>Ry>-1 20, 2(f-3) m0 ‘This formula can be proven, if we use the Lommel expansion (Section 12), with the help of which the ratio under the integral sign PACE) (@+ayr can be represented in the form of a series, and then we use the Weber integral (21.11) J this ate we obtain F dav aya gp “eraph =E5 Peay ptm) 2T(u +1) bh (-1)"(a2/2)/- 8-1" orig mIf(v— n+ m—141) YTwt)) = re a-1(a2). Consider another integral which is can be obtained in a similar manner: F Julai) (4 ay generalization of the preceding one and £204 gp = gael 2)/2 uve (he )/2q =m (h41)/2 x Dut (k+1)/2) Ip (essyia(az), where p> —(k-+1)/2, v > 2p-+k— 1/2.21, IMPROPER INTEGRALS 7 Remembering that the Lommel expansion can easily be generalized to arbitrary cylindrical function, including, obviously, the Tankel function, passing from a real argument to imaginary one and repeating the calculations given in page 76, we obtain the important formula j Sie 42u41 at Zr, -1(02), (21.33) where @ > 0, Ru > =1 Consider the discontinuous Sonin integral fone Zo [VFA] uss gy : (+n (21.34) for Ry > Ru > -1,a>0,b>0. In order to prove this equality, we use the Lommel expansion (page 31), change the order of integration and summation and use the discontinuous Weber-Sonin- ‘Schafheitlin integral (21.24) in the special case* A = v-+m—p—1. As a result, we obtain pms (VFA) Ay Crapper ‘Fm f Fle) Ivamlatere at (caymamztmge-¥— m+ pg~¥— (g? — payee cd mi"T(y — pt my aah ach, ae ach. ae atever=PH 2. Consider improper integrals of the Struve functions. In order to obtain these integrals, we can use (sce [7]) the same methods which were applied before in this section, namely, the expansion of the integrand into a power series, ot using an integral representation of the Poisson type (see Section 16) which is followed by the change of the order of integration. Here the result will be obtained with the help of the representation of the Struve function given by formula (20.3), which is followed by the change of the order of integration. First, we will, obviously, use the integrals of the Bessel functions obtained before in this section, and then use the definite integrals considered in Section 20. “See formula (21.43).78 2, DEFINITE. AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS Let us set f ‘Iu(az)®(bz)z" dz = f(a,b,n,0), where the expressions ® and f should satisfy the conditions which guarantee the convergence of the integrals and legitimacy of the change of the order of integration, which follows the application of formula (20.3). Then, obviously, [ Petone (eye de a = he 1 aapalae in8)0(b2) 2 (6in9)'-P-222 dzd0 a0 12 ifs ([Link] Loe 3) anit-rg ‘As the first example, we cousider the Weber integral of the Struve Junction [sse- vif / f Soaxpa(e sin 8) sin/2-¥ 9 2-Y 40-1? ds do T(u+1)/2) Teri ay (21.35) xf "odo T'(u/2) tan( (2) 3 = piUeaeel, epee (ve) ansg In these calculations we used formulae (20.3), (21.11) and relation (6), Consider now the integral oat wn [om apn(bt)t-¥? dt = {] (*®) #-¥/9J,_(bt sind) sin?” 0 d0 dt - fi, Rv >0, blac. (21.37) In the calculations we used the formula J tot 208 dt=b a", if Ru >0, blac, ° which is a special case of (21.25). This formula will be deduced below (page 80).21, IMPROPER INTEGRALS 79 Consider the Sonin integral for the Struve function ef fee (osin0VEEH) rae gpg Lary ~-TH (az). (21.38) Similarly, we can calculate improper integrals whose integrands contain the Lommel function. So, for instance, using formula (20.2), one can obtain after simple transformations 7, ceyntas)stv-tgy = BAAR Warsi / tes ensalae)Dalartttde = ETE, ([Link]) Rv > -1, Ru > Rk > 0. In a special case * Ten? if Supe, v-wst(02)Jpv—1(a2)edz = a (21.40) By > -1, Ru—v—1)>0. This approach can also be applied in order to obtain the integral of the product of two Bessel functions of the first kind, which have the same arguments but different, indices. Por this purpose we use the Neumann formula (18.16). Using different generalizations of this formula, we can obtain integrals of the product of two different cylindrical functions of the same argument. If the ar- guments of the cylindrical functions are different, then we can use the Neumann formula (18.17). “The improper integral of the Lommel function of two variables (see page 116) will be computed below in a similar manner, in Section 27. 3. Consider some examples of applications of the integral Bessel transforms to problems of the computation of improper integrals. [fin an integral transform rey= f1oKeya i the kernel K(f) is a cylindrical function, then the transform is called the Bessel transform. The most important among the Bessel transforms is the Hankel trans- form fu) = fr ap(ayat 0 -1/2, (21.41) 380 2, DEFINITE AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS which has the following inversion formula: f= fof eau) d 0 -1 We denote pry — v1 = €, pn +41 = p and introduce the function Joy (220088), 0 < [| < x/2, sescoey= {i 1/2< |0| 0, then, setting v = iz and using the same inversion formula, we obtain f Kicks) con at a cont, O0. Using the change of the variable v = ix and the inversion formula, we obtain I [K.