Printed niul Published by 'THE ELECTRICIAN" PRINTING AND PUBLISHING 1, CO., LIMITKD, 2 and 3, Salisbury Court, Fleet Street, London, E.<\

PREFACE TO VOL. II. FROM one point of view the present volume consists essentially of a detailed development of the mathematical theory of the propagation of plane electromagnetic waves in conducting dielectrics, according to Maxwell's theory, somewhat extended. it is the development of the waves along wires. The connection of the two subjec ts was thoroughly explained in Chapter IV. of Volume I., which should be underst ood. But on account of the important applications, ranging from From another point of view, theory of the propagation of Atlantic telegraphy, through ordinary telegraphy and telephony, to Hertzian waves along wires, I have usually preferred to express results in terms of the concrete voltage and current, rather than the forces specific electric and magnetic belonging to a single tube of flux of energy. translation from one form to the o ther is quite easy, The when understood.

As far as space would permit, I have tried to develop theory thoroughly considering every kind of wave, and including the calculation of the waves the as as possible, latest etc.) Even the theory of the produced by multiple reflections. kind of so-called wirel ess telegraphy (Lodge, Marconi, has been somewhat anticipated, since the waves sent up the vertical wire are hemispherical, with their equatorial bases on the groun d or sea, which they run along in expanding. (See 60, Vol. I.; also 393 in in this volume.) The investigations are based upon those my "Electrical Papers," with considerable extensions. My old predictions to skin conduction, an d to the possibilities of longrelating distance telephony have been abundantly v erified in advancing practice ;

and my old predictions relating to the behaviour

IV. PREFACE. approximately distortionless circuits have also received fair support in the qua ntitative observation of Hertzian waves of along wires. The reader need not therefore fear that he may be muddling himself over fantastic theories void of practical significance, whatever the scienticulist may say. The mathematical methods employed proved themselves to obtaining are those which have me by practice to be those best suited to new is results to and advancing natural knowledge. The general idea make the differential equations themselves perfectly definite, so that the differential equation of a problem is actually its full it solution, the operational or differential solution,

though may not be in obvious quantitative form. The process of algebrisation, or conversion from differential is, to algebraical form admitting of numerical treatment it of course, very important. Though may be easy when the proper way of treatment has been found, yet there has been a good deal of exploring work which makes no appearance. In Chapter VIII, I have g iven a condensed account of my researches on generalised differentiation and ser ies, a subject that grows naturally out of .the operational way of working. Although I think th is subject has a large future, yet I must warn the reader that there that persons much of who ought no pretence of logical rigour, and the matter was rejected some years ago by is

to be good judges. units. The several appendices relate to electromagnetic waves in general, save the one on rational There is some progress to report. Of the three stages to Salvation, two have been safely passed through, namely the Awakening and the Repentance. I I am is not alone in thinking that the third stage, the Reformation, bound to come. have good reason to be satisfied with the reception given to the first volume of this work. Nearly all parts of it, the outline of general theory, the nomenclature, the rational

units, the vector analysis, tion, and the waves and their applicahave been approved in this and other countries. B ut I

PREFACE. regret that I have been able to make so little impression upon British official science as expressed by its late leader. It is true that the " K.E. law," which set such unnecessary and unwarrantable restrictions upon telephony, is not much heard of now. With advanc ing practice it became so ridiculously wrong (say 1,000 per cent.) that it was i mpossible to save appearances by any manipulation of figures. But a dangerous and alarming officia l error has been pressed forward, even to the extent of experimentation with the public funds. I refer to telephone Mr. Preece's proposal to increase the capacity of cables, with a view to Atlanti c telephony, by bringing the twin conductors as close together as possible. very true that by Mr. Preece's ingenious plan of flattening the wires on one sid e, and bringing the flat sides It is, indeed, closely together, the capacity may be considerably, and even is greatly increased. But it is not the working capacity that ! increased, but the electrostatic capacity Faraday knew if it that much. And this blundering is so unnecessary. For the dignity of one who sat at the fee t of

afterwards rose to be beneath be the leading matters (according to Answers), to insignificant person, still there are other w ays. Why not ask someone else ? It may be too late to consult the family doctor; but there are Faraday and authority on electrical consult the works of an many young gentlemen going about who have been It is to to technical colleges and are quite competent to give information concerning the capacity of condensers. be hoped and expected that the late important removals in the British Telegraph Department will lead to much improvement in the quality of official science. The above two examples show how much improvement has been needed. Others could be gi ven. This volume may help. APRIL 10, 1899.

CONTENTS OF VOLUME II. [The dates within brackets are the dates of first publication.] CHAPTER SKCTION V. MATHEMATICS AND THE AGE OF THE EARTH. (Pages 1 to 29.) PAGE 223 224 225 226 [Nov. 23, 1894.] Mathematics is Rigorous Mathematics an Experimental Science ... Narrow, Physical Mathematics Bold is 1 and Broad ... Physical Problems lead to Improved Mathematical Methods " Mathematics and Ma thematics." Remark[Dec. 14, 1894.] able Phenomenon ... .. ... ... ... ... ... 4 8 10 227 228 The Age of the Earth. Kelvin's Problem 12 14 ... Perry's Modification. Remarkable Result 229 230 231 Cooling of an Infinite Block composed of Two Materials Large Correction for Sphe ricity in Perry's Problem 16

18 Remarks on the Age of the Earth 19 232 233 Peculiar Nature of the Problem of the Cooling [Dec. 21, 1894.] of a Homogeneous Sphere with a Resisting Skin ... ... 20 Cooling of a Body of Variable Conductivity and Capacity but with their Product C onstant 22 22 24 234 235 Magnitude of the Correction for Sphericity in Various Cases Explanation of the Last ... 236 Investigation by the geneous Sphere Constants ... of the Cooling of a Homowith a Resisting Skin. Effect of Varying the ... ... ... ... Wave Method ... ... ... 25 237 Importance of the Operational Method

28 CHAPTER SECTION VI. PURE DIFFUSION OF ELECTRIC DISPLACEMENT. (Pages 30 to 274.) PAGE 238 239 [Jan. 11, 1895.] Analogy between the Diffusion of Rod and the Diffusion of Charg e in a Cable ... Heat ... in a ... 30 The Operational Method assists Fourier 32

VIII. CONTENTS. PAGE Characteristic Equation ... ... SECTION 240 241 The and Solution .in Terms ... of Time33 Functions ... ... Steady impressed Force at Beginning of Cable. ... Differentiation. Simply Period ic Force Effect of a Terminal Arrangement. of a Resistance ... ... Fractional ... 35 37 40 242 243 Two Solutions in the Case ... ... 244 245 246 247 248 [Jan. 25, 1895} Theory of a Terminal Condenser Theory of a Terminal Inductance T

he General Nature of Electrical Operators 42 44 49 249 250 251 The Simple Waves of Potential and Current 15, 1895.] The Error Function. Short T able The Way the Charge and Current Spread Theory of an Impulsive Current produc ed by a Continued Im[March ... ... pressed Force Diffusion of a Charge initially at ... 52 54 57 ... ... ... One ... Point. ... Arbitrary Source ... of Electrification ... ... 252 253 254 255 256 [March 29, 1895} The Inversion of Operators. Simple Examples ... The Effect of a Steady Current impressed at the Origin Nature and Effect of Multiple Impulses [April 12, 1895.} mulae ... Reflected Waves. 58 60 63 Convenient ... ... Way ... of denoting Diffusion For... .. ...

... 66 67 69 An Cases of Simple Reflection Line Earthed at Both Infinite Series of Reflected Wav es. 257 258 259 260 261 Ends The Method of Images. The Waves are really Successive ... 72 [April 26, 1895.] Reflection at an Insulated Terminal General Case. Effect of an Impressed Force at an Intermediate ... .. ... any Terminal Conditions ... ... ... ... Elementary Waves ... The Refl ection Coefficients in terms of the Terminal Resistance 74 75 77 79 80 81 Point, with Four Cases of. 262 263 264 Operators Cases of Vanishing or Constancy of the Reflection Coefficients... Gene ral Case of an Intermediate Source of Electrification subject to any Terminal Conditions 10, 1895.} ...

... ... ... ' [May The Two Ways of expressing Propagational ... 265 266 267 268 269 270 271 Results, in terms of Waves, or of Vibrations ... Conversion of Operational Solut ions to Fourier Series ... 83 88 Ways. (2) (1) Effect due due to e by Special at A, when earthed at A and B Initial Modified 7, Way of doing the Last Case 90 [June (4) (5) 1895.} (3) Earth at Both Terminals. at a Point. Arbitrary Initial State Charge 91 ... Line Cut at Both Terminals. Line Cut at Terminals. Effect of Impressed Force

95 96 96 Effect of Initial Charge Effect of Initial Charge (6) Earth at A and Cut at B. Periodic Expression of Impulsive Functions. Fourier's Theorem 98

CONTENTS. SECTION IX. 272 273 PAGE [June SI, 1895.] Fourier Series in General. Various Sorts needed even when the Function is Cyclic and Continuous. ... ... ... ... ... '100 ... Expansions o f Zero Special Forms of Fourier Integrals. Interchangeable Property. ... Use in Transforming Definite Integrals ... ... 274 Continuous Passage from Wave Series to Fourier Series, and Reversion ... ... ... 104 its ... ... ... ... ... ... 107 108 275 276 How to Find the Meaning of a Fourier Series Operationally its [July 12, 1895.] Taylor's Theorem in tionally Considered Essentials Opera110 114 277 278 A Fourier Series involving a Parabola interpreted by Taylor's Theorem Representation of a

Row of ... Impulses by Taylor's Theorem, ... ... ... leading to Fourier's ... ... 116 279 280 281 [Aug. 9, 1895.] Laurent's Theorem and Fourier's 119 122 On Sketch of Operational Solutions and their Interpretation Way of Extending Fourier's Method to Fourier Series in General 124 282 283 The Expansion Theorem. Operational sions in Xormal Functions ... [Aug. 30, 1895.] : Way ... of getting ... ... Expan... 126 129 131 131

Examples of the Use of the Expansion separately... 284 Theorem (1) Inductance Coil and Condenser The Treatment of Simply Periodic Cases (2) 285 286 287 288 289 290 291 and Condenser in Sequence. Also in Parallel ... (3) Two Coils under Mutual Influ ence [Sept. 20, 1895.] (4) Cable Earthed at A and B. Impressed Coil 137 Force at (5) A A and Cut at B. 138 140 (6) (7) (8/ Impressed Force at A Earth at A, Cut at B. Impressed Force at y Earth at A and B . Impressed Force at y General Terminal Conditions. Impressed Force at A ... Cable Earthed at (9; 140 142 143 292 293 [Oct. 4, 1895.] General Case of an Intermediate Impressed Force 148 The Determinantal Equation 150 152 153 156

293A Subsidence of Special Initial States 294 (10) General Case of an Arbitrary Initial State in the Cable... 295 (11) Auxiliary Expansions due to the Terminal Energy. Case of a Condenser ... ... ... ... ... ... ... 296 Points of Infinite C ondensation. Exceptions to Fourier's Theorem 297 298 299 300 301 302 [Oct. 18, 1895.] 158 159 160 Abnormal Fourier Series Origin of Two Principal Abnormal Cases First General Cas e: Z - -Z x Equal Positive and Negative Terminal Resistances 162 ... ... ... 163 165 167 Equal Positive and Negative Terminal Permittances ... Physical Interpretation of the Abnormal Case of 300

X. CONTENTS. PAGE 8, 1895.'] SECTION 303 304 [Nov. Positive Terminal Kesistance and Negative ... 305 306 307 308 309 310 311 ... ... ... Terminal Permittance ... Equal Positive and Negative Terminal Induct ances Impressed Force in the Case Z = -Zj ... 170 171 172 173 Singular Extreme Case of to the Main One Z or -Z 1 being a Cable equivalent More General Case [Nov. 22, 1895.] Cable. to Elucidate the Last. ... Terminal Cable Z ... ... ... not equivalent to Main One .. 176 179

Arbitrary Initial State Singular Case when Z or - Z: is a Real Terminal Conditions. Coil. Two Ways Terminal Arbitraries. of Treatment Case of a 181 Terminal Coil and Condenser in Sequence Coil and Condenser in Parallel 184 185 312 Two Coils in 313 314 315 [Dec. 6, 1895.} Sequence or in Parallel A Closed Cable with a Leak. ... ... ... ... 186 Split into .. Two ... Simpler Cases ... 187 190 Closed Cable and Leak. Another

Operational Solution Evaluation of Energy in Normal Systems 203 322 323 324 325 [Jan. 1896. 1895.Way Split Closed Cable with Intermedite Insertion. with an Insertion . Normal Systems of a Leak.. 10. 27. into Two 192 194 .} A Cable in Closed Circuit without Constraint 196 198 201 Theory Theory of a Leak..] Initial States in Combinations 207 208 211 Two Cables with different Constants in Sequence. 316 317 318 319 320 321 Same Simpler Cases as last without Initial Splitting Closed Cable with Discontinuous Potential and Current 196 [Dec.

213 215 217 326 327 . 222 225 228 General Solutions for Sources in a Bessel Cable with minal Conditions . Normal Solutions..... ... 28.... .. 1896. 1896. Ter....Cable involving the Zeroth Bessel function A Fourier and a Bessel Cable in Seque nce ...... ...... . Cables in Sequence ... . . . .. .. ^4. . 220 Two . Bessel .. Construction of Operation al Solutions in a [Jan..... Construction of a Normal System in General . ...} Two . ..] 328 Connected System Remarks on Operational and. [Feb... Connection 329 330 331 A with the Simply Periodic Bessel Cable with one Terminal Condition.

Conjugate Property of Voltage and Gaussage . Solutions 352 333 Operational Solutions for Sources in the General Case Conversion of Wire Waves t o Cylindrical Waves... Two Special Cases of Zeroth Bessel Solutions 229 231 Ways Series 334 234 335 Numerical Interpretation of Formulae. The Divergent 236 .

1896. Construction of General Solution by Waves Construction of General So lution by the Convergent Formulae.} Rationality in p of Operational Solutions .??...CONTENTS... PAGE .. Waves in a Medium whose Constants vary a s of the Distance Power 238 241 337 338 339 340 341 . 336 [March the 7i 1896. . 342 343 344 345 The Convergent Oscillating Bessel Functions... .... SECTION XI...} Generation of th The Divergent Formulae are Fundamental... . Solutions in terms of I m and K... 244 246 248 250 Comparison of Wave and Vibrational Solutions to deduce . Relation of Divergent to Convergent Bessel Functions Nature of Algebraical Transformation from Diverg ent to Convergent Formulae [April 24. .. ..... Solutions in terms thereof Oscillating Bessel Functions Reason of the Unlikeness of the and Operational . . 252 255 . with two Boundar ies.

.... .. 256 1 .. ..} Reduced Formulae when one Boundary Bessel Series. 259 346 [May 29.. is at the Origin 260 347 348 The Expansion Theorem and to Initial The Potential due 262 265 266 Dis. Argument showing the Impotency the Origin unless !>>i> .} Uniform Subsidence of Induction and placement in ... Charge Time Function when Self-Induction is allowed for 349 The Potential due to Initia l Current 349A [July 17.The Divergent Physical Electrical Two Divergent Functions H m. K m of Restraints at . 1896.. 1896..

.. .. .. Cases by Inspection (2.. ....} Determination of the Value of p*l by a 352 353 Diffusion Problem m Elementary Generalised Differentiation... .. APPENDIX RATIONAL UNITS . Cable Problem C = (K + Sp)*(R + Lp)-i e (1... VII. .) SECTION PAGE 350 351 [June 26.. 1896... .) Elementary : 286 288 290 291 .. Value of p l when ra is Integral or Midway between .. C.. 275 CHAPTER TIATION.Combinations of Coils and Condensers 349B Uniform Subsidence of Mean Voltage in a Bessel Circuit 349c Uniform Subsidence of Mean Current in a Bessel Circuit 268 270 272 .. ... ELECTROMAGNETIC WAVES AND GENERALISED DIFFEREN(Pages 286 to 433... . .) ..

. K zero... . Convergent and Divergent Results .. Two ... ways... Algebrisation when c is constant and ... ..

.) All Cons tants Finite...) C due to steady V when S Two ways (6. and V due to steady C C due to impulsive V.. Cubic under Radical. (8.'] (12. 18. . 1896.) C due to V varying as Discharge of a Charged into (3. .] (4.. 1896.. Two ways (7. CONTENTS.) V due to steady C when K = 0. tion of Operators [Sept.. and V due to impulsive C Reversibility of Operations. . . C due to V varying as e~P t .) . $8. The Error Function again (5. Instantane ous Impedance and Admittance ..) 362 363 364 365 Empty Cable C due to steady V.........K '/ s (9..) V due to steady C when L = 0.. 293 295 296 297 298 298 an (10.. . ) V due to steady C when Sis zero. . = 0. . 299 300 300 Distribu.. PAGE SECTION 354 355 356 357 358 359 360 361 Change of Operand [Aug.XII. .) (11. Third Way.) . ..

... of C due ..) 306 308 Summary . of Equation (60). The Complete Wave .) Another Development of Equation Third Development of Equation (60) (60). C due 303 304 to steady (13... .. 301 Development .. Integration by 305 Parts (15. Origin varying as e~ K</s (16...) V 366 367 368 369 (14. to V .) Generalisation.. at the ...

9.) Derivation of the (18.. 309 310 311 370 371 The Wave of V Independently Developed The Waves of V and C due to Initial Charge or Momenat a Single Place .) Wave of V from the Wave of C ... .) (19.. 1896. [Oct.. .. . tum 372 373 .... ...of Work showing K*/s the Wave of C due to V at Origin varying as e~ (17.] The Waves ...

. V varying as e~P*.K '/ s The Current Wave due at the Origin (T-) .. of Distortion Operators in Powers to of_p-! 376 377 Example. ..... .... to a Steady Voltage . [Nov. ..of V and C due . e e. .... The Wave of Show Waves the Deformation.. .. ... ... Analysis of Transmission Operator to 312 315 317 319 374 375 Progression and Attenuation of Three Examples...] Expansion or steady .. e ~ K</s due to impressed Voltage . 1896... . ... Impressed at the Origin . 6.

and of Length Zirvjcr Development of General Solution in u m P m and w m P m Functions Derivation of C Wave from V Wave. .. with Examples. 1896. Structure of the Convergent Bessel Functions 324 378 379 380 381 Analysis Founded upon the Division of the Instantaneous State into Positive and Negative Pure Waves Simplest Solutions.... 326 329 [Nov.Impressed Voltage 322 Value of (p 2 nI (<rt) when n is a Positive or Negative" Integer.] Waves of Infinite Length. and Conversely. Condition at a Moving Boundary. Expansion 330 ofe^ 331 .. . 27. ..

.. IS. .. Examples . . of V due to any Impressed Voltage 335 Identical Expansions of Functions in I H FuncK .K of a.. SECTION XIII.CONTEXTS. ] 384 385 386 tions. PAGB 382 Deduction of the the V and C Waves when */ V is Constant from Case V n = ce. . 383 [Dec. 1896. 337 . Formula for (\x) Electromagnetic Applications Expansion of any Power Seri es in I m Functions. Construction the Wave Expansion of eP* in I Functions..

] Wave due varying as e Impressed Voltage varying as e-/**.] nt The Waves developed in w P m (z) Examples and C due to any V Functions from the Operational Form of V ofV 387 e/* 346 Remarkable Formula for the Expansion of a Function in ! and Modification of 386. 26. Effect of increased Resistance in Rounding off Corners and Distorting to 390 391 392 393 [Feb. 1897. [Jan.1 . 1897. Functions. Example 388 389 Impulsive Impressed Voltages. and the Impulsive Waves and their Tails Generated 348 350 353 Tendency of Distortion to vanish in rapid Fluctuations. 15. and Examples.340 344 Expansion of a Power Series in J H Functions.

357 to Wave due V P t log t 361 363 [April. 1897.] Effect of a Terminal Resistance as expected in 1887 and as found in 1896 Reflection at the Free Ends of a Wire. A Series of Spherical Waves 394 395 396 Reflection at the 367 End of a Circuit terminated by a Plane Resisting Sheet Long Wave Formulae for Terminal Reflection [May 21.. 1897.] Reflection of Long Wa ves Terminal Reflection without Loss.9.. 369 373 375 378 380 384 387 . in General .

.. 391 Summarised Complete Solutions [Aug. . .. Full Solutions with the Inversion of Operations. . Waves due Terminal Reflection 400 401 Application to Terminal Resistances... Critical Resistances.. General to Way Solution of Definite Integrals of Finding Second and Follow. .397 398 399 Wave Solutions Comparison with Fourier [June 18.... 6.. 1897... ... .. Derivation of First Wave from the Second 402 403 Derivation of Third and Later 390 Waves from the Second . Second Wave with any Resistance ...] ing Series.. .. ..

.] 392 393 ..... expansible in J H functions .. . V and . Initial States. 395 396 398 The States of V and C resulting from any .404 405 406 407 408 Reflection by a Condenser Reflection by an Inductance Coil Initial States. .... C .... 400 . Expre ssion of Results by Definite Integrals The Special Initial States J ((rx/v) and J m (<rxfv) 1897.

and Modification .. .. 14. 1897.... .. 18. 409 410 411 [Dec. due to Elements at the Origin The Time I ntegral of C. ..] Some Fundamental Examples The General 402 Solution for any Initial State. and some Simple Examples . . 406 408 409 410 412 413 415 ^ 417 422 424 Initial State of .. Conversion to Definite Integrals. [March 2o.... .} The C due to iutial V Operational Method.. 1898..XIV. 1898. Both kinds of Bessel Functions [Jan.. . .......] initial V and C Undistorted Waves without and with Attenuation . due to Elements at the Origin . The V due to initial V Final Investigation of the V and C due to [F eb.. 17... 1898...] Effects of Resistance and Leakauce on an . . .. . 403 Short Cut to Fourier's 412 413 414 415 416 417 418 419 420 421 Theorem The Space Integrals of V and C. . Evaluation of the Fundamental In tegral Generalisation of the Integral. ... CONTENTS.

] Positive and Negative Tails 429 . 422 423 424 426 . [April 29. PAGE 434 . Division of Charge Initially at the Origin into Positive or Negative Charge betw een them Two Waves with The Current due to Initial Figs. GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. 1898. Charge on one side of the Origin ...) 425 426 427 A Formula for ^ obtained by Harmonic Analysis Algebraical Construction of g(n)... . 1 to 13 described in terms of Electromagnetic Waves in a 431 doubly Conducting Medium CHAPTER VIII.Constant V on one side of the Origin . 428 The After-effects of an Initially Pure Wave......... . . ... SECTION (Pages 434 to 492..

. [July 432 433 434 Remarks on the Operator (1 + A" 1 )" . 1898..... .] .. ... ....] Value of 9(n)g(-n) 428 Generalisation of Exponential Function Application of the Generalised Exponentia l to a Bessel Function.[May 27. .. Remarks on the Use of Divergent Series Logarithmic Formula derived from Binomial .. 445 1898... 436 439 441 429 430 431 443 Limiting is Form of Generalised Binomial Expansion when Index -1 1. Algebraical Connection of the Convergent and Divergent Series for the Zeroth Bes sel Function. ........... to the Binomial Theorem [and to Taylor's Theorem . . ... .. 445 447 450 451 Logarithmic Formula: derived from Generalised Exponential . .

. 465 467 469 440 441 442 .. 1898.CONTENTS. Generalisation of e~ x (a.. PAGE 453 435 436 437 438 439 [August Connections of the Zeroth Bessel Functions Bessel Functions . Operational Prop erties of the Zeroth Remarks on Common and Generalised Mathematics 19..} [Oct.. 7. 456 457 1898. 463 H (a. SECTION XV.} The Generalised Zeroth Bessel Function Analysed Expression of the Divergent Generalised Bessel Functions.) and in K (x) in Terms of Two .) ..

.Rectification . 1898. Generalised Bessel Functions (a. Two . Bessel Function of the Same Order fx Product of the Series for [Dec. Power 472 474 and . 2...) n (#) in Terms of any Generalised and e~ z cos rir..The Divergent The Divergent Hm Hn (as) and and Kw K Terms . Special .. of .} Possible Transition from c~* to e* 443 444 445 446 Series for log x Examination of some Apparent Equivalences...

} DISPERSION 507 APPENDIX [Dec...} Cotangent Formula and Derived Formula for Logarithm.. 481 447 [Jan..476 478 Determination of the meaning of a Generalised Bessel Function in Terms of H and K Some Apparent Equivalences .. 20. ON COMPRESSIONAL ELECTRIC OR MAGNETIC WAVES 493 APPENDIX [Aug. 7. 1898.} D.. 1897.. . 12. E.. Various Pr operties of these and other Divergent Series 482 of 448 449 Three Electrical Examples Equivalent Convergent and 487 490 Divergent Series Sketch of Theory of Algebrisation of (1 . 13. 1896. .bp')~ l l APPENDIX [Nov. .. .. .

.... and the Properties of Inversion and Interchangeability of Operators General Transformation of Wave... Cases of Reduction to a Simplex Wave-Surface Transformation from Duplex Wave .} ON THE TRANSFORMATION OF OPTICAL WAVE518 518 SURFACES BY HOMOGENEOUS STRAIN (1) (2) (3) (4) Simplex Eolotropy ...to Index-Surface by a Pure Strain .. .. . Eff ects of Straining a Duplex Wave Surface .F.. 1893..Surface by Homogeneous Strain 519 520 521 (5) 525 527 528 (6) Special (7) . Properties Connected with Duplex Eolotropy .and Wave-Surface Equations. . Forms of the Index.

Two 529 .. (9) 530 531 APPENDIX [Jan.. 536 APPENDIX I..... WAVES 538 . NOTE ON THE MOTION OF A CHARGED BODY AT A SPEED EQUAL TO OR GREATER THAN THAT OF LIGHT . [Nov.XVI.... 1898.] NOTE ON THE ATTENUATION OF HERTZIAN ALONG WIRES.. .. 533 APPENDIX [June 11. 14. PAGE (8) Substitution of Three Successive Pure Strains for One.'] G. 1897. IN NOTE ON ELECTRICAL WAVES SEA WATER .] H... 11. CONTENTS. . Ways Transformation of Characteristic Equation by Strain (10) Derivation of Index Equ ation from Characteristic . 1898..


For =. Equation (46) two lines after. equation (91) : 22 read 2. Page 160. Equation (31) : For M/irt read (lijvt)^.si. equation (30) : For (li/irt) read (A/irt)*.sZ read = .ERRATA. line 17 " For " conection read " connection.si. Page 42. equation For = ( read =( . equation (100) For : +Es . : Page 81. for si read ." (45) : Page 97. For : : Page 115.

12 : : For Z read Page 229. llth Page 248 : line from end For " ' is .Es." Page 165. Page 246. e h(x . equation (126) Page 210." : Page 245. equation (29) For : I read I m . line 5 " For " reciprocal read " conjugate. line For ** read z. line 8 from end : " For " thus read " then. Page 163.read .

" . 8th Page 274. 3rd line line from end from end For " when : " read " where." For 339 read 340. : Page 255." stanFor " notation " read " dardization.read " is a.

we also add. because they are fundamental. n. are seen to be secondary or unessential. MATHEMATICS AND THE AGE OF THE EARTH. not because they are unimportant .CHAPTER V. an Experimental Science. accompanied with just a sufficient amount of theory to connect it together and render its acquisition easier and more interesting. Mathematics is 223. but because they become subordinate to the theory in a certain sense. in order to secure the inimitabl e advantage of a personal acquaintance with something real and living. B . however. it becomes possible to consider th e theory by itself. will prob ably be agreed with by most persons. whereby the essentia their connections become recognised. and the foundations are always hidden fr om view in well-constructed might buildings. the general experimental kn owledge has been acquired. like So it comes about that a great theoretical work Maxwell's treatise on Electricity and Magnetism contains so little explicit info rmation regarding the experimental facts Theory facts ls and of the it always tends to become abstract as emerges successfully from the chaos of by the processes of differentiation and elimination. and. to be explained by additional means. That the study of the theory of a physical science should be preceded by so me general experimental acquaintance therewith. in a great measure. After. as theory. VOL. and are ignored temporarily. whilst minor effects science. The experimental facts then go out of sight.

so to speak. to acquire geometrical ideas through his senses though now. One must actually know the practical straight line before any defi nition of the abstract straight line can be underis Then our understanding and acceptance of the de anition it states what we knew a lready. of cour se. We ma ke our geometry. disguise it as we may. with the essence geometry of nature. though we may have never openly thought stood. and we can see plainly how m uch it is to the advantage of the learner that he should continue. no strictly formal There is ba sis. periential fined per se. such as the one that asserts it. a recognition that As regards an axiom. in the first place. V. can ever be possible. For. its acceptance involves the existence of a very extensive experience of the geometry of nature as it is found to be. in accumulated experience.ij ELECTROMAGNETIC THEORY. that two straight lines ca nnot enclose a space. Geometry. CH. . like any other science. in the first place. before being set to work upon the more intellectual theo ry on a formal basis. Now take a hint from nature. apart from experience. besides the formal axioms and definitions and postulat es which would be unintelA straight line can never be intelligibly deligible wit hout it. of the and become saturated. even though this be unaccompanied by any intelligent compreh ension and expression. considerable to start with of an exalways something very nature in the background. tendency in the most abstract and logical pure mathematics. to suit the state of things in which we are born and live. has most certainly an experimental foundation. with his attention specially directed to the subject freely assisted by mode ls. then. solid and skeleton. for instan ce. We all make our acquaintance with geometry first through our There is the same of all sciences senses. How impossible.

and these are only und erstandable by must experience. independent of nature. is also no self-contained theory possible. ence associated therewith. apart from metry For a language is used in its enunciation. definitions. besides the general experidefine a thing in a phrase. even of geoconsidered merely as a log ical science. There which implies that developed ideas and complicated processes of thought are already in existence. We . We come down to axioms. practi cal meaning. and postulates at last.about it be to prove anything rigorously as absolute truth.

We are all antipodeans and upside by the gradual fitting together of the parts of a distinctly connected theory that we get to understand it. Some ways will be preferable to others. instructive not to later to obtain different views of the same country. no bottom down. It is to anything. 3 using words. in a complicated maze. of course. These words have to be explained in other There is words. It may be better to wander about.MATHEMATICS AND THE AGE OF THE EARTH. and We go over the ground in any way. and so on . and be guided by circumstances in the choice of paths. It may It may be more interesting and go by the shortest logical course from one point to another. for ever. may begin anywhere. or give broader and clearer view s. and by the revelation of its consistency. but no strict course is necessary. and vary the route not even be desirable. and keep our eyes open to the side prospects. since they may be easier. Now it is .

containing trackless jungles. and be guided by circumstances. and differen t experiences. much easie r and perhaps better ways will be found. It is o bvious. the effect of putting a learned society in the unfortunate position of appearing to exist not merely for the encouragement of research along established lines. mountains and precipices. already well explored. as sometimes happ ens (of which a case was quite lately brought to my notice). if they lead. But it is enough when the question is somewhat different when it is a case of exploring a comparatively unknown region . keep your eyes and your mind open. they may operate unfairly in effect. To attempt to follow a logical course from one You should point to another would then perhaps be absur d. However harmless in intention. Later on. so that a crowd can push along. I think. it has ing. that complaints of the want of perfection of the ways and manne rs of work of explorers on the part of men who are accustomed to more rigorous m ethods have a considerable element of the ludicrous in them. How you do it is quite a secondary consideration. no doubt. but also for the active dis- .that of guiding another plain over a well-known country. to the rejection of honest work which failed to be appreciated by the judges. who had no doubt diff erent ways of working and thinkWhen this result arises. You have f irst to find out what there is to find out. that certain distinct routes may be followed with advantage.

of course. But then. This is the natural way. he will. as the mind and i ts wayshave grown to harmonise with external Nature). get on in a manner suitable for progress in his physical studies. the probable fact that the judges were animated by benevolent motives. the properties of space along with the properties o f the matter found moving about therein. it is greatly to the advantage of a student of physics that he should pick up his mathematics along with his physi cs. especially. V. and of course. an d only desired to turn the misguided author from the error of his ways into more rigorous paths approved by themselves. but I do not suppose that it would be generally practicable. does not make the matter any the He has his own ways. There is no end to the subtleties involved in rigorous demonstration s. Rigorous Mathematics is Narrow. even though he be told (virtually) th at his work is valueless. Physical Mathematics Bold and Broad. so that exceptions and reservations track. though possibly the best course for a large-minded man. an d must follow better for the author. couragement of work of a That this is the case in reality it is impossible to believe. This would be very comprehensive. And have to be added. mathematics being fundamentally an experimenany other. pursued by the creator s of analysis. If the student does not pick up so much logical mathematics of a formal kind (common sense logic is inherited and experiential. when you go off the beaten the most rigorous demonstration may be found later to contain some flaw. by inference. To have to stop to formulate rigorous demonstrations would put a stop to most physico-mathematical inquiries. t hat he must not continue to send more of it.4 ELECTROMAGNETIC THEORY. papers are so cheap. it is clear that the Science of Nature might be studied as a whole. Again. at any rate. Nevertheless. and not worth printing. 224. For then the one will fit the other. tal science. like Now. and one more or less does not matter. . less conventional character. CH. if he can. them.

in the first place. The physics will guide the physicist along how to useful and important someby the constant union of physical and geometrical or analytical ideas. The practice- . there should be. Now.formalism. in working out physical problems no pretence of rigorous results.

as he could not put without the physical guidance . and (it m ust be confessed) sometimes with justice. problem. pleasure some re marks on this point in the Preface to the recently-published second edition of L ord Rayleigh's treatise on Sound. fairly satisfactory . is right and proper to match the But along with this has come a tendency for purists to object to the introduction of physical ideas in mathem atics. But to this question there are two sid es. but it would be to no purpose. and which commends itself to the electrician. especially in details involving much calculabut the general prin ciple should be carried out as much No as possible. But no do ubt I have seen with much there is frequently a lack of rigour. ever important it may be to maintain a uniforml y high standard in pure mathematics. It may be that there is no lack of rigour sometimes. The physics should be carried on right through. The pure mathematician will complain. and to obtain the great assistance to the mathematics. severely taken to tas k by a distinguished purist for his use of Green's Theorem in Spherical Harmonic s. This cannot always be done. but an Maxwell was increased generality and sirnpliBed treatment. probably quite rigorous. There have been enormous advances made in pure mathematics it much more.MATHEMATICS AND THE AGE OF THE EARTH. and well's treatise. in the last half century. with particular attention to dynamical ideas. howof deficient rigour. the physicist may occasionally do well to r est content with In the arguments which are view. 5 of eliminating the physics by reducing a problem to a purely mathematical exerci se should be avoided as much as possible. but only illustrates different wa ys of thought. For. to give life and reality to the which the physics gives tion. mathematical purist could ever do the work involved in MaxHe m ight have all the mathematics. which I cannot do better than reproduce and it is : mathematical investigations I have usually employed such methods as present them selves naturally to a physicist. This is in no way together to his discredit. a method which is excellently to the purpose. with a possible lack of rigour as result. as advance in physical science.

highest standard would mean the exclu sion of the subject altogether in view of the space that would be required. in many cases of difficulty to insist upon the strative. the more severe procedure of the pure mathematician may appear not more but less demonAnd further. exercised in a different order of ideas. especially in the subject ." more but less demonThis is exceedingly true.and conclusive from his point of mind. his To " not Particularly notice the words strative.

which occupies so la rge a part of the treatise in question. Thomso n and Tait.6 ELECTROMAGNETIC THEORY. how hard to read from its severe formalism. and how narrow it seems from the want of physical illustration. in order not to interrupt his work. But this purification is more esp ecially suitable for the purist to undertake. and I wish purists would take a lesson from Fourier. work done in proposi tions and corollaries ? It is bad enough in pure mathematics. cleared it is But to be away from physics it becomes remembered that the men mathematics working out who have . When mathematics set in logical form. it If one sort of work is as necessary as the certain that the physicist would get very little other. having the fear of the is What is more hateful than a physical rigourists before him. considerable part of a very big Theory of Functions in search The work is most admirably painsfor what I did not find. Perhaps the subj ect might be greatly lightened by having a physical theory to rest upon or is to illustrate. the proper time to find out the reservations and corrections is later on. and tell their tale differently and make it in teresting by letting in I have had occasion to go through a a little imagination . of the expansion of functions in series of other functions. Maxwell. work done by trying to d o both. or Rayleigh. CH. but how d ifferent from physical mathematics. V. And I would add that if the physicist do es sometimes get carried too far. taking and severely rigorous.

it only differs in degree from most . consider that a in advance of the la tter. even though he him so. an enormous and endless subject. and there is a striking difference in the ways in which it is regarded by the physicist and the rigourist. demonstrations to prove what he knows . The expansion material to be made coherent and elaborated. can have practical certainty that a certain expanphysicist sion is when he possible. of functions in series. because his physical problem tells seeks and finds the solution. already mentioned. a lthough it be a fallible kind of knowledge. It is arose physically.have usually been in the past initiated great advances in men who were employed in They supplied the purists with raw physical questions. has not investigated the properties of the functions used in the expanHe does not arrang e long and severely disagreeable sion. with the peculiarity that the former is far To understand this.

changes in the conditions the physicist gets an endless But by number of expansions of one and the same arbitrary function in circular functions. and as a matter of course. the purist carries his mathematical developpresent. as is perfectly clear from It is evident that for tion in physical considerations. methods which he knows work usefully to his assistance. yet in other respects the purist lags And this is true in other matters than that menbehind. and what becomes of the rigorous demonstraThen there is an endless number of other s orts of tion then ? functions. Given certain funcas a question of functions. "What a pretty piece of work there must be to answer this question ! This is true even in the case of circular functions in the particular case rigorously treated by purists. 7 mathematics in it But eliminate the physics. Certainly the purist is bound to complete the logical treatment some day. but Thus at the hard-bound rules of the purist make it difficult. For a physicist may use daily with success. a lthough ments so far in some directions as to be far beyond physics.MATHEMATICS AND THE AGE OF THE EARTH. every one of which can represent an arbitrary fun can endless number of ways. comprehensive rigorous demonstrations (not special) we need enlarged ideas about functions. I would it. out of sight in fact. tioned. to w ait for a proper development. The best result of mathematics is to be able to do without To show the truth of this paradox by an example. remar k that nothing is more satisfactory to a physicist than to and which have r get rid of a formal demonstration of an analytical theorem and to substitute a q . but wh ich are logically unintelligible to a purist. and this respect. an d perhaps the purist would obtain the necessary broadness of view by a study of the physics in which such comprehensiveness of expression is found. simply put can an arbitrary function be expressed in terms of them? t ions.

symbols may be . or a geometrical one freed from co-ordinate symbols. with the commonsense method of proof by The The first is latter means of the addition of circuitations. and make it be practically axiomatic. When seen to be true. and in any very tedious. and quite devoid of luminousness. kind of co-ordinates. which wi ll enable him to set the necessary truth of the theorem.uasi-physical one. Contra st the purely analytical proof of the Theorem of Version well known to electrica l theorists. makes the theorem be obviously true.

and the truth becomes an integral part of one's mental constitut ion. however this may be (and the matter is difficult and speculative). of this work. as having To complete the matter evidently a substantial existence. there is nothing impossible or incredible in the result. like the persistence of energy. But. CH. V. It is should be regarded like any other dielectric that is. as regards electromagnetic disturba nces traversing it. The plain meaning to be given to it without in troducing additional data is that the ether. Chapter II. 225. which the actual displacements in bulk represent the electromagnetic disturbance s. we come to a stop. and find that mechanical forces of the ordinary kind are invo lved disturbances pass through the ether which supports them. and the suggestion is that it should not be regarded as an elastic solid (gener alised) in . There is sions of mathematical ideas a curious analogy to be found between extenand extensions of electrical theory. But when we go to the very ve rge of the system. come to matters needing interSo it is in mathematics. Take. the p oint here is that on the outskirts of the theory we pretation and a larger theory. Maxwell's theory in the form presented in If we do not inqui re too I. when electromagnetic simply unintelligible at present. dispensed with. we may easil y be temporarily Vol. for instance. requires a theory of the ether itself. under the impression that the two circuital equations and their accessories expr ess the dynamics of a quite self-contained system. Physical Problems lead to Improved Mathematical Methods. The fundamental notions are so . supp ositional Now.8 ELECTROMAGNETIC THEORY. minutely into the consequences.

But. But we do. keep to convergent series and whole differentiations and regard . To say nothing is of the negative quantity (which interpretation imaginary). besides these. and of fractional differentiations It is customary to or integrations. and connected matters.simple that one might expect that unlimited developments could be made without e ver coming to anything unintelligible. whic h has only a sort of become understood and its properties developed comparatively recently. there are m uch more obscure and illunderstood questions. there is the imaginary. and in various of directions. such as the meaning and true manipulation of divergent series.

the latter suggesting the former. however ill-understood may be the theor y involved therein. and the a voidance of complicated evaluations of definite integrals. and that practical way s of working arise naturally. or even to ignore them altogether. When C is exwe have to find E to match through the operator E". But then the Well. See also plicitly given as a function of the time 221 later on. and the execution of the processes invol ved in E". frequently effects great simplifications. . E" being a complicated function of a differentiator. It is in the carrying out of these processes in the investigation of various ele ctromagnetic problems that we are obliged to regard certain kinds of divergent s eries as representing fully significant functions. ticularly here of the solution of the differential equations to which physicists are led by quasi-algebraical processes. I suppose all workers mathematical physics have noticed how the mathematics seem s made for the physics. for obvious reasons. as being legitimate and feasible. The latter is the usual attitude of moderate and practical mathematicians. 9 divergent series and fractional differentiations as meaningless and practically useless. The reader will see to what I refer by reference to 203. Vol. as if they did not exist. which assumes various algebraical forms. then C has to be found by the operation on E of the inverse of E".-MATHEMATICS AND THE AGE OF THE EARTH. This is really It is a fact that their use the cas e with resistance operators. If they can be ignor ed. L. when I allude to th e definiteness of meaning of the operator E" in the equation E = R"C. wh at of that rigorous logic of the matter is not plain Shall I refuse my dinner be cause I do not fully understand the in ! Xor is the matter an unpractical one. why trouble about them at all ? But when these things turn up in the mathema tics of physics the physicist is bound to consider them. and make I am thinking more par the best use of them that he can. and when it is E that is given.

satisfies himself of the accuracy of his r esults. not if I am satisfied with the result. may be fully aware of his want of infalliand that his investigations are largely of an experimental character. and may be repellent to unsympathetically .'? process of digestion ? No. by the application of tests. Now a physicis t may in like manner employ unrigorous processes with satisfaction and usefulnes s if he. At the same time he bility.

a snivelling religion. and other wicked people have it all their own way. others may readily be given. were unduly exalted at the expense of the supreme virtue of justice.10 ELECTROMAGNETIC THEORY. But what about the theo ry of functions which deliberately ignores the treatment of divergent series ? C an it really be the theory Is not a more comprehensive theory needed. charity. and if this religion we re carried into practice generally. constituted mathematicians accustomed to a different kind of work. where the sometimes prohibitive. and so forth. the strong-minded and the just allowed it to prevail. rogu es. a nd the just would and . should ever come to pass that there prevailed in world a so-called religion in which the minor virtues of mercy." Remarkable Phenomenon. If the example fails. If this it and Mathematics. it would be a very bad thing for the world. then would the liars. slanderers. CH. meeknes s. V. only fit for the For such a religion would be weakminded of both sexes . of functio ns ? including both convergent and divergent functions in a harmonious whole ? "Mathematics 226. hypocrites. Under these circ umstances the substitution of equivalent divergent series which may be readily c alculated becomes a matter of great practical importparticularly in is labour ance. Convergent mathematics often excessively unpractical in the labour it involves in I have noticed this t he numerical interpretation of results. resignation. There is is another point of view. spherical harmonic expansions.

in self-defence would assert the paramount importance of justice. instead of wasting several years It is to Camof my l ife in mere drudgery.* I regret exceedingly not to have had a Cambridge education myself. 1894. . in tacit defiance of the nominal religion i n which it was made a secondary virtue. November 23.if be crushed along with the meek and and better than their religions. Let us above all things try to be just. I cannot join in any general attack upon them. bridge mathematicians that we are indebted for most of the Do not mathematico-physical work clone in this country. 100. p. or little more. Even Cambridge mathematicians deserve ju stice. most ma thematical physicists hail from Cambridge ? Are not * See article in The Electrician. But just men are lowly.

differential equation. to say nothing of the large crowd of other and mostly younger men who se names will suggest themselves ? must take the good with the bad. are apt to take refuge in a definite integr al. and so finding out what the soluIn such cases we might as well keep to the tion means. whi ch I very much fear they do not always get. so as not to lead to unnecessary obs truction. We unpractical conservative tendency that exists (conserving the bad as well as the good. or harder. and resisting innovations). in this as in other and though legitimate and s erious objection may be raised to the distressing and soul-destroying style in w hich some Cambridge mathematicians do their work. and call that the soluIt is certainly one form of the solution. when they cannot find the solution of a problem in a plain algebraical form. tion. from the difficulty of evaluating the integral. What one has a right to expect. especially the meek and lowly. But it may be just as hard. Orthodox mathematicians. As regards their want of sympathy with l ess conventional men. it has come knowledge that a man who is not a Cambridge mathematician." I may mention a somewhat remarkable phenomenon which has lately occurred . and to the matters . it is not sympathy that is particularly wanted perhaps it would be unreasonable to expect any at all. and not unduly de preciate or make invidious comparisons. to interpret than the differential equation of the p roblem in question. Maxwell and Eayleigh Cambridge mathematicians. and t hose who long suffer under slig-ht. is a fair field. Now. For even men who are not Cambridge mathematicians deserve justice. and that the want of sympathy should be kept in a neutral state. we should also bear in mind the great volume and value of the work done.MATHEMATICS AND THE AGE OF THE EARTH. howe ver. On this question of " Mathematics and Mathem atics. and who does not pret end to be much of any sort of to . and be just as wise. 11 Thomson and Tait.

but who is a practical physicist.my mathematician. capable of discussing with prop er judgment such a question as the age of the earth. recently made the d iscovery that a certain unconventional . a higher limit to which he finds (and with good reason) to be very likely hundreds of times greater than th e most probable previous estimate (which conclusion has obviously important interest to geologists and astronomers).

question of high mathematics at all. but o f the substitution of simpler and more direct processes for the indirect and complicated processes of the highly cultivated mathematician with too rigoro us proclivities. It is the fact that as he has got to know how to go to work. to speak. maticians of the Cambridge or conservatory kind. as soon . rigorous kind but makes use of what he finds useful. so in question. he has no prejudices of the practical physicist. It is really not a . method which professes to obtain the solution in plain language directly from the diffe rential operator. without any final success beyond obtaining a definite integral of too complicated a nature to be practically discussed or obviously e valuated. It has naturally given me much pleasure to find that the solution of the problem he . and. V. he is a practical p hysicist. to evaluate the definite integral without the trouble of finding it. For this reason. CH. at least in suc h simple cases as the above method experience shows are fairly and without much trouble within the reach of practic al physicists and electricians not mathe. or go from Dan to Beersheba and say that all is barren but of the common field variety.12 ELECTROMAGNETIC THEORY. s hould receive such ready appreciation from a Of course. who look the gift-horse in the mouth and shake their heads with solemn smile. who take the seasons as they come and go. . with grateful appreciat ion. in these diffusion problems at least. that constitutes the importance of t he phenomenon. without mathematical pretensions. in a few lines in whereas by the methods generally employed he migh t have spent days over it. mode of treating the mathematics of the question (explained to him by myself) co nducted him immediately to the exact fact had in hand. I shall have no further hesitation in making use of the in question occasionally in the course of the rest of this work.

The Age of the Earth. as he has made the matter of the age of the earth interes ting at the present time. 227. I give some particulars regarding simple solutions. Therefore. . as we shall see later. John Perry has suggested to me that I should write something on t he subject. Kelvin's Problem. Now Prof. They are not s o much out of place as may appear at first. for they all represent electrical problems of interest.

*" who started this branch of inquiry in his celebrated paper on< the Age of the Earth. the same as those in the crust. the converse problem. 1& The main problem is : Given the earth initially at a constant throughout. V how long it takes to arrive at its present state. and approximations made with various d ata. The earth is and by mea ns of surface source? initially at zero temperature. find its way of cooling. with a plane boundary kept at zero temperature. . Let V be the temperatur e at distance x from the plane face the . It can then be deduced crust. In accordance with general practic e. Find the temperature gradient in the skin which results. Lord Kelvin. k the conductivity. and p means the tim e-differentiator. and. well-known characteristic of V is where c is the capacity per unit volume. much simpler problems are substituted.IATICS AND THE AGE OF THE EARTH. Therefore.MATIIE:. in temperature find the gradient of te mperature in the earth's particular. its skin is thereafter main tained at constant temperature V Now consider first . for that is t he observed datum. substituted for the earth an infinite body of uniform capacity and conductivity .

we only want the surface gradient.V = e-^V gives (2) V in terms of V . This is easily integrated. (5) the same as . between (2) and (3) we get 9 = <lVo. (4) which is the solution. say Thus. So. we have p*l and so that (4) is = (irt)->. it To turn to algebraical form. but g.

S. 1862 ." . Thomson and Tait's Natural Philosophy. or " App.. R." Trans. D. * Since the final state due to our source is V everywhere in it follows that in th e subsidence problem.the earth. Edin.. starting " On the Secular Cooling of the Earth.

only reversed in This required solution. Remarkable Result.* . which are data used by Perry. f is to assume that the capacity and conductiv ity are higher in the earth than in its That this will prolong the time of subsi dence may not skin.000C. is not large. due to Fourier. The next step. V. Taking ff V. 228. The correction necessary for the finite size I find of the earth. used by Lord Kelvin.. the gradient has the same so that (6) is still the the formula. under the assumed circumstances.14 ELECTROMAGNETIC THEORY.=4. direction. is V . = . from a uniform temperature value. that the time requires to be reduced by -^ part. That is. due to Perry. CH. other data being the same. understand in the case of an . it t akes about 100 million years for the gradient in the skin to fall to its present value. we find that t is 103 million years. be difficult to .Perry's Modification.

or by the same investigation simplified to case. where /^ is the conductivity.infinite block with a plane face but the result is most distinctly not an obvious one in the case of a sphere. Then the current of heat inwards through the skin equals that distinct case. for it may be readily shown that the time of cooling may be either less or greater. (39) below. which is kjl. This gives the condition that at #=0 we V have if is the temperature outside the skin. So * This may be proved by formula meet the present " . t John Perry. is finite. according to circumstances. On October 14. So take a Let the skin be so thin that its capacity may its conductance per unit of surfac e. whilst entering the inner earth. 1894 . B ut we want numerical results. be ignored. and I the depth. Y! that just inside.

further circulated with addit ions in pamphlet form in in revised November. p." privately circulated in MS. 1895. .the Age of the Earth. and published January 3. form with other matter ill Nature. 224.

due over initially. (9) which. by division. by the use of (5).MATHEMATICS AND THE AGE OF THE EARTH. useful. The latter is most one convergent.] V . B is the resistance of unit area of the skin. There are two ways of converting (8) to algebraical form. Thus. V = [l-Ekq+(EkqY-(EkqY + l . is "15 the solution for Va in terms of V impressed. the other divergent. is* to V constant all . . . is turned to from which the subsidence solution.

which an important conclusion. so that is term. The gradient is got by dividing by I.Here a = c&R 2 the first . with the same ^ in the skin. on account Then apply an alternative method to (8). the value of t varies as the is value of ck in the earth when t is big enough. . Now (11) is unsuitable divergency. when t is small. which is sufficient when rt/ t is big enough. So.

. cooling from an initially unifo rm temperature the boundary condition being constancy of ratio of . is turned to the algebraical +V * I a (l-c'/ ).n.. because the boundary condition on the insi de of the skin is formally the same as in Riemann's case. the rate of loss of heat to the temperature. The problem is formally the same as Perry's problem of a resisting skin. V. (14) am informed that this is Riemann's solution for the surface temperature of an infinite block with a plane face.of the viz.= which by means form of p-^l = tnj . Perfectly free escape of heat means R = 0.

besides continui ty of current at the interface of skin and earth. but with different constants. to initial CII. VSo the subsidence solution due V is The equivalence of (11) and (15) is merely an example of the generalised exponen tial. (17) where Vi is the temperature at x = and C. any distance x may be obtained and without difficulty. or . D are unknown. Going a step further. c. (16) lZ fl t. but it is interesting to see how it arises. immediately on receipt of the above. D. so far as the beginning and most important part of the solution was concerned. k. and is tre ated in the same way as the inner earth. T he boundary conditions are v = V at z = /. 229. let us examine the influence This was done by Perry o f the capacity of the skin itself. Cooling of an Infinite Block composed of The solution for the temperature at Two Materials. Let V belong to the earth. and c v kv v to the skin.16 ELECTROMAGNETIC THEORY. and to be eliminated. and v = Vi at z = 0. V = c-^V = ' *C + e* 1. and z in the skin Then we have also from its inside. but we need not similarly consider that here. Measure x in the earth from the inside of the skin. The skin may now be of any depth.

x = 0. is (20)That is. which need not be given."^ =-'"-/. the flow of heat in the the skin due to applied V on . These find C and D. to the following expression for the g radient of temperature at the outside of the skin : where Notice that s s = + (c^/cky when cJi = c^\. and lead us by ordinary algebraical work. 1 y= r.at L c' v i dV n o\ (18) = 0.

its outside same for any .

3.MATHEMATICS AND THE AGE OF THE EARTH.. Intewe obtain a series which also belongs to the exponential kind. since they are formally We have alike. (21) Only one term needs transformation. expand the fraction in (19) by division. (1 Thus.. To get the complete so lution. 17 material inside. 2. .)fcV . being (23) 7rh\t Here n has (21). in the successive terms of we obtain + 2* 9 A 4 + 2*"A 9 + where . so to be 0. using the ordinary exponential expansion.)V 0) (22) by (20). g= + 2sy + 2y + 2y + . . provided ck = c 1 k1 which reduces (19) to the elementary form (4) first considered. VVo=(?i-2^ grating by (5) + l^. . . &c. The approximate solution involving the first power of I may be got at once from (19).. and it leads to the former results when I is small. 1.

as in (11) for example. Or. So the VOL. We may get thi s by picking out the coefficients properly from the expansions of A and its powe rs in (24). (24) A= on the In fact (21) expresses the initial effect of V in the first term right. . as they involve whol e differentiations.) + &c. 26(1 + 2 2s + 3V + 4% . II. and of a n infinite series of images due to the change of medium.. (25) E when arranged in powers of q lt Here the even powers of qt are ignorable. Although (24) is a very neat form of solution. J.. from the operational equation (21) itself. which becomes . 2* (1 + 2 3s + 3V + 4V +. we may want a s olution arranged in inverse powers of **. JL 3 W)l 2 .fcV .. C .

and so It is easily seen b y inspection that the functions of s that occur in (25) are derivable in success ion from one another by So they the operator s(d/ds). term gives the factor of ~-. Or. we can see the effect of the capacity of the skin. Comparing with (11). The next step would be to go on to consider the case of a spherical earth. first with no surface resistance to the escape of heat. 230. may be written finitely if we like. <?! on. of which is the same as where R = l/ki as before. the q^ term that of t~2.18 ELECTROMAGNETIC THEORY. CIT. is the complete solution. Large Correction for Sphericity in Perry's Problem. with a . and. next with a resisting sk in. thirdly. where the A is first part short for d/ds. the first one being 2*/(l s). V.

Perry has examined Fourier's solution for a homogeneous sphere with constant surface emissivity (equivalent to a resisting skin) and a very interesting result comes out. A-x = -00595. Prof. or 4 kilom. But these are too complicated for the present purpose. internal and external. viz. with can be got through the operators in the form of Fourier series much less work th an by Fourier's way. and c = 2-86. Thus.. k = -47. according to Perry.. the capacity being allowed for. take the radius of the earth at 6380 G and the depth o f the skin at 40 5 centim.. and perhaps some Cambridge mathematicians At the same time.skin of any depth. that the plane solution gives mu ch too big a result with the same data as regards conductivity and capacity... I may remark that the solutions as well. Then the time of subsidence to the present state is 960 s years. or 96 times the estimate of 10 years given with the 8 same data except that the . . ^ = -507. the initial temperature as 4000C. and kjc 14 times that of surface rock. so that k in the earth in the earth is is 79 times that of surface rock. and would frighten timid readers. the present surface gradient 1 in 2743 centim .

for the capacity of the skin. the time of subsidenc e would. So the effect of having an infinit e conductor with a plane face to represent the earth is to increase the time of subsi dence from 960 8 to 46308 years. I find it comes to Allowing sary. By the firs t term of 7 Perry's formula. so alone for our comparison. I have worked out the corresponding formu la with the capacity for. by (12) above. Perry found two terms in the Fourier expansion to be necesallowed term is a large multiple of the second. Thus. be ck/ofa times as great as the estimate without the inc reased c and k inside. There is very little difference numerically. 7 . s or about 4630 years. first But the take it we may 9030 years. Perhaps it to note the difference does not matter in very rough estimates but it is interesting made by the finite size of the earth. In order to ascertain whether the objection made to Perry's neglect of the capac ity of the skin had any serious basis. and . or the linear diffusion of heat. But. that it lessens the time of subsidence in the ratio 463 96. according to the plane theory adapted to meet the case of a plane slab of skin instead of a spherical shell. 4-7 to 1. or . when calculated by the plane theory. earth has the same c 19 and k throughout as in the skin. the time of subsidence comes to 9020 years.MATHEMATICS AND THE AGE OF THE EARTH.

Physicists . Perry's investigadiffidence) tions. The first is by not requiring so much.* There are two evident ways of getting more. Now a few remarks (which I make with much of Prof. Tait. This seems to * It is to be remarked. of the Earth. by allowing that the earth's evolution went on at a far m ore rapid pace in former times than at present. on the practical outcome on the other hand. wholly apart from catastrophes.But I have not taken special pains to get the third figures right. offers them only 2 0. I believe. that geologists have come down in in consequence of their demands remarkably. however. It is known that geologists demand long ages of time for the earth's evolution in its geological aspect. and probably Lord Kelvin's work. Remarks on the Age 231. have offered them what geologists consider the miserable allo wance of from 20 to 400 million years. They want more. . Prof.

it is evident that the age of the earth may possibly be much extended. Perry's argument seems to me to apply with I greater and greater force the furth er we go back in time. which Prof. may be even at present much during the whole period of geological evolution has to be considered. The state the latter. especially greater in the earth's interior than in its skin." philosopher.20 ELECTROMAGNETIC THEORY. presume that the earth's birth. V. the only reasonto be Topsy's theory. whilst it incr eases one's difficulties. He advances arguments to s how that it is reasonable to suppose that the earth's effective capacity and conductivity. and Any other theory is of the elephant an d tortoise kind. . and Prof. so that the time of subsidence to the present skin gradient is prolonged. As for the origin of life able view seems to me upon this planet. CH. Perry's way. consequent upon its cooling. for geological purposes. The other is Prof. should be reckoned from the time when it became encrusted. Is it necessary to solidify the earth all through If we allow it to solidify gradually before beginning its life ? by the natural increase of depth of its solid crust. explains nothing. She was a true " she spekt she growed. be a reasonable view. But it is not a mere question of the present state of the earth's interior. or when the crust attained some notable thic kness to give some sort of stability. a sort of evasion.

is not by any means an obvious result. In connection with the above problems in cooling by diffusion and escape of heat at the surface of bodies there are a number of incidental matters of great interest.Tyndall was of Topsy's persuasion. some of the In the most notewo rthy of which may be briefly noticed. first place. Peculiar Nature of the Problem of the Cooling of a Homogeneous Sphere with a Resisting Skin. as was remarked in 228. the p rolongation of the time of cooling of a sphere from a given uniform initial temp erature until a given gradient of temperature is reached at the surface of escap e. as I firmly believe (subject to correction) in the truth of his view as to the " "promise an d potency of life in so-called dead matter under the influence of the forces of nature. 232. So am I. t hough not difficult to understand in ductivity . produced by augmentation of the conand capacity of the inner portion only.

MATHEMATICS AND THE AGE OF THE EARTH. except in the skin. when it is a sphere To show this. made where either infinitely great as it an extreme without case. we can easily make it be either a retardation or an acceleration at pleasure. changing the capacity in the of discharge capacity . le t the conductivity be that is in question. remains skin finite. 21 there the case of an infinite block with a plane boundary when is a similar augmentati on of conductivity and capacity its within skin. In fact.

