OPERATIONAL MATHEMATICS | ©“!Operational Mathematics Third Edition Ruel V. Churchill Professor Emeritus of Mathematics University of Michigan McGraw-Hill Book Company New York St.Louis San Francisco Diisseldorf_ Johannesburg Kuala Lumpur London Mexico Montreal New Delhi Panama Rio de Janeiro Singapore Sydney TorontoOperational Mathematics Copyright © 1958, 1972 by McGraw-Hill, Inc. All rights reserved Copyright 1944 by McGraw-Hill, Inc. All rights reserved Printed in the United States of America. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the publisher. Library of Congress Catalog Card Number 70-174611 07-010870-6 1234567890 MAMM 7987654321 This book was set in Times Roman, and was printed and bound by The Maple Press Company. The designer was Jo Jones; the drawings were done by John Cordes, J. & R, Technical Services, Inc. The editors were Howard S. Aksen and Madelaine Eichberg. Matt ‘Martino supervised production.Contents Preface Chapter 1. The Laplace Transformation 1 2 3. 4, 5 6 7. 8 9. 10. Introduction Definition of the Laplace Transformation Sectionally Continuous Functions. Exponential Order Transforms of Derivatives Examples, The Gamma Function The Inverse Transform ‘A Theorem on Substitution ‘The Use of Partial Fractions (Table 1) The Solution of Simple Differential Equations Generation of the Transformation Chapter 2. Further Properties of the Transformation 1 12 1B 14, 15. 16, 1, 18, 19, 20. 2 2. Translation of F(t) Step Functions The Impulse Symbol S(t ~ to) Integrals Containing a Parameter Improper Integrals Convolution Properties of Convolution. Difieremtial and Integral Equations Derivatives of Transforms Series of Transforms Differential Equations with Variable Coefficients Integration of Transforms 2 29 33 39 4 43 Sega 6 6523, 24, 25. 26. 20 Periodic Functions |. Partial Fractions Repeated Linear Factors Quadratic Factors Tables of Operations and Transforms Chapter 3. Elementary Applications 28, 29. 30. 31 32 33 34, 35. 36. 31, 38, 39, Free Vibrations of a Mass on a Spring Forced Vibrations without Damping Resonance Forced Vibrations with Damping A Vibration Absorber Electric Circuits Evaluation of Integrals Exponential- and Cosine-integral Functions Static Deflection of Beams The Tautochrone Servomechanisms Mortality of Equipment Chapter 4, Problems in Partial Differential Equations 40. 41 a2 48. 44. 4s. 46. 47 ‘The Wave Equation Displacements in a Long String A Long String under Its Weight ‘The Long String Initially Displaced A Bar with a Prescribed Force on One End Equations of Diffusion Temperatures in a Semi-infinite Solid Prescribed Surface Temperature 48. Temperatures in a Slab 49, A Bar with Variable End Temperature 50. A Cooling Fin or Evaporation Plate st 2 Chap Temperatures in a Composite Solid Observations on the Method ter 5. Functions of a Complex Variable 53. Complex Numbers 54, Analytic Functions 55. Exponential and Trigonometric Functions 56. Contour Integrals 57. Integral Theorems 58. Power Series 59. Singular Points and Residues CONTENTS 66 70 B 8 B 85 85 88 91 95 96 102 106 110 13 1s 17 19 123 123 126 130 133 135 143 145 146 151 153 153 155 159 162 162 163 166 168 170 im 1BCONTENTS, vil (60. Branches of Multiple-valued Functions 7 61. Analytic Continuation m9 62. Improper Cauchy Integrals 181 Chapter 6. The Inversion Integral 186 63. Analytic Transforms 186 64. Permanence of Forms 188 65. Order Properties of Transforms 189 66, The Inversion Integral 193 67. Conditions on f(s) 195 68. Conditions on F(t) 198, 69. Uniqueness of Inverse Transforms 201 70. Derivatives of the Inversion Integral 202 71. Representation by Series of Residues 206 72. Residues at Poles 208 73. Validity of the Representation by Series 210 74, Alterations of the Inversion Integral 213 Chapter 7. Problems in Heat Conduction 219 75, Temperatures in a Bar with Ends at Fixed Temperatures 220 76. The Solution Established m2 77. The Series Form Established 224 78. Properties of the Temperature Function 226 79. Uniqueness of the Solution 228 80, Arbitrary End Temperatures 230 81. Special End Temperatures 232 82, Arbitrary Initial Temperatures 237 83, Temperatures in a Cylinder 240 84. Evaporation from a Thick Slab 245 85. Duhame’s Formula 247 Chapter 8. Problems in Mechanical Vibrations 253 86. A Bar with a Constant Force on One End 253 87. Another Form of the Solution 256 8. Resonance in the Bar with a Fixed End 258 89. Verification of Sotutions 259 90. Free Vibrations of a String 264 91, Resonance in a Bar with a Mass Attached 266 92. Transverse Vibrations of Beams 268 93, Duhamel’s Formula for Vibration Problems 270 Chapter 9. Generalized Fourier Series 216 94, Self-adjoint Differential Equations cu95. 96. 77. 98, 99. 100. 101. 102 103. 104. 108. 106. 107. Green's Functions Construction of Green’s Function Orthogonal Sets of Functions Eigenvalue Problems A Representation Theorem ‘The Reduced Sturm-Liouville System A Related Boundary Value Problem The Transform y(x,5) Existence of Eigenvalues ‘The Generalized Fourier Series ‘Steady Temperatures in a Wall Verification of the Solution Singular Eigenvalue Problems Chapter 10. General Integral Transforms 108, 109, 110. 1 112. 113. 114, 118. 116. 117. 118. Linear Integral Transformations Kernel-product Convolution Properties Example Sturm-Liouville Transforms Inverse Transforms Further Properties Transforms of Certain Functions Example of Sturm-Liouville Transformations Singular Cases A Problem in Steady Temperatures Other Boundary Value Problems Chapter 11. Finite Fourier Transforms 119. 120. 121 122 123 124 125, 126. 127 128. 129. 130. Finite Fourier Sine Transforms Other Properties of S, Finite Cosine Transforms Tables of Finite Fourier Transforms Joint Properties of C, and S, Potential in a Slot Successive Transformations ‘A Modified Sine Transformation Generalized Cosine Transforms A Generalized Sine Transform Finite Exponential Transforms E,{F} Other Properties of Chapter 12, Exponential Fourier Transforms 131 The Transformation E,{F} CONTENTS 278 281 283, 288, 21 292 293 294 296 299 305 308 309 37 317 319 320 325 327 329 332 333 336 34 343 348 348, 350 354 356 357 360 362 368 370 372 316 378 383CONTENTS ix 132. The Inverse Transformation 385 133. Other Properties of E, 388 134. The Convolution Integral for E, 391 135. Convolution Theorem 333 136. Tables of Transforms 396 137, Boundary Value Problems 397 Chapter 13. Fourier Transforms on the Half Line 401 138. Fourier Sine Transforms f(a) 401 139, Fourier Cosine Transforms f(a) 404 140. Further Properties of S, and C, 405 141. Convolution Properties 406 142. Tables of Sine and Cosine Transforms 407 143, Steady Temperatures in a Quadrant 408 144, Deflections in an Elastic Plate 410 143, A Modified Fourier Transformation T, 414 146. Convolution for T, 416 147, Surface Heat Transfer 418 Chapter 14. Hankel Transforms 420 148. Introduction 420 149. Finite Hankel Transformations 21 150, Inversion of H,, 423 151, Modified Finite Transformations Hy 44 152. A Boundary Value Problem 26 153, Nonsingular Hankel Transformations 431 154, Hankel Transformations H, on the Half Line (x > 0) 432 155. Further Properties of Hy, 434 156, Tables of Transforms Hy,{F} 436 157. Axially Symmetric Heat Source 437 Chapter 15. Legendre and Other Integral Transforms 441 158. The Legendre Transformation T, on the Interval (~ 1,1) 442 159. Further Properties of T, a4 160. Legendre Transforms on the Interval (0,1) 446 161. Dirichlet Problems for the Sphere 448 162. Laguerre Transforms 452 163. Mellin Transforms 433 Bibliography 456 Appendixes Appendix A Tables of Laplace Transforms 458Table A.1 Table A.2 Appendix B Table B.1 Table B.2 Appendix C Appendix D Table D.t Table D.2 Index Operations Laplace Transforms Tables of Finite Fourier Transforms Finite Sine Transforms Finite Cosine Transforms Table of Exponential Fourier Transforms Tables of Fourier Sine and Cosine Transforms Sine Transforms on the Half Line Cosine Transforms on the Half Line CONTENTS 458 459 467 467 469 471 473 473 475 477Preface This is an extensive revision of the second edition of “Operational Mathematics” published in 1958. Chapters have been added on general integral transforms, finite Fourier transforms, exponential Fourier trans- forms, Fourier transforms on the half line, Hankel transforms, and on Legendre and other integral transforms. The presentation of theory and applications of the Laplace transformation has been revised. Tables of several of the most useful transforms now appear in the Appendix or in the text. Additional problems illustrate applications of the various integral transformations. The book is designed as a text and a reference on integral transforms and their applications to problems in linear differential equations, to boundary value problems in partial differential equations in particular. It presents the operational properties of the linear integral transformations that are useful in those applications. The selection of a transformation that is adapted to a given problem, by observing the differential forms and boundary conditions that appear in the problem, is emphasized in this edition. The Laplace transformation receives special attention because of its many useful opera- tional properties and the large class of problems to which it applies, in- cluding applications outside the field of differential equations. The applications to problems in physics and engineering are kept on a fairly elementary level. They include problems in vibrations or displacements in elastic bodies, in diffusion or heat conduction, and in static potentials. No previous preparation in the subject of partial differential equations is required of the reader. This book is a companion volume to “Fourier Series and Boundary Value Problems” and “Complex Variables and Applications.” The three books cover, respectively, these principal methods of solving linear boundaryxii PREFACE value problems in partial differential equations: the operational methods of integral transforms, separation of variables and Fourier series, and con- formal mapping. Generalized Fourier series and their applications are presented in Chapter 9 here, with the aid of the theory of the Laplace trans- formation. A summary of useful theory of functions of a complex variable is given in Chapter 5. All three books are intended to present sound mathe- matical analysis as well as applications. Conditions of validity of analytical results are kept on a simple and practical level. More elegant conditions may call for training in analysis beyond the level of advanced calculus. The first four chapters are designed to serve as a text for a short course in real Laplace transforms and their applications. The impulse symbol or “delta function” is introduced in Sec. 13 in an elementary and careful manner. No theory of distributions, or generalized functions, is included in the book. A satisfactory presentation of the theory would require considerable space and the introduction of concepts not needed elsewhere in the text. Neither can the author justify a presentation of the abstract theory of linear spaces in this book. An intuitive approach from vectors to functions (Secs. 10 and 97) serves as a guide for writing inner products of functions as integrals. In preparing the three editions of this book the author has taken advan- tage of improvements suggested by many students and teachers. He is grateful to them for that assistance; also to authors referred to in the Bibliog- raphy and footnotes, whose publications have influenced the selection of ‘material. Ruel V. Churchill1 The Laplace Transformation 1 INTRODUCTION The operation of differentiating functions is a transformation from functions F(t) to functions F'(d). If the operator is represented by the letter D, the transformation can be written D{F(O} = FO. The function F(t) is the image, or the transform, of F(t) under the transforma- tion; the function 32, for example, is the image of the function ¢°. Another transformation of functions that is prominent in calculus is that of integration, {FO} = f Fede. The result of this operation is a functional f(x), the image of F(¢) under the transformation. A simpler transformation of functions is the operation of multiplying all functions by the same constant, or by a specified function2 sec. 1) OPERATIONAL MATHEMATICS In each of the above examples inverse images exist; that is, when the image is given, a function F(t) exists which has that image. A transformation T{F(0)} is linear if for every pair of functions F,(t) and F,(t) and for each pair of constants C, and C,, it satisfies the relation w T{C,F(O + CoF2(0} = Ci T{FO} + C2T {FO} Thus the transform of a linear combination of two functions is the same linear combination of the transforms of those functions, if the transformation islinear. Note the special cases of Eq. (1) when C, = Oand when C, = C, = 1. The examples cited above represent linear transformations. The class of functions to which a given transformation applies must generally be limited to some extent. The transformation D{F(t)} applies to all differentiable functions, and the transformation 1{F(0)} to all integrable functions. Linear integral transformations of functions F(t) defined on a finite or infinite interval a < ¢ Q) T{F(} = f K(es)F de. It represents a function f(s), the image, or transform, of the function F(t). The class of functions to which F(t) may belong and the range of the param- eter 5 are to be prescribed in each case. In particular, they must be so prescribed that the integral (2) exists. We shall see that with certain kernels K(t,s) the transformation (2), when applied to prescribed linear differential forms in F(t), changes those forms into algebraic expressions in f(s) that involve certain boundary values of the function F(t), Consequently, classes of problems in ordinary differential equations transform into algebraic problems in the image of the unknown function. Ifan inverse transformation is possible, the solution of the original problem may be determined. Boundary value problems in partial differential equations can be simplified in a similar way. The operational mathematics presented in this book is the theory as well as the application of such linear integral transformations that bears on the treatment of problems in ordinary or partial differential equations. Later on, we shall return to the general transformation (2) and to the question of deciding upon the special cases that may apply to a given problem in differ- ential equations. First we present the special case that is of greatest general importance, the operational mathematics of the Laplace transformation. Other prominent cases include the various Fourier transformations, to be presented later.THE LAPLACE TRANSFORMATION [sec.2 3 When a=0 and b= © and K(t,s) = e~™, the transformation (2) becomes the Laplace transformation. The direct application of this trans- formation replaces the earlier symbolic procedure known as Heaviside’s operational calculus.’ The development of the transformation and the accompanying operational calculus was begun before Heaviside’s time; Laplace (1749-1827) and Cauchy (1789-1857) were two of the earlier con- tributors to the subject.” In this chapter we present the basic operational property of the Laplace transformation, the property that gives the image of differentiation of functions as an algebraic operation on the transforms of those functions. In the following chapters further properties of the transformation will be derived and applied to problems in engineering, physics, and other subjects. Appli- cations to boundary value problems in partial differential equations will be emphasized. Our study of the Laplace transformation leads to the theory of expand- ing functions in series of the characteristic functions of Sturm-Liouville systems. Such expansions form the basis of the method of solving boundary value problems by separation of variables, a classical method of great importance in partial differential equations.? Furthermore, we can use that theory to adapt the integral transformation (2) to certain types of linear boundary value problems. 2 DEFINITION OF THE LAPLACE TRANSFORMATION Ifa function F(0), defined for all positive values of the variable t, is multiplied by e~*‘ and integrated with respect to ¢ from zero to infinity, a new function f(s) of the parameter s is obtained ; that is, j eo *F(t) dt = f(s). ° As indicated in the preceding section, this operation on a function F(t) is called the Laplace transformation of F(t). It will be abbreviated here by the symbol L{F}, or by L{F(t)}; thus L{F} = f eo "F(t dt. ° ‘The new function f(s)iscalled the Laplace transform, or the image, of the object function F(t). Wherever it is convenient to do so, we shall denote the * Oliver Heaviside, English electrical engineer, 1850-1925. ? For historical accounts see J. L. B. Cooper, Heaviside and the Operational Calculus, Math. Gazette, vol. 36, pp. 5-19, 1952, and the references given there. * That method is presented in the author’s “Fourier Series and Boundary Value Problems,” 2d ed., 1963.4 SEC. 2) OPERATIONAL MATHEMATICS object function by a capital letter and its transform by the same letter in lowercase. But other notations that distinguish between functions and their transforms are sometimes preferable ; for example, $9) = L{S(O} or Hs) = Li} For the present, the variable s is assumed to be real. Later on, we shall let it assume complex values. Limitations on the character of the function F(t) and on the range of the variable s will be discussed soon. Let us note the transforms of a few functions. First, if F(t) = 1 when t > 0, then L{F} = foe dt = ° hence, when s > 0, Ll} = With the aid of elementary methods of integration, the transforms of many other functions can be written. For instance, 1 2 Lij=a Le}=S . k s L{sin kt} =a. L{eoskt} = 3 when s > 0; but soon we shall have still simpler ways to obtain those trans- forms. It follows from elementary properties of integrals that the Laplace transformation is linear in the sense defined by Eq. (1), Sec. 1. We can illus- trate the use of this property by writing when s > kands > —k; that is, L{sinh kt} = (s > [kDTHE LAPLACE TRANSFORMATION (sec. 35 PROBLEMS 1. Use the linearity property and known transforms to obtain these transformations, where a, b, and c are constants: (a) L{a + by = S42 (>0); (b) L{a + bt + ct?} = L{(a + bt) + ct?} _ as? + bs + 2c (>); s . (© Llasint + beost} = wat (> 0); (@) L{cosh ct} = 2 (s > Id): at ght a—b b. © Let) = (s > aands > 6). 2. Use trigonometric identities, such as 2cos?¢= 1 + cos2r, 2sin at sin bt = cos (a — b)t — cos (a + bt, and known transforms to find f(s) when s > 0, in case Fis (a) cos? t; (b) sin? t; (c) sin t sin 2t; (d) sin t cos t;(e) sin? ¢ = 4(1 — cos 21) sin ¢. 842 2. 4s Ans.@aray Ore ORTDETS' 1 6 One ORE 3. Show that the linearity property (1), Sec. 1, can be extended to linear combinations of three or more functions when all the transforms exist. 4. If for all functions of some class and for every constant C a transformation T satisfies the two conditions T{FO) + GO} = TLFO} +7(G(0}, T{CF()} = CT{FO}, prove that the transformation is linear. 3 SECTIONALLY CONTINUOUS FUNCTIONS, EXPONENTIAL ORDER A function F(t) is sectionally continuous on a bounded interval a k, is an example of a function that is sectionally continuous on the interval 0 0, A function F(t) is of exponential order as ¢ tends to infinity, provided some constant « exists such that the product eo F(O) is bounded for all t greater than some finite number T. Thus |F(t)| does not grow more rapidly than Me" as ¢ > oo, where M is some constant. This is also expressed by saying that F(t) is of the order of e*, of that F(t) is O(¢"). The function S,(z) above, as well as the function £", is of the order of 2" as t+ 00 for any positive «; in fact, for the first function and, when n = 0, for the second, we may write « = 0. The function eis of exponential order (a = 2); but the function e'’ is not of exponential order. If a function F(t) is sectionally continuous on each bounded interval 0 a. This follows from a well-known comparison test S4lt) Fig.1‘THE LAPLACE TRANSFORMATION [sec 47 for the convergence of improper integrals (see Prob. 14, Sec. 5). For in view of the sectional continuity of F, and consequently of the product e~*F(0), that product is integrable over every bounded interval 0 <1 < T. Also, since F is O(c"), a constant M exists such that for all positive ¢ le" Fn] < Me~"-, But the integral from 0 to co of Me~*~* exists when s > a; consequently, not only the convergence but also the absolute convergence of the Laplace integral j e-*F(e) dt 0 is ensured when s > a. The above conditions for the existence of the transform of a function are adequate for most of our needs; but they are sufficient rather than necessary conditions. The function F may have an infinite discontinuity at 1 = Ofor instance, that is,|F(t)| + 00 as t + 0, provided that positive numbers m, N, and T exist, where m < 1, such that |F(0) < N/t™ when 0<1 0). ° Vs The last integral has the value \/x/2 (Prob. 10, Sec. 5); hence L{t"#} = Qj (s > 0) 4 TRANSFORMS OF DERIVATIVES By a formal integration by parts we have L{F(0} = r oF (dt 0 = oro. + sf eo F(t) dt. 0 °a SEC. 4] OPERATIONAL MATHEMATICS Let F(t) be of order of e* as t approaches infinity. Then whenever s > a, the bracketed term becomes — F(0), and it follows that ) L{F(O} = f(s) — FO), where f(s) = L{F()} Therefore in our correspondence between functions differentiation of the object function corresponds to the multiplication of the result function by its variable s and the addition of the constant — (0). Formula (1) thus gives the fundamental operational property of the Laplace transformation, the property that makes it possible to replace the operation of differentiation by a simple algebraic operation on the transform. Asnoted above, formula (1) was obtained only in a formal, or manipula- tive, manner. It is not even correct when F(t) has discontinuities. The following theorem will show to what extent we can rely on our formula. Theorem 1 Let the function F(t) be continuous with a sectionally continuous derivative F'(0), over every finite interval 0S t S$ T. Also let F(t) be of order of e* as t +> co. Then when s > a, the transform of F'(o) exists, and Q L{F(0} = sL{F()} — FO). Since F(t) is continuous at ¢ = 0, the number F(0) here is the same as F(+0), the limit of F(t) as t approaches zero through positive values. To prove this theorem, we note first that 1 L{F(0} = lim |e“ F'(de, Todo if this limit exists. We write the integral here as the sum of integrals in each of which the integrand is continuous. For any given T, let fy, t2,---,, denote those values of t between ¢ = 0 and ¢ = T for which F(t) is discontinuous (Fig. 2). Then f e* F(t) dt = f oF (0) dt + i oF (dt bo + feo dt ° 2 5 (4.0) (0) (20) Fig. 2THE LAPLACE TRANSFORMATION [sec 49 After integrating each of these integrals by parts, we can write their sum as 2 r -ro|. + ro] ° a fot ro] + sf eo" F(a) de. ° Now F(t) is continuous so that F(t; — 0) = F(t, + 0), etc. and hence r r @) jf oF (0) dt = —F(0) + eT F(T) + sf oF dt ° ° Since |F(0)| < Me* for large ¢ for some constants a and M, it follows that leTR(T)) < Me“#"?, and since s > a, this product vanishes as T+ oo. Also the last integral in Eq. (3) approaches L{F} as T—> co because F is continuous and O(e*. Hence the limit as T-> co of the right-hand member of Eq. (3) exists and equals — F(0) + sf(s); therefore, the same is true of the left-hand member. Thus Theorem | is proved. If F is continuous when t = 0 except for a finite jump at to, where tg > 0, the other conditions remaining as stated in the theorem, the above proof is easily modified to show that our formula (2) must be replaced by the formula @ L{F(O} = sf(s) — FO) — [F(to + 0) — Flto ~ OJe™*®. The quantity in brackets is the jump of F at to. ‘We use the symbol F’ here and in the sequel to denote the function whose value is the derivative of F wherever the derivative exists. In the case of our step function S,(0), for instance, Si(t) = 0 when 0 k, but Si(k) has no value. To obtain the transform of the derivative F” of the second order, we apply Theorem 1 to the function F’. Let both F and F’ be continuous when 20 and O(e"); also let F’ be sectionally continuous on each bounded interval. Then L{F"(O} L{F()} — FO) = s[sL{F(0)} — F(O)] — FO). Hence we have the transformation ) L{F"(O} = s°f(s) — SFO) — FO). When Theorem 1 is applied to F"~ (1) to write L{F()} = SLUR" (H} — F=0) and again to write L{F"~ "(i)} in terms of L{F°”?