(2)]? sinh(v7) sin(2ut) dt = Flee sinht), (21.50) i O), Go(@) = 940 [e(1/2-4 1v5/2)] = 9H Ee), po=n/3. From the computational viewpoint the solution obtained is more convenient than the initial (21.54), Actually, (21.01) consists ofthe sum of two terms. One of them represents a linear combination of tabulated Bessel functions. ‘The second one jn-an integral which converges much faster than the initial one. Fr the convenience of the computations, the integral present in (21.61) can be represented as a sum of two integrals with the limits from 0 to a and from a to oo, where a is small compared to one. Then the fst ofthese integrals can be expressed via the tabulated function. If, for instance, we take a ~ 0.2, then *Pateyat ae) a ow f ett at 2 In this case the maximal error in the value of the integrand is equal to 1/15625; introducing the new variable z = té, we reduce the integral to the form J Kolté) at = i T atees. ‘The integral obtained is tabulated in (7, Vol. 2]. Thus, it remains only to calcu- late the integral with the limita a, 00, using the formulae of mechanical quadratures. ‘The upper limit of this integral can be taken approximately equal to 3-4 because of the very fast growth of the denominator and the even faster decrease of the Macdonald function in the numerator.86 2, DEFINITE AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS Let us note that, when differentiating (21.61) (with the purpose of obtaining the slopes and internal forcee in the plate), the integral in the right-hand oide remaine sufficiently well convergent, ‘The method of the transformation of the integral which was described can easily be generalized to the case when the numerator of the integrand contains a more complicated function; the calculations in the case when the denominator of the integrand is a polynomial of a more complicated form are more tedious. ‘Asan example, we consider the problem of the calculation of the integral present in the calculation of an unbounded plate which lies on elastic half-space and suppose that the plate in the middle plane is stretched by a uniform effort po. In this case, the deflection and the reactive pressure have the form PIF Jo(€) dd u6)= 50 J Tea +3" a P_ Ff AJo(A€) dd (21.62) = 998 | T+ ad+ 3 = po/tt. ‘The determination of the firet integral (21.62) is reduced to the calculation of ‘two integrals along the contours shown in Fig. 9. ‘We write only the final result F Jol) ad _ 2 F Koltg)at [ote = 2/3 ss a ro +G(6), (21.63) where Gg) = an 8), FolE) + iol) = HG [Can + 481)é] = HS? (pele tango =Aifar, a =0/2, f= VifP-P/A, 7=8/2-2/b, 6 = 3H, where 6 = —zo, 29 is the real root of the equation z* + az +1 = 0. Let us compute the integrals in the case when the loading is distributed on a circular domain and can be represented in the form &/b, (6,0) = > Fa(€) cos nd. 7 For the sake of brevity, we consider only integrals of the type (21.56). Let the loading qq cos n8 be distributed along the circumference of the radius R. Using the superposition principle, we find the deflection at the point with the coordinates & = R/I, g. The distance between the points in the accepted dimen sionless coordinate system (£,) and £1, 0) will be denoted by Ie? + & — 246s co8(0 ~ 9).22, DUAL INTEGRAL EQUATIONS s7 Using the addition formulae for the Bessel functions, we can, integrating with re- spect to 0, immediately obtain for & < & IPs cosmp (2 Knl(Et)Ia(6st) mfg, p) = SPSrcosne (2 f KalEO)n(E0) ay { D ® T+e + FAV 3 Fa (€) tn) ~ Bn (6)8n (61) + In) On Ea) + Inteenteath, f>&, (21.64) where the notation tin (€), an (€), Ja(€), and ga (€), is given on page 14; for € < the expression for w is obtained from (21.64) by interchanging £ and in the brackets. It should be'noted that although the function In(z) as x > oo increases as e//, the asymptotic representation of the product J,,(t€)Kn(té1) contains the exponential factor e-(&~®), which provides the required decreasing vf the nuiera- tor for £, > &. The same is valid for € > &;. It is clear that at € = & the advantages caused by the presence of the exponential factor in the numerator, disappear. For the loading q(€,) cos n@ distributed on a circular ring, it is necessary to integrate with respect to the argument . If (€) = Ag? (n = 1,2,3, the Feault will be expressed via the cylindrical functions. If 4(&1) = Aéf, then the result contains the Lommel functions of the complex argument. 22, Dual integral equations It is known that many problems of mathematical physics can be reduced to solving dual integral equations. Different contact problems of the theory of elas- ticity, problems on strained states near cracks, some problems of electrostatics and other topics are among them. The theory of dual equations recently became very developed and is the subject of numerous investigations. It in interesting to note that today, when solving, dual integral equations with Bessel kernels, different Sonin integrals and a series of other results of the Bessel function theory are widely used, First, consider the dual equations [racscen a= f(p), O -1, €>-1. Substituting * = xt, p= zein0 in this integral, we obtain COU seanlet) = Sern] Jelot\e (22 — pPEdp, (22.2) ey +0 ‘After multiplying both sides of the integral equation (22.18) by ge vg? — gaye ee integrating with respect to p from 0 to z, changing the order of integration in the left-hand side and using the transformed Sonin formula (22.2), we obtain A(t) Jusets (2t) dt = S(o)p** (2? — p)Fdp, (22.3) are i) O-1, €>-1 Now we formulate another Sonin’s result related with improper integrals: alzt) [eter ery? tery batt ds, 2 (n +1) ° where v/2~ 1/4 >> —1. Setting s? +2? = p? here, we obtain ef toe rtipt— 2" dp, (22.4) 2s Brm+h Jung (zt) = ‘After snultiplying the left and right-haud sides of equation (22.1b) by werd 2T(n+ 1) integrating with respect to p from z to oo, changing the order of integration and using the improper Sonin integral (22.4), we obtain fortawa an-a(zt) dt where d <2 < 00, v/2-1/4 >> -1. We shall try to make the index of the Bessel function and the exponent of f in the left-hand sides of the equalities (22.3) and (22.5) equal. Then, combining these two equalities, we can immediately obtain the Hankel transform of the desired function. Thus, we obtain the following equations: v4etl=v—9-1, a-€-1=-9-1, -a/2. peg? — 23)", ek Fray | owes z7)"do, (225) hence, = —1+a/2,22, DUAL INTEGRAL EQUATIONS ss Since £ + = —2, it is impossible that € and 7 are simultaneously greater than =1. We shall go to further calculations, considering two special casce separately. a) 0 1/2—a. We set € = -1+ 0/2 and n = ~a/2. Substituting the values of € and 7 into the formulae obtained above, we find [ #77 A(t)Jy4o/2(zt) at = a z j 44 (9 — pPy- thal? Piel / Foe (a8 = PP dp, (228) i O<2d, ‘Multiplying the left- and right-hand sides of equation (22.7) by z-”~¢/?+1 and differentiating with respect to 2, we obtain [PAO dosorlet at 2 [aon eo *!? dp, (22.8) rod, Using the fact that the left-hand sides of equations (22.6) and (22.8) are iden- tical, we can rewrite these equatious in the form of one equation ii 1/2 AO)vgaprlet) t= Ve), 0S 2 < 00, (22.9) where cman +1, -14+0/2, some [ 10 (2- A) Map, OS 2 d. One can easily see that, up to a factor V/Z, the function y(z) represents the Hankel transform of the unknown function A(t). Using the inversion theorem, we immediately obtain the answer for the case v > 1/2—a, @ > 0, 4 (nt? T(a/ (ery J geii-er F + Ta =a) a At) = 2h P7415, 472 zt) F(z) de 2t0l2 J, 4 /2(2t)Gi(z) de, (22.10)90 2, DEFINITE AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS where Fie)= | soe" AP p, ae (22.11) Gute) = ff store 6? ~ 2-27 ap Formula (22.10) gives the result by Titchmarsh [51], who considered the case O -3/2—a. Replacing € and 7 in (22.3) and (22.5) by their valuce, we obtain actraltt Bary apy | Mee =~ A? ao, (22.12) [OMA derapreslet) at = j [ePaeirarlen a sheer" (p? = 2-1-*0 dp, (22.13) z>d, We transform the first of these equations. Multiplying both its sides by 24*2/241 and differentiating with respect to z we obtain as a result eal g PT (1+ a/2) dz ii #77 A() Jy 4o/2(2t) dt = f Soyo’ (22 — pP)*0? dp, . : (22.14) Oa 22.7) where H(t) is a function decreasing at infinity. Tn contrast to equations (22.1a) and (22.18), we cannot obtain the solution here in quadratures. The final aim is the reduction of the problem to a solution of an integral Fredholm equation of the second kind. If 0 < @ < 2, then equations (22.16) and (22.17) can be transformed in the same way, which allowed us to obtain equations (22.6) and (22.8) from (22.1a) and (22.10): ? ah ferns HW) Sesepnl3t) = Sarat Mla) (22.18) a ose4, (22.19) where F(z) is defined by formula (22.11). Let us denote a? Frenpeye’ OS2 f dalotysueyetoe a oe) (22.22) 8 6, act, as a result, we obtain N= cRrarae | 77 0Ce? ~ 97° ae (2228) ’ In order to obtain an equation with respect to O(£), we substitute (22.20) into (22.18) O(z) + srereerr(o) [ 19/7 H(t) B(t)Jp4a/a(zt) dt = 2'/?-¥-9/2F, (2), : (22.24) O 0, where A> =(2v-+a+2) and H(t) + 0 as t + 00. If H(t) tends to zero like 1”, <0, as t+ 00, then in order for the kernel to have no singularity at = = , it i that 1 < -1; if 0 > 1% —1, then the kernel has a singularity at z= If -2 —1, then the function J,(2) has only real roots. & )” coe, repr 4mm +1)" assuming that z = ia, we obtain a series which contains only positive terms. Hence, the function J,(z) cannot have pure imaginary roots Let us prove that J,(z) has no complex roots. Suppose that a is a complex root of the function considered. Since J,(2) ean be expanded into a power series with real coefficients, 4 should also be its root. Now we shall use the integral 4,(82) eee) Ha (089) Since Ju(z) fr e z ad, H(i) 1 (6), (Bt) at = a. [rs (0 at = a> ay [Jol 80) EE whose proof is given in Section 15 {see (15.21)]. . Setting x = 1 in formula (23.2) and taking into account that # and B are roots of the function J,(2), we obtain 1 f 1J,(8t)J,(Bt) dt = a ~ 1B However, this is impossible, because J,(At) = J,(Bf); hence, [rtonaciin a= f ya (onfae> 0. a 3 ‘This implies that J,(2) has no complex roots. Hence, for v > —1 the function J,(2) has only real roots. ‘THEOREM 2. AU roots of the function Z,(z), except z= 0, are simple. Consider the cylindrical function Z,(2). Suppose that: zo is a root of order n > 1. Then Z,(zo) and Z}(20) are equal to zero. Using the differential Bessel equation, we immediately obtain that Z!/(zo) is equal to zero. Now differentiating the Bessel ‘equation and substituting the values Z"(z0) and Z/(z0), we obtain Z!!"