Pe rry's example.then like that of the is the cooling of The reof a condenser through a resistanc e. the result is 920 years. as Perry has shown. and the case is considerably 8 mixed. but 7 only 330 years when it is 1 kilometre. by reducing the internal conductivity from infinity to 79 times that of the skin. or 0'5. which is 8 only 9 times the standard res ult of 10 years found by Lord Kelvin. It is a general principle that incre asing the conductivity accelerates the subsidence of a normal system. s dard result. Thus. with infinite internal conductivity and with internal capacity as in the skin . It is raised to 960 R years. Perry's 960 years problem. and therefore the resistance. If. or a distr ibution of temperature which will subside according to the condenser law. and Pro f. 8 reduce the time to 520 years. until the present gradient of temperature is reached in the 8 skin is 160 years when the skin's depth is 10 kilometres. sistance here is the resistance of the skin. we raise the internal conductivity to infinity. will show the danger of over. or a little less. Perry's case is no exception.hasty generalisations regarding the effect of vary ing the internal conductivity and capacity. The theory capacity per unit volume also the skin value. the time required falls off to any extent. and the is that of the inner body. Now by reducing the depth of the skin. If it is 4 kilometres. making no other change. requires the skin to be only J kilometre in depth. in Pro f. the time taken to fall from an initial uniform temperature of 4000C. or 0-006. with the skin conductivity as in 230. But there are other considerations. and the internal or body of the sphere the sphere. w hilst at the same time increasing the internal To obtain the stancapacity to 5'7 times that of the skin. 10 year s. 7 as in Prof. we Here the internal capacity is . When made thinner These examples st ill. we may accelerate the discharge as much as we please.

say C. value of The result may be extended to include any number of slabs . consider another matter. It was mentioned in 229 that the flux of heat into an infinite homogeneous block due to sources maintaining its plane face at the constant temperawas not altered by changing the material under the same value of the product cJc . Now reduce it to the same still 5-7 times that in the skin. and the time falls to 920 years. V. it falls to 108 years. reduce the inter nal conductivity to the skin value. the flux of heat. 7 value as in the sk in. Or thus.22 ELECTKOMAGNETIC THEORY. by equations (2) and (3). as just mentioned. Lastly. being now Lord Kelvin's Cooling of a Body of Variable Conductivity and Capacity but with their Product Constant. and case. CII. 233. After these illustrations of the curious nature of the problem. per ture V skin to another having the is unit area. (28) This is unaltered by a change of material not altering the ck.

heterogeneous material in which ck is constant. in the limit. provided ck is the same for all or. Conversely. and ck is the same for all. then we . if we start subside by internal diffusion and free escape at the surface. homogeneity in every slice parallel to the plane face. similar result applies to a sphere. however. A due to the finite sensible. the flux of heat at the surface will be unaltered by every plane stratum is any change of material. and let it everywhere. provided the correction for sphericity. to a continuously .of different materials put together to make a block. with. provided homogeneous in itself. with concentric shells instead of plane slab s. since t he final temperature due to the impressed V is a state of uniform temperature V with V everywhere constant.

be insensible : or if it be may have approximately the same result. . this is very correction for sphericity.size of the sphere. for Sphericity in Magnitude of the Correction Various Cases. 234. Another interesting point is the magnitude of the As mentioned in 230.

the 108 years of the infinite homogeneous block is reduced by J. we do not get the Perry effect. But I confirmed the result. with the sa me values at the surface as before. years On all Now in Perry's case we have a large increase in c and a very large increase in k beginning at a moderate depth. Also. surrounding city and very much greater conductivity. the other hand.part to represe nt the new case of variable c and k in the earth. To illustrate this I have calculated a few cases of continuously heterogeneous material.600 millions problem of a homogeneous sphere of greater capa4 kilom. The time for the corresponding infinite bloc k is then 4-7 times that for the This was so remarkable that I suspected and sug sphere. when the surface values of c and k extend through the earth. rection for sphericity is quite small. so as only to become very big near the centre. let the c and k in the earth vary inversely as . as in Lord Kelvin's problem. 23 a shell of depth large in Perry's 9. as when th ey are That is. I find that the correction is reduced to -! part. as mentioned in 8 It only reduces the time of cooling by -^th part of the 10 which belongs to the infinite block. the cor227. gested to Perry an error in his calculation of Fourier's formula. instead of by part.MATHEMATICS AND THE AGE OP THE EARTH. and also obtained very nearly the same result from an enti rely different formula which allowed for the capacity of the skin. But if we increase them gradually.. to accentuate this effect. ^ constant. When c and k both vary inversely as the distance from the centre of the earth.

provided in c and k near the surface. That is. to the given 8 is 10 years. the same Similar results occur in other cases of gradual variation. we do not introduce great changes . and clearly great latitude in the law of variapermissible.000C. very large changes in c and k near the centre.the square of the distance from the centre. but very little near the surface. it may be. the time of subsidence from the initial state of gradient of temperature at the surface as for the infinite homogeneous block. or betwee n the surface and with. half-way tion is down . 4. Then I find that the correction vani shes.

the c and k in the sphere be constant. very small round face. Since the flow of heat is radial. V. and so is its total capacity. if Now the total conductance of a shell is proportional to its area. it is like the diffusion of heat We through a series of flat plates of variable conductivity and capacity.'2i ELECTROMAGNETIC THEORY. Explanation of the last. being zero at the centre. cfl. Therefore. may get an insight into the meaning of these corrections by dividing the sp here into a series of shells of unit depth. 235. and greatest and increasing most rapidly at the surSo we see that large variatio ns in c and k near the only trifling differences in the state of things and capacity of the innercentre may make . the conductance and capacity of a shell vary as the square of the distance from the cen tre. it.

at the surface. we see that to rep resent the sphere of uniform c and k. in the sphere vary inversely as the distanc e from the centre. Also. this block to be insulated at its sides and open at its ends. c and k in the block must vary directly as the distance from the far end. so little difference is made in the time of cooling to the present gradient. say a block of length equal to the radius of the sphere. as the conductance most shells are naturally low. Finally. they are so far away from the surface where the escape takes place that their c and k may become of little moment in the practical problem concerning the gradient of temperature in the skin. only it is That is. now see why the correction in the last case disappears. Besides that. . block. when c and k in the sphere vary inversely as the square of the distance from the centre. These may help one to unders tand why. and of cross-section equal to the surface of the sphere. c and k in the block must be uniform. the c and k in the block must vary directly as the sq uare of the distance from the far end (corresponding to the centre of the sphere) where they are zero On the other hand. when c and k (equivalent to insulation). when c and k vary as the first or as the second inverse power of the d istance considerations from the centre. It is the time for the sphere being the same as for the block. we have a homogeneous of finite depth instead of infinite. by substituting a block for the sphere.

We not asserted that the complete solution of the problem is the .

Then V . found in 228. and that under the circumstances the finiteness of depth of the block does not influence the result sensibly. and q = (cp/k)*. Prof. Investigation by the Wave Method of the Cooling of a Homogeneous Sphere with a Resisting Skin. Perry's case involves so large a correction for sphericity as to deserve an independent confirmation by a method not requiring the use of the Fourier expansion. I find that operational method my leads straight to the solution by a simple process. in this case. Effect of Varying the Constants. equation (8). It is very easy to make mistakes in cal culating Fourier series of complicated forms. Let V b e the temperature at distance r from the centre of a sphere of radius a and of u niform c and k due to impressed on is V the outside of an enveloping skin. 25 same in both cases. when R is the resistance of unit area This of the skin. the secondary waves to and fro along the block due to its finite depth being of insensible effect because the d epth is so great. Fortunately.MATHEMATICS AND THE AGE OF THE EARTH. k the internal con ductivity. but that the problem is reduced to one of linear diffusion i n a homogeneous medium. Now get the corresponding solution for the sphere. varying 236. In contrast with the above results with continuously c and k. that We expresses the temperature V x just inside the skin due to V impressed on its out side. in the plane problem. For it is by the con sistency of results obtained in different ways that a conviction of the accuracy of the results of complicated processes may best be obtained.

note that in the place V satisfies the spherical characteristic * -**' (31) .is given by a shin gr r shin qa la + Ekqcoihga first To prove this.

it satisfies the condition of continuity of the flux of heat there. that it ELECTROMAGNETIC THEORY. V. V . O is Y x. and. is complete. So put r =a (33^ which gives V x in terms of that 1 becomes 1 . is finite CH. at the centre . or This in (30). we may develop (33) by long division to the form . that at the inside of the skin.s (which Comparing with (29) we see less) and Ekq receives the factor coth qa.26 next. which brin gs in an infinite series of secondary diffusive waves between the centre and bou ndary. To show them explicitly. y_ I-8 + Ekqcothqa _Y _ = Ek/a. But what we want if s This makes. lastly. where r = a.

the third term results from the weaker second reflected wave. term is the result of the first wave reflected from the centre. and so on. (34) (35) Here V the result of the primary wave from the source outside the skin. as we know by the solutions previously given when the proper numbers for c and . But all these secondary waves are of insensible effect in our problem .is a trifle Y! = (a where + aft + a$* + ~^ a y= . . as modified by sphericity the second r/ V is . . .) Y .

is is We see that the practical solution in operational form. it We also know. by (11) and (12). only with a changed constant. by (33). are put in. &c. when it reduces to So.. equivalent to (29). that only the first power of q is significa nt in the earth problem. infinite in the coth function. 228. first same as making a part independent of y.k. The significant solution is This amounts to the merely the 1. This makes (36) become <> .

230. with a changed The effect of the s phericity is.)'l + \ _ZlA (l-s)lgJ (41) Now (40) is of the same form exactly as (12). where I. So . by is 27 (5). (40) if g -(l-. fore. which. v" (1-8)1 fcx(lif Vj = gl. . Now 0-0495. This makes tracting th e right member from V . therevalue of the gradient. g is of depth We may the gradient of temperature in the skin write (39) thus. converted to the algebraical form The subsidence solution is got by subthe required result.MATHEMATICS AND THE AGE OF THE EARTH. the same as changing the gradien t in the plane problem from g to g. put in the numerical values as in 5 I is 40 and V is 4000.

The method followed is an example of the theory of 12 of my " On paper Operators in Physical Mathematics. 9 and as the square of V so these may be dismissed at once. (42) which t is to be used in (40). the maximum occurs 8 being 10 years. c = 2-86. the unit when That Perry should have spotted the maxim am ^ . and is = 95-5. This increased value of g requires So the time required to make g square. With Z = 4 kilom. = 73. The formula (39) allows us to see readily the effect of The time of cooling varies directly as varying the constants. S. Vol. I find that t has a maximum and a minimum when I is under 7 ki lometres. E." Proc. to be reduced as its . Q.D. be -rrrs i g (2-1819) 2 or 4-76 times as long for the infinite block as for the sphere. c .That of s is g' = 2-1819 g.E. LIL. Varying k only. There are left / and k .. the full expression for the secondary waves can be developed from (3 4). If desired. but all we wanted was a direct corroboration of the result got from the Four ier expansion. 1893. The minimum is of no consequence. which is o f very general application.

v. We now leave heat problems.28 ELECTROMAGNETIC THEORY. Then t a smaller k will do. first. '. 237. heat. CH. as of There is. These results altogether favour Perry's view. Importance of the Operational Method. Thus with I = 20 kilometres we get = 68-5 when k/k^ = 15-95 only. But the hump is so flat-topped that Thus &/&! much smaller values = 30 makes t be about 90. Aad I = 30 kilometres allows us to have = 53-4 when k/k is only 10-65. comes in principally in two different ways. and are better than his own exampl e. so closely is (unless he is a witch) one of the most remarkable coincidences in ancient or modern history. Lord Kelvin's . Another (and perhaps physically better way than increasing k) of getting big is to increase the depth of the crust. t of k will do for big t. and pass to the theory of electrical matters involving diffusion. Pure diffusion. and = 95 when k/h is 21-3.

the electrostatic diffusion involves the sim plest fundait when is mental ideas. viz. and will therefore occupy us first. these diffusive propag ations arise from the general theory of electromagnetic waves has been explained in ChapHow ter of elastic diffusion.electrostatic diffusion in a submarine .. diffusion in a cable or other circuit. these. and a kind of diffus ion in a magnetic conductor. in considerable detail. Secondly. there is Maxwell's d iffusion of magnetic induction in electrical conductors. After that. diffusion in electrical conductors will naturally follow. the self-induction and the leakage that control Of matters. IV. including the more difficult case What we have now t o do is to consolidate . perfectly insulated and free from self-induction. There are also two comp aratively cable when unimportant cases.

Next. the practical import of which. as very uncertain and speculative data are involved. remains to be disco vered. Perry. however. to illustrate Lord Kelvin's theory of the age of the earth and its recent extension by Prof. to show that my operational method . as it turns up naturally. At present the above illustrations from the theory of heat diffusion will serve a double purpose. The physics itself will serve to guide us along to useful methods and results. First.the knowledge by actual exemplification. We shall then be able to explain the me aning of the operational mathematics above employed.

" The genera l case is perhaps hard (I did not try it). 1895. integral or was quite easy. although less it goes deeper..MATHEMATICS AND THE AGE OF THE EARTH. but I found at once that when linear. according to any power. 1895. 29 of dealing with these and similar more advanced problems is I assert that by its means problems can be of importance. saying. Nature. He s . I sent Perry some solutions then himself extended the matter by taking c and k to be any similar functions of the temperature.] After writing the above. but there may be times when it becomes a duty e. It is rather disagreeable to have to others quite practical. when mathematical rig ourists work and are obstructive. This i s also obvious when done He finds that if c and k increase (Perry. And. of this kind. 7. Furthermore. so that. f inally. says he can't do it. c fractional. 341). Feb.g. be self-ass ertive and dogmatic (especially when one thinks of the always possible risk of e rror). and is sure you can't. "X. p. doesn't know anyone who can. Prof. of the temperature. the solution and k vary together.* * [March 21. attacked and successfully solved with greater power than by any other known meth od. Perry wrote asking about the case of capacity and conductivity functions of the temperature. yet it mathematics of the complicated that it is for the above reasons and kind. that it is essentially simple in operation requires less . the cha racteristic becoming This is obvious when done.

224) that his conclusi ons were independent of the cor" rectness of R. Perry would be fully E. Jan. Feb. K. But Dr.. 14. Prof. p. allowing for other things. Weber has supplied fresh data which do not in k. Perry's is important principle really operative. or whether mathematical physicists will. 367. then Lord Kelvin's age is multiplied (l + s/5) times e. whilst that in c is though to an uncertain extent." In any case. by 121 if s 50. If correct. but also under great pressure. 1895." must be temperature. His data were due to Dr. Geikie. 1895. and so bring down the age even to 10 million years . justified. p. . be obliged to go up to meet them. 438) concludes that Perry is wrong. It will be interesting to s ee whether the geologists will continue their downward course to 24 or 10 millio ns (Sir A. it difficult to come to a reliable estimate as to how far Prof. Mar. and indicated a large inc rease in c and k with temperature. to about 24 millions. Lord Kelvin has to prove the impossibility of the rocks inside the earth being better conductors (includi ng convective conduction in case of liquid rock in crevices) than " The rocks at 20 miles deep are not merely at a high the surface rocks. 1895. Perry said (Nature. and Kelvin on the age of the sun.2 per cent. however. Weber.. is quite satisfied with only 1 00). 3. Nature. 7. So Lord Kelvin (Nature. or. show any notable increase much less than Perry assumed. in agreement with Mr. Prof. . He is also incline d to reduce the initial temperature. p. Clarence King's conclusion in comparing the calculations of H elmholtz. g. per lOOdeg. by fresh data. Newcomb. Weber's results (the old onesj.

liable and that the external disturbances to which circuits are should be removed. On this understanding the two circui tal equations. when 200-202.d -BpV. accompanied by the greatest possible distortion in transit. reduce to suitably transformed as explained in . as before explained. dx * -f=KC. to the theory of d iffusion being in slow signalling through a long cable. Analogy between the Diffusion of Heat in a Rod and the Diffusion of Charge in a Cable. In order that the problem of the propagation of electrical disturbances alo ng a telegraphic or telephonic circuit shall reduce practically to that of the diffusion of the electric displacement a fter the manner of heat in the celebrated theory of Fourier.CHAPTER VI.* making self-induction be of importance. and is therefore a state to be avoi ded by. PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. because it is a sta te of relative inefficiency. It w ould not be at all desirable to bring a practical telegraph circuit to such a state closely. ax L . it is necessary for the self-induction to be ignorable. if efficie nt rapid signalThe nearest approach ling with little distortion be required. 238. represented we make believe now forms that this of case is truly by the reduced the more general equations appropriate to elastic diffusion.

Chap. IV. . Vol..(1) 215-218.

if the flow of heat in a real rod be equivalent extent. are the conductance and capacity for hea t per unit length of rod. if not imp ossible. whilst V and C are the transverse voltage and the circuital gaussage of the more gener al theory.and S It is the temperature and C the flux of heat. if there be a pair of wires. but which may now be called the potential difference of the wires and if the current in them. to insulate it thermally to an So.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. laterally. rod. easy to insulate a rod electrically but it is much harder. 81 where R and S are the resistance and permittance per unit length of line. using only one wire. But the rod should be insulated . Or. we may From (1) we derive the characteristic g-B8pY-/ly. (2) In order to translate to heat problems. perhaps the easiest way is to consider t he longitudinal conduction of heat in a Then 1 V is whilst R. have a cable in question. rejected for want of . then we call V simply the potential and C the current.

On on account the whole. case of the linear flow of heat in an infinite is possible in is This the homogeneous to conductor.a sufficiently close similarity to the electrical problem. and prev ent the collapse that would occur were they separated. and. we may imagine an infinite number of rods Just as jerry-built by houses in a street mutually support one another. On the other hand. so will the rods prevent the lateral escape of heat from their fitted together in contact side side. of the facility wi th which terminal and other auxiliary arrangements can be imagined. there is no doubt greater scientific interest in heat problems when they concern such stupendous earth . neighbours. venient in this respect. questions . if need be. translate These remarks are to enable the reader from electrical to heat problems readily. so that a longitudinal flux of heat the same way as in a perfectly i nsulated rod. the cable is preferable in the study of diffusion. The beat problems are n ot so conpractically realised.

common mother but since . I cannot go on further with that question.as this the is age of our primarily an electrical work. but lea ve it to David and Goliath.

CH.32 ELECTROMAGNETIC THEORY. But it was more than lucid. The reader who may think that mathematics The virtues of the academical system of mathematical training are well known. proposi tions and corollaries. Its light showed a crowd of followers the way to a heap to When we of is all found and can be put in a cut-and-dried form like Euclid. But this is not to be misunderstood in the sense suggested. to be disparaging and endeavouring to therefore. The Operational Method assists Fourier. but as The limitations of mutual friends helping one another. 239. admired. we rules. No one a dmires Fourier more than I do. We have now to consider a number of problems which can be solved at once wi thout going to the elaborate theory of Fourier series and integrals. not by rigorous have to find out first how things go in the matheWe matics as well as in the physics. preliminarily. VI. and bring them into alignment with other processes. by keeping him in may find that there are other checks instead of stimulating any originality regular grooves. to work by instinct. grooves just as . in out. It wa s luminous. a ppear But it is nothing of the sort. complete treatise on diffusion Fourier's and other methods would come in side by side not as antag onists. and if he expects a similar system atic exposition here he will be new physical problems. In doing th is. one) is that the student it Outsiders may possess. Bu t it rigorous A very serious one (perhaps a necessary has its faults. is very much mistaken. And I must here write a caution. I may have to point out sometimes that my method leads to I may. solutions much more simply than Fourier's method. disappointed. In a supersede his work. and I must necessarily keep Fourier series space and integrals rather in the background. go of it we may be able to see our have learnt the an understanding of way the meaning of the processes. It is the only entertaining matheIts lucidity has always been matical work I ever saw. forbid thi s. shall have.

as my grooves ar e not the conventional ones. Now. and would not be favourable to .good. and perhaps a great deal better. for their purposes. there is Such would be quite no need for any formal treatment. improper for our purpose.

and is reader may of then be able a it . only it is done by trial and error. You would be fettered by your own conventions. Surely the author has got to know t he go of it already. find It is ever so it much it circumstances. and be in the same fix as the House of Commons with respect to the despatch circumstances. is known. 33 For it is in matherapid acquisition and comprehension.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. but he knows better. of business. and can therefore eliminate the preliminary irregularity an d make it logical. and then go on by mere deduction. stopped It is not the proper way under the unnatural way. deductive work in a sense. any formal exposition possible. But the reader may object. matics just as in the rea l world . you must observe and is All experimentation it. being an easier to the reader to the go of first. not experimental? So he has in a great measure. followed by new deductions and changes of direction to suit experiment to find out the go of Only afterwards. when the go of it is Nothing could be more fatal to progress th an to make fixed rules and conventions at the beginning. by its own rules.

consider the its characteristic equation (2) above. To this exhowever. If q . with a little assistance. Ther e is no occasion whatever (nor would there be space) to describe the failures which make He can be led into up the bulk of experimental work.little the natural way. although their mathematical knowledge the rigourists as a straw to a haystack. D . and yet know nothing whatever about the operational solution of physical differential equations. successful grooves at once. the historical method can be departed from to the reader's profit. solution would obviously be (3) V = e^A + c-^B. II. may be to that of It is possible to carry waggon-loads of mathematics under your hat. were a constant. The later to see the inner meaning tent. Of course. VOL. Now. himself. The Characteristic Equation and Solution in terms of Time-Functions 240. I do not w rite for rigourists (although their attention would be delightful) but for a wid er circle of readers who have fewer prejudices.

VI.34 ELECTROMAGNETIC THEORY. satisfaction still the are by t. there are two are any constants. test is the same. to is The constants constants with respect is. where A and B It is sense when formal are equally true that (3) is the solution in the same because <f has the operational meaning RSjj. independent functions of A and B means independence of x. That is. CH. now any to find functions of . The constancy of x which satisfy (2).

duced at a distance can be easily deduced later. that the oper ations can be effected.That (3) but they x. general solution of the characteristic. . a form of the further. convert processes implied by the exponenthe solutions from operational to volve preliminarily unintelligible ideas. To go wo have tial A and B to suit special cases. as they inexample. The easiest solutions are those relating to the effects produced at a given spot by causes acting there. to The best proof is go and do it. and then by the execution of the operators There is a lot of assumption here. for algebraical form.

even the finite size of the seas of the earth might be sufficient to contain a sufficient length for So. so by winding it about. Let an our purpose. in a large interval of time. whilst is so very far off that it ca nnot react sensibly on the near end. and to a great distance therefrom. now we treatment. and B=V in (3). which is. where x 0.Those protake some special cases to begin the It need infinitely long cable be laid in an y depth of water. not be laid straight. and V may be regar ded as the sole cause of disturbance in the e from charge. and the other to earth. Then A = . not necessarily continuous). age be regarded as any function of the time t (real. that the near end of the cable is the far end to be freely at our disposal to operate on. but cable itself further away. to line. the absence of resistan ce put The impressed voltbeing merely a practical simplification. Let the cable be its initially free operated upon by a battery of voltage = beginning. . and be then e and no resistance at That is. Th e effect is to raise the potential at the beginning to V = e. one end of the battery is may of course.

making V = -*V (4) .

The operator j/R turns potential to current. because V can be made simple. By (4). in which A is not C= the . (5) first equation giving C in terms of We also have second in terms of V . and there Other cases. the (6) if C is the current at x = . will no other imposed condition. therefore. and also The last equation is the simplest in general. Similarly.V= r"V . So the resistance operator is see at once from (7) that the current entering the cable depends only upon the ratio of S to R. and V = V at x = 0. We Steady Impressed Force at Beginning of Cable. for the characteristic is satisfied. the conductance opera tor of the cable. . Fractional Differentiation.PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. V at the place. is 35 zero. Simply Periodic Force. 241. come later. and the second of (1). It is. Its propagation in the cable itsel f depends on their product.

stated is a well-known one in Fourier's theory of heat conduction . 1*1 =Wr*. on removal of unnecessary co nstants. (A) . What is C ? To find it.Now zero before let V be such a function of the time as to be and constant after t = 0. we Now the problem require to know the meaning of p*V . and when by Fourier's methods we develo p the solution we find that it is Comparing with (7). we see that we require to have.

for the present. let it be taken as a f act that the value of p*l is (irt)-*. On the other hand. is = 0. Any problem involving p*\ in the operational form of sol ution will do for the determination of its value. and does not need Fourier. without necessarily going a step further in the direction of fraction al differentiation.3G ELECTROMAGNETIC THEORY. By and its at the first see that the current entering the line is infinite (because of the absence of se lf induction). can make use of this fact extensively in F ourier mathematics with We much advantage. the result is a simple fundamental one in fractional differentiation. The 1 means that function of the time which We are only This way of finding the meaning of a fractional differentiation of a given zero before and unity after t concerned with positive values of the time. a fundamental formula. VI. But the reader presumably cannot tak e in the idea of a fractional differentiation yet. then falls. function is purely experimental. by comparison with the solutio n by Fourier's method. according to the inverse square root of the (9) we moment At . So. CII.

to zero. because the resistance is infinite. Mathematically speaking. and the current rent is Finally. but the potential near the beginning gets to be so nearly V as to prevent the v ery distant parts of the . But as the cable gets charged the slope gets smaller. Or we may say that the final curV zero. . never stops increasing. first. we say that V=V everywhere. and so is is C . There can It really only be current when the charge is increasing. when = oo t . cable receiving their charge except at an insensible rate. the potential is everywhere zero.time. the slope of V in the line is infinite at beginning.

(5) . and gives the ultimate steady form. Another way of looking at the matter is to consider how w o get the simply periodic solution out of an operational solution. If the frequency we put p = ni in the op erator. when is n/27r.variation. the impressed force is simply periodic. or an infinitely prolonged period. Now p = is equivalent to w = 0. . can be a steady state. Putting p = in an operator destroys time. which means a steady state.V and C may also be seen from (4) and them and they reduce to V = V and = 0. This process is general. when there The final states of Put p = in .

that time must be allowed to enable this state to be arrived at. the simply periodic solution is. differentiation is so (10) it is complete when e siunt. that is. 242. Here Say i means p/n is . 37 In the present case. then V with respect to nt in amplitude and phase. given . by (7). let us = ZC be effect of a ter minal arrangement. Let . Two Solutions in the Case of a Resistance. Still keeping to the beginning of the cable. It (11) should be understood. Effect of a Terminal Arrangement. \smnt + cosnt)e. though.PURE DIFFUSION OF ELECTRIC DISPLACEMENT.

. consequently by (12) B This finds V . we have (B/S^) . Also. the potential at the beginning of the cable.examine the its V put between the cable and earth with the impressed voltage acting. operational solution (13) may be readily algebrized (or converted to algebraical form) in various cases of Z. we have V = (B/S^) C as before . . so that is Z 1 is its resistance operator. as before seen so if Z Now. in terms of e. This because the operat ors are additive like resistances. equation per that of the cable se. is is to express the current through Z and entering the cable. practical The and unpractical.

the conversion. One case will do to begin with to illustrate \ i Let Z be the resistance operator of a coil. say (14) .

(16) I-. r is its resistance \r CH. shall suppose. This may be algebrized as follo ws.. as . having previously The cases of a perbeen zero .38 ELECTROMAGNETIC THEORY.}* Here in Z we have only complete differentiations. constant after t=0.. therefore in union with the ev en powers of (Sp/R)* we still have complete differentiations. VI. and Z is a mere resistance. All these terms may be ignored when e is. Then ( ~ "I / 7 \ / /S?A* ^/^ \ 15 ) is the operational solution giving V in terms of c. By division. where and I its e inductance.

As t decreases. in the case of no is self-induction. differentiations performed upon is found by complete Thus. the final When t is smaller. smaller still another term requires to be counted. then we ca . the first t term. so the solution it. t is When big enough. when it is a mere resistance that concerned. When value. so that the initial convergency has wholly disappeared. the accuracy of When it reaches calculation becomes somewhat impaired. the only significant term is e. So (16) reduces to ?* 2 ?* )(*)> We know p l l already. the smallest term moves to the left.we mittance and an inductance will follow. This makes a series in descending powers of t\ Thus. As it comes near the beginning of the series. and so on. But we must never pass bey ond the smallest term in the series. the nex t becomes significant.

and some amateur botanists have declared that the antidote is to be . It is said that every bane has its anti dote.n only roughly guess the value of the So (19) is unsuitable when t is small enou gh to make series. the initial convergency be insufficient.

antidote is 39 The have an example here. in a different way. In the first line we have t o make the same &c. by division.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. obvious enough. This is also done at sight by (B). n is integral . viz. 1.. : -{*<)*> In the second line we have complete integrations to perform on e. This we may split into two series. found near the bane. 2. Keepgot by algebrizing (15) We ing for the present to the simple case of a resistance only 1 = in (15) we may write R (20) l+(_^_) \>*SpJ or. 3. when. as at present. when the matter of fractional differentiation . This is done b y which is viz. complete integrations upon the function p*l.

by using (A) in the easy enough. But : at present we can do without it.is understood. which first line is of (22) we i-t So. is we can calculate V conveniently when t bad when t is big. and so on. . though not mutually destructive. sider (23) the bane. obtain the complete solution in the for m We is see now But that (23) small. Then we may con(19) th e antidote. and They are complementary. and integrate directly thus r-r*-* jr-rt-JL.

Then TTT/BJM 1+ ( sjt \ R/ 1 rrr + I (24) . equation this for is C = spV. so its resistance operator is -1 (s/>) . Vl.40 ELECTROMAGNETIC TIIEOKtf. Superficially considered. Put Z in equation (13). is another simple case in which substantially the same process obtains as in the last example. ance. the problem of the effect of a terminal condenser 243. Thus. CH. Theory of a Terminal Condenser. There in modifying the action of an impressed force en a cable is entirely different f rom that of the effect of a terminal resistYet there is a very close analogy. let the Its terminal arrangement be a c ondenser of permittance s.

potential at the beginning of the cable due to the impressed force runs through the same course when the condenser is interposed as the current (multiplied by r) does when a resistance is interposed . To obtain the effect at a distance requires in both cases the .aq if a = s/S. and q is as before. only differing in the changed So. if they are equal. put of a terminal Z=r in (12). we see that the operational solutions are of the same form. to show the analogy with the resistance. that of the condenser. we see that the cons tant. making r + g q 1 =-' 1 4(24A) bq if b = r/'R. or that length of cable whose permittance equals effect. Now. or that length of cable whose resistance is the same as the terminal resistance in the changed problem. Comparing (24) with (24A). a becoming b.

current in the resistance problem.same operator Consequently. we know that the course of the potential throughout the whole cable in the condenser problem is the same as that of the introduction of the . we do not need to algebrize the solutions in order to predict this result. due to the same impressed which may be any fu nction of the time. . And. ~QX force.

Thus. expand the operator in descending powers of ay by division. makin ( 25 ) Here the even powers of q involve complete integrations. and when we find the n umber whose logarithm it is . Similarly we might say number when we take its logarithm. &c. Thus. to answer the purpose very well. When fully e is a steady force. we may algebrize (24) in two sight ways as before. People are always saying something.. tegrations performed upon p~ . merely remarking that the work can be done at after a little practice by using equations (A) and (B). and that like . beginning when = 0. to be done by (B) at si ght. extended in the latter case to fractional degrees. 41 Perhaps some people will say (as usual) that they do not "algebrize" that it is un-English. &c. and I will do it rather now. that we logarize a it delogarize and so on. to obtain a convergent solution. The odd powers involve complete inand t. a matter to be considered lat er. What is more important is that a word to express the idea of conversion from operational to " " seems a lgebrize algebraical form is much wanted.t>URE DIFFUSION* OF ELECTRIC DISPLACEMENT.

'f* + 2<w(28) . we know already that /. which is a time constant. so (26) is Also.~*1 = 2(0r)*.il. in rising powers of qa. except many terms have to be used.qa + <f? . and answers well. with limits lip A-y A // -3 where A = ESa2 . so To get the alternative solution. expand the operator in (24) V = -^L_= (1 1 + . converted to This that complete. Thus. is when t is big.


and then execute complete differentiations upon Thus. ) qae. so 9 4 4 V = (1 + fa + 2 a + 2 Here we have to find qa 1.{'-I This formula answers well for big t.42 ELECTROMAGNETIC THEORY. (29) which it. . Here the even powers of q contribute nothing (which simple a matter as it not so looks). . 7Tt 4'). is CS. The condenser the full value e . is known. G rt G + . and also when t is not so small as to rende r the initial convergency insufficient. VI.

when we put p = Theory of a Terminal Inductance. of the cable itself without much error. the current is continuous through the condenser into . It then falls the cable acquires to zero as the conOf denser gets charged. materially different from that of a resistance or a permittance. although the total charge is finite. so that the potential at the beginning instantly. the cable rece ives the same charge as the condenser that is. in accordance with (27) and (31). As a third example. and this is allowable when we which is desire to exhibit the effect of the inductance alone. But as the charge spreads itself over a condenser of infinitely great permittance.acts like a short-circuit at the first >pf moment. let its resistance be If really small the resistance may be merged in that zero. so that. by (13). That the fina l V is zero is also to be seen by the operational in it. we have '+"' . so that V = is the final state of the cable. according to Maxwell's now orthodox theory. let the terminal arrangement be an inductance coil. solution (24). For simplicity. its density attenuates to nothing. 244. the cable. The terminal Z is now lp. course.

. l p~ l in the first line. put g (fg). if 43 f3 = Z/R 2 S. *. a Thus (33) is the same as It is a time constant. we obtain and the known value of I 5. using (B) in the second line. and the odd powers complete integrations performed upon 1 To ease matters. Expand in rising powers of fq.. . a constant.. ^T. for KS/ .PURE DIFFUSION OF ELECTRIC DISPLACEMENT.so there is no new difficulty..^-r i > )^J(fjpy (34) + i +.9 .)* So...}. It is quite easy to obtain the convergent algebraical solution. (33) Here the even powers of q involve complete integrations on 1.7. thus -. V 1 fl -(' +J-+ J_+.

15 (35) It is not laborious to calculate the curve of V from this formula. at least up to : t =5 or 6 times g. I get the results in the following table t/ff .

The General Nature of Electrical Operators. The period ?n// J3.. The alternative formula is more difficult to obtain.. size.41 ELECTROMAGNETIC THEORY. It is -. which applied will then swing past its equilibrium position and oscillate The attached mass corresponds to the coil in the cable problem. the previous convergent series should be employed. the string be first in equilibrium. the string will follow suit in time. ourse. This illustrates the case of no terminal inertia in the cable problem. a matter t o be returned to. owing to the abs ence of inertia. when it is close to the beginning of the series. and the accuracy of ulation becomes impaired to the possible extent of the size of the smallest .. The descending series must be counted up to the smallest term but. is 4 of c calc term . more likely. I will merely give the result here. (86) which is useful in the later oscillatory part of the phenomenon. or. and back again to a l ittle below the equilibrium position. When the force is it will now take time to fully displace the mass. close to the fixed end. V becoming e immediately. and V becomi ng e everywhere in the cable later. The string will at once be transversely displaced to a distance V say. VI. and as its derivation from the operational solution involves more adv anced ideas than have yet presented themselves. Then apply a force e close to the fixed end. CH. of V and its passage beyond the value e to its first maximum. The oscillatory function in ( 86) arises from the infinite series of even powers of q in the operator when expanded in rising powers of q. but without any vibration. and the rest of . The table on the preceding page shows the initial rise about it. proporti onal to e. Bub next attach a mass to the string at the place of application of the force. to the extent of half its .

245. with a considerable amount . I have worked out the above examples (except the end of the last one) in a manner suited to one who has not done any work of the kind before.

45 of it But when the go is attenperceived. which admit of generalisation. <V= obviously. (37) side of and the potential on the right Z would be . then the current due to and the cable also a e would be resistance. that we first obtain the operational and that this is usu ally easily got and is of simple form at least in the examples used. the transforming work may be simplified by So now.PURE DIFFUSION OF ELECTRTC DISPLACEMENT. concerning the propagation of effects to a distance. Now. ing to problems of an elementary kind. processes. of detail in the transformations. in getting (24) for example. only replacing the resistances concerned by the appr opriate resistance operators. before passti on to certain rules which are obeyed. say Z. Later on. though treatthem as if they were still resistances that is. constants. a resistance. in the above. if the condenser were a ing . I interpolate some explanatory remarks about operators in general. say z. Thus. solution. we may be concerned with the theor y of fractional differentiation. Observe. the operational solution is got by algebraical of the same nature as if we were dealing with merely conductive circuits.

They become the resi stance operators. the current. because replaces resistance. tial through becomes V = R ZC when in the direction of C. h ave an equation V = RC for every We branch. in the real problem. V being the fall of potenNow . more generally. which connects the V and C thereof. If this there is stored electric and magneti c energy concerned. we it call Z the resistance operator. with meanings attached to Z and z.we work in the same way. Ohm's law applied to a si mple conductor. different . + V^RC.. which tak e the place of resistance in the equation V = RC viz. They are the functions of p. the timedifferentiator. That the Z's may be treated as resistances may be seen by considering the nature of the well-known problem of a conductive net of wires. (39) . and reduces to resi stance in steady states. or.

let them be arrangements storing electric and magnetic energy that is. Now. tions enables us to write down the expression for the current any branch due to the impressed force in the same or in any other branch. generally. an d the algebra of simple equaproblem the branches dut to all the of the current itself. an additional impressed force therein. in . then. VI. But 2 V is (40) zero in any circuit. and makes 2 C = there. more for every circuit in the net. When i t is a very complicated net determinants are useful but in most practical proble ms they are a useless complication. up along any path in the net. or. we sum 2* + 2V = 2RC. and the work is easier without them. true for any particu lar path in the net. instead ductors following of the branches of the net being simple conOhm's law in the above way. and is more instructive from the physical point of view. CH. arrangements of conde .46 if e is ELECTROMAGNETIC THEORY. and that is the circuital nature which connects together the values of all The the C's meeting at a junction. so 2<? we have (41) = 2RC This being the case. (40) being there is only one thing more required to determine C in all e's. is now determinate. we get If.

then. of a branch. and. There . Clearly. a complete any formal similarity between the problem of merely conduc tive is. though there may be mutual action between the constituents . 2V being zero in a 2=2ZC in circuit in the net.nsers and coils but still such that the current in any branch is the same at bot h ends. But the equations of voltage are changed. We now have 6 + V = ZC Z has to be found from its (42) in any branch. detailed strucAlso 2*+2V=2ZC along any path in the net. (44) therefore. (43) circuit. the currents. where ture. are subject to identically the same conditions of continuity. and such that there is no mutual action between one branch and another. though now variable with the time when the forces are steady.

Nor he stop there. for the very im portant case of simply need periodic variations can be fully investigated by a c ontinuation of the algebra from the operational solution to the algebraical. work out the given problem as i f solutions. Every The equations which find the C's in terms a Z. e'a are. we obtain an algebraical solution which may putting p cerned. It is then f ully realised. the result is the operational solution. It anybody can work out electrical problems of an advanced nature so far as the ope rational solutions are concommon algebra. rotation of vectors in a plane. The se equations are the operational So the rule is. only with the R's replaced by Z's. i signifies Geometrical methods are sometimes used. assisted by electrical ideas. therefore. For .PURE DIFFUSION OF ELECTRIC DISPLACEMENT. where the p/n or d/d(nt). the independent branches were mere resistances then give to the Z's their actual functional expressions in terms of p . circuits 47 R becomes of the and the general one involving stored energy. when a single simply periodic impressed e acts with fre2 quency w/27T. the pow er of _p in the operators is ri* so. by = ni. cation of the diagrams that arise when . involving the Their value seems to me to be is Their drawback principally illustrative. and constants follows that . by . be reduced to the simple form (a + bi) e. identically the same.

and can be used with advantage therein. Nor is it confined to electrical probl ems. even in complicated problems. with waves in conductors space. It is. howe ver. but applies generally to the mathematical sciences involving linear equatio ns. on the other to work out results. works with admirable simplicity. Returning to the network before cons idered. hand.the great compliwe depart from very simple problems. operators is not confined to condense rs and coils. and the hard thinking and labour required The algebraical metho d. only a special case of resistance operators. the e's being any functions of the The application o f these time in the operational equations. so as to lead to a resultant equation . in the general use of which we are not confined to simply periodic variations. its Z must be got by properly eliminating all the internal V's and C's. and dissipation in general. if a branch is itself complex. but extends to electromagnetism in and dielectrics. the ultimate reason being the linear nature of the equations.

(45) Bp put between two cables. by (B/Sj?)*.V l =e always. Thus we shall have A The impressed . It is therefore necessary that the current should be the same at both ends of Z. where V is the voltage on and C the current in the branch as a whole tha t is. Ej and Si being the constants of the new one. is in treating p and functions of p as constants for the time. The (Bj/Sijp)*. multiply by the negative of the resistance operator of the second ca ble. viz. however. a further generalisation is required. we can attack more genera l problems. force is put between the Z a nd earth. at the terminals. But should there be mutual influence between a branch an d some other one. which presents no difficulty save in the extra still of the same nature work involved. If Rj/Si^B/S. on the lef t side. . They are pr oportional to the displacement. VI. as it would introduce extra energ y. the current is halved by the substitution of the second ca ble for direct earth. of course. To find V 1} the potential of the beginning of the se cond cable. V = ZC. But if we put the e between Z and the cable. remark should be made here about the figure in 242. constant can be added to this kind of V. like the electrostatic potential. though V . If E 1 /S l = E/S. having no connection with our impressed force. so that the diffusion of V in the cables is repreNo sentative of the diffusion of the charge on the wires. CH. \e and numerical value. It should be understood that these potentials have nothing inde terminate about them. the potentials are e is .. which will make no diffe rence in the state of the cable in the examples above. a complicated arrangem ent of condensers and resistances like the cable itself.48 ELECTROMAGNETIC THEORY. when But in general they will not have the same \e. For Z may now have many branches for example. and to the charge. for they are really transverse voltages in the dielectric of the cables. in our problems. changed sign is necessary on account of the current being f rom one cable to the other. which. To find V multiply by the resistance operator of the first cable.

an d ginning <?. I remarked befo re that the solutions concerning the effects produced on the spot by an impresse d force were the easiest to investigate. provided the terminal conditions are of the simpl est kind. which is kn . 49 The Simple Waves of Potential and Current. bewhen t = 0. write V = (1 shin qx)e and since this involves complete differentiations performed upon ql. Go back to 240. due to V at #=0. Taking V = constant. To find have the operational solution This expands to V at a. effects now pass to some simple cases concerning produced at a distance. so there is nothing in the way of operati ons.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. This is true when the constraint on the spot (or terminal condition) is not 246. Let us But some cases of the effects produced at a too complicated. we may discard the even powers of ^/. distance are quite easily examined operationally. = (cosh qx new shin qx) V . (2 ) Here we have even and odd powers of q. we (1) V = e-"V .

II. both and as the basis of other fo rmula. If we differentiate to x we shall obtain the formula for the current. Thus e VOL.own. so we may as well give some details about it. . the full algebraical result follows at once : + This is _ ' in itself an exceedingly important formula in diffusion.

(9 ) So we get EC = cosh which is . CH. we obtain (7) which gives rise to (5) or (6) on development. We may But if we please we got (5) by differentiating (4) to x. viz. excepting 1. from (3) tional forms. Or we may start from the initial operational solution C.. because (on account of th e q factor) the even powers of BC = (cosh qx . this makes In developing the constant term q go with it. thus. which is another important formula in diffusion. for EC = <r^KC = <r**ge.50 ELECTKOMAGNETIC THEORY. But now we must reject the shin function.shin qx)qe. V we rejected the cosh function. so we may write it finitely. get it in the same way from any of the previous operaFor example. VI. (8) On expansion. Here we recognise the exponential formula.

the terminal condition . that (2). but that there is a consistent fitting together.qx . 247. i s troublesome numerically. on account of the very slow convergence when t is sma ll. qe. the time conV and C are zero initially everywhere except at the origin. The Error Function. is =y . But the functions are well known. dent. : the characteristic dition. so there is no need to be frightened. and C are the solutions T hat these solutions for V may (1). however. Short Table. be tested by their satisfying the necessary conditions . (OA) I give these variations to let the reader see that the solutions do not arise by fortuitous acciequivalent to (7). The last. (3).

2 . e a function of RSit-2/4* * 3 5 II 711 . So we may write it say. observe. The V/* formula (4).

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. enough." Vol.." going by steps of 0-01 from y = to 2. pretty full table was given by De Morgan in the Ency. III. and denoted by erf y. 51 which is the same as (11) = l-erf y. (12) A is comparison of (11) and (12) defines the function which sometimes called the erro r function. A Met.. ** Theory of Probabilities . table is all that is tracing. This table is reproduced by Lord Kelvin in his " But a much briefer Physical Papers. p. and for curve Perhaps even step O'l would be y- . say with step 0*05. needed for general purposes. 434.

OZ ELECTROMAGNETIC THEORY. or 2 kilom.. and the ordinate is the The curve 1 is got by making KS = 4tf. . 1 is represented the way the potential (or the charge) extends itse lf into the cable when the potential at its beginning is raised to and maintaine d at a steady value.000 we t' like. things length from the beginning. CH. ing x... They therefore serve to plot the curve of the deri vative. 248. represented by formula (6). The Way the Charge and Current Spread. The abscissa potential. VU equal steps 0-05 in y itself. which may also be used directly. and varyOr if t' = 4/RS. then curve 1 represents the state oi is t' when = 1. We may use any unit of length we please. that is. so that the base line kilom. if may be 2 centim. the curve of current. or 2. In Fig.

Doubling x again for potential distribution when the same values of V/e. the same values as before wit h doubled This is shown by curve 4. representing the values of x. showing the Now V/e is So when = 4 we have ' = 16. showing the potential when t' = \ and ' . potential when we obtain the curve J. The initial potent ial curve is simply potential when t' the vertical line OV The line final potential curve is the vertical (up and down) and the base line 012. OV and the horizontal V A. = 4.2 unchanged by altering x in the same ratio as t'. by halving x in the curve 1. . we get the curve 16. showing the = T\-. Similarly. halving x again brings us to the curve f$.

On removing the impressed force. and the shapes which follow are to be seen by inv erting the diagram and looking at it from behind. The When t' = l. 2. moving in a medium which resists its transverse motion with 0. and 16. we have the curves marked with these numbers in Fig. we have the shape given by OV A. 4 is got from curve 1 by halving the ordinates and doubling . 4. 2. It will be seen by (6) that by So curve doubling x and quadrupling t we get a ha lved C. and be initially along 012. But the exact mechanical analogy before employed requires an elastic massless string. The string should be fixed Then a force applied close at to FIG.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. The curves of current may be g ot by plotting the A's in the string will table. inertia as will stop oscillations. when the e nd of it is there being enough friction or as little to lifted from We may imagine the V . the fixed end should stretch it (like a piece of " elastic") to V . then move through the different forms shown. 53 by a rope curves to represent the shapes assumed resting initially on the base. a force varying as its velocity.

init and earth. the impressed force we generate instantl y a definite amount of magnetic momentum at the beginning of the cable. For a purpose to be seen presently. Theory of an Impulsive Current produced by a Continued Impressed Force. the reason being is constant. The pot regards the other curves. The Similarly as belongs to t' = ded by the curve In another form. let the imforce be given by pre ssed <"> where Q is a constant. that is. An exceedingly interesting and instructive case arises when serted between the impressed force at the beginning of the cable. and attenuates to zero density amount. the line-integral of the current that . Before = the cable is to be understood to be uncharged. because we may suppose L to be finite. unnumbered one. (13) at That is. is variable w ith the time in a certain way. Vo-V-fRCda J and V = x = oo . These curves all bound the same area. becau se we have But that of the use of the idea of magnetic is nothing in the way momentum. although so small that the diffusion law of propagation is followed practically. ELECTROMAGNETIC THEORY. when we apply by Ohm's law. It is true that the zero in amount.54 the abscissae. CH. 249. VI. which th en diffuses itself along the cable without alteration of total magnetic momentum is assumed L = in the theory. the area boun and the vertical and horizontal axes is constant. with the top cut off. J.

by (A). it is where the second equation definition of f. Q by (15). By (5). of course. and the third by the constant for any finite value of That is. (15) Now find the current entering the cable due to the impressed force.ential V is raised to the value e. It is the same as V = ^Q/S. Since time. the result is zero. 241. 240. there is no current entering arises is .

any function of the time. (16) are not wanted by an experienced worker the operational solution BC = qV of 240 bein g sufficient to show that if V is ql 9 then C vanishes. and it is essentially the same here. . of total amoun t 1. or at least indicative of som e modification of treatment being desirable ? Not at all. That which is wholly concentrated at the moment t=Q. then then. It is an impulsive The idea of an impulse is well known function. . 55 the cable under the action of the continuously-present impressed force at any fi nite value of the time. pQ is t amount is Q. so to spe ak. Unlike the It is = 0. as in the present case.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. It is the fact under the circumstances stated. Is this nonsense ? Is it an absurd result indicating the untrustworthy nature of the operational mathematics. (15). pQ is zero except when to note that if We have is its Q is rate of increase. For the above details concerning (14). in mechanics. Q zero before and c onstant after t=0. and the principal remarkability is the instantaneous arrival at the result.

If. impressed force at the first moment the amount of the charge is Q. pi means a function of functional. that a charge. taneously given to the cable at its beginning. but involves only the ordinary ideas of differentiati on and integration pushed to their limit. and that ther e generated by the of its application that . which charge then spreads itself without loss anywhe . is no subsequent It is the same as saying that the charge Q is instancurrent. the functional does not involve appeal either to experiment or to ge neralised differentiation. But its total t is to say. then infinite. Our result C =pQ is therefore is means that an impulsive current.

for V and C anywhere in the We have.re. so the (18) By differentiation to x. equations result is (8) and (6). by (15). (17) This was algebrized before. the current is r_ _1 dV _Qx "" /BS\* -RS^ . Next work out the solutions cable.

CH. its expansion.56 It will also ELECTROMAGNETIC THEORY.. Thus which leads full directly to the formula (19) above or. be useful to obtain the current formula directly.\qxQ. We have. rather. by (16). v li li (22) involving only complete differentiations upon ql. . VI. gives We may also note that (20) or (21) tiation C by a time differen- . and we get which expands C=-p(l + M\(3*f+.. C = c~^ x C Here reject the ~ *x < vQ (20) even powers of to q.

(1) and (4) This again leads to the result (19). But if we introduce in the new wire an impressed voltage equal to and acting against the former voltage. That 6-9*1. and two points A and B thereof be joined through an external conduct or. If there be current in a network of conductors. the removal must be done after its initial e ffect in in the above case charging the cable. everythi ng will go on as before. Applying this to the cable.upon the function above. Of course. and C are conthe previous special results for V tained in the general formulas for V and enough that is. they make V =0 and C anywhere is C at x=Q. there will be a derived current in the new wire usually that is. there will be no current. . already obtained. no current as clear We is have now only to explain why there initial charge. is after the It is for the same reason why there no current in the galvanometer in Poggendorff's way of comparing two battery voltag es. if we remove e and insulate the beginning of the cable. and everything will be the same as if A and B remained insulated. if there be any voltage between A and B due to the origina l arrangement. we see that the impressed force is so artfully gradu ated in its strength as to be exactly equal to the potential at the beginning of the cable due to the charge Q when redistributing itself without terminal loss that is to say.


according to the curves in Fig. Arbitrary 250. If the cable be infinitely long both way. and have a and C charge 2Q sudden ly introduced at # = 0. It has been mentioned already that areas are conserved in that diagram. dismissing the impressed force altogether. Source of Electrification. In the present case this means that the electrification is conserved. one of whi ch will go to the right and the other to the left. On the . so that where x there is no current at the origin. The potential (or the charge) redistribute s itself in the same way as the current in the previous problem of a steady impressed force. and then immediately insulate it. because the charge 2Q will split into two equal charges. p. that is. the potential and current which result will be given by (18) and (19) above. we see that to the beginning of a cable. 57 if we suddenly communicate a charge Q So. Diffusion of a Charge initially at One Point. V is positive. and obviously in a symmetrica l manner. the resulting will still be given by (18) and (19) on th e right side of the origin.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. 2. 53.

electrificat ion is at the rate 2k per second. Now. terminal datum sequently for the positive half of the cable. Thus. Con(23) C = e-^ and since gives the current at distance x.left side. let a doubly have an auxiliary wire attached at the point = 0. V will be the same. EC = aV. without the use of infinite cable . This splits generated at z = and left. and C the same negatived as at the corresponding We shall also obtain these results points on the right side. Here h may be any function of the time.i* . (24) is the corresponding V. through which current is artificially sent into the cable. It i s equivalent to a source of electrification of strength 2h that is. shall if h is impulsive. e. so that C Q =h may be taken to be the equally right hi another way. Let 27* be this current. acting only at the ' have .

we .^^ moment t = 0.

may. and the p otential and current anywhere have been derived from . YI where (24) Q is the time-integral of h.53 ELECTROMAGNETIC THEORY. regard the matter and C) or. formally the previous statements about th e The Inversion of Operators. however. . Putting this in (23) and we obtain the e. V V . given differently and C) . 251. given C find C (and as. Simple Examples. in all cases on th e understanding find (and that the state of the cable depends entirely upon the state afe . This proves behaviour of expressions (17) and (20) for V and C. so far as the general electrical property of balancing voltages entered into the explanation. We V . ifc. We have hitherto usually supposed that the potential at the beginning of the cable is produced by an impressed force there. CH.

one expressed by whilst the other cons ists of therefore not subject to change. with a node between them at t he origin V=e side of the itself. and allowing it to discharge freely into the cable on the positive side. So the potential at the origin will be made permanently equal to e. and distributions. Then we may find the . and the current is uncontrolled save by its natural connection with the potential. and that the cable is not subjected to impressed force or allowed to receive electrification anywhere The state at the origin need not. but may result (for instance) from its connection with a continuation of th e cable on the other side of the origin. This dependence upon the state at the or igin implies that V and C are initially zero except at the origin. impressed force th ere. at the origin between the line and But suppose it is the current at the origin that is controlled and led through a sequence of values. this continuation being initially charged. be due to an else. the case of a st eady impressed voltage e between line and earth at the origin may be imitated by having the whole of the imagined continuation on the negative side initially charged to the unifo rm potential 2e. therefore the result on the positive side must be the same as that due to e considered as an impressed force acting earth. V on the negative and V= . Thus. For the assumed distributi on is equivalent to the combination of two = e constant all over. however.V the origin due to a cause acting there.e on the positive origin.

and V co mes out constant. we can immeV for example.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. problem . and is . to -*. at the origin. the q in the denominator of (26) is cancelled. . C conveniently. We found before that V conSo now we solve the inverse stant ma de C vary as t~*. diately find So. that is. we have in (26) general C is proportional to ql. by taking If. Thus. that t is. if C is zero before t = Q. potential that results there 59 and elsewhere.

given The result was that C was initially impulsive. where The V Q is constant. to find C . we see that the state of thing s everywhere is determined as well by the condition C oc t~ as by the other cond ition V = constant.-p/. since to vary as 1. Next. (14) to (16). that results is This is the inversion of the problem beginning 249. (28) is the measure . where / (when regarded merely as a constant) of the impulse. suppose C varies as </ before seen. as A gain. therefore pQ. Then C is impulsive. say V. which was. equations varying as t~t. V are derived from C and V by the operator t-* *. 2 l.then made the potential V will be suddenly raised to and be maintained at a constant value . so that we may call it being the charge in the impulse. and zero later. suppose that V V is impulsive. Furthermore. and 1 They are equivalent under the circumstances mentioned. that is.

and that it goes. and transforms functions in a remarkable manner. . It has. The reader may have noticed in the above.the time-integral of V . a very wide application. howeve r. The result is (29) which indicates the current coming out of the cable initial after its charging. as in f(p)<f>(p)l = <(p)/0p)l. I do not assert t he universal validity of this obviously suggested transformation. that we change the order of operations at conBut venience. and perhaps previously.

the solution arising from a steady impressed force. f being the timeThis process is general.CO ELECTROMAGNETIC THEORY. then C impulsive. of course. the moment of time of solution . Or. VI. and substituting the strength o f the impulse for the intensity of the former force or source. giving C in terms of any e through the operator Y. or a steady made source. of * = Yp/ is the operational total/. we obtain that for a momentary force or source by differentiating the former solution to the time. If we have worked out tota l of e. The present example is. Reservations should be learnt by experience. elementary. if C = Ye is the primary operational solution. CH. Observe that the current due to impressed e when it is impulsive is got by writi ng pf for e. in another fo rm.

to the latter by a simple diffe rentiation. is of little use unless the integration can be readily effected in the case of the special function of t he time that e is chosen to be. It is generally better to work out a solution due to e constant than that due to an impulse. at the origin. Effect of a Steady Current impressed at the Origin. the solution for a continued force varying anyhow with the time is at once expressible by a definite integral. knowing the developed impulsive solution. which allows of a ny . there is to be no c urrent before. That is. But. and others may The be given. Clearly this wou ld be impossible were it not for the permittance of the cable. is an obvious example. 252. let C be made constant.when e is the impulse being = 0. however. or by derivation from the other. as above. either directly. because the former leads. whereas there might be some trouble in rising from the developed imp ulsive solution by integration to that for a steady force. it is not able uncommon more for the result of the integration to be obtaineasily directly The simply periodic solution from the operational solution itself. and a steady current after e = 0. The de finite integral. beca use the continued force may be regarded as consisting of an infinite series of successive infinitesimal impulses. Moreover. Returning to the mutual dependence of C and V and the inversion of operators.

practical limitations finite strength of the d ielectric being nowhere .amount arising from the of electrification being sent in.

if sufficient time be allowed.PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. it is worth while algebrising fully the solutions when the current at the origin is steady. in a theoretical discussion. The potential at the origin is by using the known value of jy~H. varying intermediately as *-*. which leads to zero final current. though it is initially infinite. From (31) . 61 rises Of course the potential infinitely. the as ti. what is the resultin g current ? This contrasts well with the case of a current is steady. we may put it thus impressed force at the orig in between line and earth varying The answer is. As a t ypical example. : steady e. For in our case we get a steady current from the first moment by present an e rising from zero to infinity according to $. Given that there is an Inversely.

1. Those curves now show how C . From there being initially the final state that it is C everywhere. distributes itself in the cable. to repeat it. due to steady as same law as was worked out for V was exemplified in Fig. As regards Y. we have . no current anywhere save is at the origin. The operational form (31) is to develop is unnecessary fully explicit and understandable.C = r^C we see that . where tended to is a constant current it The formula being the same. C follows the e.

V = --<*C 9 . which is obviously not the case in the present problem. (82V Here we get reject. it would partly express the final steady state were such a state possible. since the next term is (R/g)C Q which tends. as in* previous cases. C . the even powers of q. .IfcC + 9 cosh qx . and V. (33) where the term involving the first power of x comes from the term in the previou s equation which is independent of q\ that is.

CH. tion to x. . is nothing new in the way of operations Written fully. obtain t he current by differentiainto the cable.. (84) which only requires the complete differentiations to give the full development. and how it spreads As a check. it is the same as L_ .) g<y 2Co \b7T/ . VI.. to be effected 3 o & I \ O I Q O 7 / i I tt A where yJ -BSz?/4t. as before. to infinity.654 ELECTROMAGNETIC THEORY. There in equation (33). This formula shows the infinitely great ultimate increase of V. namely.

Jx (37) so to get the missing term (that involving ti ) we require to evaluate a series . on the other hand. and attempt to obtain that V with the last formula by integration with respect to x. If. Comparing with (10). the law followed by we confirm C now being the previous statement about the same as that followed by start for Y when e is steady. The term indefinite integral is the formula independent of x.We get o. we should for C. (35) without the Now y( *ECdx. there might be some difficulty initially.

already known. The work is done automatically. That is. and also when t = at any x.for an infinite value of the argument. it is known by the operational method otherwise it would have to be The operational method usually avoids specially obtained. a nd made to vanish at x = oo for any t. as was V . which But instead of that we may note tha t is inconvenient. V which we may see by the way (35) is got. as it were. at the origin. (38) so by this formula we get is what we want from the value of . auxiliary evaluations of this kind.

G3 Fig. the new formula (35) requires ca lculating to show how the charge Without making a fresh table fully. Write (35) < where II N/2 + the former to 2= RSz 2 /4. the "error function." the following special results which I have calculated in the may draw the curves roughly. should he care form be quite sufficient to allow the reader to to do so.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. 1 serving to show how the current spreads. and the 7/ 2 of those of ?y are as Then z is a z function of values of follows : corresponding some y 2 1 z . like that for spreads.


In our present case. and Maxwell showed how to carry on the process with m ultiple poles. however. or the magnetic be maintained finite. way of making the magnetic force of a magnetised molecule out of the combined ma gnetic forces of two equal poles of positive and negative magnetism. But if. the differential effect will decrease with At. so as to generate solid harmonics of any order. The fina lly resulting C is is . VI. Now. But if the product of It is like the old strength of pole into distance apart. equations (23). it will nearly undo the effect of the first one. If Q be finite. Q is infinite. but a time distribution. (25). whilst their product finite QAt the result takes a special simplified form. if this impulse be followed at time At later. as before examined. where At vanishes and is finite. the differ ential effect is maintained of finite size. and in the limit. by the impulse ->Q. and takes a special form in the limit. and become zero when it does. whilst decreasing At. their joint effect vanishes when they are bro ught to coincidence.64 ELECTROMAGNETIC THEOHY. the negative of the former. The " moment" of the differential imp ulse is QAt. 250. This is the way to form a solid harmonic distribution of potential. it is not a space distribution of poles that we ar e concerned with. the resultant magnetic force moment. If the poles are kept of finite strength. is finite. where Q the strength of the poles (or impulses). and At distance apart in time. we simultaneously increase Q in the same ratio. CII.

= QAt.?. if Q. The potential corresponding to (42) is obtained in th e usual way by multiplying by the operator B/.is their where At is infinitesimal.. This makes let (43) . It is clear that we may extend the above to impulses of any degree. but at present us keep to the middle one. say &c. (42) C= is ~VQi the current due to a differential impulse of the first degree. Or.

But we must understand pql more is literally. t Thus. = Qj/S 1 for simplicity. which is its rate of increase. is zero before oo. however. then. acting when t is and. ql is the function of the time which later. as before explained with respect to the impressed force e = ql. serving (in Poggendorff's way) to preven t current this Now at the origin. by algebrising pql we get Let finite. merely represents the dregs of e. 65 from which we see that the impressed force at the origin between line and earth required to produce the same effect is given by This. needs close and literal interpretation. zero before = and to is is (B8/*)* So pql. .PURE DIFFUSION OF ELECTRIC DISPLACEMENT.

going back to the impulses without any residual effect. there cannot be a complete neutralisation. p*ql is to be interpreted as the represents it all. pql Similarly. only the last part that developed time-function in explicitly (45). and (42) from (20) or (19). and . it is unnecessary to write the full developments. r ate of increase of pql just considered . the same time the positive charge in front di ffuses itself forward into the cable. (43) from (8) or (6). is therefore a double impulse. ^Q! we see is the time-rate of increase of ^Q!. These tend to neutralise and do neutralise But since at one another by m utual diffusion to a large extent. then jumps through zero and lastly rises to zero again gradual ly. and consists of a positive wave followed by a negati ve one. all to the right thereof being positive and to the left negati ve. whilst the density Since the formulae (43) and (42) rapidly attenuates to zero. viz. . for V and C are der ivable from previous formula by simple time differentiation. Of course it is = 0. VOL. As regards the cha rge in the cable produced by the impulse 2 p Ql5 just after the first moment it is confined practically to the region close to the origin. first positive and then negative and so on. The result is that th e region occupied by the charge is continuously enlarged. On represented by the the other hand. and so on.then jumps to oo. And. ^Q has been already inte rpreted as a simple impulsive current next. F . n. the node between the positive and negative charges advancing along the cable.

since the operator potential KSp is equivalent to a double differentiation to x. From it follow vari ous other results. it is . so far. CH. where g is the square root of a differentiator. . sufficient merely to point out.66 ELECTROMAGNETIC THEORY. how they may be got. Our results. Moreover. Similar remarks apply to multip le impulses of higher degrees. as is indicated by the characteristic of the and current. Convenient 254. as regards their development by differentiation s o. VI. the new curves of V and C may be found by making their ordinates be proportional to the curvatures of the old ones. as we do not want them at present. Way of denoting Diffusion Formulae. notably the functions ~ qx l and^e~ sx l. as above. depend upon the fundamental function ql.

Since they represent the elementary waves of potential and current due to a steady impressed force. let for shortness. Equation (1) bel ongs to Fig. beginning at = 0. in more advanced problems that the preceding. 248. when V is steady. their meanings should be carefully no ted. Then we have 1 -erf y which gives the V/V curves. they are of particular importance and since they form the elements . and which gives the C/V curves. relatively simple nature. Mathemaworking with the short expressions that when they get solutions in terms of one or a finite number to accustomed . the series functions receive special names and have conventional short ticians get so expressions. Most solutions of problems in mathematical physics are in the form of infinite s eries. and the moment equation (2) to Fig. Finite solutions are quite When of fundamental importance and of a except ional. collecting some formulas for reference. Thus. 2. 1.

through the operators generating them. had not received special names. proper forms for the easy and immediate perf ormance of operations on the functions far more easily than upon either the fall series or the exp and erf forms. Now. in (1) a relatively unfamiliar function which see ms a great deal more transcendental than the other. be an undesirable d eviation from common practice were no advantage gained. we shall see that they are made up and e. 67 of such functions they sometimes say that they have got the But if the functions were such as solution in a finite form. in fact. Considering all these things. but as structural formulae. and by using erf we have a condensed empiri cal form for the functions we are concerned with. The name erf?/ for the deficit from unity was invented by Glaisher. they are the forms which present themselves Furthermore. serving not merely as empirical a bbreviations. . the case would be substantially The difference i s merely conventional. granting the desirability of having special short ways of representing impo rtant functions. Practice confirms this conclusion. when we pass to more complicated of the e~ 9Z l cases. for example. shorter than the other ways. we may remark first that x (Tqx l and e~v ql may be themselves regarded as the special This would short ways. as will be evident in the following.qx ql functions. which of e"5*! is its slope. In (2) we h ave a case of the (1) familiar exponential function. Lastly. we see that there may be great advantages i n using these forms in their naked simplicity.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. But in the present case the forms in question are actually indicative of the functions themselves in their structural meaning. Compare the same. occurring in t hese particular forms. Moreover. and (2) above. they ar e the naturally in the mathematics.

and exhibit the essentials in the simplest manner along. the easiest are those in only one reflected Let. When we employ an infinitely long line we do away with reflected waves. conditions prevailing are the same all of the V and C waves generated wave. so feasible. The no change occurs in the behaviour at a source. passing which there is to more practical cases. 255.Reflected Waves. Cases of Simple Reflection. the impressed force e be situated at F2 . Now. for example.