\(0)}, and so on, the follow- ing result is indicated.10 SEC. 5) OPERATIONAL MATHEMATICS Theorem 2 Let F and each of its derivatives of order up to n — 1 be con- tinuous functions when t = 0 and O(e%); also let F(t) be sectionally continuous on each bounded interval 0 0 and it has the following algebraic expression in terms of the transform f(s) of F(t): (6) L{F@(t)} = s*f(s) — s"-F(O) — s"-7F'(0) FO) — ++ = FO) (5 > a). -s Theorem 2 can be proved by using Theorem | and induction. Under the conditions stated, suppose that formula (6) is valid when nis replaced by some integer k where 0 0), or L{l} = sL{t}. Since L{1} = 1/s,it follows that Lit} (s> 0). Example 2. Find L{sin kt}. The function F(t) = sin kt and its derivatives are all continuous and bounded, and therefore of exponential order, where a = 0. Hence L{F"()} = s*L{F(t)} — sFO) — FO) (s > 0), or —K*L{sin ke} = s?L{sin ke} — k.THE LAPLACE TRANSFORMATION [secs 11 Solving for L{sin kt}, we see that k + L{sin kt} = (s> 0). Example 3. Find L{t"} where m is any positive integer. The function F(t) = ¢ satisfies all the conditions of Theorem 2 for any positive a. Here F(Q) = F(Q) = +++ =F") = 0, FO\) =m), Ft) = 0. Applying formula (6) when n = m + 1, we find that LAF MQ)} = 0 = s™*tL fem) — ml, and therefore a) (s > 0), This formula can be generalized to the case in which the exponent is not necessarily an integer. To obtain L{t*} where k > —1, we make the substitution x = st in the Laplace integral, giving [oecwa ; The integral on the right represents the gamma function, or factorial function, with the argument k + 1. Hence Q) Ly ee k> -1,s> 0) f xte“* dx. (s > 0). ° Formula (1) is a special case of (2) when k is a positive integer (see Prob. 13). Example 4 Find L{fi, F(t)de} when F is sectionally continuous and of exponential order. The function 8) ow = f "FG de : is continuous (Sec. 14), and G(0) = 0. Also G'(t) = F(0), except for those values of t for which F(t) is discontinuous; thus G'(t) is sectionally continuous on each finite interval. If the function G(t) is also O(e*), then according to Theorem 1, L{G'(O} = sL{GW} = L{FO} (>a;12 sec. 5) OPERATIONAL MATHEMATICS thus, if > 0 so that s > 0, @) uff F(t) act = ° To show that the integral (3) represents a function of exponential order when F is sectionally continuous and of exponential order, we first note that constants «and M exist such that |F(0)| < Me” whenever t 2 0, and if the number « is not positive, it can be replaced by a positive number. Then f(s) (s>a> 0). Go| s freer dt< Mf ed = Mee -1) (#>0), o o « and therefore eat < Ma — ea eM (@>0) a a This establishes the exponential order of the function (3). PROBLEMS 1. State why each of the functions 1 (@ Fe = e* and (6) Go) -{: i When O 2, is sectionally continuous on every interval 0 < ¢ < Tand O(e”) as t + 00, where a > 2 for F and a 2 0 for G. 2. State why neither of the functions (a)(¢ — 1)~* or (b) tan cis sectionally continuous on the interval 0 < t < 3. 3. If $,(# isthe unit step function (See. 3), draw graphs of the following step functions and find their transforms when s > 0: (a) F() = 1 — S,(0; (6) G(t) = S\() — S,(0. Also, note that both F and G vanish except on bounded intervals; thus their Laplace integrals become definite integrals that exist for all s(—oo 0); (b) L{sinh ke} = ks? — k?) (5 > Ik). 5. Given the transform of e!, use Theorem 1 to show that 1 Lite} = (>h. (s—kPTHE LAPLACE TRANSFORMATION [sec 5 13 6. IfG(0) = Owhen 0 <¢ < kand G(t) = ¢ — k when > k, draw graphs of Gand G’. Given the transform of S,(:), (a) apply Theorem 1 to prove that g(s) = s~7e~** (s > 0); (6) show that G(e) = f',5,(0) dz and find g(s) from formula (4), See. 5. 7. Prove that the function F(t) = sin (e”) is of exponential order ( 2 0) and that its derivative F' is not of exponential order. Show that Theorem I ensures the existence of the Laplace transform of F’ when s > 0, in this case where F’ is not of exponential order. 8. (a) Derive Eq. (4),Sec. 4, (6) Illustrate that equ: of the unit step function S,(¢). 9. (a) Ifa function F and its derivative F’ are both O(e*) and continuous when ¢ 2 0 except possibly for finite jumps at a point fg, and if F” is sectionally continuous on each interval 0 < ¢ < T, apply Eq. (4), Sec. 4, to derive the formula jon by using it to find the transform L{F"(0} = s*f(s) — SFO) — FO) — se“*°[F(to + 0) — Fito — 0)] =e "LF (ty + 0) — Fito — 0)) (s > a, to > 0). (b) Show that the formula applies to the function FQ) =sintwhenOStSx, Flt)=Owhent2 x to give f(s) = (I + e-")/(s* + 1) when 00 <5 < 0. 10. Let J denote the second integral in Eq, (1), Sec. 3; then ea eaxf er ay= f j e749) de dy. 0 0 lo Jo Evaluate the iterated integral here by using polar coordinates and show that J = \/r/2. 11. Prove that each linear combination AF(t) + BG(t) of two functions F and G of exponential order is also of exponential order. 12. Use properties of continuous functions to show that iftwo functions are sectionally continuous on an interval (a,b) then (a) each linear combination of the two is also sectionally continuous on that interval ;(b) the product of the two functions is sectionally continuous on the interval 13. As noted in See. 5, the gamma function is defined for positive values of r by the formula re=f xe dx (> 0) 0 (@) Integrate by parts to show that the function has the factorial property Te +1) = 11. (&) Show that (1) = 1, and hence that P( + 1) = n! when n= 1, (©) From the value of the integral J found in Prob. 10, show that T() ‘Then use the factorial property to find (3) and formula (2) to show that L{,/t} = 4,/n/s!. 14. Improper integrals Consider only functions that are sectionally continuous on each interval 0 <¢ < T.14 sece) OPERATIONAL MATHEMATICS (a) If 0 < g(t) < (e) whenever ¢ > 0 and iff h(e) dt exists, use the definition of the improper integral i g(0) de = lim f g(t) dt to prove the existence of that integral. [Note that the value of the last integral is non- decreasing as T increases. It never exceeds I where I = J; h(t) dr because {7 g(t) dt S JEM) de < 1, so it has a limit.] (b) If f@ |p(o)| de exists, prove that p(t) de exists by writing p(s) = [p(¢) + |p(Ol] — Ip(e. Note’ that 0 < p(t) + |p(e) S 2p(0)| and apply the test in part (a) to those two components of p(é) (0) Prove this comparison test: If |q(t)| S r(t) whenever t > 0 and if [jr dt exists, then the improper integral {~ q(t) dt is absolutely convergent, and the integral itself exists, 6 THE INVERSE TRANSFORM Let the symbol L~ "{ f(s)} denote a function whose Laplace transform is f(s). Thus if L{F(O} = f(s), then F(t) = L“{f(s)}- Using two of the transforms obtained in the foregoing sections, we can write, for instance, wife Z if =e, uf tat = sin kt. This correspondence between functions f(s) and F(t) is called the inverse Laplace transformation, F(t) being the inverse transform of f(s). In the strict sense of the concept of uniqueness of functions, the inverse Laplace transform is not unique. The function F(t) = e is an inverse transform of 1/(s — k); but another, for instance, is the function (Fig. 3) Fy) =e when 0 <1 <2,ort > 2, 1 when t = 2. For the transform of F(t) is 2 2 5 f oF) dt = jf ee dt +f ee dt, 0 0 2 and this is the same as L{et}. The function F,(t) could have been chosen equally well as one that differs from F(t) at any finite set of values of ¢, or even at such an infinite set as ¢ = 1, 2,3,..THE LAPLACE TRANSFORMATION [sec.6 15 Fig. 3 A theorem on uniqueness of the inverse transform will be proved later (Sec. 69). To state it, we first define a class of functions of exponential order. Let the class & denote the set of all functions F(t) defined on the half line ¢ > 0, sectionally continuous on each bounded interval, and defined at each point fo where F is discontinuous as the mean value of its limits from the right and left, F(to) = 3[Flto + 0) + Flto — 0)) (to > 0); also, for each individual function F of the set, let constants M and a exist such that |F(¢)| < Me* when ¢ > 0. As we have seen, F then has a transform J(s) defined on some half line s > a. The theorem states that no two functions of class & can have the same transforms. Thus if f(s) is the transform of some function F(t) of class & then F(t) is the unique inverse transform L~"{ f(s)} in &. For example, the only function L~!{e~*/s} of class & is the unit step function S(t) defined in Sec. 3 if $,(1) is defined to be 3. As another example, the only function L~1{1/(s — k)} of class & is e It is well to note here that not every function of s is a transform. The kind of functions /(s) that are transforms of functions F(t) of broad classes are limited, as we shall see (Chap. 6), by conditions of regularity that include requirements that f be continuous on a half line s > « and that f(s) > 0 as sa. We have noted that if L{F} and L{G} exist, then (y) L{AF(t) + BG(0} = Af(s) + Bels) whenever A and Bare constants. Let us restrict our functions of t to those of class &, and functions of s to transforms of such functions. Then unique inverse transforms exist, and the linearity property (1) can be written Q) L”'{Af(s) + Bg(s)} = AF(t) + BGC) = AL'S (9)} + BL” {ais} 5 that is, L~! is also a linear transformation of functions.16 SEC. 7) OPERATIONAL MATHEMATICS The most obvious way of finding the inverse transform of a given function of s consists of reading the result from a table of transforms. A fairly extensive table is given in Appendix A. But we shall take up methods of obtaining inverse transforms of certain combinations and modifications of functions of s, as well as methods of resolving such functions into those listed in the tables. With the aid of such procedures, we shall be able to make much use of the transformation. In addition, there are explicit formulas for L~*{f(s)}. The most useful of these formulas involves an integral in the complex plane. To use this integral, we must let s be a complex variable and we must be prepared to employ some theorems in the theory of functions of a complex variable. 7 A THEOREM ON SUBSTITUTION Let a function F(t) be such that its Laplace integral converges when s >a. Then, replacing the argument of the transform f(s) by s — a, where a is a constant, we have f(s — a) = f #1 F(t) dt = fp eterna, 0 ° when s — a > a. Therefore @ Sls = a) = Le" FO} (s>a+a). Let us state this simple but important property as a theorem. Theorem 3 The substitution of s — a for the variable s in the transform corresponds to the multiplication of the object function F(t) by the function &, as shown in formula (1). To illustrate this property, let us recall that (m = 1,2,...58> 0). m! Hence qceFFn L{t"e"} (s >a), As another illustration, s L{c0s kt} = 35 (s>0), and therefore L{e™* cos kt} = +4 (s> —a). (s + a)? +k?THE LAPLACE TRANSFORMATION [sec.8 17 8 THE USE OF PARTIAL FRACTIONS (TABLE 1) A few examples will show how the theory of partial fractions can be used in finding inverse transforms of quotients of polynomials in s. In the next chapter, a more systematic use of this procedure will be introduced. Example 1 Find L~'{(s + 1)(s? + 2s)}. The denominator of the function of s here is of higher degree than the numerator and has factors that are linear and distinct. There- fore constants A and B can be found such that stl A, B Ms+2) 5 *s42 for all values of s except 0. and —2. Clearing fractions, we have s+1=(A + B)s + 24, and this isan identity if A + B = Land 2A = 1. Thus A = B= $,and hence s+1 11,11 S425 25° 2542" Since we know the inverse transforms of the two functions on the right, we have the result s+ 1 1 Lo aby lyn {a + xt 272 The procedure can be shortened for such a simple fraction by writing s+1=Hs42) 44s, s¢l tt 1 and hence ee W(s+2) 2s 254+2 2 Example 2 Find L~'}—* | s(s + a) In view of the repeated linear factor, we write a A B c a= + tang. stapes sta (stay Clearing fractions and identifying coefficients of like powers of s as before, or else by noting that a? =(s + a)? — fs + a) — as,Table 1 A short table of transforms Fe) S09) a(s> a) ria a ° 2/e a 3 ) | ae ° 4 fet oetnod | oe a 5 | sink S 0 sin ke mae 6 | coske — ° 7 (| sinh ke 1 8 | cosh ke Mi 9 | em sinke —— = “ ora ee “ sta 10 | e-*cos ki a - con craree “ " ° 2 ° Tk+1) 3 | A@>-1) “a o . Tk +1) a 4 feeds | Te 15 | s4@ Gee. 3) ° 16 | #-& (a>) a 1 1 aw 17 | tsinar - + a ° aim 5h | Tai eT = a 18 | cosar—cosbe | =o ° cosar— cose | GOSTHE LAPLACE TRANSFORMATION, [sec 819 we find that a 1 1 a sis+aPs sta (s+ay" Referring to Table 1, we can now write the result 2 ja ty em — ate. s(s + a) s Example 3 Find i where a? # b?, Since s (s? + a7)(s? + b?) (8 +a) — (5? +b?) BF (s? + a?)(s? + B) 1 s s “Rata 24 ey when a? # 6’, it follows that i = pra gileos at ~ cos bo. Example 4 Find F(t) if Ss +3 (s — 1s? + 2s + 5)" In view of the quadratic factor, we write S543 A, Be (s — 1)(s? + 2s + 5) +2545) f(s) = After clearing fractions and identifying coefficients of like powers of s, we find that A = 1, B= —1, and C = 2; thus 1 s-2 oy) "arr 1 stl 3 =1 G+ +4 Geet Referring to Table 1, or to Theorem 3, we see that F(t) = & — eos 21 — }sin 21,20 SEC. 9} OPERATIONAL MATHEMATICS 9 THE SOLUTION OF SIMPLE DIFFERENTIAL EQUATIONS The application of the Laplace transformation to the solution of linear ordinary differential equations with constant coefficients, or systems of such equations, can now be made clear by means of examples. Such problems can of course be solved also by methods studied in a first course in differential equations. Later on, when we have developed further properties of the trans- formation, we shall solve problems of this sort with greater efficiency. We shall also be able to solve more difficult problems, especially in partial differ- ential equations. Example 1 Find the general solution of the differential equation ()) Y"() + KY) = 0. Let the value of the unknown function at ¢ = 0 be denoted by the constant A and the value of its first derivative at t = 0 by the constant B; that is, YO)=4, YO)=B. In view of the differential equation, we can write L{Y"(O} + PL{Y(O} = 0, assuming that Y and Y" have transforms. If the unknown function ¥ satisfies the conditions in Theorem 2, then L{Y"()} = s*y(s) — As — B, where y(s) = L{¥()}. Hence y(s) must satisfy the equation sy(s) — As — B + k*y(s) = 0, which is a simple algebraic equation. Its solution is clearly s B_k Sa ek ee Now ¥(t) = L~*{y(s)}, and the inverse transforms of the functions on the right in the last equation are known. Hence ys) = A 2) Y(t) = Acos kt + Fain kt, = Acoskt + B'sin kt, where A and B’ are arbitrary constants since the i ¥'(0) are not prescribed. To verify our formal result given by formula (2), we need only to find ¥"(i) from that formula and substitute into Eq. (1) to see that the itial values ¥(0) andTHE LAPLACE TRANSFORMATION [sec.9 21 differential equation is satisfied regardless of the values of A and B. Thus it is not necessary to justify the use of Theorem 2. However, our function A cos kt + B’ sin kt does satisfy all conditions in that theorem, and the order of the steps taken above can be reversed to show in another way that our function satisfies the differential equation. These remarks on verifying the solution apply also to the other examples and problems that follow in this section. Example 2. Find the solution of the differential equation @) Y'() — YO — 6Y() =2 satisfying the initial conditions (4) YO=1, Y@=0. Applying the transformation to both members of the differential equation, and letting y(s) denote the transform of Y(0), we obtain formally the algebraic equation s*y(s) — 5 — sy(s) + 1 — 6y() = =, where we have used the initial conditions in writing the transforms of ¥"(t) and Y(t). Hence ye penne (s? — s ~ 6)y(s) ra s-s+2 A, BC ss— 3642) 5 s—3 542 Evaluating the coefficients A, B, and C as in the preceding section, we or ys) = find that a W= 354155373542 Hence ¥(t) = —4 + fe® + $e°*. Itis easy to verify that this function Y satisfies the differential equation (3) and both conditions (4). Example 3. Find the functions ¥(t) and Z(t) that satisfy the following system of differential equations: ¥"()- 2") + ZO - YO =e 2¥"() — ZW — 2¥') + ZN = —4 YQ) = YO) = ZO) = Z(0) = 0. 2,22 Sec. 9) OPERATIONAL MATHEMATICS Let y(s)and 2(s) denote the transforms of ¥(1) and Z(t), respectively. Then in view of the differential equations and the initial conditions, those transforms formally satisfy the following simultaneous algebraic equations: s?y(s) — s?2(s) + s2(s) — y(s) = z 2s7y(s) — s?2{s) — 2sy(s) + 2(s) = These equations can be written (s + Dy(s) — sas) = — 1 ass) — (8 + Des) = —SE ay Eliminating 2(s), we find that 225-1 ‘s(s — 1)? With the aid of partial fractions, we then find that (s? — 2s — Dyls) = WS) ao = “s (s— 1)?" Therefore ¥() =1-e +te. Likewise we find that y=, = S(s— 1? 2?" (s— 1)” and therefore Zt) = -t + te. Example 4 Solve the problem ¥"(@) — 2¥"( + SY = 0, Y0)=0, YO=1, (= Let C denote the unknown initial value ¥"(0). Then s*y(s) — s — C = 2s*y(s) + 2 + Ssy(s) = 0,THE LAPLACE TRANSFORMATION [sec. 9 23, so that (262248 WO 2545) oy ae P+ 4 10 (s—1P +4 Cc c-2 Thus Y= + [3 sin 2 - cos 2t 3 10 Since ¥(n/8) = 1, it follows that - 8 12 Fo? 43-2044, 5 10/2 or that C = 7. Hence the solution is ¥(0) = 1 + e(sin 2t — cos 21). PROBLEMS 1, Use Theorem 3 to (a) obtain entry 9 from entry 5 in Table 1 and (b) show that LM(s = a) #} = ent), given L~ "5-4. 2. Use partial fractions to find the inverse transforms of 2 2 wera © Ans, (a) 1 — e~*; (b) 1 — cos at; (c) 1 — (1 + at + $a?) 3. Use partial fractions to obtain the inverse transforms shown in (a) entry 16 of Table 1; (b) entry 17 of Table 1. (a) (b) sea’ s+ ay Solve the following problems and verify that your solutions satisfy the differential equations and any accompanying boundary conditions. 4. YQ) — Y(t) =0(k #0). Ans. Y(t) = Cie" + C,e™. 5. ¥%(t) — 2kY(d) + KP Y(Q = 0. Ans. Y(t) = &(C, + Cyt). 6 "(H+ RYH =a Ans. ¥(t) = C, sin kt + Cz cos kt + a/k? 7. Y"0) + 2¥(8) + 2¥() = 0, YO) = 0, YO) = 1. ‘Ans. Y = e™'sint. 8. Y(t) + 4¥(0) = sins, YO) = YO) ). Ans, Y(t) = $sint — sin 2¢. 9 YO+YW =e +2, YO) =4, 10 Ans. 3¥(t) = 0 + 67 +6. 10. Y(t) + ¥() = 10e”, ¥(0) = (0) = 11. Yt) = YO, YO) = YO) = YO =0, Y") 12. X() + YH) + XW) + YO = 1, YO — 2X — Ye) Ans. X(t) =24 SEC. 10) OPERATIONAL MATHEMATICS 13. ¥"() +2Z'(e) + YO) =0, YO —Z( — 2¥(0) + 2Z() = 1 — 24, YO) = YO) = Z(0) = 0. Ans, Y(t) = Al — e' = te", Zi) = YO) —t. 14, YH + YO =0. Ans, Y(t) = t + eos t. ~ sin x. 15. y"(x) + yO) = 1, + Gn) = 0. Ans. yx) 16. y"(x) + 2y() + yx) = 0, y(0) = 0, y(t) = 1. 