(20) = 0. In the same way we can show that all derivatives of the function Z,(2) vanish at z= 29. This implies Z,(z) = 0, that is false. Hence, the function Z,(2) can have only simple roots. ‘THEOREM 3. The roots of the functions Z,(z) and Zp41(2) alternate each other. Let us write the formula for the differentiation a vy, E20) Zvsa(2). Consider two consecutive roots of the function Z,(z), which are, as has been shown, real. Then, according to the Rolle theorem, at least one root of the derivative should be located between these values; hence, a root of the function Z,41(z) is also located between these values.23, ROOTS OF THE BESSEL FUNCTIONS 98 ‘TuEoREM 4. The roots of two cylindrical functions, which have one and the same indes, alternate each other. Consider two different cylindrical functions Z$(2) and Z,(z). Since they are linearly independent, their Wronski determinant is non-zero, ie, 2,25' - 12, where c# 0. Let 21 and z2 be two consecutive roots of the function Z,(z). As has been proved, 2; and 22 are real and Zl) #0, Z)(22) #0. Assume that the function Z,(2) is positive in the interval 21 < x < zz. Obviously, we have Zi(zr) > 0, Bilan) <0. Using the Wronski determinant and consecutively setting z= x, and ‘we obtain 22, “ni Zie1)" i2s(gi) ammonia eee Pe ee eee, that the same conclusion will be obtained, if Z,(z) < 0 for 21 << 22. Hence, ee Since at least one root of the function Z,(z) lies between two roots of the function Z¢(z), this implies that Z3(z) has only one root between , and 22. Let us give some properties of roots of the cylindrical functions without proofs. ‘More detailed information is given in [7] and [4] (see also the footnote on page 59). Zila) = 1. The functions J,(z) and J,4m(z), m = 1,2,3,..., have no common roots ‘except = = 0 (the Bourger conjecture). In this connection we note one result due to Siegel, who has proved that if v is a rational number and ¢ in a non-sero algebraic number, then J,(=) is not an algebraic number. 2. For the smallest roots 7», 14, and 7! of the functions J,(z), Js(z), and Ju (a), we have VUUFD << VAFDEFS), v>0, Ve Fd) <9, < Vv), v>0, Veal) << ve? = 1, v>0. 3. The function IL{az)¥.(bz) ~J,(b2)¥,(az), a>0, b>0, has only real simple roots. The set of these roots is infinite.96 2. DEFINITE AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS 4. For v > 0 the function K,(z) has no roots such that | arg z| ‘I. Any positive root of the function Z,(2) can be considered as a continuous increasing function of the real argument v. 8. Any positive root of the transcendental equation Jolke)¥o(2) = Ile ¥o lke) = is a continuous increasing function of the real argument v under the condition that v is positive. 24. Series of FourierBessel and Dini Jn this section we consider the Fourier-Bessel series which plays an important role in applications as well as some other similar series which are their natural generalization If the functions u(z) and o(2) satisfy the equations then Je-oywi= ( (24.1) Setting P = a? — (4v? — 1)/2?, Q = 6? — (4v? — 1)/2, we have usz¥?Z,(a2), v= 2/2,(62). ‘Theee relations can easily be checked by differentiating, From (24.1) we obtain i: (ap) f 22,4a2)2,(92)d2 = —|a2Z}(az)Z,(82) — 822; (6z)Z (a2)\. (24.2) Let us prove that the Bessel functions satisfy the orthogonality property which can be written in the following form: J #4elimt)dotin2) de = 0, kém,v>-1, (24.3) a where z is the real variable; Jm and j, are real roots of the Bessel function.24, SERIES OF FOURIER-BESSEL AND DINI Fa Setting Z, (az) = J,(az), Z,(B2) = J.(82), @ = 0, b= 1 in (24.2), we see that the right-hand side of (24.2) vanishes at a= 0 for v > —1. ‘This implies [see also (15.21)] that J #tla2\Jo(6e) dz= eee Bpetentet6) (24.4) ‘ ‘The right-hand side of (24.4) is ambiguous at # = a. Expanding it by the L’Hospital rule and using (6.3), we obtain : f ei2le2)de= J243(0)/2 (24.5) i Ifa and # are distinct roots of the Bessel function of index v, i.e., a= jm, B= in, mek, then the right-hand side of (244) vanishes, Hence, the Rome Functions are orthog: onal with the weight z. The functions v2 Tulimz) 141 (3m, are normed on the interval (0, 1) as one can see from (24.5). ‘The functions Jm(Ym2), m = 1,2,3,..., form a complete system. A proof of this fact can be found in different courses, in particular, see [30], [7] and 80 on. ‘The properties of the Bessel functions formulated above allow one to obtain the following expansion of the function DY am Jeima) (24.6) where jm, m = 1,2,3,..., are as before the roots of the Bessel function. If, assuming that an expansion of this kind existe and admite a term-by-term integration, we multiply the right- and left-hand sides of (24.6) by zJ,(jxz) and integrate, then we obtain the following expression for the coefficient ay: A a= Tomy [ 2f(2)Ju(inz) dz. (24.7) ‘The series (24.6), whose coefficients are defined by the formula (24.7), is called the Fourier-Bessel series. Let us formulate the following theorem without proof. ‘Turon, Let f(t) be an arbitrary function given on the interval (0,1) such that the integral {/2(t) dt exists and converges absolutely. Suppose that ay are a given by the formula (24.7), where v > —1/2. Let = be an internal point of the interval (a,6), 0 0. (24.12) In order to expand the power function 2# into the Fourier-Bessel and Dini series, we should to caleulate the integrals 1 1 fe Joliee)de and [ese a a 2 which can be expressed via the Lommel functions by the formulae given in See. tion 16° for » # v; in some cases, when the right-hand side contains exponential and trigonometric functions, the coefficients of these series can be expressed via the Lommel functions of two variables and partial cylindrical functions (see Sec- tions 27, 28). The series considered are a special case of series in eigenfunctions of the Sturm-Livuville problem for the Bessel equation. ‘Consider the expansions into the series in cylindrical functions which are suit- able for positive finite integrals. If the function f(z) is defined in the interval a < z 0, then the expansion obtained by Titchmarsh is HE) = YD ambSo(4m2)¥e (amb) — Yo (mz) Jol Ymb (24.18) mat here Ym is the mth positive root of the equation [Link])¥. (bz) ~ Yi(az)J,(bz) = 0, mi, om = {( ema]? = (ermB)? 