The currents. The potential V generated by e is V1= + je-*"-**. and 2 to the left.68 ELECTROMAGNETIC THEORY. CH.. Ed = i?e-^-<% which follow from (3) RC = \qt-* a-*e. and the two potent ials +^e and -\e are sources of elementary waves. left. the point x = a. however. Vi to the right. V= 2 on the on the right. (3) (4) It-^-^e. and the beginning x = be earthed. are the same at equal distances V Y from e on either side. viz. whilst the end of the line is at an infinite distance. 2 . VI. because there is a rise of potential to the amount e at the the place of e.

\^-qae. The positive wave Vi to the right suffers no change. is the reflected wave. or by . attenuated to . wave at the origin. except what is involved in its known expression. this would mean that . which.(5) and (4) by space-differentiation. side and to the source on the left side. therefore. and is. 3 (6) This right. Since . and because C is reckone d + from the source e on the right left. But with the negative wave V 2 it is different. When it re aches the origin it has gone through the length a. But the potential is constrained to be zero at the This requires origin. by the previous.~qae raised to potential +\ at the origin. means a wave V = -^x Je-^.q on the using the operator q on the right side of e. the line is + ^c~ qae Taken by to be superposed on the negative itself.

only the constants being different. we of the time. These waves are of the former kind precisely. or reflection. If e is steady. 3 and (8) V = V.)' on the left side.is it suffers and is no further positive. from left to the complete potential on the right side of the source. + V = -i e -^-*WK*(a+a. The emay be any function may employ the developed .

(8) ) 69 The and is V=e final result in this case (by 2 side.PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. Note that when e is shifted up to the origin (by making the two positive waves a re made equal and coincident. their making a single wave of doubled size. This makes C = there. V= V= +J -*-- . the case of e at the origin (earthed) previously treated may be regarded as the rt = 0).tx = a are the same as before. but the reflected wave differs. Another easy case is when the line is cut at the origin. because it has to cancel the current due to the prim ary wave V2 So co-existence . The initial waves "V^ and V 2 from e a. It is the negative of the former reflected wave. = in left on the right and V= on the side. formula (7) (1) above. whilst the negative vanishes. That is. tically th3 case of instantaneous generation of a reflected wave of idensame nature as the p rimary wave to the right.

e on the left and = f inal state when e is steady is The on zero. After the above easy cases in which only one reflected is generated. = . (right) (left) (9) je-*"-*^J-* V fl+a!) (10) are the potentials when the line is cut at the origin.fl j +*. pass t o a case more nearly allied with . *. V the right side of the source. Both Ends. The final current is These steady results are obvious. Line Earthed at wave 256. An Infinite Series of Reflected Waves.

The (11) initial wave from A V --*. and has attenuated to ~^<?. Call these A and B for descriptive convenience. Have We will first build up the is result in a physical e on manner. and be earthed at both beginning x = and at the end x = I.c^e has to be superposed to cancel the effect of the first . in all of which we have an infinite series of reflected waves the same type. But when V l reach es the end B. were it not for the constraint forcing V to be zero. So V 2 = .practice. it would raise the potential there to that value. Let the line be of length I. X and there would be no other if the line were infinitely long. at A.

This is .wave.

r^ ~ x) x c^e ff . CH.x. so that. and therefore has attenuated the second wave is .. VI. The third wave is therefore + e~ . I wave starting from B. this It When . kind of reflection as at B. and At distance x from A it has traversed the to . viz. generally.70 the value at distance so ELECTROMAGNETIC THEORY.~Zgle. So we have the same But the potential is kept constant. V =-e-^-*V 2 (12) wave reaches the origin A. it has attenuated to would lower the potential there to that extent. z<jl with reversal of potential. B of the reflected going towards A. at A.

like V^ When it expresses the third wave. .. The fourth wave is th erefore . The current to correspond is given x by ( RG = (qt+ qt< *W. reaches B. because the the same for all the succeeding wa ves.)e.~*qle at B. and (in the same way as the first 3<2 wave) generates a fourth wave with the potential reversed.. and becomes V= 4 . (16) .e It is positive.x ) + ge. The is complete potential therefore V=y1+ v +v +v 2 3 4 + . it becomes e~ ^.. It is is process unnecessary to elaborate further.lW+^+..-*-*>* (14) at x.

but the waves are alternately currents to correspond are If e is steady.where it should be noted that all the signs of the waves are + That is. . above. and The successive waves are smaller and But the above reasoning is the same whether e be steady or be any function of th e time. so the above results are fully expressive of the solutions in general. (15) gives all the same way. of course. smaller. in a series of elementary waves (16) gives C in terms of the V of the kind (1) functions (2) above. if the i nitial wave produces positive V lf the reflected and + as regards potential.

we get . summing up.-_. because the operational method gives the waves automatically and easily. The potentia l is kept down by the reflections. and should certainly be But when understood. to obtain (17)._.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. much of it may be take n for granted. . followed sometimes. we The above way is . tional solution. may readily transform the above solution to a compact form. whilst at the same time they allow the curren t to as is rise to a steady value. Putting g = steady state we see that a steady e gives the final obvious by Ohm's law when the cable has become permanently charged. We and the even terms make another.-(-> !--* This is S hin. So. instructive. Thus.yZ the condensed and most convenient form of the operain (17). The odd terms in (15) make one geometrical series. 71 understanding that the waves will have different meanings quantitatively in different cases.

They and are determined by the terminal conditions V = at x = I. Expand it by division. We arrive easily and speedily at the condensed form. and inserting them in (19). (20) Finding F and G from these. and in fact have to be determined. and the developed wave solution (15) results. = F + G. This i s the practical way to work in more complicated cases. (19) F and G are time functions (constants as regards x) to be determined. because t hey are determinate. which give e V=e at x = 0. we obtain the solution (17) at o nce. That the development of the .have the general solution V = e**F + where -<*G. We should not call them " indeterminate. develop it if and may then we like. = *F + z G." as is sometimes done in similar cases. or deterniinable.

x having to shows the wave train in an it is of finite length. VI. the value of P being by using P = (JRSn). We have e-** sin nt = e-ra: sin (nt . Equation (21) cable. may sum up this infinite series if we please. CII. we obtain an infinite series of simply perty p = trains of waves. i'=-. If the 2 is simply periodic solution be got from (15) by the pro. . It may be that only the first.72 ELECTROMAGNETIC THEORY. We . are wanted. When members infinitely long the type of the individual of the infinite series of wave trains . operator in (17) into elemei<tary wave operators by division legitimate.n2 orp = ni. receive the series of values indicated in (15). or the periodic first two or three. is obvi ous from the above.Vx) (21) when p = ni. If so. this way may be useful.

It is usual in heat problems involving refl ections. act simultaneously in the same direction. it in the condensed complicated way. due to the pair of forces at 2 and 3. to consider the extra terms to be due to images. 257. In our present problem . another way of regarding the matter. each of strength 20. But if this can be got much more easily by making a operational solution. because the summation has been already The Method of Images.and obtain a resultant formula in p = ni resultant be wanted. So is that due to the pair 4 and 5. Let the dots divide i t into equal lengths I. Then the potential at A. is zero. The potential at let and . Let the straight line represent part of an infinitely long cable without any external connections. and so on. There is The Waves are really Successive. say from left to right. effected in it. it would work out thus. At every second point 531246 AB I i I ! ! ! ! I impressed forces.

Now conside r B. Nor do the forces at 3 . . is.A is right at 1 therefore that due to the 2e there only that and e on the left of A. e on the forces The and 2 produce no potential at B.

3. It will also came into existence one after the other. The waves are successive. but is only a mathematical fiction in reality. it as easy to follow. But the way previously followed I think. This per accou nts for the contemporaneity of the waves in the diffusion described. if not easier. 2. (17). preferable. is. in the way i< The speed of propagation of disturbances is is the inductance or (LS)-* approxim ately. Our ignoration of L makes v be infinite. and 4. says This is right enoug h for the formula. It But more importantly. 5. on the other hand. In the figure these mark the places of the is 1.PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. A and B. be noted that I described these waves as if they The formula (15) or that they are contemporaneous. but introduce the reflected waves as it is. but earthe d potential is the same as if the at both A and B. and do not leave it. 73 and so on. But the real sources are at source and its imagined images. &c. We The order of the waves just as they arise at the terminals. the line were AB only. and with an impressed force e inserted at A. = (^ c )-* L . This is is a mathematical equivalence. The result is that between A and B. where centimetre. 4. take the line represe nts the true physics of the matter.

how small. be rational as well as practical. property but should be taken with salt in making app lica. It is very great in some telephone circuits. however little difference it may make quantitatively in general. with Hertzian waves. great. To make the theory of heat diffusion tions to real physics. of In the more advanced treatc ourse. some modification of the equations is needed to remove the instant aneity. But give L a finite finite. we have merely to take L into account. All diffusion formula (as in heat conduction) show instantaneous actio n to an infinite distance of a source. and v is By taking L small enough. no matter are successive. We then change The change may be little or the type of the waves as well. though only It is a general mathematical to an infinitesimal extent. wh ilst being successive. . and. to rationalise the theory in our immediate problem. they will differ as little as we please from the above waves in type. and the waves value. Of course. So we are justified in using the above natural way of des cription.formula.

or without reversal of the potential. VI. including CH. and the waves are of another kind. Therefore V3 is negative. ment we have is only one exception to the rule lectrical arrangements made up of t is possible to comtheory of the trary wave by means of a suitable ed waves. L (and K the leakance as the same kind of wave analysis as above. So the V2 differs from the old one of that an infinite waves results from terminal e a finite number of parts. Then there are no reflect This V2 is equation (12) in sign only. and since it is reflected positively at B. namely. makes a series : cycle of signs. 258. This So we now have the following V= writing identically the ++ -++ . where th e impressed force e is situated. earth at the beginning A. V4 must be negative too. just th ough q has a different meaning. If the line is insulated at the far end B. reflected negatively at A. and that is in the I distortionless circuit. but it must now be reflected at B. other things remaining the same. pletely absorb an arbi terminal resistance. in order that the current may be maintained zero. instead of being earthed. There series of Reflection at an Insulated Terminal. The initial positively sign of the second wave wave (11) from A is the same as before. as in 256. the change made in the waves is very easily settled. being now positive. of the subject.74 ELECTROMAGNETIC THEORY. well).

which are otherwise same as in equation (15). The final V is e at all over. down only the signs of the waves. arrangement Summing up obtain as before. by the law of geometric series. V and C *= .-++-&c.. we which is the condensed form of the solution. as is obvious. generally. and the final C is zero. where they make the H h .&c. To test (22) = e at x = 0. observe that it makes x = l t which are the two termina l conditions.


the case is a little more complicated. going both ways. = a. as two earths . 75 General Case. At an earth. of these considerations it is eas y to write down . or one earth .PURE DIFFUSION OF ELECTRIC DISPLACEMENT. with any Te rminal Conditions. say at x not at the terminal A. The primary wave is dou ble. Here a and /3 are the terminal operators given that . or two insulations. it will be as well to take a more compr ehensive case for illustration. As there is no difficulty in this process. and each of its members suffers reflection at A and B. then /3v is the reflected wave and that if v is a wave incident at A. V without further calculation the full solution in terms of the any case in which the terminal conditions are earth or insulation. because there are two series of waves. 259. When the impressed force is but is at some intermediate point. the reflection of By means waves for is positive at a disconnection it is negative. and one insulation (two cases) and these solutions are fully ~ qx realised algeb raically by the functions e-^l and qi when the impressed force acts steadily. Let the terminal conditions be unstated except that it is if v is a wave inciden t at B. th en av is the reflected wave. Effect of an Impressed Force at an Intermediate Point. .

to obtain the value of the last at distance . the reflection. the coefficients ot Now let find its effect at a point there be an impressed force e at x = a. to obtain the value of the reflected wave Finally.x from B.defining the nature of reflection. going to the right is Vi = -<K*-a) J tf> (23) In this put x = (3 I to obtain its value at B. so to say. Then multiply by r 2 at B. that is. The result is I - . and let us x on the right side. z) multiply by e-^. at x from A. A a x B The wave from e or between the impressed force and B.

going towards A. .> J*. (24) which shows the second wave.

The fourth wave is therefore = C -* -*)ae-^-a) le = ajSV-* 41 -*-*^. Then multiply by e~ q(l^x} to get valu e at x. of the reflected its Multiply by /3 to get value wave i\ at B. = /Sr-^^iv 3. Then multiply by a to the value of the reflected wa ve v3 at A. ViL (26) z After this. fa So we have the series i\ = e-tf*-^ J.a+a!) J*. CH. VI. to get the value of vz at A. (25) showing the third wave. multiply b y f-a Put x = = e-^a/3e-< M 0) J* = a/3effl . Put x = I to get its value at B. it is the same over and over again.78 ELECTROMAGNETIC THEOEY. produce ~qx The result is to get its value at x. Finally. (27) ] .

ae. So.2 * the potential at x due 1 . if this new set of and A precisely in the former waves be called w we have t (29) = /3e-** and so on. we ha ve Je a succession of reflections at B manner.a^e. by the law of geometric series. So. The total is showing the part of going to the left. But there is also the initial wave to the left to be con~ qa sidered. becomes ^e at A. the sum of the v waves is first The wave is initial which represents the potential at x so far as it arises from the wave to the rig ht. the f's and w's. After that.ae~ aft j <? by reflection. . and the rest are obtained in succession by multiplying by one or the other of two factors in turn. (l -*hc 5 . given explicitly.(a+z) when it reaches x. which becomes . It is -~q(a x] ^e between a and A.and so on. generates the new wave .

. the sum . therefore. (so) to the initial wave of The real potential is.

But when the point x lies between A and the impressed the case is somewhat diffe rent. we only gets to x by producing a reflected have the series of waves and so on.gr^gH^9 K and its (32) Similarly. If we consider the effect at x of the initial it wave to the right. making allowance for the fact that wave at B. 77 force. whose sum is .PURE DIFFUSION OF ELECTRIC DISPLACEMENT. the initial wave to the left are represented by the series consequences 8 J and so on. whose sum is 2f .

or sidered.1. There are several casea in which these solutions are reducible to the elementary waves of the kind before conThese occur when a and / 3 are + 1 or .Finally. and the latter the sum of (32) and (34). of (28) and (30). Four Cases of Elementary Waves. denote by and by sum That V 1 the potential on the right side of e. or zero. equivalent to the previous deve lopment in terms of waves. . represent the complete solution in compact form. The former is the the potential on the left side. V 2 is to say. 260.

) Both ends earthed. VI.1 and + 1.1 = /?.) _shmga shin ql sh g (S-a). and v + cosh a i ghln shin (l _ x) (37) ql y (2.78 ELECTROMAGNETIC THEORY. constants lying between the limits . CH. so a = . a or fi being (1. Thus. Under these circumstances the re flected waves are copies of the incident either full-sized or reduced. with or w ithout reversal of sign. g . 1 makes four important cases. The V waves are reflected negatively.

(39) ql shin (3. and cosh (4. Then a= 1.. The V waves are reflected positively.) ql Earth at B. Here a = . and cut at A. (3g) Both ends cut. so a = 1 = /?. and shin aQ. and cut at B.g.) ql Earth at A. /? = 1.

(46) . and cosh These solutions and the more general cases may be easily converted to Fourier se ries. (36) may be got by assuming (45) and determining the four time functions by the two mediate conditions interVx-V. C^C.. In the meantime it may be noted that the general solutions (35). if required.1. /?= 1. = *. by a method to be explained later.

either What the terminal conditions are in general. The Reflection 261. In the case of equations (37) to (44) these terminal conditions are simply = or C = at and B. selves. Let 7 = 2^ This says that Z x is the resistance operator of the at B. will now be pointed out. Coefficients in terms of the Terminal Resistance Operators.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. Now let i^ be a wave incident upon B. and by two terminal conditions. /?. and r 2 be the reflected wave. Their sum is the real potential. 79 &ix = a. terminal arrangement at B. a s may be readily tested. The reflection coefficients are usually operators is themThere no difficulty in finding them. . in relation to V A the reflection coefficients a.

j/R to ge t that belonging to va because the first is a positive and the second a negative wave. So q(v 1 -v. from B respectively. (47) Also. J (49) giving ft in terms of if Z^ the resistance operator of the terminal . (48) division.) = EG. we multiply by g/R to get the current belonging to v lt and by . therefore = V = iz/?. we get * . by ft (48) Eliminating C between (47). going to and coming .So Z 1 0.

(50) the incident and r 2 the reflected wave.--*=*. to be multiplied by correspond. to be multiplied by whilst v 2 is a pos itive wave. we shall have Vl if t\ is + r2 =-Z C. we obtain * therefore . (51) Eliminating C. Now Vj is a R/j to get the current to negative wave. R/0. so ?(r 1 -i. Z is arrangement at A.Similarly.) = RC. (52) .

This because R/<? To infinitely long cable line. . By inspection of (49). Cases of Vanishing or Constancy of the Reflection Coefficients. (36). CH. at x= and V = ZjC at x = I. VI. is we is see that fB vanishes when Zj = R/#. 262. that the cable either does not stop at or else that if it the resistance operator of an say that it equals Z 1? asserts . in Using these expressions for a and (3 giving a in terms of Z the solutions (35).80 ELECTROMAGNETIC THEORY. they are made completely exThe terminal conditions they satisfy are V = Z G pressive.

We now have Vj = (<+ j8r-< ffl l> flB> )J. or else there is a te rminal arrangement copying the continuation. Again. In this case /? = reduces (35) (36) to V = (t qa . (55) V= 2 e**(e. there is a terminal arrangement which is exactly equivalent to the infinitely long continuation.ae-^e-^le.n x 2 (53) e. because either the cable is continued past A indefinitely. ')e. so far as the real cabl e itself is concerned.(* + a-*x)./?<r*2Z . does stop. (54) showing two waves only. the Finally.a (56) showing two waves again. V = . There is no reflection at A . for there is no reflection at B. if Z = R/g then a = 0.j3. and therefore can be bu t one reflection at A. if . but goes on to infinity.

. freed from p* (besides in the previous ways). (3 cont ain q. (59) . to have the coefficients involving p . /3 Il/j =Z =Z 1. . It will be observed that the expressions for a. without subsequent interference. therefore. WAV.= (A-y. we must introz.reflection being at B.--W*fc (58) which are simply the primary waves from the source. duce p* in Z x and Z Say. and we (57) reduce to Vi-'r**-^ V. for exam ple. both a and vanish. } In order.

put between two other B S on the left of A. (36) are in action. General Case of an Intermediate Source of Electrification subject to any Terminal Conditions. on ly with a and /3 constants instead of.* If we do not abolish the reflections. AB is of the type cables of different types and of the type E 15 8j on the right of B. w hich creates a is . In conection with the preceding. operators containing p. the other kind of The treatment is similar when source should be mentioned. We can get rid of the reflection at A by having R /S = R/S. 263. This is somewhat more general than the previous way of ha ving continuations of the same type as the real line. . The differentiator p then disappears from a and /3. and may evidently have any values between (and including) 1 and +1. the whole series of waves summed up in (35). and of tha t at B by = RJ/SI R/S.PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. as in the general case. which reduce to (B /So) + (R!5) These are constants. 81 These mean that the cable . the source at as = a is not impressed force.

left. Thus. not necessary for the terminal cables to be homogeneous. if corresponding property in the diffusion of heat. creating a discontinuity in the current. though not in the current. then t he current in the line increases by the This in passing from left to right past the source. but a source of curre nt. (61) (62) c2 = -e-5 '"-*^. n. case has been already briefl y considered (250). so far as the reaction on the real cable is concerned.discontinuity in the potential. G . we want R/S = Ri/S1} we may make it go by having any number of cables in sequence. let h be the strength of a source of current at x = a (say led to the cable by an auxiliary wire). on the on the right. Comp are with the * But it is For example. as described in 233. The primary waves are amount h Cl = r-**' J/i. in which the value o f Ri/Si is constant. though not in the potential. VOL.

So we may employ the previous efficients as above. (64) When h is steady or impulsive we have already indicated the at present in question is the extension to include terminal influences. wave to the left is made the same by writing ~R>h/q for There is no other difference. VI. on the left. to CH. = e-(-*> 2q . The the right is made the same as in 259. Have the same reflection coresults. and the potentials Vi correspond are = e-a(*-a) BA on the right> v. by substituting ~Rh/q for e.82 ELECTROMAGNETIC THEORY. whilst the initial e. wave to . eq uation (23). The point initial and apply them to the present case.

.dV + d V* = m. as modified by the terminal influences.* dx dx (67) Take h =pQ when the source is impulsive. V x on the right side and V 2 on Test that V! = at x = a. /? regarded series. but to the numberle ss other expansions which occur naturally in the physics of the matter. (66) represent the potential due to a charge Q which is initially all at the point x = a. In particular. as functions of p. (36) become (66) which express the potentials the left side of the source h. The conversion to a Fourier series of this result leads to the expansion of an arbitrary function in all sorts of Fourier not merely the period ic case which rigourists have tried so hard* to demonstrate. Then (65). * It is .results fully. V a> and . associat ed with different forms of the coefficients a. the solutions (35).

no easy matter from the restricted I demonstrations do not mean that they have not succeeded. hard to follow and not very convincing. but that the rigorous are. from a phys ical point of view. .

the time differentiator. as. or of Vibrations. or two. The effect is t o express the solution in the form oi the sum of a series of waves. though they have a wider application. it would have this important one that in considering the effect due t o a source it . Some ways of algebrising such operational equations have been already given. which applies very generally in physical problems concerning continuous media. howmust come The Two Ways of expressing Propagational Results. especially applicable to diffusion problems. This is a very simple and powerful method. The operator Z is to be co nstructed in the way previously explained 245 (or in any equivalent way).PURE DIFFUSION OF ELECTRIC DISPLACEMENT. to be generalised in the final result to the in functions of p. the resolution of the operator Z by algebraical division int o wave operators. a cause e. Why shouldn't it ? This matter. These may be either simultaneous or successive. the current in one part of an electrical system due to an impressed force in another part. 8D rigorous mathematical point of view to answer the question. in terms 264. say e = ZtC. in the same way as if the elements of t he combination were mere resistances subject to Ohm's law. Given an electromagn etic operational Here C is some particular effect due to equation. which are appropriate to the real natur e of the elements. namely. later. according to circumstances. for instance. of Waves. Having accumulated a stock of formulae in the last few sections. that is. One way in particular should be noted and remembered. There may be but one wave. If the method had no other recom mendation. is rather. Why should an arbitr ary function be capable of expansion in such or such a series ? But from the phy sical point of view. though it is not necessary to restrict their meanings in this way. the question ever. or an infinite number. it becomes necessary to explain their connection with Fourier series.

whereas an alternative and equivalent formula might completely disguise it. which lea ds to a strikingly different functional expression of the developed G2 . But there is another very d ifferent way of resolving an operator into other more elementary ones.in the imitates nature by directly expressing the course of events way it happens in ac tual fact.

if w e produce evident and visible progressive waves. that is. fixed at its ends. it circumst ances which of the two ways is to be depends upon employed. CH. vibrations the be frictionally resisted. for instance. may express our results either in terms of the waves or in t erms of the vibrations We of normal systems of disturbances. in the latter case. It is not a matt er of indifference which way of development is of the It may be that in some particular problem one two is far more manageable than th e other. Only by the familiarity with them. analogy. They are widely separated. preferred by its natural recommendations. The trunk corresponds. of course.84 ELECTROMAGNETIC THEORY. the case of a long stretched cord. A perfect analogy in every respect anything but would require an identity for nothing is wholly like itself and an identity woul d be useless for an The present one is good enough as far as it goes here. We may compare the two developm ents to the branches and the roots of a tree. Apart from this. VI. which their universal existence in physical probl ems involving propagation (subject to certain restrictions) makes possible. to the operational solution. Take. solution. if in terms of the subsidences or decadences of the normal systems. Or. it is natural to express the ma . or more amenable to numerical calculation. If we it so as to make a hump or a number of humps run along the string. No doubt the analogy will fail if pushed much further. do they become commonplace. but hav e a common bond in the trunk which joins them. The results are so entirely unlike the wave results that their equival ence produces algebraical identities which are rather astonishing at first.

that it is not natural wave in the form of normal . Here it is obviously natural to express mathematically th e visibly a complete may be form of normal vibrations. Again. and let it go. expressed entirely in the quantitatively.thematical results in the form of waves. it will vibrate in the same disturb form over and over again. or ha ve intermediate nodes. But in the former case the progr essive wave We to express the simple progressive see at once. however. whether the sine curve be only half wave length. and in the latter the vibrations may be expres sed entirely in the form The methods are perfectly equivalent of progressive wav es. evident simple vibrations as vibrations simply. if we displace the whole cord to the form of a sine curve.

with nodes at beginning and end. when. or increasing the self-induction. By reducing the resistance. so the disturbances simply subside or decay. The waves. or the 85 normal vibration in the form of progressive because the results would disguise t he reality. vibrations. and the other f orm later. tory subsidence. under the same circumstances. however. examples used here are extreme. and conversely. we make the vibrations last longer. though decaying at the same time. if earthed terminally. more like ly. owing both ways being complic ated. if the potential in a submarine cable. imagined to be qu ite free from self-induction. be numerical advantage in using one form or. that in pure diffusion problems We we are not concerned with true vibrations. and the potential will This is oscillavib rate. should remark. The normal vibrations may involve much sion problems. one form may be useful in one part of the history of events. however. of which there are many examples in diffu. The resistance stops the vibrations. and be left to itself. though conBut self-induction will make pass through the equilibrium position. The . In intermediate cases it may to be quite difficult to say which way is preferable. and resemble those of a stretched co rd. the curve of potential will preserve tinuously falling to equilibrium. say V = V sin ax. the waves are easily done . or in both ways to gether. it its sine form. be distributed according to a sine curve. labour in calculation for some particular value of the time. as of a string. There may. Thus.PUKE DIFFUSION OF ELECTRIC DISPLACEMENT.

perhaps. different matter when the etherial th eory is in question then r . w hich makes the Of course. The stretched cord makes by the best analogy for a telegraph circuit when it is desired to have a simple analogy. and reflected. without resistance. and how pulses are tra nsmitted along it. Now all these things have their close representatives in a telegraph circuit. The loss of intensity. suitable a different nature. it is an entirely simple analogy be very useful. bec ause every one knows something about how a cord vibrates. and so on until finally killed by friction. be as difficult to follow as the electro magnetic theory analogies are of . when far friction ignored.limit vibrations would be reached if there were no would continue for ever. like those of the string in acoustical is theory. Th ey are visibly evident. which may.

of course. we have of As an example vibrations or vibrational subsidences or pure subsidences of normal systems in l imited number which do not admit of very . Progression is the essence of a wave. (if we include subsiding And. tically. or vibrations only. Why these failures occur is evident in practice by physical considerations. there can be at most only two waves from one source situated in an i nfinitely long cable. limited number of degrees of freedom in an electrical arrangement. operational equations refuse to go m ore than one way.86 itself. Then we cannot have a vibrational form of solut ion normal systems under vibrations). FitzGerald. Now the reader must be cautioned against supposing that every operational equation convertible. w hen there is only a except artificially. but not both. the stretched cord fails. however. by employing the vibrations of air in a pipe in the analo gy. But the pipe analogy is not so easily followed least in in most respects as the stretched cord. VI. into waves or normal vibrations. of course progressive waves are referred to. as in the theory of condensers and coils. the continuous medium first imagined by MacCullagh. at any simple manner. as in 255. It does not.. CH. instance. waves are somewhat deceptively so called. waves only. if simple norm al vibrations be included therein.* however. was able to include leakage. represent the effect of leakage on a telegraph circuit. since compressions and condensations of the air are less easily pictured than the motions of a cord. is Some described. it is due to peculiarit ies in the form of the MathemaZ operator. In speaking of the diverse modes of representation as being i n terms of waves and of vibrations. in which rotationa l elasticity is involved. In one respect. for ELECTROMAGNETIC THEORY. as. They make waves only. whether undistorted or distorted as they Ir " Standing progress. Prof. as save perhaps by artificial expedients.

when there is a continuous medium for propagation in some part The distinction is usually connected with the of the system. 1894.expression in the form of continuous waves. or absence of bound aries. May 25. . If we create a disturbance presence in an elastic medium which * is quite homogeneous and is The Electrician. 106. except perhaps artificially. General ly speaking we may have both. p.

PURE DIFFUSION OF ELECTKIC DISPLACEMENT. 87 unbounded, the disturbance goes out to infinity in wave fashion, and are possible. dissipate no free vibrations in definite and separate modes But if the medium is bounded, the wave cannot itself freely, but is reflected and re-reflected any times. number of We can now express the disturbance in terms normal systems of a complicated of the vibrations of definite nature, depending upon the nature of the medium and on the The possibility arise s from the form, &c., of the boundary. of an infinite number of progressive wave s, superposition and we may express our vibrations. results either in terms of waves or of This brings us to the reservation made about an artificial way of representing a progressive wave by vibrations. If in the last case distinct we imagine the boundary shifted further and further away, however far in force. we go the normal systems remain of representation are the boundary is removed Now, to an infinite distance, the expression for the sum of the normal systems becomes a definite integral. Thus a single and separate, and the two modes in the limit, when progressive wave may be expressed by a definite integral as the sum of vibration s of infinitely numerous normal systems Instead of a differing infinitely little from one to the next. definite sequence of separated periods, the periods run into one another. Of course it is simpler to think of the progressive

wave tban of the vibrations in the definite integral which equivalently expresses it, thou gh not in a desirable manner, and to derive it directly from the operational solution when possible, instead o f through the integral. After these general remarks about the chief peculiaritie s, we must proceed to show practical methods of manipulating the operator so as to convert operational solutions to vibrating or subsiding normal systems. It is not difficult by the general to be explained. Perhaps it is too easy. That is, too hi execution, for the theo ry thereof is more difficult. As, easy however, it applies to all sorts of norma l systems besides those Z method concerned in Fourier series, it will perhaps be best at first to show the connec tion of Fourier series with the operational forms in more special ways, which, t hough longer, may be more immediately intelligible methods. to the unpractised in operational

88 ELECTROMAGNETIC THEORY. CH. VI. Conversion of Operational Solutions to Fourier Series by Special (1). Ways. at Effect due to e at A, when earthed A and B. 265. Take, to begin with, an interesting special case, namely, the effect due to steady impressed force e at A, where x = 0, when earth is on at both A and B, w here x = l. As before shown, the operational solution is shiuql and many other simple cases can be done by using well-known trigonometrical to a Fourier series in this identities, assisted The conversion by an elementary operational result. Thus, we have -1M qlJ ,.. Here noting that means ESp, we see by trigonometry. 1 that we have merely to alg ebrize (1+P/^)- !, where P is a constant. in full, Thus, 1+ P P

p p \p (2 [8 or, in the usual brief expression of the exponential function, rt P This result should be noted, as it to (2), we obtain it often turns up. Applying where the constants plt p^ &c., are given by . 2 Pn = - fiV/KSZ (5) Comparing (1) with (4), we see that V will be found by operating on (4) by ql

PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. This but that e pt 89 is differentiations. is a function of ff also, so involves only complete They may be carried out at leng th if desired, on unnecessary. For the power of p in the operator (6) simply P. It is therefore ze ro on the first constant term on the right of (4), then p1 on the second term, a nd so is on. That is, these values are to be put in (6) for (1) is converted to p to express its effect. So V= where the through 2 <? q's are RSp. put q the constant values corresponding to the p's, Equation (7) is the full solution, but for trigonometrical convenience, since the values of <f are negative, it is best to 2 = - s2

. This makes or, since sinsZ = for every s, A ) (9) where the s's are the roots of sin si = 0, which is the same as saying that the p's are the roots of shingZ customary form (9) is the same as = 0. In the I/ r n I (10) This is "When the equivalent of the wave solution before got. t= x the time factors vanish, an d there is , left the

final state expressed when t = 0, by the outside term. the time factors are unity, and On the other hand, L J TT n (11) I We know where it is e. that V is initially zero everywhere except at A, Therefore l-^ZJ-Isin^, ~ n l I I ' V (12) That is, except at x = 0, where the right member gives zero. ihe steady state, w hich is e(l - x/l), has become expanded in a

90 series of sines. ELECTROMAGNETIC THEOKY. CH. VU But we have got more than we wanted. For So the right is periodic, with period 2 1. member represents 1 - x/l between and 21 then the same values are repeated fr om 21 to 4Z and so on. Similarly on the Fourier series ; ; the left solution side of the origin. of the problem of Equation (11) is, in fact, the finding V due to an infinite number of equal impressed forces of strength 2e acting all the same way at the p oints 0, 21, 4:1, &c., - 21, - 4:1, &c., in an This was shown before as regards the infinitely long cable. and wave solutions. operational (2). Modified is Way of doing the Last Case. 266. There another similar way of getting (9) which IB worth noticing. Use the trigonometrical identity + . .

.,(13) differing terms. from (2) on the right side only in the signs of the Let the operand, as before, be 1 in the usual manner, ql coth ql and we obtain . e = e(l + 2e pl + 2<?* + 2e^ +...), * (14) instead of (4) above, with the Now (1) is the same as same values of them's. V = (coshqx - coth ql shin qx) e = cosh gx.e- shingaXl + 2^ + 2<?* + ...), ql (15) by (14). Here the power of p is zero in the cosh function, and also in the other so far as the constant term goes, and p l on ePi*, &G. So, putting q 2 = - s 2 we get , U which is .*, si (16) the same as (9). Since circular functions are finally to be employed, we 2 - s 2 at the beginning if we like, and write put q = sins(/ may x)^


PURE DIFFUSION OF ELECTRIC DISPLACEMENT. instead of (1). 91 Here of course s 2 is RS/>. It finally takes the meaning a differentiator, namely, of a constant, or rather of an infinite series of constants, merely because p has the power of a constant on the exponential function. The determinantal equation shin ql = 0, or, more = 0, (shin ql)/ql strictly, is really the equation finding the admissible values of p in the normal systems. To every p belongs an s\ so if we choose to c onsider the determinantal equation in the form as finding the values of s, we need only attend to the positive values. Notice a lso that the operator in (1) does not contain p } but (sinsZ)/s , = only the complete p. The elementary waves which make up the solution (1) do depe nd on p*l, but the union of all the wave operators to make the solution (1) caus es p* to be eliminated.

length a rational equation in p, provided the terminal arrangements themselves a re finite* combinations whose resistance operators is, that The determinantal equation for a cable of finite bounded both ways by some restr aint, is always are themselves rational functions of p. (3). Earth at Both Terminals. Initial Charge at a Point. Arbitrary Initial State. last, 267. A more comprehensive case than the bringing a fresh peculiarity into view, is that of the effect due to an initial charge at a single place when the cable is earthed at both ends. There are two solutions in terms of waves, one for the region to the left, and the other for that to the right - 1 =/3 in To fi nd them put side of the source. (65), (66), Then we obtain 263, and also h=pQ, a s there described. a= on the right at the place side of the source, which 2, is a charge Q initially left side, x = y. To obtain

V the potential on the interchange y and x in the last equation. We see that we have the same denominat or shingZ as This is clearly because the terminal conditions, earth before. * There is an exception to finiteness in the case of a distortionless circuit used for a terminal arrangement, because it behaves like a mere resistance when it is infinitely long.

92 ELECTROMAGNETIC THEORY. CH. VI. on at both ends, are unchanged. the reflective action of the ends. The denominator sums up But it is not to be supposed that changing the terminal conditions necessarily changes the determinantal equa tion. For that depends upon both terminal conditions, and a change in one may be balanced by a change in the other. Thus, if both ends were cut, the denominator would be unaltered, since the signs reversed. rates, as It is their of both a and ft would be we see by (65), (66), 2 product that determines the subsidence 263, which show that, in the " general case, = a/3 (4), (19) is the determinantal equation. (18), Returning to V, and using the result -

we get (20) = siring shin 2 (l - x) p bt (1 2^ + 2e^ -...), The final steady with the same values of zero obviously, so there as before. V is once is no outside turn, and we get at by putting <f = - s2 , V= S or, since sin si sins?/ = 0, . ('22) ol because for have in the last two equations written 15 That is, (22) the intercha nge of x and y makes no difference. represents V at a; due to Q initially at y, whether x be to the right or left of the source. Only one Fourier series is required to represent th e essentially different results to the right and left

We V V of the source. t Putting = and dismissing constant factors, we see that if L Li is infinite. (23) the summation including all integral values of n from 1 to oo then u represents everywhere between A and B, except at , the point y = x, where it because the space total of But is its space total is 1, SV in (22) left Q at the first moment, before any of the charge has the cable.

PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. 9B The function u like the p\ with which we were concerned is therefore expresses an impulsive function, before, only now the t, variable x instead of or the function is distributed in length instead of time. Such functions can be represented in various ways. Every Fourier series involves one form. If we multiply a continuous function of ?/, say/(y), by u, an impulsive function which exists only at the point y = x, the product is obviously zero except at t hat point, where it is infinite. But if we take the space total of the product exists at #, ttf(y), the result is /(#). For u only and its total is 1. Thus, if This the limits include the point x. If not, the result is zero. is the property made use of in Fourier and other series when employed

u, to express arbitrary functions. The function in the above or other special form, spots a single value of the arbitrary function in virtue of its impulsiveness. Suppose, however, that the given function f(y) had no happens when there is a discontinuity, or sudden jump from one value to another. Under Here/(//) has any value between the two extremes. definite value at a particular place, as what should we expect the impulse to do, can only, through (24), lead to one val ue ? To be quite This is an excellent fair, it should show the mean value. reaso n why it ought to do so, though not a proof that it does. these circumstances, if it Why we it must may become (24) to the clear later on. (23), Applying see that problem in hand, with u as in /(*)-* Z^^/Msfe^ g (25) expands the arbitrary function f(x) in a series of sines, so as and I (where, of course, if /(x) does not to vanish at x = We also see that vanish, the formula fails). in!!!^ i (23) represents the potential at x at time the potential was represented by V t . after the moment when is The former Q

represented now by <hj. S\V///, that is, the charge on the clement of length

'94 ELECTROMAGNETIC THEORY. OH. VI. The formula (25) may, as we have seen, fail either at the But the terminals or intermediately at a discontinuity. formula (26) does not fa il at all for any finite value of the time. then. Initially it is the same as (25), and may fail But the presence of the time func tions makes V continuous. zero, there matter how small t is taken, if not actually no discontinuity. Sharp corners in V are instantly rounded off by the time factors, though if graphically represent ed there might be no observable difference. We may take t so small that the time functions differed from 1 insensibly for millions of terms, yefc go far enough, and we must come to time -functions differing insensibly from So V is necessari ly made conwith the same value of t. is No tinuous in the smallest interval of time after the first moment, Thus, by if it be initially discontinuous. is Another way of getting (22) (18),

by using the coth function. V = ql shin gy (cosh qx x shmga? ql gl \ coth ql) / Q bt (27) Noting that the the last to first part is inoperative, the result (13) turns V= - shin qx shin qy . St (1 + 2e^ + 2ep2 ' +...); (28) and now, noting that the power of p is in the first term, in the second, and so on, and putting <f= -s 2 we make , p^ V = ?5 2 sins* sins*/ S as before. Finally, a third way. *,

(29) We have, by (18), Now the .(1) and part after the x was algebrized in Therefore (9) there. 265 ; see equations l-f)and now, on sight by the power of 2 ; *} which is (31) effecting the differentiations, done al p being constant hi every term, we

PUBE DIFFUSION OF ELECTRIC DISPLACEMENT. obtain (29) again. These variations will, the reader to operate quickly and safe ly. (4). 95 if practised, enable Line Cut at Both Terminals. Effect of Impressed Force. 268. Next, in further illustration of simple cases, let the line be cut at both ends. First, with an impressed force e at the point x = y. Then, by equations (39), (40), 259, we have to Fourier series, put the part involving sin&r. Then To convert <f= - s 2 in (32), and reject . (si) Si In this use (14), making Si which finally makes, by the power of

p being a constant in every term, Vx = "This is y 6 + 2 ?! 6*6 sinsy cosftv e*. (35) on the right side of e. On (1 the left side, (33) makes (36) V = -*+* 2 + 2e^ + 2c* + cP*, . . .) si = _e+ which only expansion differs _ ^ +2 N

sinsy cos &* (37) from (35) in the outside term expressing the steady state. is The values of s are as before, but the represents the negative of the final state, which is discontinuous. As regards t he final state, we have two condensers in in cosines, not sines. The summation sequence, of permittances force S?/ and S(Z y), with the impressed is between them. The of the effective permittance of the the reciprocal of the sum

reciprocals separate permittances that is, Si/(l y/l). Multiplying obtain the charge, and then dividi ng by S(J-?/), by e we we obtain

96 ELECTROMAGNETIC THEOEY. CH. VI. ey/l, the left side it is the final potential on the right side of the force. lower by the amount e. On (5). Line Out at Terminals. Effect of Initial Charge. 269. We may Let it same way. do the effect of a single charge at y in the be Q, then by (65), 263, with h=pQ. This is on the right side of Q, and on the left side th e formula is obtained by interchanging Put qz = s 2 and reject the part involvin g sin sx. x and y. , Then Vi = cos sy cos sx . si cot si ^ bt + 2e Pl + 2ep2* + * . (89) = S Q. cos sy COS sx (1 .

.) (40) =Q bt {1 + 2Zcoss2/ cos sx e*" }. (41) In passing from (39) to (40) the result (14) is used, and the further passage to (41) involves only a formal change by the power of p being a constant, as in fo rmer cases. By the last formula we see that t^i + ^cos^cos^ Li (42) I I is a unit impulsive function existing at the point y = x, like (23) in fact, so tha t the formula (24) applies, and an arbitrary function may be expanded hi cosines thus cos : cos dy. (43) in the The introduction of the time factor *", where p= -w summation will, by the prece ding, represent what becomes at time t the initial potential f(x) later

by diffusion in the cable with its ends insulated. The final state is a uniform mean (6). potential. Earth at A and Cut at B. Effect of Initial Charge. Next take a case involving a different determinantal Say there is earth on at A and disconnection at B. equation. 270.

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. 97 x = y, we Then a = 1, /? = 1. 263, So, with an initial charge Q at have, by (65), in which the interchange of x and y will give V 2. Or, putting = si, q Y! = si sin sy (cos sx + sin sx tan si) -*-. (45) The first part is inoperative as regards a Fourier series. sinsx)(sl tansZ) So (46) M. Here we see that we have a different operator to produce the series of exponenti als, viz., si tan si or ^than^. The expansion of this function is trigonometrica l Remembering the meaning gl

of q < z , namely, RSp, this makes * . . . , than ql . 1 = 2e Pl + 2ep2 + W* + (48) set, in the way (14) was got, only now or the p's are a different namely, the roots of cosh<?Z = 0, &C. cossZ = 0; (49) say sl = ir, ITT, TT,

So, by (46), Vi= which is ^ sinsy sin sx Z bt sin sx P ', (50) at once transformable to V = -^ 2 sin sy bt **, (51) by making s be constant instead of a differentiator. In this example the normal functions of the type sins# vanish At B they have maxima, so that their slopes a t A only. This is to make the current vanish at B, where there vanish. The chang e in the first normal function is very is insulation. great. Being sin (irx/2l), it is just one half of the first normal n vov ii.

98 ELECTROMAGNETIC THEORY. CH. VI. The function for a line twice as long, earthed at both ends. corresponding p is the same as that for the line of double length, so that the subsidence of the first normal system in the present case of insulation at B takes four times as long to reach a given stage, as the subsidence of the first normal system of the same li ne does when it is earthed at both ends. most important. the only one worth counting. practically big t is smaller, the s econd one acquires significance, then the third, and so on, up to the first mome nt, when the whole But these Fourier series are frequently series is required. The first normal system it is, of course, the When When t is is very unpractical for numerical purposes on account of the slow convergency. when t is taken small Periodic Expression of Impulsive Functions. Fourier's Theorem. 271. The above examples the are sufficient to solutions nection between

operational show the conand the corresponding Fourier series in the simple cases of terminal direct earth connection and complete disconnection, at least so far as the particular method employed is concerned, bearing in mind that principles are often as well illustrated by simple as by more complicated problems, if not better. The above process consists in first converting the reciprocal of the determinator shin^Z o r cosh ql, &c., to the sum of partial fractions by using known trigonometrical formulae. These are easily algebrized, and the re sult is to give V or C or other quantity at some particular place. Thirdly, we o perate on this result directly by whole differentiations, which merely means giv ing p a constant value in any term, and the result is the solution at any place in the form of a Fourier series. Before passing on to another way of considering Fourier series operationally, some remarks about the functions called u above may be useful. They were considered merely as expressing unit impulsive functions located at a point x y and I. So far as that goes, they are quite situated between similar. But in other respects they differ. They are both Every time x is increased by the periodic, however. 2Z, the sines u and the cosi nes u repeat themselves. amount But the

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. sines 99 function. an odd function of x, whilst the cosines u is an even So the sines u is negative d by changing xio - x. The complete meaning of the sines u is therefore an infin ite is u row &c.,xof positive unit impulses at the points y=x, x + 2l, x + M, - 1, &c. ; together with an infinite row of 21, x negative unit impulses at the points y= -x, -x + 2l, -x + 4l, &c., -x -21, -x-4:l, &c. On the other hand, the cosines u represents positive unit impulses at all the above places. Neither of the U'B is a periodic repetition of a single impulse. if we take half the sum of the sines u and the cosines the negative impulses wil l be cancelled. There is left simply an infinite row of positive unit impulses a t the points The diagram will serve y = x 2nl, where n is any integer. But u, to fix this plainly : TTTTT Cosines 1.1 1.1 i.i 1,1 i. Both -21 Q x is 21 u -x

and The 21, horizontal straight line vertical lines at divided into equal lengths and the x and x2nl, -x2nl (23) represent the distribution of the impulses in the three cases of the sines u, th e cosines u, and half their sum. By and (42), the new u is represented by < 52) its meaning for our purpose is a function of y which is zero everywhere except a t x 2nl, where it is infinite, but so This u is the periodic impulsive that its total is unity. and function concerned in the periodic form of Fourier's theorem. For, using (52) in (24), we have f(x) = , 1 + 22 " cos T (y ~ which expands an arbitrary function f(x) whose values are and 21 into the sum of a series of given between the limits H2

100 ELECTROMAGNETIC THEORY. OH. 71. sines and cosines with constant coefficients, and with a constant term expressin g the mean value. Besides that, it makes the given values repeat themselves over and over again, with the That is, the impulse operator fu... dy, in virtue peri od 2Z. of its periodic nature, makes f(x) periodic as well. When the period is infinitely long, if we puts = W7r/Z, we have ds=7r/l to represent the step from one s to the next, which is infinitely s mall in the limit, when the impulsive function becomes -if TTJ cos s(y x) . ds, (54) and its meaning it So, applying a unit impulse at the point y=x only. through (24) to a function of y, we get is (*) =-

Try-* f (55) which is the form of Fourier's theorem applicable to any function given between - oo and + oo . The it be finite, may be considered separ ately. mean value, should Fourier Series in General. Various Sorts Needed even when the Function Zero. is Cyclic and Continuous. Expansions of 272. Fourier's theorem as exhibited by equation (53) applies to the diffusion of charge in a cable in two cases only. First, let the cable be infinitely long, a ll in one piece, and let be charged initially in a periodic manner, say arbitrarily between x = and 21, a nd with repetitions of the same state in it all is the other sections of length 21. Then the expansion (53) The function /(a?) repr esents the obviously the proper one. The state at any time t later is got the time factor c pt as in previous cases, in the by introducing summation on the right side. This allows the harmonic initial state of potential. , initial state terms to subside or decay, leaving only the mean value of the behind finally. Bu t, secondly, it answers the purpose equally well to have

finite a cable, of length 21, with the ends joined to make a closed circuit. We may take the origin (x = 0) anywhere, and

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. 101 is x = 2l, 4/, &c., mean the same as a? = 0. Equation (53) the appropriate expansion to employ, with the time factor introduced to indicate the resulting state at any time. The action of the time functions is always to make the later terms in the expans ion insignificant compared with the earlier, so that we may generally say that t he successive terms summation, not counting the outside constant term, decreasing importance. And, i ndeed, in making practical applications, it will be found that this is of the are of regularly usually true for the initial state itself. usually find that the natural order o f the harmonic terms hi the expansion of We a function is the order of their relative importance inversely. But this property is by no means a necessary one. One or any number of terms may be wholly absent, or if not absent may be very small, and followed by bigger on es. There is then an exceptional state at the first moment, and for a certain ti me But when the decadence has progressed sufficiently after. we must arrive at a state of things in which the order of the terms that exist at nitude. all is also the order of their magWhen a cable is closed upon itself, say as described real potential,

above, the whatever function of x. Now fully periodic form. fact of the periodic character of the potential in the case of a closed circuit made the periodic form of Fourier's theorem be, is necessarily a periodic Fourier's theorem in the form (53) is the It might therefore seem that the mere it may obligatory. But this is only accidentally true, as it were. The expansion (53) i s only one of an infinity of expansions applicable to a closed cable, each in it s proper place. For (53) to be applicable, it is not only necessary for the cable to be that is, there must closed, but also for it to be self-contained be no external connections imposing some restraint upon the " terminal" potentia l or the current at a particular place. The V at x = and V at x = 21 are identical, and that C is also continuous at those p laces, because they are united, and in such a manner as not to interfere with th e conWe may, indeed, choose the initial tinuity of V and C. state to violate the se conditions. But the violation is only conditions are that momentary, for continuity action of the time functions. is instantly established by the

102 ELECTROMAGNETIC THEORY. CH. VI. But suppose we put a simple leak on at the place x = 0, The charge in the cable everything else being unchanged. will redistribute itself now in an entirely dif ferent manner. For one thing, it will all leak out of the cable, instead of settling down to a state of uniform distribution. It is obvious that the periodic form of Fourier's expansion of the initial state is unsuitable, alt hough the real potential of the line is periodic. There is no difficulty in obtaining the proper series, though it would be too gr eat an interruption of continuity to give it just now. The peculiarity is that a lthough V is continuous at # = and 2 (except, it may be, at the first moment), C is not continuous. The discontinuity in C at the joined terminals is proportional to the potential there, by Ohm's law applied to The proper Fourier series must be found to satisfy the leak. this condition as well as that of continuity of it V in order that the time factors are introduced, represent the The initial state may be chosen a t any moment. potential and so as to violate the conditions stated, but this arb itrarily shall, is when of no consequence any time. at all in the

complete solution for V at leak may be generalised to any combination a definite resistance operator, by su bstituting the having Thus we obtain any latter for the resistance of the leak. The simple number of Fourier series for a closed circuit, which may represent one and the same arbitrary function (the initial All are continuous as regards V, but discontinuous state). at the junction x = Q as regards C accordi ng to the special law imposed by the nature of the leak. That is, when t is = 0, there may finite for, going right back to the moment t be failure then, viz., by choosing the arb itrary initial state so as to fail to comply with the conditions. ; tive of the potential), Instead of a discontinuity in the current (or in the derivawe may make it be a d iscontinuity in the potential inself. Thus, we may put a coil in circuit with the cable, doing away with the leak. And, as before, this be variously generalis ed. Or we may have the potential may and the current both discontinuous at the junction of the beginning and end of t he cable by suitable electrical devices. In every case there is one and only one appropriate Fourier series,

and there is no difficulty (save complication in

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. detail) in finding it 103 by seeking the solution of the physical an initial state subject to the imposed problem On the other hand, if we eliminate the physical conditions. ideas and al so the time variation, and look at the matter from of decadence of a purely mathematical point of view, there are serious culties introduced. diffiThe where. initial state That is, may be one of equilibrium, say V = we may expand zero in any number everyof ways in a Fourier series. To do this, we require an external source of electric or magnetic energy. may, f or example, insert a charged condenser and coil in sequence between earth and th e cable at the junction of x = Q and x = Zl. Let the initial state We be a charge in the condenser, but none in the cable. The initial arbitrary funct ion to be expanded in a F. series of a very special kind finite, is zero. The time factors t and express the potential at time is make the series be due to the condenser's we come to zero Finally, if the condenser potential again in the cable. charge.

leaky, may also require to determine a Fourier series not merely so as to express /(x) (including zero as a special case) between certain limits, but also so that cert ain functions of the Fourier series may have special values in addition. And then connected systems o f Fourier series are sometimes required, as when the line is divided into connec ted sections, with external conditions imposed at the junctions. Or we may have a network of cables, of the same type or different, We with imposed conditions, and have to find a set of Fourier series (or other kind of series if a cable varies in its resistance, &c.), to satisfy all the conditi ons. extent. And so on to any There is no difficulty in obtaining the operational solutions by the method of resistance operators in a clear and definite manner, and when t hat is done we are virtually in possession of the real algebraical solutions in the cases of simply periodic impressed forc es, and steady and impulsive forces, and arbitrary initial states, for the conversion from operational to But algebraical for m involves only formal transformations. we must now return to Fourier's theorem. The above remarks originated in the departures from Fourier's theorem required even when the funct ion under consideration is everywhere

104 ELECTROMAGNETIC THEORY. CH. VI. continuous and periodic. In the treatment of Fourier series to be found in certa in works, the reader can hardly find a glimmer of a notion of the subject in its general and comprehensive aspect. Special Property. Forms of Fourier Integrals. Interchangeable Use in Transforming Definite Integra ls. two special forms of Fourier's theorem in form of a double integral which should be noted. Eeferring to the diagram of impulses, 271, we see that when I is oo t he sines u, say u2 is reduced to represent merely a positive unit impulse at y = x, and an equal negative one at y = x. The cosines u, say ult on the other hand , represents a The formulae are positive impulse both at y and at y. 273. There are the , ^ = _l 7TJ 2 r cossycossxds, * (56) 2 f u^=-\ 7TJ So, if sinsysins#c?s. (57) we use these in the fundamental formula (24), or the result on the left side, which is f(x) when u is a single impulse within the limits as before stated, will depend upon If the limits in (58) are complete, v iz., the limits employed. oo to oo and we use ult we shall obtain f(x) +/( #). B ut

, if we use u^, we shall obtain f(x)f( x). For the impulses are double, and positive or negative as described. may, of course, choose the li mits so as to exclude both We x and -x. Then the result is zero. Or we may include just one of them, and get either f(x) or The most f(-x). useful way is to exclude the whole of the negative values of x. Then (58) applied to Uj_ and u 2 makes /-co /-oo I f(x) =_ Q f(y) cos sy cos sx dsdy, J f \

(59) 7T J f(x) = 2 [* -\ TTJ J f(y)smsysmsxdsdy. (GO)

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. 105 These are very useful in the mathematics of physics, The perhaps more so than th e complete formula (55). Given that following way of looking at them is interest ing. Then the symbols F and /are interchangeable. also That is, we have (62) There is a corresponding interchangeability in the other That is, if case. {68) then g(x) = (?)*f*G(y)*uixydy. (64) The corresponding property connected with the zeroth Bessel function is that if F(*)=f/fr)Jo(ay)ydy, Jo (65) then the symbols F and / are interchangeable. To prove them, put say (62) in (61 ), and we get (59). That is, we get an equivalent, the notation being changed. W e have dy both in (61) and (62), but of course in the double integral it is desirable to have them different. The above

is practice. can at once deduce another one. We Fourier's theorem often turns up in have an integral of the form (61) say, and w e the way It is not necessary to go the work of making up the double integral, and so thro ugh using Fourier's theorem explicitly. But it may not be convenient to have the (2/7r)* coefficient. Then, partially sacrificing it the perfect interchangeability, we may put thus. If (66) TTJ then it follows that (67)

106 ELECTROMAGNETIC THEORY. CH. VI. Similarly when we write sin for cos in these formulae. Examples are numberless. One or two will suffice here to illustrate. Say that we know that It follows that (69) to Observe that the limits in (68) are not complete. From y y = s, the f(y) func tion integrated has no existence that is, y must be greater than s. Consequently , in (69), y must be If y is less than s in (69) the value of tho greater than s . integral is zero. (We have assumed that x is positive in (68). - sign must be inserted.) we reversed the process and passed from (69) to (68), we should require to know that the (69) integral is true as given when y>s, but is zero when y<s. This will exclude the If negative, the If region to s from the if first Similarly, G Q (sx) is integral (68). the companion function to 3 Q (sx), the second kind of oscillating function, and we MU I know that G o

(sx) cos xy dx, (70) J when y is greater than s then we deduce s, but that it is zero when y is less than This is like being excluded. Reversing the process, we observe that the conditio n y>siu (71) has to be preserved in (70). Else we get zero. to y the region from y = =s going from (67) to (66). As an extreme case of (61), if we take /(//) to be a single unit impulse at the point y z we obtain F(#) (2/7r)*cos xz. Next, using this function in (62), we co me back to the impulThere is a sive function, in accordance with (56) above. = t = similar property involved in all normal functions, since they are capable of exp ressing arbitrary functions, and, therefore, a Observe that the (69) integral co ntains two single impulse.

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. 107 normal functions, so that we may not only deduce (68) from it, but also another integral, by employing the interchangeable property involved in (65). Continuous Passage from Wave its Series to Fourier Series, and 274. Reversion. Now, of getting Fourier series, a few words on the reversibility of the process Given a Fourier series operation ally. say f(x) = 2 A sin sx, its tionally ? series itself. Of course, can we find its meaning operaimmediate meaning is given by the F. But what the resultant meaning of the sum of the harmonic functions may be is no t evident, and it may be rather a serious matter to find it by numerical calcula tion, though after all the meaning may be quite simple, a is straight line, for matter for regret here, for whilst the example. solution of problems by F. serie s may be easily elaborated, the ease is confined to the formulae, not to their c alculation,

There which gets more and more difficult. One way here. of finding the meaning may be briefly mentioned By the introduction of a time-factor to every term of the F. series we may make it (if of a suitable kind) represent a real physical problem in diffusion. Then /(a-) means the have already pointed out the identity of the wave solutions and F. series. It wi ll have been observed that we can pass continuously from the waves to the F. ser ies. For we can construct the wave solution of a diffusion problem expressing th e effect due to a given source, subject to terminal reflections, by writing down the waves themselves in operational form, first the initial waves and then thei r consequences 255 to 260. Adding these together, we in order, as done in obtain the condensed operational solution, which may be independently obtained without thinking out the waves in detail. Finally, wf. 2an convert the result to a F. s eries, as explained in 265 to 270 in special cases, and in a general way that wi ll initial state. I Now come later. It is certainly reversible on one Is this process reversible ? side, for we conv ert the operational solution to the wave form It is not only in the simple cases of terminal earth or disconnection that the meaning of the indiby algebraical division. vidual waves can be ascertained, though I have not given any

108 ELECTROMAGNETIC THEORY. CH. VI. we can advanced examples in detail yet. As regards the other side, plainly pass from th e diffusion solution in a F. series to the operational solution by simply reversing all the steps. Keferring to exam ples before given first, the constancy of pt power of the time-differentiator on e enables us to put the circular function occurring in the F. series outside th e summation sign, by taking s 2 to mean ES(d/dt). Then the summation that is lef t, or 2e p ', may be converted to the operational form 2(1 P/7?)" 1 This series of fractions may then be summed. The final result is an operational formula with unit operand. . the reader will apply this reversed process to the he will find that he will not examples be able to do it safely with his eyes shut. This is because in if But in detail hitherto given, going from the operational equation to the F. series, certain So in the reversed operations are impotent, and are omitted. This can only be done by process, they should be restored. careful inspection of the problem, and is a question of expeThe terminal conditions satisfied by rien ce and judgment. the solution in F. series should be examined. satisfied all the They must be reversal to be effected. complete But, though there may be troubles of the above kind to be surmounted, i t is not a useless way of transforming from the F. series to waves when the sour ces way through, and this will enable the On the other hand, should the sources are of a simple nature. be not simple, as in the case of subsidence from an " arbitrary" is the case is altered, on account of the integration which involved in the determi nation of the coefficients of the F. We may series, which complicates the matter considerably. state, therefore leave this question on one side.

How to find the Meaning of a Fourier of Series Operationally. way regarding F. series operationally will be conveniently introduced by conside ring an elementary example. 275. Another Given ,-(*) = ?Srlsin^. make a t (72) Find what f(x) means, without numerical also without introducing time functions to calculation, and diffusion problem. We see that /(x) is an odd function of x and that

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. it is 109 from x = periodic. to It is therefore sufficient if we know its x= meaning I. of x in the for djdx. Regard the right member of (72) opera lionally as a function same way as our pre vious functions of t Let A stand . Then first This replaces the former />, the time-differentiator. we have to convert the ser ies to operational form as a function of A. This is easy. For we have 11 _ _ y A Similarly, in passing, we see that cossx i+(s/A = same with

/-(*) this formula. namely. have already had occasion to employ So equation (13).1 for numerator. and makes = 5 . That is. Next. we imagine the operand t o start at # = 0. Now this equation (76) expresses the full algebraical meaning of f(x) on the ass umption made. From x = 4Z to 6Z. and makes f(x) auxiliary = 3 . +2 +2 + .21 We we produce nIK l 2 . (75) with unit operand understood. Next. expand the coth function in (75) by division. Thus. the second auxiliary exists as well as the first. with gl instead of /A. and be 1 afterwards. How these jum . = cothZA--!. the term exists. first From x = 2l to x = 4Z.. and so on regularly. It is l-x/l from x = Q to x = 2l. All the following terms are then zero.. hi getting (73). using (78) in (72) . (76) where we have also algebrised the (ZA)-1 1 in (75).x/l. that it begins at x = 0. This adds on another 2.x/l. Its value is 2.. The value of f(x] is explicitly r epresented by (76) from z = up to # = x .

ps come f(x) about through the auxiliary terms will be explained next. .

tance of . Imagine any wave form to be transmitted at a constant speed undeformed. Let it be as in the figure. CH. narr*ely. position at the moment of time t =Q A. a hump running along a flexible cord. for instance. 276. It is to be imagined to so that the distravel at uniform speed v from left to right. as. a triangular is hump whose some vt. and at later time t is B. VI. Taylor's Theorem in its Essentials Operationally Considered.110 ELECTROMAGNETIC THEORY.

It is a positive wave. Differentiate equation (77) with respect to x and t separately an d compare them. and let it become They only differ in the change of origin. Let f(x) represent the function when at A. We see that a* . the same as the value of / at a distance vt to the left. and the minus sign must be ch anged to plus. " function the axis of x. Their F(x) at B. Equation (77) expresses the characteristic property of an undistorted and unattenuat ed wave when the speed is constant. the travell ing form is our x" value being the ordinate of the waveform. that is. relation to one another is F(*)=/(*-rt). When it goes the other way. it is a negative wave. the value of F at x at the moment by definition.AB its is base line is If the increasing uniformly with the time. (77) t is.

dt (78) This is the characteristic partial differential equation of a It is easily under stood by watchsolitary undistorted wave.dx v i. . Equations (78) and (77) have pra ctically the point. ing the transit of a particular part of the wave past a fixed and considering how the speed v of transit combined with the slope determines th e time rate of increase of F at the fixed spot.

That (80) is the solution is ay be seen by putting t = 0. when considered to exist only on the positive s ide of the origin. What / is m F =/. Then herefore the same as the former/. Now write (78) in the form ^=-"AF. we found that a certain F. series.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. understanding that / expresses an arbitr ary function of x. Ill same meaning. Thus. translate it bodily through the distance h to the left. This makes F = e~ r' A/. given (78). was identical with is to (l_z//) to t. its solution is (77). when applied to a function of This is A is ar. (80) where /is a constant with respect f x. In our present case it is any function o verifiable by its satisfying (79). and solve it as an ordinary linear equation. and is t Comparing (80) with (77). we see " vt ^ e . where (79) A stands for d/dx. so that the universal property of the operator e^ when h a constant. So / is the initial state of F. "What it does. Applying (81) to the res ult (76) which gave rise to these remarks about uniform motion of a function. that the operator is the translation operator.