10 GENERATION OF THE TRANSFORMATION We can show ina manipulative way why a linear integral transformation of functions F(t) defined on the half line t = 0, w TiFO} = [" KesFlode = f0, must be essentially the Laplace transformation if it is to have the basic operational property of that transformation, namely, that it replaces the simple differential form F'(t) by an algebraic form in f(s), s and the initial value F(0) of the function F. Let primes denote differentiation with respect to t. Then by a formal integration by parts we can write Q TF} j 7 K(t)F(O dt Keoro]. oo f- K'(ts)F(t) dt jo Jo = =KOS\FO) = f° KUIFO dt : provided that K(t,s)F() > 0 as t+ 00. The kernel K must involve some parameter s if the transform f is to correspond to a unique function F of some large class; otherwise, f is merely the numerical value of the integral (1) of the product K(:)F(#), a number that is the same for many functions F. In order that the final integral in formula (2) will represent f(s) except for a factor A(s), where 4 is some function of s, we require that Q) K(ts) = —A(s)K(ts). Thus K(t,s) = ce~™. But it is convenient to write c = 1 and choose A(s) as the parameter, so we write A(s) = s here. Then @ K(ts) T{F(O} i e “FO dt = L{F(O},THE LAPLACE TRANSFORMATION [sec 10 28 and formula (2) is the basic operational property of the Laplace transforma- tion: 6) T{F(O} = ~ FO) + sT{FO)}- A corresponding procedure will be used later on for the generation of integral transformations that will reduce other differential forms in functions defined on some prescribed interval, in terms of the transform of the function itself and prescribed boundary values of the function or its derivatives at the ends of the interval. Readers acquainted with the concept of a function F(t) (a < t < b) as a generalized vector! can see that a linear integral transformation ° 6 T(F(O} = K(ts)F(0 dt = f(s) is a generalization of linear vector transformations. Our transformation (1) is a special case of (6) in which a = 0 and b is infinite. A generalized vector, or function, F has an infinity of components consisting of the values F(t) for each 1; that is, the components are all the ordinates of points on the graph of F(t) over the interval a b, we see that foe = fem “z) dt. ° The following property is therefore established. Theorem 1 If f(s) = L{F(t)}, then for any positive constant b, (2) eo f(s) = L{F(0}, where F,(t) is the function defined by Eq. (1). The function F,(¢) is illustrated in Fig. 4. Its graph is obtained by translating the graph of F(t) to the right through a distance of b units and making F,(¢) identically zero between t = Q and t = b. We can refer to Fy(t) as the translated function. Our unit step function S,(¢) is the translation of the function So(t) = 1 (t > 0). It serves as a familiar illustration of the above theorem, since its transform is s~'e~"*. This step function can be used to describe the transla- tion of any function F(t) by writing FO) = SutF(t — 6) (c > 0), provided that F(t — 6) is defined where ¢ > 0; that is, provided that F(t) has numerical values for those values of ¢ in the range t > —b. For example, the function F(t) = sin kt is defined for all ¢, so in this case we can write Fd) = Ss sin k(t — b) = 2 (t > 0, b = 0). ‘t) = Sy(t) sin = 77k , b= 0). — 0 @ 0 60) @ Pe Fo) Fig. 4FURTHER PROPERTIES OF THE TRANSFORMATION [sec 12 29 On some occasions it is convenient to define F(t) as zero for all negative values of t. When that is done, the graph of F,(t) is simply a translation of the graph of F(t), and F(t) = F(t — b) for all t. Now consider the linear substitution cs for sin f(s), where c is a positive constant. If the transform f(s) of a function F(t) exists whenever s > a, then Ses) = j eo F(x) de whenever es > a. ° The substitution t = ct enables us to write that formula as if? t ==] e*F(-} de fies) ad, 7 () in terms of the transform of F(t/c), to establish this theorem: Theorem 2 If L{F()} = f(s) whenever s > a, and if c is a positive constant, then > amet fm We can write a = 1/c to obtain an alternate form of (3): (4) L{F(at)} = (a > 0,5 > ao). Given, for example, that L{cos t} = s/(s? + 1) when s > 0, it follows that sla 1 Meosat) = 2 agra 7 Fa Fre s>o The effect of a general linear substitution for s can be seen from formula (3) and Theorem 3, Chap. 1, since (5) flas — b) = sels = Hl = fterre(4)] (a> 0). 12 STEP FUNCTIONS When £ 2 0, the bracket symbol [1] is used in mathematics to denote the greatest integer, 0, 1, 2,..., that does not exceed the number t. Thus [7] = 3 = [3]. The function [t] is therefore a step function of the type that is some- times called a staircase function with unit rise and run:30 SEC. 12) OPERATIONAL MATHEMATICS (J=0 @s¢t<1), =1 (lSt<2), =2 (2St<3).... The staircase function with an arbitrary positive run h, and with a unit rise beginning at the origin t = 0 (Fig. 5), is then represented by the symbol [1 +h] or 1 + [t/h]. The function e[1 + 1/h] has the rise c and the run h. To obtain the transform of the function shown in Fig. 5, ) Y= [: + ‘ (h> 0), we may describe the function by means of a difference equation of the first order together with an initial condition, as follows. ¥() = ¥(t— hy +1 (20), =0 (<0). The function ¥(t — h) is then the same as the translated function ¥,(t) and, in view of Theorem 1, the transform (s) of the function Y(t) satisfies the equation 6) = My) +4 (>0), Therefore the transform of the staircase function (1) is 2 (3) = uff +i] -! Another useful step function is the unit finite impulse function (s > 0). 1 Q) Mh, t ~ to) = 7, when fo <£< to +h, = 0 when ¢ < to and when t > to +h, Fig. 5FURTHER PROPERTIES OF THE TRANSFORMATION (sec. 12 31 Fig. 6 illustrated in Fig. 6. Let us introduce a variation of our unit step function S\(0), one that is defined to be zero when t is negative; namely, @ So(t — to) = 0 when f < fo, =1 when t > fo Then our function J can be written So(t — to) — So(t — to — h) o Tht =) = ; It follows at once from the transform of So(t — to) that 6 LU(ht — tg)} = e7#04 (to 2 0,h>0,s > 0) It is clear from the definition (3) that a lim I(h,t — t) = 0 when t # to, a Also, the area under the graph of the function I is unity for every positive h; hence (8) tim [10h = t9) de = 1. n+0d oe The order of the positions of the limit and the integral here is important, for in view of Eq. (7) f lim I(h, t — to) dt = 0. = h0 Ibis interesting to note the behavior of the transform of the finite unit impulse function as h tends to zero. By evaluating the limit of the right-hand member of Eq. (6), we find that 0) im L{U(h,t — t0)} = e*° (Fo 2 Oh > 0, > 0) But note that Lilim 1(h,¢ — t0)} = L{0} = 032 SEC. 12) OPERATIONAL MATHEMATICS Equation (9) presents the exponential function e~*, or exp (— sto), not as the transform of a function, but as a limit of transforms of functions of the set I(hyt — tg) consisting of one function for each positive value of the parameter h. Thus exp(—sto) approximates the transform of the impulse function I(h, t — tg) for small positive values of h, when to = 0. Example A particle of mass m, initially at rest at the origin X = 0, is subjected to a force in the form of a unit finite impulse 1(h,t) beginning at the instant fy = 0, in the direction of the X axis. Its displacements X(hnt), which satisfy the conditions (10) mX"(ht) = ht), X(h0) = X'(h0) = 0 (h> 0), where X’ = dX/dt, could be found here by integration But if we use transforms to solve for X(ht), we write em AL ? x(h,s) = L{I(h,)} = (ut) ms*x(h,s) = L{I(h,t)} is? assuming that X and X’ are continuous functions of ¢ of exponential order. Thus 2 hs] os and therefore Lia 2; (12) mX(h,t) = alt = (t — hPSo(t — hy); that is 1a = sis (13) X(h,t) Son! when0 0 to give the limiting displacements X(t): (14) X(t) = lim X(h,t) = a (t > 0). ao m The displacements X(t) correspond to the idealized unit impulse given to mat the instant = 0. Asa verification of the solution (14) of that idealized case, we note that X"(t) = 0 whenever t > 0, and X(+0) = 0.FURTHER PROPERTIES OF THE TRANSFORMATION [sec 13 33 Furthermore, the particle has the momentum mX’(t) = 1 whenever t > 0; thus if it is initially at rest, then its momentum jumps suddenly from 0 to 1 at the instant ¢ = 0. The instantaneous impulse has the same effect as an initial velocity 1/m. 13 THE IMPULSE SYMBOL <(¢ ~ t) The above example is a simple illustration of problems in differential equa- tions involving instantaneous impulses. In order to reduce the number of steps in formal solutions of such problems, we shall occasionally use the unit impulse symbol 6(¢ — fo) to a limited extent indicated below. Let a function F(2), defined for all real t, be continuous on some interval to StS to + k. Then according to the basic law of the mean for integrals to each h(0 < h < k), there corresponds a number 6(0 < @ < 1) such that © to+h f I(h,t — to) F(t) dt = if F(t) dt = F(to + Oh). Consequently the impulse function I has the property a) lim | I(h,t — to)F(0) dt = F(to). nod. sometimes called the sifting property of selecting that value F(t) of F(t). We abbreviate that property by writing (2) f A(t — to)F(t) dt = F(to), where the entire left-hand member is a symbol that denotes the operation represented by the left-hand member of Eq, (1). That symbol is selected so as to suggest an integral of a function, because if 4(¢ — t) is replaced by I(h, t — to), the symbol does represent an integral whose value is approxi- mately F(t) when h is a small positive number. If ty = 0, then I(h, t — fo) = 0 when t <0 and @) him I(h,t — to)F(t) dt = F(t) (tp = 0,h > 0) according to Eq. (1); that is, in symbolic form ) Jf de sR at = Fe) (ta 20) When F(t) = 1, formula (2) becomes (3) f- &{t — to) dt = 1,34SEC. 13) OPERATIONAL MATHEMATICS the symbolic form of Eq. (8), Sec. 12. When F(t) = e~* and to 0, the symbolic integration formula (4) can be written (6) L{O(t — t0)} = exp (~ Sto), which is the symbolic form of Eq. (9), Sec. 12. The letter 6 used in our symbolic integrals has further significance in differential equations. We use the example treated in Sec. 12 as an illustration. Let us write a mX"(t) = 4); X(0) = X(0) = 0, to signify that X is the limit as h + 0 of the solution X(h,.) of problem (10) there, the problem with 6(¢) replaced by 1(h,t). In the transformed problem (11), Sec. 12, we may let h tend to zero and assume that x(h,s) ~> x(s) where x(s) = L{X(}; the result is (8) ms?x(s) = 1. But that result is obtained at once by formally applying the operator L to terms in problem (7) if L{6(t)}} = 1, in accordance with formula (6), while the symbol L{X”(#)} means the limit as h -> 0 of the transform of X(h,t) so that we may be justified in replacing it by s?x(s). Thus mx(s) = 1/s?, and our result mX(t) = t follows. That symbolic method gives limits of solutions, as h > 0, of other problems in differential equations with I(h,t — to) as a forcing function. We shall use the method occasionally in the sequel. Results obtained by such manipulations clearly need to be verified if they are to be relied upon. More often we sacrifice the brevity gained by using the 5 symbol and use instead the function [(h, t — to), then find the limit of the solution as h > 0. The symbol 6(¢ — t0), often called the Dirac delta function, is not a function. In particular, it is not the limit of I(h, t — to) as h > 0; that limit is a function with value zero everywhere except at the one point f = to, a function whose integral is zero (Sec. 12). The symbol 6(t — to) may, however, be replaced by the function I(h, t — to), when h is small, to give results that approximate the effect of 6. Generalizations of functions, called generalized functions or distribu- tions, have been developed since 1950 which include the symbol 6(¢ — to)and its generalized derivatives. There are now a variety of theories that give sound developments of distributions. But even the introductory presenta- tions of the theory of distributions seem too lengthy to present here.’ + See, for instance, the article entitled From Delta Functions to Distributions, by A. Erdelyi, in “Modern Mathematics for the Engineer,” Second Series, 1961, or A. E. Danese, “Advanced Calculus,” vol. 2, pt. 6, 1965, or references listed in those booksFURTHER PROPERTIES OF THE TRANSFORMATION [sec 13 96 PROBLEMS 1. With the aid of our theorems and known transforms find the inverse transforms F(t) tabulated below and draw a graph of F(t). So) Fo (o>) 0@<1) (0) Jas" te“2* 2i—2 S40) em js mn =S4osint ite" Os sint(@ << m);0(¢>m) Soft ~ 4)sin ne 2. From the Maclaurin series that represents (1 — x)~' when |x| <1, or from the sum of an infinite geometric series, show that 1 2h Apply the inverse transformation term by term to this infinite series, formally, to obtain the result shown in Eq, (2), Sec. 12; namely, 1 cs r i> 0. 31 3. Show that Eq. 2), Sec. 12, ean be written in the form Afra t]b = 21 + con’ +E Tf a(t + eoth Ss]. 4. (a) Use the formal method indicated in Prob. 2 to find F(t) when fi) = (>0 1+ Draw the graph of F to see that F can be written in terms of the bracket symbol [¢] in the form FQ) =1+(-1)"36 SEC. 13) OPERATIONAL MATHEMATICS (b) Show that the function F in part (a) is des equation of the first order, when t > 0: id by the following difference FQ) + Ft 1)=2 where Ft) =0 when t < 0. Transform that difference equation to verify the formula given in part (a) for f(s). 5. Apply Theorem 2 to the transformation found in Problem 4 to show that uft +0} 12 (h>0,5> 0). 6. Find the function Y(t) that satisfies the following difference equation of the first order and the accompanying initial condition. Y(t) ~ c¥(t ~ h) = Fe) yy =0 when ¢ <0, where c and h are constants and h > 0, and where F(t) = 0 when ¢ < 0. The expansion of (1 — ce)" in powers of ce~™ is helpful here (compare Prob. 2). Show that the solution can be written ¥() = ¥ cF(e - nh); =o but for each fixed value of t this series is a finite series because F(t — nh) = 0 when t— nh <0. Thus an alternate form of the solution is Y(0) = Flt) + cF(t — h) + c?F(¢ — 2h) + +++ + c™F(e — mh), where m = 0, 1,2,...,when mh <¢ <(m + I)h. Verify the solution. 7. Find the function ¥(¢) that satisfies the following difference equation of the second order and the accompanying initial condition. Y(t) — (@ + B)Y(t — h) + abY(e — 2h) = Flo) Yo) when ¢ <0, where F(t) = 0 when ¢ < 0. The constants @ and b are such that a # b, and the constant his positive. Note that the solution of the transformed problem can be simplified, with the aid of partial fractions in the variable e“™, to St) { a b\I = ae" ys) Ans. Y() = br FE — ni a-b 25 8. Solve Prob. 7 when b = a. Ans. Y(t) = x (n + l)a"F(t — nh). 0 9. Solve the difference-differential equation ¥(0 — a¥(e— 1) = Feo,FURTHER PROPERTIES OF THE TRANSFORMATION (sec. 1337 where F(t) = b when t > O and F(t) = when ¢ 0. when ¢ < 0, under the condition that ¥(«) = 0 Ans. ¥(0) = of! +H IP wenitt - on} wheren S¢Sn+landn=0,1,2, 10. Initially a particle of mass m moving along the X axis is at the origin X = 0 with a velocity v. The only external force that acts on that particle is an instantaneous impulse po at time fo, in the X direction. Thus the displacement X(t) satisfies the follow- ing symbolic equation and initial conditions: mX"(O) = Po Alt ~ to, XO) = 0, X'@) = vo. Find the formal solution (writing L{X"} = s?x — vo) X(0) = vot + P(e ~ te)Solt — to) (620, t)>0) and show X(t) graphically. To verify the solution, show that X"() = 0 when ¢ # to, that X(0) = 0 and X'(0) = vo, and that the momentum mX"(2) undergoes a jump pp at the instant fo 11. Replace the instantaneous impulse po 4(t — f.) in Prob. 10 by the finite impulse Pol(h,t — to) and then find the displacements X(h,t). Show that the limit of X(h,f) as h > Ois the function X(t) found before. 12. Problem 10 can be stated in terms of functions as follows: mX"() = Owhent # to, X)=0, XQ) =n» mX (to + 0) — mX'(to — 0) = Po (to > 0), where X and X’ are continuous except for the specified jump in X’, Use our formula (Prob. 9, Sec. 5) for the transform of the second derivative in this case to obtain the formula found before for X(t) 13: Show formally that the solution of the symbolic problem X(t) + KPX( = podlt — to), X(0) = XO) (t 29) where k and po are constants, is (if we write L{X”} = s?x here) X() = Psu = to) sin k(t — t9). 14, Show that the formal solution of the symbolic problem (if we write L{X"} = sx here) X(t) = det), XQ =0 (29 is X(t) = So(t — to). In this sense, then, 4(¢ — fg) corresponds to the derivative of the unit step function So(t — fo) as we could anticipate from the symbolic formula L{dle ~ to)} He = SL{Syl 0}.38 SEC. 13] OPERATIONAL MATHEMATICS 15. Let AI denote the change in I(h,t — tp) when At = —h, Draw the graph of the function of t represented by the difference quotient Mht = to) = Mt ~ h to) If F and F’ are continuous on an interval that includes the points f, and tg + 2h, show that oO Fe +H = FO) 9 Al Godt = 7 a i pet = =n Fc + 6h) de where 0 < @ < 1, and hence that lim ar dt = —F (ty). m0) A symbolic form of that formula is f © BUC = tohF(O dt = ~ Fed): which is a symbolic integration by parts: f O(t — to)F(Q) dt = — f Olt — to)F'() dt = —F'(to). In particular, if 9 2 0, note that 16, Let Fbecontinuous(¢ = O)exceptfora jump jo at fo (to > 0). Then F(t) — joSo(t — to) is continuous whenever ¢ 2 0 if F(é.) is properly defined. If F is exponential order and F’is sectionally continuous, then {sino — JoSolt — van} = [ro ~— Jo | — F(0). ‘Show that this formula agrees with formula (4), Sec. 4, and note that its symbolic form may be written (Prob. 14) LAF (0) ~ jo H(t ~ to)} = f(s) ~ F(0) ~ joe **. 17. Since I(h,t) = 1/h when 0 <1 h, describe the function I(h,—1) and show that lim f F()Mh,—0) dt = FQ)FURTHER PROPERTIES OF THE TRANSFORMATION, [Sec 14 39 when F is defined for all t and continuous over some interval |t| 0) the particular Laplace integral ay wos) = f xe"* de ° exists; it has the values p(0) = 0 Q) (x)= 1 when x > 0, It can be written as a definite integral plus a remainder, r 0 P(x) -f xe* dt +f xe" dt = 1 — e7*T + R(x,T), r ° where R(x,T) = [7 xe™™ de; then R(T) = 0 (3) - R(T) =e"? when x > 0. Uniform convergence of the integral for all x in a prescribed interval means not only that the integral exists for each x in the interval, but also that the remainder can be made arbitrarily small in absolute value by taking T sufficiently large, uniformly for all x in the interval. Thus when x 2 1 and € is any small positive number, we see that [ROO] = when T > log(I/e), a value that is independent of x, so that the integral converges uniformly with respect to x when x > 1. In the same way we can see that it converges uniformly when x > x, whenever x, > 0. But the convergence is not uniform in the range x = 0 or x > 0, because the remainder (3) when x > 0 is small only if xT is large. Since x can be arbitrarily small, the value of T that makes R(x,T) small must depend on x. A consequence of this lack of uniform convergence in the range x = (is the discontinuity of the function p(x) at x = 0 that is exhibited by Eq. (2) The Weierstrass test for uniform convergence of the integral *TsebK0) and such that the integral f° M(0) dt exists, then the integral (4) is uniformly convergent with respect to x(0 < x Sc). This test is established by first noting that the integral (4) exists according to the comparison test (Prob. 14, Sec. 5), then that the absolute value of the remainder for the integral, for all x, in the interval, does not exceed the remainder {° M(t) dt in the integral of M, The latter remainder is independent of x, and it tends to zero as T+ co. The Weierstrass test states useful sufficient, but not necessary, con- ditions for uniform convergence. Under those conditions the integral (4) is also absolutely convergent. If the integral (4) is uniformly convergent with respect to x and if its integrand f(x,t) is continuous by subregions over the rectangle 0 S x $ ¢, 0ST for each positive T, then the integral represents a sectionally continuous function q(x); moreover (5) f q(x) dx = f I S(x,0) dt dx = f fi foo dxdt; thats, the order of integration of the definite and improper integrals here can be interchanged. In particular, let f(x,t) be continuous in each rectangle 0 < x Sc, 0.S¢T, except possibly for finite jumps across a set of lines ¢ = 1, 1,2,...,n). Then q(x) is continuous if the integral (4) is uniformly convergent. Suppose that f/0x as well as f satisfies those continuity requirements and that the integral {® (@f/@x) dt converges uniformly and the integral (4) exists. Then the derivative of the latter integral exists and the order of differentiation with respect to x and integration with respect to t can be interchanged : df” a) 6 i, flst)dt = I Eflostyat 9). It will be convenient to define G(t) to be zero when t < 0; but our final result depends only on the values of F(t) and G(t) when t > 0. According to Theorem 1, for each fixed x( = 0), e-*g(s) = L{G(t — 1)} = i e-"G(t — 1) dt,44sec. 16) OPERATIONAL MATHEMATICS where s > a. Hence = ° Fl a dt= Fy * "Gt — t)dt dr; Foss) = fF Foie "ato ae J (of e*Gte— oat de that is, 7 pe () F(s)g(s) = Jim f, fi F(tje~“G(t — 1) dt dz. Since f(s) and g(s) exist when s > a, the limit here exists. The integrand of the inner integral in Eq. (I) is continuous by subregions (Sec. 14), over the rectangular region 0 < t S$ T,0 St S R, for each pair of positive constants T and R. For if the jumps of F(t) and G(t) occur at t= 1(i= 1,2,...,m) and t= t{j=1,2,...41), then the jumps of the integrand occur at the lines t = t, and ¢ = t + ¢;. In view of Eq. (5), Sec. 15, the order of integration in Eq, (1) can be interchanged, provided that the improper integral there is uniformly convergent with respect to its parameter 10 StS 7) The uniform convergence is seen by noting that, owing to the exponen- tial order of F and G, a constant N exists such that Q |Ficje""G(t — 2] < Nee™*e"9 = M(D, where M(t) = N exp[—(s — af]. The function M(0) satisfies the conditions of the Weierstrass test (Sec. 15); that is, M(t) is independent of t and integrable from zero to infinity. Equation (1) can now be written in the form ee 8) foo) = jim Fe f Feeyate ~ ayaa a J iim (A(T) + 17), Toor where nin) = f ef F()G(t — 1) de dt, jo do © r 1,(T) -f enf F(t)G(t — 1) ded. T ° In view of condition (2) e r mT I(T) < vf ewe drdt = NT g--arr, 0 s—a therefore lim 1,(T) = ToeFURTHER PROPERTIES OF THE TRANSFORMATION [sec 16 45 The region of integration for the integral 1,(T) is the square 0 < tS T, 0StST. But Gt — 1) = Owhent > ¢. Therefore roo I(T) = f ef F(t)G(t — t)dtdt and Eq, (3) reduces to (4) f(s)g(s) = lim 1,(T) = frenf F()G(t — 2) de dt. Tro ° ° The convolution F * G of the functions F(t) and G(t) is defined as the function (5) F(t)* G(t) =} F(t)G(t — 1) dt, so that Eq. (4) can be written (6) S(s)g(s) = L{F(O) * GO}. We summarize our result as follows. Theorem 3 —_Iff(s) and g(s) are the transforms of two functions F(t) and G(t) that are sectionally continuous on each interval 0 St S T and of the order of & as t tends to infinity, then the transform of the convolution F(t) * G(t) exists when s > a; it is f(s)g(s). Thus the inverse transform of the product f(s)g(s) is given by the formula 7) L7"S()g(s)} = Fl) * G0). The functions F(t) = t and G(t) = e* for example, satisfy the conditions of Theorem 3. Consequently wif Ec eere flere a ; = ef re de = A(e" — at = 1), : a Partial fractions can also be used to obtain that result. When G(t) = F(t), we have the formula (8) Lf)? = L{F * F}. Asan example that will be useful in finding inverse transforms with the aid of46 SEC. 17) OPERATIONAL MATHEMATICS partial fractions, we note that - 1 0 eer} 1 g@ sin ke * sin kt = zl sin kr sin k(t — t)de ° apsesin kt — kt cos kt). The conditions stated in Theorem 3 are narrower than necessary for the validity of formula (7). If F(e) = ¢~# for example, F is not sectionally con- tinuous onan interval <1 < T, but the Laplace integral of F(t) is absolutely convergent and formula (7) is still valid if G(t) satisfies the conditions stated in the theorem. 