2 Hi votre? f Ustam8)¥ Comb) —Yu(rmt) Ju (md) (#) dt The boundary conditions of this problem are analogons to the boundary condi- tions of the problem which generates the Fourier~Bessel series; from the physical viewpoint they correspond, for instance, to the problem of free oscillations of a membrane fixed along the contour, which occupies a domain in the form of a cir- cular rectangle. In the membrane under consideration the curvilinear part of the contour is represented by segments of arcs whose radii are equal to a and b, respectively. The In some cases (see page 42) they can be expreseed via Bestel functions.100 2. DEFINITE AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS segments of the radii which connects the ends of the arcs form an angle which is equal to vx. Only such forms of oscillations are considered the nodal curves of which are arcs of circumferences which are concentric to the circumferences containing the arcs of the contour. The forms of these oscillations coincide up to the factor C'sin v8, where C = const, with the expressions in the brackets in (24.13). Let us formally obtain the Dini series for a segment. For this purpose we first write the function which we have considered in Section 17: ¥i(aa, az) = Telegu. (az) — J',(aa)J,(az)), ¥o(aa, a2) = ~5*"_(J,(aa)J-y(az) ~ J-v(aa)J,(a2)}; Tein vm fore ma Nal, Y=0, Y=0, Y=1. Assume that the function w(z) which is a solution of the Bessel equation with the index v and parameter a, satisfies the following condition at z =a ote (24.14) and at z w+ int =0 (24.15) Obviously, the first boundary condition is satisfied by the function &,(au, az) = aff Yi (au, az) — ¥2(aa, az). (24.16) From the second boundary condition we obtain the following transcendental equa- tion with respect to @H2{aHy¥; (aa, ab) — ¥;(aa, ab) + aH¥;(aa, ab) — Yo(aa,ab)=0, (24.17) whose roots are denoted by m. Similarly to the reasoning above, in order to prove the orthogonality of the functions $,(im2, ima) and ®,(i4e2, p20), k # m, we ete ea : =) [ £9. me Bo (b4 2) de = [me (td, pm) (40, Hu) rs — pas, (tna, Het) Bo (Has Hme)| = Hb B, (dm 4, md) ®. (Hea, pb) = Heb! (Hea, 4b) ®, (stm, mb). (24.18) ‘The condition (24.15) implies that the right-hand side of (24.18) is equal to In order to norm the functions obtained, we consider the integral ; . . ‘ [ s2utozytae= = {(24(azy} + (1-245) 22(00)} i (24.19)24, SERIES OF FOURIER-BESSEL AND DINT sor Here Z,(a2) is an arbitrary function which satisfies the Bessel equation with the index v and parameter a. ‘The expression (24.19) can be obtained either by integration by parts or as a result of the application of the L’Hospital rule (7. For Z,(az = %/(aa, az) we obtain . if 2{@,(aa, a2)? dz = § {e(o0,on" + ( on) @2(0a, wy} _< [(: = oa) Hla? + | = (24.20) ‘Taking into account (24.15), we transform (24.20) to the following form: J #1800, 02))*de 1) fhaJi(aa) + J(aa)}? {( + get a) Hgalilad) +100) fake (0 - a) Hi .} = Ba). (24.21) Denoting the right-hand side of (24.21) by B(a), we rewrite the expansion into the Dini series in the following form: Ste) = Fey jolie, (24.22) where Cm [10% nts a) at Consider the expansion into the series of eigenfunctions of the problem de- scribed by the differential equation AvAyw, ~ atu, =0 (24.23) and the boundary conditions wy(aa) = 0, where102 2, DEFINITE AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS ‘The eigenfunctions of this problem are, in many senses, analogues of the beam functions which are considered in the theory of oscillations of elastic systems in connection with the investigation of free oscillations of a rod of constant cross- section and are widely used in different applications. ‘We need to find the eigenvalues ax and the eigenfunctions, which will be de- noted by ®,(ax6, axz). In order to obtain the eigenfunctions, we will use the Cauchy functions con- sidered above in Section 17, which satisfy the property of the identity matrix. In order to satisfy the boundary conditions at r = a, we assume (axa, a4) Yo(axa, xz) + AYa(ax0, a42), where A is the value which should be determined. ‘From the boundary conditions at 2 = b we obtain a transcendental equation, whose roots are the eigenvalues a4. Furthermore, from these conditions we find the coefficient of the second term; thus, we obtain Yo(axa, ond), Ya(ana, ab) We turn our attention to the proof of the orthogonality of the functions ob- tained. For this purpose we multiply the equations which are satisfied by the func- tions ©, (ama, Amz) and ©,(a_,4,aq2), by z@/(ana,anz) and 20,(ama, amz), respectively, integrate with respect to z from a to b and subtract one equation from the other one: (aya, an2) = Ya(ana, onz) — Ya (ana, 42), (24.24) (ot, - 08) [ 28,(0ma, m2) ®(0n0, ane) de J 29, (md, Omz)O/4,0(Ona, aq2) d2 7 if 20, (an0, an2)AyA,®(ima, Om2) dz. Integrating by parts, one can easily obtain J £8,(0md, dm2)A,A/®(dn0, 0n2) dz ‘ = 28,(on2,0n2)£A,8(008, 002) eh o(emajame)S/ (ane, ane) ce [ le + if 2B,9, (m4, Am2)A,®(an0, 042) dz. For the given boundary conditions, the terms outside the integral vanish.24. SERIES OF FOURIER-BESSEL AND DINE 103 ‘Taking into account the relation which can be obtained from the previous one by interchanging @m and a, we obtain ¢ (eh, —af) f 20.