. Denoting the auxiliary terms in (82) are equivaf(x).+ 2( e -'2' A is + . the The result is that the next when x r eaches 4Z. (82) where the operand 1 this operand by only positively existent.)l. (83) The value of every one of these functions is 2 when it exists.4/A +e .6ZA + . . and so on. 1 which itself down .. 6/. as passes the #=2J.. But the first one only begins to exist when x reaches 2. constant 2 points is added on every time the variable 4Z.. lent to 2/(*-2J) + 2/(tf-4/) + 2/(z-6/) + . falls &c.

The same series of values is then repeated between x = 2l and 4Z. when x = 2l. liary to its starting value oo is lifted up by the first auxi+1. and which would by .to The function (l-x/l) therefore. Then the second auxiliary begins and lifts up the functio n to its initial value for another go on decreasing to .

and If it is sufficient on the gative e. work has been already done equivalently on the (81).112 repetition. or. ELECTROMAGNETIC THEORY. For it is merel y the condensation of the form obtained by expanding the exponential. to note that positively since the the other way . : (85) . of values of the function. is the well-known so-called " symbolical" form of Taylor's theorem. and series it is negative side also. Equation which is the same. The F. an odd function of x. series therefore represents (1 xjl) between followed by unending repetiti ons of the same series 2Z. positive side. The ne side may be done operationally by itself by ignoring the part already don reckoning x to exist we allow the F. VI. to see its nature there. CH. viz. but this is quite unnecessary.

They would. is not the usual way of leading to theorem. But there is no difficulty of the kind in the physical way of loo king at the matter. But now comes the In fact. and hedge round it. It is Taylor's a quasi-physical way.The form (84) is much more convenient. are largely devoted to the discussion of modifie d forms with remainders occurring when there is some discontinuity. One shape of the function is just as easily . by which its truth becomes obvious in a ge neral sense. or infin ite curvature. and the meaning is simply a bodil y translation performed upon it. for which see works on the Calculus. When matheinvolving an in finite differential coefficient. for example. maticians come to an infinity they are nonpluss ed. so far as the act of bodily translation is concerned. stick at the three sharp corner s in the function above used. we get solutions of the form c^ acting upon something or other. which involve discontinuity in the slope. It only involves the idea of translation of a The way followed above form so far as the essential part goes. It is the form that That always occurs naturally in operational mathematics. A is. the proofs of Taylor' s theorem question of failures.

the whole thing breaks down if the s lope is infinite. remark is particuand their propagawhen discontinuous functions have frequently to be dealt with. The in the discussion of waves larly impor tant tion. Although. and t his tie does not break when they are momentarily infinite. not uninterpretable . the upright piece But (78). and has all its derivatives finite and continuous. The variations in dF/dx and those in dF/dt keep pace together precisely. and requires no special interIf f(x) makes a jump suddenly pretation involving a t any place.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. which is taken to be it equivalent to (77). not merely (77) is unexceptionable. It was long ago remarked by Sir G . and we are not limited to the angelically perfect function which is finite and continuous itself. they are n . so does F(#) precisely at a corresponding place. (The translation o f the form should be thought of. Stokes that it is important for physical mathematicians to conceive of functio ns wholly apart from their possible symbolical expression. of course. in a popular sense. inasmuch as utilises differential coeffishould not say that require interpretation. Now that of a value. infinities. behaves just cients. like the rest. 113 conceived as another. because may We It requires interpretation. but is (78) becomes meaningless.) If the curve is upright. infinities are immeasurable.

to enlarge the scope of the VOL. for example. Consider. it may be no bigger than zero. and therefore trification. If suitably measured. Starting with a finite volume density. i . n. doubly infi nite volume density. It is now to be measured by its surface de nsity. Again. we have infinite volume density. or at any rate in the conception of the suitable way of measuring. A suitable standard is required to measure an infin ity. or else quite small. Again. though it involves the idea of an infinite force. the notion o f impulses in dynamics is exceedingly We useful. Now I think that similar ideas may with advantage be introduced into pure mathematics.ot necessarily unmeasurable. imagine it condensed from finite surface electrification to linear elecThis means infinite surface density. must not be afraid of infinity. But we now have a finite linear density. the case of electrification. But that causes no trouble. so there is no dif ficulty in measurement. if we imagine it condensed on a surface.

is but just jump over. the subject of opera tion is real and single valued.. no matter how many di scontinuities may be in the way. definite series of values. Let it be applied to any infinite values. A Fourier Series involving a Parabola interpreted by Taylor's Theorem. concerned with the real or supposed failures in Taylor's theorem. it will be worth while to consider another example of the same sort. perhaps rather to be considered failures in the In the operational mathematics the translational operator. we are not in finities. CH. coth/A f(x) = (1 + 2e put it thus ~ 2/A ~ 4^A : = and if +2 _ _ /M A ' + 2e . Since the investigation in 275 shows one the working of the inner mechanism of a Fourier series in an interesting way. 277. In physical applications. mathematicians' investigations. though it need not be finite and continuous.114 ELECTROMAGNETIC THEORY. or perhaps to the case of a mu ltiple pole. We do not have to consider the mathematical machinery. We may . VI. say to waves. involving a collection of infinities. But I do not see any reason why the translational idea should not be carried furthe r. and enable them to tackle However. instead of evading them. or of multiple values.

Next take /(a) =x. . .)/(*) (86) (87) ' I we like to confine ourselves to the region to 2.x/l considered 275. we may omit the translational auxiliaries. and write Take in f(x) =1 to make up the expansion of 1 (88) .+ . Then becomes since z Ar 1 .

2 !. Continuing the algebrisation we obtain (90) . a: = A~ 2 l.

and these are t he same as x 21. and so on. and is positively existent. they are zero. Now/(#) is born when a = 0.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. &c. After x = 2l. that in (92).4Z). with a particular value x. namely. namely. (87). to 2/ we have the curve The ordinate y is J at x = Q and 2/. and has its minimum. it would rise up i.. according to (86). which is the same. is where the terms with square brackets are the auxiliaries. values. So f(x-2l) is born when # = 2Z. when they exist. x 1. midway between. They are to be underst ood in this way.f(x . to infinity by itself. The final result is of simple enough. TThis is true between x = But the full and 2Z. &c. meaning. or. just as many of the auxiliaries exist as have positive so the auxiliary functions (not doubled) are/(# . but the first auxiliary . f(x) is to be x. The In the region x = rest are not yet born .2l).. including the auxiliaries. In (86).

and causes y. = and 21 repeat themselves between x = 2l and Then the second auxiliary comes into play. (95) Comparing the last with (93).is now existent. we see that they have the same So the values of fo rm. only x in (93) becomes x . another repetition of the same series of values of And . y between a.2Hn (95). Adding it on. we get * = a2 Sx 2T*-T + 13 ir (94) This is the same as -^-(^-H x= 4l.

i2 .observing that it refers to the positive side of the x axis.so on to infinity. On we see the meaning of the Fourier series by is an even function of x. The above the negative side .

We between certain limits.. or by formulae derivable from them. may be expressed in series exponentially (e. may be effected by the formulae for the expansion of coth JA. VI.116 ELECTROMAGNETIC THEORY. general way. in partial fractions.g. g iven manner of 275. The trigonometrical functions of /A . likee" &c. is CH. The coefficientsThe summations are supposed to be known funct ions of n. Thus. comes into play.. if you can. (96) This may be converted at sight to . Now sum up the partial frac tions. in its ge neral translational sense.&c. and then Taylor's theorem. and the full meaning of the F. and endless repetitions of the same values caused by the existence of the auxiliary terms involving the translational The example in 275 is fully descriptive in a operators. combined with the explanation . comes another important . serie s becomes evident. by the use of (73) and its cosine companion formula. When this i s done. The reader who curious in this matter will find it useful to treat various simple Fourier series in the to discover their meaning. 1 (shin /A). step.).

41. by some e xtension of the process. turns it into the applied sum of f(x) and of twice the sum of its va lues at the points278. Probably it can. Representation of a Row of Impulses by Taylor's Theorem. to a positively existent fun ction of x. because the summation of the partial fractions may be a troublesome matter. and the But the application may be made to a variety of formulas. . determine it to represent a certain function method is not meant for general use. t &c. so that they may be regarded as functions of n. But it may be necessary to discover first the law follo wed by the coefficients. By equation (86).in 276. we see that the operator coth /A. x-2l x. The question arises whether a Fourier series with its constants given numerically can be done in the same way. leading to Fourier's.

117 From and 21.f(x). Therefore (98) represents f(x) and periodic repetitions of the same (not is doubled). this. as in (86). from x = zero except between to oo . and then an infinite series of repetitions of the values of the same function d oubled.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. if f(x) is a2 we get . in regions of length 21. the series coth/A . in its proper place. represents f(x) first.4) 2 and so . every one by itself. on. For example. 21 f. we conclude that it if meaning thereby that is f(x) exists only between zero outside those limits. first #2 next 2(x . then 2(x .

And by a slight extension we see that left then F(#) represents f(x) and periodic repetitions to right and through the whol e range of x from negative to positive infinity.same series F(x) has the meaning as Fourier's . as than ZA. It will be sufficient here to derive Fourier's theorem in its general form from the above. By construction the . But if f(x) be so chosen that its initial and final values do no t agree.. and their powers as operators on a function. there is a jump in the manner before alluded to.r) is momentarily violated periodic property F(x + 2l) unless we repreas re gards a single value of the function sent the vertical part of the curve by the mean value. and the reciproc als of shin A and cosh JA. The = F(. Only one term exists (or is fin ite) at a time. Go back to (99). the reader may amuse himself by considering the exponential expansions of o ther trigonometrical functions of a differentiator. Now. still understanding that f(x) if and 21.

and (100) shows how a single value of the function is isolated. . (101) Here u is expressive of a unit impulse at the points x 2nl. as before explained.(100) where -l + 2Seos".

. this u represents the periodic unit impulse.x.)Ai._(i+-a*+ e -a + . thus : Wl = i(l + 2e. we see that (102) must be the operational form of (101) itself.118 ELECTROMAGNETIC THEORY. and nothing anywhere else..2?A + 2 e = i(l + 2 2'A . . by construction. . positive and negative regions separately.)A1. . But to algebrise (102) it is convenient to consider the Sp lit u into halves. CH. (103) (104) ^2 +2 4'A +. . VINow Al is a unit impulse at the origin of the variable. like (101). = So. . Here u^ is for the positive meaning half a unit impulse at at .)A1. by (99).4ZA + . (102) Since. . Let Then Al signifies a unit impulse at the the variable be y . we have point y x.+**+*+. .

/A. u^ Now (103) is equivalent to %= fJL coth/A . But it h as not the same meaning. &c. x at the origin for the negative. and whole ones at x 21. &c. and whole ones = x + 21. the algebrisation is immediate.1. and u 2 y = x. is understood. side. y = x. Eemeniber the reservation that y-x ranges from to oo on ly in connection with u r In (105). whilst u 2 means half a unit impulse y . (105) This is equivalent in partial fractions to (106> Finally. making u This result shows the same expression identically as Fourier'sin (101).. x + 1. (106) the positively existent unit operand side. So (107) is limited to the positive .

so that u z is represent ed formally by u lt we find . from the symmetry of the complete u in (102) with respect to the orig in y = x.Similarly. or by reckoning y-x positively the other way.

that w 2 is 119 given by the same expression (107) . this v function is continuous.2 alone. right side. It seems somewhat unnatural to consider the right and left And yet it is necessa ry. the impulse periodic both ways. i f at y = x. we me the range of its application. The complete Poisson's demonstration is interesting in this connection. we find that the u of (101) mea ns the same as that in (102). So. Evidentl y Wa is inoperative. the function is A on the left and B on the is. we ed in 2 has the same expression as t is an even function of y . w x will give JB. Its total . Again. and uz will give JA. Suppose f(y] is zero on the left side of y = x. Instead of u he used the function (108) It becomes the same as u when h = it l. and finite instantly on the right side. that and u r alone works. It gives only /(#). the trigonometrical formula because i its use to represent w x or it. evident at a discontinuity. This becomes halves separately. When unrestricted. In rely restrict this case as w r Only in y has to be not greater than x see that the sum of w x and No change is need either. removing the restriction. u will therefore spot the mea n value. Now when between h is is < 1.x. J the value on the right side.PURE DIFFUSION OF ELECTRIC DISPLACEMENT.

close round the point y = x. if h is very close to 1. So . Now its is symmetrical with respect to y = x half on one side. 279. we sha ll obtain i (A + B) approximately. area make h approach unity no reason why a reached.and 2Z is 1. I have said rather first. more about Fourier's theorem is than I meant at adding one more . and has a hump at the point y = x. and half on the other. like u. and we use v in the formula (91) instead of %. I know failure should occur just as perfection is Laurent's Theorem and Fourier's. and the more closely we the curve of v is . the closer the approximation. The closer h made to approach to unity. the more the v function is heaped up at and In the limit it i s all there.

That it. a good enough reason for it. about in order to have done with .

120 ELECTROMAGNETIC THEORY. have necessary periodicity of funct ions of the position of the point. round . r and Assume that/ of z. The way Fourier's theorem presents functions that the theory of manifestly is. and may be repr esented by the position of a But a circuital journey round the origin point in a plane. (109) Then we may easily determine the coefficients. In this in the plane brings us back to the starting point. in the manner of Fourier's theorem. round a circle centred In at the origin. VI. except zero. which is therefore position essentially involved in the expansion of complex functions of in integral powers of the variable denoting the position of a point in the plane. variability. which is confined to limited notions of a f unction is The complex variable has a double worthy of notice. So. So we process the variable 9 increases by the amount 2?r. there is Laurent's theorem for the expan sion of a holomorphic function. where z is the complex re^. is 9 being the polar co-ordinates. Let /() be the function. capable of representation in integral powers positive f(z) and negative. The mean value of zm regarded as a function of 0. itself in CH. say = A + Aj = ZA n3B oo . multiplying equation (109) by z~n and integrating with respect to 0. . Thus. in the theory so-called thereof. is zero when m is any integer. that exceptional case the result is 1.

So. the value of An is the respect to 9. (110) That is. by (109) and mean value of (110). a circle centred at the origin.) \Z / (in) the powers of Fourier's theorem generalised by the introduction of r. we get (.) 5-^ = An . make r = r'. we have f(z)z-* with 's This is <D -co (*. so as to . To obtain it explicitly..

be concerned only with a function of Then we get Fourier's theorem : .9.

the n's following some special law. Say. theorists would be would of some interest to know how the function treat the subject of Fourier series in general.*". In the above. and when the function may is tions with r controlled by the restricted functional notions. (113) . For a rigorous demonstration of Laurent's theorem see works on the theory of functi ons. 121 0'). In general. the expansion is the rudimentary one in i ntegral powers of the variable. that f(z) is given on the circ le Then we require to have of z'. (112) substitution of the cosine for the exponential function occurs because the imagi nary part of the latter contributes th th nothing to the result.w in this The But the imaginary part must not be omitted in the more general result (111). The variaand with are not indep endent. The values of It the function on one circle control those on another. fractional powers are required. The n term canc els the . referred to the complex. for example. be any circle within the holomorphic region.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. so Laurent's theorem is the natural sequence to Fourier's when the distance from the origin is not invariable. The line of integration respect. /W-ZA.

and another similar equation on the other side. make a satisfy certain boundary conditions. . where = 2-. and. where derivatives with respect to 6 = 0. conjugate property. to For example. and let the function have to satisfy a given linear differential equation between it and its on one side of t he slit. series of n's. not that of the vanishing of the mean value of zn round a ci rcle as before mentioned. but of a more I general character determined by the co nditions at the slit. in addition. radial slit in the plane at 6 = 0. The infinite which may be real or imaginary. This case cor responds to that of a cable subject to terminal but not to extra intermediate conditions. will follow a law depending upon the conditions at the slit and the determination of the coefficients A may be effected by a def inite .so as to be right for the given circle.

and the corresponding cases in the function theory would. is . especially when one thinks of the various extensions that may be given by the introduction of interMatters of this sort da mediate differ ential conditions. At the perhaps. Of course. VI. have looked into Forsyth's work to see if there is anything about these expansions in fractional powers of the v ariable. not present much difficulty in strictly physical problems. is. same time. ledge of the complex is necessary.122 ELECTROMAGNETIC THEORY. some useful knowsupposed or real necessity. there is room for much expansion of the subject of the determination of functions. CH. of It is a serious question whether the study of the theory of functions ought to b e taken up by any ordinary physical student. and no one could fail to but that pick up som e as he goes along by ordinary algebra . involve little more than fo rmal changes. may be much which perhaps. quite different from the theory of functions as elaborated. but without success. the general tone of which is quite unlike that of the usual The frame of mind required is not physic al mathematics. I must say that a matter which sufficiently clear in physical mathematics may become obscured by a deliberate r emoval of the physical ideas. If the matter has not been investigated. the reason of the existence of such elaborate and disagreeable demonstr ations for the sake of rigour.

by which (in one way) they become amenable to numerica l calculation. tions must be treated in order to convert them to the ap propriate Fourier series. It may be remembered that I have insisted upon the definiteness a nd fulness of meaning of an operational solution. . and that it contains within . he had to worry over all the niceties of logical mathematics under severe restrictions bound say. that you are through a gate.one that its is conducive to progress in mathematical analysis in A man would never get anything done if physical aspect. little 280. although that action would at once bring you to th e goal. but must on no account jump over it or get through the hedge. After this of duplicity excursion to the borders of the realms and fearful rigour. to -go On Operational Solutions and their Interpretation. for instance. we may return to the proper We have to see how the general o perational solusubject.

standing of the subject impli es the existence of a full theory . with place though. sion. that the solution is the sum of a number of that No external aid type. itself 12 & not only the full statement of the problem. and is in a great measure independe nt of logical demonstrations. Thus.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. that is a matte r principally of observation and experiment. parts of the A compl ete logical undercourse of examination of the How. rather than the first . The conversion to algebraical or quantitative form may be But in any case it easy or hard. tion. but also is its solution. self-evident or very obscure. by the prior construction of the operational solution. Why. the conversion furnishes a distinct subject of which is of great practical value from the physical study As regards finding out how to effect the converstandpoint. of a special type of soluassump tion. fully. is possible. with subsequent determination of the constants required The work of satisf ying the imposed to complete the matter. cf cannot fail to be perceived in the coarse. conditions has been done already. It is the How. no and therefore required to algebrise it for instance.

so does the theory improve. This is a most pernicious doctrine. and even then it is bound to be only imperfectly logical. you do not prove anything . As But the subject open s out. it can only become logical when the subject i s very well known indeed. and. Does anybody fully understand anything ? Three of the pernicious results of overmuch rigour may bementioned here. It is come first. First. I feel inclined to be rather emphatic on the matter of the use of experiment in mathematics. when applied to imperfectly explored regions. unless you follow regular paths. others not. and that you are bound to fully understand an d rigorously prove everything as you go along. even without proper understanding. for the reason s mentioned at the beginning of this volume. Observation of facts and experime nt. important to note that it is just the same in mathematical research into unknown regions as in experimental physical research. suffi- .which we are mainly concerned in the Why to account for why certain ways of working are successful. For there is an idea widely prevalent (though it may not receive open expression) that in mathematics. its enfeebling action on the mind. with merely tentative suggestions of theory.

should it be unconventional. Secondly leads to the omission from mathematical works of the most interesting parts of the subject. and be devoid of rigorous pretence.124 -ciently indicated it ELECTROMAGNETIC THEORY. It may not even be desirable to do so. . and the produces complication. Thirdly. if t he meaning is sufficiently plain in operational form. (a + bp where e is the operand and p the ti me differen- . it leads to an inability to recognise the good that may be in oth er men's work. It may not be necessary t o algebrise an operational solution. merely direct ones. as. . because the authors cannot furnish rigorous pr oofs. for example. CH.)e. VI by the analogy at the end of 279. -operations are This is very often algebrisation true when the + cp 1 + tiator. .

we do not require to algebr ise these forms in order to see their meaning. or (X/Y)*<?. in general. for i nstance (XY)*0. operand e is a simply periodic function of the time. Again.But it is also true. . Consider. in a reasonable time. by means of a specially constructed diagram. and how it went forward with g reat bounds when algebra came to its assistance. I see that Mr. method of representing it is suitable for calcuFor geometry by itself is rather a heavy and clumsy machine. or B m0. are operators of the form a + bp. Remember its history. a geometrical Sketch of Way of extending Fourier's Series in General. however. Kennelly has lately dispens ed with the use of the full development in the more complicated cases of the shin cosh of a complex (the com plex corresponding to a + bp). up a certain point. in more advanced forms. and then see if. We can do that by geometrical con struction in the now well-known manner founded upon the property p 2 = . or K. to do the work and analytically first. It is best. Method to Fourier special ways which I have so far given of turning solutions of the diffusive kin d to Fourier series operational 281. or p = ni. desirable to hav e the full algebraic formulae. representing X and Y by rotations of vectors in a But if we desire close calculation of the results. at least as far as the operational solution itse lf. the assistant became the master. it is plane. where X and Y. I do not doubt that certain cases in which e is impulsive or steady can be done by d iagram. lation. Later on.?i 2 If the .

The .

determined except as regards size. by applying the terminal co nditions of diffusion. to the above assumed solution. or JTT. as TT. they may be (as regards s in the The elementary expancircular functions) r eal or imaginary. There are now three ways that present themselves of attacking the question. to follow up Fourier's way by natural extensions of the kind to be found by experience. Next. But. 125have all been applied to elementary examples. are not equally spaced. First. the roots of the determinantal eq uation ql af3.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. Ij7r r c. Apply it to the actually stated problem. and we are allowed to use well-know n trigonometrical expansions. &c. sion formula are useless. and find a and I. the terminal conditions being such that the roots of the determinantal equation have been equally spaced. Find the connection between P and a so as to satisfy the characteristic equation Next find. is . in general. We thus get an eleme ntary type solution fully their values. 2 7r. assume that where the summation includes all the values of a n . what relation exists betwe en Then find the law followed by the a's. and they may not be all of the same kind tha t is. and w e know that every admissible value of a gives a solution. where a and ft are functions of p depending on the terminal conditions. a normal solution of the form v Assume the existence of = A sin (ax + 1) e pt . Algebraic simplicity is thereby attained..

observe that t = Q makes The summation on the state. we requireto determine the A's in accordance therewith. any To find them. Only the A's are arbitrary. property allows us on V by quadratures in such a way as to isolate any . to operate the conjugate it When known. the sum of number of the type so lutions. namely. This may be done by finding possessed by the normal functions.the form of the solution of the special problem. Given the initial state right therefore represents the initial Y as a function of #.

that is to say).v is. with energy involved in the terminal The conjugate proapparatus. Finally. The integral of the product of two normal s ystems of vanishes save when they are alike. it is quite another thing. process of finding A. VI. works out more difficultly. sufficiently obvious. But with terminal conditions in general. electric over the the complicated process. is not the same. to discharge itself in the above manner. requiring terminal arrangements themdepends upon the vanishing of the excess of the magnetic mutual energy of a pair of different normal systems (with different values of j). may be done The and allowing that are similarly by considering the final state it produces. introducing the time factor . single coefficient and determine its value. Operational . the two terminal arrangements have themselves .128 ELECTROMAGNETIC THEORY. if any part of the energy is located therein. when the terminal conditions the simple kind previous ly considered. The Expansion Theorem. In d oing this. CII. and alth ough fundamentally simple perty in meaning. 282. provided the effects are And the effect due to an impressed force enti rely due thereto. The dete rmination of the A's It may become of a careful investigation selves. we arrive at the complete solution as regards the effects due to a given initi al state.to be included.

Way of getting Expansions in Normal Functions. First. since the theory of partial fractions may lead one into obscure regio ns of the complex. As a second way.a/3) which occurs in the operator. Finally. Com plete the solution by carrying out it The expanleft in the way of direct operations. required are analogous to those o f coth^ and other But this process may be rather functions already used. It does not require special investigation of the pro- . fractions follows. there is a third and very general way of converting operational solutions to the form of the sum of normal solutions. wherein p has to have all the values which make The integration of the partial (the determinator) vanish. Next investigate the proper expansions in partial fractions of the reciprocal of the determinator (e 2<2Z . what may be sions difficult. find the operational solution. we may follow up the special ways of operationally working alre ady given in connection with simple terminal conditions.

that C is the current due to an impressed force. more generally. which covers the rest of the work. magnetic problem at a certain place briefly (though imperfectly) stated as be the operational solution of an electro say. But i n Z is entirely algebraical. beginning at the moment t expressed by dp to be understood thus: In the first place. Z in the operational solution is an function of p the time differentiator. a large amount of investigation of the conjugate proIt applies to all kinds of series of normal functions. when e is = 0. Z is the equation (1). or. e.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. in general. with a finite or infinite number of vari ables . Let the form of Z be such as to indicate the existence of normal solutions for C. perties of the 127 normal functions. The unnecessary work. for definiteness. Thus. is wholly avoided. as well And it applies ge nerally in electroas to Fourier series. it to the determination of the coefficients. the C due to e is steady. magnetic problems. and of the terminal apparatus in detail in order to apply perty. It It is very direct and uniform of application. Then. The method may be follows : Let e = ZG . It is the algebraical function obtained by putting p = Q in Z. at the same or some other place. the operator. avoids. and upon which he bases his discussion of general properties. to the system of dynamical equations used by Lord Eayleigh in th e first volume of his treatise on Sound. .

a effective voltage. which determinantal equation. and many other varia tions might be mentioned. itself much more. and steady resistance to e when. though Z as explicitly. in the summation. as supposed. it may be V due to e. differential coefficient of Then. dZ/dp is the ordinary Z with respect to p as a quantity. e it is at the place of C. all the roots of this is the the equation Z = 0. Lastly. Instead of a in respect. or V or C charge at a point. more general. But it must not be inferred that equaC due due to an initial . the algebraical summation ranges over is. Otherwise i t is is a. These special values of p are to be used in dZ/dp as well to an e.

subject to terminal conditions. certain ways. to one of which the expansion theorem applies full y. Dynamical conditions restrict the form in But it is not even true that the for one thing.ql . immediately excluded from the expansion theorem. I have already pointed out that we require barriers or boundaries to produce definite a nd finitely separable vibrating normal systems.a/3 which occurs as a denominator in the general operational solutions arises from the coexistence of successively (though apparently simulSo all sorts of problems are taneously) generated waves . at least in the form of equatio n (1). which . Even s upposing that an absolutely perfect knowledge of the subject made it possible todo so. tions of an obvious nature.128 ELECTROMAGNETIC THEORY. Inspection of the operational equations will show this. CH. VI. it would be very unpractical. whilst the other requires a different treatment. It would be worse far worse than that very lengthy enunciation of a theorem in the 5th Book of Euclid. which may be read an d re-read fifty times without properly grasping . It may involve direct operaNow it ciation to would be useless to attempt to state a formal enunmeet all circumstances. tion (1) expresses a general mathematical property independent of the form of Z. the real Z may be the sum of two operators. or else physical considerations. Also that in the cable problem. its meaning. expansion theorem above holds good in strictly dynamical problems. the determinator e. Again.

just to show how the expansion theorem goes.is not. to begin with. and how to discriminate different forms of Z. it may be possible to At present all that need be said is see its inner meaning. we can apply it to diffusion problems in the same way. After that . So. to learn the nature and application of the expansion theorem by actual experience and practice. much. I will give a few elementary cases concerning on e or two degrees of freedom. afcer all only something in compound proportion that It is better the mode rn schoolboy does in a minute or two. And when the reader knows how to work the theorem. A theorem which has so> wi de an application is a subject for a treatise rather than a proposition. Other than diffusion pr oblems coming under the same theorem may occur later. that in its theory it contains the first method of evaluating the .

C the current and First. Start with the characteristic equation of a coil. namely. Say = (R + Lp)C. e the impressed voltage. : 283. let it it. and zero previously. and also the second partial fractions. coefficients 129 method involving by the conjugate property. it First we the . e2 is an impulsive force of total LC. e l = RC to keep up the current. and e2 = LpC to start Since pi is a uni t impulse. Only it jumps over most of the difficulties connected with both. be given that C is steady from the initial moment. Examples of the Use of the Expansion Theorem (1). Inductance Coil and Condenser separately.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. being the form taken by t he second circuital law when applied to a linear circuit of constant resistance. where (2) R is the resistance and L the inductance of the circuit. We want two impressed forces. What is the corresponding e ? The answer is immediate.

But now find the C due to steady e . Here Z is R p is . But the treatment is sufficiently to keep Now If we tried to force it into the expansion it. so. of a proper form for the expansion theorem. . the operational way of also be indicated. not of the working of the expansion theor em. theorem we should want the root o f (R + L/?).momentum LC this is require the impulsive force establishing instantly. by (1). showing the gradual Since this case getting the result rise of is may C to its steady value.1 = 0. and then the steady force RC up against the resistance. an example. but of its failure. By (2) the answer is clear without TO which is .R/L dZ/dp is L . We have . fundamental. which has no finite root. which is quite a different problem.

R + Lj. _e_ e _ \ fR R \Lp . n. LX1 + B/Lp) VOL.

let e will be observed that the case is that of equation (3). by the meaning of the exponenThe steps (5) and (7) to obtain (4) may look c umbrous. perhaps. comes slightly modified. for it is.130 ELECTROMAGNETIC THEORY. applied e to (5). ^l/By_ (3VL/ B (6) fB*_l/Be\ [2_U/ \ (rj} &1L which is j' (4). once done. VI. It course there is no further need for the intermediate work. But there is deception here. and in However. the expansion of tial function. This is got by expanding the fraction by division. in in the customary use of the short expression of the exponential function to d enote a certain infinite series. Then equation be impulsive. 265. of taking its proper ties for granted. say = pE. CH. done by The rest is which makes. where (3) above makes . a The deception shorte r and more direct way than any other. To obtain it as in the equation referred to.

The rest is as in it up. >E B + Lp _ ~ E L (1 =E _R. The impulse E produces the equivalent There momentum E = LC.E is the impulse total. . so there is no steady term. ~ Q_ as in (3). so equation (4). it no impressed voltage to keep decays according to (8).t/L + R/Lp) is L /gx 265. only multiplying by p or R/L. In doing (8) by the expansion theorem. notice that p = makes C = 0.

(10) the conductance and S the permittance. or one have (equivalently) by a shunt. ^changed. and permittance. We ' C = (K + Sp)e. made conducting "Where is a conducting leyden or condenser.Thus. by (1). and inductance . but cur rent and voltage are interis K as well as conductance and resistance. where 'p has the special value which makes Next consider another fundamental cas e. that of a leaky condenser. The form the same as for a 'coil. that is.

either by translation of symbols or independently. and an initial impuls ive current of total that is. 284. 2 . the current will be is the This current constant.PUEE DIFFUSION OF ELECTRIC DISPLACEMENT. the total curren t in the condenser and shunt. disappear. It is like So equation (3) above. and then it must be maintained constant by the steady current Ke. Say and inquire what e is required to make C jump from zero to constancy. supplied The Treatment of Simply Periodic Cases. we require the steady current ~Ke. a nd maintained constant. Here the e xpansion theorem applies. during the whole time. In another form. of course. the rest is in the condenser.. which goes thr ough the shunt K. current. to constancy in the if way the voltage be made to rise from zero indicated by (12). part hi the shunt is The For I\e . is convenience we assume that the conductance externally. Now invert the problem. Without the steady current the charge of the conden ser would lished. first the charge Se must be impulsively estabS<? . viz. C. 131 In order that the voltage e should be instantly raised from zero to constancy. answers the question.

say. the perty i> = working is perfectly simple. by means of the pron* which obtains in simply perio dic states. when e is given instead of C. given as a simply periodic function. or. applied to reduce the resistance or conductance operator to th e standard form a + lp. obvious. already. which the same. for (13) is in the standard f orm In the other case. (13) when it is C that is is G=csm(nt + 6). we have . The case is e = (R + Lp)C.As regards the simply periodic solutions in these cases. p = ni.

it should be noted that i is not the imaginary of algebra. Thi s is essential. follows that a simply periodic e produces a simply periodic C. as I have poin ted out before. perhaps often. we bring C to the standard form by multiplying the numerator and denomi nator by the operator conjugate to R + Lp. C= This it Lj9 is =--^-. .n 2 in simpl y periodic But it must be finally interpreted correctly. so that ' e = ~LpC . w hich may be that is. then if e iswe obtain. by p = ni.Lp) This realises the de nominator. L/i 2 . Whether this is done by^> 2 = -ri2 or by p = ni.132 ELECTROMAGNETIC THEORY. There is so metimes. Another point which may not be always understood is the assumption involved that both operand and resultant are simply periodic. some confusion or indistinctn ess of ideas on this point. is no resistance. as a cases. VI. differentiator. given that there given. of course. but. .. the differentiator d/d(nt) It is a specialised differentiator. because there is resistance. that is (B . That is. But if it be It d oes practically. treated like the imaginary because p* = . p/n. But it by no means . Only . is indifferent. CH.

starts when t 0. and & C is. in a long time. by taking e = eQ smnt. (15) the solution if would never reach G is simply periodic. lead to the but if e = ~LpC is rigidly true. p signifies integration from easily show the complete establishment of the simply periodic state from (5). e state (15) approximately . of course. But is. the reciprocal of to t. Lp which is Lin (16) quite different. Then the powers o f p~ l are We may successive integrations . the resulting would. but not assuming C to be simply periodic. if e = eQ sin nt. . Here.e. That given a little resistance. as well as this state without resistance.

and the result is (-'* )-()'(!?-* )* .from to t .

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. final periodic state. or verifiable by differentiation. used in (14) is fully valid without escape may be seen thus Operate on (13). and then add on a term to cancel it . 133 integrals of sin nt. inde pendent of R however.~n\ Here the second part cancels the first initially. By collection of terms. (17) may be put in the form H p_ Rsinn-Lncosn ~TF+LV~~ ^ 'B + L 7/ Lnc" 2 2 . whi ch is general. though it But ultimately vanis hes and leaves the first in full play. when R initially and subside at the proper rate. by R Lp. which is the beginning part As regards the establishment of of (17). but have the state (16) in stead. is zero we never arrive at the simply periodic state. making : (R . The above way oi successive integration shows the internal working. where the functions of t in the brackets are the successive which are obtainable at sight. That p specialises itself to ni when there is simple periodicity That the transformation of e and C rests upo n p l = -n~. it is quite sufficient to first write down the (18).

Now by periodic. (13). we can generally bring it usual form of a differential equation. say if we to the A* = BC. we make the power of p n be conand . (19) which is also general.1ip)e = (R 2 . specialise. and C stant B are operators involving only direct operations Then. where that (21) A is. by specialising integral positive powers of p. not necessarily true. In the case of a complicated operational solution. We by letting C be simply though the converse is now have _//2 C= -n?C. e also. to be simply periodic. and therefore. and (19) (20) becomes which is (14) slightly changed in form. choose to clear it of fractions.LyjC.

and be p x constant when odd. So we reduce (21) to the form (22) .is when n even.

The resistance operators are additive. Coil (2). In the above cases of a coil or of a condenser. as regards their reduction to the s tandard form. then by B x and it as well. as he can use electrical ideas throughout rather than purely mathematical. multiply by A x . the structure of the formulae is plainer or p = ni.2 p instead.B. and Condenser in Sequence. A similar treatment applies in the case of irrational operators. 285. CH.A 2 _p. by either p*= O f course. . If Finally. This work is just in terms of e. we may p ut the is but one degree of freedom. n* stretched-out form of a differential equation.134 ELECTROMAGNETIC THEORY. VIr where the A's and B's are constants. solution may be much better done in the operational without preliminary conversion to the usual itself. and simpler in the operational form. Also in Parallel. so l_c is (23> the equation of voltage. if want e in terms of C. This is a very important matter. there To make two. coil and condenser in sequence. especially to the electricia n. C .

(24> The impulsive voltage. on the coil.arrangement. and the steady C(R + K~ 1 ) are clear enough. and steady after see that the corresponding e consists of t hree parts. We e being impressed on the complete need say nothing here about the simply By inspection of (23) we see that the operator periodic case. as th e expansion theorem applies to one part of it. It is merely operative (24). an impulsive voltage of total LC togenerate the current inst antly a steady voltage EC to keep it to the other. though not the initial If C is zero before. impulse does nothing to the condenser. Then. we . The case K = more simply by is interesting. moment. and the variable voltage (l-*K< /3 ). The exponen tial term is required to keep the total The initial current constant in the cond enser and shunt. up against B . by or is . is peculiar. namely.

(23). .. the voltage on the condenser increasing uniformly with the time.much Ctf/S.

135 Only one root into the play. then two roots come into Thus. and next So.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. (27) C is c= a where. equation that (1). to be Z. But if we inquire a steady e. instead of (K + Sp)~ . we see first that and applying the expansion Z is R + K"1 . by C= Taking the denominator theorem. Z= it . of if pl and _p 2 (K + JSpf denote the two roots of Z = 0. is C produced by (23). the expansion clp pf^-Lp _S^_. concerned in the above.

that the terms involving the time depend. if C = (Y/Z)e. In (14). with the specialised meaning of p. in detail would need a chapter to itself. /o + . and next as regards the terms variable with the time. it is on the infinities of Y/Z (1). we see that a common factor is also introduced. written (R + Lp) 2 because Into a discussion of (28) in detail we have .makes them equal. and the su bsidence constants of the normal systems depend upon the vanishing of C= ?. Another way is to convert (26) to first This will give the same result by the expansion theorem. the object here being merely the The theory of circuits application of the expansion theorem. then two values in succession which make the denomiThat is. the algebrisation was effected by in troducing a common factor into the numerator and denominator. though .2^J rfz dp (30) expresses the expansion theorem. In the numerator. Z. Comparing (29) wit h (26). being equivalent to equation Putting it in anot her form. as regards the steady s tate. p has to receive the nator vanish. is unnecessary to enter.

&c. that it did. we multiply numerator and denominator in (2 6) by r + lp. concerned in Z. nator. there is no loss of generality by the simplifica tion of Z mentioned. there are m degrees of freedom. it may be. roots. under the circumstances. however. say. This means that the combination behaves. by introducing special relations amongst th e resistances. even down to one root. thus introducing a fresh root p= -r/l to the modified Z. and m terms in the C expansion.. which may be referred to in passi ng. the expansion of C = ejTi goes.136 ELECTROMAGNETIC THEORY. and is. Now. besides p ossibly an outside steady term. The same applies to simply periodic solutions. It make no difference in theorem. VI. towards the impressed force. Suppose. The new term is will the working of the expansion of size zero. CH. inductances. Let C = */Z be the operational solution of an electromagnetic problem. we may be able to reduce it to a simpler form. involving a comIf Z = has m plicated combination of coils and condensers. . involving a smaller number of roots. The reduced form used in the expansion theorem will give th e proper result. therefore. for instance. Suppose. It does not alter the roots of the denomiunnecessarily. automatically excluded. There is an important principle involved here. So far as just as though it were some simpler combination.

viz. from an We = 0. for But then it is done in such a way that th e larger bination. those (if any) of to a larger combel ongs really enlarging the problem. Thus. is equivalent. to be able to eliminate simply the unessentia l and inoperative parts of the work. we are roots of the denominator.. C = e/Z also see. with unit operand. or as regards free subsidence It is very convenien t arbitrarily given initial state. The extra terms i n the expansion of C are of zero size. and intro duce thereby a number of new Y YZ combination e and C. as regards a differently situated impressed force for instance. by reversing the reasoning. . the combination does possess a greater generality than the simplif ied equivalent in other respects. and the Z operators allow us to do this. as regards the relation between to the smaller.Nevertheless. that when we alter to C = Y0/YZ.

to the current is steady after the initial moment. the vani shing numerators make the terms be zero . and therefore Sp) The conducC = {(K + is + (B + L^)-1 }* (32) the equation of current. Returning to the co ndenser and coil. To have e steady after the initial moment. a steady current Ke. "We therefore get the original form of equation (1). additive. with the of Y = only. roots and subject them to the tance operators are now same impressed force. but this does not happen with those of Z = 0.gradually rising current by what has preceded. Now. 137 where the accent means differentiation to p. On the other hand.PURE DIFFUSION OP ELECTRIC DISPLACEMENT. and a . we require an impulsive current of to tal S0. Only one time function if is concerned. put them in parallel. we have expand = .

con286. c^ (34) p l} p2 (3). . of two degrees of freedom.Comparing with (26). -we see that the result in the present case c The roots . in (28). sider As another example two linear conductive circuits. c^< . . Coils Two under Mutual Influence. n -' and is its expansion. are as before. The second circuital law applied to them makes E = (R + Lp) C + Mpc.

except in the case of M. circuit. which is co mmon to both." of course. (36) small for the other.e (35) = (r + lp)c + MpC. This is "ironless mathematics. the mutual inductance. Young using big letters for investigators need not be discouraged by the contempt which is too often poured upon "ironless mathematics" in one .

we require the voltages indicated explicitly. making e = MpG. Sometimes the contemne rs only indicate their own ignorance. the equation and behaviour of the first circuit are the sa me as if the second were taken away.138 ELECTROMAGNETIC THEORY. we may The equation elimi nate C or c in turn between (35) and (36). the steady EC in one and re in the other circuit. of C is C= which.. the forces are given. (37) That is. Thus. on the other hand. and very useful.M /' 2 (1). The principle of the destruction of mutual induction by a suitable impressed force is a useful one in theoretical reasoning. c If c is to be kept zero. by or (30) . When. CH. put E = (B + Lp)C. It is fundamental. we should solve for the currents. At the same time. by cross multiplication. engineering journals. If both C and c are to suddenly become steady. following t he initial impulses LG + Me and Ic + MC respectively. viz. = 0. we require the voltage MpC in the second circuit to prevent current in it. VI. .

are sufficient to show how the expansion th eorem is to be applied. The above examples. expands t to E + E L^(r + . They are not particularly good examples. going on ly as far as two degrees of freedom. Cable Earthed at A and B.2M Pl ) + lPl (E + lPl ) PaV + ditto with Here P) and p2 are the roots of the denominator in (38). 287. Passing to the consideration of the application of the expansion theorem to diffusion problems. (4). as well as on e of the most .more conveniently. and when it does not. and see when the expansion theorem applies. take first what is one of the simplest cases. Impressed Force at A. but are chosen rather to show how to readily discriminate different operators.

Being operational solut ion was described on the same page. the final steady state is e(l So the . we can see that it is right by noting that it makes V = at x /. we may use either of the forms given in (40). namely _ string siusl The practical way of getting this where <f= -s 2 = KS/>. and a = steady impressed force e be inserte d at # 0. x/l). 256. The potential produced is given by (17). But gives the result in terms of the values of p. because to are those of shin ql. these are negative. 139 Let a cable be earthed at both ends. wh ich is an odd function of ql (and therefore This may be formally done by of j)*) . Take Y toThe roots of be the numerator. making Z be Noting this. therefore the second form is more confirst The venient. rationalise shiuql.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. an easy case. putting d /Z = kl cos si. we must exclude it. . and Z the denominator . ds (11> p = 0. So we have p ^L = i/ dp Also. for by omitting the zero root from shin ql = Owe come to the same result. and V = e at x = 0. interesting. But not the zero root. Take the form (30) of the expansion theorem. (shin ql)/ql but it is unnecessary.

it is really the values of p that control matters. . square of s. we see that only the positive roots of sin si = 0need be counted. .x\)+ ^ sms(l-x}t T I J ^sl cos si where the summation ranges over the finite roots of sin si = = 7r. / r which wa. STT.expansion theorem (30) gives > /. 27T. &c. \ pt 1.. and^> in the exponential function is.3 obtained before. .. Equation (42) may be at once reduced to (43> '. sl 2 is given by RS/j= -s Owing to the de pendence of p on the that . T hat is.

and the expansion theorem (30) makes V=e at V = . as a function of p. cossl . CH. Impressed Force at A. the potential is giv en by at with the sole x = I. Now in this case. y _ cosh q(l cosh which. Then. (5). x) e _ cos s(l cos si x) e ^ gl we see to be correct by the formula making # = 0. the denominator is rational. Cable Earthed at 288. VI. last Next take the same case as change made of insulating instead of earthing by (22). A and Cut at B. and C = at x = l. + e2 coss t^. so no res ervation is needed. that is to say. 258.140 ELECTROMAGNETIC THEORY. as before.

and its specialised algeThe latter only by the same symbol. and also of the p's. occurs in algebraical summations solutions. (46) which of the The summational part is in (43). same form as 2> fusion between braical value denoted the differentiator. . we reduce (45) to V = 0-2?2 si--e p si '.si (45) l %dt ranging over the positive roots of cos &Q. = . Impressed Force at . Cut at B. 1 JTT. and simplifying the numerator by the vanishing of cossZ. the former in operational (6). There is no conis the expansion required. or. si = JTT. Earth at A. But of course there is a different series of va lues of si. Effecting the differentiation in the denominator.

Say that there is earth on at x = Q. . and the terminal conditions. though still with the simple when cases being sufficiently illustrative force is situated terminally. and an impressed force e at x = y.y. Then. 289. The above two the impressed terminal conditions are of the simplest type. take next the case of an intermediate impressed f orce. making either the po tential or the current vanish. insulation at x = I.

putting s = 0. 141 -s 260. (42). putting for convenience in the expansions to b e found. First. So applying we obtain V = e . are given by by (41). ^= a cos si V = ^lsin S (Z-7/). we see that the final steady state is (30).*. the resulting potentials. Va = 0.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. 2 Here. of course. s is the differenRSjj. cos si 2 t (48) V when tiator Vj being the potential when x is on the right side of y and 2 on the left side.e^ x COS * V cos s(l . Vi = .

e* (51) V2 = -2*Zsina?. may now simplify by means of cos si = 0. The transition from the operational solutio ns (47).. and the addition of the outside term. (52). (49) V= 2 . by the omission of terms which vanish in the preceding equations. (50) involves two formal processes. where is a positive integer. 4*. t he change in the denominator.y) <. leaving We Vi = e . In going further. done by the differentiation which occurs when the summational sign and the time factor are introduced. (50) n = subject to cossZ 0. (52) directly . we reach (51).*2 -frslsiusl sin S (Z .. The same occurs Now this omission may be done in the previous examples. . as in the last case. in the act of applying t he expansion theorem. (48) tothe expansions (49). viz. thereby deriving (51).2*2 C SSy sin sx Si . or sl = (n J)TT.* M (52) the simplest form of the expansions.x) .

from (47). in through the expansion theorem. wemay make the omission in the operational soluti on. at (51). (52) is concerned. so far as arriving cated intermediate equations. But the . (48) without the more complifact. And.

first CH.142 ELECTROMAGNETIC THEORY. (7). steady part of the solution must be extracted case. If both terminals are earthed. smsl . (53) = --. and in any made nonsense of (considered as a general operational solution) by the too early omission. VI. at the point y. we have. Earth at A and B. whilst the impressed force is intermediate.a?). by (37). . Impressed Force at y. (38). 290. Vsin 2 soc /1 \ / 1* A . So the omission is properly done in the act of conversion to the expanded series. the operational solution is . 260. the results Vx =+???? sin s(Z.

which The determinantal equation depends only upon the nature of the terminal conditio ns apart from impressed force. (56) . But make it be a function of y. there will always be the same set of normal systems. for example. Vl = e l I/ . In (55).2.*> (54) on the right and left sides of the impressed force So the expansion theorem makes respectively. in virtue of sins = 0. ^s(d/ds). the outside terms got by putting and then write the summational changing the denominator by the operator is equivalent to p(d/dp). and omitting terms which. There may be any distribution of impressed force. general term. (55) (56) where we s first write down = in the previous equations. rep resenting a distributed Then the integration of (55). so as to include all the i . (56) with respect to y. e is entirely at the point y. vanish from the result with changed denomina tor.\ -COS *(*-#). but unless we change the terminal arrangements. intermediate 01 at the terminals as well.2 Si *.

will The outside terms are altered. .mpressed forces. impressed force. and likewise the size of the normal systems in the summation. give the resulting potentia l.

we have (58) by V = ZjC ment have Z Here Z x is such that 261. terminal reflection coefficients are a and # = 0. (8). and the same series of s's and p's remain in force. in which put a = 0. Let. at x due to e at ft. General Terminal Conditions. 291.PURE DIFFUSION OF ELECTRIC DISPLACE JIENT. Then we only want the formula on the right side of the f orce. Impressed Force at A. say sins^e'''. 259. is not changed. expansion is. and . x = 0. equations (49) and (52). the expanded solutions are It is somewhat differea dily calculable numerically at once. when the at x = Q and I In terms of the terminal resistance operators Z and Zj. which arises from the fact that the terminal reflection coe fficients are of size + 1 or-1. for a first case. and require preliminary determination from trigonometrical tables before The process of the formulae can be subjected to calculation. but the type of the normal systems. the impressed force be a t the beginning. as the roots become unequally spaced. is the equation of voltage f or the terminal arrangeat x = I. however. the same. making which shows the potential respectively. Owing to the equidistant spacing of the roots of the determinantal equation in the above cases. This is (35). rent when we depart from the simplicity of terminal earth or insulation. as the case may be.

although it will come to the same thing in the end if we is infinite. When the terminal arrangement is a mere resistance. Going farther. by means of the given expre ssions for a and ft. at every . to convert it to circular form first. in passing. before applying the e xpansion theorem. the singular case a = l. when Z which needs an infinite terminal resistance and no terminal permittance. This happens V .V= -ZC for that at x = 0. By (57) it makes = that is. Notice. the cable canno t be charged at all. . moment. The best way is. let us expand (57). and Z 1 = R 1 these b eing constants. we = R say.

no further reductions of the kind just made are required for them in any case allowing of t he application of the expansion theorem. y= (1 ___ -Z Z lS {sin + (Z 1 . put q = si in (57) and (58). and then change to circular funcAdopting the former co urse. The result is V= Now if the and . and reduce by ordinary algebra. and therefore of . that although Z and Z x are uns pecified. Observe.. by its construction.144 ELECTROMAGNETIC THEORY. also. CH. only another form of the operational solution itse lf. So (59) is the operational solution in a form which is convenient for the imme diate application of the expansion theorem. The result i s tions.. because they are then rational functio ns of p. expand (57) as it stands. and si or sx for sin si and sin sx. the meaning ofs 2 being -ESp.s/R)cos}s(?-^) ' 2 /R 2 ) sin si + (Z + ZiXs/R) cos si g> This is. s 2 Put s = to find the steady state. Practically. put 1 for cos si and cos sx. VI.

and in steady resistances. (60) terminal arrangements are mere resistances. have finite produces them. Thus. then s = (61) is still the result. (60) may be written V= . But it may happen that the term contain ing s 2 in (60) does not vanish. It may be infinite. where Z and Z^ R and R a denote the effective steady resistances. this becomes which is obviously right.__ Bg-oQ + fr .. say R Bj. if the terminal arrangements. though not simp le resistances themselves. and the numerator is that part of it which lies to th e right of the point x. Similarly. because the denominator is the total resistance of the circuit. so caution is needed when t he effective terminal resistance is infinite.

The p is cancelled. is Now suppose the terminal arrangement at a condenser. making Z = (Sop)"1 . But A .in terms of p.

So1 .say. provided the resist and BO are infinite.1 Z =(S . E beinoSo if R! is finite. the state ultimately when charged through a e resistance at the further end. the resul t is V = 0. making Z = (S^). On the . ment is (64) condenser. cable settles down to a neutral say (62) Z 1 = (S I /?). at the 145 infinite. the result is V ance at A is finite.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. the full value. same time the denominator is made infinite. if both R! s at both ends. may be verified by the law of displacement in conOn one side of the impressed fo rce is the elastance and on the other side (Si + SZ)-1 So the total displace. . Lastly. when the condenser 1 is at B = e. the result of putting = in is This result densers. 1 ^). That is. by having condenser . provided there is not infinit other hand.

the outside term to express t he steady state * when there is one. infinite. the final steady potential of the cable and further condenser. being lower than the above V b y the amount e. If we compare these cases with the formula (61). or we must go back to the more complete formula and interpret it instead. on the side of the condenser at A w hich is next the cable the potential is negative. and the result is V as in (63). we see that they are in agre ement. use of Maxwell's disp lacement current. the result (61) is ambiguous. denser at reverse is A next the earth that is charged positively. The the case in the dielectric of the cable and the B conThe simplest mental realisation is obtained by the denser.* Having thus settled VQ. In such a (60). except that when R and R! are both case (62). On th e other side of the impressed force that is.Divide by SI + S lf the permittance on the right side of e. understanding that R and Rj are the steady terminal resistances. It is the side of the con. we may apply .

but in any case = in (59) will lead to the required result. L .the expansion Of course condensers may occur terminally in other ways than the s above. VOL. II.

--&( iU when Zi = 0. VI. any form of the same that Another form is . and to (G8) tansZ=. which is the same. tan^-J^f^. and then dividing the numerator a nd denominator in that equation by cos si. ("0 4)^i( 5 (67) This simplifies to tanl. (59).isZ. or to CH. z . may be more 1) _ cos sx (tan si + Z^/R) + sin sx (tans? ZjS/R obtained from (59) by expanding the numerator. theorem to convenient. JL\tt (69) when Z = 0. The determinantal equation is or.146 ELECTROMAGNETIC THEORY.

The values si or both. whi ch includes the extreme cases of terminal earth and insulation. It is So the specification of the initial st ate wasted and done for. but that does not count at all. as the case are then very easily determined by means of a table of tangents. we have the simple terminal conditions considered by Fourier. by the use of the property of the vanishing of the integral of the product of any two different normal systems. producing a constant ratio of the potential and current at the term inals. Th e reason is because there is no energy in the terminal arrangements that can by itself affect the state of the cable.In these equations the Z's are functions of being mere in virtue of being functions of p in general. Z or Z of 15 may be. merely requires a statement of the . are constants. (sl) In the case of the terminal a rrangements resistances. is the only one that admits of the expansion of an arbitrary function after the manner of Fourier that is to say. This kind of terminal condition. It is true that there is waste of energy in the terminal resistanc es.potential in the ca ble .

not by mere waste.. so the complete normal systems must include the termi nal arrangements as Now if the terminal arrangements consist well as the cable. only of condensers and resistances (or equivalently). But should we introduce magnetic energy into either or both of the te rminal arrangements.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. itself. the energy in the cable itself will set up energy in the terminal arrangements manner. this terminal energy must be allowed for by widening the conjugate property. . that implies electric energy. we only introduce electric energy. is like property of the vanishing of the total mutual energy of a pair of complete norma l systems. Fourier's method fails. But even if it does not. upon which the later state of the cable when left to itself depends. with a wider range. in the absence of an impressed force to introduce fresh energy. however. This that in the cable (its self-induction being ignored here). The conjugate proviz. involving merely a waste of energy fr om the cable. or in other ways. the energy of e lectric displacement. If it does. The initial state may or may not include any energy in the terminal arrangements. The integral of the product of two normal systems along the cable is no longer zero. however in a reversible complicated in detail. perty which has to take the place of Fourier's is then sub stanIt is the tially the same. But whenever the terminal arran gement departs from the above simple type.

or of the Thus. following the formula (30). and of the k ind of conjugate property required to effect the expansion after the manner of F ourier extended. But the expansion theorem goes straight to the final si mplified result. of a pair of normal systems equals t he mutual magnetic We energy of the same. the terminal arrangements being included. applying it to (65). we obtain V=V + cos sx (70) L2 .. that the m utual potential energy (or electric. irrespective of the absence or presence of energy power of receiving and storing energy in the terminal arrangements.the last property breaks down. This may work out in a complicated manner. and using is (dlds) instead of p(dldp). viz. have to go to a still wider one. here).

Remember in doing so that this is not really a complex . we see that its vanishing does not make the denominator vanish. if we please. (36).148 ELECTROMAGNETIC THEORY. may substitute the right member of (67) for tansZ in (70). we (66). its roots are out of the General Case of an Intermediate Impressed Force. and therefore question. by performing one differentiation upon the denominator in We cannot carry it fur ther profitably without specifying (65). fun ctions by putting q = si. Since. unless t he Z's are given. When the impressed force is point x = y. keeping it in the denominator. and 2 But there is no par similarly put sec s in terms of the Z's. and we desire to make possible simplifications of expression. Or. however. to the numerator of (65). no notice was 2 taken of the factor (1 ZoZ^/R' ) in the denominator in (65). (9). 292. In obtaining (70) by the expansion theorem. We Now circularise the 259. ticular advantage in doing so. VI. but they are in operative. in which put y for a. si in the summation are fixed by the equation by the vanishing of the denominato r in (65). (35). CH. what the natures of Z and Z x really ar e. the treatment detail need be given. It has its roots truly. the values of that is. They lead to The factor may therefore be regarded as belonging nothin g.

because the operators 2 This are rational functions of p alread y. To x on the right obtain the corresponding x is on the left side of y.s(7 ' a?) (R .ZoZjS 2 tan ) Jf e + B *< z o+Zi) expresses the potential due to side of y. is intermediate. and therefore of s fact makes the transformation easy. and observe that 2 is got from Vj . so that very little have the operatio nal solutions that occurs. 2 (Z^cos + si Rsin). at a point where e is . refer again 2 259. by the cancelling of t erms . situated. say at the quite similar.transformation. The result is that y _ (Rcoscos si Zossin)gy.

we get tiving the result. V when V y= _ (Rcos Z^sinX^ y) (Zpscos + BskQac same denominator . doing the sa me to (71).y. (36).by interchanging a and /?. and then negaSo. expression for to (35).y and I . . x and l-x.

(& cos . and we must go further. finite. is are to be regarded as we the effective steady permittances of the combinations that S and Si in the place referred to. denoting the steady states by t\ and v2 and applying the expansion theorem to (71). which it is 291. The outside terms for the steady state may be p erhaps most easily got by elementary considerations by Ohm's law and the condens er law. should t he terminal arrangements include several condensers.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. may get the steady states directly out of (71) and (72) by However the work seeking their limiting values when s = 0. In general. 149 Now apply the expansion theorem. But when R there are terminal condensers concerned. that is When the steady terminal resistances (effective) are to say. however. we obtai n at once . and apply the condenser law. a special case of the present. say R and R we 1} shall have ( by Ohm's law. y= c t< y. equation (64). be done. say as in and R! are both infinite. Or.

And when the Z's are explicitly given. we may use the alternativ e special values of si. . formula if we like.~Z s sin) sy . to be got from (74) is by changing Vj to V 2 and the same. and changing the denominator by the operation ^sl(d/d(sl)). in virtue of the expansional part Of course. which finds the steady states on the two sides of e. by inspection of the original 259. As regards V 2 we may get its formula by making the interBut changes already indicated. identical for all the values of p given by equations. (Z: s cos + R sin) s (I . So the V formula is The ^ to 2 r. in getting (72) from (71). of two . because. That is. the time factor. we may see that when operational formula e (35). For.x) by introducing the summation sign. making the interchanges before descr ibed. the denominator vanishes V x and Y a become identical in those this is quite unnecessary. This is not a matter of indifference. (36). the determinantal equation.2 . simplifications of form may be readily carried out. if numerical c alculation is wanted. though not in the case of^? = 0.

which is the most impo rtant in general. and a the revised value to the tables. VT. may show that the roiun value is several degrees Estimate the correction by rule of three. it is the first root. tangent tables readily find the roots. the curves (76)* . Saygraphically-obtain ed values as a first approximation. draw them very roughly.150 ELECTROMAGNETIC THEORY. Only a few general at the beginning are wanted. one the other. just to find the situation and rough values of the roots . The determinantal equation is (66) or (67). Apply It will be nearly right. Here <(sZ) is a rational function of si.equivalent forms. OH. yz Draw = <j>(sl)'. except for very small values o f the time. 293. I mentioned before that when the Z's are mere It is sufficient to Use theresistances. 2/i^tan**. The . (75). Say. may be much more manageable than The Determinantal Equation. The interthe abscissae being si. and the ordinates yl or yz section of these cur ves indicates the admissible values of si.

and the coils may be The trouble with complex roots se ts treated as resistances. second revision a second. Of course. nevertheless. so that the above described process can be followed. may bring it right to a minute. and a third to Another easy case is that of terminal condensers. And it may be remarked here that when the terminal arrangements involve only ele ctric energy. the inclusion of the effect of the self-induction of the coils is not an important matter. in slow the inductance be of a suitable value. This a single coil is concerned at eith er terminal. without magnetic. have an interest of a scientif ic kind. as showing what.tables wrong. when cable working. Also the case of condensers an d resistances does not trouble much. occurs even if There may then be complex roots as well. But it is not so easy when magnetic energy is involved. the roots are always real. may be done if the necessary trouble be taken. a limit to the desirability of elaborating formulae beyond the point of ready sy stematic calculation. The formulae. .

+ tanK 1 1 ^ h> / 1 = 0. (66). Equation (78) may be r eadily converted to a power series. Special formulae to any whether they repay the labour of obtaining them. depends upon the terminal operators. perhaps with occasional advantage. . 259. (77) which makes ntr = si + tan^ HI + tan" ^.PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. and (65). when tables of circular fun ctions are so useful. rU (78) This form shows how the difference between si and the standard value mr. as it shows itself in the ope rational solutions (35). The original form of the determinantal equation. (36). General principles may become smothered by overmuch detail . n being a positive integer. 151 The dcterminantal equation may be written tan si + tan ( tan\ 1 ^.

(79) If the left-hand operator be applied to a wave originated at a point. apparently done on a normal solution. and we get where the operator rational. and reflection there according to a. l written or ae. according to P.r = 0. followed by a journey to x=0. refl ection there. say at . But i t is original state. of course. but it is questionable 253. any use of p*. chosen to be any one of the values given This equation does not imply by the det erminantal equation.may also be constructed exhibiting the roots explicitly desired degree of accura cy. The right-hand member would show that the effect of the cycle of opera tions was to restore the disturbance to its This is impossible. maybe & = ap. Thus (80) apc-* c if l vt = ltt . Cl ear the constant is it P of fractions./?z l = l. . but only abbreviates a rational equation. the resultant would represent what it became after a journey to x = l.

is an even function of q. and is therefore .

unsupported by impressed force. -2M *.2 as the case may be. the state of potential v is expanded in the special Fourier series concerned . CH. the state at time t later will be given by . if we write (74) and its compani on thus. 293a. For. v is left to itself. If. (81) to be ^ or v. Subsidence of Special Initial States. we have. V= where v has initially. but also the connected question of the subsidence of that state to equilibrium when the force is removed. then. In the above equation (74) we have not only solved the question of the est ablishment of a certain state by an impressed force.152 ELECTROMAGNETIC THEORY. VI. That is. (82) v = ^u.

electrically up the v in the cable . well. many expansions expressing the same functions tions. and subject to the same differential condiThis is consider what For quite clear in the physics of the matter. except when only mere Now the same state of po tential resistances are concerned. where v There is (83) is what v then becomes. for it may be only one of v.i-ZtU*. the impressed force does in the ac t of setting It energises the terminal arrangements as and magnetically. this remark to be made about (82). It cannot be called without limitation the expansion of the function v in a Fourier series. s ubject to the terminal differential conditions. state v.

the Z's a re only given by formula!. cable. but also let the te rminal energy act on the and in various manners according to circumstances. Equation (83) exhibits the po tential at time t due to the initial state u in the cable. and to the accompanyi ng initial This is true even when states of the terminal arrangements. and what the terminal energies are.may be accompanied by various states of energy terminally. and we do not analyse them to see what particular arr angements are really represented thereby. and all due to the same impressed force. So when we take off the imp ressed force we not only let the charge redistribute itself. .

apply the expansion theorem to arbitrary initial states. the limits of an arbitrary function of x between naturally rests upon the expression of the That is. it may be noted here that an integration applied to (74) and its companion ena bles us to express the effect due to an arbitrary distribution of impressed force along all starting the line. Now it is just in these cases that the examine the expansion theorem shows to best advantage. In connection therewith .PURE DIFFUSION OF ELECTRIC DISPLACEMENT. But this is a mathematical development which is useless for our pres ent purpose. 153 It will be seen that the expansion theorem is a labour-saving agent of a remarka ble character. we require to find how function for a single point. For it is not from formulae representing the expansion of an arbi trary initial state that we can most readily learn the general course of events in the We should rather prefer to physical problems concerned. Then to I. Put edy for e. We obtain our formulas in a very read y manner. using one or other integrate with respect to y fr om formula. . General Case of an Arbitrary Initial State in the Cable. result of some special initial state. at the same moment in the simplest be a function of y. as an impulsive or a continued source. case. and let e (10). or the result of a disturbance initiated a t a single spot. according as x is to right or left of the elementary impressed force concerned. without the circumbendibus connected with arbitrary initial states and the conjugate proWe shall see presen tly how to perty of the normal functions.