1 1 Thi a “ 7m - a Ym” lea If we make the substitution r = \/t here, we find that = 1 a r ye = at where the error function erf (x), also called the probability integral, is a tabulated function defined by the equation oP dr =e erf(/0), 2% 2 2 ena Yale It was shown in Prob. 10 at the end of Sec. 5 that erf(co) = Since the substitution of s + 1 for s in a transform corresponds to multiplication of the object function by e~', it follows from Eq. (10) that | = erf(\/i). (uy) erf (x) = (12) 17 PROPERTIES OF CONVOLUTION By substituting the new variable of integration 4 = ¢ — + in the convolution integral (5), Sec. 16, we find that the convolution operation is commutative; that is, a F)* Go) = Gi) * Fl) = f Fl — AGA) dd. 7FURTHER PROPERTIES OF THE TRANSFORMATION (sec. 1747 The operation is clearly distributive with respect to addition 2) F(t) * (G(t) + H(d)] = F()* Gd) + F(O)* H(0. Also, F * (kG) = k(F * G) if k is a constant. Properties (1) and (2) are valid whenever the functions are sectionally continuous over an interval (0, T) that contains the point t. For such func- tions the operation is also associative. @) F(t) * [G(e) * H(] = [F() * G(0)] * H(0. To prove this we first write, in view of Eq. (1), F*(G*H)= frof ce —1— AHA) dade lo Jo and observe that the iterated integral here represents an integration over the triangular region shown in Fig. 8. When the order of integration is reversed, the integral becomes ‘ 1a f maf F()G(t — A — 2)dr da, jo do which represents H * (F * G) or (F *G)*H. Note that for each fixed value of ¢ the integrand of the iterated integral is a function of t and A that is continuous by subregions over the region 0 $ 2 < ¢ — 1,0 0). 0 But if ¢ is any positive number and if M, represents the maximum value of the function re~“ when ¢ = 0, then Myte = Myte“#e = Meer, where M = M,M,. Consequently F * G is of exponential order (5) IF(t) * Go| < Mes" (€ > 0,1 > 0). Now, if three functions F(s), G(1), and H(t) satisfy the conditions of Theorem 3, then the function G*H satisfies those conditions and its transform is g(s)h(s). Therefore S(sa(s)h(s) = L{F(e) * (G(0) * H(9)} ; but since the convolution operation is associative, this equation can be written in the form (6) S(s)g(s)h(s) = L{F(o) * G(e) * HO}. A similar formula can be written for the product of n transforms. When G(t) = 1 in Theorem 3, we have a result noted earlier (Sec. 5): Theorem 4 Division of the transform of a function by s corresponds to integration of the function between the limits 0 and ¢: @ cs {ty wh = f F(a)dz, aft He (8) ie {24h = f fi F(a) dadz, etc., for division by s*, provided F(t) is sectionally continuous and of the order of & (a > 0), where s > a.FURTHER PROPERTIES OF THE TRANSFORMATION. [sec 17 49 As examples, we note that - k 1 1 = lage L ta r a I sin ke dr = 7(1 ~ cos ko), k of var he — si lata} = f f sin kA da de = jo(kt — sin kt) PROBLEMS 1. Use transforms to show that (@ I*l*t=4e; (6) tetera, © J reose = 1dr = 2t = sin. 2. Show that an alternate form of the transformation (8) Sec. 17,8 me oh = ce ro=e[ Foae s F(t) de. 3. Show that 1 - and in view of formula (10), Sec. 16, that 1 1 rn Lo) =e + at eerfy/t 7; = 4 Vat Wa 4. Generalize formula (10), Sec. 16, by replacing s by s/a? to obtain transform 40, Appendix A, Table A.2. 5. Generalize formula (12), Sec. 16, to obtain transform 44, Appendix A, Table A.2. 6. Find L~'{(s + 1)~'s~#} and generalize to obtain transform 41, Appendix A, Table A.2. 7. Obtain the inverse transform i i = etl - af 8. With the aid of transformation (9), Sec. 16, show that wf 8 YL spine — teose) — (+p i: 9. Let ¢=h and ¢ =k be the points of discontinuity of a sectionally continuous function G(t) on an interval 0 < ¢ < T, and let j, and j, denote the jumps in the value of G(®) at those points. With the aid of a graph of G(®), note that the function Gt) = Gl) — ISD — ASAD. @stsT) L in t50 SEC, 18) OPERATIONAL MATHEMATICS is continuous if its values at t = h and ¢ =k are properly defined. Write F *G in terms of F * G, and of convolutions like the one in Eq. (4), Sec. 17, to show that if F() is sectionally continuous, then F * G is continuous. Also note how the result can be generalized to permit G(t) to have n jumps in the interval. 10. Show that the integral eo x wx) = f mae has the values u(x) = —x/2 when x < 0, u(x) = 2/2 when x > 0 and u(0) = 0. Examine the remainder R(x,T) for this improper integral and show that the integral is not uniformly convergent with respect to x over any interval that contains the point x = 0 in its interior or as an end point. 11. Prove property (5), Sec. 15, under the conditions stated there. This can be done by first showing that for each positive number T faa =f ["yarae + J Rownax and hence that, corresponding to each positive number «, a number T; exists such that when T> T, . a | faa [seca c, and J? @ dt is uniformly convergent with respect to x(b < x < 0) Suggestions: Show the region graphically. Introduce a function f to which formula (6), Sec. 15, applies by writing Slt) = Gxt) when b x, fe) =0 when either x be; () when b Ans. (a) Y(t) = ab(b? — be)~* sin (t./B? — be); (b) Y(@) = abe. 13. Solve the nonlinear integral equation 2¥() + f. Y@¥(t— de =e +2 Ans. Y(t) = 1 14. Write ¥"(0) = V(0. (a) If ¥(Q) = a and ¥'(0) = b show that Y(h=leVintb, — ¥()= ea V(t) + bet a; (®) then convert the ini "+ oY + BOYO = FI, YO) = YO) = in Y(t) into the following integral equation in V(t): value problem V+ f (oe + ple = HIM de = FO 19 DERIVATIVES OF TRANSFORMS When the Laplace integral (I) fis) = f eR) dt 7 is formally differentiated with respect to the parameter s by carrying out the differentiation inside the integral sign, the formula f= [eM -Fwd = {HF} :FURTHER PROPERTIES OF THE TRANSFORMATION [SEC 1955 is obtained. Another formal manipulation with the integral (1) indicates that f(s) > 0 as s + 00. We shall establish conditions under which those formulas are valid. First, we note that if F(t) is of the order of e* as t + 00, then the function "FO, where n = 0,1,2,..., is of exponential order. Let € be a positive number. Then constants N, and N, exist such that Ie"F(Q) < Ne = Ny tte “tet SN yN act" (¢ > 0), where N, represents the maximum value of the function t"e~*' when t > 0. Thus the function t*F(t) is of the order of e, where a = a + « Also let F(t) be sectionally continuous on each interval 0 <1 < T. Then ¢*F(t) has that property. The absolute value of the integrand of the Laplace integral of 1"F(t) satisfies the condition 2) Ie"F(je~*| < New &~ 9" (n = 0,1,2,...), where N denotes a constant. Consequently when s > a, @) if PFE" at | < Nf comma = 0 0 therefore L{t"F(t)} 0 as s+ co. Moreover, if s = a, where a; > a, then according to condition (2), lnF(Qen"| < New" = MD, where this exponential function M(t) is independent of s and integrable from zero to infinity. It follows from the Weierstrass test (Sec. 15) that the Laplace integral f PF(jen* dt (2 =0,1,2,.-..) 0 converges uniformly with respect to s when s = a, > %. Theorem 5 If F(t) is sectionally continuous and of the order of &', then each of the Laplace integrals L{F(t)}, L{tF(0)}, L{t?F()}, ..-, is uniformly convergent when s 0, where a, > @; moreover (4) and lim L{e"F()} = 0 (n= 1,2,...). The function F(t)e~* and its partial derivative of each order, with respect to 5, satisfy our conditions for the validity of formula (6), Sec. 15. Hence differentiation with respect to s can be performed in Eq. (1) inside the integral sign, and the following theorem is established.56 sec. 19] OPERATIONAL MATHEMATICS Theorem 6 Differentiation of the transform of a function corresponds to the ‘multiplication of the function by —t (5) SOs) = L{(— FO} (n= 1,2,.-.)5 moreover, fs) +0 as s+ 00. Those properties hold true whenever F(t) is sectionally continuous and of the order of &, ifs > «in formula (5). Since a function is continuous wherever its derivative exists, it is true that f(s) and each of its derivatives is continuous when s > 2. To illustrate the last theorem, we can note that since ara = L{sin kt} (>, it follows that —2ks . wey 7 El-tsin Thus we have a formula that is useful in finding inverse transforms with the aid of partial fractions: 2ks o Lt sin kt} = app (5 > 0). Transformation (6) can be obtained by another method that is some- times useful, that of differentiating a Laplace integral with respect to a parameter k, independent of s. When s > 0, the Laplace integral ” —st s a L{cos kt} = J e" cosktdt = = converges, and the integral a =o, 6) fs cos ui} = Seler" 208 kt) =- f e”“tsin kt dt 0 converges uniformly with respect to k for all real k, according to the Weierstrass test. All integrands here are continuous functions of k and t. Hence formula (6), Sec. 15, applies to show that the integral (8) represents the derivative, with respect to k, of the integral (7); that is, ~Lftsinke} = oles 5 sae 2ks Ok\s? + k? (s? +k)?’ and this is the transformation (6).FURTHER PROPERTIES OF THE TRANSFORMATION (SEC. 20 87 20 SERIES OF TRANSFORMS A useful method of finding inverse transforms L~'{ f(s)} is that of repre- senting the function f by an infinite series of known transforms, then applying the operator L~' to the terms of the series. For certain types of series in powers of 1/s, conditions for the validity of the procedure will now be established. Theorem 7 Let f(s) denote the sum of an infinite series of positive integral powers of 1/s which is absolutely convergent when s > a, where x = 0: S 1 oO) SW) = Vas = Oe See. > a20) ey ets FS Then the power series in t obtained by applying the operator L~* to that series term by term converges to a function F(t) whose transform is f(s), when s > a: @) Fy = ¥ a5 = LF} (20, where 0! = 1, Also, F is continuous when t = 0, and of exponential order O(e*) whenever a > a. Since series (1) is absolutely convergent when s >a, whenever a > % 2 0, it follows that |a,|/s"*" < |a,|/a{*! + 0 asin > co. Therefore a constant M, independent of s, exists such that |a,|/s"*! 020). If we write 5 = ay here it follows that, when t > 0, ' ; ro @) ay | = lay! = Ma, 2 int | = Mey a! The series of terms (2,1)"/n! converges to ¢", so by the comparison test series (2) is absolutely convergent, and its sum is of exponential order, FOS ¥ lal s Mayen" (20,0 > 9) anni Also, F is continuous because it is the sum of a convergent power series in t. Let s be fixed (s > a) and let T be a positive number. The series of the terms a,e~“¢"/n! is uniformly convergent with respect to t over the interval 0 < ¢ < T according to the Weierstrass test for infinite series of functions, for in view of condition (3), Heys lle“ al (5>0,0<1 00; that is, to each positive number « we are to exhibit a corresponding number T, such that r © (9) -f oF) dt 0 T, It will follow that f(s) = L{F(0}; then Theorem 7 will be proved. The series in Eq. (5) is absolutely convergent because (U) = mJy and series (1) is absolutely convergent. Hence if N = 1, 2, "sila (8) | 709 5 [errioal "5 lad fre ~#1" dt + Ry) 0 zon! where, in view of condition (7), Ry(T) = x a f estas Y la ws Since series (1) is absolutely convergent, it now follows that to each « (« > 0) there corresponds a number N;, independent of T, such that R(T) < de whenever N > Ne Now consider the sum from n = 0 to N ~ 1 in condition (8). Let be some fixed integer greater than Ne. By successive integrations by partsFURTHER PROPERTIES OF THE TRANSFORMATION [sec. 20 59 we find that ttre pet rl rf) ence Fe +4] Since |a,| < Ms"*?, it follows that, if sT > 1 andn SN — 1, lal syn nar] STP, OTK wre "dt < Me [en +aq-pt +1] < memory +5 z 7 bok | < Me“"(sTYN. Thus the sum in condition (8) is less than MN?e~*7(sT)*, which vanishes as T+ 06 for that fixed N. Therefore, there is a number 7, such that MN?e~S"(sT)% < ¢/2 when T > T,, and condition (6) follows to complete the proof of the theorem. Theorem 7 can be generalized to the following theorem Theorem 8 Let g(s) be represented by an absolutely convergent series when 5 > a(x = 0) of this type: 0) 26) = Fay Oa20) So Then L{G(t)} = g(s) when s > a where the function G is continuous when t > 0, is O(@") if om, > a, and is represented by the convergent series of inverse transforms _ Bp eifaeerd 2 eka (10) Gt) = ye {sah S % Ta Em (t> 0). A proof can be based on Theorem 7 by writing 1 $ a, -r 0 The last series here represents the transform of the function - ay Fo = ¥ an. a function that is continuous (t = 0), and O(e*") if a, > a. Also we know that (Sec. 5) r& Litt (k > 0,5 > 0).60 SEC. 20) OPERATIONAL MATHEMATICS Since k < 1, the function ¢*~! is bounded when = T > 0. Then * F(t) ( a. In view of Eq. (11) we can write (12) fh F() = fe — 1) F(t) de 0 = San (Gein =a KG a lamde = 4G, = ¥ tee, ym where the term-by-term integration is valid because the power series for F(z) converges uniformly with respect to t over the interval 0<1t <1 (Prob. 15, Sec. 21). By using known transforms, we find easily that tag TUM ge Ta+1+h and therefore et « ete TH PO> 2 resem It now follows that L~"{g} is the function G represented by series (10). That convergent series is the product of t*!, or «~~, by a power series in t. Hence G is continuous whenever t > 0, and Theorem 8 is proved. Example To establish transformation 78, Appendix A, Table A.2, when k =I there, we write the absolutely convergent representation a aste t= FCN (s > 0) According to Theorem 8 and the form 6, Appendix A, Table A.2, of the elementary transformation L~'{s~"~+}, then, L{G} = g whereFURTHER PROPERTIES OF THE TRANSFORMATION sec. 21 61 The terms of that series reduce to (—1)'(2\/0)?"*4/(2n + 1)! and therefore enue s/s. V Git) = uf } Lsine/p Jr 21 DIFFERENTIAL EQUATIONS WITH VARIABLE COEFFICIENTS We have seen that Levi) = (—I-Ze ero} = (9, and therefore we can write the transform of the product of t* and any derivative of ¥(0) in terms of y(s); for instance, @ L{?¥'(0} = gi) — YO] = sy") + 2y'(), d LEY} = —Gls°v69) — s¥O) — YO) = —s?y(s) — 2sy(s) + YO) A linear differential equation in Y(t) whose coefficients are polynomials in transforms into a linear differential equation in y(s) whose coefficients are polynomials in s. In case the transformed equation is simpler than the original, the transformation may enable us to find the solution of the original equation If the coefficients are polynomials of the first degree, the transformed equation is a linear equation of the first order, whose solution can be written in terms of an integral. To find the solution of the original equation, however, the inverse transform of the solution of the new equation must be obtained. Example 1 Find the solution of the problem Y()+tY(-¥H=0, YO)=0, YO) The transformed equation is d — glsvl — v5) s?y(s) — vo, [2 _ or ve + F~J = >ee sec. 21) OPERATIONAL MATHEMATICS which is a linear equation of the first order. An integrating factor is 2 2) 2 exp f — s| ds | = exp (2logs — 452) = ste-#*, 5 so the equation can be written dio ae , [ste #¥*y(s)] = —se7#" a Integrating, we have where C is a constant of integration. But C must vanish if y(s) is a transform since y(s) must vanish as s tends to infinity. It follows that Y(t) = and this is readily verified as the solution. Example 2. Solve Bessel’s equation with index zero, 1¥"() + Y(0 + t¥() =0 under the conditions that ¥(0) = | and Y(t) and its derivatives have transforms. The point t = 0 is a singular point of this differential equation such that one of the solutions is a function that behaves like log t near that singular point, and the Laplace transform of the derivative of the function does not exist. The transformed equation is d d glen) — 5 — YO] + sy(s) — 1 - gO = or (s? + Dy'(s) + sy(s) Separating variables, we have dy___sds yo +l and upon integrating and simplifying, we find that Q) 3(s) ar where C is a constant of integration.FURTHER PROPERTIES OF THE TRANSFORMATION [ste 21 63 Expanding the function for (s) by the binomial series, we have, when s > 1, Cc, 1 v9 = (+3 c > (l) 1) y (= 'en)! s[+ Ee }-eS ae where 0! = 1. The ratio test shows that series in positive powers of 1/s are absolutely convergent when s > 1. Theorem 7 therefore applies to show that the operator L~! can be applied term by term to that series to represent the continuous function whose transform is y(s) as a convergent series in powers of ¢: -1y" ) (9) =¢ ¥ CV a be ay It is not difficult to verify that for all 1 the power series (3) is a solution of Bessel’s equation (1). If that function is to satisfy the condition Y(0) = 1, then C = 1, and the solution (3) can be written @ YQ) = Jo) where J is Bessel’s function of the first kind with index zero (1 ( \" e, rs 5 t) = = w Lo 5 (ny \2 Be We have shown above that, when s > 1, 6) Li Tol} a result that is actually valid whenever s > 0, because Jo can be represented by an integral! that shows that |Jo(t)| S 1 The differential equation () PY) +tY( +(P = W)¥() =0 is Bessel’s equation with index n. It can be verified that the function gt 1k [eyez S 2 en mero: known as Bessel’s function of the first kind with index n, is a solution of that equation. In the problems to follow we shall establish the transforms of J4(0 and 0"J,(0. ‘Churchill, R, V., “Fourier Series and Boundary Value Problems,” 2d ed., p. 175, 1963. (8) IAQ) = (n = 0,1,2,...),64 SEC, 21) OPERATIONAL MATHEMATICS PROBLEMS 1. Use the transformation (6) and Theorem 2 to find L{J0(at)} listed in Appendix A, Table A.2 (transform 55) 2. If ¥, ¥', and Y” are of exponential order and Y, Y’ are continuous (¢ 2 0) while Y" is sectionally continuous, and if L{¥(0)} = y(s), show that L{PY"(O} = sy") + Asy(s) + 2918). Solve the following differential equations for Y if Y and its derivatives are to have transforms. 3." + at) — 2a¥(t) = 1, ¥(0) = YO) =0,a> 0, Ans. Y(t) = 2/2. 4. 1¥() + (¢— NYO + YO =0, YO) = 0. Ans. Y(t) = CPe™ 5. 1¥"() + 2t + 3)Y(O + (t + YO = 3e Ans, Y(t) =(C + de™ 6. PY") — 2¥(0) = 26, ¥2) = 2. Ans. Y(t) = =, 7. When k is a constant, show that the equation PZ") + UZ + kz) = 0 leads to the same differential equation in the transform 2(3). 8. With the aid of transformation (6) show that s MIO} =< Ta Use formula (8) to show that Jo(‘) = —J,(), and hence that +1 — Ld; =v {J} Jeri 9. Use the transforms of Jo and J; (Prob. 8) to show that (@) f stone — ddr = sine; lo o f IoltW y(t — 1) dr = Jolt) — cost. 10. Expand the function s~' exp(—s"*) in powers of s“! and apply Theorem 7 to verify that wife) — ravi and thus obtain the transformation 75, Appendix A, Table A.2. Apply Theorem 8 to verify the following transformations: 1". {enh = Feel (CE. transformation 76, Appendix A, Table A.2.) 1 i 12, L~'{—e"*)} = —~cosh (2,/0 a | Fac evi. (Cf. transformation 77, Appendix A, Table A.2.)FURTHER PROPERTIES OF THE TRANSFORMATION [sec 22 65 afoot Yl = 13.1 far “DOG al (n= 1,2...) (CE. transformation 57, Appendix A, Table A.2.) 14, (> 05> 0. 15. Given that a power series °_,4,«" converges uniformly over an interval 0 St <1toasum F(2), that is, the series converges to F(t) there and Not Foe) = Y Aye" + Rule) @stso where Ry(t) > 0 uniformly with respect to t as N+ co. If 0S ¢ <1, prove that eR) =D Agee, a result that was used to obtain Eq. (12), Sec. 20. Suggestion: Write Net ree (QL Age # = 0-8 * RYO) aS and prove that the right-hand member vanishes as N+ 00, when t and c are kept fixed. 22 INTEGRATION OF TRANSFORMS When a function F(t) is sectionally continuous and of the order of &, then its Laplace integral fl) = f e-*F(n dt A is uniformly convergent with respect to x in every interval x 2 04, where a > a according to Theorem 5. It follows from formula (5), Sec. 15, that when r > s >, f sores = ff erro -f Fy fem ava s Is fo 5 If the function F is such that F(0/t has a limit as t tends to zero, then the latter function is also sectionally continuous and of exponential order. Under those conditions the last equation can be written [fora = f° , 5 where g(s) = L{F(0)/t}. But g(r) + 0 as r > co (Theorem 5); hence (a) [foods = Pena ear — feat = a1) ~ a 0 (>a)66 SEC. 23) OPERATIONAL MATHEMATICS and we have established the following theorem Theorem 9 Division of the function F(t) by t corresponds to integration of the transform f(s), in this manner: Q) {fe oh f Se) dx, Sufficient conditions for the validity of formula (2) are that F(t) be sectionally continuous and of the order of e*, that s > « in formula (2), and further that the limit of F(ojt exists as t > +0. The function F(t) = sin kt, for example, satisfies the above conditions when a = 0; in particular, ¢~! sin kt + k as t+ 0. Hence when s > 0, sin ke ek dx T Ss k 3) L = = == - f= = @) { ; i f yp = 3 - atotan? = arctan Recalling how integration with respect to t corresponds to division by s, we can now write the transform of the sine-integral function (4) Sit= fe 0 7 This function is of some importance in applied mathematics. Its values are tabulated in the more extensive mathematical tables. If k = | in Eq. @), it follows from Eq, (4) that (3) L{Sit} = t arecot s arctan (s > 0). As another illustration of Theorem 9 we note that L. = x+al® a “}- Plea ail® log elk when s > —a ands > —b. Hence © fem oeh = logit? 7 bea When a = 0 and b = 1, we have the special case lees o 4 : } =v0e(1+4 (> 0) 23 PERIODIC FUNCTIONS Let a function F be periodic with period a over the half line t > 0 and sectionally continuous over a period 0 0, or =F) = Fa) whent >a, The function is bounded over the half line and sectionally continuous over each bounded subinterval, so it has a transform f(s) when s > 0. For convenience in examining the transform, we write F)=0 whent <0 and introduce a function Fo that is the same as F when 0 < ¢ < a and zero elsewhere; thus Fel) = (1 = Sot = AF, fs) = LAF) = [trae 7 Then F is described for all t by the difference equation a F(t) — F(t - a) = Fit), as we can see either directly from graphs of F(t) and F(t — a) or else by noting that when ¢ > a, the equation becomes F(t) = F(t — a), and when t 0, and sectionally continuous, then §ge-*F(o) dt (2) S(s) =r (s > 0). Let us apply that formula to the function M(ca) = 1 when 0<1 0). ‘The integral of the function M from 0 to ¢ is the function H(c,t) defined as follows: He,t) = t when 0 0). Let G denote an antiperiodic function when t > 0: Gt+c)=-G) whent>0, or G()}= —Gt—c)_—_ whent>c, which is sectionally continuous over the interval 0 0, then Sue""Gt0 ae (6) a(s) = (s> 0). For the periodic function G, with period 2c such that Gi) = GO) when0 0. For example, when G(t) = sint and c = 2, g(s) = (s? + 1)~! and the func- tion G, shown in Fig. 11 has the transform sint + find _ 1 8) LG} = x -@ypase) Similarly, the periodic function Gz such that (5> 0). Galt) = Gt) when 0). When G(t) = sin t, for example, then G,(¢) = Isin dl if ¢ = x and 1 as (10) L{lsin tl} = ETO (s > 0). In case the antiperiodic function G has nonnegative values over the interval (0,c), then G(t) < 0 when c < t < 2c and the function G, described above is the half-wave rectification of G. It replaces the negative values of G by zero. Also, in that case G, is the full-wave rectification of G, and we can write GO) = 3160 + IG), G0) = 1G). 