(0ma.dm2}(0n0,dn2)de=0, men. When calculating the coefficients ay of the expansion $(z) = and, (ana, 0x2), (24.25) where fef(2)®.(oxa,ay2) dz eee the following formula can be eel : . J seodony? te = Ef w.(aey? + [av.(oe)? = dus (a2)I'w, (az) + 2%5us(az)Aoue(a2) . + 2 faiws(az)- a.m(as)w(as))} > (24.26) where the differentiation is fulfilled with respect to the argument az and, further- more, we set in (24.26) ae ee Oe ¥ Baz)?“ az Gaz) (az)? ‘This equality is analogous to (24.1) and can be obtained in the same way. Under the given boundary conditions we have * 2 * / 218, (a10,0,2)]" de = = (A,9,(axa,a42)) de (wazr) 4 ‘Taking into account the relation Yalan, 049) y, (a4, an2), ALS (axa, az) = —Yi (ana, ane) + Yaexacend) we obtain after elementary simplifications ® Jf =1foxa,ox2)7 de Sf weak iy (aybhd-vfone)~ Jolene) (oud))? 4 laste 2Ayysinue)? xo Foose, :104 2, DEFINITE AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS where T; (ax0)K. (axa) ~ Ky (and) (axa) Ase = TF ayB)F-v(ana) = J (an0)Je(ana)” Bay = Hilent Ebon) ~ KE (oudyI5 (an) Foland) J, (ana) — I", (and) Solana)” “The results of this section can be used without essential changes, when consid- ering different particular problems on integrating differential equations which lead to the Bestel equation or to a system of the Bessel equation (see Section 17)? 25, Schlémlich series Usually, a series of the form 0/2+ J dm J{mz), (2) where the index does not depend on the number and the argument is proportional ‘to the number, is called the Schlémlich series. First, consider the Schlimlich series for zero index of the Bessel function. We prove that a continuous function f(z) for 0 < z < m admits an expansion into the series fle) = 2+ Yemsaime) (25.2) where a an = 24(0) +2 J wf twsing) dou, oa. 2 (25.3) cn=2] Partusingscormdpcs, if In this case it is required that f‘(z) is continuous and of bounded total variation for00,a dmJolim), (25.17) 4 where aor J :9(0)Jobime) du, 3 m2 g(u) = f(0) +u if (usin @) do. 3 ‘Then from (25.4) we obtain wo xy f Tolima sin p) dip. (25.18) mad f(z) = Using the formula Piseinn aes 5.3( which can be obtained from the formula on page 59 after the substitutions v = n= 0, Z,(a) = Jy(a), we finally find Se) = 3 00st (32) (25.19) mt However, this expansion in squared Bessel functions which is obtained following the method of Schlémlich, does not satisfy the main property of the Schlémlich series, which consists in the fact that the argument is proportional to the number of the terms of the series. Let us introduce an auxiliary function F(2) = 92 + 5 an cosms, O end(masin)] 0 an = B42 Som f se(mesineyas = x ons (2). (25.24) mat Now suppose that the function g(z) can be expanded into the Fourier-Bessel or Dini series on a segment [a, 8] (the formulae (24.13) or (24.22)). In this case the function g(z) can be expanded into the series whose general term is a linear combination of cylindrical functions of a common index. Restricting ourselves to the case when this index is equal lo zero, we obtain that if g(z) is expanded into a series with the general term Cen Jol jen) + de Yo(im2), (25.25) then the expansion of the function f(z) has the form fe) = a7 [cmd (2) + dn¥o (=) Jo 3) (25.26) Here we have used the formula m2 J Ytesing) ao= 7()*(3). 3 which can be obtained from (18.15) at n = 0. Consider the Schlimlich series in which the argument of the cylindrical function is equal to z= VE¥-+ mz, where m is the number of the term of the series and k is a non-zero constant.102. DBFINITE AND IMPROPER INTEGRALS. SERIES IN BESSEL FUNCTIONS ‘The expansion of the function f(z) into the series He) = }eolo(ke) + 9° ame (VEEFme) (227) oes, where | 12)Gm(e,) de, Gm(e,k) = CD osha =a) f mz cosh —aF) sin me dt vata at Via aT . huay eeu ubtained in [07]. ‘The proof is analogous to the one given before when obtaining the expansion (25.2), the difference consists in the fact that the Schlimlich equation obtained in [67] ae fe) 2 f (ein) coo(ke cos 8) d0, (25.28) 3 which can also be reduced to the Abel equation, is considered instead of (25.4). Representing the function g(z) in the form of a Fourier series in the cosines and using the Parseval integral, we immediately obtain (25.27). Let us give following Cooke's method [4], another proof which is based on the using of the Sonin integral Consider the Sonin integral (sce Section 20): *h Af Falosind)s,{bc08 0) sin" Boos 040 HO (a? + YPN, ass (Var +H), Ru>-l, Rv> Let us set a=2Vm? +R, ike, v=—p-1/2. ‘Then =p if Jp [Vent + Be sind] J, 1/al ike cond) sin #0 cos ##9/? 040 [2?(om? + 22)]""? (dke)-#-V9(me)-Y?Jyya(me). (25.20) Using the formulae Ji2(mz) Vi[(ama}sin me, (ne DI Tepaapalke 00), 12(ikz cos 8) = exp f25. SCHLOMLICH SERIES m we rewi (25.29) in the following form 2 if Jn [ Vin? + Basin] J_4-1/2(k2 0086) sin** Boe" #*4/? 9 dB = fe it 7 (hay oo [Zain (25.0) = me Suppose that we have the following expansion of the function f(2): -< 3p) vma(kay+¥/2 f(2) = Lowe (Vat+ 2) Vilet(mn? + Hyp? (25.31) O