263. the construction of the opera tional solution for a point charge is like difference. in which h is the current introduced. by . to be given as a function of the time . The operational solutions to rig ht and left of the source are given by (65). at the place of application. Taking h=pQ makes the source impulsive. A continued e at a point has its analogue in a continued source of current. diffuses itself when controlled by giv en terminal conditions. initially all at the point r. with a latter produces a jump in the po tential. led in by an auxiliary wire. (66). and the case is then that of an initia l charge Q.294. The that for an impulsive impressed voltage. As was before pointed out. So. The expansion and I a charge Q. the impressed force former in the current.

where 263. on the right side of the point Q is impulsively intr oduced. terminal . Converting (84) to circular functions. the potential Vj at x.154 ELECTROMAGNETIC THEORY. is given by On the left side of y the potential V 2 is to be got by simply interchanging x a nd y in the last formula.s(Z . VI. It must be 1 there is conductive connection with earth at either is. (65). by means of q = makes the equivalent form ula si. y The zero if _ _ (Z S cos + R sin)sy . y.x) sQ final state must be either zero or constant. CH. (Zjs cos + E sin).

the last distributed along the cable. . B! are the effective permittances of the terminal This result may be dug out of (85). the potential of the cable is . useful test. or is divided between the cable and the terminal condensers. it is the permittance occurs as a factor. so that In the former case. that controls matters. (86) if S and combinations. but is sufficiently clear without doing that.infinite . equation shows the result at once. on account of the s w hich In the latter case. The charge Q is finally either uniformly the charge are infinite. if there is no terminal permittance. that when at least one of E and E is not and it must be finite when both Q cannot be got rid of. which (85). Thus.

We have This formula the same form of determinantal equation. a Applying the expansion theorem to we obtain pt -x) 2Q coss . both as regards the For the steady state is consteady v and the summation. (66).. however. where <f>(sl) is as in (75) and <f>'(sl) is its derivative. 263. . become the same when t he denominator vanishes. stant. applies on both sides of the point y.is. and t he original operational solutions (65). We may.

.. The ultimate resul t is a state of uniform potential viz. for a great deal more than t hat. then. Given. outside the limits there are periodic repetitions. the total initial charge divided by the total permittance. including the effective terminal permittances. when we have the usual Fourier series with regularly spaced roots. and make other changes controlled by the dete rminantal equation. a function /(a?) representing an initial st ate.*} becomes at time t. though not periodic in t he usual sense. by introducing thetime factor in the s ummation. Then + 2" where u is v0 . except in the elementary cases of total reflection sive function of at the terminals. As regards the meani ng of the expansion for V outside the limits for <e. its expansion is using the previous expression for u\ and.lUNr PURE DIFFUSION OF ELECTRIC DISPLACEMENT. x) /QQ\ initial state. exchange x and y in (87). Let u be the result. Here it represents the the value of v with Q/S = 1 Therefore u is the expansion of the unit impulx concentrated at the point y. j2 ' . viz. . It represents and I. Put Q/S = 1 and t = 0. we find what the potential /(. 155 however.

free 2 and I.. its t in the cable.problem. when the ter minal reflecV = 0. we equation. is from rf] impressed general characteristic (A RS the space differentiator and <f is x the time differentiator. we make V . which are imposed in the physical some knowledge from the determinantal g et If V is the potential at x at time force. Also. where A is tion coefficients a and $ are introduced.

we have. It is to is V we write A for q. Compare with So. 276. V is f(x) t . or A /BS substituted for^. by Taylor's theorem.satisfy where in the operator on (80A). if be understood that in the two Z's the same2 change made.

If. The expansion (89). and which only come int o operation when t is finite. because it is the one that depen ds only upon the initial state of the cable. therefore. may not be the It will still proper one to use in the physical problem. has some right to be termed the expans ion of /(a?) in the Fourier series in question. resistances. specification of the initial state must include not merely th e of independent ways ments may be energised. be finite or infinite. would lead us too far {11). but will fail later. the Z's are mere initially unenergised. the initial state. CH. showing the primary kind of periodicity. When the Z's are zero. subject to (88). away from the proper subject. independently of the above expansion. for example . the Fourier series (89). it reduces tof(x + 1) =f(x . The number of such auxiliaries is determined by th e which the terminal arrangemay. should the represent terminal arrangements be initially energised. Further examination of (91) to see how this is modified by the terminal operators. others which represent zero initially.I). So there are. there is a coil at A and a condenser For the full at B. therefore. Auxiliary Expansions due to the Terminal Energy.156 ELECTROMAGNETIC THEORY. Case of a Condenser 295. there are just two aux iliary expansions. It gives the potential at any later time on the understanding that the terminal arrangements are Unless. in It . (8 8). VI. This is the equation showing the functional connection between the values at poi nts distant 2Z from one another.

A case in which there an infinite number of auxiliaries is got by letting the terminal arrangement at either end be a second cable. Initially. of finite length. will introduce an infinite number of auxi liaries.number charge in the cable. however. would introduce one auxiliary. but also the charge of the condenser and the current in the ceil. The metal. state of they are all connected. but can be trary. After the first moment. Another one is got by putting Without the metal the coil a piece of metal inside a coil. because the initial state will require a specification of the magnetic force in parts of the metal. is they are independent. all . and this initial state may be arbiThis case may seem fearful ly complicated. which will produce a quit 3 d eterminate rational terminal operator.

_-_ cos sinW . by (85) above. a \ cos .JL Ss \ impulsive introduction of Q at y. this expands to Jsy sins(Z x) v-s^.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. Say that Z = (S p)~ and Zj = 0. the operational method. So. j (92} represents the potential at x on the right side of y. 157 without great diffi(in the case of a round core).-r.-arT-g^ 2L lslcossl(sec?sl+ \ b()S (93> C/ } . Then the cable is earthed direct at B. due to the And . let there be a l terminal condense r. culty. by To show the action of a single auxiliary. sins(Z x) cosdftand. and th rough a condenser of worked out permittance S at A.s sin .

Or. . with no initial charge in the condenser. (94) and an integration according to (89) constructs the potential due to an arbitrar ily given state in the cable. Initially V at time t in the cable due to V<> we have .3 beginning. The determinantal equation tansZ = S/SoS. its initial represents the potential alone. if V is So we may substitute it for Q/S potential. put y = in (92). Then V= 1 sim(i-ar) Q That is. Now as regards the condenser.which is valid on both sides of y. to an initial charge Q of the condenser. represents Vj in the cable due to Q at it. is There is no outside term.

This exhibits the auxiliary The complete expansion.0=V 2 -for same denominator =_ (97) any finite x in the cable. expansion. V being any constant. .

It is true that in (92) Q may be finite. inasmuch as the same operational solution serves for the elementary charge in the cable and for The distinction is that Q is finite in on e the terminal charge. VI. case and infinitesimal in the other. showing V due to the initial V l of the condenser and V =/(#) in the cable. CH. is therefore V+ I f(y) ( cos s sin) sij dy This case is peculiar.158 ELECTROMAGNETIC THEORY. but when we it tion of potential. .

the Q for the condenser is f inite. which is therefore infinitesimal. hand. these points being separate. the charge on the element dy. and so on up QU /S at yn . Points of Infinite On the other Condensation. Another consideration presents itself here. common Fourier It is a 296. Exceptions to Fourier's Theorem. at y QJ8 at } Go back 7/2 > to (88). That this is wrong may be seen by the above Let there be to initially investigation. QX /S . . error that only finite functions can be expanded in series.is pass to an initial finite distribureplaced by Q dy.

where fini te charges exist.Then v = 2. so that Qz In the limit we have a . v is zero except at the n points. Combine with (98). closer. except at the n points. V + v is initially finite and continuous.Qn un b (99) by (88). increasing Q in the same ratio. indeed. If f(x) is finite and conhowever. In this But we may respect the solution would be of a useless na ture. introduced at a point y would in an infinitely short suddenly time raise the potential of the whole ca ble infinitely. represent the initial state of the potential at due to the n point char ges. but there it is infinite. The simplest case is first to have two finite point charges Q and .Q at -will. combine the infinities so as to have finite results. Nor need these condensations be themselves finite. tinuo us. with finite space totals. time t distance s apart. The It is. true that an infinite charge Q's may be infinite. Initially. Then bring them is finite.

series. Nor does the matter end here. is error is Another common There are other uniquely determined by Fourier's theorem.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. and the extra functions. extra normal systems. as in Fouri er's theorem. There whilst a connected series ex presses the initial current. and such a series ha s to express the initial potential. involved. and a corresponding expansion of the potential in a F. If the self-induction of the line be included. or si = mr. the initial state must include a specification Fourier series are of the initial current. expansions. To show this (in one w ay). or R = Zjfo Z^s . But now 2= 2 2 R2 determines a distinct set of also. that the expansion of a function in a series of sines and cosines of mrx/l. let Z t = . n be ing integral. as well as of V. The result . of the type **. 159 point source involving a double infinity in Q.Z in (85). where h is constant. Then tan si = 0. Similarly as regards poles of higher multiplicity. So we can expand a gi ven function in terms of the usual sin sx and cos sx. may be points of initial infinite condensation in both series.

if one end be earthed and the other insulated. making 1 a special normal function. of potential of this type becomes reduced to sin sx t pt at time . Abnormal Fourier 297. vanishing termiLeft to itself. with equally-spaced roots. represent another function of x expanded in terms of the usual sines and cosines . ends are p both insulated. Series. Now. fixed . normal function t later. its When a cable is earthed at ends the type of the for the potential is sin sx. but with the same series of values of s. -s by RS/> = 2 If the zero value. therefore. the normal function is cos sx for the potential.is. The roots of the determiriantal equation are equally spaced. I will give one or two examples in illustration of this. to though it is equally true. a distribution nally . But it is not the expansion got by Fourier's theore m. the normal functions are . with the addition of a being negative. s being fixed by sin si = 0.

with roots of the deterrninantal equation are equally spaced in this case also.$l sin sx or cos sx. In all other cases whatsoever. as the case integral. n being The may be. = (nJ)TT. so far as I know. of real practical problems. the series of s's in the general 'no rmal .

acting at the terminals. the important and essential fact that there ar e extra normal We functions of an abnormal kind. to be produced by late nt impressed forces. toact in concert in such a way as to bring the determin antal equation back to the original simple type. VI. n being in tegral. differ from those given by sin si = or cos si = 0. of Fourier'3 And so we come to violations theorem regarded as a unique expansion. set going by the real electrical mechanism. resistances.J)TT. though applied toentirely unpractical arrangements. It does not seem to b e possible for any two arrangements of real condensers. with. does not uniquely determine expansions of functions. indeed. + a sin sx. or rather for similar reasons in other physical problems involving Fourier serie s. It is. They would be concerned with professedly rigorous proofs of Fourier's theorem. involving instability. On the other hand. perhaps. Origin of 298. for the abovestated reason. It seems natural. CH. a nd of the convergency.160 function cos sx ELECTROMAGNETIC THEORY. by physical ideas. said to do so. and coils. in which the s in the is mr/l. and are unequally spaced. . pure mathematicians would. that it has apparently normal functions escaped notice that Fourier's theorem. that mathematical phys icists should not come across exceptions. Guided. however. owing to the peculiar way they have of regar ding the subject of the expansion of functions. but that is ano ther story. it is quite easy to adjust matters so that the terminal ar rangement at one end of the cable shall combine with that at the other in such a way as to bring the roots of the determinantal equation back to the original simple kind. or (n. however. rather than with the discovery of exceptions. not arrive at them. come back to sl = mr. It is perhaps for this reason. where a is a constant. which are and necessary in order that real an d practical permissible terminal conditions may be satisfied.

(Z r s cos ' i . equation (85) shows the potential at x on the right side of //.Two Principal Abnormal Cases. Tims.x) sQ S -B' )Bin*/ + Es (Z + ZJ cos si . We may write it . due to Q i nitially at y. subject to the terminal conditions V= . . and V = ZiC at B.Z C at A.s y _ (Z cos + R sin) sy sl (Z Z 1 + R sin) s(l .

161 In general.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. So sins = 0.s-Z = (102) the determinantal equation. . save in the limiting cases of terminal earth o r . disconnection. with practical resistance operators. the denominator only vanishes b y the sum of the terms in it vanishing then we have the determinantal equation ( 75). giving sl = n~. (101) Then is (Z Z lS R 2 ) sin. But let Zo 2 + Z^O. n (103) being integral and also. 2 Zfo . with unequally spaced roots.

we get (104) R = R=RS/>. positive. 2 Thus. in case of a resistance 2 R at A. or (105) coil R S ^ = RS. again one root. where # = -s = RSp. with Z^Rb + L^. (106) There are now three roots. 2 In case of a condenser at A.2 = R2 . So there is an odd number of extra special normal funct ions. showing one root. we get R = (R + LoP) 2 RS^. Here. But two of them com . the left member is an odd function of p. positive. In case of a 2 at A. and three extra normal functions. when Z is a real resistance operator.

bine to make an oscillatory function of a mixed kind. physical interpretation will be considered a little later. in (109) of the cable were earthed addition. The If. or Z^-^L-. So cos sl = 0. VOL. we substitute Z Z r 2 = R2 s . II. just as one end and. or sl = (nJ)TT for the regulars. '/A (107) the determinantal equation becomes 0. . In general. (108) if . the number of extras is unlimited. instead of (101) as t he fundamental relation connecting the Z's.

and the other insulated Z 2 72 = R2 M .

Zr . and if Z is a permittance.Z x Equatio n (100) s y _ (R sin + Z cos) sy 2 . for the extras. This had better be done separately for the regulars . VI. a negative resistance. First General Case: 299. Z = . (Rsin Z scos)s( (Zfo Resins/ -x) sQ ""S" Apply the expansion theorem. if Z is a resistance. We Therefore reduces to first cannot consider both cases at once very well. with the extras as in the former case. a negative permittance. Zj is . CH. Z x is same Z there are the same But now. examine the case Z = .162 ELECTROMAGNETIC THEORY. So.

scos)s# pt nin R' + Zfr* The outside term represents the mean potenQ/(SZ + S + SJ. working as . 2 i?! is for the reg ulars.and the extras. where S and Si are the effective terminal permittances. we may transform (110) to _ (Rs hin + Z gcosh)yy (Rshin Z ?cosh)g(l . or x = 0. in many . '~S/ where tial _Q + 2Q 2 (Rsin + Z ~ SI si cos) (Rsin +Z . For the extra terms. But the ou tside term is only required when the that is. we find j/. in which it is . previous cases. terminal resistance is infinite at A. however. cancel one another. Say V = v i + v 2 where i\ for the extras. which.x) . and Then. = nir.

to be noted that the _p's determine the normal systems. so far as the extras are involved. We have d 1 \ "i / I dp \ . expands to (114) ranging over the roots of Z^ 2 = R2 . and putting it. using (113). Hence.y Z dp l / \ / when R = Z^ 2 2 . R=Z ^ in the numerator of (112).


specially simplified cases. 163 The = v l + v2 is the complete expansion. Finally. nit The simplest case of . making _1 "~ 2 (R sin + Z s cos)sy . unit impulsive fu nction is got by dividing by Q/S. has the f ailings as well as But we can easily interpret the merits of general results. (Rsin +Z s cos)sx +2 where the extras are in the second line alone. 300. Equation (115) being general. Herein Z may be any rational resistance operator. Equal Positive and Negative Terminal Resistances. and V omitting the time factor. It follows that (116) is the expansion of an arbitrary function /(x).PURE DIFFUSION OF ELECTRIC DISPLACEMENT. as before said.

start w ith . Therefore. Let f(x) = 1 for example that is. There is but one extra.all is that of a positive resistance A. _^ shin/^ + ? 2(5 7 s* n + s cos ) sy ' (' i s^n + sc s ) 5a. as per (104). potential in the cable. since dZ Q/dp = Q. Thus an easy integration O 7. following expansion . /i y^ + s2 a uniform gives us the expresses the unit impulsive function. Then h = R/R the reciprocal of the effective leng th of the terminal resistance R in terms of the cable. and an equal negative resistance (to be interpreted later) at B. extra q by h for distinctness. Denote the .

increases with the time.. STT./uc : -where sl = ir. The introduction terms with even multiples having of the time factor to every normal function shows the tional part result later as usual. ZTT. the .zero coefficients. &c. The summaThe outside extra term ii 2 . ultimately vanishes.O.

One normal That function is e**. It illustrates a property by the conjugate property of normal an absolutely sound method. the type of the rest is (h sin + s coa)sx. We may integers : if sl/ir is confined to the odd _i ~ 4 ' y A sin&g 2 4 As cos sx not a Fourier series in the ordinary It absolutely violates the conditions settl ing the size sense. Fourier series. It is only of the coefficients according to Fourier's theore m. the extra one is conjugate to the rest" we see thus verify (118) is We may . provided we have all the normal functions at command. however. of the wonderful function e*. VI. 2e 7ia: CH. write (118) thus. The function on the left But it is side is apparently expanded in a one example in a million.164 ELECTROMAGNETIC THEORY.

we may (121) I ivnv2 o dx = 0. J If. we assume that we can isolate any coefficient by quadratures. if wz denote any two of the regulars. dx =P J o dx = (t^siusx) dx 0. therefore. (120) MI and verify that Also. **rf. This : Pepsin + scos)sx /o .r .systems.

The quantity Changing its sign merely makes the positive and negative resistances change plac es.Thus 2 finds A and any . one of the rest ri is found by A -./o Comparing with h (118). . we verify that result. we have the case of terminal disconnecmay be positive or negative. Between the two. witk 7t = 0.

Unt o physics should we return for fresh inspiration. the physical assumption would be such as to make the normal systems be cos so. . There is but one extra q. So. we find that the new _1 ~l 2 I y. Equal Positive and Negative Terminal Permittances. and this would correspond to an entirely different state of things. The next simplest case is that of a terminal con1 Say that Z = (S^)" at A. its value is S/S A lso.R 2 instead of + B 2 as in . unit As regards the expand m of the left member of (119). FO as to verify that formula. (113). it by h. by (111) unit impulsive function is and (114). by using the new Z Denoting in we see that the result is . denser. 301. (R sin . if done according to Fourie r's theorem. tion. Out of physics came the subject of expansion in normal functions. the last case. and sins^. 165 The conjugate property may be readily applied to the impulsive function (117) it self.PURE DIFFUSION OF ELECTRIC DISPLACEMENT.

and we may write (125) more simply. (R sin + 2 Z s cos)sx _ R+ZV shinW (125) where the extra function is the last term. (ssin 2 hcos)sx _ h<? lx+y-l) /^Q) A- . (ssin I this /i cos)*// .h/s. Here sZ /R = .+Z s cos)*// . _1 I 2 _.

including or a constant term. 2?r. 'lying obtain u to the special state /(#) = !. sin infinite series of the type (s /tcos) sx... . and the coefficients of the terms belonging to 2-. &c. on account of as before. STT. &c. vanish The formula fails a t the terminals. the condensers. But to verify (127) by the normal functions an case 0. The and the s = are now **. we which gives another expansion Fourier's theorem.+6 shin/i// initial Ap. 47r. the sole extra function. of c** not conforming with In (127) sZ has the values TT.

CH.. if w = (s sin 7i co$)sx it ( 1 23) be taken as the normal function along the cable supplemented by must be wn = . Thus. we must take count of the terminal condensers. (130) which. Wi = -h cos si. Similarly. Soi^o'S^rrte + Sic"^. the normal function . . thus proving the in dependence of the system e hx &c. we systems is zero. (129) Similarly. and any one of the iv systems. may be seen to vanish. by the previous two equations. at the terminals. YI conjugate property.h.166 ELECTROMAGNETIC THEORY. is therefore . The mutual energy of hx &c. may show That .** must be hl supplemented by 1 at A and e at B. and iv. &c...

we find the coefficient of any normal system by fo rming the expression for the mutual energy of the initial state and the particul ar normal state. if V and Vj are initially charged as well. quadratures find the coefficients when we assume I /(a ? ) = A + 2A H ir n + B**. have not done it.is. Thus using (130) or (131) to isolate the coefficient. b ut have no reason to doubt it.lll tP l B +/o S'P. (132> according to whether /(a?) is the complete initial state. all the normal systems being proved to be i ndependent. Then. ways of doing it.m and w n are any two different regular normal functions. that the mutual energy of two w =S if H.>.w V^ + Si"V. or whether the condens ers In the latter case. (X=l} (131) (1=0) u. different But there are are their initial potentials. .

the coefficient of which is what w reduces . if WQH w ln are the terminal values of wn . if it be taken to to mean when s = 0.A ^S and finds An . And y V^ 1 finds B. A . The formula for An also answers for /*.


and and evaluate. except at the terminals. and (133). 167 Doing this. we shall arrive at (127). To find the proper exp ansion to represent V in the cable due to the the charge initial charge S V of the real condenser at A. in the case of (127). developing and V^ may also use (133) to obtain the effects due to that is to say.PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. (134). Initially we must have zero expansions Now. must make V and V t zero in f(x) alone is existent. put/(#) = f + l? The in (133). S^ at B. the results. (134). and we We V expansions which represent zero. and verify it. The results are I hi to be used in (132). result is therefore = 2pV* + 1 .

special interpretation is needed. This we multiply (127) by V it expresses V all along the Adding the result to (137).L_l + 2 III I ( V. we obtain 1 = 1. on account o f the everywhere. except condensers. because there . cable. cos a .V. (137) where only the odd values of n in corroborates (127). the potential of the cable at the terminals.cos ri). If we ass ume the mean term disappears. is it will remain constant. s = mr/l are effective. and we left side The on the V^V^ reduce the last result to = Z (1 . If .) (136) expresses /(#). This is But it means more than that here obvious by arithmetic. that if the initial state is V constant in the cable and namely. where. both condensers as well.

If 300. and further say that R x is is a positive constant. the most obvious interpretation The equation then asserts R ! is a mere resistance. say at B. that Ohm's law simply. But this is not necessarily the interpretation. we say of the cable.nothing to disturb it. is that the equation to be obeyed at one end "V^RiC. Physical Interpretation of the Abnormal Case of 302. .

does not require any specialisation. The distortionless cir cuit will behave to an impressed force precisely Similarly.RiC. still is V = . is what is impl ied by that equation of condition. either with or without waste of energy by conductive resistance. we say that the terminal condition Any arrangement according to the commonly understood electrical that will. it is To save circumlocution. Electric and magnetic energy are then involved. There do it.168 ELECTEOMAGNETIC THEORY. however. when acted upon by an impressed voltage. to be understood as above. if like a resistance. law. that the ratio of V to C is positive therefore be any electrical arrangemay A di stortionless circuit will that will produce this result. a positive resistan . compel the current to obey the law V = . we do not assert that That would make for a n egative resistance. of course. wh ich. the equation asserts is CH. obviously convenient to speak Now put of a negative resistance. VI. where R! is R! stands nonsense.RiC. and in many ways. it is to be done. since the resultant effect is embodied in the law. a positive constant. We are not obliged to enter any further into detail as to how A source of energy is involved. What ment and constant.

the latter but into the cable at B. The B end of the c able therefore becomes charged faster than A end. The current to match . and let the cable be initially charged at one pl ace. say at A. a current from the cable at A. say with a charge Q at y. when the normal effects due to the initial charg e have subsided. abnormal. The ultimate result. /ix e for the poten tial. and an equal negative resistance at B. therefore. the cur rent due to the initial charge would first rise to a maximum and then fall But t he continuous entry of fresh charge at slowly to zero.ce at one end of a cable. It spre ads all along the cable. This charge instantly begins to spread. raising the potential of the whole cable. The former is norm al. due to the terminal resistance being negative at B. is a state of positive charge increasing continuously with the distance from the A end. and the higher its potential is raised the stronger becomes the current i nto the cable. from B to A. slightly. At the A end. and raises the potentials at A and B There is. B alters matters. on the other hand. and increasing the normal function This is the meaning of the extra continuously with the time. where things are normal. therefore.

An B will depend on the sign of the initial point charge. left to itself. the current and .PURE DIFFUSION OF ELECTRIC DISPLACEMENT. current will leav e the cable at B. and the cable will become charged negatively. will in time either wholly disappear. z Both are associated with the time factor c*. or will be Equation (117) replaced by the dist ribution e**e'* x constant. is 169 directed from B to A. where p = h /RS. and continue to do so. If V becomes positive cur rent will enter the cable. when s applied to the initial An initial state of the type 4w = (h sin cos)sx subsides to equilibrium. at An arbitrary initial state of electrification in the cable. as may be readily seen by its satisfying the abnormal t erminal condition at B. determines the behaviour state. state of zero potential is accordingly infinitesimal charge arriving at B is suf ficient Which way the current will set in to disturb the equilibrium. fully. If negative. The equilibrium unstable.

case. first of It is zero if specialised so as to exclude in (123) settles the coefficient equation the integ ral of the product of the initial must be state is and e^ is zero. I n order that no term of this sort should enter. the initial state it. Thus. the sum of any number of s uch distributions. the **. of any sizes. however. if The simplest positive. Thus. will subside to But they cannot make up an arbi trary distribuequilibrium. the other negative. This is the case when the initial state the sum of a number of systems of the regular type w. distribution of the type e'tx . is to have two point charges.the potential having opposite signs there. because the abnormal distribution is left out. Similarly. with their size s suited to their distances from B. An initial tion. instead of subsiding. one . increases with the time.

at y l and y z respectively. Elliotson comes in useful. After so much detail concerning this simple abnormal case.Q x and Q 2 being the charges. the expression for u in (117) shows that the abnormal term c** does not exist. One of his students said he did others of a similar character may . be very briefly treated. an anecdote abou t Dr. As for why they are considered at all.

^)TT. and in addition. the solution. CH. the set corresponding to (Z ?) = R for a point charg e. unnatural. 303. of health can be ascertained. in +Z scos)sy . which indicates that we h ave the set of normal systems corresponding to si = (n . is reduced to . VI. where U is the regular part. we \V the e xtra part. (R cos + Z s siu)s(l x) . > qft at y itself. not see the use of studying morbid physiology it was so The doctor told him he w as a blockhead. 2 e subject to cos pt ." Positive Terminal Resistance and Negative Terminal Permittance. In the other way of simplifying the general determinantal equation describe d at the close of 298. we reduce it to the form (108). showing the potential at x on the right of // due to Q initially Let V = U + W. anl Then by the expansion theorem (30).170 ELECTROMAGNETIC THEORY. 2 2 So (100). obtain U = ?Q. adding. " It is only by studying the morbid that the true conditions .

we find ranging over the roots of Li*=(Z Q q)*. And. is The sum t the expression for the potential at time the source. take Z = R is . there is a pe rmittance of amount . Integrating the product of u and f(y) w ith respect to y therefore expands /(a?) in the Now. of U and on either side of W As an example. of value R/R = /t. and = U .B/h resistance R at x= I. in It need not be w. + W. as its form is obvious. a mere resistance.^ s* n + sl = 0. Then there To balance the just one extra q. if we put Q/S = 1. by the same process. written out. at x = 0. say. the appropriate unit impulsive functi on. it represents 0.


. if /(a-) = l. The expansion (141) may be v erified by the conjugate property of the normal systems. . we in which the in the evens. assuming = Aoe** + 2 A(/< sin + scos)sx. as usual.i.. + be used in the odd terms. and the Introducing the time factors. Remember that there is one terminal condenser.+ (143) and also + which finds A. /-'. (142) we must have the identity + which gives A S i?. S^ 1 x = A*s& + Sx. 1 So. 171 normal functions in question. shows sign is to what the initial uniform state changes to later on. obtain the expansion In particular.PURE DIFFUSION OF ELECTRIC DISPLACEMENT.

tion in Sj/isinaZ x = A/S S^sin^j. (ssin . 1 . Z = (Sort" which makes find thatand a negative resistance at B. Say the one extra q be h = S/S Then.va? ~~~ . An Carrying out (143). (144). of the terminal condenser. instead of the above-used unit impulsive function. (144) Here w stands for the general normal funcThe which occurs stands for the initial (142). . real condenser at A. Put a interesting modification may be mentioned. potential =S//t. coefficients in (141). Si is its permittance. we shall 2 " * (ssin 7tcos). we shall arrive at the and verify that expansion.

let without resistance. turning to the more interesting case in which Zi.7*cos)sy /-. Equal Positive and Negative Terminal Inductances. Z be L p. ~x is the proper one to use. Z= coil 304. . where (115) shows the proper unit impulsive function. indicating an inductance of the The determinantal equation . treated in 299. E.

say. (146) finds the three extra JD'S that p=y. (106) shows that s P= it = /. in virtue of (146). where = 9( -* VI)So. is.. a cubic. Thus. we may write E/L p for q. and the assumption of no resistance makes manageable. CH.172 extras is ELECTROMAGNETIC THEORY. and i //.. by Also. the extras is given by TZ L(}ff R . the part of the unit impulsive function depending upon (115). 9* f/i. VI.

shows the V that res ults. t0lt e^.2 ). The part of u dependin g on the regulars is given by the first line in (115). = . Of the three terms in (147) only the first increnses with the The other two unit e to make a real function indicating oscillatory subsidence. (147) L _ o'/ 7 . 291. The time factors to be introduced later are e *. given.Z^ The same is extra functions are. 1 he full development need not be time. Put Z . For example. because it depends on two inductances separated by the permi ttance of the cable. an impressed force. but it is of a comp licated mixed kind. it. let it Equation (59). concerned when be the source e at # = 0.+ (same shin with g^) + (same with y. omitting the 1 jl term. Impressed Force in the Case Z 305. however. because there is terminal earth connection. of course.

by making Z =K .= -Ziin Then v= gives -_O_V in terms of e. . Specialise. Then where h denotes R/R . a resistance.

(151) Multiply by ** . That is. ( (150> without the time factors.. . develop by s Put = The result is *cos)g ' . In the summation. We may Assume that verify the result by the conjugate property. and then the expansion theorem (30). we have si = rnr.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. and at the This is because there is terminal resis tance. as in previous cases. everywhere between x = and Z. representing V = terminals as well. it is the initial state.. 17B to get the part independent of the time.

Singular Extreme Case of Z or -Zj equivalent to the Main being a Cable One.. + s cos) sx and integrate to find A. It is got which will be useful by differentiating (150) to x. . I give this case for the sake of a result a little later. Theresults are as in (150).Multiply by (h sin and integrate along the cable to find A<. We (7 t obtain ju-x) = l hi g shin A J + si sin) sx (A/) + (/)* cos 8 Q . 2 which will be used in another problem.

of such functions depending upon These abnormal series are merely special cases of the general series got previously by leaving Z and Z x independ ent of one another. be made negative. inasmuch We as the wave lengths are the same as in real Fourier series. which constants are real and p ositive in real problems. where h is a constant. real or complex.abnormal Fourier we 306. . If one of the any electrical constants involved in these operators. It will be seen from the preceding that the subject of series is quite a la rge one. find that require to use series that look like Fourier series. but that the complet e set of normal functions includes extra functions of the type **. as the case may be. the number the terminal conditions.

result.). I will therefore conclude the present remarks on .Some cases of this kind were considered by me in 1882. time functions. But it would be out of place to elaborate further in this work. I. and.rigorous mathem aticians are sadly lacking in demonstrativesufficient information to enable ness as well as in comprehensiveness. and earlier (" Electrical Papers. as a p< oc cur in which p is positive. . . I think I have given any competent person to follow up the matter in more detail if it is thought to be desirable. because there is a reversal or of the law of displ acement. CH.abnormal series with the consideration of a very singular case indeed. in of the electrical system concerned. or of induction. It is obvious that the methods of the professedly . some part . VI." Vol. introduce something abnormal. or . In none of the previous cases did we cause the terminals to be nodes of the regular normal functions.174 ELECTROMAGNETIC THEORY. as e of we Ohm's law.the The peculiarity of the cases above treated lies in the simplicity produced by reducing wave lengths of the regular normal functions to be the same as in the original Fourier series.

"V = terminally in them. s (153) Z is a cable identical with the one in question.make this.Z x above. and make Z = . to (59). let Z This says that =-tans-/. This produces 191. So far it expresses V due to e at x = Q with any (148) terminal Z provided there is its negative at x = I. In the following we shall do . earthed at its end. Next. Go back . x=Q x=l . We have therefore Z e I i__5 i x--l $ 4. or equivalent thereto.

e is started and maint ained constant. It is of -equilibrium. extending I to + 1. It is just at . is. without other change. two of which are alike. with a neg ative cable added from x = I to 2Z. little consequence to follow up the later effects.x=2l from that three cables in sequence. a cable in which the permittance and resistance are This arrangement being in negative.

This is highly The (154) by the expansion theorem. if existent. because the is (155) initial state is where sin *l = 0. For it makes the final continuity. to (155). Now this is incorrect. apparently V= .result is V Expand permanently zero at # = 0. . should The corresponding be the current through a . this. The condition (153) -reduces (148) to the very simple re sult (154) si tan which makes anomalous.(/ 2 SL sin** e*. something missing.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. according ?. There in But there -current be is some sense it. 175 the beginning that we require the behaviour. so as to obtain the proper expansio n of the initial state.

r egarding the rest as a mere terminal arrangement. So e establis hes V in the Z cable just as if there were an earth on the other side (the right side) of e. so far as the cable Z is concerned. According to (155) a complete normal system of potential would consist of sin sx * To find this operator is x = 0. state of the middle ca ble is in question. which is where the at x After this partial explanation return to (154) and (155). By The zero potential at # = is accounted for. . reckoned at # = 0. equivalent to a short circuit.state of potential would be then as in the figure. It will be found that the resi stance operator of the two cables* from x = to 2Z. The resistance operator to the right is zero. is zero. when Z zero. We may corroborate this by calculating the rise of potential in the cable Z due to e.ll three cables.(R/s) tan si. we may use (59). but then the potential = is zero. C = <?/R/. . 291. so t here is compensation. and Zi is .

It is the ratio V/C at .

It = is the equation belonging rep resents sec si x sin si. in fact. VI. (155) proves that the sum of the whole of the missing It is certain that may ask.x) from I to 2Z. The normal functions to correspond Thei r are therefore unmanageable from being ungetatable. is in full operation. however. and =/(2Z the result of expansion by quadratures is ze ro.1. up to x *= Z. so to . make it represent the we should have. infinite in to number. V V And. But it is to Z. and sec si the extra functions. where are the abnormal functions necessita ted by the negativeness of resistance and permittance in Z x ? The answer is to be found by seeing how tan si in (154) arose. zero from instance. from x = to 21. combined with sin sx from x = I up easily to be seen that we cannot expand an For arbitrary function in terms of th ese systems alone. . That is. I CH. For the failure of influenc e. But the roots of secsZ. functions required to is e. The e nergy is zero. we (155) is incomplete. initially. are at infinity. let =f(x) (any function) from .176 ELECTROMAGNETIC THEORY.

Then '(148) x} 'makes .-e . Terminal CableZ not equivalent to Main One. The cable Z is no longer an exact or equivalent copy of the middle one. Take.initial state truly = e . 307. then.e 2 I sin sx. More General Case to Elucidate the Last. (157) instead of (153). Si (156) where the first e is the sum of the missing terms. But to make this conclusion plainer. so that we can manipulate them. we should bring the roots at infinity to a measurable distance (like the abolition of that fraud 4?r which has been pu rblindly inserted in the electrical equations).

Here = Es /EQ s or (RS /R S)*.c~ 1 1 tan.9^ instead of (1 54).s n ? cos)s(^ i 2 6+ c~~ tan' c e.=i Vfsin ( . ^ ) sin . . The abnormal roots are now finite.

c . and W the being the part of extra part. U sin si = 0.PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. or . We get (with = 0) < 159 > where reduces to the right are given by 2 (RS/YR S Z )*. This U evidently member of (154) when c = !. Develop V U t depending upon in the usual way by the expansion theorem. The developmen t of The extra roots requires more care./=!. / is the constant W tan2 s Z = -c2 or . 177 Let V = U + W.

than'yA-c*. is indifferent, (160) Whether we take c we work consistently. which is to be than gQ lQ provided Say, then, that ihwqQl = c, Q (161) or, the same, **j =/, tf say, (162) which makes 3<fo = ^ + inTr, 1. l * = log0, >1, then (163) provided c is less than q But if c is =X + i(n-%)7r real constant.

(164) instead, where X is the same is, by W when expanded At the point c The form assumed two is cases. =l therefore, quite different in the there is a jump, and special care required. Let us use (161), with c<l. Then 1 . p tan 2 Vo = Vo tan S (1 2 + tan\Z ) (165) '

)application of the expansion theorem to (158), so far as the extras are concerne d, therefore makes The W- ^ In this, the relation between q and <? is ql=fqQ lQ and in (163) n has to r eceive all integral values, positive and negative, The n ih term of (166) pairs with the - 71th including zero. , to make a VOL. n. real function. But, owing to the exponential N

178 ELECTROMAGNETIC THEORY. CH. VI. function in the denominator of (166), the resultant takes a rather complicated f orm, which need not be given. It is sufficient for our purpose to see that if / is an integer, the denominator simplifies. We then get * W = l_ that is, ^ c y/_l Z4*; qj,o (167) by (163), e w_ Zee fl ^ 9V A2 + 2 , . TTpr^vrr-jx . (wr) j' all posi where the summation is with respect tive integral values from 1 to oo to n,

which has Comparing the last equation with (152), we see that lil becomes A and x becomes fx. The quantity in the big brackets in (168) is therefore (169) by (163). This reduces W to 2 . (170) ] provided / is an integer. Let it be 1, i.e., RS = R a S L2 then , W= Finally, c l. (171) e, as was to be proved. have proved that in the singular case of identity or equ ivalence between the terminal cable Z and the middle one, when the extra normal functions apparently go out of exis= 1 makes

W= We tence, infinite, by the roots of the determinantal equation becoming they are nevertheless virtua lly existent, and must be allowed for in the expansion of the initial state. In the more general case, in which there is not the equivalence mentioned, no such straining of the expansion in (168) may be tested theorem is needed. For instance, W in other cases in which/ is integral, in (159). the negative of the to doubt the equivalence of in (166). U U and be found to represent There is, therefore, no reason with the more general f orm W

PURE DIF1USION OF ELECTRIC DISPLACEMENT. 179 Arbitrary Initial State when Z or Singular Case. 308. If there Zx is a Cable. is is no impressed force, but the middle cable charged arbitrarily instead, there i s a similar curious limiting case when the terminal cable Z and the middle cable are equivalent. The general solution (100), initially through (110), reduces to indicating the potential at x, on the right of y, due to Q When on the left side , interchange x and y to initially at y. get the formula. The Z operator is (R/s)tansZ, as in 306. the expansion theorem, according to the Developing by roots of sin si = 0, and denoting the result by U, we obtain U = ^2 sin sy sin sx 4*. &v (173) Now this is exactly the I. same as if the cable were earthed at x = Q and to exist See (22), 267, for example. But the clusion

normal functions, which we know The congeneral, have not been considered. from t he above result is, therefore, that in the latent auxiliary in complete expansion of the initial state, the sum of all the extra terms is zero. Only when t is finite can they be existent in the total. They may be infinite q uantitatively p* then, should the p in e be infinite for the extra terms ; but that is of no consequence To verify the above, we Z be not equivalent to the middle cable in the first pla ce. Then Z is as in (157), and Zj is its negative. The general equation (110) is as regards the initial state. may proceed as in 307. Let now l V = ( c gm + tan S(A cos) sy (c sin - tan sQ Q cos) s(l- x) sQ ' instead of (172) above. Developing this by the expansion theorem, regular part, and TT ' let W the extra part. U be the Then we get ST


180 in ELECTROMAGNETIC THEORY. sin si CH. VI. which = 0, or si = TT, 27r, &c. And, in addition, ' w= s 2 . (i(176) the notation being as in 307, equations (161) or (163) The unit impulsive functi on is the finding the values of qQ sum of U and W, with t = 0, and Q/S = 1. subsides to equilibrium, whilst the the time. W 2 The part part increases with ^, U

Simplify by taking /=!, making RS/ The unit impulsive function becomes =E S <x and ql = qQ Q l . (177) Here the first part expresses the unit function concentrated x = yt with repetitions outside the limits conWe conclude that the rest cerned, which do not count. zero between the same limits. The verification is represents at the point 1 allows easy, because the simplification made by assuming of the reduction of t he extra part to a recognisable form. The extra part of u, say u^, is /= where n has all integral values, Or, which is the same, positive, negative, and zero. os T (x + y} }(179) This, by 271, represents a row of impulses (not of unit size) Within the limits, therefore, outside the limits in question. u.2 is zero, as required. t is Let u become v when finite.

Then (180) is ^- 2-rfn55rin^.I Li + "( where n. H is nV/RSZ and , 2 the summation with respect to And for the extras, we have x> = exp ^ + inw _ nV + 2mirA) , (181) I

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. 181 which makes Va = J^_ l + 22rcosrG.e- H , '} (182) where H is as before, _ and, for brevity, additional letters F, G are introduced, given by + + Whilst i\ subsides to zero, r a augments. In the singular case of failure, real or apparent, which occurs when c = l, the extra part va jumps from zero to infin iteness instantly. Real Terminal Conditions. Terminal Arbitraries. a Coil. Two Ways of Treatment. 309. Returning Case of now to real electrical conditions, one of is the minor matters that remains to be considered the influence of the terminal arrangements on the state of the The existence of cabl e, when they are initially energised. auxiliary functions, one for every independent kind of energiAlso how, by means of the sation, has been pointed out. conjugate property of complete normal systems, their size may be determined. But the example given in 295, relating to a terminal condenser,

and the later examples, were not sufficiently general to illustrate the matter fully. So now take some other arrangements. Say the x l end of the cable is earthed. Then Z 1 = 0, and, by (85), 294, *-smsV-*^^^^ S +Z '(Rsin scos)sZ (184) is the potential at x (>y) due to Q at y. Interchanging x and y makes the potent ial at x (<y) due to Q at y. Differentiating the new expression with respect to x and dividing by - R produces the expression for the current on the left side o f y. Put x = in it. The result is A) C= . S (185) This is, therefore, the total current in Z due to Q, reckoned

from left to right.

182 ELECTBOMAGNETIC THEORY. CH. VI. We established in conclude, by reciprocity, that a current momentarily Z will produce potential at y, determined by the ft same operator inertia in Z . as is concerned in (185), when there is magnetic This is easily corroborated. Pu t an impressed voltage e at x = 0. The C due to it, on the spot, that is, in Z , is (186) to right because the denominator is the sum of the resistance operators and left of e. Th is expands, when e is steady, to O-Z^.

= (and a steady term), ranging over the roots of Z 0. But let e be impulsive, sa y = ;?L C so that L C , (187) is the momentum generated. Then C= Putting t L C =L C 2^~. So (188) = makes the initial state. < 189 ) must be the expansion of C the current in , regarded as initially given to be e is Z . Also, the potential at x due to e in (59), got by putting L ^>C for

and Z 1 = 0. This makes operationally, or -j derived. It will ' 9QQ68l(dZ/dp) by the expansion theorem. From this, again, (189) may be p be convenient to consider Z to be E +L and inductance in the above, meaning a coil of resistance At the initial moment, therefore, (191) is the zero expansion required to suit the case of an energise d coil, an K L . limits, expansion in a Fourier series representing zero within the such that V shows wha t it becomes at time t later, as itself into the coil discharges the cable.

PUKE DIFFUSION OF ELECTEIC DISPLACEMENT. 188 viz., Next consider the same matter from another point of view, that of independent no rmal systems. Put r = sin s(/-#) x = I, these c=-eoss(l-x). R (192) for Since v = at are the proper s is is, normal functions V in the cable, provided - Z at x = 0. That satisfy v/c and C suitably determined to = Z + ?tansZ=0 = Z s (193) is Provided, then, we include every possible normal system subject to the last equa tions, we see that the initial state is fully specifiable by the determinantal e quation. V = 2 Av = 2 A sin s(l - a), C = 2 Ac = 2 A 1 where (194) A

is a constant. B Given V cos si, (195) , and C the conjugate property finds A. That property o asserts that ^-L emcn = 0, cn (196) where vm, vn are any two normal systems of potential, and cm1 the currents to ma tch at x = 0, because there is magnetic energy in the coil, and electric in the cable. It follows that S/V A _finds the coefficient is t;<te -L C c A belonging to any system v, c when V given as a function of x, and C has any va lue we like. Attending only to the part dependent upon C the result is , ( 198 ) which, used in (194), completes the solution so far as C is concerned, since the state at time t is deducible by the introduction of the time factor. It will be found that the solution thus got agrees with the former one, viz. (191), on expanding dZ/dp. Also, that (195) agrees with (189). As regards the

184 ELECTROMAGNETIC THEORY. CH. VI. conjugate property (196) which has been used, its proof is to be found in testin g that it is true, when the proper expressions for v and c are used. This remark applies in all expansions in normal functions. I shall, however, give later a general proof of the property. In the numerical execution of the above case, it may be as is sometimes a pair o f irnaginaries concerned in the determinantal equation, which might be well to remark that there overlooked, if we merely drew yi the curves 2/ 2 = tanrf, = (-s/R)Z , (199) and determined the roots by their intersections. Terminal Coil and Condenser in Sequence. 310. Now take a case in which both kinds of terminal energy are concerned. Say ZO-BO+LOP+* b so that there are a coil z 1= o, p in sequence at

(200) and a condenser The determinantal equation is x = 0. Z=0 = 5tan^ + B + L oP + -L, s bp where p = the coil. s 2 (201) /RS, as usual. as before, equation (192), the current at Taking the normal functions # = is the current in terminal arbitraries Therefore (195) is the expression for one of the The viz., the initial current i n the coil. other arbitrary is the charge (or the potential) of the condenser. Say that the condenser is next the earth, and that its potential is v 1 in a normal system and V x altogether. We have (202) ^-^(Ro + Lo^c, where v and c belong to the terminal of the cable, and also *!---. OOP respectively. (203)

these being the equations of voltage for the coil and condenser So =* ! sin si + (R + Ljp) ... B cos si (204)

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. is 185 the normal function for v r This may, by (201), or directly by (203), be simplified to r 1= =JLcosaZ. (205) 80S The initial state is therefore fully expressible by , V = ZA*, A is C =2 Ac, V = ZAVl T (206) a constant, the same constant for every three where connected functions, but dif fering in the different complete systems. arbitrarily, Given, then, V as a function of we evaluate any coefficient A by x, and C and Vj A=s because the conjugate property is = 8fcmvndx + S vln p' - L cmcn,

(208) When L =0, the terminal energies being of different kinds. the value of C ceases to have any influence on the value of A, as we see by (207), although the expression C = 2Ac gives the coil current for Coil (209) any finite value of the time. in Parallel parallel, the case and Condenser 311. If the condenser is somewhat different. and the coil are in We now have (210) because the resistance operator Z is the reciprocal of the sum of the reciprocal s of the resistance operators of the coil and condens2r or the conductance operator is the sum of the separate conductance operators. Th e potential v at the terminal of the cable in a normal ; system is the potential i\ of the condenser, so V = 2Ar = ZAsinsZ 1

(211) so expresses one of the terminal arbitraries. But only a part of the cable current enters the is coil, it not the full current c that is concerned in the other arbitrary. We have

186 if c is ELECTROMAGNETIC THEORY. the proper function to associate with v. CH. VI. Therefore (213) the expression for the current in the coil initially. But to determine A to suit the initial circumstances, we may still use the formula (207), taking care, how ever, to use is the proper expressions of the present case for v, v l and c. The last has now be come the C Q in (212), whilst v l is to be the v in (211). Two Coils in Sequence or in Parallel. 312. If the terminal arrangement consists of two coils in sequence, there is but one auxiliary function, because the current is constraine d to be the same in both coils. Suppose an impressed voltage, and there is mutual inductance m The equation of voltage i s between the coils rlf ^ and ra l z e is , . e = (ri + IIP) C + mpc2 + (ra + Izp)c2 + mpcu I (214) to c, where c and c2 are the currents.

Both being equal say, this reduces to = (n + r the 2)c + fa + Z2 + 2m)pc, (215) same as for a single coil. So there is nothing new here. But if the coils are in parallel, there are two arbitraries. For we shall now have e e = (n + (ra ty)^ + mpc2, I2 (216) ; + p)c2 + mp^ (217) from which the resistance operator of the combination and the separate ones have to be deduced. Solve for the currents thus : These are the conductance operators separately. Their sum is the conductance ope rator of the combination. Its reciprocal is the resistance operator. Therefore Z ^fa + Mfo + ^-roV is /oio\ the required Z , to be u?el in the determinantal equation. Cj Also, the previous equations for

and c2 give the proper

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. 187 normal functions for them in terms of e, which becomes v, the normal function fo r the potential at the cable end. Thus determined, the two arbitraries are ZAcj and ZAc2 , to be associated with cable. ZAr, expressing the potential in the In determining A by the conjugate property, be forgotten. Twice the energy of c x and ca is I m must not A+ Z 2cl + Zmc^ = F, say, ; (220) or fa + MC^C! + (lcz + mcjcz (221) and the mutual energy of C^ C a and (Ic^ clf c2 is + wic2) G! + (lc.2 + mcj G 2 = G, say. (222) along the cable, and also the values Consequently, given of C and G! initially, the value of A is V which renders the solution of the problem complete, when the normal potential fu nction v for the cable is properly determined to suit Z as in (219). The principle underlying these determinations

Take an explicit example.is quite a simple one. Let the length of the cable be At . pass on to another minor matter. It is may be readily inferred from the preceding. and in the treatment is. Some difference not great. with or without an earth connection at the same place. Split into Two Simpler 313. and may be applied to the most complicated cases. A Closed Cable with a Leak. . thereneeded. Cases. we If the cable ends A and may B are joined. either directly or through apparatus. the two condit ions expressing V/C at the terminals independently of one another are replaced b y two conditions of a different kind. fore. The last examples tion between the main sufficiently indicating the connecsolution for the cable itself and the terminal arbitraries.

.B. without 2Z. there is simple continuity. where x = I.

As regards Q at y. however. and a little thought. we may split the problem into two. and the effect of terminal energisation. VI. the effect of any initial state of the cable. so we need not cons ider it specially. of which the solutions are already known. imposed condition. four charges is simply Q at y. P air the other with an equal charge of the point 2 The sum of Ihe opposite sign at the corresponding point. We At A. when Z is of a suitable kind. the effect of an impressed force anywhere . where x = or 2Z. Split Q into halves. Pair one of them with an equal charge at the corresponding . defined by a resistanc e have to see how this arrangement will *1 behave. It is operator Z . JQ at y and JQ at they cause no current at B. so the solution is the same 'or is. CH. or leak to earth. By inspection of the above diagram. We may consider three sorts of problems: the diffusion of a point charge initially at y. which will give. by integration. may be i nferred from the solution for an initial state in the cable. The last.188 ELECTROMAGNETIC THEORY. there is a shunt any sort of leak.y. and behave At A the currents s .

The no . to Z are equal. So the JQ in the lower cable discharges as if it were earthed at A and B. that 2Z . and so does the . with another Z .JQ in the upper cable. charges in its own cable in the same way as if it were insulated at B.imilarly in the upper and lower cables. There is no leakage current in potential at either A Z and consequently continuity of current between the upper and lower cables. on the other hand. each member of the positive pair disthrough Z the four as for the original Q. positive and negative pair. therefore unite and leak out They going Thus. This part of the complete s olution is also known. . . and were earthed at A through the arrangement whose This case has been already conresista nce operator is 2Z 2Z may be constructed by putting Z in sequence sidered.y. produce or B. or in other ways. By symmetry . Consider the positive pair alone.

189 Putting these together. on the right of y. . =Z . Then 294. and 2Z for $m)sy C<3 ^" a) ? . in the lower cable. Interchange x and y when x<y. we obtain the complete result. or the 267. . and can use pr2vious form ulas without more work.PURE DIFFUSION OF ELECTBIC DISPLACEMENT. In the case of the positive pair we have this arrangemen t w diagrammatically. In the case of the positive and negative pair we h ave this arrangement. sQ ' ( expresses the potential at x. Put Z x = oo . Use (85). -JQ 1Q and we may use the quoted formula with Zx = special formula (18). Z and iQ for Q.

whilst v a is negatived. and their negatives in the second. when greater than y. In the upper cables the potentials are the same at corresponding points in the f irst case. The same formula is valid all the w from y to # = 2/.x. we see that V =v +w l l l (8) is x ay l the potential at x in the lower cable.The result is that expresses the potential at x on the right of y in the lower cable. . as we may see by altering x to 2f . and observing that v is unchanged. and the same with and y interchanged when x is less than y. Uniting the two formulae.

x) *pt 2B s cossl + (R + 2R //) sin si In the . When we make R = oo we have the two cables connected We then reduce the to gether without external constraint. we have first summa(5) for determinantal equation. The result for the upper line need not be written separately. The conversion . In the limit the root is zero. VI. on both sides of y. Notice the way the equation has a very small root when B is big. . we obtain in the done as usual by most simple case of V = Q 2 sin sy sin sx ep bt Q^ -y (2 *~ * $1 R s cos + E sin). Doing Z = E a constant. the expansion theorem.190 ELECTROMAGNETIC THEORY. and the corresp onding p is very small. In the second. being the same for mula. to Fourier series is to be it CH.* ' which is now valid tion sin si = 0. cos s(l .s?/.. and the result is a const ant term showing constant term arises. .

If Start with Another Way. Continuity of potential at A makes (2). (8) . notice that the problem admitted of two in the above way. This is an impulsive external source of as explained before. and four conditions which find them.above to Fourier's periodic formula. we would proceed thus. V x is the potential at V 2 when on the left side. Then let x when on the right side of Q. is a uniform state of potential in the comClosed Cable and Leak. 314. where (1). The last that the final effect plete cable. Continuity of potential at y makes 'I. we should not splitting into current Q pQ = h. at y. and There are four time functions to be determine d.

y) ~~ (14) equations readily as the That the determinantal equation splits into two distinct is shown by the denomin . T> V~ 07. A makes (9) at* = 0. Or -C^C^pQ.sin s(2Z ziQ ~~ \ y) sin s# + s sin sZ cos s(Z + x . (H) (12) These four conditions determine the time functions and make. Discontinuity in the current at 191 (3). finally. Discontinuity in the current at y (regarding h as a function of the time) m akes h at# = /.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. Or (10) (4). i .

and taking the difference. w hen it becomes multiplied by the denominator. One of these partial operators produces vt and v 2 th e other W7 and w The expanded solution in Fourier series is the X 2 same as befo re got. But this difference has no existence in the Fouri er series. the other the bracketed part o f the denominator in (13). one having the denominator sin si. We can see that by putting the extra part of (14) in the numerator. applied. sum . we have h any t . p instead of the impulsive source of current ^Q. this extra operator is not of no mom ent hi general. We also see the same by observing that only even powers of But are involved in the extra operator in (14). It is equivalent to exchanging y and x in the preceding formu la. of course.ator in (13). If. and it may be shown that the operator in (13) may be represented two operators. The difference between V x and V 2 is operationally exhibited in equation (14). and therefore gives zero terms whe n the expansion theorem is of .

or else zero potential. CH. with similar effects at A. Z produced a discontinuity in the But if we introduce current. i t is the current that is continuous. because equal simil ar charges at y and 21 y will produce no whilst equal unlike current either at B or A. in Z will produce no potential at B or in the middle of Z charge s . discontinuous at A. and have no earth connection. though none in th e potential.ri*. In the above. (14). p roduce zero current at B. Split into Two Simpler Cases. for example. When we have an impressed force e at y instead of h or pQ. . we see that the problem splits into two connected simpler cases.192 ELECTROMAGNETIC THEORY. and the algebrisation is effected by p* = . Z in circuit with the double cable. The question arises whether the problem Zo admits of splitting. noting now that C is continuous and V discontinuous at y. Closed Cable with Intermediate Insertion. we can treat the matter similarly by either of the a bove ways. Since \e at yt together with -J0at the corresponding point in the upper cable. . It clearly does when Z is a resistance. the solutions to left and right of the source are got by w riting h/p for Q in (13). that is. and the difference between the forms of Vj and V 2 cannot be neglected in general . function of the time. and the potential 315. when h is simply periodic. VI.

simple electrical arrangement which is divisible into similar halves. That a similar split takes place when Z is a condenser is The same may b e said of any also evident on consideration. in one of which there is insuA and B.So the problem lation at both splits into two. and earth through resistance J R at A. whilst in the other there is earth at B. which may have no sort of sy mmetry with respect to the It is sufficient to think of a coil and cable termina ls. But it is not immediately evident that a spli t is possible when Z is any electrical arrangement. .

193 a condenser in sequence. The same is true of any regular combination. time. all in may Then one of the new condensers and one of the se quence. So. It its be symbolised by an operator. entering into condensers. Where is the middle ? The condenser and At the same coil are not compa rable in the way suggested. we say that equal unlike charges at y selves in the same way as if earth or. without detailed structure in the form of coils an d may The proper interpretation is to substitute for the combination symbolised by Z t wo combinations in sequence. witho may and the corresponding point redistribute them were on at B. Given a charge at //. If it be said that if the charges are unlike there is zero potent ial in the middle of Z . the property must be true in a certain sense.PUKE DIFFUSION OF ELECTKIC DISPLACEMENT. because the combination of condenser and coil behaves to external it may voltage as a single whole. be asked. A similar process is applicable to ut actually doing it in detail. each coil This is easily done. any as in imagination. as well . The condenser and symbolised by JZ be replaced by two conde nsers and two coils. new coils in sequence makes JZ provided the resistance. inductance. without . Z has now a middle in reality. and earth on in the middle of Z . combination. . and elastance of the original coil and condenser are all halved. Similarly as regards the remain ing new condenser and coil. and an equal charge at the corres ponding point.

we have either by (85). we have (2Ksin VOL. Q and . as if ments each equivalent to JZ we . Also. if w : is the potential at x on the right of y due to at y Q . n.speaking of the middle. B were insulated.y. Q at 21 -y. . substitute for Z two arrangeand then put on earth between them. with Z = oo =Zi and Q instead of or by the elementary theo ry in 269. 294. They If v: is the resulting potential at x on the right of y. So let move as if there be a charge JQ at y and the ends A.JQ at 21 .

+ Zi sms(t -x] .scos)sJ .

Z . the operational formulae ar e much easier Note in the present case that there is a conto manipulate. with Zj = 0. VI. stant term in Vr It arises from . it is the special expansions required here. the other hand. JZ instead of only CH.191 ELECTROMAGNETIC THEOKY. and JQ instead of Q. (30) is very easy to unnecessary to exhibit The expansion theorem On remember and to apply when required. the resulting potential Vj at x on the right of y is Vj-^+ufc The expansions of i\ and w are have given so many examples that l (17) I to be effected as usual. It follows that if there is initially Q at y. by the same formula (85).

is the mean potential which is finally existent because there no leak anywhere. But as the above reasoning about the splitting of Z not be wholly convincing. that l is.v lt and its value is Q/2S. G x = %l. let . as last without Initial Splitting. Say there is initially Q at y. Let may . from y up to Also. we may justify its accuracy by a direct investigation. (18) at x on the right of y. Same 316.

(19) on the left of y. (22) (23) V = V -sms(x-y). b 2 1 . V 2 = V + sms(x-y). Doing this. Cj and C 2 being the currents corresponding to Vj and obtained by the oper ation . The condition of continuity of the potential at y satisfied. we get (20) V2 . (21) making and C2 = C x coss(x y) .-. }>Q. at y. is already We also require jjQ^-Cj.(d/dx)/H>.'K V meaning the expression 1 in the previous equation.

PURE DIFFUSION OF ELECTRIC DISPLACEMENT.pQ -tv .x + y) + (Z s/B) cos sy cos s(2l . G) = ^ .cossy . we obtain 2 sin si cos s(l . . G to be The condition JL(sin 2sl. found. of continuity of current at the ends of Z gives R F .cos 2s/ . 195 There are now only the two time functions F. (24) and the condition Vi-V^ZoC. . Using the results in (18) and (23). which is (25) gives the equation of voltage for Z (26) These equations find F and G by the usual algebraical work.

These are the complete operational solutions without splitting.0. Now. add tog ether the vi and w1 of (15) and (16). The result is (27). And if we expand according to the other we shall obtain the expansion of w 1 in (16). we shall obta in the expansion of the i\ in (15). as required. To prove this. and tan si = -Z s/2R. uniting the two fractions to make one with a denominator which is the product of the former ones. by the expansion theorem. when in the fully expanded form. '28) initial We see that the determinantal equation does split into two. attending only to sin si = 0. condition in (29) Thus ^7 1 = v 1 + u' 1 . The expanded formula for V2 . But the equivalence is also true in the operational form.sQ/S.. (-29) proving definitely the validity of the previous reasoning.x) _sQ Z s/R) 2S ^-j) ^. giving sin si . if we expand the above Vj.

The same is therefore true for The reason is that there is not the instantaneous introduction (except at the initial moment of o2 .is the same as for V^ The complete solution for the whole cable from x = to 21 is therefore given by a sin gle Fourier series. although it is separable into two of any initial state of different types. the cable.

(31> (32) V 2 = Z 3 C l -(Z 2 + Z 3 )C 2 In these equations at x . V -V = Z. V = Z 3 (C1 -C 2 ). VI. we should want 21. the current. 317.196 ELECTROMAGNETIC THEORY. say a constraint at x = z. CH. but there would be a break in the normal syste ms. 1 We 4their junction. two distinct Fourier series.C2 2 . have done one case involving a discontinuity in and another involving a disconti nuity in the But in general both will be discontinuous. one for each side of z. the connection between x = through Z + Z 2 in sequence. Closed Cable with Discontinuous Potential and Current. and x = 2 be made for example. we have Let V be the potential at the junction. these voltage equations to satisfy : Then Vj-Vo-Z^. (80> and therefore have the two terminal conditions V = (Z + Z 3 )C -Z 3 C 2 1 1 1 . Let. potential. w ith a shunt to earth put on at. The determinantal equation would be the same for both sides. to of the charge) any sort of constraint all the way from x = If there were.