24 PARTIAL FRACTIONS We shall now systematize the procedure of finding inverse transforms of quotients of polynomials in s. Let p(s) and q(s) denote polynomials in the variable s with no factor in common, and let the degree of p(s) be lower than that of q(s). We shall see that the inverse transform of the function (8) = p(s)/a(s) exists and that it can be found when the elementary factors of q(s) can be determined. Consider first the case in which q(s) has a linear factor s — a, not repeated. Let $43) denote the function that is left after removing that factor from the denominator of f(s); that is, Pls) _ ls) (ly) Ss) =a) 73a Note that (s) may be a quotient of polynomials. According to the theory of partial fractions, a constant C exists such that (3) c soa t scat ho Q) where h(s) represents the sum of the partial fractions that correspond to the other linear and quadratic factors of q(s), any of which may be repeated.FURTHER PROPERTIES OF THE TRANSFORMATION [sec. 24 71 In order to determine the value of C, we multiply both members of Eq, (2) by s — a, when s # a, to obtain the equation Hs) = C + (s — a)h(s), which is satisfied identically for all values of s in a neighborhood of the point s = a, except possibly at that point. But both members of the equation are continuous functions of s at that point; thus their limits as s > a are the same as their values when s = a. Therefore C = $(a). The inverse transform of the partial fraction corresponding to the factor s — a, or the term in F(t) corresponding to that factor, is (ae In view of Eq. (1), we can also write s-a Jim @(3) = lim n[ m9 = ]- Pa). ar where we have evaluated the limit of (s — a)/q(s) as the limit of the quotient of the derivatives of s — a and q(s), since q(a) = 0 and q(a) # 0 because s = ais asimple zero of q(s). Consequently $(a) = pla)/q'(a) When all the factors of q(s) are linear and not repeated and when qs) is written in the form G3) as) = (8 — ay)(S — ag)-+-(S — ay) where all the constants a, are distinct, we can write the inverse transform of f(6) in full, Let g,(s) denote the product of all the factors on the right of Eq. (3) except the factor s — a,, so that the function 4(s) corresponding to that factor is p(s)/q,(s). Then ~1 fps) Pa, ) oe PCA) oa, 4) LPO S ° {ee - 5 aan) = Ba 7a,)” The second sum here is sometimes called Heaviside’s expansion. The principal results are stated in the following theorem. Theorem 12 If f is the quotient p(s)/q(s) of two polynomials such that q(s) has the higher degree and contains the factor s — a which is not repeated, then the term in F(t) corresponding to that factor can be written in either of these two forms: Pla) 5) aye" or PS at ¢ Ha) qe) where ¢(s) is the quotient of p(s) divided by the product of all factors of qs) except s — a. Theorem 12 is valid when the constant a is any complex number.72 sec. 28) OPERATIONAL MATHEMATICS For complex arguments the exponential function is defined by the equation (Sec. 55) er) = cosy +i y) (x and y real). Example Find F(t) when 2s? — 4s Qs + Dis? + 1) We display the factors of the type s — a by writing L(9) = s?— 2s ~ (s+ As — d(s+i Using the first of the two forms (5), or the first of the expansions (4), Ss) we find that 3 L+2i -14+2i ry ade? : _ Oe Gp” * CIF DR =e? 28 “2 — 2sint PROBLEMS 1. Show that Theorem 9 applies and find the transform of: (a) (1 — cos at)/t; (6) (1 = cosh at)/e; (c) (e — cos tt Ans. (a) (b) See transforms 105 and 106, Appendix A, Table A.2; (0) log ((s? + Ds — 17s > 1). 2. The condition that F(¢)/t has a limit as ¢ + +0 was used in proving Theorem 9. If it is replaced by the condition that for some positive constants k and M, | F(t) < Mt‘ over an interval 0 < ¢< T, then g(s), the transform of F(t)/t, exists and vanishes as s+ 00. Thus formula (2) in the theorem is still valid. Use this fact to establish trans- form 36, Appendix A, Table A.2: Salve vet (> aands>b), by showing that |e" — e"/,/t < M,/t on an interval (0,7). 3. (@) Use Theorem 11 to derive the transform (3), Sec. 23, of M(c,t). (&) Apply formula (7), Sec. 23, to obtain the transform 68, Appendix A, Table A.2, of the half-wave rectification of M(k,t). (c) Use formula (9), Sec. 23, to verify that the transform of the full-wave rectifica~ tion of M(c,t) is 1/s. 4. Sketch the graph of the periodic function F for which F(t) = t when —1 0).FURTHER PROPERTIES OF THE TRANSFORMATION. [sec 25 73, 5. Let F be the periodic function such that F(t) = r when 0, ©) 869 (>, © 0 = 2, >) 6. Use Theorem 12 to find the inverse transforms tabulated below, where a, b, and ¢ are distinct constants. Sf) FO @— s ae" — be © ae-5 ab a (= det +(e ae + (a — et 6-a6-HE-9 (a= b)b— ole = a) 7. Use Theorem 12 to find the inverse transforms of the following functions. 4s+1 © 10= aay Ans. F(t) = &? ~ et? + 6" s 0) £9) = 3s? 2 4. 3en2! @ 10-353 ery Ans, F() = 2 + 3e Use Theorem 12 in solving the following differential equations. 8. YO — YO=1+e% Ans. Y(t) = Cie + Cye™' = 1+ be™. 9. ¥"W + YO —4Y'W — 4¥( = FOO, if YO) = ¥"@) = 0 and Y@ =2. Ans. Y(t) = sinh 2t + ayF (0) +(e + 3e°% ~ 4e~, 10. Yt) — 2¥"@ — Y"@ + 2Y'@) = 6F(O, if Y(¢) and its first three derivatives are zero when t = 0. 25 REPEATED LINEAR FACTORS We now consider partial fractions for the case in which the polynomial 4(s) contains a repeated linear factor (s — a)**", We write ) s w _ 7) _ 4) qs) (s—ayt*t?74 sec. 25) OPERATIONAL MATHEMATICS where (s) is the quotient of polynomials obtained by removing the factor (s — a)"*? from the denominator of the fraction p(s)/q(s). As before, the degree of the polynomial p(s) is assumed to be lower than the degree of q(s). Note that (s) and its derivatives are continuous functions at the point s=a, The representation of f(s) in partial fractions now has the form o0) Ao AL 4, An Goat scat qcart tear tpt. Q) where the numbers A, are independent of s, and h(s) is the sum of the partial fractions corresponding to the remaining factors of q(s). It follows from Eq. (2) that (3) G(s) = Ao(s — a)" +--+ + Als — al'h +--+ + Ay + (8 — a)" Th(s) in a deleted neighborhood of the point s = a. When s— a, we see that An = (a). To find the remaining coefficients 4,, we differentiate both members of Eq. (3) with respect to s,n — r times, in order to isolate the number 4, When s ~ a in the resulting equation, we find that g(a) = (n= n)!A,. Equation (2) can now be written in the form » gq) @ $9) = Yn erat where 0! = 1 and $'"(a) = g(a). If H() = L~"{h(s)}, it follows that the inverse transform of f(s) is ne ginny 6 ro= 5 (nm — rir + h(s), Ye" + H(0). This equation can be simplified by recalling the formula for the derivative of order n of the product of two functions u(s) and v(s), namely, Sw = S a Soin Hirt "(sos When wu = (5) and v = form , then d'v/és" = te" and Eq. (5) reduces to the 6 Fe = ai gatooen} + HO) This result can be stated as follows.FURTHER PROPERTIES OF THE TRANSFORMATION [sec 26 75 Theorem 13 If f(s) is the quotient p(s)/q(s) of two polynomials such that (s) has the higher degree and contains the factor (s — a)'*', then the term in F(t) corresponding to that factor is ®,(a,t), where o nl as" ™ ®,(s,t) = [o(sJe"] and $(s) is the function indicated by Eq. (1). The term in F(t) corresponding to a factor (s — a)? in q(s), for instance, is (8) ®,(a,t) = [$(@) + Hale", and the term corresponding to a factor (s — a)? is (9) A(a,t) = 3[4"(a) + 2G(a)t + Pla)t)e* ‘As an example, if —— 6-6-2” then the term in F(t) corresponding to the factor s — 1 is e. To correspond with the factor (5 — 2), we have $(s) = (s — 1)! and $(s) = —(s — 1)? so that (2) = 1 and $(2) = —1. In view of formula (8) the term in F(0) is (—1 + de". Hence f(s) = F(t) = e + (t — Ie" Since the number a may be imaginary and since a factorization of every polynomial into linear factors, real or imaginary, exists, Theorems 12 and 13 provide a systematic way of finding inverse transforms of quotients of polynomials in all cases where the factors of the denominator can be determined. If, however, imaginary factors are present, the results are given in terms of imaginary functions. The reduction of the results to real forms is sometimes tedious. To obtain the real form of the inverse transform directly, and to observe the character of that function in general, we may proceed as follows. 26 QUADRATIC FACTORS In this section we assume that the polynomials p(s) and q(s) have real coefficients. The imaginary zeros of q(s) then occur in pairs, each pair consisting of some complex number a + ib and its conjugate. The corre- sponding linear factors of q are a U where a and b are real numbers. The product of those factors is the real quadratic factor (s — a)? + B? —a-—ib, 42 =s—a+tib (b 4 0),76 SEC. 26) OPERATIONAL MATHEMATICS Let q have the factors (1), not repeated, and let , and $2 be the quotients of polynomials obtained by removing the factors q, and q2 in turn from the fraction p(s)/q(s); thus _ Pls) Gals) Ga qs) s-a—ib s—a+ib (2) £3) According to Theorem 12 the sum of terms in F(t) corresponding to those two linear factors is 3) G(t) = ep ,(a + ibje™ + (a — ibje- 7), This is the component of F corresponding to the quadratic factor (s — a)? + b?. We shall represent it in terms of real-valued functions. The rational function $,(s) is continuous at the point s =a + ib. Ina neighborhood of that point excluding the point itself (4) G(s) = (s — a — ib) f(s). Similarly, @, is continuous at the point a — ib and (5) 2s) = (s — a + ib) f(s) throughout a neighborhood of that point with that point deleted. From elementary properties of complex conjugates we can see that the conjugate of f(s), a rational function with real coefficients, is f(S). Consequently $26) = (5 — a + ib) f6) = (6 — a — 1) = 4,09. In view of the continuity of ¢, and 5 it follows that 6 x(a — ib) = Fa + Since e~ = cos bt — isin bt, it is the conjugate of e®. Thus formula (3) can be written Gt) = eb (a + ibye™ + bila + ibje™) The factor in brackets is twice the real part of the first term, 2 Re[@ (a+ ibje™]. So if we let r, and 6, represent polar coordinates of the point representing the complex number $,(a + ib), which is never zero, ) ,(a + ib) = r,e = r,(cos 0, + isin O,), we can write the component G in the real form ® G(t) = 2e*r, cos (bt + 04) (b 40,7, > 0). If a=0, the component G(t) is the simple periodic function 2ry cos (bt + 04), and if a <0, it is a damped periodic function. Those components, corresponding to cases a < 0, represent stable oscillations inFURTHER PROPERTIES OF THE TRANSFORMATION (sec. 26 77 the theory of control of mechanical and electrical systems where t denotes time. Stable oscillations are bounded as ¢ > co. When a > 0, the com- ponent G(t) represents an unstable oscillation. To illustrate the use of formula (8), we find F(t) when WOE 2-2 2s -2 165) = DEF ETD CF DS+D +4 541-2 2s —2 ee 9) = Gee 1 Foy hd Iai V2 in Then oI +2) = San t= Se and G(t) = \/2e~*cos (2t — n/4) = e~(cos 2t + sin 2). After adding the term corresponding to the linear factor s + 1, we find that F(t) = e~ (cos 2t + sin 2t — 1). In case q contains the square of the quadratic factor (5 — a)? + b?, we write ps) ___ gal) bas) qs) (s—a— iby a+ ib? and apply formula (8), Sec. 25, to get the corresponding component of F(t) G(t) = e*[b4(a + ib) + thy(a + ibyje™ + eM [ps(a — ib) + tx(a — ibje™™. But $2(s) = (s — a + ib)f(s), and as before we find that (5) = (5), also that $4(3) = 6{(), so that (10) Git) = 2e* Re [hy(a + ibje™ + th,(a + ibje™). 0) f(s) = We use absolute values and arguments of the complex numbers ,(a + ib) and $4(a + ib) displayed by the polar forms (iy ya + ib) = rye, h(a + ib) = pye™, where r; # 0 because ,(a + ib) #0. Then formula (10) becomes (12) Gt) = 2e"[p, cos (bt + Wy) + rytcos (bt +O,)] (ry > 0). Our results can be summarized as follows. Theorem 14 When f(s) = ps)/q(s) where p(s) and q(s) are polynomials with real coefficients and q has the higher degree and contains a nonrepeated real quadratic factor (s — a)? + b*, where b #0, the component of F(t) corresponding to that factor is the function G(t) given by formula (8).78 sec. 27) OPERATIONAL MATHEMATICS If q contains the square of that quadratic factor, the component of F(t) is given by formula (12), We note that the component (12) represents a stable oscillation only if the real part of the imaginary zeros a + ib of q is negative: a < 0. Ia = 0, the component contains the term 2r,¢cos (bt + 6,) representing an unstable oscillation. When the quadratic factor appears to a degree n + I (n = 2,3,...),we can see from the procedure we used when n = 1 and from formula (5), Sec. 25, that the component in F will contain a term of the type Cet" cos (bt + 61), where C is a constant. When p and q are not polynomials in s, partial fractions in s cannot be used. Such cases will be treated in Chap. 6 with the aid of residues and contour integrals. 27 TABLES OF OPERATIONS AND TRANSFORMS Appendix A, Table A.1, contains a list of operations on F(t) with corre- sponding operations on f(s). That table of operations summarizes several of our results on the theory of the Laplace transformation. The table of Laplace transforms (Appendix A, Table A.2) gives a fairly extensive list of transforms of particular functions. Derivations of a number of them have been presented above. References to some more extensive tables can be found in the Bibliography. Transforms 82 to 84, Appendix A, Table A.2, of functions that are prominent in problems of diffusion and conduction of heat, will now be derived. A more direct derivation will be made later on in Chap. 6 with the aid of contour integrals and functions of a complex variable; but we wish to use those transforms earlier, in Chap. 4. First we perform formal manipulations that will suggest the inverse transforms of the two functions a vs) (k20,s>0), Q) 2s) =e = (k > 0,5 > 0). We see that (\/sy)' = —kz/(2/s) and 22’ = —ky, so that y and z satisfy this system of differential equations with coefficients that are linear in s GB) 2sy(s) + y(s) + k2(s) = 22'(s) + ky(s) = 0. The corresponding system for the inverse transforms Y(t) and Z(t) is therefore A-t¥y + V+kZ=0, UZ +k¥=0.FURTHER PROPERTIES OF THE TRANSFORMATION (sec. 2779 Thus —2rY’ — ¥ + kZ =O and Rol k 4) ") = [= —<| = =y, @) 2vi=(B ; Yo, AH= 5 YO. The solution of system (4) is i KR ¥() = wee ( -4), ZX) But when k = 0, then y(s) = 1/,/S, and we know that Y(i) is then 1/,/nt, so if C is independent of k, it follows that C = 1/,/7, and our formal results can be written oem ate \ a k We 6) Haga? (“i } seh (k > 0). Let us now prove that transformations (5) and (6) are correct when k= Lands > 0. The function ° (k20), 1 1 =— — ¥(0) = 0, (7) ¥,(t) sae | al when t > 0, ¥(0) is continuous when t 2 0, and bounded. Hence when s > 0, © 1) at =f? L\?] a i= fereo(—2] nfo -(va st) vavi) = | Pla) 7 , P| - (V5 - 5 Jl Fi We substitute t for 2,/t to write eI 12 (8) Sre*y,(s) -{ exp [- vs, | |e 6 25 and make a further substitution }/st = 1/A to see that ° i a wy (3) = 2 (vs, —1)7]¢% 6 vie n= ef evo | S13) |S By first adding corresponding members of equations (8) and (9), then substituting x for 4\/sd — 1/2, we can write - 2° Js, 1? Fy (5) = _ a 2/ney,(s) = ah oo 5 A i) | i exp (—x2)dx = “Fl.80 SEC. 27) OPERATIONAL MATHEMATICS Therefore (s > 0), yils) = Lge (- a} and when k > 0, transformation (5) follows by writing y,(k?s) and applying Theorem 2. Formula (5) was established earlier when k = 0. If k > 0, the function Z, defined as follows, Zt) = when t > 0,Z(0) = exp ( -‘) 2/a 4 is continuous and bounded when £ = 0; also Z 2 Y(y = see (- ie and y(sy=—he“HF (5 > 0) VJs Vat 4t| Theorem 9 therefore applies to give the transformation 4kt~'Y where as) =k eM ent (s > 0), 7 vx which is transformation (6). Finally, in view of transformation (6), we note that ert 2 eta val 2Ja Vadisi 2f, - fe aah ace a2 Fado Therefore (10) ified =1- «(5 ({k 20,5 > 0), where erf(x) is the error function defined in Sec. 16. Equation (10) can be written k 1 ay) tere ()h =e (k = 0,8 > 0), Wiss where the complementary error function erfc (x) is defined as en? da. wh (12) erfe(x) = 1 — erf(x) =FURTHER PROPERTIES OF THE TRANSFORMATION [sec 27 a1 PROBLEMS 1. Find the inverse transforms tabulated below: Ss) FO @ sta b oem ab |ow C+ DEP o-oo 6 o-F SEs) Hi got O BINE+ BD ems eeese 264 22 © ee 2cos bt + b*? — 2 sib © aa tos bt O) ee (5t — 8)e~! + 6 sin 2t + 80s 2t oF e+ 4) 2. Without finding F(0), determine whether the oscillations of that function are stable or unstable when its transform is 2s? + 1 @ S90 =s558Es Ans, Stable. s Wass ee Ans. Unstable. (b) f(9) = 3. Solve the following differential equations: @ Y"@ -2¥ + YO =1 Ans. Y(t) =(C, + Cxbe +1 (b) ¥"(0) + Y(e) = 2sinr, if YO) = 0 and YO) = —1. Ans. Y(t) = —teost. (0) 4¥"(@) + 4¥"() + YO = FO. @ ¥%) +2¥ + YQ =0, if YO=0, YO=1, YO ¥"@) = -3. Ans, Y(t) = t(sin t + cost). 4. In Sec. 26, if we write gy(a + ib) = cr, + if, show that formula (8) for the com- ponent G can be written G(t) = 2e*(a, cos bt — B, sin bt). 5. With the aid of transformation (5), Sec. 27, find L~!{s~# exp(—k,/s)} listed as transformation 85 in Appendix A, Table A.2. 6. Solve for Y(t) (@) (= PAY" + 2") + 2¥() = 61, YO) = YQ) Ans. Y(t) = (b) YO + 245 ¥(x) cos (t — x) dx = 9e” 0. = 8 Mt = 1S(t = D. Ans. Y(t) = Se* + 4e~' — 61e™™ 7. Draw the graph of the ramp function ¥ such that, Yo —n when 2nSt<2n+1, ¥()=n+1 when n+15 -¥-1) and YQ =0 when 1 <0, and show that ¥(082 SEC. 27) OPERATIONAL MATHEMATICS (l+e) OE 8. When F, G, and H are sectionally continuous, show that Fi) * GQ) * Ho = f Fu af G(x — YH) dy dx. 9. When Fis sectionally continuous and G and its derivative G’ are continuous over an interval 0 St S T, show that, at each point (0 << T) where F is continuous, sro * GO] = FQ *G@ + GOFO. 10. Under the following set of conditions the limit of the integral a(x) = i Sls) de @>0, as x + 00, is the same as the integral of the limit. (a) Let f(x,t) be continuous over each rectangle ¢ < x S C,0 StS T, except possibly for finite jumps across a finite number of lines = 1, in each rectangle, and (b) let the above integral converge uniformly with respect to x when x > c. Also, (c) let f have a limit FO = lim f(x) uniformly with respect to 1(t # 1,) on each interval 0 < ¢ < T; that is, for each positive « there is a number N; independent of ¢ such that |F@ — fol <« whenever x > Ne @ I (cf. Prob. 10). Show that lim q(x) = x/2 while the jegral of the limit of the integrand, as x + co, is zero. ae 12. If Flt) 2 G(e) when t > 0, then f(s) 2 (3) for every real value of s for which the transforms f and g exist. Prove that property, and illustrate it with some particular functions. Note that the case F(t) = 2sin t, G(t) = sin 2r shows that the converse of the property is not always valid.FURTHER PROPERTIES OF THE TRANSFORMATION [sec 27 83 13: Initial-value theorem If F is continuous over the half line t 2 0 except possibly for a finite number of finite jumps, while F’ is sectionally continuous over each bounded interval there, and if F and F’ are of exponential order, then the limit of the product 3f(6), a8 5 -» 00, exists and equals the initial value of F: lim sf(s) = F(). Use a formula for L{F'} and Theorem 5 to prove that theorem. 14, Illustrate the initial-value theorem (Prob. 13) with these particular functions: (@) F()= Az (6) FO) =Solt ~ to); (6) F()=t3 (@) FQ) = cos kt; (¢) F()= {1 — Sot ~ to). 15. If f(s) = p(s)/a(s) where p and q are polynomials whose terms of highest degree are a,s" and b,s"*', respectively, use the initial-value theorem (Prob. 13) to prove that FO) = a,/b,. 16. Let F and F’ be continuous while F” is sectionally continuous over each interval 0 StS T, and let all three functions be of exponential order. Then if s?(s) has a imit as s + 0, prove that lim sf(9) = FQ) and F(0) = 0. 17. Illustrate the modified initial-value theorem presented in Prob. 16 in case F(t) is (a) sin ke; (6) at + br?; (c) 1 — cosh. 18. 4 final-value theorem Let both F and F’ be bounded over the half line ¢ 2 0 and continuous there except possibly for a finite number of finite jumps; also let |F(| be integrable from t = 0 to t = 00. Then lim, sf(s) = lim F(0. Prove that theorem. In the formula for L{F"} note that our conditions ensure the continuity of that Laplace integral with respect to s when s = 0 (Sec. 15). 19. Illustrate the final value theorem (Prob. 18) in case (a) F(t) = A + Be™'; (0) FQ) = Sole — to); (€) Fld) = t = (¢ — 1Solt — 1). 20. Generalized convolution For a function F(t,t) of two variables, defined over the quadrant t > 0,2 > 0, let f(5,2) denote its transform with respect to z. Using the same parameter s, let f (s,s) denote the iterated transform: Hiss) = i e-"f(5.2) de = fen f eo" F(x) dt de Indicate formally why /(s,s) is the transform of the generalized convolution F *(0) of F(t.) defined below Sis) =L{F*()} where F() = fre-sode Itis convenient to write F(¢,t) = 0 when either 1 < 0 ort <0. (See Prob. 21 for condi- tions of validity of the generalized-convolution property.) When F(t,t) = F()G(0), note that the property reduces to the convolution property (6), Sec. 16. 21. Prove that the generalized-convolution property written in Prob. 20 is valid when s > a under these conditions on F: |F(t,t)| < Mexp[a(t + 1)] for all positive " Similar theorems with conditions on f(s) rather than F(Q) will be found in Chap. 6.84 SEC. 27) OPERATIONAL MATHEMATICS t and t, and F is continuous by subregions (Sec. 14) over each rectangle, 0 < 1 < T, Ost SR Sec. 16). 22. When F(t,2) = So(t — t) in Prob. 20, show that /(s,s) = 4s"? when s > 0, also that F *(t) = 41, and thus verify the generalized-convolution property in this case. 23. When F(t,1) = Jo(2\/t2), we found (Prob. 10, Sec. 21) that f(s,t) = s~! exp (—t/s). Apply the generalized convolution property (Prob. 20) to derive the integration formula fsa T(t — 1) dr = sine (2 0). o 24. When + is fixed and positive, let Z(t,t) denote the solution of the differential equation with constant coefficients (independent of t and 1) nw Za) + Y a gaZlest) = FO) such that Z and d"Z/dt"(n = 1,2,...,m — 1) all vanish when 1 = 0. Write the corre- sponding equation for the iterated transform 2(s,s) of Z(t,t) (Prob. 20) and compare it with the equation for the transform y(s) of ¥(0), where Yo + x a, YO) = FQ, YO) = YQ) =- ye" 0) = to show that y(s) = s%(s,s). Thus deduce the following formula for the solution of the problem in ¥ in terms of the solution of the simpler problem in Z: Y= [fae —41)de. 25. Use the formula in Prob. 24 to find the solution of the problem Y"(O-— YO= FO, — YO) = YO) =, in the form Y(t) = f',F(2) sinh (¢ ~ t)de. Note that —F(-) is a particular solution of the differential equation in Z(t,z).3 Elementary Applications The properties of the Laplace transformation that we have derived up to this point enable us to solve many problems in engineering and physics involving ordinary linear and partial differential equations. In this chapter we shall solve a number of problems in elastic vibrations involving ordinary differential equations. They are problems in which our method is very convenient, although not essential. We shall also treat some simple applica- tions of integral equations. The next chapter contains applications that involve partial differential equations. The solution of problems of this type is a primary objective of this book. In later chapters we shall extend our treatment of such problems. 