You might also like

• Euler's Textbooks vs Modern Curriculum

  

  No ratings yet

  Euler's Textbooks vs Modern Curriculum

  15 pages
• Integrals of Bessel Functions

  

  No ratings yet

  Integrals of Bessel Functions

  13 pages
• Mathematical Analysis - Apostol (2E)

  

  100% (1)

  Mathematical Analysis - Apostol (2E)

  256 pages
• Solitons and Chaos I. Antoniou and J. F. Lambert

  

  No ratings yet

  Solitons and Chaos I. Antoniou and J. F. Lambert

  340 pages
• Variational Methods for Equilibrium

  

  100% (1)

  Variational Methods for Equilibrium

  23 pages
• AA Affine Spaces

  

  No ratings yet

  AA Affine Spaces

  22 pages
• Very Basic Lie Theory PDF

  

  No ratings yet

  Very Basic Lie Theory PDF

  24 pages
• Material Stresses and EM Boundary Conditions

  

  No ratings yet

  Material Stresses and EM Boundary Conditions

  12 pages
• Geometric Modeling for Engineers

  

  No ratings yet

  Geometric Modeling for Engineers

  502 pages
• The Hyperbolic Theory of Special Relativity

  

  No ratings yet

  The Hyperbolic Theory of Special Relativity

  109 pages
• Algebraic Geometry

  

  No ratings yet

  Algebraic Geometry

  331 pages
• Edward A. Spiegel - The Theory of Turbulence - Subrahmanyan Chandrasekhar's 1954 Lectures (Lecture Notes in Physics, 810)

  

  No ratings yet

  Edward A. Spiegel - The Theory of Turbulence - Subrahmanyan Chandrasekhar's 1954 Lectures (Lecture Notes in Physics, 810)

  136 pages
• Galois Theory Without Abstract Algebra PDF

  

  No ratings yet

  Galois Theory Without Abstract Algebra PDF

  35 pages
• Solid State Physics: Problems & Labs

  

  No ratings yet

  Solid State Physics: Problems & Labs

  103 pages
• Lectures On Quasiconformal Mappings Ahlfors

  

  No ratings yet

  Lectures On Quasiconformal Mappings Ahlfors

  147 pages
• J. E. Akin-Application and Implementation of Finite Element Methods (Computational Mathematics and Applications) (1982)

  

  No ratings yet

  J. E. Akin-Application and Implementation of Finite Element Methods (Computational Mathematics and Applications) (1982)

  381 pages
• Aiwa XA-008-Service-Manual

  

  100% (1)

  Aiwa XA-008-Service-Manual

  13 pages
• John Earman Primer On Determinism

  

  No ratings yet

  John Earman Primer On Determinism

  31 pages
• Continuous Groups For Physicists (N. Mukunda, Subhash Chaturvedi) (Z-Library)

  

  No ratings yet

  Continuous Groups For Physicists (N. Mukunda, Subhash Chaturvedi) (Z-Library)

  299 pages
• PDJ Endre Pap, Arpad Takaci

  

  No ratings yet

  PDJ Endre Pap, Arpad Takaci

  415 pages
• Ordinary Differential Equations - G. F. Carrier and C. E. Pearson - SIAM

  

  No ratings yet

  Ordinary Differential Equations - G. F. Carrier and C. E. Pearson - SIAM

  231 pages
• What Is Group Ring - Donald Passman

  

  No ratings yet

  What Is Group Ring - Donald Passman

  14 pages
• FW Lawvere - State Categories and Response Functors 1986

  

  No ratings yet

  FW Lawvere - State Categories and Response Functors 1986

  32 pages
• Polyhedral Geometry

  

  100% (1)

  Polyhedral Geometry

  442 pages
• Ziemer-Modern Real Analysis

  