Cable in Closed Circuit without Constraint. The result is. the re is an initial charge Q at y. case of a cable forming a closed circuit in itself without external constraint d eserves some notice. If. then. Z2 are equivalent. applied to (31). if Z x and It sometimes divides into two sets. we get symmetry. we can obtain the operational solution by means of equations (18) and (23) above.. If we have an . a single Fourier series to represent both V T and V 2 For example.. of course. A 318. to det ermine F and G. as before. = 21 . V x and C t mean the potential and current and C 2 those at x = 0. V2 . (32). but it h as a somewhat different application to the The one before made.. Fourier's periodic theorem is involved.

in every interval of time equal to 2Z/f. each containing JQ of electrification. at any point sources. so that the effec t at any point x is the resultant of the overlapping of the right wave on itself . 197 an infinite series of uniformly infinitely long cable containing point sources o f equal size. There is no reflection anywhere. The apparent multiThe matter is best underplicity arises from overlapping. the time jump s round the circuit. it splits in to halves. one to the right and the ot her to the left. Then. there is one source only. in the corresponding case. whilst making the speed of propaga tion be finite. which are at distance from the source. and of the left wave on when two The potential at x makes a very little jump itself. and which therefore rush round and round the cable in opposite directions. this diffusion has been described before. Every source produces a double wave. diffusing and the effect produced to the ri ght and left respectively . say v. stood by introducing self-induction to such a small extent as not to alter sensibly the shape of the waves. an d only two waves. if there is a charge Q at the point y. .PURE DIFFUSION OF ELECTRIC DISPLACEMENT. we produce periodicity by spaced symmetry. There are either of the wave -fronts passes the place. The manner of The point here is rt that the two waves have fronts. is the sum of the effects due to the individual But in the closed cable. and initially two diffusive waves result.

the self-induction. though attenuating by leakage. the charge Q represents the result of an impulsive current pQ at y. which divides . but with the proper allowance made for the overlap ping. We in a distortionless circuit After the initial splitting. which is that due to the initial waves in reality. and the little jumps we remove now have continuous variation of the potential. the potential changes slowly (relatively) by the natural As time goes on. v becomes infinite. and the steady state is approximated to. This infinitesimally. disappear. is SV = Q/2/.of transit of a wave-front Between the jumps. The contrast with the behaviour is striking. If the mean value. But keeping to the diffusion problem. the two half-charges would go on travelling round and round at constant speed without any diffusion. the jumps decrease diffusive process. The initial waves therefore overlap inst antly.

on the right of y. the third is e^fy. and so on. in the initial wave . - . CH.2QZ That is . the factor for a circuital journey is e. j. the second term in Vj is .198 ELECTROMAGNETIC THEOBY. equally to the right and Therefore (33) represent the current and potential at x. V at x so far as the wave is to the right is conNow. To allow for overlapping. the same points distant Zl-x + y from it is y. The total is representing cerned. left. reckoned the other way round. VI.

the same point. V at x so far as the wave to the concerned. we obtain (37> where we recognise the unit impulsive function of Fourier's when we put t = 0. . That y = cosli7(Z-a. when So 1-c-sai represents -ty left is is. . Tiie complete potential V is V + V. x 2. theorem. + y) RpQ = coas(l-x + y) shin ijl sQ 28* 2q sin si Algebrising this by the expansion theorem. T here is only one such function instead of an infinite row because x and x %nl me annow.Otherwise /5 the same.

something. Passing now to a third minor matter.Theory of a Leak. and there So all that . 319. treatment. and by combinations of cables. Normal Systems. These cases can be constructed to any extent by means of intermediate conditions. But a whole book would be required for a are full more interesting matters to be considered. should be said about the tr eatment of cases in which there are two or more connected Fourier series concerned.

to be Let the V" cable be earthed at x = and I.. Therefore There is = (Y + Y 2 + Y 3)Vz 1 (38) z. and each current is the potential at z multiplied by the conductance operator of the path concerned.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. will be 199 done here will be to take two or three cases of the kind. to write down the determinantal equatio n. It may r. its resistance operator be any sort of leak. Say there denoted by is a leak at 2. For the sum of the currents leaving the point z is zero. is . and to exhibit matte rs of principle as well as practical working. to be suggestive of what is necessary in other cases. one thing that can be done at once viz.

when complete combination subsidence solutions. and the values of p satisfyregarded a lgebraically.the differential equation of the potential at where the three conductance operators are r If there are terminal K *). Y= 3 . are the values of p in the pt time function . Y and Y2 t must be suitably modified. r (39) arrangements at # = It follows that and I. Y satisfied 3 (40) is the general differential equation of the by all ing it.

a's tions are of the That is. the soluof the normal systems. The above process of finding the determinantal equation when any number of circu its of any kind meet at a But it is important to note that any point in the poin t. But the are functions of #. and require further investigation. the forms are intrinsically the factor tions. combination may be taken for origin. whose in containing a common vanishing determines the p's in the time funcsame . '. which may be simple at one pla ce and complicated at another but all . form 2 ae p where the /s are known. The difference may be applies very great in the form of the resulting equation.

(42) To find B. Then /. make continuity. Going further. Here v functions. 9 (41) on the left and right of Since z. consider the normal systems of potential.200 ELECTROMAGNETIC THEORY. VI. or v =w at z. we require x). V = ZAw. CH.m (43) iv . vanishes at x = and w are the normal and Z. Let V=2A V V .

si coss(lz}\ fi ~\ (45) or R _ sms2 + . we have the (44) final condition d-Ca = V/r at z. sin si . B\ sms(/ z)J . Applying this to a normal system makes sinsz r =--5 (cossz + smss -r --{). sms(t s x) ^. and A.= smsz sins(l-r-. z) Two things remain namely. To find 5.

and is equivalent to (40) above. The numerator is the mutual energy of Q and the normal system. that is.=A 0. Lastly. part for the leak is denoted by X. this requires the construction of A in the case of a point charge. to be found as before explained in special examples. itself. in a changed form. by the conjugate property. Q multiplied by the potential of the normal The denominator is twice the excess of the system at Q. system The . Say there is Q at y. Then. / tn\ (40) rs This is the determinantal equation. to find the A's to suit any initial state. subject to (39). electric over the magnetic energy of the co mplete normal The part for the cable is explicitly shown. less than z.

\ /7 ~ sin 2s(/ (48) z)\ \ ~~^ / This completes the problem so far as any to z is concerned. * / sin sz 2 W^-V z. f rom side of initial state of V . Then A A. by an integration. 201 But X vanishes when r becomes is a mere resistance.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. (2Q/S)Biny sin 2ss -*r on the other numerator.

. well. therefore. now exhibit a charge the more general way. provided that there is no allowance to be ma de for the leak itself. r. The same remarks about facility and the reservation apply in more complicated ca ses. detailed examination of This will be explained presently. on the other hand. Operational Solution. and also in evaluating A. get the operational solutions first. bad failing. It will be seen that there is considerable facility in obtaining the normal systems. 320. a nd other sorts of solutions as we We may. which may be a troublesome matter by the above But there is another way. safely deduce the general operational solutions. from the above normal syste ms. Theory of a Leak. But there is. If.is But when Q we must use w y instead of vv in the Else it is the same. a rather We cannot. nevertheless. we can at once deduce the Fourier series. whereby the evaluation of the energy difference is done by a differentiation without process.

we require 1 Next. Here (50) B is found by Co-C^Q (51) .A. Start at x = 0.Put Q at y.B. as there is to be V 8 = V + sin(a?-y). continuity in V at y. we have Vj-sinftr. and see the effect. First. (49) satisfying the terminal condition.

and (53) V= 2 at x I finds D. so (52) to be continuity in 2 V at z.y) Then there is . ELECTEOMAGNETIC THEORY. VI. CH.D.202 at y. and makes V _v3 V v V V 2 3Z -77 r-t \ sms(l-z) . and makes V^V^ sms(x . V = V3 + sins(^-c). sQ/S.

and makes V1= . Finally.where V means V 3 3f at x= I. sins/ S (56) B . we have the discon tinuity in the current at zt or V _?.sins* sms2 + This _sr .= ~ r B _ s V3 * sins(/-z)' This finds A definitely.

we may u se the modified V 8 in (54). which may in It is made impulsive general be any function of the time.x) ?Q V2 = . specially to suit the treatment of an initial state.S "*"<*-) rintt +g sins* sin s(l S z) (57) B There are two other for distinct V2 . . though not in the Fourier ser ies solution for an It will be found that V 3 differs from V l5 cominitial state . It will usually differ from Vi.sin s(l z) is the complete operational solution. merely in the interchange of x and y. in the usual . Finally. (57). They the 8 expansions. As regards V 8 it is found by (52). one for Vi and follow from (56). paring (52) with (56). and obtain h is . Q is lijp.sin. where an externally introduced source of current. sr sin s(l .-/ .

. so that D = is the deterrni- . by the expansion theorem. by D.V way. But. denoting the common denominator sin sz + ..

When fails. the active determinantal factor.2Qs sin sx sin sy QTo . The . the But the operational solution (56) remains true. ( 40). object of altering the form of the denominator of so as to make the function in the big brackets in (58) be (57). r being a rational function of p. r is expansion (58) A definite integral is required. this relation. be of any kind. and may therefore of s 2 . nantal equation. Evaluation of Energy in Normal Systems. According to (58).andV^S . as in (38). 321.PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. 203 of we can simplify the numerators by use Then the expansion theorem makes V. was to introduce a point not noticed in the previous. -Cotsz + -cots(l-z)} sz-( + R / K ds\r ---y (58) The form of the denominator was first Similarly for V2 changed in order to bring the relation (40) above into view. The leak a mere resistance. not a rational function of^. It may be verified that (58) agrees with (48 ) when the leak is But (58) is far more general. sin --j-^ .

Now the general f ormula to find from the conjugate proLet there be selected any two normal perty goes thus. belonging to p l and electric. let U .e.. i. .2. say. the sum of the conductance operators of the three paths meeting there.. p.the value of the coefficient A is given by d /I r v (59) ' sm This is . A systems. 2 szx Us -2s (Is (\r the same as (60) if Yz is the conductance operator at the point z.

From which it follows that (62) A.12 be their mutual and T 12 their mutual magnetic energy. =H .1^ = la expresses the conjugate property. Then (61) U .

but. which is here the first one. keeping to one normal system. represents the denominator in (62). and reckoned at the place of v. i f the initial state involved magnetic energy in the leak r. wh ich may confuse. (64) dp Y is this property is general . The numerator in (60) agrees with (62). the coefficient for the term involving p^ if U 01 is the mutual and T 01 the mut ual magnetic energy of the given . On-*n-?* for 168) any normal system. is CH. may be any point in the system. initial state Tn are double the electric and the normal state in question. (60). Y . VI. electric. because of the confinement of Q to a point. Now. z v its potential there. We need not use the suffixes. write U-T = * understanding that 2 . The denominator of That is. There would be.ELECTROMAGNETIC THEORY. and because there is no mutual magnetic energy. theref ore. of course. whilst U u and and magnetic energy of the norm al state itself.

divide Y into its three members . To illustrate. c's. The variations in v will be exactly compensated by those in Y. then (65) TJ-T = ^(Y + Y 2 + Y 3 ).the corresponding conductance operator. Y x is 2 for the second. the ratio (conductance operator) c/v for the first Y3 for the third. all with th e same v. but with different And c d ir = v dc dp dv . 1 djp Here Y path.

c = 9 dv .. But the last two c's may . z that is. tdp v dp dp c (66) We may therefore write -(U-T) = if Zj. dp dp (07) At the point terms of the are the resistance operators. ^ dp + rffi + . first.H''p. there are three c's and Z's to be allowed for. at the leak. &c.

be expressed in bringing us to a result of the form .

205 If. It will somewhat simple case of terminal earth connection and no intermediate Zl Z2 condition. though Take then the abstruse. two Y's. Then _(U-T) = c'^. and dY/dp a permittance be as well to illustrate these remarkable. where Z and left dZ/dj> is is the sum of the resistance operators to right of the point in question. which may be any point. or two Z's. proper ties explicitly.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. an inductance operator. dp (68). operator. any other point in the cable be taken. and the c's are equal. so that the v normal system c is simply (69)* . however. there are only two c's.

on the spot.I S cos si COS SZ COS 8(1 . Also d/dp= so dZ dp because R .A cos sx. Z.are the resistance operators to left and right of the point and their sum is . Zj-^tanaz.= sin sx. since 2 RSp = -s . s s Here (10) zt Z a = 5 tan *(*-*). = . n s = having any value making sin si 0.Z) to e (71)so that e/Zi would be the current there due (RS/2s>//f/s.

. if there is initially at y. Therefore C 2 rfZ ^ = s 2 2 SZ K ^Q = SI 2it .2 S/ R 2 Sl 2 1 2 cos sz 2s2 cos 55 cos s(2 R 2 2s sin si = 0. . expands to Consequently.

v = y Qsinsysin&r .

independent of the Here Y is the sum of the two con ductance and right of *. If the evaluation of The practically useful point is this. to be right. VI. and it goes simply. Of course the best place resistance operator is for * is Then only one concerned. Here "" dp I 2s ds\n~~ K" 7 Vj which also position of z. .206 which we know at a terminal. do in terms of the conductance operators. ELECTROMAGNETIC THEORY. it Again. CH. operators to left works out to the result JSZ.

To be very particu lar. It if the may and assuming be objected in the above that in taking v = sins#. by integration when there is we can avoid it troublesome. though Asin& be the normal system of potenti al. constant in a certain sense. by the way). We do not. But this will make no difference in the end. need to form the operational solutions and apply the expansion theorem. we may let rather superficial. size though and any dimensions. which is particularly the externally connected energy to be allowed is altogether by means of the identity (64). except what may be contained in the actual verification in the examples used. choosing the point of reference so as to make the work as nimple as possible. and the . A more important objection is that I have given no proof of the general properti es used. only object is to develop the proper Fourier series. we The objection is valid. A being the same and so have a symbol to put the dimensions in (though they are not known.U-T case for. and as above. may have For any unity. are violating dimensional properties. v to be potential. Or we may let asinsx be the normal system. and will complicate the form ulas somewhat. therefore. expansion theorem. which is suffiBut the fact is. that the cient as far as they are concerne d. let a be represented by a unit factor. or the tra nsverse voltage.

and the ways of working it. . will be considered separately. out of the general equations of dynamics.equivalence of the integration conjugate property. differentiation and the in their application to electromagnetics. which. are best proved in a general manner. being general matters.

Initial States in Combinations. or the sum of the conductance operators of the paths open to the impressed is the connection between Yy is current.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. Let the current be A h. if 7* and V the resulting potential at y the conductance operator there. 207 of regarding the matter is useful in combinations. If The path . and or its point of entry be denoted Then = Y. It has the advantage of compa ctness.V. The following way upon by y. * rt (76) y. Imagine any combination to be disturbed by the impression of a current 322. it. f V .

But to represent an initial state of charge or electrification. where Q is constant. Then V. combination. =2 * + steady term. = Yy is f Xy is (77) expresses the potential at y due to the impulsive entry of the charge Q . known. This is exceptional. and its expansion V. But . the accent meaning differentia There is a stead y term if _p/ Y y is finite when p = 0. Xow as in to find Va . (76) may be algebrised as it stands.of entry is not counted. ranging over the roots of Y = 0. we require to multiply the potential at some other point x in the Vy by the operator VZ/VV . let h=pQ. (78) tion to p.

. such as the general term of the if ?/ expansion refers we must multiply by ux/uyt is the normal function. state.in a normal to. This makes Vx = 2 Xow. the coefficients 5^ ul * / t pt = 2 Awx * (80) A are subject to the conjugate property.

and determined by denominator is (62). VI. CH. It is fixed represents an integration extended over the whole when u is fixed at any point. So the V. The numerator Qi^ Or agrees. so the position of Q.-Stf But 1 . nothing to do with the value of fore the alternative form of U-T y. (81) UT system.208 ELECTROMAGNETIC THEORY. U T. Thereshown that in (80) must be independent of the position of . and has> y.

It may. . magnetic energy. further. as that will require a compensating infinity. and Y g the conductance operator there.-S-^. have various forms of as a single function . Example should supplement precept. 188) z. Two 323. An integration over the whole combination will determine If the value of A so fa r as it depends upon initial charge. we have the equivalent form V. with an insertion. indeed. precede the above to the case of two cables of differen t types joined V x=Q "NV ..-SL-*-. where (82) z is any point. let us apply it. by (66).V.) Lastly. if t here are only two paths meeting at 2. (G8). correspondThis form isperhaps more generally useful. wh ich may be better still. there is initial but that algebraical expression in different parts of the combination. it must be allowed for It is important to think of the normal f unction u as in (62). (But do not choose a pla ce where u vanishes. Cables with different Constants in sequence. is of no conseque nce in the present connection. if it does not Therefore. where cf is the normal current at ing to uz and Z2 the resistance operator there.

resistance operator of the insertion (containing Let r be the no leak). which . with an intermediate insertion.x=z x=l in sequence.

z. is 209 made at the distance z from x = 0. e/Z is the current at z due to e there.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. If the normal potential v in the first cable and w in the . Here Z such that Now put Q at y. less than is function u of the last section z. is The resistance operator of the combination. reckoned at Z = r + 5S1 1 tan v+ * tan s 2 (l z). (84) is S2 the cables having different E and S.

iv = a sin s. w. whilst Q refers to y. 2 . V and W refer to the variable point a?. We cable is have next to specify the nature of v. and t he second at x = I. v.second. we have (85) V-Zf*. Since the first earthed at x = 0. then (86) In these. and the resulting potentials are V and W. w. But if W = 2u* Q is in the second cable. we may put v = sin $!#.

it remains to Since we know Z at the point z . 2 way is by . is to be at "" < 88 > This makes .(l x) . This is. the for it.--f*_S!** coss RjS-2 2 (Z rfn*(l-*> z) (89) The normal function being complete. done by the continuity of the cur rent at z that .c Z'. Thus sinssins evaluate shortest U-T is / ( -1 cos s : z Z } (90) VB. the potential at x due to ) dp . (87) provided a is settled so as to harmonise them.

The summation ranges There is no steady term.at ?/. because of over the roots of Z = 0. Z being given by (84). p . n. Q when both x and y are in the first cable. VOL.

210 ELECTROMAGNETIC THEORY. CH. follow (85). If y is in the second cable and In short. tiation to For d/dp may be substituted differencome if for trigonometrical convenience. use wx and wy in the numerator. we use the resistance operator at the place. need to allow for that unless there is initial energy in r itself. and Z is the steady resistance. x in the first. It is the ratio of the normal current fun ctions at x and where Z is C at zt and is If e is at . To find other place x. introduce the factor cx/cz in the general any term. use iv y and vx . we must put w for v in (90). and so we Sj to the fully-developed formula. If x and y are in the second c able. Thus. If x is in the second cable. then as in (84). (86). VI. But we do not required. the earth connection. when e Z. su itable for calculation or s2 In general r is a function of p. y being still in the first. If the effect of a localised may is at impressed voltage is wanted.

the case under the sa me formula by treating it as a cable (the second one) with a terminal arrangemen t Z 2 = _J 1 tan S-.known. say Z 1? given by 2 Z^r + 5: S tans 2 (Z-*). a different Z is needed. x = 0. (92) 2 It is we may bring then a special case of (59). (93) It may also be noted that the same formula allows us to .Z. if e is at x = l. We may regard it as the resistance operator of the first cable with a terminal arrangement. 291. Again.

1 y Z . a of cable 1 with Z 2 at its end.construct the resistance operator of any number of cables of the same or differe nt types put in sequence. It is R . we know the resistance operator 1 1 at i c 3 6 2 a. with intermediate insertions. For instance. Call it Z2 . (98x) s (Rcos-Zjssin^ .

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. the recip rocal of the sum of their reciprocals. 211 using the proper B and s for the first cable. without essenBut leaks require a little change of tr eatment. come If to the next leak. is a leak at a. say at we may proceed as before b. its end Using this modified Z 2 . then Z2 is to be the resultant resistance operator of the leak and that of the cable 1 with Z x at that is. Say there . which soon assumes gigantic size. we know the resistance operator at b of cable 2 Then we know the resistance with Z 2 at its end. Then. the resistance operator at c of the complete combination. and results in a continued fraction. we know operator at c of cable 3 with Z 3 at its end. an intermediate resistance operator wanted. where we make a is till we similar modification. Call it Z 3 That is. tial change. by the same formula. There may be intermediate insertions {without leaks). to be got by the s . This process may be co ntinued to any extent. that is merely the sum of the resistance operators to right and left. .

for simplicity an ordinary cable. Cable involving the zeroth Bessel function. Still keeping to two cables in sequence. we can continuously deform the simple normal function (sin sx + a cos s x) so as to make it assume the shape of any other kind of normal function. we need only let the conduct ance and permittance per unit length of There cable both vary directly as the di stance from x = Q. 324. which applies when the cables are not each of uniform resistance an d permittance. it may be observed that by causing B and S to change gradually from uniformity of distribution to various other arrangements of the same total resistance and perm ittance. series for the other. properly In order to . however. and when Of course. In passing.ame process. let one be and S varying in such a the othe r with R way harmonised. and a Bessel of illustration of broad principles. the results will be exhibited in two series. In ca se of two kinds of cables. the inductance and leakance are not ignored. say a Fo urier series for one cable. as to produce another kind of normal function. when B and S vary in any one cable. its resistance operator requires special investigation. have the zeroth Bessel normal function.

S=S (9 being constant. VI. S. (07)i where . Thus. dx unite to ---BpV. but this the simplest. . . are other ways. is CH. the circuital equations _^ = EC. Therefore. if 1= KS . dx (94) make the characteristic without assumption of constancy of R and *. we produce the characteristic dx sJV.212 ELECTBOMAGNETIC THEORY.

itself. But we do no t want that is The meaning V a. and I is is denned by the expansionThe corresponding C where Ij is the derivative of I of . clearly the potential at x given as a function of the time. V is a time function. since x=\ is In .=0 particularly.RS is the constant RS The proper solution for our purpose is (98)' where given. x fact.

is if e is impressed at A. our immediate purpose.the conductance finite zero at x = 0. We have (100) at x. Therefore. the resistance operator . to any there is and we should avoid the point x = Q no current there due for impressed voltage.

Or. we see that A) Io(<7 = 0. By (102). (102) from which C follows by a differentiation according to Ohm's law. That is. if there is earth on the other generally. or . The formula for the potential at x is simply got by (98) . /Z the current entering at A. Thus. = observing that it must reduce to e at x A. due to sources on the side of A beyond the cable in question. the current positive from \ to 0. if 213 we now reckon is e.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. side of more V/Z is the current at A when the potential there is V variable anyhow.

(103) the determinantal equation. function would be cos to Applying the expansion theorem to (102). Why we should be seen on remembering that there is no so that. is Z g*= -SQ Here J is an have it so will current at x = 0. if B and S were c onstant. oscillating function analogous to the cosine. the normal sx. when e is steady. .if J (* A) = 0. expands v= \ -<if Jj is the manner of This shows the negative of the derivative of J establishment of the final state .

325. and natures. We on can. Pass to the matter immediately in question namely. put on any terminal Z in place of earth the other side of e.of V. . of course. owing to the vanishing conductance at the far end. V=e all over. which is . the harmonisation of the solutions for two cables of different Let the cable on the left be of the regul ar kind. would be premature A Fourier and a Bessel Cable in sequence. and solve in a similar way. but that at present.

if that is the point of xQ vanishing conductance. x =o vT" We measure x from iX r left to z ' right in v Lc\ the in which the potential is V. is inserted any left cable. CH. and z from right to Between them. .. in the first ca ble. in the other. left combination r without leakage. at x=-l in one and = A. in the other cable. wh ere the potential is W.214 ELECTROMAGNETIC THEORY. but the z = Q end of the Earth is on at second cable is virtually insulated . operator at r is The complete resistance Z =r+tan si + _y ^. VI. that on the right be of the kind considered in the last section.

and evaluate U = sinsx. current Cx = subject to = + 21 (106> Z = 0. we harmonise v and obvious.(105) by (101). to distinguish it from x in the So the impressed voltage e at r p roduces tha first cable. and le aving them to themselves. with the form of Z shown by (105). iv. or v-w = rc. and c is . so v The formulae are (85) and (86). This being merely a special case. w = bIQ(qoz). except th at x is now 2. w fit (107) Thiswhere b has to be found so as to make may be done by v properly. let us examine the result of charging both cables initially in any given way. The form of v ia ic. provided . where x = l in v and z = A in ponding to v.T. the voltage equation at the junction. using the notation of the last section for the Bessel cable.

(110) . and their equalisation produces the determinantal equation (105) already found.(108) the current corresIt is also the current corresponding to w. Now (108) is the same as sin si H (q X) =-~ cos Q si. (109) which gives b} and makes -$4S ainrf + L'ooBrf .

subject to Z = i n (105).PURE DIFFUSION OF ELECTRIC DISPLACEMENT. Bessel cable. we may employ the -c 2 Z' form. except as regards the internal variation in r itself. When U. the initial state is given to be that the potential is . the tion. 215 This w is the other part of the normal system u. Substitute w. use is one and y in the other cable.wy . and therefore get. using (110) when they are in the and vxwy or v y w t when z or 2 is in Also. T. There is only left the evaluation of As before. for v and w together make it co mplete. the resistance operator at r being known. it being vxvy that is used in the numerator. U _ Qainaesimy ( when x and y are in the -first cable . which we do not wa nt. instead of Lofez).

where in the first form u it is whilst in the second cables. A normal system of potential is one which.coefficient of the : normal function got by an integraic thus A _/SUM dx _ /S Vt. when left to itself. with an y terminal arrangements. Th en. the complete normal function. siding. and any number of intermediate insertio ns or leaks. It should be regarded as a single function of position in the combination. and more particularly the case in which one cable is uniform whilst the other is of a variable type. imagine any number of cables put parts. any way of variation of resistance and permittance. pre serves its form when subTo find it. The above case of two cables of the ordinary kind having different uniform resistance and permittance. . separately exhibited for the two Construction of a Normal System in General. are meant to lead easily to a b road understanding of a normal distribution in general.dx +/S AY " dz U-T U-T ma . in sequence. no matter how many forms it m ay assume in different To illustrate this. 326. the initial is U becoming the initial V and W respectively. sections containing we may first divide the whole cable into no intermediate insertions or leaks.


and so on. is the values of *. say to a section stants. having a single characteristic. introducing all the necessary The result is a single normal cont inuities and discontinuities. represents a complete normal system at time t. represented by w x in the first section.216 if ELECTROMAGNETIC THEORY. The size of the normal then settled by their conjugate p . Thus Awep where A is a constant. Regarding p a s a constant. v come down and has its u iv the two solutions. It the conditions fur nish a general determinantal equation. we must subdivi de into smaller sections. u2 in the At the same time second. with its own s pecial v and w. The normal u may. the characteristic of the whole combination. say Z = 0. sponding to the different >'s. it may be treated b y itself. be continued over all the intermediate an d terminal insertions make it fully complete. They have next to be fitted together. but of arbitrary size. there u = av + bw. so far as the cable is concerned. R. VI. functions of x. It is are two equation (95) above. CH. This is to be done by the intermediate and the termi nal conditions. and p satisfying it are the values of p permissible. function u. a and b being any conindependent solutions. if we please. Every section of this kind. and therefore 2 A?6 tpt including all the norm al systems corre. But if there are two or more formulas So we concerned. represents the solution arising from any initial state. S vary in any section each according to a single formula.

said about these things in other electrical applications.roperty. tho ugh the latter does not occur In intermediate cases. The normal functions are always entirely real when energy alone is concerned. may be taken to be the conductan ce operator at any point. its detailed consideramay be omitted. to is combination contain no energy initially. in the above diffusive appl ications. a nd its influence on the rest allowed for This applies to parts of the cable a re sistance operator. and other f orms. whole combination. electric . and also when magnetic energy alone is concerned. all other fo rms being derivable from it. There is more to be the p's may be either real or complex. and leaks. which systems If any part of the leads to the formula (62). but the present remarks are suffi cient for the immediate purpose. whose vanishing settles the admissible rates of subsidence. specialised by regarding p as a constant. by tion itself as well as to insertions The characteristic of the and leaks.

though made up of many parts.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. before. provided the proper operators are employed to sum up their reaction upon the retained parts. We may use same symbols to indicate position. as a single system. must be continued over different all the branches. We still have to regard the normal system. as System. of any degree of compli cation. To make the normal system complete. or we may use the symbol with a different range. We should . it There is connections. by means of cross we convert the above arrangement of cables entirely in sequence into a network of cables. T hese things are immaterial. Construction of Operational Solutions in a Connected At the same time. contained. 327. selfany parts thereof may be left out. '211 no material alteration when.

as above specified. We have to find V (and C) all over due to e or say V = Xe. and these constitute the effective sources of disturbance. Let the electrical arrangement be th e same. . changing its form of expression as we regard pass over the combination. the latter in the current. function (in its expr ession) in different branches. Since e is at a definite point. but now consider how all to the operational solution expressing the effect over the combination due to impressed force at any point therein. it is a function of p. either a single circuit of cables or find a network. but since all the different forms have to be harmonised. It may be impressed voltage e. whilst V is anywhere. The former will create a discontinuity in the potential. We have here a difference from the se lf-sufficiency of a normal system.r. and varies in form as concerns p as well as . X is a fu nction of 7i . or impressed gaussage h. . where by energy is brought into the combination.now point out how a normal solution. It is not the same specifying position in the combination. In this respect it i s like the X But at the same time. There are likenesses and differences.y. the time differentiator. differs from what I have usually called the operational solu tion. because p enters along with the electrical constants. for example. The form of X is wanted. we may still as a whole.

normal function. .

if we wish a differentiator. It remains what it is fundamentally. V We to explicitly indicate that a and b are operated cart But this is a matter of co nvention. the a and b are not now c onstants. in the general solution of any section. To X fully. operator. So. upon by u and w. Here there is but The other is ja. A may be pulled or pushed. of = (v +jw) a. but differential operators. may write V = va + wb. we proceed as in the case of a normal function. to divide the combination into sections involving only one characteris tic. find CH. Another form is V one time function explicitly. = av + bw. and v. but time say functions. where j is an Now supposing we V for every one of .218 ELECTROMAGNETIC THEORY. But we do not assume that p is constant in the characteristic. VI. w are not merely algebraical functions.

and X by itself is not numerical. In the case of the nor mal system. for the constant^ It ia in it may have any one of a particular series of values . up to a monise process found the general solutions the sections. harmonising difference comes in when. say finally (though it may be done treat the section containing the impressed force. we know the tion. operator X completely Or. nothing special there in a nor mal solution. is made two sections. on the other hand.them by the is. . in another form. and a jump in V produced by e+ The result is that all the time functions in all the branches become known in terms of e. in V = X<?. as. It is one of a set. and there are two auxiliary coninitially) we There is ' ditions. But in The section the operational case we have an additional spli t. Apart from this. The operand e is any different in an operational solution. But the function of the time. we have next to harThis conditions at the junctions. throughout the combinaand there is nothing left over to be fixed by extra data. formally like that of The the different parts of a normal functio n. have certain point. the normal system constitutes a succession of values making it be a numerical function of x. continuity in C. its size is left quite indefinite.

of the type #*/(*) This brings us to . For if we put = 0. If it is.determinantal equation e is implicitly involved. then X-1 = finds special solutions. the latter is the general characteristic when free of the solutions from impressed force. producing = X^V. a rational equation. as in most of our applications.

from which the subsequent states arise .PURE DIFFUSION OF ELECTRIC DISPLACEMENT. more and strictly. voltage in an impulse at the moment t = 0. equivalent to an external source of current. The The greater generality of the form of X normal system cannot be more than it is. where P and Q when t are constants. by making p in it become It then also constant and assume one of the critical values. the total of the impulse being P. The other has to cover all possible cas es. 1 in normal functions whe n they are appropriate. constant functions of the time is when is positive. but the data consist in the spec ification of an initial state of V and C. viz. This brings us to the normal functions again from another point of view. Then e is merely the impressed negative. Similar remarks apply when the source is h. it will be seen that the operat or Y^tf/X" itself X may be regarded as expressing the type of the normal function (apart from size). when there are no i mpressed forces given whose effect is required. h is the impressed gaussage in zero an impulse of . is essential.. simplifies in form. imp ressed gaussage. From the examples given. There are two sorts. 219 I have shown how to expand the normal functions again.. and is constructed to do it. Let e = pP and h=pQ. Similarly. We convert this general problem to a special case of the former by the use of im pulses. or. in virtue of the det erminantal equation. viz.

But. P and Q are given. the left Q is instantly introduced. where L is the concrete in ductance connecting the momentum P and Similarly as regards P. and then to itself. included. is P = LC. That is. required to represent them. in general. because the assumptio n L = destroys the momentum. as before explained. magnetic momentum. that is. The analogue of Q = S V. when L is and h to obtain the proper The addition of the impuls ive e is important e . to unit length.charge total Q. e and h. and the expansions the current C. we do not require to know C at all. In our diffusion problems. we must use both expansions. if S and L belong Q and P are the length integrals o f SV and LC. Now the initial state is fully specified by V and C. e equivalently. That is. It represents But in the case of a cable. That is why only h = pQ has been employed in discus sing initial states. or which is the same thing. where S is the concrete permittance connecting the charge Q and the voltage V.

and from the operational standpoint. The most serious d rawback to the method of analysis into normal systems is that. Objections founded upo n want of rigour seem to be narrow-minded. ot her special ways of working have been given. CH. the general nature of normal modes and then may be quite unnecessary to resort to operational methods.220 ELECTROMAGNETIC THEORY. and are not important. will give unless passive indifference should be replaced by active In making them rigorist s make confesobstructiveness. and if we do not. involving the reduction of where Some V = Xe to V = (X + X^)0. operati onal ways are much the best in dealing with the effects of localised impressed f orces. but it is tional a great practical objection. though their use will often settle matters which are obscure without them. wanted e. and if so. the numerical examination of general results becomes impossible in any re asonable time. There are other be used. beyond a certain point. . usually the simply periodic. We may turn them to normal systems if we want to.. depends upon circumstances. or m odes of vibration. or of subsidence. and the study of such cases is often most instructive. We should go to the operasolutions for information.g . It may be that only very partial knowledge is methods. and I e is more later on. The discovery of practical methods of manipulating operators is a matter of importance to the future of physical analysis. and m ap out the field of knowledge. because it enables us to treat the electromagnetic problems with generality as r egards initial states. Connection with the Simply Periodic. VI. we can The most important is app ly other methods to them. This is of no consequence in theory. simply periodic and i is the differentiator d/d(nt). Remarks on Operational and Normal Solutions. 328. On the other hand. The object of the mathematical investigation of p hysical matters being to ascertain what happens under given circumstances. so to speak. or of time variation in more mixed manners. the method of normal functions. how it should end. is one of the most important ways of working to the desired But whether it shou ld be used. from the wide application possible.

It would be more useful for them to .sion of ignorance.

may be regarded as made up of a succession of infinitesimal impulses. Returning to the simply periodic solution just mentioned. of course. it may be worth while to point out two ways of connecting it with solutions in s eries of normal functions. and desire to express the end (or anywhere else) in terms of e may do this operationally. we see that the last solut ion in normal functions. try to extend the matter. by being used as the element of a time integral. we can convert the result to a it normal functions. is merely parenthetical. observing that if e is given variable in any way with the time.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. Now. given that series of e is impulsive. Say we come to C We = Ye. Next. Suppose we have a telegraph current line of C at the distant at its beginning. any sort. But this if they want to. . 221 and remove the want of rigour. enables us express toC when e is variable.

of all the know the simply periodic soluvaries anyhow periodic variations. but its executio n and is = ni. tion. (113) by The simply periodic solution for the element271. The process But if steps. (54). the impulsive voltage e=pE is cosntdn. C= AfY TT cos nt dn J o . is reversible. by analysing an impulse into simplyThus. by choosing e to be be equivalent to that which = Ye. by the transformation a very complicated way of getting the simply periodic solution. if we only obtained directly from the operational. The method described verification is quite feasible in some cases. the result must may be obtained directly from C l> Lastly. if not. there is another and muc h shorter way of passing to the solution when e we are in possession. equivalent to namely.simply periodic. So (E/TT) cos nt dn is ob tained by operating on it by Y.

= 2L f (YO + ^ A. the evaluation of the definite .} J o cos nt dn \ d nt/ (114) expresses C due to an impulsive voltage. by a time integration. as before. in general. we express C due to any varying e by the second method. But I am afraid t hat. in some cases. and therefore. by the evaluation of the integral. This process can also be practically carried out.

any more than the some years ago. it converse process should be noticed. and which proved to be relatively In connection with this matter. equivalent. integration. CH. . It may The resulting integral may not be or may be equivalent only within a certain ran ge. Considering the extreme complication introduced into normal sys tems by the inclusion of gradual propagation into the wires as well as along the m. integral will present the process (114) seems at first sight rather enticing. which had its foundation in the physical ideas involved. much difficulty. This But the transformation from C = YpE be done to (114) requires to be the case that the physics is such as to show that the process is legitimate. So i t is. the integral does not tell us So I found operational solution itsel f. r w ith caution. when I desired to trace the progress of an ela stic wave along a resisting wire. viz. may then present itself. with a view to their is evaluation by any means that often a very useful practice. But the operati onal solution itself was amenable to treatment by a specially-invented process. But the legitimacy cannot be asserted when Y is an arbitrary operator. the transformation of definite integra ls to operational form.. It looks so simple. That there should be failures in this application of (113) will of unknown meaning. The integral was unmanageable. VI. the simple in execution. in idea but if we cannot effect the .222 ELECTROMAGNETIC THEORY.

which has been principally useful as a convenient basis upon which to rest the theory of x=0 diffusion. however. 329. x=\ The application. particularly if it is infinite itself there. s ince the function may . There is little more to be said about the submarine cable without self-induction. Two Bessel Cables in Sequence. by considering the application or misapplication of Fourier's theorem in the definite integral form (55). at least partly.be understood. A Bessel Cable with one Terminal condition. 271. of the zeroth Bessel be continued in a generalised form. to a function whi ch does not vanish at infinity.

223 Bessel functions are of great importance in mathematical physics. Then we make a single cable ABC. is choose Z x to be another Bessel cable exactly like the but turned the other way. Take the problem of 324 . w hose conductance and permittance are greatest at B. and initial state is it follows that if the Q at the point y. Z (115). and Z=^0 being the determinantal Now first.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. The resistance operator at the point A. applied to the present case. and linearly to zero at left fall off A . 322. but now put a terminal arrangement Zj at jc = A. is now (115) by (101). the resulting potential is by the formula being given by equation. using the same notation. instead of having a short-cir cuit there. (83). and have many applications in electromagnetism. so that the zero conductance at the far end.

because it charges the t wo cables symmetrically but oppositely. n. the right one being operator at B is twice that of either cable.and C. that result is erroneous. But it on consideration that Z be used in (116). or solution (16) relates to the The resistance uncharged. for the cables may be charged . The one. so that V = at B when e is removed. But it cannot be right in general. For every norma l system then has a node at B. This is such that e/Z is clear is the current produced by if this e at the point B. if limited to the roots of = 0. is perfectly right lo(^) This as regards the normal systems concerned when e is at B.

we requi re another condition giving nodes to the current at B in a set of normal systems. approximately In those of the sum of the operators and of their difference. with current nodes at as the other term. the Ii function being the derivative of I is solely . but not quite. and the second lot to I 1 ( ^ A)=0. with potential nodes at B. CH. VI. Besides IQ (q Q \) = 0. say same B. the by the length of the right cable different from that of the left one. if in (115) we Z1 be nearly. the first lot lead to Io(q \) = 0. the limit. symmetrically so as to produce no current at B.224 ELECTROMAGNETIC THEORY. it will be being slightly found that the roots of Z divide into two sets. let In fact. The first set of is normal systems concerned when the initial state .

representing may initial c vanishes at the point B for all the missing normal functions. The reader has a lready been cautioned not to choose a point where v vanishes in the alter2 nativ e form v Y. . and and V We K R instead of B . This result is on the left side.such as to produce no potential at B. course.c 2 Z'. require. find the resulting in the usual way.. y interchanged serves is explicit in making when x and the same expression with x. It peculiar cases that th e peculiar be noticed that in the denominator in (116). It is a necessity for the validity of the expansion of an arbitrary state in normal systems that every possible normal system should be included. (^) . the quantity there In the treatment of this problem by the operational solution. on the right side of y . of to use the second solution. and the second set when such as to produce no current there. The result is. Put Q at p in the left cable. there would be no failure. betweenB and A or C. (118) at x. be the determinantal equation when . I have introduced this example to illustrate the need of caut ion in the treatment of normal systems in peculiar cases. along with I (qx). Bu t it usually happens in these results can be foreseen. is no room for hesitation. writing q instead of qQt for convenience. If we chose Z at some other point.

(118) is expanded in a. .

The K (2z) appears altogether. The n cresult is v__9_ S/Y \ Q total permittance' > 2 the permittance per unit length is S# in the first cable. Passing now to the general case involving the zeroth Bessel functions of bo th kinds we can obtain our results very . so it is only needful to recor d the results as above. But General Solutions for Sources in a Bessel Cable with two Terminal conditions. theorem. and symmetrically in t he second. and shows the two sets. then disDeveloping (11 8) by the expansion v 2~ v 0~^ y V V first summation is subject to I (^A) = 0. and the The accent means differentiati on to qX. series of the 225 normal functions I function. 330. though required in general. we get (gx). The change to the oscillating 2 functions J G (SJJ) and J^s-r) by mea ns of s is obvious.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. second to IjfeA) = 0. There is a steady term V because there is no ear th connection. in the operational It is obtained as usual by making p = where the solution (118). if f= this is a special case of the next investigation.

/ x dx V and C being the voltage (transverse) and current at z.Let shortly and symmetrically by the operational method. terminal arrangements Z and Z x be put on at the points # = A and I. Q . VOL. in a cable subject to the circuita l laws ' (/. L. II. and are preferentially of the forms Y-E + Sp. (2) R. where the resistance an d leakance operators per unit length are Z/# and Y. and S being constants. leakance and permittance It is merely the introduction of var iation per unit length. being the values at the point x = \ of the resis tance. inductance.r. Here Z and Y are independe nt of x. K. Z = R + L^.

Zp V. VI. is x=\ xl ^ characteristic of V. CH.226 ELECTROMAGNETIC THEORY. obtained by eliminating C from Ij? x dx Later on. according to The two solutions of the characteristic are ') 2 . with x in the manner specified that introduces Bessel functions. ' V2 y . q. ^ dx 2 = YZV = 2 2 V. Zl ^ The (1). (3) we can those assumed for Y give different meanings to and Z.

(qxY 2242 6 . solutions directly or i ndirectly. follows that they are sufficient to express all That is. j being . (ga) . easily it characteristic is That these are solutions of the and since they are not in verified .2 where 7 = 0*5772 K. is (5) A and. V={I (^)-/K where (^)}A. qx is the general solution is of the characteristic.2. 2 2 4'2 6 2 2*4*6* * . Q(qx) +(1 is a certain constant introduced to make vanish at infinity./A are any time functions. constant ratio.

howBoth I and K are ever.any operator. finite . is on the left. Now two forms. always positive when qx is positive. and V 2 when on the right . so we have point y. not finite. When like e qx . It is. numerical and positively real the I function at the origin and increasing continuously to The K function. say (6) (7) where V\ holds when x side of y. but infinite at the origin. infinity like <rqx in falling continuously to zero at infinity. on the other hand. is at infinity . construct the operational solutions for e and h at the The characteristic fails at that point.

as due to the source e. l. (7) (8) at the point y.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. These conditions. Say V=Z C 1 at x= ft. we have e = V 2 -V1} C^Cj. and produce v_ ~ iy ~ aK j y) which are quite general. (6). V=-Z C at z = 0. The operators a and /3 enable u s to introduce terminal conditions. In the case of 227 e. applied to determine A and B. (11) These determine a and .

^ and K! are the derivatives of ID and K .(<?AZ /Z)KlX Equations (9). 5 KoV . and may be developed in Bessel series. ' (10) are now made fully explicit in terms of the terminal operators. 1 . and. I (qx) is denoted by 1^. As regards the no tation. to ease matters. V =V A and B. and but instead of (8) similarly for the rest. we 25 still use (6).thus. (7). When h is the source. we have the conditions These find at y.

an d the e solutions those of current. Q2 . them use was made of the identity =-. and make * (13) V. The determinantal equation of normal systems = /? The h (16) solutions are required to represent initial states of In obtaining potential.Ca-Cx-ft. TTZ (17) which often turns up in these investigations. = fr-Zfe " y " (14) _" V= The expressions for a and J3 are the same as is before.

is an example. N Now let vlt (?! and r. per unit length of circuit. since d . time. It is often to be observed that a property.228 ELECTROMAGNETIC THEORY.-. as well The characteristic of V is functions of x. of V and G. CH. K dx p dx be any two solutions subject to (18). VI. (18) Here p and K are. .2 Then. if it be really a general property.dC .2) c. Thus. They may be any functions of x. admits of simple proof. n T7 Q .differentiator. It is a particular case of the known conjugate relation between t he two solutions of the ordinary The proof is simplest in terms equation of the second order. functions of p the when variable states are c oncerned. Conjugate Property of Voltage and Gaussage Solutions.<A tion dx be the connections. first noticed in a special case . as More generally they are operators. \ dx =K\. then in others. = PC. in the simplest case. let r . 331. the resistance of the wires and the cond uctance of the insulator. . Pe rhaps the more The last equageneral the property the simpler the proof.

K and p become identical functions of x in the last two equations. This is the conjugate property. (22) and consequently where m is constant. . and the same constant in the two systems. by (18). M wc^ = = ( (20) __i dx . Therefore djo l(^2 .Vl) = 0. _d_ Similarly.dc9 di\ we have./l) Now if we make p be a constant. so that the right m embers are the same.

or combinations I (<?2) and so on to other kinds. c ross products are equal. When easy to generalise the property by extension to space distributions of electric and magnetic force. r's only. making v=v 1 + vst and G C! + c.^. . the sum of the leakages from tyi. limits us any two solutions of the ordinary equation which the characteristic (19) constant. or the = activity per unit length of circuit. o r combinations In becomes. the best way is it to be real. p is imaginary. v. to That p is to be the same constant in the two states. the systems are real. and to It is imagine their energy. For example. p is replaced by a and e"9*. and K (<?z). but we do not want that at present. &c. 229 If the two systems coexist. and its rate of decrease with x is its leakage. It consists of .PURE DIFFUSION OF ELECTRIC DISPLACEMENT. for the sake of the electrical argument. the product vc is the total energy flux.2 be the voltage and current. i\c% and r^ and the reciprocal property in the form (22) asserts that the parts depending on When p is real.

it is <*4 -</! = "< (25) Thus (24) relates to the two solutions of (19). = p. when the differentiator terms of the the conjugate property *v'2 is *y-. and . (24) to where the accent indicates differentiation of the c's x . and.t qx . . in terms only.

instead of i\ and i? 2. 332. ever. By means of the conjugate property we are enabled It to obtain the operational solutions for sources e and h. of a limited nature. Operational Solutions for Sources in the General Case. m vanishes. another can be deduced by the conjugate property. howwhen one is given. will be convenient to d enote two independent solutions by u and tr. since. The independence is.(25) to those of the corresponding characteristic for c. When not in constant ratio. they are said to be independent. so that the property is u iv fc u . If the t wo v solutions taken specially are in constant ratio. of course. which is not the same.

(26) .= mp.

which makes C^C.230 ELECTROMAGNETIC THEORY. 2 (27) where Vj is on the left side of the source and V2 on the right. These are operat ional. a and A and B unknown time-functions.pwxjB. ~_fei h m(a-ft) V^k^'^fa-^'lh. VI. so two forms of solution for the potential are required. we similarly pr oduce the results V\! --V= The value K~ alc *) . When the source is e at y. Now there is a discontinuity at a source. conjugate property (26). V = (u t . p meaning the time-differentiator.Ci at y. using the . ft unknown operators. This makes V l = V 2 and Specialise the source to be h at y. we produce (28) u V = i *^?*l&. Applying these conditions to (27).aw x) A.^ and e = Va Vj at y. CH. say Vx = (MS . m(a-p) /3 (29) which are explicit when a and are known. h = C 2 .

Only the commencement of the series need be used. depends upon the standardisacan be found as soon as they are given. the operational s olutions are complete for the region between X and I. Say V/C = determined by V/C at any point on the /3 similarly by V/C at any point on the Z at A. and = Zj at I. by inserting them in the conjugacy equation. The operator a left of is the source. and right.w. . That is to say.----mp(a # K ~ -f ir) et ro> ("<-') ft K mp(a-() of the constant m It tion of the functions u. then /QON of Z and Z x we allow for the reaction of all the parts of the electrical system beyond the limits upon the part between them. so far as the source e or h is By means .

to be 231 is The conversion to series of normal solutions done in the usual way. by means of certain equivalent transforma . It is constant as regards x. h=pQ . Comparing the present investigation with that in 330. which most of the fundamental formulae involving the zeroth Bessel functions of both kinds. Cylindrical Waves. we see that u is represented by I (qx) and w by K^a?).PURE DIFFUSION OF ELECTRIC DISPLACEMENT. concerned. and initial states of potential and current bye=joP. Conversion of Wire Waves to Two Ways. Returning to the Bessel investigation of 330. It contains p (or may do so). contains 333. and that the constant m is 2/7rZ. as before described.

We may conveniently take the tube to be of square section for plane waves. circumstances the internal plane wa ves are made self-. desirable to point out that the formul ae are not merely those concerned in the theory of a telegraph line of a variabl e results. Then the lines of force are straight. Moreover. But in the present applic ation. and the magnetic induction upon perfect magnetic conductors forming the other two sides of the tu be. I. with lateral spreading prevented. in the transmission of plane electromagnetic waves through a conducting dielec tric. by proper interpretation of symbols. but belong to cylindrical electromagnetic waves in general. they do not do so by mere analogy. of no practical use. any tube of energy flux may be isolated from the rest and made independent . ordinary nature. a beam of light. Vol.. whilst the two forces cross one Under these perpendicularly within the tube. Also. But before doing that. It increases in section regularly as we . guided by the tube. and the magnetic force to the two electric conducto rs. another The internal state is like supporting. I have shown in 206. the electric force is tangential to the two magnetic conductors.tions we can develop the whole body of it is and extraordinary. Thus. but. the two theories may be seen to be one and the same essentially. The electric displ acement in the tube must terminate perpendicularly upon perfect electric conduct ors forming two sides of the tube. the tube must be imagined to be of the shape of a wedge. by enclosing it in a conducting tube made of two materials. how .

The enlarged meaning of potential. VI. with an being the transverse ga ussage. or be perpendicular The E lines must then circular. but two of its whilst the other two separate. the intensities of E and H. each the same at circumstances Also. along its length. Under these E we shall have propagation of electromagnetic waves in the tube according to the formulae given. For. at the tube in section and AC Now let straight lines of to the plane of the paper. centred at A. flat CH. consider how the inductance and permittance of the x depth of tube. and may be called simpl y the sides. let H join the sides. and bottom. w hilst the others are parallel to the plane of the paper. and be diagram. tube va ry. sides represent the traces of the separating sides. to be called the top and bottom. as at join the top in the E and H. AB Looking remain equidistant. is E x by length of arc con- . be the same distance from A.232 ELECTROMAGNETIC THEORY. The current C is the same as " potential" V. pass along it.

because the section of the corresponding tube of displace ment is constant. the permittance is constant. The circuital equations are therefore -_=L^C. whilst its length varies as x. -vr Tx /oo\ (33) length of tube. the values of the The above make s no allowance for any conductivity within the tube. S/o? L# and are the inductance L Next. let the medium in the tube be uniformly conducting. Then the magnetic conductance will vary like L. same at # = 1. and permittance per unit and S being constants. length because the section of the tube of induct ion belonging to length unit length of the tube of energy flux varies as x. or the transverse voltage.H The inductance (per unit cerned. the distance from A. both electrically and magnetically. if d\ -r r\ S . that is.dC --pV. of tube) varies dire ctly as x. varies inversely as x. which is itself merely a gu ide. as x\ and the electrio . whilst its On the other hand .

We shall therefore make the circuital equations take the form -JJ_(B if + L/. and the distance x not too great.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. K The depth of the wedge is immaterial. now. The magnetic flux is endless. If we remove R from the inside of the tube. If we remove K too. and substitute equivalent electric resistance in its top and bottom . equivalent to being circuital.)*0. But since K no use doing that. in addition R is zero. or as arl . This is also the the angle of the wedge becomes immaterial. we must substitute equivalent magnetic resistance in the other pair pair. (34) Ex and I\/x are the unit length of tube. or the part relating to the transformation of the circ uital equations in general to the special forms for plane waves along wires. magnetic and electric conductances per So far is an exact theory. The reader should study Chapter IV. It is the same theory now. It is now the theory of waves along a pair of finitely resisting electric conductors through an electrically conducting dielec tric medium. conductors. Make it infinite. 233 conductance will vary like S. it is Therefore r etain within the tube. of. and the interpretation of the quantity called magnetic conductivity. the side expresses a really existent property. . -fg = (K + S/>)lv. only appli ed to a tube of energy flux of variable section.. The wedge now consists of two planes (top and bottom) inclined at an angle. and do away with the sid e magnetic conductors altogether. which is If. the equations remain approximately true when the angle of the wedge is small.

case flux. a conducting dielectric. as we please. deal with a finite depth along the . when R is not zero. we have merely to the permittance as x. We may th en remove the top and bottom The result is complete cylindrical waves in plates altogether. the electric an d magnetic forces in the above. the conductivity being of either kind or of both kinds. The voltage V is now circuital. and finite. on the other hand. however. We need only. is infinite. when C will be finite too. interchange . The measu re of C. Increase the angle until it becomes 360deg. But in 330 the resistance of the wires varied as x~l and To obtain this case. axis. provided it be in the tube of energy and not at top and bottom.

So. by letting the The process employed is not confined in circuit be variable. Spherical waves may be simila rly treated. The same allowing for the change in position of the electric and magnetic forces. but w e must not H run away from the matter immediately in hand. and be circ ular. VI. The electric force must be straight and perpendicular to the paper. The gaussage C is now reckoned once round the circle of magnetic force. the sides of the tube must be perfect electric conbegin ductors. CH. we may partially discus s some of the operational results. along of this case is much simpler than the other. 334. leaving complete cylindrical waves in a homogeneous dielectric. Notice first that the results for e in . and the top and bottom perfect magnetic conductors. To make V finite we may consi der only a depth along the axis. Both E and will be pairs of sides separating. its application to cylindrical waves.234 ELECTROMAGNETIC THEORY. figure. the theory arcs of circle s. conducting in either or both ways. by employing a tube whose section varies as finite We #2 both . the quanti ties K and S will vary as x. centred at the axis. In fact. Also. The magnetic Then. After the enlargement of ideas of the last section. force must be in the plane of the paper in the and parallel thereto. t he bounding conductors may be wholly removed. Special Cases of Zeroth Bessel Solutions. see that the problem of a telegraph circuit has real appli cation to electromagnetic waves of much greater gene" constants" of the rality t han plane waves. if the medium in the tube is uniformly conducting in both ways. substitutions of resistance in the bounding conductors for the internal conducta nce may be made as before. and the 1 quantities R and L as or . to with.

Interchange x and y when V a is wanted. Thus. take equatio n (14). or h.830 are produced from those for by substituting eqy(d/dqy) for Z/t. We may keep to the h set. at present Q*\ a-fl giving Vi at (35) x<y due to h at y. .

K oz is zero. if /X = 0.. If / 235 This (36) is infinity.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. by (12). Again. and p is infinite. reduces (35) to -aKJKo. So (37) . and disappears. by (12). K oX is infinite. We see that the terminal condition at I becomes impotent. and <x = 0.

. of is a source at the origin. .. and . K^. in the region reaching t o the origin. weobtain infinite results at the origin wit hout using K^. there is no doubt that the K function is excluded a source there. That is. unless there is a source at the origin. as we saw above. because to do so would make V infinite there. when A = 0. becomes impotent and Also note that disappears by shifting it to the origin. It is usual to consider that we cannot have t he K function hi solutions extending to the origin (the axis of the cylinder in cylindrical waves )._ V-(. unless there making a. which is given by . X ' .AZ be finite /Z)I. unless there can be something very p eculiar about the Z operator. Nevertheless. does not appear. That thisreason is inadequate and unsatisfactory will be evident if it be allowed that V can be infinite at the origin without havingwhich I will give examples.Thus the terminal condition at A.

I If both X = = oo . and (39) which represent an important fundamental solution. (40) If y = A = 0. p = oc . Both terminal conditions are impotent. If y = A. and (I x-aKoX). we have only the V 2 solution under control. we reduce the last to (41) . then a = 0.

(41) reduces to 0. is CH. and the Z condition is I = oo . /? = oo and . with the source at the tion alone if (42) origin.' we have the K functhere . VI. V = j7rZAK Thus. Finally. The source impotent.236 ELECTROMAGNETIC THEORY.c . if at the origin.

because of the vanishing conductance there. Similarly as regards Z By (9). o n the other hand. reflection. 2 In the last case the finite e produces no disturbance beyond the origin. x (43) (44) A= makes ^=oo. In That it needs (42).A = </ make make (45) (46) =A= V = 0. V = ^qyel^Kiy.is no e. = oo makes V^^qye^-aK^K^ V1== J *qyeI0x (Kly -/^I^. but the I function as well when there is reflection at a finite distance. h is given to be finite. as in (41). infinite .

stant. without Thus change of for m.. In these direct numerical applications. to make Z and Zj represent the steady resistances terminally at A and I. which is conThen 1^. Putting p = makes <f = EK. Numerical Interpretation of Formulae. are all algebraical functions and are is given. So. qx is . when a and /? are concerned. The Divergent Series. knowing the general nat ure of . when the source gives rise to a steady s tate we can use the above at once. But they are strictly numerical results in certain cases. K we can see by inspection in some cases the nature of the steady states on both sides of the source. of course. The above results are usually operational. We have. use the formulae (4) when big. or perhaps with only nominal change. 335. numerical when x the curves I and &c.impressed voltage is immaterial.

Then use But they are quite unsuitable the approximate formulae when qx is .small.

and ignore the rest. enough to introduce a large error. convergence. which is at a distance along the series when qxis big. that we are obliged The difference between to go t o the convergent formulae (4). only one of d egree. the smallest term. In the convergent series. use the formulae (48) (49) The first terms represent (47). or the place wher e The error is limited by the size of the smallest term occurs. 237 which show plainly how the functions behave ultimately. comes to the beginning of the series. and. especially if half the The above are the practical formulae smallest term is counted. se ek the point of convergence of the series you are calculating. to obtain the proper values. That is. series. and the degree varies. These are divergent formulae. for numerical calculation. stop when you reach the bottom. Bu t the terms at the to end tend zero in .PURE DIFFUSION OF ELECTRIC DISPLACEMENT. the divergent and convergent series is. the error is not very gre at. The quantity 8qx should be several times unity to make (47) be practically true. More generally. and may be far less. the point of convergence is always at the end of the which cannot be reached. except when qx is so small that the smallest term may be big Even when qx is as small It is when the point of as 1.

and of fair approximative ness so long as the point of convergence is not too near the first term.size. if 8qx is greater than 1.. a very large number of ter ms must be counted before we come to the convergent region. the formulae are convergent from th e very beginning. trouble we may reduce so that by taking enough the error to be smaller than any quantity that can be named. There is & . But then. The connections between the convergent and divergent form ulae will be given in Chapter VII. when qx is big. as well as their connections with other form ulae. so the convergent fo rmulae are On the other hand. It is a question of practice which formulae to use. divergent and allow of rapid calculation with an accuracy whic h i& unpractical. practically unexceptionable when qx is a large number.

VI. After that (48) may be used. Look at (39) for example. and therefore . it is not necessary to rewrite them.238 iable of the I ELECTROMAGNETIC THEORY. both with q finite first. by the vanishing of K or K. to indicate the special results. and it will not be a bad thing for the reader to look into a few cases in detail. p. CH. or both vanish. The constant impressed current h at y produces V as there given. going up to I cannot find any on. as has been said. Observe. Since. The corresponding current if = E. noting But table of that Ii is the derivative of I the important K functi use (49) myself. and I x functions 282. with <f = RK in the I and and K functions. that should either B or K vanish. to get rough values. or even (47). when steady states are concerned. Now. qx=5-l. and th en with q = 0. or . R (and others) in Gray and Mathew's work on Bessel functions. else (4). how ever. and Z sides. the above formula become numerical when steady states a re concerned. the formulae simplify co nsiderably. with *JRK substituted for q.

take a source at a point. one after generates. h all goes to the right ultimately. and construct the double wave it We Then we add on. leading to Fourier series.is zero. another. C 2 = ft. rep resenting the case of no waste of energy by resistance. but by transference to a great and conDetails tinuously increa sing distance of the variable state. on the right side of the source. The next thing to be done is to show how the Bessel This was done solutions in general are generated by wave s. V is zero on both is C x = 0. going to right and left. Waves in a Medium whose Co nstants vary as the nih power of the distance. The complete set of waves. when Z = Lp and Y = S/>. terms rep resenting the reflected waves generated at the terminals by the incident first waves and their successors and overlappers. or of motional statics. but they disappear ultimately. setting up of the steady state involves electromagnetic waves. on the left.] Generation of The divergent Formulae are fundamental. The actual cessation. when in opera- . before in the case of a uniform cable. 336. of course. [Not by point. and there is no The absence of e lectrification is the striking electrification. later. It is really a case of statics. That is.

Also. Conexpresses the operational solution in condensed form. the mathematical side. let plainer by . tional form. the convergent analysis the Bessel waves of The wave and the divergent. the series in terms of To do this for the Bessel series is not a useless because in Maxwell's theory all netic force are propagated at a speed v settled tivity disturbances of electric or magby the inducand permittivity. which property is not interfered with the simultaneous existen ce of conductivity (electric or by magnetic). when summed. we make v be infinite. by artificial In such limiting cases restrictions. all the same. normal functions are equivalent. for they are made simultaneous. Therefore. mathematical but is of fundamental importance. but there is no essen tial difference in the The waves are there. and the extension. From the physical side physical and the mathematical side. we do not see any su ccession of waves in time.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. This is also true when. and synthesis is nearly as easily done is for order as for the zeroth. All solutions involving change are . which. ultimately of wave nature. 239 makes a geometric series. version to series of no rmal functions then follows by the So we prove that the series of waves and expa nsion theorem. from principle. therefore. both from the complication. the question is important in casting light upon the conne ction between the two kinds of Bessel formulae.

dx or f^ + dx* x dx f= 2 *V. so that the resistance and inductance per unit length of circuit vary as or n and the n The characteris tic of V is leakage and permittance as x .n th power specific prope rties vary as the n or is medium whose when of the distance the variation from a fixed plane . (2) where This 2 = YZ. xn dx I^-^V. . Y and Z being preferentiall y K + Sp and R + Lp.mih made dx Xn dx (1) be the circuital connections of voltage and current.th is as the is dimensions. . the theory of the propagation of plane waves in a ih .

. and so on to other (n l) not restricted to be a n integer it may .But n if or of cylindrical waves . be fractional we like. power.

first CH. VI. It may be readily These are to define the functions x-m R m (qx) and arm 'K m (qx) satisfy the characteristic provided n = 1 H + 2m. They are. It is difficult to see th is. actually the fundamental not Bessel formulae. and K. The step is to find the wave operators. because they are the wave generators. when qx is numerical. are divergent series. that m= is. Now H and K or J (n 1). Let tested that (2). nevertheless. Suppose there is no waste of energy by .240 ELECTROMAGNETIC THEORY.

qx G. if there else to interfere with this qx simply proportional to t and K to eindicating propagation at constant speed w ithout distorIn other cases. Then to t q turning t Similarly ~qx + x/v turns is a mere translation operator. the operator to an H outward wave. Then H flx . nothing Therefore. and the is K 276. G are tion. is t qx F. t qx to t .J. and in any time function on which it works. in the property. where F. T he translation is still done by the exponential as in (3). and K is e. pjv. by operator belongs to an inward wave. Now. and m = .x/v.resistance.

from the fundamental fact that disturbances travel at speed v. If this general reasoning is not convincing. eP*/17 operators may and cj are other operators also concerned in producing distortion. but the operand is altered these operators do the distortion. (4) with an operand added is H H quite easily effected in various simple cases (and the extension to the general case is only a matter of complication) with the results as above described. wave. or wave s. I may add that the algebrisation of (3).. of So the deformation When there is resistance. case of a is homogeneous medium n = 0. . when/ and K prove that they are forms of for an inward. Thus. first by F or G. H operators. the be reduced to But the exponential / and t-P x/v g. and K for an outward the wave operators. still q is not p/v.