28. FREE VIBRATIONS OF A MASS ON A SPRING Let a body of mass m attached to the end ofa coil spring (Fig. 12) be given an initial displacement and an initial velocity and allowed to vibrate. The other end of the spring is assumed to be kept fixed, and the spring is assumed to86 SEC. 28) OPERATIONAL MATHEMATICS pete Fig. 12 obey Hooke’s law, so that the force exerted by the free end is proportional to the displacement of that end. The factor k of proportionality is called the spring constant. We also assume that the mass of the spring can be neglected in comparison with the mass m and that no frictional forces or external forces act on m. Let X denote the displacement of m from the position of equilibrium; that is, let the origin O denote the position of m when the spring is not deformed. Then according to Newton’s second law of motion, ex 0 "ae kx (k > 0), Let xo denote the initial displacement and vo the initial velocity, so that the function X(1) satisfies the conditions Q) X00) =x XO) = 0. We can determine the function X(t) by applying the Laplace trans- formation to both members of Eq. (1) and using the conditions (2). Thus if x(s) denotes the transform of X(0), it follows that m[s?x(s) — sxq — vo] = —kx(s), s and therefore x(s) = *o2 5 Gm + Pos my Hence » 3) X(0) = x9 Cos wot + —* sin wot @ The motion described by formula (3) is a simple vibration with angular frequency @o, called the natural frequency of this system, and phase angle a, and with the amplitude [xp? + (v9/@o)"J*. If a viscous damping force proportional to the velocity also acts upon the mass m, as indicated by the presence of a dashpot cin Fig. 13, the equation of motion becomes (4) mX"(0) = —kX() — eX",ELEMENTARY APPLICATIONS [sec 28 87 i ol —>x where the coefficient of damping c is a positive constant. Let the mass start from the origin with initial velocity v9 X()=0, XW =r%. The equation in the transform x(s) becomes Fig. 13 ms?x(s) — mv = —kx(s) — csx(s) or, if we write 2b = c/m, and again write @? = k/m, _—_%e = 8? + 2bs + Wo? (s + b)? + wo? — B? (5) x(s) If b? < @p?, that is, if the coefficient of damping is small enough that c? < 4km, then the formula for the displacement is © X(0) = volo? — b?)-#e™™ sin (¢ 07 — b’). In the case of critical damping, that is, when wp = b, or 2 = 4km, it follows from Eq. (5) that @ X(0) = vote. We can see from this formula that the mass m moves in the direction of vp until the time ¢ = 1/b, then reverses its direction and approaches 0 as t tends to infinity. When c? > 4km, a similar motion of the mass takes place. The dis- cussion of this case and the case of other initial conditions is left to the problems The mathematical problem treated in this section can be interpreted also as a problem in electric circuits. This well-known analogy between problems in vibrations of mechanical systems and electric-circuit theory will be observed for other problems in this chapter. Naturally, the notation and terminology differ in the two types of problems. In the electric circuit shown in Fig. 14, let Q be the charge accumulated in the capacitor C at time t, and J the current in the circuit, so that (8) 1) = Qo.88 SEC. 29) OPERATIONAL MATHEMATICS Fig. 14 This equation requires that 1(t) be positive when Q(t) is increasing. If the positive sense of flow I(t) of positive charges is taken in the clockwise direction in Fig. 14, then Q(t) measures the charge on the upper plate of the capacitor. The circuit has a resistance R and a coil of inductance L. Since the sum of the three voltage drops in the circuit is zero in this case of no impressed voltage, then (9) LI) + a) + RI(t) = 0. The system of first-order differential equations (8) and (9) can be used directly to determine I(t) and Q(¢) in terms of initial values of those functions. By eliminating [(:) from the two equations, however, we see that (10) LOW + ROW + 201) =0. Except for the notation used, this equation is the same as Eq. (4). When the resistance is negligible, R = 0, the equation reduces to our Eq. (1). The initial conditions in the electrical problem can be made the same as those in the mechanical problem. For example, if the capacitor has an initial charge Qo and if the initial current is Ig, then Q(0) = Qo, Q'(0) = To, which are the same as the initial conditions (2) in our first mechanical problem. 29. FORCED VIBRATIONS WITHOUT DAMPING Let an external force F(t) act upon the mass in the mechanical system of the last section, assuming there is no damping (Fig. 15). The displacement X(1) of the mass m then satisfies the differential equation a mX"(t) = —kX(t) + FO. If the initial conditions are 2) X(0) = Xo, X"(0) = boELEMENTARY APPLICATIONS [sEc. 29 89 pow Fn Fig. 15 o—-x the equation in the transform x(s) becomes m{s?x(s) — sxo — vo] = —kx(s) + f(9), where f(s) is the transform of the force function F(t). Let @ again denote the natural frequency of the system, Then we can write 3 s) = “0S tM 8 x9) = 5 Hence the displacement for any F(t) can be written, with the aid of the convolution, as 0 ae (4) X( = xocosagt + Hsin wot + wm J, sin wo(t — 2)F(2) dz, a result that satisfies conditions (1) and (2) above. But the motion of the mass under particular external forces F(t) is more interesting than the general formula (4). In these special cases it is often easier to refer to the transform (3) than to formula (4). When F(t) is a constant Fo, as in the case when the X axis is vertical and the force of gravity acts on m, Eq. (1) can be written mX"() = —k [x = %). If ¥ = X — Fo/k, this becomes mY” = —KY; so the motion is the same as free vibrations if displacements are measured from a new origin at a distance of Fo/k units from 0. Note that Fo/k is the static displacement Xp of the free end of the spring due to the applied force Fo, since Fy = kX. In case (6) F(t) = Fo when 0 < t < fo, =0 when f > fo, _ pf _ pets then fi) = ri a90 SEC. 28} OPERATIONAL MATHEMATICS If x9 = 0 = 0, it follows from Eq. (3) that © x(s) = Fe lata EXP (= to) ] . m| ss? + @o2) (5? + @2), Now 1 1 2 1 i -y- aa {aa + ast ail! COS Wot) arsin? 30h and if we write W(t) = sin? dat when t > 0, =0 when t < 0, it follows from Eq. (6) that 2Fy 7) XW = SSW — We to). The graph of this function can be drawn easily by composition of ordinates. When to is approximately $/o, the graph is the full-drawn curve in Fig. 16. When tf = 21/0, it follows from Fig. 16 that the mass m performs one oscillation and then remains at the origin (Fig. 17). The step function (5) is proportional to the unit finite impulse function (3), Sec. 12, with impulse starting at t = 0. The area under the graph of our function is Foto. If this is kept fixed, Foto = Mo, while tg +0 and Fy > 00, the force F(t) becomes the impulse represented formally by the equation (8) F(t) = Md), where 6(¢) is the unit impulse symbol (Sec. 13). The constant Mg represents the increase in momentum of the mass m at the instant ¢ = 0.ELEMENTARY APPLICATIONS [sec. 30 91 0 (0) 7 Fig. 17 The formal transform of M, (1) was found to be My. This suggests that, in view of Eq. (3), when xo = 09 = 0, M, 1 On FEO and hence that M 9} X() =—* sin wot. 0) (0) mang 10" It will be left as an exercise to show that formula (9) follows rigorously as a limiting case of formula (7). But the formal solution can be verified as, well by noting that, according to formula (9), the momentum of the mass m, M(0) = mX'() = Mg cos wot, satisfies the condition M(+0) = Mg. If the mass started from rest, then its momentum jumped to the value My at the instant t = 0. 30. RESONANCE Let the external force in the problem of the last section be F(t) = Fosinot, where Fy and @ are positive constants. Then according to Eq. (3), Sec. 29, XS + 09. Fy o q oe Sg Y 9) = FF att m+ aye + oF) and, if # wo, 1 2) X(t) = Xo COS wot + al + @®o92 SEC. 30) OPERATIONAL MATHEMATICS That is, the motion is the superposition of two simple harmonic motions, one with frequency @y and known as the natural component of vibration, and the other with frequency « which is called the forced component of the vibration. Note that the natural vibrations are not present in case vy = Fo? 0 = to? a? However, if the frequency of the periodic force F is the same as the natural frequency of the oscillator, « = wp, then X08 + 09, Fo 34 wy?) m+ wy @) x(s) The presence of the repeated quadratic factor in the denominator here shows that X(t) will contain an unstable component (Sec. 26) having the form of the product of t by a cosine function. In fact, Fo me 1 R (4 X() = xocos mot + arove + 3 sin wot — 2 £08 wot 0 Im, In view of the last term here, the amplitude of the oscillations of m increases indefinitely. In this case the force F(t) is said to be in resonance with the system. We note in particular that if xo = 0 and v9 = —Fo/(2mao) the resonance type of motion reduces to Fy Xi) = - = feos Wot, shown in Fig. 18.ELEMENTARY APPLICATIONS (sec. 90 93, PROBLEMS. 1. A good mechanical oscillator for demonstrating vibrations can be constructed by selecting a length of coil spring such that when a weight of } 1b (= mg, g = 32 fy/sec*) hangs at rest from the free end, the spring is extended by 0.2 ft. Show that the spring constant then has the value k = lb/ft, and that the natural frequency of the oscillator is = 4\/10 = 12.65 rad/sec, or about 2 cycles/sec, and hence the period T of oscillation, T = 2n/av, is approximately $ sec. 2. Let the supported upper end of the spring in Prob. 1 be moved vertically with displacement Z(0), where Z(0) = 0. Let X(t) denote the vertical displacement of the mass m from its position on the lower end of the spring at time t = 0 when the spring has its natural length. Also let Z and X be positive for displacements downward. If damping forces are neglected, note why X(¢) then satisfies the equation mX"(t) = —K{X() — Z(0)] + mg. Let Y(t) be the displacement of the mass m from its static position when supported by the spring while Z(t) = 0. Show that mY") = —K¥() — Z2(0] = —kY(O) + KZ, and hence that Y also represents the displacements in a horizontal oscillator with an external force kZ(0) applied to the mass. 3. The upper end of the spring in the oscillator described in Probs. 1 and 2 is suddenly lowered 1 in. at the instant ¢ = 0 and kept in that position during twice the natural period 27 = 4r/on9 sec (about 1 sec), then suddenly returned to its initial position, so that Z(t) = {1 — Solt — 27) a) If the mass m (of weight 5 1b) initially hangs at rest, show that it then makes two oscilla- tions 2 in. downward from and back to its initial position, then remains there after the instant ¢ = 27, since Y= bsin? when0StS27, Y()=0 whent 227. 4. In Prob. 2 if Z(t) = vot and ¥(0) = ¥'(0) = 0, show that ¥% Y (0) = vot — 72 sin wot (0) = vot 5 and describe the motion of the upper end of the spring and the mass. 5. Let the upper end of the spring in the oscillator described in Probs. 1 and 2 be moved vertically by means of a rotating eccentric or cam with the periodic displacement Z = dysin wot ft, with the natural frequency «wp = 4/10 rad/sec of that oscillator. If ¥@) = ¥(0) = 0, show that Lo & Y(O = 34 sin wot — Fer cos wet. Note that the vibration of the mass is unstable under that periodic 2-in. oscillation of the support, and that within only four cycles of the oscillations (¢ = 87/ay < 2sec), displacements of the mass mount to approximately 1 fi94 Sec. 30) OPERATIONAL MATHEMATICS 6. If the coefficient of damping for the oscillator shown in Fig. 13 is so large that [c? = 4km 2 > 4km and if X(0) and X"(0) "9 (@) show that X() = 2 sinh (ar) exp ( - sh. where a = (6) Find X(o) and show that the mass m moves in the direction of vg until the instant J igge 2am 2a'°Fc— 2am’ when it turns and approaches the origin. 7. Let the mass m in the damped oscillator shown in Fig. 13 be released from rest with an initial displacement X(0) = xo, and write 26 = c/m. (a) If c? < 4km and @, = (@o? — b?), show that X(t) = xo22e-™ cos (wit — a) where sina =>, Osa<% Oy Wo (b) if c2 = 4km show that X(t) = xoe7™(1 + bt), and hence that the mass never moves across the origin. 8. When the force on the undamped oscillator shown in Fig. 15 is the periodic function F(t) = Fo sin (wt + 0), where @ is a constant, show that resonance occurs when « = @o. 9. When ¢ = 0, let the current in the circuit shown in Fig. 14 be zero and let the capacitor have a positive charge Qo. If R > 2\/L/C, derive the formula [=a 10. The current J, and the charge Q on the capacitor in the circuit shown in Fig. 19, are functions of t that satisfy the conditions dQ 12 4S4Rr= arc when Q and / are initially zero, where ¢ is the time after closing the switch K, and the electromotive force E> is constant. If b= R(@L)"! and «4? =(LC)"! — b? > 0, derive the formula for the current. Show that I(¢) < 0 and I(co) = 0. Q= j Ita) dr, 10) = 0, if Fig. 19ELEMENTARY APPLICATIONS (SEC. 31 95 31, FORCED VIBRATIONS WITH DAMPING When an external force F(t) acts parallel to the X axis on the mass in the damped oscillator shown in Fig. 13, the equation of motion becomes 0 mX"() = —KX( — eX) + FO. If X(0) = X’(0) = 0, the solution of the transformed equation is 1 1 (2) (s) = — f(s) —_..—3.— ce k Q x= OS RTae TE (>- 50-4. Again, consider the periodic force F(t) = Fy sino and the case c* < 4km, Then b < @o, and Eq. (2) becomes Fo o m (s? + @ f(s + bP + @,7) It follows from Theorem 14, Sec. 26, that X(t) consists of terms of the types Acos (wt + 8) and Be~™ cos (w, + ,), where A, B, 0, and 6; are constants. Consequently the component of oscillation with frequency @, is nearly damped out after a sufficiently long time, and the periodic forced component of vibration (4) X,() = Acos(wt +) (A> 0) (01? = wo? — 6»). (3) x(s) = remains. Its amplitude A is, according to Sec. 26, 2p,(ic)| where (8) = (8 ~ ic)x(s). Thus IFol 1 6 4 “m [ia + bY +o, The square of the absolute value in the denominator can be written as log? — w? + 2ibal?, which is (o — 92)? + 462002, and by expanding and completing the square, we can write that polynomial in @ as (w* + 2b? — 2)? + 4b*(w9? — b?). Thus 6 lol m{(w? + 2b? — @o?)?? + 4b@,7F* ide A of the periodic vibration depends on the frequency « of the force Fo sin, and it varies directly with the amplitude |Fo| of that force. When Fy is fixed, the value of w for which A is greatest is the resonance frequency ©, of our damped oscillator under that periodic force. Let the coefficient of damping be small enough that c? < 2km. Then 2b? — wo? < 0, and we can see from formula (6) that A is greatest when96 SEC. 32) OPERATIONAL MATHEMATICS 2 ©? = wy? — 2b?. Thus the resonance frequency in that case is a) o, = Jon — 28 2b = Se < 2km|, a frequency only slightly less than @y when ¢ is small. When w = @,, the amplitude A of the undamped component of vibration has the value Fl _ WF ® ~ 2bma, ~ coo, We can see that the expression (3) for x(s) cannot contain a repeated factor of the type (s? + 7)? in its denominator unless c = 0 so that b = 0. Hence no value of the frequency @ of our periodic force will induce a com- ponent of X(t) of type tos (wt + 8) as it did for the undamped oscillator. The amplitude A of the forced component of vibration is still given by formula (5) when the force Fo sin wt is replaced by a force Fo sin (wt + a) (Prob. 3, See. 32). In case c? = 2km, it turns out that the amplitude A of the forced vibration tends toward an upper bound |Fol/k as @ tends to zero. When @ is small and ¢ is large enough that ot is near an odd multiple of 1/2, the magnitude of the force Fo sin wt is then near |Fol. Under static conditions the force |F,| would displace the end of the spring a distance |Fol/k. The formula for X() when c? = 2km, found in Prob. 4, Sec. 32, shows that when @ is small the mass may oscillate slowly with an amplitude near | Fol/k. 32. A VIBRATION ABSORBER We have seen that for the simple damped oscillator with an exciting force Fysint the forced component of vibration A cos (wt + @) remains un- damped. Let another spring and mass be connected in series with the original mass (Fig. 20), where that second oscillator is undamped. We shall see that if the spring constant and mass of the auxiliary oscillator are chosen so that the natural frequency of that oscillator coincides with the fixed frequency « of the exciting force, then the forced vibrations of the first mass will be eliminated. Fa a 7 y—j » }—-yELEMENTARY APPLICATIONS [sec. 32 97 Let X and X, denote the displacements of the masses m and m,, respectively, from the positions they have when both springs have their natural lengths. If the exciting force is F(t) = Fy sin ot, and if m and m, are initially at rest at their respective origins, then the displacements X(t) and X,(f) satisfy the following system of differential equations: xX > =~ RX + kX, - X)- ty Fo sin cot ee = Tw (X1 — X), X() = X'(0) = X,(0) = X4(0) = 0. The transforms x(s) and x,(s) of X(t) and X(t) therefore satisfy the simul- taneous algebraic equations (ms? + es + k + ky)x(8) — kyxy(s) = 7 Kyx(s) — (mys? + ky)x4(s) = 0. Eliminating x,(s), we find that a x(s) = Fog ae, where (2) P(s) = (mys? + ky)(ms? + cs + k + ky) — ky?. In view of the presence of the quadratic factor s? + «? in the denomina- tor of the right-hand member of Eq. (1), it follows that X(z) will contain a term of the type @B) Cos (wt + 0) unless k,/m, =o, in which case the numerator of the above fraction cancels with the factor in the denominator, leaving 1 (4) = _. 4 x(s) Foomay The inverse transform X(t) of the function (4) represents a damped oscillation of the mass m if each of the four roots, real or imaginary, of the98 SEC. 32) OPERATIONAL MATHEMATICS equation p(s) = 0 has a negative real part (Secs. 24 to 26). Now the poly- nomial p(s) is the determinant of this system of homogeneous equations in xand x;: a (ms? + es +k + ky)x — kyx, = 0, —kyx + (mys? + ky )x, = 0. When the determinant vanishes, that system is satisfied by numbers x and x,, both different from zero. Let s be any one of the four roots of the equation p(s) = 0 and let 2,2, denote a pair of nonvanishing roots of the system (5) that corresponds to that value of s, where z and z, may be either real or imaginary numbers. We now write x = z and x, = z, in Eqs. (5), then multiply the members of those equations by the complex conjugates Z and 2,, respectively, and add to obtain the equation (6) (mzz + m,z,2,)s? + cz3s + B= 0, where B= kz + ky(@z — 22 — 212 + 242) = kz? + kylz — 24)? > 0. If we write 4 = mlz|? + m,|z,|?, then Eq. (6) becomes fa) As* + cld?s + B= 0, where the coefficients A, clz|?, and B are positive numbers that depend on the value of s. However, the number s is given in terms of those positive coefficients by the quadratic formula (8) poe (=z? + ./c*|z\* — 4B). 2A Whether the radical here is real or imaginary, it follows from formula (8) that the real part ofs is negative. This completes the proof that the oscillation X(t) is damped Thus the forced component of the vibration of the main mass m is eliminated by the system m,, k,, if the natural frequency of that system coincides with the frequency of the exciting force: ‘k Since all components of the vibration of m are then damped, that mass approaches a fixed position as t increases. This is the principle of the Frahm vibration absorber, which has been used in such practical appliances as electric hairclippers.! Note that, in * Den Hartog, J. P., “Mechanical Vibrations,” 4th ed., pp. 87 ff., 1956.ELEMENTARY APPLICATIONS (sec. 92 99 view of Eq. (9), the absorber is designed for a fixed frequency @ of the exciting force Fy sin ot By solving the above equations for x;(s), we can see that the mass m, has an undamped component of vibration of the type (3). PROBLEMS 1. If the initial conditions X(0) = X'(0) = 0 used in Sec. 31 for the damped oscillator with exciting force F(t) are replaced by the conditions X(0) = Xo, X’(0) = v9, show that the additional terms in the formula for X(¢) represent damped oscillations. 2. When F(t) = Fod(t) and X(0) = X'(0) = 0 for the damped oscillator considered in Sec. 31, where c? < 4km, show formally that in o,t b= 5.07 - Fe _ pl x0 = 78 ov], m and note that the jump in the momentum of the mass m at the instant t = 0 is Fo. 3. Show that formula (6), Sec. 31, for the amplitude A of the forced component of vibration for the damped oscillator is valid (a) when the exciting force F(t) is Fy cos wt; (b) when F() = Fy sin (ot + 4. If c* = 2km for the damped oscillator considered in Sec. 31, and if F(t) = Fy sin wt and X(0) = X"(0) = 0, show that X( = Fotos 6, cos (ot + 6,) % cos 8, cos (bt + 6)) where 2b = c/m and @, and 6, are these arguments of complex numbers: 6, = arg(—2bw + Ho? — 267), 0, = arg(2b? — ia) Hence when q is small and ot is large, show that X(t) is approximately (Fy/k) sin a. 5. In Sec. 31, if c? = 4km, F(t) = Fy sin wt, and X(0) = X"(0) = 0, show that X(0) has the form A cos (wt + a) + (B+ Coe, and when w is small that |] is near its upper bound |Fo|/k. 6. Let the oscillator described in Probs. 1 and 2, Sec. 30, for which k = $Ibjft and mg = }1b, be subject to a damping force —c¥(t) where ¢ = de lbjft/sec. (a) If the upper end of the spring is fixed so that Z(1) = 0, and if ¥(0) = yp and ¥(0) = 0, show that after about nine cycles of oscillation of the mass the amplitude of the displacements Y(0) is reduced to approximately 0.Ilyo|. (b) When the upper end is oscillated so that Z(t) = 29 sin wt, show that resonance occurs if @ = w, where w, = /159.5 radjsee, a frequency slightly less than 9 (cf. Sec. 31). (c) When @ = c, in part (b), show that the amplitude A of the undamped forced component of vibration exceeds 1229] 7. Let the two masses in the system treated in Sec. 32 have arbitrary initial displace- ments and velocities. When condition (9) is satisfied by the elements of the absorber, show that the vibration of the main mass again contains no undamped component. 