  No ratings yet

  Ziemer-Modern Real Analysis

  403 pages
• The Kind of Motion We Call Hea - Brush, Stephen G. - 4438 PDF

  

  No ratings yet

  The Kind of Motion We Call Hea - Brush, Stephen G. - 4438 PDF

  338 pages
• Peter Topping Lectures On The Ricci Flow London Mathematical Society Lecture Note Series 2006

  

  No ratings yet

  Peter Topping Lectures On The Ricci Flow London Mathematical Society Lecture Note Series 2006

  134 pages
• Kubo - Statistical Mechanics PDF

  

  No ratings yet

  Kubo - Statistical Mechanics PDF

  434 pages
• Solid State Physics - Kittel

  

  No ratings yet

  Solid State Physics - Kittel

  408 pages
• Handbook of Elliptic Integrals

  

  No ratings yet

  Handbook of Elliptic Integrals

  370 pages
• GENERAL MECHANICS Being Volume I of Introduction To Theoretical Physics - Macmillan (1949)

  

  No ratings yet

  GENERAL MECHANICS Being Volume I of Introduction To Theoretical Physics - Macmillan (1949)

  281 pages
• Differential Equations Overview

  

  No ratings yet

  Differential Equations Overview

  165 pages
• First Course in The Theory of Equations

  

  No ratings yet

  First Course in The Theory of Equations

  207 pages
• Introduction To Koopman Operator Theory of Dynamical Systems

  

  No ratings yet

  Introduction To Koopman Operator Theory of Dynamical Systems

  24 pages
• How Euler Did It 59 PDEs of Fluids

  

  No ratings yet

  How Euler Did It 59 PDEs of Fluids

  7 pages
• Algebraic Vector Bundles: Problem List

  

  No ratings yet

  Algebraic Vector Bundles: Problem List

  12 pages
• Numerical Solutions of Second Order Boundary Value Problems by Galerkin Residual Method On Using Legendre Polynomials

  

  No ratings yet

  Numerical Solutions of Second Order Boundary Value Problems by Galerkin Residual Method On Using Legendre Polynomials

  11 pages
• Classical Fields Course Notes

  

  No ratings yet

  Classical Fields Course Notes

  203 pages
• Schur Complement and Positive Definite Matrices

  

  No ratings yet

  Schur Complement and Positive Definite Matrices

  12 pages
• Understanding Shannon's Information Theory

  

  No ratings yet

  Understanding Shannon's Information Theory

  13 pages
• Dog-Rabbit Chase Problem in Kinematics

  

  No ratings yet

  Dog-Rabbit Chase Problem in Kinematics

  5 pages
• On Darwin Lagrangian

  

  No ratings yet

  On Darwin Lagrangian

  4 pages
• Data Driven Model Discovery and Coordinate Embeddings For Physical Systems by Nathan Kutz, University of Washington

  

  No ratings yet

  Data Driven Model Discovery and Coordinate Embeddings For Physical Systems by Nathan Kutz, University of Washington

  43 pages
• Finsler-Randers Spacetime Dynamics

  

  No ratings yet

  Finsler-Randers Spacetime Dynamics

  20 pages
• Resolving The Banach-Tarski Paradox

  

  100% (1)

  Resolving The Banach-Tarski Paradox

  27 pages
• Thermodynamics With Internal State Variables - Coleman - Gurtin - 1967

  

  No ratings yet

  Thermodynamics With Internal State Variables - Coleman - Gurtin - 1967

  18 pages
• Periodically Loaded Line Phase Shifters Report1

  

  No ratings yet

  Periodically Loaded Line Phase Shifters Report1

  7 pages
• (Analytical Methods and Special Functions v.8) Korenev G - Bessel Functions and Their Applications-Chapman & Hall - CRC (2002)

  

  100% (2)

  (Analytical Methods and Special Functions v.8) Korenev G - Bessel Functions and Their Applications-Chapman & Hall - CRC (2002)

  286 pages
• Bessel Preview

  

  No ratings yet

  Bessel Preview

  19 pages
• Treatise On Bessel Functions and Their Applications To Physics

  

  100% (1)

  Treatise On Bessel Functions and Their Applications To Physics

  316 pages
• Bessel Funtions - Epelde García

  

  No ratings yet

  Bessel Funtions - Epelde García

  69 pages
• Bessel Functions With Pure Imaginary Indices 0910.0365

  

  No ratings yet

  Bessel Functions With Pure Imaginary Indices 0910.0365

  10 pages
• Gray Matthews Bessel

  

  100% (1)

  Gray Matthews Bessel

  304 pages
• Treatise On The Theory of Bessel Functions

  

  100% (1)

  Treatise On The Theory of Bessel Functions

  815 pages
• Special Functions For Scientists and Engineers - W.W. Bell

  

  100% (2)

  Special Functions For Scientists and Engineers - W.W. Bell

  258 pages
• 7491 Dmth517 Special Functions and Integral Equation

  

  No ratings yet

  7491 Dmth517 Special Functions and Integral Equation

  162 pages
• Bessel Function

  

  No ratings yet

  Bessel Function

  18 pages
• Bessel Functions Overview and Applications

  

  No ratings yet

  Bessel Functions Overview and Applications

  40 pages
• Web Chap4

  

  No ratings yet

  Web Chap4

  40 pages
• Sneddon Special Functions of Mathematical Physics and Chemistry

  

  100% (1)

  Sneddon Special Functions of Mathematical Physics and Chemistry

  99 pages