Some cases will be given in Chap. . VII.

that C 2 . Thirdly. the operand is h. but if we take it to be constant. and K to show the primary Let there be a source h at y. which is (HK'-H'K). The last is done by the conjuga te property of the H. K functions. beginning at the In (5) moment from y speed v. that V 1 = V 2 at?/.C 1 = h at y. which may be any function of the time. and V2 to the right. 2 m m 4 y x (5) show the waves. To prove this. note f irst that the proper operators containing x are used for the waves. 241 Having settled this point. use waves generated by a source.PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. V x to the left of y.--i. Next. TTOG (G) if the accent means differentiation to x. t . H Then V^-Z&ls??"* mt xm 4 y V-^-Z/il"*??.

primary and secondary. It is the same when the origi n is not in question. and so on for ever. the 2 wave goes on to infinity. It is now easy to construct the wave series. generates a reflected wave. If there are no terminal impositions that is. both the outward waves. if the characteristic is valid (except at the source) all the way from x = to oo . There are no more waves But should there be a boundary on the right of than these. and travelling to right and lef t at Construction of General Solution by Waves. which also travels out to infini ty. 337.= 0. on the T and y. generate reflected waves there. it y. n. left of R . VOL. But the Vi V wave. when reaches the origin. then at that V x and V2 represent waves of V starting moment. say at Z. and both these inward waves generate outward wav es when they reach the origin. but the place of inner reflection is at x = A.

)/(A. is Let # = A and I be the boundaries. and the point x where V wanted be on the left side of the source h at y. because this operator The result is it. to produce the value of the reflec ted wave at the same place.242 ELECTROMAGNETIC THEORY. is 1 at A. VI. (8) &' Put x = . and has the right x K mx / if di is *' rt 4 y m ' . CH.K TOX ) to expres s the reflected wave at any function in point. say a'. Then multiply w by (or K wa. The first is wave ^. Multiply by a reflection coefficient.'a^-rSm xm 4 y (?) by the first of (5). Put x = \ to produce its value at A.

m m mZ ) to produce t he complete third multiply by (x~ ~H. I. The result is (9) mZ if the operator in the second brackets. V . I to give the value of v 2 at &'.rhe sum of all these waves is which expresses from Vi.the operator in the brackets. Thus. it is V 6 ==v 4 a fv r &c(10) . reflection coefficient. \ is After this. say to give the value at the .mx )/(l~ H wave anywhere.2 Finally. the same over and over again. Then multiply by a same place of the third (inward wave) generated by v.

but only by reflection at 1. But there It sidered. It is (12) ^ } is also the initial wave to the right to be condoes nothing at x itself directly . . to produce the first inward reflected wave. so far as it arises Put x = l.at x (less than y). Initially the primary wave is the V 2 in (5). and multiply by V(x-m H mx )l(l-m Rml ). say w 1} which can act at x.

and the rest are w8 = 'iai&i. tF = ir8a 6 1 1. &c.. for both and Wi are inward waves. W4 = w o 1 6 2 1. After this vl it is 243 to be treated in the same way as vl was.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. Therefore 4 is y m xm (12A) the next outward wave. (13) Consequently the sum of the w waves is .

The corresponding \ .aA Let Va a.2 ^ = ^Jl^. only remains to find a 1 and b lt or coefficients. that is Vl It 1 . multiplied by a factor not containing currents are +1. the reflection and consider any outward wave superposed. . x a?. They are rep resented at Z. 1 aA (14) and the complete Vj at x is 2> + Zw . V = ZjC wave and the by reflected a' and &'.

and using these properties of the functions.. ml and therefore (Pg/ZKl^-bJI^)' (20) R2 . i. xm ' m q dx x which are m q dx x _ = __ xm ' The easily verified by the definition of ratio of (16) to (17) is therefore Zj when H and K x = l. = __ K. above. So Z.~ ^Hm+1> x\ (17) by the first of (1). + bJ3.

244 ELECTROMAGNETIC THEORY. and b for the reciprocal of 6 X in (20). by The divergent series are the addition of the reflected waves . so t hat a is derived from b by turning I to A and have only to interchange x and y t o obtain Zj to Z . and the results are explicit in terms of the terminal resistance operators. then symbols that V= . We V2 on the right side of the source. VI. therefore essential to a proper understanding of the matter. : r/ A n Zi-fl We may therefore write (15) symmetrically thus y _ vZh where a stands for a v as in (21). u A/. . let CH. They are derived entirely from the initial waves from the source.Z C be the condition at the we shall find by the proper change of . In a similar manner. other boundary A.

These I and infinite when m is negative. Now got by turning functions are always finite at the origin zero. and L^ or I_ m (qx) is m to . Thus x~m lmx and x~ m l_jnx series satisfying the c haracteristic.Construction of General Solution by the Convergent Formulae.m. positive or It follows that when m is Vi = -A. lt V2 = m -B p (24) . are convergent solutions of the characteristic (2). if 338. consider a radically different way of viewing the Besides the divergent series t here are convergent subject.

V x = V 2 at y. s.my (25) . the place of the sou rce h. and represent V on left and right of a source h provided r. A x and Bj are properly found. makes my T *. First.are general solutions.

and the value of the quantity in the second brackets in the denol is easily seen to be x~ x constant. writing down the fi . we find A So. so 245 Cj = h B T is eliminated. h condition just mentioned. in the manner of 331. involved of voltage a nd current. Now Next. A x is to be found by the C's corresponding to (24) are Ca C i= C2 = because Int-fl "* ' (26) 1 <f I_ m m ^I_ m _! m ' . at y. applying the = -S In the denominator a conjugate property is at the point y. To find the constant. * } as may be quickly tested by the series (23).PURE DIFFUSION OF ELECTRIC DISPLACEMENT. use the series (23).

m -i t minator &c. . and . and pick out the terms involving or1 result is The I I_ ?n _i-I-. T= -2 = ^ \ irq x sinmTT. I.rst terms of I m. There is only one.H I l +i = 21 -j qx \tn\-mm "*~ . (29) because 1j-m= -m\ -m-1.

-m y m7r sinwiTT (30) come are elementary properties of the \m^ function, which we shall across again in Ch apter VII., to be there proved. Thus A = \^ m 1 ' y mv (s r) siuiinr' and therefore, finally, x m y m (s -my) r) sin (S^\ mir in a symmetrical form, is another we determine before, way of expressing V r If and r by the same terminal conditions as - Z C at A, we find namely V = Z 1 C at I, and V = s (33) -m-l, X

246 where ELECTKOMAGNETIC THEORY. CH. VI. A and B is, are as before, (20) and the solution like the old one, explicit in So now (21) above. terms of the terminal operators. Comparison of 339. Wave and Vibrational Solutions to deduce Relation of Divergent to Convergent Bessel functions. (22) and (32), obtained two entirely forms of solution of the same problem. In o ne way we built it up with the primary waves and their reflections different We have thus, in and it is certainly right (barring The other way ignores waves possible working errors). altogether, or indeed any sort of elementary component solutions, but is entirely operational, using formulas whic h are convergent, when numerical. But the form (32) may be expanded by the expan sion theorem into a series of kind. normal functions of the subsiding or vibrating therefore for the present regard (32) as representing this normal expansion, in the same way as we may regard (22) as representing the series of waves, for ever y wave given in operational form may be algebrised if we like.

We may We therefore prove the strict equivalence of the series of normal solutions and the series of waves, arising out of (32) and (22) respectively. But, besides that, we may regard the investigations algeFor q may be a positive constant, braically and numerically. when Z = B and Y = K. It is then the steady state namely, due to h that is in question throughout, which is instantly assumed when h varies, because the speed v is infinite, and t here are no time differentiations concerned in the various So we operators, whic h become constants or functions of x. prove the numerical equivalence of (22) an d (32), when qx is positive number, apart from their equivalence as operational formulae. What are then the relations between the divergent functions H, K and t he convergent functions I m and I_ m ? They are involved in (22) and (32) of cou rse, but it is not clear at first how they are to be exhibited. We must either r earrange (32) to show identically the same form as (22), or else the other way. But a trial with the sum and difference of Im and I_ m used in (22) shows the wa y. Thus, keeping entirely to (22) at present, let = Imx + I_ 7Jl ,, K mx = ^-:J=?. sm WTT (84>

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. 247 These are merely to define H reference to the former meanings. Then and K, without present (32) becomes ^^ mi,}, (35) which is expressible as Zfr (H 4sinm7r ' (37) in the numerator of (36). Here r and s are given in (33), according to which (37) become Next, there is the denominator in (36). Putting it in terms So of a and b in (37 ), we find its value is 2(& a). _ iirZh (H b -a This being merely a modification of (32), compare it with They are identical; fo r the present a, b given by (38) (22). are identical with the a, b of the diverg ent investigation, viz., !

and But in (22), in (21) and the reciprocal of b t in (20). are defined by the diver gent series, whilst in (39) they H K are defined in terms of the convergent series, through (34). It follows apparent ly that (34) express the equivalence between the divergent and convergent formulae. But it is not a rigorous proof. For there sibility is just this posby comparison. However improbable it it is possible (unless proved to be may be, impossible) for some other combination of the functions I m and !_, to behave in a proof same way as regards reducing (32) to identically the form (22). If it did, we sh ould soon find out something anomalous by the impossibilities which would arise on further However, I may mention here that in Part 3 of my pursuit. " " Operato rs in Physical Mathematics Paper on (May, 1894), in the

248 ELECTROMAGNETIC THEORY. CH. VI. I have given an investigation which transforms the Im and I _ m functions to m a nd K m functions according to (34). It is an H the above, effecting the transformation by algebra alone, without any differenti ations or Of this I will give some account in Chapter VII. integrations. entirely different process to use the second of (34), we shall arrive at the same result, equation (39). Going further, if we use , Hm = 2Im instead of Im + I_ m Im The above suggested ambiguity does and still occur. For if we put = pRm + a- sin IIITT K m, (40) leaving /> and <r arbitrary, in place of the first of (34), still using the seco nd, the result is that the right member of (39) is multiplied by 2/>, so we requ ire p = \ to harmonise the convergent and divergent formula. As for cr, in the result at all, so it looks at f irst as minate, and that it does not appear if it were indeter(41) Hm = 2I TO -2o-simmrK m, with any value of a-. But there is another consideration. The last formula must

not contradict the second of (34). Now m and K m are even functions of m, so the last equation H makes Hm = 2I_ By addition, ?n + 2o- sin mTr K m . . (42) we is obtain the first relation in (34) atraction the second relation, provided = - J, and by suband only then. ; So the matter made square. Nature of Algebraical Transformation from Divergent to Convergent Formulae. 339. But I have previously given an investigation which covers the case of integrali ty of m. I have shown that the operain one tional solution of a certain physical problem, when algebrised way leads to the convergent form of the zeroth Bessel way leads to the divergent form; Now, given a Bessel function of any order, all those differing from it in order by an integer may, as is known, be derived by c omplete differentiations, as in (18) and function, and in another thus,

H = 2I . is any integer. Therefore m = 2ITO when (27) above. the rigorous mathematicians have to say To see what , H m this matter, I have referred to the latest treatise on (Gray and

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. l249 On p. 68, equation (142) is a result given, which, Mat-hews). allowing for the d ifference in notation, harmonises with the second of (34). But the other result given, equation (143), is discordant. is It is equivalent to mx = ^-mx in my notation, and therefore true only when m is an integer, in which case the functions I ni and I_TO become identical, as (23) shows. No is H proof given in either case. algebraical transformation from the diverLet series, it goes thus. As regards the gent to the convergent Bfo*r)~ \m + r-l \r-I is \m + r\r [m + r + l \r+l + ....' (43) This Bm function mth order. of r. In it, r is the generalised Bessel function of the any number positive or negative. If we

increase r by 1 The Now reproduces itself, so it is a periodic function be continued both ways, unless i t stops. in the H, K formulae, put for qx and e~qx the following it series is to generalised expressions, .., r (44) Jr+1 (qx) r+1 \r-l where r cations, is r -^+r " ; ,.-\ as before. they are alternately the + and : H, K In (44) the signs are all + in (45) On performing the multiplifunctions are turn ed into B functions . according to the following Change the sign of m

to obtain a second formula. The two . formulae then give m andK m in terms of B m and B_ OT "When r is zero or any int eger, B m reduces to l m, and we obtain (34) above. In the generalised formulae (43) to (46), H qx should be a real positive quantity, except in special cases. The identity of I m and I_^ when m is integral, makes the second of (34) assume the 0/0 form. Th en take the limit. Thus, K.. =-! 1 7TCOS W7T

250 ELECTROMAGNETIC THEORY. CH. VI. when m is integral. Or, by (23), [0 Q^ i m if f(-) + (M- m+ *f(-m+i) + is its is (48> 1 /(m) = ([m)- and /'(m) formula before given for general case presents no K 0x The convergent derivative. a special case of this. The difficulty, but requires /(m) to be explained, which belongs to Chapter VII., along with related matters concerning the development of wave formulas. The reader should be cautioned against conclud ing that equivalence, as of means identity. The a divergent and a convergent formula, fact that they are different shows that they are not alike in all respects, and cannot be interchanged under all circums tances. I am inclined to think that this is true even when the equivalence exists between two convergent formulae of different types, in fact, what rigorous mathematicians call an identity. Or there may be equivalence when the argument

negative. is real, but not when to it is The extent which equivalence imaginary or even persists is an interesting matter, but is better observed in the practical concrete examples th an theorised about upon in complete data. Experience and experiment must precede theory. Rationality in p of Operational Solutions with two Boundaries. Solutions in term s of I m and K m . 341. In order to convert the operational solution to a series of Bessel normal f unctions we naturally use in tha first place the form (32), involving the conver gent functions We that are virtually in possession of the unit impulsive function, and developing by th e expansion is, by putting hpQ, theorem, the result is a formula showing how the charge Q, initially at y, subsequently behaves. But to allow of this development, (32) should be a rational differential equation. necessary failures are obvious. First, if Z = oo, which does away with the infin ite series of reflections, leaving in general only two waves, these waves themse lves constitute Two

the practically significant result. The set of normal functions nowith distinctly separated periods or rates of subsidence

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. 251 functions approlonger exists, although the series of normal when I is finite has its ultimate representative in a priate definite integral, the limit of the series. Secondly, if either of the terminal operators be irrational, we have a similar failure, and a simila r resulting definite integral. But, assuming that I is finite, and that both the terminal Z in the circuital equaconfidently expect that the operational tions), equation (32) contains p the time differentiator rationally, in m where spite of the pres ence in the functions concerned of q operators are rational (as well as Y, we may , m may of fact, be fractional, or of log q in some cases. As a matter such irrationalities are i noperative by appearing in a suitable manner for cancellation. Thus, in (32) it will be found that r and s bo th have the factor <f m whilst I m has q m So the numerator and denominator in a nd I_ m has q~ m both have the factor q* m and, therefore, (32) is a rational (3 2) function of $ 2 itself a rational function of p. This being true for any valu e of m, it follows that in the case of m being integral, when we need to employ the 0/0 , , . , , K m function, which brings in the logarithm of q, a similar elimination of this logarithm. It goes out = In the results in this fashion log \ql log \q\ log //A. form of the there

is : we have only logarithms of numbers. The alternative form of C32), in terms of I OT of I m and K m instead or (49) and I_m, is got by using the second of (3-1), I- m in (32). = Im + K m sinwr This brings us to V1 _ j^ZA m x ym (Iwx -/Kins) (I??/ f_ g where /= is *""-^W* , (51) Here the congot by turning A to I and A to B. function may be understood, though the same vergent result is true with the divergent form. and g Km infinite, it will be found, by using the divergent oo according to the g becomes either + oo or nature of Z r In either case Z 1 is impotent, and (50) reduces t o If

I is made series, that ,

252 ELECTROMAGNETIC THEORY. reflected one. CH. VI. showing just the primary wave and a lished, as to This is the form suitable for simply periodic states when fully estabwhich more later. normal functions is merely a formal one by the expansion theorem, but there may be a good deal of work needed sometimes to bring the result to a form suitable f or calculation by such tables as exist. In of (50) to The conversion complex. That is too bad. a relatively simple nature. finite. general, the roots of the determinantal equation f=g are But there are important cases of Say L = 0, is Then only electric energy "medium," whether with wires or retaining E, S, concerned in the If also K is not. there energy concerned at the terminals, the roots (for p) of the determinantal equati on are all real and negative. In like mannei, if S = 0, retaining R, L, K finite , only magand if this is also netic energy is concerned in the medium only electric ;

the case terminally, we have again roots of the same nature. These are two ext.r eme cases of diffusion, with infinite speed v of propagation, though the practic al result may be slow R, K are zero, and L, S finite, we have propagation without waste by resistance ; In this case the therefore undamped vibrations can occur. roots are p = 0i, wh ere 6 is real, and corresponding terms can be paired to make real vibrations. To this may be added finite enough. Also, speed of if that if R and K have the last case again, are not zero, but are properly balanced, we and with attenuation due to R K superimposed. This nearly exhausts all the applications of a relatively simple n ature; though if we do away with the terminal considerations, by taking Z and Z : to be either zero or infinity, which makes either the current or the potential vanish terminally, we can extend the matter further. The convergent Oscillating Bessel functions, and Operational Solutions in terms thereof. 342. In all such cases, it is convenient to convert the funcand K m to the oscillating functions J m and G m,or The cases of integrality of to J m and J_m instead. perhaps m are perhaps more common then it is better to u se K m or G mt and ignore I_m and J_m tions from I m ; .

PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. 253 Thus, (53) The where oscillating functions are got U<F) = (**), = si. q = ^J-mOw), I- m(^) by putting and J_ m sx is the same with is real, this is w put for??i. When the argument an oscillating function, the original Bessel But the functions involving q are more function in fact. primitive. Now as regards J m . Use the second of (34), with q = si. (55 > Then, Here i 2 = (cos + i sin) imr. So we may write K m (qx) = *-{GU0 - i3 m (*x)} denned by

9 (56) where G m is snmir Or, if we expand i~ m in (55) as well, we get m-m g _ J-m-J _ But the form (56) is the one to take note of. So now, by the use of (53), we con vert the equation &rZ/t ( J mx - pj _ mx ) (Jmy - o- J _ my) (32) ta y_ where -l.X , A' A= ^ sZ = n /ftn\ , m\(bO) . -m-l,\ I and o- is got by turning A to and A' to B' in p.

We have Also, by the use of (56), (50) to the form we convert the alternative equation V1 where and /? is x my m aJ ^+ a= \im\ + A got by turning A, ^-^, ljm+l,\ to (62) B' in a. to J and A'

254 ELECTROMAGNETIC THEORY. CH. VI. /, Observe, comparing the forms of (50) and (61), and of a and that the transformat ion is the same in effect as turning In to J m and K TO to G,H with the change f rom A to A' , in the denominator, it to a. (59). q = si, - A' in the numerator of/, to turn as regards the transformation from (32) to Si milarly, The symbol i does not appear in (50), but when we put it appears in m t hrough (56), and therefore in /and g. and to K It cancels out on further reduction, and then the strikingly That i ought to cancel out is clea r similar form (61) results. because since (50) is rational in <? 2 (61) must be enough, > rational in i goes out is remarkable, depending, as it does, upon the It will not take place when existence of two boundaries. I = oo then g~l is zero, and the form there i s only one. Say s . 2 Nevertheless the way the symbol somewhat , Put q = si, and we do not get a result rational in s 2 and we ought not. In conn ection with this, there is sometimes a bit of hocus pocus. If we like we may mak

e the functions Jm and G m the (52) results. , primary objects of attention, so that V^J^-aG^X is (63) s 2 the initial form of operational solution, having the meaning - YZ. Determining a and X by the conditions as regards h and terminally, we shall arrive at the result (61). This may, in fact, be the best way to work, when the ultimate results are to be simply periodic or normal solutions, and the development of wa ves is not in question. But there is a curious irreversibility sometimes concerned. m to Jm and pass from the primitive Im and K We can always G m, but we cannot always go the other way. For instance, (61) leads to 1 (50), by substitut ing ^~ ^ for s, only when there are two boundaries. It fails when I = oo . For example, suppose that the condition at I is that Y 2 = 0. Then 3 mx -f3Gjnx

must vanish at /, and this shows that ft must be J,nZ/Grml Introduce some other kind of terminal But how find condition, and we get some other form of (3. If we have worked entirely with its value when l=cc? J TO and G m, and know nothing o f I m and K m there is appaFor J,M/G Ml the rently nothing to show what /3 shoul d be. ratio of two real oscillating functions when the argument is . , ,

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. real, 255 infinite. has no particular limiting value when if I is Nevertheless, we write J^/G^ = - i, we shall come to the proper result. For it is the same as /?= - i, which The explanation has been already virtually given. function that is is correct. It is the then (G m -iJ m )i~m takes its place that is, -ifim + iGm)^", or ft = - i. But, quite independently of this determination of /?, a physicaUy-minded man, who was working in terms of ; Km concerned alone when I is oo ; when q = si, J

/rt and G m arrive at the correct result for convenience in the practical application, would by considering the flux of e nergy. In a simply periodic state produced between the source at y and infinity, with n o reflection, the flux of energy must be outward. This necessitates fi = -i. The divergent Oscillating Bessel 343. functions. We know that the operational solution in terms of the divergent H m and K m is e quivalent to that in terms of Im and I_ m or of I m and K m Therefore, the same transformation . q = si in the divergent operational formula should lead to a result equivalent t o that in terms of J m and For dism G . tinctness, put a bar over the divergent functions. Then, q = si in H,n , Km produces (sx) m (8x)}, (64) (65) w (iO - i3 m (sx)}, when 5,,i

and G OT are given by * 3 m(sx) = (}* (P cos + Q sinW \ TTSX/ Jir \ Im*), (66) / 5,() = (^)*( p sin + Qcos)(S2 J* iiw), in which (67) p P and Q are the divergent functions 224m2 ) (I - 4m2 4m 2 ) (3 2 - 4??i _ (I (^ _ (5 2 2 ) " ) / l 1.28a? 2(8a?) a \ 3.48s ^ 2 ~ _

' ' ' ' (68) 2 _ " (7 -4m )(9 2 2 -4?/t2 ) l.(8taj) (69) obtained from the J m one by turning sin to cos and cos to - sin. When x is larg e, P = 1 and Q = 0, so is The G m formula

COS SX ~ CH. They be have just like cossx and sins#. recognise. but with amplitudes varying inversely as the square root of the distance from the origin. and so are G m The first was proved by Sir G.256 ELECTROMAGNETIC THEORY. J m and J m are equivalent when sx is real. some of which are very inconvenient. owing It to the use of several forms of the second has been standardised in different way s. " 71 ~ W?r ' > G m (sx) = -Y (. function.SOC/ \ 7T sin (aj JTT J^TT). though somew hat difficult to . (71) show the ultimate nature of the oscillating functions at a considerable distance from the origin. VI. Stokes the second and G m I find in Gray and Mathew also. If in the .

instead of the proper form (64). two same There This is in fact. But it is not an equivalence w hen q = si. we put we shall.n in (22). divergent still oscillating functions. Hm and Km We do w = 0. and G m instead of G m alone would not prove the equivalenc e of the convergent and . result (61). for instance.iJ and we . This only with J m instead of J m. arrive at the is. One makes J iG On the other hand. The other makes 2J and sx is real. an essential difference between the divergent functions Say have H m = 2I m when gx is an important case. . K = G . For example. .iJ and K = G . (65). we shall arrive at (61) above precisely. K m formula (22) we make the changes according to (64). real and positive.H m. if H m = 2*m J.

Km H .. 344. We . both when q is real positive. I have given elsewhere* an algebraical explanation of the distinction. and when s is real positive. have K =K . . and suppose that the source h is simply periodic. But it will be more satisfactory here to regard the matter physically. Physical reason of the unlikeness of the two divergent functions m.. by consideration Go back to (5). initia l waves. . Part 2.. for example. expressing the of the physics. and we ought to have the other discrepancy. in P. * 0.M. ought to have the one agreement.

257 and that there is no waste of energy by resistance. then q = nijv and s = n/v.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. The wave to the right is then y (72) when E and if e e are simply periodic functions of the time. These solutions indicate a train of waves travelli ng outwards.GmJm+l) x is . This makes q = p/v. = eQsiuntj Or. _ /f then . Its mean value over a period is (V2 C 2) me . (75) . and the correspondin g current is (74) Here q/Z = (Y/Z)* = constant. The flux of energy is V2 C 2 .J mx cosnt Qv This represents the ultimate result of the periodic source. n = \el This is (J m GTO+1 . so that if the frequency of the source is H/ZTT.

and this is as it should be. TTSX That (76) is to say.Gm =A. because there is no barrier. J OT and Gm. In a similar mann er. s . there is a steady average flux of energy from the source out to infin ity. the potential on the left side is /> f where F and / are simply periodic. again be found to be constant on the average .constant. on account of the conjugate property of is (if which the accent d/dsx) -JmGm + j. and to be But this state of things is an imdirected to the left. and this represents a The flux of energy will wave train travelling to the left.

The case m= is peculiar. the result between the barrier and the source must be a stationa ry vibration. and then travelling out to infinity. involving no average flux of energy. there must be one on the left side. VI. regarded as the ultimate state. possible one. and the frequency is so great that very many wave lengths can exist between h and the left barrier before it is Then the above solution may be nearly true in a reached. Its nature w ill depend upon the kind of condition imposed at the barrier. equation (7). new state of things will begin to prevail. It is second wave equation then generally necessary to consider the (8) as well as the first. If the barrier should be such as not to take in energ y continuously. Say that h is at a great distance. When the barrier is reached a as before. But. whereas there is no b arrier on the right side. first at the barrier.258 ELECTROMAGNETIC THEORY. or at the origin itself. It can only exist approximately initially. when 0a. either at A. great part of the region occupied by the disturbance. CH. and superimpose them properly. the barrier . like the other.

that is. In th e region occupied by it = y/v. that is. preceding state ( J )e.is at the origin. the equivalent 2I reached. . The matter is made plainer by considering h to be steady. There can be no current there. . later.0 or (J (G0x -iJ 0x)ie. as so on as the origin is and then. add the out ward wave the inward wave H^. The inward wave from h at y is calculable from H0a unti l the origin is reached. 0. But after that. The result is 2J 0ze. result is t convergent. Here 2J 0a e represents the ultimate stationary vibration which repl aces the . valid between y and the front of the return wave. The . 0a. beginn ing at the moment t = 0. after the moment only a partial solution.iG Qx The operator H is valid at first. So to iG 0x)e. 2Iox e. property Hm = Im + I_ m should be true when s is real as well as when q is real.c 0a. and no flux of energy. is . We see that we h ave no right to expect that the .

. identical at the junction. we can calculate the wave H0x The two forms of solution be come . a cause which is not op erative when the q by one or two special examples. H Km function is in question. This matter will be made plainer The present remarks are directed to the cause o f the failure of m = I m + I_ m when = si and s is real.the return by 2I0x instead of wave and the primary.

becaus e any terminal condition V= the origin.Z C of the usual nature. some noteworthy peculiarities. But we can throw cally. finite values. when A. = 0. We know that in o ne special case (viz.. some light upon the matter physically. there is no interference with the power of imposing . under This is rather tedious mathematidiffe rent circumstances.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. or vary as the (or inverse first) waves in a medium in which the constants power of the distance from condition its is a fixed plane.J). n = Q.1. we may impose any condition at the functions concerned have there are 345. or m= . the Z impotent. unless l>?i> . that of cylindrical waves in a of plane first another case homogeneous medium. from the absence of luminosity. that of plane waves in a homogeneous m edium. Z condition is potent. and see that the . to inquire when in genera l the voltage. is at a finite distance from the origin. But when A. This is to be can be no current at We done by examination of the limiting form assumed by a in (22). So long as the barrier at A. Electrical 259 Argument showing the Impotency of Restraints at the origin. = l. '(n or wi We have also observed that in = 0). is made zero. therefore. and when impotent. because there place of application under any finite have. or by r in (32) o r by /in (50).

Here the distance So we see that to # may be a very little bit at the origin. which is finite when n is positive. consider the permittance of a little bit from the origin. and also when n i s negative. This is f x Sxndx. Under what circumstances can there be a current at the Plainly the origin when under an impressed voltage ? Now the resistance per unit resistance must not be infinite.1.dx. length ing as is 7i ~Rx~ n which . The resistance of the length from n x This is finite when n is negative. when it becomes s2 is + or . must be impotent whe n n is 1 or >1. down to n= . if true also as regards the source h. = 1. applied. there can be no current there when n is 1 or > 1 This is The termina l condition. .mathematical results are justifiable. and also when is f Ex. and it is positive. when it becomes infinite. to # at Next. up to remains infinite for all greater values of n. . is infinite or zero at the origin accordBut the resistance per unit length is not the circumstances) the same as the res istance of unit (under to a? length at the origin.

260 infinite. when + . If.m b eing usual kind is operative if In the first case. This makes C be zero at the origin. and the Z condition of the potential at the origin above zero.1.1 and + 1. or tt = or >1. applied. CH. mx. to the nature of .1 or 1. ELECTROMAGNETIC THEORY. . no finite charge can raise the < The terminal condition must be again impotent. both the resistance Thirdly. In the intermediate case both functions will or may occur. is or require to use I mx only. when n is between and the permittance of the little bit at the origin a re finite V and C may then have any ratio. In the next case. according 1 or < 1.1 or infinite for greater negative values of n*. when n is . . we must use I. VI. n and remains is . then. we m < This makes V be zero at the origin. .

In the original formula (22) put A. or Im and K m. Ifn isIf n is not less than 1. we get a=+ sinm7r.sin irnr. In the internot greater than 1. the oscillating functions.1 and + 1. At th e same time it does not absolve us from making the examination. that is. we get . 346. when n is between . it 2AJma. we get a results. but also the form .).-aGma. mediate case. The last is leads not only to expansions of the form 2C(Jma. There are three . or J m and Gm . The preceding electrical reasoning will enable us tounderstand the results produced in the formulae when the inner boundary is shifted to the origin. both I m and I_ m. = 0. Reduced Formulae when one Boundary is at the Origin.Z . because without it we cannot say what special form is assumed when the terminal condition is potent. when the origin is one of the barriers. if we use rather remarkable.

in the first case the x function in (22) reduces tc2Ima. '. or they may be done separately.. Thus.where +u -ggj(gV -m-l Z V . (78). in ttieshown From . but in the se cond case to 2I_ m:r whilst in the intermediate case Z remains potent. That is. the above may be derived the changes in the other formulae . and a has the special value in the last equation.

u being as in (78). which are connected by . (Im or . n>l. is the reduced form of the determinantal equation when = ur or # = 0. (80) l or b + sin mir = 0.~Q^ we have r = when n>l. intermediately. and r . alternative formulae (32) __ 261 and imr ' (50). r=oo when Therefore s = 0.PURE DIFFUSION OF ELECTRIC DISPLACEMENT._ r sin _ ( _s sin mir 1 +r T+T' n< -1.

sin imr (!when n< = oo.t + BI m+1) = 0. or b . 1. In the se cond case the x function is J. . when w< + B'J_^1 ) = 0. g + sin nnr = 0. is s Also.^)^0. or (J m -B'J m+1)z = 0. we have (88) and = is the determinantal equation of normal systems ft the same as (80). Then a = tan imr. and . in the first case. Intermediately (82) _ . or + 31. As regards a in (61) when A = 0. or (J_TO n > 1.Pin imr COS imr + v So. the determinantal equation or . first we have a = l (81) when 1.^/COSTWTT.

using a as in (82). + ( J w .m + B' J _ m _ 0. or is m tan */ (1 . (85) where v As make as in (82).J in the last equation. It reduces to n = 0. Intermediately. (86) . the determinantal equation a = fi is represented by ( J . a test of avoidance of error in the way of wrong factors. = 0.v =l 7" r-sL COSTTITT zj ( W= or<-l) (84) whilst the determinantal equation is equivalent to (81). as before.B'J m+1 ). = . and B' is ZjS^/Z.Z Z lS2 /Z 2 ) + = ZjS/Z + Z s/Z 0.

allowing for the present gen eralised meanings of Z and s compared with their meanings in that place. we find proper agreement. . 293.Comparing with (75).

by external restraint. through Z = 0. Similarly. The peculiarities of infinite resistance or permittance at the origin with suita ble values of n place restrictions upon the restraints possible. down to just ov er 1. up to just less than + 1. also ha ve the potential zero there when nWe namely.. CH. through of wires. the current at the origin due to the source at \j vanishes naturally when n = or > 1 We can also make it vanish there when n is less. VI. . but then it must be done by external restraint . however. being equivalent to a short circuit if a pair of conductors be in question. may. It is equivalent to a disconnection in the case it is when n . But is greater. The source at ? / cannot raise the potential at the origin above zero when n equals or is less than. when n= or >1 it is no use trying to make the potential vanish.1.262 ELECTROMAGNETIC THEORY. Z =x But .

1. it harmonises with the proper forms outside those limits. we impose the condition V2 = at I. to make the current vanish Equation (85) only applies when n is between. the other form. = at origin by external restraint. Vi This is reis the determin antal then J_7ni = equation. (Z = 0).but equals or is less no use trying than .1 and + 1. (Z . If. if Z is infinite (85) reduces to (80).. and (85) (87)becomes (J+ J)i = 0. provided n is intermediate. thenB' = 0. If Also. and then (85) rt reduces to (81). (-l<w<0). if the current is zero at the origin by J mZ = plac ed by external restraint. which makes the potential be zero terminally. in addition. Thus v vanishes when Z does.

The development is in Bessel series when h steady or done in just the same way as for Fourier which department has been somewhat elab orated. Series. . using. Soseries. . The Potential is initial Charge. solution. First find the final steady state. The Expansion Theorem and Bessel due to 347. little need be said about i t. Let this be V to the denominator a-/? (or other form) in the operational Thus. impulsive to be when there is one. as is nearly always the case in practical Then apply the p(d/ dp) operation problems.= oo ). according to the expansion theorem.

for we naturally want to have the result entirely in terms of s. It can only differ in the expression of the steady part V In passing. It ^ = R. we have diffusion. Also s 2 is a function of p. we The values of p being the roots of a = /3. then the effect o f h accumulates incessantly. over which the summation ranges. and there must be one with a steady sour ce unless there is perfect insulation in the way mentioned. Fo r instance. is impulsive. and there is waste of energy in some part (no matte r how limited) of the connected system. There must be one when the source steady st ate tended to. it may be remarked that cases may arise in which there is no tending to a steady state. Therefore the formula for V 2 on the right side of the source at y is just the same as for V x as far as the summation . . and Y = Sp. and the values of s are settled by the determinantal equation. wires. The further development therefore rests upon the nature of Y and Z in the original circuital equations. Now in (88) the J an d G functions concerned have sx for argument. and and But such exceptional cases need not delay us here. The outside term then contain s t.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. beginning when = 0. 263 t and supposing h expand it to (61). if the above refers to a circuit consisting of a pair of parallel goes. . with one value of p for one of s 2 Then . The interchange of x and y makes no difference. to be steady. Vj rises to infinity. we s ee that either a or /? may be eliminated from the numerator. and the insulation is quite perfect intermediately terminally. but can be At present assume that there is a treated when they arise. and so is Z.

and a = 0. 2s ds (89) This brings us to <rf/*)(-fl 2 the value oip in the time function is-s /RS. and dp = . the last V V -V "*5w reduces to V ^Q* 5 ( Jm aG m)x( J m . where If \ = 0.ffGTO) y fVt ' /am is 1 or .** = .?.A.BSj>. which occurs naturally when n more.

in /?. VI. making J ml = So ^mx^my ^t 2 be V = "X * if 7?"Qs vmx myGml "S^/T ^T f pt _V " _^ W(xy) m ($%} (3'na ) the accent denotes differentiation to the argument second form is derived by the conjugate property . = finds the values of s 2 and p. If the potential is B' = I. CH. constrained to be zero at = now. .264 where /3 ELECTROMAGNETIC THEORY. then Also V the determinantal equation.

as before explained. n 1 or less. The (98) remembering that J OTZ = 0. with any . by (84) and (57). Put Q = Sy*Wy. provided Z = oo But is if imposed terminally. We get. down to just over 1.si. the solution takes a different form. Equation (92) expands an arbitrary function of x to suit the conditions stated. and integrate with respect to y from to I. The result is the V arising from the initial distribution of potential v. This also applies when n is is smaller.

See equations (66). jugate property. V2 = at I.Zr So Zj = 0. (71). Equation (96) does not look right. making B' = 0. (97) A companion formula is G. (98) . V2 being got by interchanging x. 1-^ m had J . (67). The last expands to = 0.m = J w cos m:r + Gm sin WITT. makes where /? is J mZ/Grwj. which makes y. The cases in which is integral subject to The connection is better be kept in terms of J wl and Grm .m =Gm cosm7r.Jm sinw7r. (70).

But we have the con=~ 2 -sin?7T= .J_ m J' m + J w J'_ m. (99) TTA't .

S. are finite. K. It is valid when = the range may be extended up to just under at the origin. L. q* (101) 348.PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. there are two ^'s for every single thus. /nnox (102) ^= dp The sum = S(R + Lp) + L(K + Sp) = tip (103) .s2 = (R + L/>) (K + 879). since J_ mZ =0 at (29). 265 which may be derived from present. except in the reversal of the . (Z = 0). 2LS^. /R + K\ 2 s) RK + s 2 )* --LS-J. is sign of m in the J functions. more general case in which R.1 or less. but m m by a terminal all Tune Function when Self-induction is allowed for. eliminated from (96). This brings it to that is. restraint making Vj = the same form as (92). K\ s> HH 7 I. and = . J* maybe So. In the s2 .

Then in (8 8) Also. an d remembering that a = /?. and <^() . ( (a -^)= d/3mx . take the simple case of the potenvanish both at A and I. therefore k -. Therefore (88) takes the special form mv ~a 2 -(Gm . ty using the conjugate property (93).)where a is y( f) n 06) 3 m}. d application.JmA = 2/1 1\ (105) si.of the two time functions of the form Zfpt/(dq z/dp) is t got by using the above two values of p. Jo (104) To show the tial constrained to V = 0. with argument sA or as the case may be.'Gm \.

and integrating. . by putting Q = Si/nUrfy. initial state Its value is 1 when t = 0.is the time function in (104). so we can expand any U of potential by (106).

349. CH. now have a. We operative. VI. we need not tinuity of current.106 ELECTROMAGNETIC THEORY. . e = 2 . thus. specification of therefore incomplete if the potential alone require to know the initial current as well. but with con= C^ To find the result . We the zeroth Bessel function was concerned. let The Instead of the above the source at y be e. or C 2 V V go through the work in when shall detail. producing a jump in the potential. self-induction in the last makes the magnetic energy the initial state is given. The Potential due to The inclusion of the is initial Current. h.x at y. but generalise the former result 330.

. a and b being the same as before. by the conjugate property Hm Km+ to it. i + K m H m+ i = A=K )H H' W Hm K' M . and the results are easily got J and G . may but this is tedious.bK m+1 ). is got by interchanging x and y.b-a . we Similarly.aK m+ i) y .xn 7r\dldx) finds the current. The operation . V= b-a Vi being on the left. It is easy to test that these satisfy the condi tions at y and terminally. putting H. V 2 on the right of y. aIi m+lUBm+l So (109) C. to find the solution in terms of derive th e results from the last. but by interchanging a and b and negativing the result. We turn V x to Va not by interchang ing x and y.= . (110) the argument being qy. and the accent denoting differentiation in terms of J OT and m.l^f-e and now C 2 (H ro+1 .

Thus. V= 1 T ?/ * ~a x m+i ~ m+i y ^ ju m ( a. rzc .(3 J (111) VV~ i-?/m+1 m-^GJa! . independently.K G .

PURE DIFFUSION OF ELECTRIC DISPLACEMENT. #m s </ xm xm at ?/ the argument being The discontinuity in The continuity in C is .r. #' s. The current on the a?/ left of y is _ 2Z and interchanging x and produces C2 . In deriving this we use s rfj. 267 show C] V = due to e at y.

rjt again. V is easily tested. The sum of the two 1 is . using this. So. where P the momentum Then we generated. For an initial is state of current. say e = pP. . The time function is now e + (<l* 2 /Jp). the when A = 0.obvious. same (115) Since a. let e be impulsive.(LSA-). we convert (116) to 1 .*'* shin 7rt. V= dp d* Here we see that if the potential is made to vanish termiand s 2 is as in the la st section. j8 are as before. for we get -v = V2 i &'!/< JmGOT M G m J m+1) = *. making two/>'s to ones 2 the two time functions to be added are not the same as before. and (103) and (105) nally. as explained in get 327. The development limitations mentioned apply in Bessel series is similar.

is c introduces another term. imposed terminally.V. a nd integrate from A to I to show the potential due to the initially given state c of current. (117) showing the potential due to the initial momentum P at y. r* shin ^v LbA (G.-(tt^)x' > t ht.Rt / L at time t. In finding the current due to the initial current. . and we have also V = the result is a current c. Put P = "Ly.B/L if (see the the initial state = constant. depending upon The interpretation is that p = .x).n ct1y. Its vanishing sometimes next Section). =2 JL^l (J-qVJ-t -i-Q^i). say by the oper ational solution (113). the presence of Z in the denominator should not be overl ooked.

VI.R */ L It is the elemenis its value at time t. Uniform Subsidence of Induction and Displacement in combinations of Coils and Co ndensers. advantage. series to same way when on short circuit separately. all. Any number of coils of different II and L. rent in say C at time whose time-constant t = 0.268 ELECTROMAGNETIC THEORY. there is a 349A. but with the same time-constan t. will behave in the influence. and without mutual The same is true w hen they are all connected in make a closed circuit. simple physical property of wide generality to test the formulae Such a property can be applied to the preceding formulas with L/R. When a simple it. so that the avoidance of error rests upon the mechanical accuracy of workwhich can only be e ffectively confirmed by repetition and Under the harmony of results obtained in varied ways. way that C = C coil. has a curon short circuit without impressed force. natural tend ency to treat them mathematically only. . and is left is to itself tary case of the destruction of momentum by a resisting force varying as the vel ocity. the current subsides in such a . When formulae become somewhat intricate. CH. these circumstances it is satisfactory to be able to utilise some . by ing.

arising from the inductance not being quite very turn of wire. Anyway. This residual . it is true. They may therefore be joined together through any unenergised arrangem ent of effect does coils and condensers (without introducing mutual influence across the air). (It is also possible for resistance independently of the external combination to be energised in special ways without interference. but we do not want that at pr esent. fference of potential There is. the same sense in not occur in the application to be made.) The above being a purely magnetic property. there is a A leaky consimila r one concerning electric displacement. denser (or a condenser with a shunt).If the initial current is C in then the current in all subsides as if they were short circuited. if initially charged to There is no di of the termina coil but that the same for e . generated between any ls. and the current in the original circuit will behave in the same way as before described. Every coil wastes its energy against its own the rest. the terminals are at the same p otential. some difference of potential between parts of any one usually is a residual effect.

First. : have the same timeThey may be put inparallel without any alteration if their potentials are the same. The particular way we want now but is this. and then left to itself.. number of condensers separately if they constant. condenser will still discharge itself through its own conductance. and waste its energy therein. . and no differ ence will be made by joining the various positive trical terminals together through any (usually unenergised) elecEvery arrangements. and also the negative terminals. the discharges will be alike.K */ s is the potential at time t. The two properties may be co-existent in one combination in many ways. the ratio of the permittance of the The same is true of any condenser to its conductance.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. potential 269 V . V=V e. discharges itself so that Here S/K is the timeconstant of subsidence.

in any case. where V is the voltage of the condensers.Kt / s . and C the current in the conductor consisting of the series of coils. and C . rows of similar coils may be employed. circuit. Then the condensers are to be joined acro ss from one row to the other. But this somewhat complicates the description. if w e like. and leakance are coll ected in lumps. subject to the constancy of the ma gnetic and electric time-constants. in which the resistance. so to speak. inductance. One series is Two enough. but all with the same time-constant S/K. it follows that if the initial state is V=V constant. earth through condensers of any permittances. But the actual distribution may be continuous. and C = 0.have a long series of coils of any resistances. all withThen put their junctions to the same time-constant L/R. and is. By the preceding. permittance. quite arbitrary. . then the state at time t V=V ment sense) e. The result is a g eneralised telegraph.

be only of and this is zero for every condenser. quite local. Either the e nds must be insulated. The waste is of energy later is simply is in the leakage conductance. But if it we must take care that the terminal arrangements obey the same law. or we may put the ends to earth throughfini te length. making C = 0.= 0. and of is magnetic the force. little or big. The circuit may be infinitely l ong. There is no developThe true current (in Maxwell's sum of the conduction and displacement current. .

it is to be further noted that if the circuit is not infinit ely long. the analogy is partly only a mathematical one. . and V = 0. CH. The waste if the initial state is . or else terminal coils possessing the same timeconstant L/E. them initially having the same time-constant S/K. . the terminal arrangements must be We require either a dead earth at th e ends. as explained in Chapter IV. VI. and initial ly charged with the same current as the rest of the circuit. The energy starts magnetic and remains magnetic.270 condensers ELECTROMAGNETIC THEORY. " true current " analogy as regards is. making It follows further. and have charged to the same voltage V . and is quite local.. that if in the first case. suitably chosen. Leaving out the completion of the analogy physically. the state at time t will be C = C c-l L and V = 0. Similarly. In default of that. we require to introduce the idea of magnetic conductance to replace the real electric resistance of the circuit. V = 0. C = C a constant current. that by making a possible case of electromagnetic wave propagation in a conducting medium. in a physical manner. of energy is In order to complete the in the conductor of the circuit.

unless the mean value of the initial potential should be exactly zero. wholly there electric. because they involve constancy of the electric spite of the va riation of the resistance. the determinantal equation must conpr esently.. Also. concerning magnetic energy only initially. unless its mean value should How to reckon the mean values will appear also be zero.where the initial energy stant factor is . Moreover. and Z or C is R + Lp in with. . the last Section. &c. per unit length of circuit. 349 B. the initial state be not one of conmust still usually be a term involving the time -K/S i n the resulting potential .Kt / L in the resulting current when not constant to begin with. i n the second case. when the above conditions are complied Uniform Subsidence of Mean Voltage in a Bessel Circuit. the other. The Bessel result s previously given come under and magnetic time-constants in test the results. there V must be a term involving e . tain the isolated factor Y or K + S/> in one case.

We may Q therefore If and exhibit the y. a charge we . initially at y. solitary terms concerned. the source is h =pQ at that is.

271 may write the operational solution for the potential which results thus. we can write = Y0. If v is the partial solution for any pt then where <f>* means d<f>/dp. <j> Now. and the values of p which make l (118) infinite are the constants <jr p p in the time factor ep *. Y = gives an where & does not (120) . V-^pQ. if we know </> that isolated root of = 0.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. .

in terms of J and funct ions. 342. The denominator S0 is evidently the permittance concerned. G . QO\ (p-<r)smm7r where . 342. or V =2^^ 1 ( Jn ~ P J -m) ( Jn ~ ^-m)y A= < /-. So is the partial solution. but it is easier to evaluate in terms of J m and J_ m So use the equivalen t form (59). Then Y'0 = S0 if is the value of 9 when Y = 0. The full operational solution in this case is (61).vanish with Y.

x (124) where m^ and m^ are positive constants. We 2 shall then obtain.-+ J_m is m+1 A J_m_1 Z _o. . 1 Y = iw^K + Sp). so that there are terminal condensers with the proper time-constant. (123) . . . on the understanding that Y = Zr = m (K + Ejp). against those in the denominator of (122). as required. but Z/s a root of the that is. But It does not look as if Y = then i t has to be remembered that Jm has the factor sm So. . B 1 also congot by turning A to I and A' to taining I instead of A an d Z l instead of Z were involved in (122). m put JTO = s P m and balan ce the powers of s in the numerator and a.B' in p. .Y"1 . making Y = give . not t he factor Z shown in (122).

^ These are included in (125). determinantal equation. In the other case. and v = 0. in fact. the source write is e=pPt or the .272 ELECTROMAGNETIC THEORY. Remember that #nY and ar*Z are the leakance and resistan ce operators of the circuit per unit length. Then v = 0. because Y = makes s = 0. and give finite v. if Circuit. because m and m^ gives a root. and reduces the J functions to their first terms. The evaluation of is now easy. OH. or = 0. = 0. exist. or the P The result is that (121) functions to constants free from s. VI. or both. mean simply K = oo at the terminal. how. Practically. the above case. Then Y = really and gives a root. the charge divided by the total permittance of the circuit. we sh all obtain v The exception is when Y and Yj vanish are made infinite. Uniform Subsidence of Mean Current in a Bessel 349c. w . a short circuit will and no condenser. if it be assumed (w ith or without warrant) that Y = = usually. including the terminal condensers subsides in the proper way. In case of terminal insulation. be comes showing that the mean value of the potential that is. The special term does not Observe that if m Q and m l are defined to be Y /Y and Yj/Y. But the term v will still exist finitely if the initial state includes charge of one of the terminal condensers to a finite potential. At the other extreme we have m Q oo or m 1 = cc or both. ever. for then the mean potential will be finite. when Y doe s.

we have '* C=e =e ~ 1"'1 ' < 127 > . we have $ = %$. (126) using the proper $. we may C = <-^P. as the last. if c i s the partial solution depending upon Z = 0.momentum P initially at y. not the same Also. Then if we know that Z =0 gives a root of </> = 0. not the same as before. of course.

. the value of 6 or <f>/p when s = 0. (113). In they are (p-cr) sin WITT = m ^ ( J H+i + pJ-m-i)x(Jm+i +<rJ_ m _ 1 )y (/> .PUBE DIFFUSION OF ELECTRIC DISPLACEMENT. itself. xm or G. The signs of p and ahave to be carefully attended to. o-. The formula of derivation used is the first of (114) for J m in V r But it is different with The companion formulae to (114) a re J_ m .o-) sin m?r This is on the left of y.^. . The potential and current solutions are ter ms of J m and J_ m and the previous p and xm C.in the numerator only of (12 8). (111). The C 2 on the right side is got by interchanging x an d y in C r The V 2 is got by interchanging p and o. 278 and now we have only to evaluate . because Z = makes s = 0. 1 s d J_ m orG_ m xm dx J.

Z x = B! + Ljp = ^(B + L/>). put J 7n = 8mPtn The factor Y then becomes Ys~ 2 that is. provided Z = B + L p = * (B + Lp). we have one or more terminal earths or short circuits. initial state includes a finite current in the terminal c coil. Lastly. If ei ther of nQ or n is infinite. or both.1 Applying (127) to (129). . it expresses more than a disconnection. . . including expresses the that of the terminal coils. we see first that Z" does not show But as before. (131) so that there are terminal coils having the proper time-constant. for L (for instance) is The result is c = 0. mean The denominator is the total inductance of the circuit. we obtain where 2m = n-l. evalua ting by the first terms of the J functions as before. or n^ = 0. Z" 1 showing that Z = giv es a root. When n = 0. . and c value of the current throughout the whole circuit at every moment. but c can be finite if the also infinite.

Practically. . a disconnection produces = in another way.

equation (123). It will be observed in the above evaluations th at the quantities A' and B' which occur in p and or. ( apart from the special factor conditions. this part of the subject it may appropriately be pointed out that this property can be generalised in many ways by appropriate terminal conditions involving ele ctric and magnetic energy. if = ra --A o w2 Z .274 ELECTROMAGNETIC THEORY. we Divide by obtain Z . There is no terminal coil as a reservoir for magnetic energy. Z or Y above considered). and the special te rm involving Z = does not exist. CH. then. are reduced to s x constant or s~ l x constant by the special terminal This makes p . (134) a condenser in sequence E +L = Z O ?J ?i and K +S p=m Y. . This says that between the terminal and earth. Z = R + L p + (K Z is a coil and +S 2>)-i.cr. We have merely to make<Z /Z and Zj/Z be functions Let of s 2 It is suffi cient to illustrate by an easy example. become a fun ction of s 2 In general 2 but of p. So now in concluding it is not a function of s . VI.

(4).is made a function of s 2 (with the . To be a function of s 2 n and HIO must be constants that is. Similarly we may make Z X /Z a function of s 2 In all such ca ses p a. and a complete development c an be obtained. so that the time functions are known. j known manner.. for the later notation. I have endeavoured to smooth matters. in function has the opposite sign to that employed " Electrical Papers. possible extra factor). In 330 and 334. p. 243. See equations ~K m p. [NOTE. w = R /R = L /L.s . K (^). which makes K^qx) be the negative of the derivative of (3). the function K^qx) is defined to be the derivative of J^. and (18). and from 336 to the end have The Gm (sx) ." for good r easons. Q(qx). The function (gx) is always positive. 240. and its roots are calculable by tables of Then there are two p s to every s 2 in a Bessel functions. an d m = K /K = S /S. (135) . in 336 and But there are good reasons after.

] .my employed that standardisation which experience in the complicated relations of B essel functions has shown me to be the best and the easiest to follow.

to the best of ray ability. paper to the Physical Society. Chapter me if I II. to revive an old labour by di recting attention to the irrational nature of the B. Mr. has strongly emphasised the importance of the change and the sim plification that can thereby be made. Heaviside. system of units (once so much praised). by reason of the occurrence of 4?r as a facto r in the specification of quantities which have no obvious relation with circles or spheres. for instance. that is better than 47rnC. In 1891 I endeavoured. In theoretical investigations tliere seems simplified formulae may come to be ad opted fj. [Xaturt. I.275 APPENDIX C. but that the practical units should be reformed. see. Lodge wrote asking had any practical proposa l to make. : resulted THE POSITION OF There . RATIONAL UNITS. a growing body of opinion that the present system of and magnetic units is inconvenient in practice. July 28. p. that the ampere turns electromo tive intensity outside a charged body might be <r instead of and similar changes of that sort . Williams's recent 4. and advanced arguments to show that not merely should the presen tation of theory be altered. some probability that the . A.. It is felt that the number of lines from a pole should be m rather than electric " " the present Qirm. and the Preface). 1892. in his articles in The Electrician and elsewhere.] is I believe. The following letters . 4?r IN ELECTEOMAGNETIC UNITS. 292. In 1892 Prof. Mr. (S ee Vol.

&c.being written instead of 4n>c.. even though it be only a numerical change but in order to find out what practical proposal the supporters of . the volt. and is Tc instead of ~ Iv . but the question tical outcome of whether it is or is not too late to incorporate the pracsuch a change into the u nits employed by electrical engineers. For myself I am impressed with the extreme difficulty of now making any change i n the ohm. .

this has happened. which pervades it from top to bottom. construct a consistent system of units. come up at Edinburgh for OLIVER J. Heaviside toHis repl y I enclose and would merely say further that in all. we c ould. irrational one. a very important question. VI. CH.. whi ch must. but to the fundamental principles electrical units will MY DEAR concerned.276 ELECTROMAGNETIC THEORY. I have considered in my . but also a very practical scienc e. becoming not only a master science. and how to cu re the evil. LODGE. Itsunits should therefore be settled upon a sound and philosophical basis. I am glad to hear that the question of rational be noticed at Edinburgh i f not thoroughly discussed. involve the quantity same peculiarity. come to a head and lead to a thoroughgo ing reform. the redistribution of 4?r had in their mind. If we were to define the unit area to be the area of a circle of unit diameter. I do not refer to practical details. probability the general question of units will discussion. which may be varied from time to time(Acts of Parliament notwithstanding). LODGE. It is. I No volume of a parallelepiped would and various derived formulae would possess the one would deny that such a system was an absurdly maintain that the system of electrical units in present use is founded How irrat ionality. in my opinion. or the unit volume to be the volume of a sphere of unit diameter. sooner or Electricity is later. on such a basis. inquire. But the area of a rectangle or the IT. I wrote to Mr.

I think. after giving a gen eral outline of electrical theory in terms of the rational. when. This is the main reason why I attach so much importance to the matter . and of the dimensions of the electrical quantities. if it continue. the question How proceed to the third stage. mass and time. this is quite easy. is an encouragement to shirk a duty ." I adop ted rational units from the beginning. unfortunately. it themselves should be is. likewise. and again in 1891. in my Electromagnetic Theory.upon a similar papers first in 1882-83. the common feeling of respect for the labours of our predecessors. But taking a far-sighted view of the matter. But the duty we owe to our followers. pointing out their connection with the common irrational units separately. presuming provi sionally that the first and second stages to Salvation (the which arises. when. and theoretical papers and treatises may. it is not merely one of abstract scientific interest. Apart from the size of the units of length. we have the following relations . as is. not at all certain at the present time. however. with great advantage. This must involve some temporary inconvenience. I thought it was hopeless to expect " a thorough reform . very desirable that the practical units ration alised as speedily as may be. and irrespective of the reform of the practical units. as it merely means working with rational formula? instead of irrational . Reform ation? Theoretically. but of practical and enduring significance . th e prospect of which. be done in rational formulae at once. is. should be paramount. to lighten their labo urs permanently. Now. go on multiplying with ever y advance in the science and its applications. Awakening and Repentance) have been safely passed through r however. for t he evils of the present system will.

[LjJxlO [S.PUBE DIFFUSION OF ELECTRIC DISPLACEMENT. if we let bsolute rational enitors. an that ^=1 in ether) . [CrJ~[Qr] The next question is. First. Also let the is referred to. Then we have [[V][[B. that I think there is no finality in things of this sort. electric force Let x2 stand for 47r. . what multiples of these units we should take to make In accordance with your request I give my ideas on the practical units. induction B. and second. electric current C. magnetic force H. suffixes r and i mean rational and " " presence of square brackets signify that the absolute unit irrational (or ordinary).gramme. then we d the convention 9 [R r ] x 10 =new ohm =x? times old. permittance S. 277 between the rational and irrational units of voltage V. resi stance R. premising . and let the E. the subject.] x 9 the rational practical units be the same multiples of " the " a units as the present practical units are of their absolute prog would have (if we adopt the centimetre. electric charge Q. however. inductance L..][C ]_[Q i ] .

or reciprocal of permittance. and the unit of elastance. . I do not. x old amp. mac.9 =new farad = a:. vo lt. however. the multiplier varying roughly from 1\ to 3i. 10 x old watt. but not greatly larger. and consider that the following system is preferable : 8 [Rr] x 10 = new ohm = =new mac = =new farad -x old ohm. amp. x old farad. x old mac. = old watt. 8 [Lr] x 10 8 [S r] x 10= i? = new amp = 8 = new volt = [VyJxlO 108 ergs = new joule = 108 ergs per sec.= new mac = x2 x~ l 10. = ne w watt= [Cp] x 1 x x old volt.2 1 [CJ x 10~ = new amp 8 [Vr ] x 10 107 7 = new ergs = new volt --x joule 10 ergs per sec = new watt = old joule. think it at all desirable that the new units should follow on the same rules as the old. all larger than the old ones. 10 x old joule. It will be observed that this set of practical units makes the ohm.

this is done away with.278 What. of. it seems rather an improvement. of course. 10 7 10s and 10. viz. in passing . ELECTKOMAGIs'ETIC THEORY. I do not see any objection to the change.. I attach CH. (But transform ations of units are so treacherous. VI. power units . the old joule and wat t by 10 is. however.) I . instead . 108 . system have described. no less than four powers. particular importance to is the use of oneof 10 only. that I should wish the whole of the above to be narrowly scrutinised. a necessary accompaniment. Though not important. 101 . needless and (in I my regard this peculiarity of the common system as a In the 108 experience) very ve xatious complication. and still the practical The multiplicati on of electrical units keep pace fairly with the old ones. as in from the absolute to the practical the common system.

~ =ne\v farad =x~. farad. .. 9 = new amp = __ x = new volt =lQx = new joule =10=new watt x old amp. x old volt.It are : is suggested to make 1C 9 the 9 multiplier throughout. x old joule. I think this other objections. =10 2 But detail. x old mac. x old watt. and the results 2 [Rr ]xl0 rL r ] x 10 = new ohm = =ne\v mac 9 x old ohm.x old x 1 9 [Vr ] x 10 10 ergs 10 9 ergs p. 9 x[S r ]xlO[C. sec.

and consequent change of op inion and formation of opinion. so far as I know. as some of the following will show. as may be seen in contemporary journals. they were so ill-advised (in my opinion) as to persist in their e rrors and announce that there did not appear to be any reason why their practica l units should not be legally adopted Thia (I have not the document by me to giv e the exact words). A. I do not know that thsre Nothing particular seemed to result. between 1891 and 1895. Paignt on. things I would emphasise But : 8 Next. Au gust myself. 1892. as. The question of rationali sation was apparently nowhere. The discussion arose out of Prof. was any discussion of the matter at the Edinburgh meeting. Abo ut 1894-5 too. was accordingly done. A. by proclamation. so to speak. including. took no f ormal notice of a serious matter in which they should be so much interested. Committee. for example. however.Two system makes the ohm inconveniently big. Lodge's Report on Magnetic Units. July 18. the Awakening. had taken place. and has some I do not want to dogmatise in these matters of First. Committee and others. . for opinions. Devon. It was reprinted in The Electrician. The development was apparently only in its first stage. The B. In the meantime. . employ a single multiplier. under th e Royal Arms. a remarkable diffusion of knowledge on this subject. which was printed and circulated amongst members of th e B. 10 OLIVER HEAVISIDB. rationalise the units.

along with letters from Mr. to be a little mor e enlightened than they are. suggestiveness is obvious. I believe. G. January 28. Everett. beautiful state of properly. and perhaps many more. meetings and at Chicago. Lodge : MAGNETIC UNITS. by open admission of error. 1895. 279 1895. who wrote on Units in the Electrical Engineer lately. you began it. system will be Even in something for them to laugh at and damn. if it is not already. DEAR LODGE I bad some idea of marking your paper all through. A. D. 449.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. but I gave it up when you got immersed in the 4?r muddle. Swinburne has suggested in Nature that I tive is that am very likely wrong in this matter. Perhaps . and so is FitzGerald. 2. and by properly Th ere is no way out discussing it at your B. It was not considered proper to circulate my own opinions of the Report amongst the members of the Committee. in the way of simp lifying it mainly. But then you are a rationalist. A. 1395. When prac ticians get of the muddle than by my radical cure. and recommendation of a revision. Bailey and Prof. and Larrnor. of which I could give examp les in the theories of eminent men. p. F. the B. it has been the cause of much mischief. Hence p. in the their separate publication in The Electrician. J. August 16. 511-12. form of the foll owing two letters to Dr. of Glasg ow. Committee I mean) are in a : muddle by reason of refusing to complete your work say you did not know you were wrong till after the " cannot put it on to International Conferences . A. Paignton. you lega lisation . What is more suggesMagnus Maclean. and the blame is yours all the more so from your refusa l to put it right. " You cannot pure theory. or even to make the beginning of an attempt to put it right. though perhaps equally unwarranted. You (the B. the condemnation that my had the assurance to dismiss my reform with The geographical reasons were unwarr anted.

Gaussage especi ally.. " on account of the age . and in fact often become Practician s are quite up to circuitation . and complete You cannot get out of the mudd le in any other way. (Sausage !) I do not admire them myself very much. because the latter are exceptional. " " I think may be dropped altogether. &c. however (or ot hers). not for falls of potential (electric or ma gnetic). then. specify them as H the electric force and the magnetic force.M.M. meaningless and quite wrong. A. not gau ssfall.M. and for M. or along a line. and you are all right." but I took voltage as I found it. or from a to b. that they should speak of parts of the real electric force. and dyna- .F. I dare say practicians will not like them. is the sum of the elementary effective I think. Then why do they not do the proper thing. for instance. in a circuit. Committee are rationalists. the E. which I do not like at all. their work properly ? Voltage and gaussage. and extended the meaning. (How about voltat ion and gaussation ?) Accepting these words voltage and gaussage. I mai ntain that intensity of E and are forces. it should be noted that they stand for E. dynamically (generalised.F.a majority on the B. of course) .F. the gau ssage or the voltage in a circuit..