8. (a) If the exciting force Fy sin wt in the system of Sec. 32 is replaced by the force Fo sin (wt + a), where « is any constant, show that when condition (9) is satisfied the vibration of m is again entirely damped.100 sec. 32) OPERATIONAL MATHEMATICS (b) [the exciting force is replaced by A, sin,t + A, sin o2¢, where the A’s and 's are constants and «, % 2, show that the values of m, and k, cannot be adjusted so that all undamped vibrations of m are absorbed. 9. Let the force F(t) be removed from the mass m in Fig. 20 and let the end A of the spring k be moved horizontally so that its distance from the wall is F(0)/k. Show that the differential equations of motion of m and m, are then the same as in the original problem 10. Assystem of two masses and two springs in series, with no damping, is shown in Fig. 21. The weights of the two masses are m,g = 8 1b and m,g = 21b (g = 32 ft/sec’). ‘The spring constants have the values k, = 24 1bjft and k, = 8 Ibjft. If periodic force F = Focos 2nct acts on ma, find the frequencies c (cycles per second) of the force for which resonance occurs in the vibration of m,. Assume zero initial displacements and velocities. Also show that resonance occurs in the vibration of m, for those same values ofc. Ans. c= 4x7, 4/3 x7" eycles/sec. hy ke fm pw 9. 21 11. For the system shown in Fig. 21, ifm, = my = 7s, ky = + ky = fs, and if both ‘masses start from rest in their equilibrium positions, find the formula for the displace- ment X,(t) of m, when the force F(t) on m, is arbitrary. Ans. X,(0) = (6sint — \/6sint ,/6) * Fie. 12. Initially the two masses shown in Fig. 22 are at rest and the spring has its natural Iength. Then a constant force F = Fo(t > 0)acts on m,. Ifthe system is free from damp- ing, find the formula for the displacement of m, and note that the displacement consists of a simple harmonic motion superimposed upon a uniformly accelerated motion. Also show that if Fy > 0, the spring is always compressed by an amount proportional to 1 — coseot, where «? = k/m, + k/m, Ans. 2o*(m, + m,)X (0) = Fo(w?t? — 2 + 2cos wt). +X X_ Fig. 22 13. If the force F = 0 in the system shown in Fig. 22 and if the masses are initially released at rest from positions X, =a, and X; = az, show that the frequency of vibration of the masses has the value w = (k/m, + kim)’. When a; = —a,m,/m,, show that X, = a, cos ct and Xz = a; cost and describe the vibration. 14. Let the force in the system shown in Fig. 22 be the periodic force F = Fy sin wt, where o? = k/m,,and let a viscous damping force —cX;(1) act on m,. Find the steady- state vibration of m, and note that it does not depend onc. Ans. —Fok™! sino. 15. In Fig. 23 the end E of the first spring has a periodic displacement Y = A sin wr. ‘The two springs are identical. Find the value of « for which this damped system is in resonance, when c? < 4km, Ans. 2m?@o? = 4km —ELEMENTARY APPLICATIONS [sec. 32 101 Z| m h 2 ty Ly In Fig. 24 the two masses are unit masses (m = 1) and the two springs are identical. Initially, the springs have their natural length, the first mass has velocity vp, and the second mass is at rest. Find the undamped component of vibration of those masses. Ans. Xv9//k) sin t \/k. 17. For the undamped system of two equal masses and three identical springs shown in Fig. 25, the initial conditions are X,(0) = a, X2(0) = a2, X\(0) = X40) = (@) If |a,| 4 |a,l, show that the components of vibration of each mass have frequencies Jkim and \/3kj/m. (b) If a, = a;, show that both masses vibrate in unison with frequency ,/k/m. (c) Ifa, = —a,, show that the masses vibrate in opposite directions with frequency /3k/m. x m bY Fig. 25 tex, LX 18. A damped absorber Figure 26 shows an absorber with elements m,, k,, and c in which the damping is located in the absorber itself, When a force Fy sin ct acts on the main mass m, the undamped vibrations of m cannot be completely absorbed here, but the coefficient of damping c can be adjusted to give an optimum range of the ampli- tudes of those vibrations. Let mand m, beinitially at rest in their positions of equi Derive the formula Fo ms? + 05 + ky —— i est MO = Fy ima? bes + kilns? + cok TR) STR for the transform of the displacement X(t) of m, then show that the amplitude A of the forced component A cos (wt + 6) of X(t) is given by the formula A? (mo? — k,P + 2a? Fo? [(ma? — ky(m,o? — ky) — m,k,o7F + 70"[m + m)o* — KP (A graphical examination of A? as a function of @* would indicate a value of ¢ for which the range of values of A? will be as small as possible for all «. ‘ee Den Hartog, op. cit, pp. 93 ff, for a detailed discussion.102 SEC. 93) OPERATIONAL MATHEMATICS Pp sin wt m o ey { 7 On m ‘X Fig. 26 33. ELECTRIC CIRCUITS In Sec, 28 we used Kirchhoff’s voltage law: for each closed circuit in an electrical network the impressed voltage equals the sum of the voltage drops across the elements in that circuit. His law of currents states that at each junction point of branches of a network the total current into the point equals the total current away from the point. Let us now present additional applications of these laws for instantaneous behavior of voltage and current, in setting up differential equations for networks. An electrical analog of the Frahm vibration absorber discussed in Sec. 32 is indicated in Fig. 27, where the impressed voltage is V=Wsinar. According to the law of currents, applied at the junction P, a) 1) = 1) + 10. We apply the voltage law to the circuit on the left to see that 2) V=LI(t) + Ri) + £00 + F200 where Q'(t) = I(t) and Q(t) = 1,(¢); then to the circuit on the right to see that (3) Q,(t) = LiL. Fig. 27ELEMENTARY APPLICATIONS [sec 33 103 Let Q,(t) denote the difference between quantities of charge on the two capacitors, so that @ 20 = OH) — Qi), O20) = HO) — 1) = In terms of Q and Q, Eqs. (2) and (3) take the form A(t). LOW = ~ 200) + L109 ~ Bo) ~ ROW + Yosino, 6) i 1,03) = — C1220) — Q0). Except for differences in notation the system (5) of differential equations in Q(t) and Q,(t)is precisely the system written in Sec. 32 for the displacements X(f) and X (0) of the masses m and m, in the absorber (Fig. 20), Since all the electrical coefficients here, R, L, etc., are positive constants, the signs of the coefficients in Eqs. (5) match those in the equations for X and X,. The initial conditions for the network match those chosen for the absorber, and the analogy is complete if (6) (0) = 2,(0) = 1(0) = 1,0) = 0. Under conditions (6) the transforms of I(t) and 1,(t) are sq(s) and sq3(s), which correspond to sx(s) and sx,(s) for the absorber. According to the results found in Sec. 32 therefore, the current i(t) has an undamped com- ponent of type B cos (ct + a) unless 1 = LC, But when the elements of the LC, circuit satisfy condition (7) for a prescribed frequency @ of the impressed voltage V, the current I(t) contains only damped components; thus I(t) 0 as t-> 00. The currents I,(t) and I,(¢) in the L,C circuit, however, do have undamped components of type B cos (wt + @). Note that Eqs. (1) to (3) with conditions (6) can be solved directly for the currents either by replacing Q(¢) by J I(z) de or q(s) by i(s)/s, etc. Thus the transformations of Eqs. (2) and (3) gives the equations 2 ra) oo. Woot (1s +R+ aie + est) = os), (8) fi Lysi(s) — (us + ai =0. Upon eliminating i,(s), we find that (Lys? + C, (9) i(s) = of) ify) = mS)108 SEC. 33) OPERATIONAL MATHEMATICS where p(s) is the polynomial introduced in Sec. 32: is) = (Lys? + C,~4)(Ls? + Rs + C784 CV — C7, The function y(s) depends only upon the characteristics of the network, not upon the impressed voltage. Since the formal transform of the unit impulse symbol 4(¢) is unity, it follows from Eq. (9) that »(s) = i(s) when V(t) = 5(¢) formally; that is, Y(t) represents the current J produced by the voltage d(t). When the network is considered as a system with input V(t) and output J(t), then, in the language of systems analysis, y(s) is the transfer function for the system; in view of Eq. (9) the output corresponding to any input V(0) is given by the formula (10) I(t) = Vio)* ¥(0. PROBLEMS 1. If 1,(0) = 1,(0) = 0 in the network shown in Fig. 28, while the capacitors have the same initial charge Qo, show that 10 = 1) = -Q.u sin ot, where @ = (CL)~*, Qi, Qa IK Fig. 28 2. If1,(0) = 1,(0) = Ointhenetwork shownin Fig. 28 and ifQ,(0) = a, and Q,(0) = a, find the undamped component /,(¢) of the current I, (0), the steady-state current through the first inductance coil Ans. 1, = —Ka, + a3) sin wt, @ = (CL)*. 3. Show that the network indicated in Fig. 28 represents an electrical analog of the ‘mechanical system shown in Fig. 24 where Q, and Q, correspond to the displacements of the two masses. 4. In Fig. 29 the currents 1,(2) and I(t) and the charge Q(t) are initially zero. (a) Show that the current ,(t) is unstable under the impressed voltage V = Vosin (ut + a) io? = (L,C)"* + (L307 (b) Show that the system indicated in Fig. 22 is a mechanical analog to the network here with X ,() and X(t) corresponding to Q,(t) and Q,(t), where Q(t) = 1,(0), 920) = 1,{0), and Q() = Q.(0) — Q.(0). Fig. 29ELEMENTARY APPLICATIONS [SEC 33 105 5. A prescribed current (t) is supplied to the network shown in Fig. 30. The initial values of f(t) and Q(t) are zero. Use the following notation: V(t) is the resulting potential drop from the junction A to the junction B, «? = (LC)~', 2b =(RC)"!, and 4? = eo? — b?, given that B® < @9?. (a) Show that the transfer function from input [(t) to output 1,(¢) is eg? 7 Dbs + ag? vs) thus that /,(0) = I(t) * ¥(#), where ¥(0) = c9?0,”1e™™ sin ont. (b) Find V(2) when I(t) = Io,a constant. Ans. V(t) fo(Cov,)'e-™ sin wt. Fig. 30 6. Inasimple LC circuit with inductance coil and capacitor in series, let the impressed voltage be a periodic function, with period T, of the type x V(t) = a9 + ¥ (a, cos not + b, sin net), where = 2n/T. Show that the current I(t) in the circuit is unstable if the frequency of V(t) has any one of the values 1 °* m JIE provided that a,, and b,, are not both zero. 7. In the network shown in Fig. 31 all currents and charges are zero at the instant 1 = 0 when the switch K is closed. If the impressed voltage is given by the equation V = VYosinwt, where w = 2(LC)~', find the steady-state value 7,(t) of the current 1,(0) and note that it is independent of the value of the resistance R. Ans. 13(t) = —VoCo cos ot. Q & pe] wht eat bt tt les R c ee Fig. 31 8. In the system shown in Fig. 25 apply an exciting force to the first mass and viscous damping to the second mass. Show that the resulting system is a mechanical analog, of the network shown in Fig. 31.106 SEC. 34) OPERATIONAL MATHEMATICS 9. Initially the currents and the charge on the capacitor are zero in the network shown. in Fig. 32. The voltage drop from 4, to B, caused by the mutual inductance of the two coils is M1,(2), and the drop from A, to By caused by that mutual inductance is ‘MI;(t),where M? < LL. Theimpressed voltageis given by the equation V = Vocoswut. When the values of C and L are adjusted so that L,C = ”?, show that the current 1,(t) in this idealized resistance-free circuit has a simple periodic variation at all times with a frequency greater than «. Ans. 1,(0 =VC2o, sina, 0 Ly Fig. 32 10. Show that the network in Fig. 33 represents an analog of the damped absorber shown in Fig. 26 if initial currents and charges are zero, where charges Q and Q — Q, correspond to displacements X and X, and the voltage V corresponds to the force Fo sin oot c 1 A PG ety i Fig. 33 34. EVALUATION OF INTEGRALS Laplace transforms can be used to evaluate some types of integrals containing a parameter. The manipulative procedure is often direct and simple, but, as, the examples and exercises here will indicate, proofs of the reliability of those formal results can be challenging. In our proofs we use this special case of the theorem on uniqueness stated in Sec. 6: A function f(s) cannot have more than one inverse transform F(t) that is continuous over the half line ¢ = 0 and of exponential order. Example 1 Evaluate the integral a) rio = f lo +aELEMENTARY APPLICATIONS [sec 34107 Here F(t) is the Fourier cosine transform (Chap. 10) of (x? + a?)-'. Note that F is bounded: |F(0)) $f (x? + a2)! dx = m/\2a): also, F(-0) = Fo. Let us first proceed formally to transform with respect to t and interchange the order of integration with respect to x and t: = 3 e sdx form [Leos ha a= [Praia Ts) f 1 1 een 1 ~@ Jo (Ha HS] Dasta ifa > Oand s > 0. Thus the integral has the value 8) Fo) = ze (@>0r20. To see that the formal step (2) is sound, we first note that, whenever T and s are positive constants, 7 pw = 7 a f ef Sets axat = f =a “cos tx dt dx, lo do o +a lo +a? because the absolute value of the entire integrand does not exceed (x? + a?)"', a function independent of ¢ whose integral from x = 0 to x = 0 exists, The integral F(t}e~™ therefore converges uniformly with respect to ¢ when t = 0 by the Weierstrass test and (Sec. 15) the interchange of order of integration with respect to x and t made in step (4) is valid. Also, F iscontinuous(t = 0) and bounded so, according to the theorem on uniqueness, if L{F} is given by formula (2), then F must be the exponential function (3). The final integral in Eq. (4) can be evaluated by integration by parts, and the equation can be written 1 0 (5) f e*F(t) dt = f g(x,T) dx, 0 0 e-*(x sin Tx — scos Tx) + i ee EE Jeb ean) where aT) sare sy This continuous function of x and T has a limit g(x,00), B(X,00) = Jim g(x,T) = s(x? + a?) (x? + 52)73, uniformly with respect to x when x > 0 since x+s > S Me™” @raerrsy se where M is the maximum value of the quotient of polynomials in x. la(x,T) — g(x,00)] < e-*7108 SEC. 34] OPERATIONAL MATHEMATICS To ach positive number e there corresponds a number T,, independent of x, such that Me-*7 < « when T > T,. Moreover, the second integral in Eq. (5) converges uniformly with respect to T, according to the Weierstrass test. The limit of the integral (5) as T— 00 is therefore J8 9(x,00) dx (Prob. 10, Sec. 27), since the latter integral exists; thus Eq. (2) follows from Eq. (4). Example 2 Evaluate the integral al Ft) = fr sin tx ay lox The formal procedure is simple. When t > 0 and s > 0 Oh f(s) = f L{sin tx} = i a” x x +s hence F(t) = 1/2. In view of Eq. (6), F(t) is an odd function: F(—t) = — F(t), and F(0) = 0. Therefore (8) FQ= Nia (>; FQ)=-F (<0; FOO) =0. The Weierstrass test cannot be used here to justify step (7). In fact, if our formal result (8) is correct, then F is discontinuous at the origin ¢ = 0, and the integral (6) cannot converge uniformly over an interval that includes the origin. Now the function S(r) = sin r/r, r # 0, S(0) = 1, that is, 0 Ser) (-~w 1, then |sinr/r| < 1, To see that the integral (6) exists, we note that its integrand 15(tx) is continuous in t and x and, if to, Xo, and x, are positive constants, an integration by parts shows that “sine _ cOStxy _costx, — f* costx f dx = 2 _ sos _ de tt) Io X 1X0 xy ko x and this has a limit as x, + 00. Thus F(t) exists because “osintx costxy [cost ro f dx +228 of ax (6>0) lox 1% dey and because F(0) = 0 and F(~1) = — F(t).ELEMENTARY APPLICATIONS [SEC 34 108 When t > 0, the substitution r = rx shows that our integral has a constant value, F(t)={% r-'sinrdr. Hence F is bounded, con- tinuous when t > 0, and F(+0) exists. If we establish formula (7) for ‘Jf(s), then F must be the step function (8). ‘The remainder for integral (6) can be written dx sin tx costX — (* cos tx (eX) fi x x I 1x? and we can see that, when t 2 to > 0, for each positive number ¢ there is a number X, independent ct such that 1.2 ata Six when X > X.. IRE Saget eg S « Thus integral (6) converges ‘vifoemty with respect to t (¢ = to > 0), and we can write, with the aid of an elementary integration, [fernia= [Eff orsnecaas 0 10 X Jig = fo nrwax - jf bltoxx) dx where h is this continuous function st sin tx. hx) = — al" a + cos | (s>0). ee x The integral f@ h(to,x)dx converges uniformly with respect to to(0 S to < 1) according to the Weierstrass test, so it is continuous in tg at the point tf) = 0; thus (10) fleceroa = fuer ax + f a (>, 0 lo Jo XP +? Let N(s) denote the maximum value of (sT + I)e~*7. Then ay Wr s POG and ito s Pew The uniform convergence of the integral f' h(T,x) dx and the fact that N(T,x) + 0 as T+ 00, uniformly with respect to x, follow from con- ditions (11). Thus as T+ oo, that integral vanishes, and Eq. (10) leads to step (7), o the bounded continuous function F has the transform 4n/s, and therefore F(t) = 42 when t > 0. This completes the proof. Incidentally, we have shown that Si(oo) = 4x, for when ¢ > 0, we found that F(e) = 4x, and Eq. (6) can be written sinr a2) F() = SOT dr = Si(oo) = 7110 SEC. 38) OPERATIONAL MATHEMATICS 35. EXPONENTIAL- AND COSINE-INTEGRAL FUNCTIONS Convenient forms of the exponential-integral function are () £0 = [Say = fax (> 0). x Formally, the transform of the second integral is me dx refi 1 1 x * a Patol (e-ee) pees | S>o. 1 The proof that expression (2) represents L{E,(t)} is left to the problems. The basic form of transform 100, Appendix A, Table A.2, is, therefore, ) L{E,(0} = Llog(s +) (>) In Sec. 22 we found that the transform of the sine-integral function, Sit, is s~" arccot s. Now let us write the transform of the cosine-integral function (4) cits -f OST -f SOS iy (e> 0). | or 1 x An integration by parts, first over a bounded interval (1,X), shows that 6) cir = EL? sintx (t> 0); thx thus Ci is continuous when t > 0, and Ci (co) = 0. The procedure used in Example 2, Sec. 34, can be applied to the final integral in formula (4) to prove that © L{Cign=— sf a 1 Dees 3, losts* +1) (s > 0). Next let us derive the transform (No. 98, Appendix A, Table A.2) of the function cos t Sit — sin t Cit which can be written * sin r . cosr cost far + sint f° ae or t r sin r a _ dr =cost| “dr — | (sinrcost — cos, sin t)— o © it r when t > 0. By replacing Si(oo) here by 47 (Sec. 34) and substituting tx for r — t, we can write the last expression in the form x © sin(n— 1) x © sin tx 0 cost [ar = Foose — fa,ELEMENTARY APPLICATIONS When we transform that final form (7), we find that [sec 38111 (s > 0). (s > 0). (x > 0). (s> 0). (s>0). ae logs (8) L{cos Sit — sin Ci} = Sp In like manner we can derive the transformation (No. 99, Appendix A, Table A.2) . slogs ) L{cost Cit + sine Sit} = — PEt Finally, let us evaluate the Laplace integral (19) acs) = f 0 Its transform with respect to x is found to be _m os logs 8-32 yT Fei In view of formula (8) we have the integration formula (ly g(x) = (bt — Six)cosx + Cixsinx, which is transformation 120, Appendix A, table A.2: 1 1 Ls =(tn- is si (12) {to (= si] coss + Cissins PROBLEMS Establish the integration formulas in Probs. 1 to 3, where r 2 0. = sin tx mo f Weep ite. the: oe sine transform of x~ (x? + 1)" (Chap. 13). = f. Tera lo ~ a c "19-0, a Obtain formally the integration formulas in Probs. 4 to 7. ‘ 4 fT sib ae ne ge 5. | — nar = net erk lo e+e ete Js (cf. transformation L11, Appendix A, Table A.2.) (a> 0,t > 0) (x 2 0),112 SEC. 35) OPERATIONAL MATHEMATICS oP = : 6. j. sy aemeral sineytay = (5) (> 0), [Note that y* + s? = (9? + 3)? — 2sy? = ° 2) _l fk 7. [ept-oreosanae= f/f 8. When T > fo > Oand s > 0, prove that fle" [aw _ pa + a “2 then prove that the limit of that integral as f + 0 and T + oo is the integral (2), Sec. 35, which represents L{E,(0}. 9. Use the transformation (3), Sec. 35, to show that, when a > 0, I I L{e*E (at) = 285 — 084 sa /28 + 8)? + v/2s + 9) ‘ (t > 0), (5 > a; cf. transformation 97, Appendix A, Table A.2) Obtain formally the transforms given in Probs. 10 to 12, if a> 0 and b>0 (see Prob. 9). 10. i : aro u }- Jo ta tra (cf. transformation 118, Appendix A, Table A.2.) b " es aye + Bi 12, ea 13. Prove that the remainder for the integral Cit = — Jf x7? cos tx dx satisfies the condition |R(t,X)| < 2/(toX) when t = fy > 0 and hence that the integral converges uniformly when ¢ 2 fo. Then complete the proof of transformation (6), Sec. 35. When a > 0 and t > 0, obtain formally the Fourier transformations of (x + a)" stated in Probs. 14 and 15, 14. f sin gy = (¥ — siat xe (2 E ,(as) ' = &E\(bs) — eE (as) (ab). = ase"E (as) (cf. transformation 119, Appendix A, Table A.2.) 0s at + Ciatsin at. "© COS tx |) x+a 15. dx = (5 - siat) sin at — Ci at cos at. 16. Complete the derivation of formula (11), Sec. 35. 17. Obtain formally the transformation (9), See. 35. 18. The beta function is defined as follows: Bixy) = f Pod nyt dr (> 0y>0.ELEMENTARY APPLICATIONS [sec. 36 113 Since Le? = Mixjs™*, note that TOMO) gat (ety) ee J Write the convolution here in terms of the integral representing B(x,y) to show that the beta function can be written in terms of gamma functions (Sec. 5) as Let ee} = PegP ys? (> Oy > 0). 36. STATIC DEFLECTION OF BEAMS Let ¥(x) denote the static transverse displacement of a point at distance x from one end ofa uniform beam, caused by a load distributed in any manner along the beam (Fig. 34). Under certain idealizing assumptions, primarily that displacements Y(x) and slopes ¥'(x) are small, it is shown in mechanics that the internal bending moment M(x) exerted by any span of the beam upon an adjacent span, through their common cross section A,, is propor- tional to the curvature of the beam at position x. In fact (ql) M(x) = EIY"(x), where E is Young’s modulus, the modulus of elasticity in tension and com- pression for the material, and J is the moment of inertia of the area A, with respect to the neutral axis of the cross section, the line in A, about which A, turns when the beam bends. Note that ¥"(x) is approximately the curvature since ¥'(x) is small. If F(x) denotes the internal shearing force at A,, it can be seen that F(x) dx = dM(x); that is (Prob. 14, Sec. 39), F(x) = M'(x), of Q) F(x) = ELY"(x) Let W(x) represent the transverse load per unit length along the beam. Then W(x) dx = dF(x), and it follows from Eq, (2) that (G3) YA%x) = aW(x) = El). ze wa2e Fig. 34114 sec. 36) OPERATIONAL MATHEMATICS The shear F(x) is a continuous function except at points where con- centrated loads or supports act. When W(x) is sectionally continuous then, in view of Eqs. (2) and (3), Y"(x) is continuous and Y‘(x) is sectionally continuous. These are just the continuity conditions that are implied by our formula for the transform, with respect to x, of ¥(x) Although the simple differential Eq. (3) can be solved by successive integrations, the requirement that the solution Y(x) and its derivatives of the first three orders be continuous at all points often involves tedious labor when H(x) is sectionally continuous. The simplicity of Eq. (3) enables us to use the Laplace transformation even though that transformation is not adapted to the two-point boundary conditions which will accompany that equation. Certain Fourier transformations specified by the boundary con- ditions and the differential form d*Y/dx* are properly adapted to such problems (Chap. 11). Let us determine the displacements ¥(x) in a beam whose end x = Ois built into a rigid support while the end x = 2c is free, or unsupported (Fig. 34), The load W(x) per unit length is zero over the span 0< x 0, together with the first pair of boundary conditions (5). In view of the continuity conditions to be satisfied by ¥(x) and its derivatives, we may expect that s*y(s) — s°(0) — 2(0) — s¥"(0) — ¥"(0) = amyre®, We have used a convenient extension, awoS,(x) when x > 2c, of the function aW(x) in the right-hand member of Eq. (4) where 0 < x < 2c. The par- ticular extension used is immaterial because we first seek a function ¥(x) that satisfies Eq. (4) when 0 < x < 2c and conditions ¥(0) = Y‘(0) = 0, and contains two arbitrary constants A = ¥"(0) and B = Y"(0). Since AB 1 Ys) = 3 + G+ awoze“, a function that satisfies those requirements is 6) ¥(x) = 44x? + £Bx? + Ahawolx — c)*S,(x).ELEMENTARY APPLICATIONS [sec 37115 When x > ¢, it follows from Eq. (6) that Y"(x) = A + Bx + fawo(x — ¢)?, ¥"(x) = B+ awo(x — o). The last two of conditions (5) are then satisfied if a A=4awye?, B= —awge, and the displacements in the beam are given by the equation (8) ¥ (x) = awelfe?x? — dex? + A(x — 9S.09] 0S x S20). This function satisfies Eqs. (4) and (5) and the continuity conditions. Since ¥"(0) = B = —awoc, it follows from Eq, (2) that the shear at x = 0 has the value —woc, or F(0) = —W as we should expect, since the magnitude of the vertical force exerted by the support must be the same as the total load on the beam. Actually, the displacements ¥(x) for this can- tilever beam can be found by first noting that the shear — Y"(x)/a at each section is the total load on the span to the right of that section. The bending moment at the end x = 0 is ¥"(0)/a or A/a: M(O) = 3cW% (Wo = ewo). If the end x = 2c were pin-supported (hinged or simply supported) so that it can rotate freely about a fixed axis, then the conditions at that end become ¥(2c) = ¥"Q2c) = 0. 37. THE TAUTOCHRONE We shall now discuss a problem in mechanics that leads to a simple integral equation of the convolution type The problem is that of determining a curve through the origin in a vertical xy plane such that the time required for a particle to slide down the curve to the origin is independent of the starting position. The particle slides freely from rest under the action of its weight and the reaction of the curve on which it is constrained to move. The required curve is called the tautochrone. Let o denote the length of arc of the curve, measured from the origin 0, and let (x,y) be the starting point and (é,7) any intermediate point (Fig. 35). Equating the gain in kinetic energy to the loss of potential energy, we have 1 [do\? _ aa) = may ~ 0) where m is the mass of the particle and t is time. Thus do = —/2g,/y — ndt,116 sec. 37) OPERATIONAL MATHEMATICS Fig. 35 and upon separating variables and integrating from n = y ton = 0, we have where T is the fixed time of descent. — o = Hn), where the function H(y) depends upon the curve, and therefore ’ a, TS -f & = 1) 4H) dn. ° This is an integral equation of convolution type in the unknown function H'(y), We may write it in the form 2g = y+ * Hy) Let h(s) be the Laplace transform of H()) with respect to the variable y. Since H(0) = 0, it follows formally from Eq. (1) that es = = swane{ 7 she = oms),/* That is, sh(s) = 1/3, hence T 1 a HU) = — Je avy We can see that this function does satisfy our integral equation (1) by sub- stituting it into that equation and performing the integration. Since Hy) =2ELEMENTARY APPLICATIONS [sec 98117 the differential equation of the curve in terms of the variables x and y is, according to Eq. (2), (a) 5 where a = 2gT*/n*. Separating variables here we have dx = [7 ay, and the necessary integration can be performed easily by substituting y = asin? 46, for we then find that dx = acos? 3040 5 sul + cos 6) d0. Noting that x = 0 when y = 0, we see that the parametric equations of the tautochrone are therefore 8) x 5 +sin0), y sa — cos 6) These equations represent the cycloid generated by a point P on a circle of radius $a as the circle rolls along the lower side of the line y = a The parameter 6 is the angle through which the radius drawn to the point P has turned, where the initial position of P is at the origin. Our tautochrone is of course just one arch of this cycloid. Since a = 2¢T/n2, the diameter of the generating circle is determined by the time T of descent. The above problem can be generalized in various ways $0 as to lead to other interesting questions; in fact, it was a generalization of the problem of the tautochrone that led the great Norwegian mathematician Niels Abel (1802-1829) to introduce the subject of integral equations. If the time T of descent is a function F(y), for example, our integral equation (1) becomes » @) V28F(y) = f (y = 14H) dn. 0 38. SERVOMECHANISMS Simple integral equations as well as differential equations arise in the theory of automatic control. As a special case we consider servomechanisms that force the angle of turn @o(t) of a rotating shaft to follow closely the angle of turn ©(t) of a pointer or indicator, where t denotes time. The shaft and ' See Bécher, M., “Integral Equations,” p. 6, 1909,118 SEC. 38) OPERATIONAL MATHEMATICS material rigidity attached to it have a total moment of inertia J that is much greater than that of the pointer. Mechanisms of that general type are used, for instance, in directing antiaircraft guns and in aircraft control. Let ®(1) be the angle of deviation between shaft and pointer or the difference between output angle and input angle: qd) Ot) = Oo(t) — OO). An auxiliary system or servomechanism can be designed to measure ®(¢) and feed back to the shaft a component of torque that is proportional to the deviation (1). The servomechanism may contain its own source of power, motors, generators, and other electrical equipment. In order to provide damping in the system, let the servo also supply a component of torque proportional to the rate of deviation ®(f). Then, since the product of I by the angular acceleration of the shaft equals the torque applied to the shaft, 2) 106(t) = —k®(1) — co), where k and c are positive constants. If the shaft is initially at rest and its angle of tun is measured from the initial position, then ©,(0) = ©4(0) = 0. Also, in view of Eq. (1), (0) = — (0) and in terms of transforms Eq. (2) becomes Is?89(s) = —(k + c)$(s) — cO,0), where we have assumed that ©o(t), ©o(¢), and (2), and therefore ©,(0), are continuous when t = 0. Since @o(s) = $(s) + 0,3), it follows that _ 1370 (3) + (0) 3) o ) 9) Ist +s +k Under an input ©,@) = Ar, it follows from Eq. (3) that ~ A gw =£ ke oH) = -Ae™sinor [b= 5.0% =F gp If k > c2/(40), the angle of deviation (0) has a damped oscillation with initial value zero. Since |sin «t/(ct)| < 1 and (be)~! is the maximum value of te, that oscillation is small at all times when b is large, because Al [®() < |Alte"™ s be If in addition to the two components of torque shown in Eq. (2), a component proportional to the accumulated angle of deviation is produced by the servo, then @ 10419 = ~ 040 ~ co ~ 6 f " ae) de, :ELEMENTARY APPLICATIONS [sec 39119 where b is a positive constant. When ©¢(0) = ©4(0) = 0 therefore, Bb Is?[ G(s) + 0(s)] = — (« tes + ‘| (s) — cO{0). In terms of positive numbers B, C, and K, where IC=c, IK?=k, 1B the last equation can be written _ 8°0{s) + CsO,(0) i) 9) = FE Ca RK + BF In the special case @{t) = At, K? = BC, Eq, (5) becomes As (s + B)[s? + (C — B)s + B?] and since the polynomials in the numerator and denominator here are of degrees one and three in s, respectively, it follows from our earlier observa- tions (see the problems, Sec. 27) that (0) = 0; that is, the initial value of the angle of deviation is zero. When C > B, all values of s that make the poly- nomial in the denominator vanish have negative real parts, and con- sequently (Sec. 26) ®(t) contains only damped components. But when C < B, ®t) has an unstable oscillation. 6) 9) = — 39. MORTALITY OF EQUIPMENT Let the function F(1) denote the number of pieces of equipment on hand at time t, where the number is large enough that we can consider it as a con- tinuous variable instead of a variable that takes on only integral values The equipment wears out in time, or is lost from service for other reasons, so that, out of mo pieces of new equipment introduced at time t = 0, the number N(1) in service at time t is given by the formula (1) N(1) = noH(0, where H(i) is a function that determines the surviving equipment after t units of time. Note that H(0) = 1, necessarily. If R(z) is the total number of replacements up to time z, then R(c) de is the number of replacements during the time interval from t= to t= 1+ dr and the number of survivals at any future time 1, out of these replacements, is R(OH(t — de The total amount of equipment in service at time ¢ is the sum of these survivals from the replacements during every time interval dr between120 SEC. 39) OPERATIONAL MATHEMATICS + = Oand r = 1, increased of course by the survivals from the new equipment on hand at time ¢ = 0. Therefore Q) F() = FO)H() + f R(2)H( — 2) dr. We have assumed here that the equipment F(0) on hand at time ¢ = 0 is all new; then R(0) = 0. If the amount F(s) that must be in service at each instant is known and if the survival factor H(¢) is known, then Eq, (2) is an integral equation of convolution type in R'(). Its solution gives the formula by which replace- ments must be made. The equation is an integral equation in the survival factor H(e) when FO and R(®) are known In either case, the transformed equation is (3) f(s) = F(O)h(s) + s r(s)h(s). Then _ f(s) = FOA(s) (4) 19) =e and R(i) is the inverse transform of that function. Suppose the mortality is exponential in character so that H() =e and that the amount of equipment on hand is to be a constant, F() = b. Then h(s) = 1/(s + k) and f(s) = b/s, and it follows from Eq. (4) that 1 1(s) = bk Therefore replacements must be made at such a rate that the total equipment replaced up to time ¢ is R(o) = bkt, a result that is easily verified as the solution of Eq. (2). Thus replacements must be made at the rate of bk pieces per unit time PROBLEMS 1. In the example solved in Sec. 36 let the end x = 2c of the beam be pin-supported, rather than free, with no other changes in conditions. Find the vertical force exerted by the beam on the pin, and the vertical force and bending moment exerted on the support at x = 0. Ans. Sy: BWo 3 sycW.ELEMENTARY APPLICATIONS {sec 39121 2. Both ends x = 0 and x = 2c of a beam are pin-supported. Find the vertical force ‘on each support when a transverse load Wo is distributed uniformly over the span e 1, Show that the curve of descent is the line x = y,/B? — 7. Ie =k = 21 in Eq, (2), Sec. 38, find the output angle @4(¢) corresponding to the constant input angle @(t) = I(¢ > 0), and compare them graphically. Assume that ©,(0) = @{(0) = 0. Also, note that the value of the output lags behind that of the input until ¢ = 31/4. Ans. Oo(t) = 1 = /2e~* sin (¢ + 1/4), 8. For the servomechanism corresponding to Eqs. (4) and (5), Sec. 38, consider this. special case: (1) = At, B° = CK?. If a represents the argument of the complex number K + Ci, derive the formal 0) = a aigic He SCTE R? sin (Ke + @)] for the angle of deviation, and note the undamped component of ®(0). 9. Let the servomechanism discussed in Sec. 38 supply only a corrective torque proportional to the deviation angle (:) while the shaft itself is subject to a damping torque proportional to its angular velocity. If ©,(0) = ©,(0) = 0, show that is), 808) = Tae es + RO where c, k, and I are positive constants. When @{¢) = A, show that @,(t) approaches Aas t increases. 10. When H(t) = e~™, where H(¢) is the survival factor in Sec. 39 and k is a positive constant, (a) find the replacement function R(¢) corresponding to an arbitrary amount F(Q) of equipment on hand; (b) find R(¢) when F(t) = At + B. Ans. (6) RQ) = (A + BRyt + Sake 11. When H(t) in Sec. 39 is the step function 1 — 5,(¢) so that every piece of new equipment survives for k units of time, (a) show that the number of replacements R(é)122 SEC. 39) OPERATIONAL MATHEMATICS up to time ¢ required to maintain F(t) pieces in service at that time is RO) = FO) ~ F(+0) + F(¢ = k) + Flt = 2k) + Fle — 3k) + if we define F(t) to be zero when t < 0. (b) When F(t) = A(t > 0), show that R(t) = A(t/k], where [1] is the bracket symbol, and draw the graph of R(). 12. A particle mass m moves on a vertical X axis under two forces: the force of gravity and a resistance proportional to the velocity. If the axis is taken positive downward, the equation of motion is mX"(0) = mg — kX"). Show that its solution, under the conditions X(0) = 0, X'(0) = uo, is X() = Gallon — g)(1 — e°*) + beth where b = k/m, and discuss the motion. 13. Each one ofa set of radioactive elements E , E,, Es, and E, disintegrates into the succeeding one at a rate proportional to the number of atoms present, except for the end product E, which is a stable element. If N,(t), where i = 1, 2, 3, 4, denotes the number of atoms of element E, present at time ¢ and if the distinct positive constants 1, Cz, and c3 are the respective coefficients of decay of the first three elements, then Ni) = —eNilt, NG) = —e2N2{t) + eNO, N4() = —cyN3(t) + c2N3(0, — Nul0) = egN3I0) When only M atoms of E, are present initially, derive the formula eee" M (cz = exes = 1) (x = €2)Mes = €2) (ex = €3€2 = 3) 14, In Sec. 36 the shear F(x) is the internal force in the direction of the Y axis exerted at a cross section A, of the beam upon the span to the right of that section, and M(x) is the bending moment exerted on that span there. If F(0) = Fy, M(0) = Mo, and W(x) is the load per unit length, apply equilibrium conditions to a span extending x units to the right of the section Ag, at x = 0, to show that Fyt f WO dé — Flx) = 0, My + Fox + fe = QW) dé — Mex) = 0. Thus derive these relations between shear, load, and bending moment: F(x) = W(x), M(x) = F(X).4 Problems in Partial Differential Equations 40 THE WAVE EQUATION Several functions in physics and engineering satisfy the partial differential equation ey jy =a > co) a known as the wave equation in two independent variables x and 1. We shall use literal subscripts to indicate partial derivatives; then Eq. (1) can be written. Yast) = a? Yen). Brief derivations of this equation for some elementary physical functions will now be given. The derivations are helpful in writing modifications of the equation and in setting up boundary conditions for specific problems. 123124 SEC. 40} OPERATIONAL MATHEMATICS Fig. 36 First, let ¥(x,f) denote the displacement at time t away from the x axis of a point (x,Y) of a string in the x¥ plane under tension P (Fig. 36). The string is assumed to be flexible enough that all bending moments transmitted between its elements can be neglected ; thus each element pulls tangentially on an adjacent element with a force of magnitude P. Suppose further than conditions are such that the magnitude H of the x component of the tension remains essentially constant at all times for all points (x,Y); in particular all displacements Y(x,{) are assumed small compared with the length of the string. Finally let the slope angle « remain small in order that each small element of length of the string may be approximated by the length of its projection Ax on the x axis. All these idealizing assumptions are satisfied, for instance, by strings of musical instruments under ordinary conditions of The vertical component of tension is the vertical force V exerted by the part of the string to the left of point (x,Y) on the part to the right of the point. It is proportional to the slope of the string, — V/H = tan a (Fig. 36); that is, 2) V(x) = —HY,(x0). This is the basic formula for deriving the equation of motion. Now consider an element of length of the string whose projection on the x axis is Ax. If p denotes the mass per unit length, the mass of the element is approximately p Ax. Ifno external forces act on the string, the application of Newton’s second law to the element gives, in view of formula (2), GB) PAXY, (xt) = —HY (x,t) + HY + Ax, 0) approximately, when Ax is small; that is, Yeo) = HE Get Bet) = Yu) ? Ax When we let Ax approach zero, this becomes Eq. (1), where 4)PROBLEMS IN PARTIAL DIFFERENTIAL EQUATIONS [SEC 40 125 If in addition to the internal force a force F(x) per unit of mass acts in the ¥ direction along the string, then the additional term p Ax F(x,t) appears on the right in Eq, (3). The resulting modification of Eq. (I) is (3) Y(t) = a? Ye(x0) + Fd) When the Y axis points vertically upward and the weight of the string is to be taken into account, for instance, F(x,0) = —g, where g is the acceleration of gravity. As another physical example consider the longitudinal displacements in a cylindrical or prismatic elastic bar. The values of the variable x are marked on the bar so as to designate the cross section that is x units from cone end when the bar is neither stretched nor compressed (Fig. 37). For each value of x the longitudinal displacement Y(x,t) is measured from a fixed origin outside the bar, an origin in the plane occupied by the cross section at x when the bar is unstrained and in some position of reference. Thus, if the bar is moved lengthwise as a rigid body, ¥(x,) is a constant at each time Since Y(x + Ax, () is the displacement of the cross section at x + Ax, an element of the bar whose natural length is Ax is stretched by the amount ¥(x + Ax, 1) — (x,t) at time ¢, According to Hooke’s law the force exerted by the bar upon the left-hand end of the element to produce that extension is ¥(x + Ax,0) — Yess) ~ AE Ax where A is the area of the cross section and E is Young’s modulus of elasticity of the material, the modulus in tension and compression. When Ax tends to zero, it follows that the internal force from left to right at the cross section is, © F(x) = — AEY,(x,0. This basic formula corresponds to Eq. (2) for the string. Let p denote the mass of the material per unit volume. When we apply Newton's second law to an element of the bar, a pA AXY, (xt) = —AEY,(x,!) + AEY(x + Ax, 1), } ¥ix,t) Fig. 37126 SEC, 41) OPERATIONAL MATHEMATICS Fig. 38 we find as before that Y(x,t) satisfies Eq. (1), where (8) When the elastic bar is replaced by a column of air, Eq. (1) has further applications in acoustics.' Again, the function Y(x,:) may represent the angle of turn of a cross section x units from one end of an elastic cylindrical shaft under torsion (Fig. 38). Let [ denote the moment of inertia of the cross section with respect to its axis, E, the modulus of elasticity of the material in shear, and p the density of the material. Then, by steps analogous to those used above, we find that the internal torque t acting through a cross section at position x is (9) a(x,t) = —1E,Y,(x,t) and that Eq. (1) is satisfied by the angle ¥(x,t), where (10) Finally, it should be remarked that Eq, (1) is a special case of the telegraph equation (11) ¥, 4050) = KLY,(x,0) + (RK + SL)Y(x,t) + RSY(x,1), where Y(x,0) represents either the electric potential or the current at time ¢at a point x units from one end of a transmission line or cable? Here the elements of resistance, inductance, etc., are distributed along the cable or through the circuit, in contrast to the lumped elements in elementary circuits (Sec. 33) that lead to ordinary differential equations in the currents. The cable has resistance R, electrostatic capacity K, leakage conductance S, and self-inductance L, all per unit length. When R and S are so small that their effect can be neglected, Eq. (11) reduces to Eq, (1), where a? = (KL)~* 41 DISPLACEMENTS IN A LONG STRING Let (x,t) represent the transverse displacements of the points of a semi- infinite stretched string, a string having one end fixed so far out on the x axis * Lord Rayleigh, “Theory of Sound,” vols, I and 2, Dover, 1945. *A derivation of Eq. (11) is outlined in Prob. 6, Sec. 93.PROBLEMS IN PARTIAL DIFFERENTIAL EQUATIONS [sec 41127 that the end may be considered infinitely far from the origin, and having its other end looped around the Y axis. The loop, initially at the origin, is moved in some prescribed manner along the ¥ axis (Fig. 39) so that Y = F(t) when x = 0 and ¢ > 0, where F(t) is a prescribed continuous function and F(O) = 0. If the string is initially at rest on the x axis, let us find the formula for ¥(x,!) The above conditions on ¥(x,t) can be written () Yulx,t) = a? You(X,0) (x >0,¢>0), Q) ¥(x,0) = ¥(x,0) = 0 (x > 0), 0) YO.) = FQ, lim ¥(x,) =0 (120), where a? = H/p (Sec. 40). The equation of motion (1) implies that no external forces act along the string. A problem composed of such conditions is called a boundary value problem in partial differential equations. We shall use a formal procedure to solve the problem and then show how our result can be verified as a solution, If y(xs) is the Laplace transform of ¥(x,t), then, in view of the initial conditions (2), L{Y,,} = s*y. Also, ca ge al [e-“¥(x,0)] dt Urbs} = f 0 are, = all eo Y(x,0) dt = ys(oo5), if the function e~“Y¥(x,t) satisfies conditions under which the indicated interchange of order of integration with respect to 1 and differentiation with respect to x is valid (Sec. 15). When both members of the partial differential equation (1) are transformed and conditions (2) are used, we therefore obtain the equation s*y = a?y,, in the transform of our unknown function. From conditions (3) we find that y(0,s) = f(s), where f(s) is the trans- form of F(9), and lim,..,, y(x,s) = 0, provided that the order of integrating with respect to rand taking the limit as x + oo can be interchanged. The Fig. 39