. How put B in terms of H ? It seems to me . thus. in which. VI. I would rather not concern myself with such a bad job. when you take in iron and do not keep to very small forces or small variations in big Her e the practical requirements of the practician have to be consulted undoubtedly.280 mically sound. ELECTROMAGNETIC THEORY. H=F+h the current density. About the mea ning to be given to inductance. when you bring in ATT and 10~ x and 10/4ir. it would be necessary to discriF only to be free. I would not presume to criticise. In 192 of " Electhat there is no definite connection between " tromagnetic Theor y I have tried to indicate how we may perhaps come to H a good magnetic theory. Their factors CH. such that the curl of F is minate . and that we have -curlE = B. that they will be sorry for afterwards. . but if they are let alone they may do it in some way forces. I think your opening part could be simplified a good deal. &c. corresponding fluxes. The difficulty seems to me to be and B. But as regards the extended meaning of ft : suppose we do t ake a definite connection between H and B. however. D and B in the energy density are the After that. and so forth. ignoring hysteresis. permeance.

and H=c-V dt . It goes well in the Then. dC so that . so that dv H=fcdv. v the temperature.) is (This is like saying that the volumic heat capacity of a body c = ^5. circuit. (Similarly D=/cdE. H being the heat diffusion of heat.) per unit volume.that the best theoretical way would be B=*! *?=/* dB (It so that = /A clH and B=//idH. we should have in a (N = total induction). similarly.

not presume to say. we should get for a T =/C But whether this is likely to dt =/CdN =fCLdC. The HB per unit volume would give T=HB = EACH. . I would Perhaps they do not want any of these quasi-scientific ways of trying to represe nt facts which are not definite in themselves . i.e. and by volume integration.dt activity L= dC =andN= 4/LdC. or T=/HdB =/Hf?dH = fu rfH coil. <2B/c?H is not a mere func tion of H. be convenient for the iron people..

is not a new opinion. others. found that they agreed with me. on the precise theory. not suffer ill effects. It On consideration.PURE DIFFUSION OF ELECTRIC DISPLACEMENT. but only the beginnings of one. There is (That. so Practicians swallow more important considerations in general theory. so I suppose they do If I were a practician. I again obviously theoretically recommendative. I should very_jmuch_ like to see a goodjjiegry of magnetism . I do not think that quasi-official or conferit. Now the effect of theory only a side matter. though. is It is an excellent theory. Paignton. 1895. and sacrifice rigour to expediency. in fact camels in their "predetermination work. DEAR LODGE In my last. a Commercially. They do their swallowing wit h complacency. is doubt no theory of magnetism. commenting on your report. like many much so. percentage allowances for wast e by hysteresis and leakage (which is a big camel). Further than t hat. I had doubts as to whether the dB/dH=fi definition would be convenient for practicians. ential decisions are desira ble on moot points involving unsettled theory. February : 281 3. " It is based on theory. too.) For the case is peculiar. secondary phenomenon. however. it assumes that one is apt to overlook other and exaggerated impor tance. only the application iron on the magnetic field is in gen eral limited. I would swallow came ls too. Not being a practician. but is modified empirically by characteristic curves.

as a secondary matter ? together of independent variables. It when the and smaller seems quite absurd to take B/F ticians call . I want to know how the waste comes in in different parts thereof. steel. with energy stored and no waste.}. in a paper by but make the loop become of insensible area. It is imaginable that quite pure iron would Say as in the figure.with variable yu. and leave practicians to work any way they if I like. with fictitious make-believe permeances and reluctances. But owing to the extreme steepness of the curves we see that there is a very large intrinsic B when loop is made of smaller H is zero or small. and take the median curve . but large hyst eresis. not enough to know how much waste there is in a cycle . system. a regular p = dB/dH. waste of energy. and on what it depends. is There is a curious case in Mavor.er. It is suggested to ignore the loop. Is it invariably associated with a change of the intrinsic m agnetism (intrinsic pro tern. The stuff is called last week's Electrical Er><jine. As it is. Hysteresis is the theoretical trouble. but if we do that we have B vanishing with H. and lumping Along with this. and large p in the ordinary ssnse. Now it is said to be chemically pure iron. and this will be so even area. It gives relatively small waste. or only accidentally.

H=h + F. between H and F in theory. But say . or dB. Here F is what the pracit belongs t o the coil . because. I think this is the right way to look at it. H much is independent pro first. we must have at least one other variable quantit Perhaps h would be enough for a practical theory under limitations. curl F = C . We B y. As is not a function of F only. theory (greatly simplified in expression) of induced magnetisation. ed F so must allow for distinction (H -h) = curl F = C. and also the more modern view of the this way satisfies Poisson's old . not curl H. or how much of B at any moment is connected with and how that curl to represent permeability. tern.'dF either. But we ne to know how h varies with F.

not in a regular or constant way. L proper realisation of the should think Ewing ought to have the material at his disposal. letters . and different and to get a sort of normal true inductivity curve (quite from the commonly assu med) and so come to a sort of theory. CH. It comes cal ly we do not have A as a given datum (constant or variable). and if h could be considered given (as a function of the time f or instance) with a given connection between B and H.induced and intrinsic together (intrinsic constant though) (also simplified in expression). and it seems.282 ELECTROMAGNETIC THEORY. same. a workable theory. in through the aqti on of F. putting the iron into a peculiar state. still For the symmetrical loop is only got after many reversals. we should But practihave. and the facts. VI. but I am inclined to think there must be waste in the initial settling. a generalisation of the present. to be able to discriminate between B total intrinsic. even if there is none finally (in a suggested pure iron). If there were no waste (or it were insensible) it might b e the same.

of which I give a few notes. He also misused words. SYDNEY EVERSHED was apparently put i n a state of fever by the Report. Mr. I condense. THOMPSON encountered the great 47r. A. he thought it impossible to EWING discussed the Report.The Report and above discussion in TJte Electrician were followed by an interesting (summer and autumn of 1895). J. I therefore only notice (in general) the opinions on the question of . rationalisation. ATT. S. P. Lodge's procedure was wrong. KINGDON said engineers would be dismayed by the Report. and not about rational units. Prof. abolish Prof. . and seemed to be amazed at my audacity in actu ally proposing to Mr. and that the units should be rationali sed first. but did not overcome it. Of course. I hol d that Prof. Prof. Also. Lodge's report mainly consisted of an attempt to systematise magnetic relations and units without employing rational units and t he discussion was mainly upon it.

t the new units. it was inch within 2 per cent. to be ad opted. formulate a theory of magnetic induction.PURE DIFFUSION OF ELECTKIC DISPLACEMENT. 283 my Mr. producing a feeling of helplessness. No importance t o Used to it. T. and do it rationally . A. . he c alled me a ringleader. had proved successful for electrical units was somewhat [Xot at all. Prof. an unmitigated nuisance and added another complication in " the adoption of a system to be the standard . ESSON was thoroughly unsympathetic with Prof. A.] Mr. See below. thus practically declining invitation. A* " for me. he says. Lodge. B. W. S. dynamo and transformer. H. LODGE said the reluctance to extend to magne tic units the same ! " " the 4?r question. E." and said I ought to be both proud and that my work (whic h. W. that most practicians or engineers only -want to be let alo ne whilst more scientific persons want to go further still. SNELL thought 47T/10 was really of little importance in dynamo Mr. V. F. He asked whether my system was ever likely practical men. will endure) would never be degraded happy by the practician by application to things Mr. ANDERSON remarked that the C. G. useful.] Mr. remembering the two objections. ATKINSON said he th system was best or not though that we were threatened with in which air or ether is not . EAVENSHAW thought. with regard to not generally known that 2H = ampere turns per He was offensive also thought practician Mr." [This is quite new.] ought it was too late to djscuss whether the B. [There seems to be something brain-paralysing in the sort of treatment as surprising. is not a rational system. subtance. . L. B. W. SUMPNEU said 4?r/lO can't be got rid of.

] Mr. but there are many And Prof. but a this little men very important one. ADDENBROOKE agreed with me thoroughly on the 4nquestion. [Truly there Besides. B. it. FITZGERALD entirely agreed with worth doing. but did not at all agree that a change is possible. W. it is is much to do. me that it is a great misfortune that the units have been wrongly based. SAYEES did not remark on it. and discouraged men from wasting th eir time in endeavouring to bring it about when there are so many other things b etter Prof. Mr. FitzGerald's argument agaiust doing little seems weak.work. L. Said that in 10 or 15 years it will probably be found advisable to sta rt with a complete new set of units on a rational if basis. G. . F. G. not such a to do matter matter in the long run.

even though it be H. Mundurn and that some day we shall wonder how th e B. discussion. called attention to some aspects of the mat ter of magnetic units which might be overlooked. the writer was not sure that I am not right aft er all. and turn to the man who tried to save u s from ourselves. i t must not be. system -was ever blundered into. A. and before the B. v. No it shall not be. the inclination prompts to action. A. The Electrician summed up in a leader. At the same time. LODGE. and says.[Much easier now. . . referring to the above discussi on. The practical man holds up his hands in horror at my proposal.] Prof. .

FEDERICK BEDELL said it was too soon. AYRTON considered the question was not what ought to have been done 30 years ago. A. SILVANUS "One of the most more progress was made. It is of the above cannot fail to notice the gradual change Then came the B. EVERETT agreed with me iu theory. The reader in tone. Prof. JOHNSTONE STONEY'S printed remarks do not bea r on the matter. but what would be best under present circumstances. discussion. Committee did not boldly face a difficulty which became greater by delay. rationali se now. in his opinion. gettin g quite favourable. Moreover. withou t units. [So much the worse for the astronomical units but I am unable to see that there is much contact at present between astronomica l and electromagnetic quantities. VI. . Not quite ready to ta ke up Prof. or make preparations. and ad opt at once my suggesHe thought it was quite possible to make the tions as to ra tional units. the rationalisation question yet. J. [I think he meant as regards the proposals in the Report. They are practi. the change But because we might soon have to remodel would not b e very difficult. when striking features Prof. Dr. CH. was the leaning shown by many of the sp eakers towards rational units. D. physically doing it rationally.] Dr.:284 ELECTROMAGNETIC THEORY. but objected that the harmony betw een astronomical and other [so-called] absolute systems of units would no longer hold.] E. [Exactly so . the whole system was a very strong reason for taking a minimum of action now." THOMPSON was impressed by the importance of rational and practically. W.

said some object to the presence of TT. TREMLETT CARTER did not think posterity would admire the present " " All agreed that the rational system was better. PREECE. without any arbitrary and unnecessary c onstant]. discussion. Perhaps the best time would be when the real nature of the ethereal constants became understood.. A.R. and " would re legate it. and applied to el ectric and magnetic and astronomical units. There was a little more in The Electrician. LODGE feared be brought in. had so arranged the units that it should be so. if astronomers object.] Prof. F. Maxwell. Consider what heaps of old weights and cally independent. Committee who objected to E = RC and said it should be C = E/R. H. He. Mr.change now.S. and measure s." to a less intrusive place. A. Mr. by mere artificialism. and the other members of the Committee. If done at all. and Prof. Dr. measures there are.] This finished the B. But I don't see why astronomy should It may mean the Greek Kalends. . W. and should be system. [Much easier. [This reminds me o f the member of the B. It is not necessary. [This is cold water indeed. but was interested in seeing how many seem ed to favour it. it should be done thoroughly. PERRY was sorry to think that the and that they enter into the daily life of the multitude !] it was too late for so radical though desirable a change. No mo re difficult than to introduce the metric system of weights adopted.

had been exaggerated. is the matter to end here ? Sure ly not." Nothing required to be added to the reason s adduced by me. W. HAWKINS expressed his belief in the ultimate and perhaps not very far distant victory of my "radical cure. I differ. do not let the matter drop after such a s uccessful beginning. C. based upon his dimensional views. 285GRAY thought the inconvenience of PERRY had no doubt whatever. SAYERS asked whether the distinction between H and B is not a relic of the action at a distance idea. Now. first. Prof. air. from initial ignoration to th e consideration of details of Reformation. is about all.] Mr. F. and it is instructive as well as somewhat amusing to see how rapid ly the three stages to Salvation were run through. JOHN " The change must come. sinner should reform Then matters would be greatly smoothed for the others. A." [Bravo !] Mr. Prof. I would say to all and sundry. and also whether it should not be M= in air. beginning with an [But that is international I think the agreement. but keep pegging away till the actual demand for the reform . BAILY proposed yLc=47r/109 for plest way out of the 4ir trouble. original a very doubtful point. and and thought that the simBetter soon than syne. Mr. He also sketched lightly his idea of the way the change should come. W. though sensible.PUKE DIFFUSION OF ELECTRIC DISPLACEMENT. B. Mr. 4?r. G. G. WILLIAMS pointed out an auxiliary argument in f avour of the rational system.

J. Onefor way them advanced somewhere. both cases. rational example. of expressing . there have b een many other expressions of opinion. Committee will do anything without It meets every year. Prof. after all. believe. has commended the simplicity of theof displaying the electromagnetic re lations .r is must come in also curious to note the action of the dynamo and transformer. "But. That Thomson. new light for some people. Of course." As of this if it argument. Dr. Kingdom. than the above on the question of rationa l units. and parallel modes * them.is pressing. J. the same ideas should be in. and practice to diverge. Of the progress of rationalistic principles outside the United I have next to no knowledge. It is not likely that an old. is a marked exception. 1. didn't matter where! It 4. But the following the rationalisation question in a a natural tendency for theory is may put There keep this action in system. To divergence within bounds. but I doubt whether Cambridge men are favourable to a change. however. institution like the B. He was an early conv ert. A. pressure. This can only be secured by the ration al There may then be identity of ideas. It is difficult to advance any new argument. Fleming.

See Chap. Vol. II. I.. .

.286 CHAPTER VII. In order to avoid introducing the idea of fractiona l differentiation from the theoretical standpoint.*!. I took the value of pU There is no as known experimentally.. . according to This ordinary notions of differentiation. was unintelligible. it may be any function of the time. and normally subsiding then series of reflected waves. At the very beginning of the treatment of the subject of diffusion there pr esented itself to our consideration the execution of a differentiating operation which. the generation of a wave of It is necessary and inevitable . 241. square. The the operand being 1. was the operation concerned in the function . it is found to facilitate working. diffusion. waves and normal series. ELECTROMAGNETIC WAVES AND GENERALISED DIFFERENTIATION. root of a differentiator occurs in the fundamentals of the physical subj ect. equation (A). 350. For the general formulae for diffusive waves were obtained. Determination of the Value of p$l by a Diffusion Problem. without this reference to a know n result. But. when studied. where p is the time differentiator. and 1 is the special function of t which Instead of is zero before and unity after the momen t t = 0. namely. also. we should be justified by the consistency of the results obtained by the assumption that the function j &*l was of the form employed. and finally these were converted to series of also carried out for Bessel The same process was states. question as to its value that is settled by Fourier's investigations in the theo ry of the diffusion of heat in conductors.

will not be overlooked. made occasionally of when it turns up^ The But theory of the matter. Then. I intro. Now. formal and logically-arranged treatment may seek it elsewhere. S (resistance and permittance per unit length). equation (7). of length /. we have to find what p\\ means. we can work out this prob lem in Fourier series. with is constants R. (1) If e is constant. and then proceed to the limit. and be earthed there. and find it if th ey can . i C duce one example cf the experimental discovery of the meaning We found in of />* !. let the cable be cable.RSp. the current produced on the spot was expressed by C = (S^/R)^. subjected at its be ginning to impressed voltage e. that when an infin itely long cable. or else go and do it themselves. first for a finite Thus.ELECTROMAGNETIC WAVES. At present. and the generalisations will t ake a subsidiary Those who may prefer a more place as they are suggested. if s 2 = . founded upon the old diffusive methods. 240. merely for the sake of com ple eness. the primary object of th e chapter now beginning is electromagnetic waves. 287 There will be a good deal of use fractional differentiation in the following. . also.

2. Observing that the step from one s to the next is TT/I.siusl is it the potential V at Z due to e. in the summation. The algebrisation by the expansion theorem ( 287. and that it becomes infinitely small L when / is made infinitely great. &c. s has the values ir/l. 3?r//. or in any equivalent way) makes makes V=e at the beginning s-f/RS /Q\ (O) where. we see at a glance that ./I. equation ( 43). as in 265. equation (1). since and V = at the end.

= cc converts (3) to v That is. the previous finite step becoming the . * 77 J there is a conversion of the Fourier series to a ir/l definite integral.

we . and -o then putting x = 0. taking the square root. x = 0. This makes (5).ELECTROMAGNETIC THEORY. CH. also well-known). it evaluate Perhaps the easiest way to an ingenious device. Comparing with see that (1). is thus (by : *2 < o dx x Try o T* (7) So. we (6) which is a well-known integral. infinitesimal step ds. and removing unnecessary constants. is got l by C = .R.(dV/dx). VIU The current at the beginning.

that the operator C/e is rational when / is finit e. in passing. reduces to the irrational form in (1) just when the Fourier series passes int o the definite At the same time the infinite integral. we have whole tive or negative. Elementary generalised differentiation. by making I infinite. should be noticed. when On m . in a thousand. The above is only one all way any formal proof that sarily lead to the same It ways I do not give properly followed must necesresult. the basis of the result just obtained. posipossession is positive.arrive at the required result. we are in of the value of pm +*l. Value of p m l when m is integral 351. or midway between. when m is any integer. Thus . which series of waves involved in the Fourier series reduces to a single wave. (8) 2>'l = (*)-'.

3. T.differentiations to perform upon p*l.8 r* .5 (9) 2.2 ** & . 1. 79*1 P= 1. For example.

8. tions from When m to t.5 . 289 integraand so on. IT* and so on.ELECTROMAGNETIC WAVES. i>l 1 = 1 i. is negative we must perform Thus. We also an integer operand be know the value of p m +-t n where n is integral.3 TT" 1. 1. 1. or is + J. Say it is integral a nd positive. Let the t n j n where jn 7i .

*>-"!=-. /n . / !. . \n (11) and now introducing the index we get p m+ ^ = tfii-K+J J = n ( !-_!_ (*ty (12) Here in (9) ?w -n is integral. Then if.3. n. so and . 1 w 1 ?. .2. *----=.\n is the factorial function 1.= >i .

and introduce the datum that the above results in generalised differentiation are valid with the extended mea ning of n.(10). For example. as well as when it is an integer. for instance. &c. Consequently and (12) are also valid when n is = 0. the t is also zero. 0 (13) with the addition that value of n. that n (11) is oo its 1 value must be fixed for any one = 1.1.&4 f-f s-*. It is really an impulse at y>l moment = 0.l (15) We . and for all negative integral values of n._*. -. and this is zero. it will be observed that if we use the formula (13) when n is an integer + J. provided t is positive. Comparing results. Thus. Now we have the former cases again. as the fundamental property of \n is w = wp-l. Also pi = t~ 1 / . tAic i. It follows that = 1 also. unless t 1 negative.

shall now have (16) .

mathematical excursion we may 352. L. is not difficult. cable have the four constants R. the factorial notation being convenient for showing the system netic problems without going further into generalised differentiation. in the notation prebeing the inductanc e viously employed. or is midway between two consecutive and the same as regards m. Then the . K. and for the expression of results.290 ELECTROMAGNETIC THEORY. merely a convention or defi nition. but Let the generalised to include self-induction and leakage. but we do not want it at The above can be applied to numerous electro magpresent. : C = (K + little Sj*j*(R + L^p Elementary Cases by Inspection. as far as the above is concerned. S. That j-J is TT\ involved. any integer. when n integers the meaning of is. The extension to the case of n being any real number. positive or negative. CH. Cable Problem (1). is . VII. After the above return to the physical problem out of which it arose. the additional L and and leakance per unit length.

The previous mode of found to be enormously complicated. without the circu mbendibus involved in evaluating complicated integrals by rigorous methods. only R and S are finite and S are zero. Now this is a far more developed case than the former. equation (12). of algebrising attack will be we have and Notice some special cases first. Ways it have to be found. (18) increases to infinity according to the square root of the time. and is purely . This is a curiou s case of leakage conductance and inductance only.current produced by e impressed at the beginning of an infinitely long cable is K that is. whilst L we have a similar case. I. 221. But if K are finite. See Vol. Thus If R C = (K/Lp) = 2(K*/Lir). the former case. by using the value of p~~*l. Bu t we can find what (17) means pretty straight out from itself. E is generalised to R+L and Sp to K + S/>.

The current a magnetic problem. .

again It now holds good always. putting p = ni in (17) will produce the simply periodic current that resu lts when e is simply periodic. Again. of C when e is the speed of propagation. no steady state is reached. whether e is steady or not. p Here (R/K)* is the s tate C = (K/R)*. in I. of course. p = x in (17) gives the initial value suddenly jum ps from zero'to a finite value. In the extreme case R = 0.. This is the distortionle ss case. Thenp goes out from (17). which Further. and need not be repeated. reduces to C = (K/R)*e. the cable behaving towards the impressed voltage as though it were a mere resistance. let R/L = K/S. . Finally.ELECTROMAGNETIC WAVES. by reducing it The developed formula was given to the form C = (K' + &p)e. the steady ultimate effective steady resistance reckoned at the beginning of the circuit. or (LS)~* In the theory of a plane electromagnetic wave the equation corresponding to (17) is if ? ^-(T^V. Vol. all four constants being finite. 221. viz.. where E and (20) c H are the electric and magnetic forces. 291 = in (17) produces Next. 209.

the constant g. same way. vanishes. It can be treated in the electric and magnetic respectively. except as regards -explained in Chapter IV. and. e be constant after t = 0. it represents the same problem physically. in and consider the case which K is the only constant that Then _. Now let Convergent and divergent results. (21) . Two ways. and zero before. 353.and /* the permittivity and inductivity. Algebrisation when e is constant and K zero. in fact. This was (2). k and g the conductivities.

ifa = R/2L. The suggestion It will to employ the binomial theorem is obvious expand the operator in powers of p. an d so substi- .

tute a series of easy operations for an Thus. But the binomial theorem furnishes an other way of expanding the operator in (21).. 1<"> So the immediate by the third of equations (11) above. merely to perform whole differentiations upon it to produce the solution with the same directness in this We form : _ Gmf J_ f . Thus. The straightforward and rapi d way of getting the result is remarkable. in rising powers of p. VII. CH.. unintelligible one. e/(Barat)l. viz. a convergent solution in rising powers of the time. result is. " Here we have only whole integrations. viz. have. _ i ~Lv Here we know already the value of (j?/2a)fe.292 ELECTROMAGNETIC THEORY. expanding in rising powers of a/p. therefore.

C= JUt! c-*J r*). I 1+ I2 .{fcraft)*. (27). . Equations (23) and (25) are equivalent. should represent (af) is H to be numerically equivalent to 2I (<). It is easier to calculate except when at is so small as to bring the point of convergence too near the beginning. I 28 a a Lv This is Sat 2( (8a) 8(8tt*) a divergent series. So much the better. we see tha t (25) is the same as C= where ^U-<H (0. Comparing (25 y with (3) 336. (26) shown the divergent zeroth Bessel function which was Therefore (23) r being convergent .

I . remark. The explanation is to be seen in the other way . Besides the v ery direct way of getting the results (which. take the special case of the same got by making L = 0. But both are is di vergent when a/p is numerical.ELECTROMAGNETIC WAVES. and cannot now escape from pi. may noticed. This is curious. there are several points to be Thus. or (22) valid. The question will ari se again. we find the us e of the binomial theorem is justifiable. the As regards the practical use of the other to a divergent one. Next. are quite correct*). that the convergency of the series in powers of p Either (24) obtained does not enter into the question at all. The lesser seems to contain more than the greater. 293 That it does may be verified by multiplying together the ordinary series for the expo nential and the convergent Bessel function (27) then becomes (23). to substitute an infinite series of separate integrations or differentiations for the operator involving the radical. latter. though one gives rise to a convergent final result. we we reduce (21) to the form (1). see Notice further that we obtained (23) from (21) through But if (22) without any use of fractional differentiation. 335 for the present.

This makes the last at become + *-)*<-. Change of Operand. proobviously. when function. by putting in the exponential factor at the beginning. Observing that in the form (27) we have an exponential factor. The reduced form of (24) is (1)./*!. / (29) t * How know ' that 2 )i iu the I (27) is right ? Because. 354.of expanding the operator. We have . This is the way. (1). a third mode of algebrising (21) is suggested viz. vided we simultaneously c hange p to p -a. that we can escape from . the result is the complete . and both involve 7>i It still remains remarkable. Third way. (28) Now. and so evaluat e it by generalising the problem involved in (3).. however. here t may be shifted to the right.

terminal condition. as well as the This generalisatiou will occur further on.is turned to wave of current entering the cable. . The partial characteristic is satisfied.

e at CH. there are points to notice. Compare with it 265. Now expand by the theorem and algebrise. The transformation (28) to (30) is. .294 ELECTROMAGNETIC THEORY. far only and it is sometimes an advantage to omit it. Here ag ain. t he unit operand may be omitted in general. p-a with unit operand understood. reduces (30) (3). Introduce the omitted factor equations (28) to (32) to produce the working of the electrical problem. e/liv into See (23). For . to unit operand again. (If an operand always understood to be at the end. <---_.) So makes a pretty change but we ca n go further. not 1. We get with binomial as required. that is the operand. in (29). just as in arithmetical and algebraic al l or e and now it is a* is operations. VII. 338. Substituting (30).

from That a. (p) is We Having changed the operand and obtained change it (29). So (33) to the left if is p(u**) = **(p + a)*u. Similarly f(p)(u**) if = e*f(p + a)u. find that this / is justifiable results) in t he case of the irrational process (by functions of p we have had in question. . simply n a< n at . (34) a rational integral function of p. depends upon the property p e = a t the potence of p is. when n u is integral. under the circumstances . Thus e at may be shifted by increasing p to p + a. we then to unity again. a function of t.

ay* = -*Pf(P. This is expressed by (35) f( P ) I = <r"f(p . of p under the radical sign. we have . p-a to The result is to come Instead of the first power an entirely different kind of operator.a )i.through (30).

But there is another thing to be noti ced. had p with the + sign under the sign.ELECTROMAGNETIC WAVES. the square of 295 p in (31). unnecessary constants. it will method again. we have The operand is e f(t or c-"' when at . results be found that the alternative This point will be noticed disconnected from Collecting as regards ^(at). The subsequent expansion and Previously we it has the correct algebrisation in (32) are obvious. of expanding (31) fails. Now Furthermore. radicals.

with p. A. when no time So far there is only viz. when K is not To this developneglected. So the developed solution . Instantaneous Impedance and Admittance. At the beginning functhey are factors only. making (87) The form provided S and L. one constant essentially concerned along But in the general case. (4). is the same as (21).the end. V A due to steady C when R = 0.. a. similar to that of 353 occurs when B is zero. The operand tion is written at the end. case which is 355. is 1. I ment now proceed. there are two c onstants involved.

the voltage required is to produce the steady The Sv that occurs here the instantaneous admittance. It is the reciprocal of LP. the electric and magnetic quantities For instance. ( to C impressed C due to V impressed. if the source is the current and zero before. then (27) is transformed A to (38) Dl? which shows current.as for is the same for V due K. we interchange R and steady after = 0. . the instantaneous impedance.

In .

R obviously. the ind uctance per centim. The error function again. In the corresponding plane electromagnetic waves.296 ELECTROMAGNETIC THEORY. E = /u-H. when undistorfced. If OD. C/V becaus e the resistance of the conductor is infinite. is 0. i mpedance always. accordthe distortionless circuit the in common units. begins to alter by reflection due to the unbalanced action of g and k upon the m agnetic and electric fluxes. then In this case no finite voltage can produce a current. and the leakage is stopped. . C due to steady V when S = 0. say from 10 to 80 usually. (5). unless But Lv is the instantan eous interfered with by reflection. and L are both zero.v. both elastic and conductive. VII. then 356. then V/C is 0. tube of flux of energy. and ~Lv is the total effective impedance ing to the size of the wires and distance apart). that is. whatever values E and K may have. using the ele ments of V and C That is. . the impeda nce immediately for the total flux. If S is alone zero. But if R and K (y and k in the proper wave problem) are not balanced. impedance remains ~Lv always 30 ohms x the number representing (that is. It is only a different way of reckoning. a nd the impedance is fj. CH. and C/V is But if K and S are both zero inst ead. pv is the impedance for a unit instead of the totals..

is of The admittance is zero. method We get Now point in geometry. as we see by putting p = oo and As regards the state at time t. there are zeros and zeros. An absolute zero is like the that the fail. as the schoolboy said. plus an infinite series of zeros. steady conductan ce.This initial is for a leaky circuit in effect insensible which the permittance compared with the other influences. But when you use a be magnified. I do not think that is the e xplanation here. the final is (K/R)*. which you cannot see even magnifying glass. the effective in turn. when V is st eady. and an infinite number of finiteness. though. observe of expanding in rising powers of p appears to a constant term. because the other some zeros can them might make . But we can pass the matter over at present.

3 (at)t _ 1 . _ (ai)? vffv " 1. ' " This is pnl = t ..ELECTROMAGNETIC WAVES. presents no difficulty.. "* \ . 3 . 297 viz. powers of p.. in way of expanding by the binomial theorem. descending _ /2K\t ((at)!. ii 1 r 5 (atM " n '.. T hus.

n complete. as explained function. in terms of the er ror /\n. See 247. is Two ways. and then . We have. before used. the integrations being done at sight by in 351. It is curious in how widely different a manner this function arises now. Or 1 to e6( . 357. Then when C is steady. by changing the operand from in the manner of 354. as we see by comparing with (39) and (43). But it will be more instructive to vary the method. There a similar case when L is alone zero. when C is constant . and interchanging symbols. (6). V due to steady C when L=0. Or.

we shall obtain the result (44) again. Multiplying them together. .to 1 again. '"' showing the result in terms of the product of an exponential and another series.

358. we obtain B(YB\-i rt . VII. is zero. is Two ways.298 ELECTROMAGNETIC THEORY. Connected with the last examples the inversion of the same.1. (7).1/B\! 1./B\l 1. CH. .3/R L . 3(2\L/ 5[8L/ 7l4\L (48) is This still. V due to steady C when S Say. By the process of 856.

(R + L/>)/(R-"Li>) and Thus pp~ l l = pt = l. B+IW (43). If we do the differentiation first. but there a simpler way For V -0" by using (48). -= f 1 . We did the integration represented by the denominato r first. but p~ l pl =pO = 0. unless we say p~ l pl =p~! times. is it K is -TT^BTir which r/RA* V R+IWKVerf = -K-(S) (r) ' rentiation now is But the solution of (39).is straightforward enough. it will make a difference. consider it to represent 1. the result is not a general t ruth that we may introduce concerned. Only one diffeWhen executed.

C due to V varying as e~P*. 6 = K/2S. p=a . 359. or p and a-.-= () 1. Now let all four electrical constants be i Then O _/K+SP\* V where there are only two connected thus. This property has to be remembered some(8). (51) oa = R/2L. effective constants. All constants finite. a and b. finite.

= /> the case of no leakage. cr = a-b. their product is then a series in rising powers powers 1 The algebrisation is then immediate. by integr ations of p" .+ b. and when When (r = we have we have the distortionless case. can expand each of the two radicals in (50) in 1 of p.. Now we . 353.

in a way that d oes not show plainly how to simplify it. without illumination. If e is constant. we can solve at once. Then (50) becomes f -P/ _. and the resulting series contains p and a.ELECTROMAGNETIC WAVES. ducing the factor e~^. or the voltage is the special ~^. Thus by (36). el* Introduce into the operand. But this is a mechanical and complicated process. 299 n according to p~*l = t /' n. and turning^? by simultaneously introto p + p. E one if e = VtP* . Follow (35). So vary the method. Here one integration upon I (<rt) is .

we fall back upon the case (27). Lv \p + Liv or/ (55) by (36).wanted. = ee. This solution is important thus. Imagine a cable which is infinitely long both ways. in having p Here we have a compact solution. Discharge of a charged into an empty Cable.Ktl*e. because the v oltage falls to zero.J.Kf/3 . to be charged initially to transverse voltage 2e o n the left side and zero on the right side of the will instantly fall to e at the origin.K '/3 . 360.-"^). (9). C due to V varying as e . Th is value will be V and . differing from (27) and o. A more decay in a different way.instead of both o-.. Say Then the first of (52) becomes significant case is got by letting the voltage = e. where e is constant. V O-lL^^Y^. When K = 0. origin. The result is The initial current is */I>. of course. The final current is zero. as in the above.

is. later to represent the complete wave of current. the voltage impressed . its later V therefore. It will be generalised right. upon the initially empty cable to the The resulting current is (55).

\p a-/ (56) which shows the voltage needed ase-P*. (10). by what has been done. that in the course of the transformations employed we are engaged incidentally upon other problems than those under discus sion. . (52) shows that if e C= Lv e. For example. and V due to steady C. CH. then V . 361. we have integrati on.1" " (5V.800 ELECTROMAGNETIC THEORY. to make the current vary C due let to steady V. in passing. Then we have This can. Notice.. be reduced to a single First. VH. Now the impressed voltage V itself be constant.

P-P (P* ~ 2 1 P ^*-pt 2 )* P-P (P ~ P o-' 2 ) 3 ' (58) by the process (35) . by (36) -pt (59) Using this in (57) we obtain TT \ pt nf IQ (<rt). .P {(P + P) 2 ~ ^^*-pt or.

integration from Similarly.all (60) Thus the C due properties function. is when in four electrical are active.is E . expressed terms of a known and its time-integral. is 1^ = bv (l + ^ ) \ C L. the residual p~ l meaning to t.J / -<* I (<rt) (61) the V when C is begins the operand). to V constant. constant (always understanding that 2 = We get the last by interchanging symbols but o.

. the space integral of LC. and V due to impulsive C. is an impulsive voltage. L and S. If C due to impulsive V = joP. where P V. makes no difference in I (at). 362. the case is that of (11). though this The symbol p does not change. a constant. generating the momentum P. total P.and K. negatived.

ELECTROMAGNETIC WAVES. (62) Or. 301 Then.. negative It is when ois positive. C =(. and positive when ais negative.. (63) This shows the current due to the impulsive voltage.> + P Lv Ltv =(p a-) -P I crt = Ltv (*> cr) Io(crO. That is . by (60). executing the differentiations.

From these we conclude that if the cable is in finitely long both ways. C = pQ is the datum. then momentum P be suddenly V= -c (p + . and V follows.r)]^). V= ov t -rt(p + <r)lJ<rt) = by c-<*<r{I. if R is back from the cable in excess. by interchanges. the current following the impulse if in deficit.is. followed by insulation. into the . it is forward.is negative. represents V due to the sudden inject ion of the charge Q.(<^) -f y<rt)}. That is. Similarly. (64) obtained from (62). C = --c -"'-(p a)I (<rt) . (63). It is positive when o.is positive and negative when o. cable. and the charge Q or the introduced at the origin.

Cubic under radical. into halves which se parate. Distribution of Operators. if A = 2/a and + Ap + B)-^?. we have z algebrised (p For it is the same as That is.l. B = p. (59) above. true for a time. (66) . It should be noted that (65) split represent V or C in the middle of the tail connecting the twoheads or waves at d istance vt from the origin. until in fact the first reflected wave arrives at The halving is done because P and Q immediately the origin.o-2 .. these are still the other. The complete wave formulas will follow. Observe that Reversibility of Operations.(65) represent the resulting V at the origin in one case. See 208. and G in If the line is not infinitely long. 363.

and put e* at the beginning and r-** at the end. say u= (p -3 under the radical sign is therefore can reduce it to a quadratic by making the operand be where x is one of the root s of the cubic. to a cubic CII. The extension suggested. VII. or (1 + x/p)-1 "**. Thus = that is.802 ELECTROMAGNETIC THEORY. J1 p+x {p* + __ P (Sx + a)p * r + ap* + bp + c)f . (67) + 3# 2 + 2ax + &}* (68) (69) when . or earlier. Turn p to p x. We at the end.

F and it G further. would be out of place here. But it is also true that )-!. but This may be carried are known. not differentiations. (71) Expand by the binomial theorem again in rising powers of p~\ and execute the wor k on I (<r).cr 2 )* (p is clear by inspection of the form of the T function. provided we understand that in expanding it. the binomial theorem allows us to write in the form (67). leaving the result 1. The coefficient of every power of t vanishes save the zeroth. having no im meis diate relation to the matter in hand. by (63). and then We get a series which putt ing it in terms of powers of p~*. That (70) IoK> = rir^i 8 . we are to employ integrations. The following more to the point. and w ith some of the preceding which is work. it is to be noted that if we have an op erator . In connection with this. concerning the reversibility of the operations used.

(73) a series in rising powers of p~l and if all the <'s are of this kind.+ c&-* + 1 1 1 .the product of any number is of others. . and in <l> 1 = a + b p. w= 01 = &&&!. then the same is true for the resultant 0. . say. and that if (72) the type of </>x . is . .

It to is is to integrations. not always necessary to keep to positive powers of p-1 that Sometimes a series o f differentiations may . Practice is more important. We have to find the time integral of with side matters. properly followed.ELECTROMAGNETIC WAVES. But sufficient and the final Development of Equation (60). equivalently replace a series of integrations. by chang es in the operand and operator. The matter is not so limited as it was just now stated. 303 whatever order the <'s are written. a nd a first way is to shift the exponential to the (12). left.. if So there are all sorts of which are bound to lead to the same function. 364. ways of algebrising (72). The utility of such changes or (< 2 <&j)<il &c all is bring operational solutions to more convenient forms for algebrisation.)( 3 l. though intermediately we may We may regard u as sorts of func tions. There are several ways of obtaining a full development of the solution (60). thus P P . one or two of which may be do ne here. be led to (< 1 c. integration involved in (60) remains to be done. of th eory now. C due to steady V. c-P e I (o-).

. _ + .P-P say. making ) pt\ i + +^ + . every set being the integral of the preceding one.) / (76) where the law of the coefficients is obvious. This u= is integrable at sight. . like the p' from which it arose. .p)~ l to find u must be done by integra1 So tions. . (74) P Here the operation (p . + P"H _ + _! !__ + 224 2 5 6 ..

The part involving u goes out when there is no leakage.Therefore. . expresses the complete development in one form. by (60). showing the C due to steady V wh en all four properties of the circuit are in operation.

VII. may be indicated. 365. and then operate on the result by ( obtain (77).804 ELECTROMAGNETIC THEORY. after a little rearrangement of terms.<r/p). so that the result is in terms of e^ and its integrals. There is to notice in the process. We first of (52). another way CH. do (78) without expanding e^ first. Here r2/> 2 )~* expanded 1 . But with a view to finding possible better forms of the function u. . so the details need not be given. As a check upon have (77). then operate on it by (1 < in rising powers of p~*. Another Development of Equation (60). We shall nothing particular by the (13). We then get where e n is the sum of the first n + 1 terms of the series for write out the operand e^ in full.

/ LA by the p/\ v L*\ a-.eP* Now. p/\ (79) A 2 / ' (80) ' ( definition of p and So becomes *?*)} Another modification it <"> is got by arranging the function u as appears in (79) in powers of pt. Thus. in (79) the inside exponential is cancelled by the outside The part of C not con taining t may therefore be exhibited separately. one. We get .

o2 //o ) sum of the first n "* in rising powers of + 1 a-'/p terms in the expansion by the binomial theorem. .where fn of (1 is 2 the .

( which is the same. 305 and the coefficients of the fa are the successive terms. of eP*.ELECTROMAGNETIC WAVES. . in It It is a curious form. we require this identity. + (60).+ ff + f>. (f This + f )/. *('->{ or.. pairs. of the expansion and <r p. may . be seen to be co rrect in the two cases cr = can = Comparing (82) with (77)..

Third Development of Equation Integration by Parts. f* We _ have i (<rt)rf*=rtili L HT+ -/> J fL^pi^dt. rearranging terms therein. 366.be verified by means of (76). P o (85) Repeat this process again and again. We therefore make . Lastly. (14). compare the results with those got by the process of integration by parts. and we get -p It is to be understood here that (1 -p/p)~ l is entiations.

verified by carrying out the differentiations also Now we know that (89) _fifc_If(ot) = u. done by integrations. this being as above in (76). provided that -pp (87) and this may be on the left side. .(78) or (60) done by become differp This agrees with (81).

(91) the current produced at the origin (# = 0) when the there is V = e e. including The results are all consistent. we can go on at the origin Consider (55) for exto generalise them in certain cases. then the result co uld we must rest satisfied complete wave of current. So. with one (15). Say.806 ELECTROMAGNETIC THEORY. Having now got ample. But it is unfortunate that the additional terms brought in by leakage do not seem to be If it could be reducible to a single kno wn simple function. VII. V Ks at the origin varying as e. 367. C= This is e~^Io(o-t). then or other cumbrous form of series. may be If infiinserted. be readily generalised to express the so. adding the results (88) and (89). If the cable is short-circuited at the origin.K '/s e be ing constant. this mathematical result follows : p p all the summation including zero. If not.*/ full results as regards V/C and C/V under different circumstances. Generalisation. beginning when voltage = 0. The Complete Wave of C due to . as a variable impressed voltage . integral values of m. CH. .

V regarded the impressed voltage must be nitely long both ways. . = ^{(p + p) 2 -^. (93) . we have where q = -* Co. Now. V = 20 initially impressed voltage but by an on the whole of the left side of the origin. namely. then But we may al so regard V as being produced not by 2V initial state. (92) The current must be the same . at the same distance on the therefore negative side as on the positive. So cosh qx for . at distance x.~ qx We may put C = cosh 2* Lt e-<*Io(<rt).

33 7.i. I. equation (2) Applying (96) to (94). Shift e-P* to the left.l m (a?). since only the integral 2 2are involved . (96) V ' (97) ^ ' of This general property is proved by the differential equation m see also (18).o. oNow ) powers of (p 2 (o-p*) I (<rt) = rWrt) = o^i^.ELECTROMAGNETIC WAVES. (94) This gives C by direct differentiations. 807 Then cosh Liv v C= (? . (98) and. . (95) and generally.)* 2 . if . 336. we obtain . I (<rt).

n (o-^) is Now (99) is nothing more than the expanded form of (101) as may be seen by arranging its expansion in a series of 2 Therefore. io (v-t* X-)* in the I function. 338. Thus where P. the last equa tion is the generalisapowers of tion required. the convergent formula is used. x2 . equation (23). so that the f irst term of P. we see that vt is turned .i.we use (97) we get ! the function got by dividing I m (o-) by its first Of course term. Comparing with (91).n itself is unity.

CH. whose front travels at speed v. the current is zero. without prior knowledge. VII.At the time t the region occupied by the current extendsBeyond that distance onl y to the distance vt from the origin. Seeing that the preceding portion of this Chapter contains the . the Since the I function has been tabu(e/LvJe-P*. if it the wave solution. Summary of Work showing the Wave ~ of C due to V at origin varying as K^ S . at successive moments of time. shape of the current curve along # . is already known that the formula satisfies the characteristic partial differential equation of C. The above is by far the simplest way of obtaining Of course. 368. But it would not be reasonable to expect people to possess theprior knowledge. can be readily calculated. is (16). At the wave front thecurre nt lated.308 ELECTROMAGNETIC THEORY. there is no difficulty infinding the nature of the problem of which it is the solution. Equation (101) theref ore expresses a wave.

. V= is 9 *A + -*B. which express So are themselves connect ed by the third equation. and th at the e~ qx part is the only one wanted when we send disturbances into an empty cable . . it is in the process of deduction.treatment of various other cases than this important one. from the conn ections of voltae and current. V . C= . (104)' the type of solution required in general when A and B are time functions.Af we deduce the characteristic /72V 11 = (R + L/))(K + S^)V = ^V. from which we conclude that (108). desirable to exhibit collected separately the various steps First. (102). (105) and C in terms of V and C at the origin. so that -*V 0--K3 .


ELECTROMAGNETIC WAVES. 809 Pt VQ = ff Tt et where e is ^_ = e_L-I . Then let K3 This makes cons tant. generalise for the complete wave. by making the operand be V ^./> + CT/ p CT LlV (^). or V = ee. e C = 6-o'Co = cosh qx C = ^r cos -(o2 p*? . Finally. (107) fully realised. I (<rt). (108) which develops to If .'/ .

instead of as in (108). Derivation of the Wave of V from the Wave of 0. we have C = cosh qx . but the additional part will not bring in any new terms. without the property of equality of the discontinuity at the origin.CQ.current at x and we do not make use of . 369. because its first term will involve (j?-<j*)*i (<rt)= P i. rest involve complete differentiations upon (17).ar. To obtain the formula for the . C . first and the term.shin qx . (no) by the (71).

there are many ways and the C formula itself. of working. may use We have ax v by (109). Now. provided value 6~ K '/ 3 at the the negative of the x integral of the right it is standardised properly. to have the origin.V wave corresponding to (109). we the second circuital law. First. looking at the expansion (99). We therefore get . Therefore V is member. So the real value at the origin must be added on. we see that its x integral will vanish at the origin.

o ' \V + 2*8 _____ i ^yave * Jj i (112) .

but work V . / m (<rt) Using this in (112). 247. The argument This is of all the I functions is a-t. Another primitive. (18). development of the V wave. See 246. CH. VII. we obtain the equivalent form. the voltage by the error function. From this. The Wave of 370. But there is a similar peculiarity in the corre sponding formulae in pure diffusion. Do independently developed. We have There is one time by (100). the V formula is more not use the developed C formula. not so compact as the C formula. differentiation concerned. the current being given by an exponential f ormula.810 for a complete ELECTROMAGNETIC THEORY.

Now for &*.shin)jar. Thus.way of getting from the beginning.P*(cosh . and t he potence of p . J^^ . and r-* put shift ~P* to the left. .shm)*V . (116) ~ V-r^Vo-srW'-'* because the voltage at the origin (cosh is e ^ ff p ^.is zero.a-2)* v (117) Here the cosh part involves only complete differentiations^ 1 2 upon e "*.o. So we get shin5(/ _ .o-2 )* 1 But here by (36). V = e e . Only the first term of the cosh function does anyt hing. Then .

..(a(^-o*)VrtJ.-/*" .?(p + r)l -F) + . o^). (120) . (118) So V = .

First. once. according to the formula (97) above. at time t later. Then. so we may derive the C formula from that for V by a similar direct way derived it process. just as from the C formula by the second circui tal law. which may be any point in a cable continuous formulae arising both ways. 811 2 2 carry out the differentiations (o. We are now in virtual possession of when electrification spreads and when momentum spreads. (19). and the development (112) immediately results. tail Momentum at a Single Place. twice. The Waves of V and C due to Initial Charge or the complete 371. upon I (( rt). This is the Now most we of obtaining the V formula. using the first circuital law. thrice. &c. the voltage and current are ..-^ ). And. let there be the charge Q initially at the origin.ELECTROMAGNETIC WAVES.

2 V = JPir-** The last pair *)T first . Then at time t later the current and voltage (*** x are (123) C= ^f \P ~ p ) ~ a8)1 Io j I 2 * ].38).V= -e * (p + I or) V*2 . (121) (122) Similarly. let there be initially momentum P at at the origin.

because of the change of sign of <r in passing from (121) to ( 123). the ot her origin and up to distance x . the result is simply V *= V ft .V on the whole of the left side except from gi ven by V = the origin to the distance x = . If we superimpose two distributions of electrification. P and Q. R and K. Observe that the spreading of current due to initial charge and the spreading of charge due to initial current take place in But the charge due precisely the same way. to initial charge an d the current due to initial current do not behave similarly. (122) and (124). . can be written down from the by interchanging V and C.\a.(124) pair. L and S. one given by V = V constant on the whole of the left side of the = la on the positive side.

or (115). qv dx (125) This is fir-t q is done by introducing equivalent to qv/qv. In this way the solutions (121). VII. say Q. (122) for a point-charge are derived from (109) and (111). which neutralise one anothe r save at the origin. = c" QX C if C is . the result is a finite charge Q at t he origin only. We know that the current impressed at the origin. and then noting that the is . Therefo re . Take C == JpQ then C is due to the charge Q impulsively introduced and dividing into two equal parts at the origin. and decreasing a indefinitely.312 ELECTROMAGNETIC THEORY. or the charge SV a. along distance a at the origin itself. CH. The waves of voltage and current which arise from it are therefore given on the positive s ide by the difference of the waves arising from the two uniform distributions of V before described. But the following way is interesting. Keeping the last cons tant.

Now let us tackle the more difficult problem of the waves of voltage and cu rrent generated by a steady voltage impressed at the origin. Now qv the function in the brackets already known. equation (91) So (125) produces ( 122). leading to (101).djdx. and e-^^e-^o^-ajs/flS)*}. The Waves of V and C due to a Steady Voltage Impressed at the Origin. The voltage formula to match may be d erived thus. But we do not want them at present. at a single point. It is the combinati on of resistance . qv which is (121). 372. Knowing the momentum voltage and current due to initial charge or we can at once write down the integrals expressing the voltage and current arising from any given initial stat es of voltage and current. For ( Z^e-^Vo-O. qv 12 6) by (59) and its extension in 367. or (105) to (109). before done.

as in (131). . But it must be done.ELECTROMAGNETIC WAVES. In the rest. Pf .skinQE/W-O'y _ ^. Algebrising. and separating even and odd by making powers of x. to complete the investigation. (128) the operand. 313 with leakage that makes this case complicated. If e is the steadilyimpressed voltage. we get -where en = sum of first n + 1 terms of e^A point to be noticed . started at the initial moment.cr2 )* .1 ) I (jr-o*y < _ (129) part has been algebrised. So ^ V = e -*( oosh (*/r) tf-jy . in the above that in (ISO) we introduce pp~l tiation should be done afterwards.shin) (x/r) (p2 . the shin operator divided by the ra dical involves even powers of p only. we have V=e -*! = e e ^(cosh . and the potence of p is therefore p. Also A (130 ) by the binomial theorem. Here the cosh op erator contains only integral powers of p.

_ (/ a*)* <P t & &'. Therefore (129) becomes . So a second part is algebrised. Putting the two algeb rised parts together Now when we shall obtain e e~ rx . (133) the shin operator works on this. v KK)i + pU ( . in a more convenient form. the potence of p is p again on the first term i n (133) on the right side.The Let differen-then by (130). (134) or. (XS5) . where r = (RK)*. (131).

whilst the rest are derive d from it by 2 2 is as in It is quite easy to illustrate. 4 o- . of which the (132). (136) Thus. first (o2 Un = -^ )U derive U n because p*en = p*en _v (o. first CH.. to .pje. p*)\ = 0-% . &c. where the U's are time functions.S^e.814 ELECTROMAGNETIC THEORY. 2 2 (cr 2 (tr jPfe< = o-<V4 . . and then.3o-y. (oSo we have &c. VII. + 8<rV. = <r*e 2 ./> 2 n ) l = o2 '1 .

there are several ways : of The one analogous to the above goes thus C= = by first ~ ^-P\P making -)(cosh - . o 1 . 1 . or an equivalent In it. rest ultimately vanishing.2 1. or else a definite integral which is of no use until it is evaluated. and so on. all the one. of current.1 o2 . But the only alter natives are other equivalent infinite series. the term e e~ rx represents the steady state. 3 oI regret that the result should be so complicated. when the result must be the series (135). As regards the wave obtaining it.

( ^r^. (137) e/>* the the radical equivalent top.shin) V cr 2 )*. We now operand. This algebrises operator divided by use with the cos h operator. we have a part. For (/>-<r)/(p-<r). and then introducing have the potence of p in the shin ( 138 > .

perhaps. V = . also sometimes in the exec ution way Say we require to find the wave of voltage generated ~ qx e. . and as regards the rest. (HO) So. and thereof. Progression and Atten uation of Waves. finally. in passing. take to ease the work. e is suitably selected As an example. Practically. where e is by impre ssed voltage at the origin that is. 315 Using (138) in the previous equation (137). we o btain -' As before with the developable. a given function of the time. of algebrising the the simplest possible is. we have a series of functions derived from according to W W n = (o-2-^)W . we see that the potence of p in the cosh operator on the first part is again p . U functions.ELECTROMAGNETIC WAVES. which in the ideas concerned. the W functions are easily Analysis of Transmission Operator to Show the Deformation. putting together separately the parts independent of the time. 373. I shall now explain another operational solutions.

Therefore. steady after see at once that is Q = Q). t (142) constant (zero before. so that cr = still. =/(*). But if there is r esistance. That is. and . and also leakage to balance. the voltage impressed at the origin travels out at speed v without any change whatever.=* where e e-/*Io(<rt). if also R = 0. qx already found the effect of t~ on the function above. then qv=>p+p. in the distortionless case. x/v). See (126) when if e <r = 0. (144) K = 0. We can because we have in (142). we have V = r**/"/W -/( by Taylor's theorem operationally considered. Now. we have q = pjv.

V = t-Ve = -P*l'<-P*l'e = t-Px v f(t-x/v). l (145) .

absorption. CH.. as agreeably complicated. occurs to every part of a wave i n transit. At the same time the wave of current is C = V/Li?. in the same phase and in conThis is the beginning of the theory of the diswhich I recommend every electrician waves in It general. who aims at telephony through methods which violate the conditions occurs by absorption in transit. viz. distortion. . since it casts to understand electromagnetic waves formulae. I beg to refer again to the po rtions of Chapter IV. which are devoted obvious that an electrician very long cables by under which it is possible. Thus the disturbances impressed at the origin now travel out at speed v and atte nuate according to -& in the interval. allows us mathematically not merely as a collection of circuit. and the uniform attenuation which to the subject of waves along wires. If o. but in terms of physical ideas of translation. is only wasting his labour. sometimes distortionless in full detail. the curre nt and transverse voltage being stant ratio. because t = x/v is the time of transit fr om the origin to x. attenuatio n.* The re are only two ideas concerned in (145).. Vol. and so on.316 ELECTROMAGNETIC THEORY.. or the wave shape is deformed or distorted.is not zero. VII. the progression of the wave. there must This is the parti al reflection that be a third process involved. whereby its parts are It redistributed. we still have propagation at speed to study an introduction to electromagnetic light in the most obscure places. reflection. It is I.is not zero. follows from the d istortionless theory that when o.

* marine Cables. _ _ ff ^ ^x/v = -rxiv Y (p-*) 9 (147) This refers to the proposal of Mr.e v. H. September 25. Both results are entirely antagonistic to telephonic transmission in long cables. A.. separ ated by a piece of paper. in fact. (146) the operator containing p should be expressed thus.Pl . 1896. Is there a . W. telephony. It therefore -*** suggested 2 2 te -"itself to me that in r *-* )-^ A. to bring the two conducting leads as near together as possible. in his paper at the " Electrica l Disturbances in SubLiverpool (1896) meeting of the B. Preece." printed in full in The Electrician. and still have the attenuation factor . for the furtherance of long cable p. 689. This will reduce the inductance to a minimum. Mr. and increase the permittance to a maximum. Pree ce seriously proposes.

who will support Mr. . theoretical or practical. Preece ? Howe ver.single electrician. the paper is not more remarkable than some of his previous ones on the subject.

Say . 817 1 where F(J for its first a series in rising powers of p. . First carry out the op eration F. It The F's are simple functions of goes well. ) is . (148) so that V = c*e = e-P*e -F* (l + F> T + F ^+ . (149) If this is possible.. But it translation. This will effect the rearrangement of parts. F = 1 + F^cr/p) + F. does not appear from the opera tor in (146) that the expansion in inverse powers of p is a natural one. F=l. (o 2 2 + F. the factor e~P* will attenuate properly.having unity term. . the algebrisation will be immediate. . first three are o-x .}j*e.(<r/p)' + 2 .ELECTROMAGNETIC WAVES.. Then carry out c -**/*. because when <r = 0. This will do theFinally.

n (<r//>) The l is simplest case the same as (o-t) n/n. The Wave of f V due to impressed . Voltage. varying as 374. Supposing at present. if . or steady.#. Three Examples. we can show the working of (149). of which the oV <rx <rV How them known to obtain these will be explained separately. so the t e~^ f""^8 is eef* = eQ or . e = e e~ pt . ThenF * lf series is algebrised.

This is be cause we have chosen V at the origin so that it shall be so. at any moment. The natural attenua tion is V from origin to wave front is e P 1 when t = x/v. is V . (151) expresses the complete wave of voltage. where V a straight line all the But it drops to zero suddenly. Or. to ^ means t-x/c. ~ this In the distortionless case the curve of way up to the front. f/2r Observe that [2 has the same value at the wave front as at the origin. .Then the next operator turns Therefore. and just the law assumed for th e variation at the origin.

together the terms involving -*> e' n ^ as before explained. &c..318 is ELECTROMAGNETIC THEORY. VII.. <& e ? xlv . let e less.. according to (149). Collect . origin . then we have to integrate d* once. twice. not generally a straight line. where we have put ^ for t.. Next. ~ that is. and conversely. steady after t = 0. be constant. It is ea sy to see that if e subsides faster than *~ pt the voltage at the wave front wil l be greater than at the . (154) c ptl . Thus. CH. Then --<>< where F^+F+F^^ of the first . . even when V is the same at the extremes. zero before.

As a may be simplified to this extent.. -rx We (155) conclude that = px/v{l + Y^lp + F2o-2//)2 + . When cr = />. third case. let etP* = ee fft . then therefore.}.. where .means the sum n+ 1 terms of This is an alternative form of the solution (135) above. so (154) r = 0.

}.. (157) or.eQ is constant. K '/ 3 instead of (153) we have the following 1) : V-^er^c^^P^ + Fa e ( ff ^-1\ o-tj + . subsiding according to the natural leakage law..0 = <? We is. Then in the first portion of (158) we have and therefore . collecting the terms involving where e" n is now the sum of the first n + 1 terms of e "^. instead of e^. But use (156). Then . have now to integrate t fft in ( 149). That .

Kt s -^-^{F + F 2 .. " + F '' 1 1 3 .oC . . (159) This is an alternative form of the solution (115) of the same problem.2 " + .}.V = ..

instead of (157) or (159) we obtain where is constant. Then (160) If we of 1 desire the expansions to be entirely in rising powers fft in fact. We get V = e . />.. and is steady.ee ~ ^{F. and integrated term by term. of (153) and (154) we shall get Thus instead Now rearrange in powers of t v and we obtain This is when e is constant. Expansions of this sort.ELECTROMAGNETIC WAVES. which are so eas .}. = e. like (152) be expanded first.inside the square brackets. It follows from the previous equation by changing e to e Q and putt ing p = a. 319 < An important reduced case is cr = or no leakage.. then the oper and fP t or t should xjvt. And in the other case. + F^" + F/2" + .

l . This is obtainable by expanding Put -signs. in a variety of cases. it in powers of Simplify by removing the . then. 2 2 -x/v = y. . to exhibit the e~ term separately. Expansion of Distortion Operators in Powers of p~ l 374 we require the expansion of 375. and-o.^ _ fD^+rt . rz It is not desirable. A= ^. In the method of the operator ~ e 2 (P ~ *?)**/ in powers of p~ o2 . whether due to impressed force or to an initial distribution of charge or current.= s Then.ily obtained. are useful in calculating the initial stages of the devel opment of waves. by Taylor's (164) Theorem. and <? .


CH. in the operand before the differentiations. This process will be found to be t D very laborious. D may s 2 may is. put = Then we see that signify d/dp*.320 if ELECTROMAGNETIC THEORY. be s a d ++-1 s 2 il That . means d/ds* and we put s = in the differential coefficients after they are devel oped. VII. But the waste work c an be avoided by observing that differentiation to z p produces the same result as differentiation to s 2 .. that is. owing to the rapid elongation of the expressions for the differe ntial coefficients. several of which are required before the law of the expansion can be recognised.

which causes failure.s 2 . the other But we are not guided in our choice of the form is divergent. we cannot use th e other form of expansion 2 in our application. and the result is + 2/> [2\2^/\ i(OYi-L yp where the function in the { } contains p inversely only. Numerically considered. that consideration. but by its having p in the denomi(164B) by nators. Besides. What is important here is to recognise the law of formation of the coefficients of the powers of s 2 . d 1 (164A> There is now no waste work. because s has to be negative the n. if one is con vergent. If we interchange s and p we shall obtain another expansion of A. so that the integrations can be at once carried out as in 374.

y/2p. So s2// if we write 1 * A = ^ll X -i/^Yr 4. s ' ( y \r2 * + : \ 2/? ?'3 . by 3 3 \2/?/ '"/' this simple process 12 . We respect to obtain the rule by observing how a differentiation with p derives any term in th e expansion from the preceding one.. we shall obtain ra 133 ?' [2\2p/ &c.

1 .

and multiply them by 3. 7 and added on. We start with the coefficients in ra . &c. is py PY Guided by + W r =110 105 + -^ py this rule. It 821 almost explains itself. 4. 5. 6. we can (164c) * 336. Thus 105 r=lc. These. again. produce the coefficients in r4 .. multiplied by 4. and add th em as shown to produce the coefficients in /g. 5 respectively. And so on. producing the second line.ELECTROMAGNETIC WAVES. really a series of .

we shall get vided we use the operand yp/p instead of yp B by (164A). (164D) and so on. We also require the expansion of A/(/r +s . derive the A series (164c) from the B series by differentiation to y. Comparing the two series. This makes (165) But the above numerical process is sufficient. 8 )*. Then working in the same way.H m functions. by that formula. we see that the series for B/> i s obtained from that for by shifting all . see that the expansion See equation (3) For. proThe result is As a verification. Call this B.

_ =l+6 +-. Now. (Ittj .A the r functions one term to the left. (T 2 X 1 2 V2/W '3 V'A (167) P where * 2/' + 44^1 IT. -x' v and t for s 2 put -o-' 2 . for y put are expressed by ..}. in the Then A and B application..

(1G9) .'i.1 ** '.-+=*'3 ^^ + ^.

-. in powers of o-/p. if ELECTROMAGNETIC THEORY. VII. . The Current Wave due to Impressed Voltage . we arrange p p p p F (170) then the F's and G's are functions of x only. (172) where the Example.322 &c. coefficients are picked out from the r functions. Finally. given by F 3 = 2*{8 . so that CH. _ 3 2 e. All signs are now positive in the r functions. 2 ^-.

376.(p" qv pt -pt . It is when q f Lv . ffi where e is constant. so that the voltag e impressed etfi = e c. at x =Q is e. operand. with the expansion B that we are concerned with the wave of current generated by impressed dealing For the current due to e at the origin is voltage.Kf / s .. such as would arise from the initial distribution of transverse voltage V = 2^ on che whole of the left Then we may put pl( p a) for fft in the side of the origin. and produce (174) .<? e~ Ki/s at the Origin.v^xlv (173) This is developable by the B expansion in the same way as One case will do to il lustrate. Take the wave of V in 374. .

or <rt and zoM'/i'. We may also interchange cr 2 /2 .ELECTROMAGNETIC WAVES... But for the reason just mentioned in-(t .. This makes o-^ become icr(t + a:/v). (177) Since the sign of x may be changed. s (175) or. \Ye conclude that 1 + o-^G. -which develops at once to 323 G ..o-V/v 2 in (177). + ^G + 2 .*>*)}.. have started us ing cosh gar instead of -9*.K.. the coefficients of all the odd powers of x must vanish in (176) and previous equations.. That is ia-t^ S o by (177). in fact. ^Ye might. = U<r(<? . (178) where the G's are known functions of g's are functions of ...x/v) will do as well. and . in terms of x and t. (176) This is an alternative form of the compact solution (101).

They are obtained from the and then multiplying by t/i. for ff lt y.to 2 Thus. it may be mentioned here that if we arran ge the solution (175) in terms of powers of o-jc/v as they occur in the G functi ons. as above. &c. i. by (172). of solution arises directly in another which will be referred to later.t x. The modified form of developing solutions. So of in the t expansion (175) we may for the functions G of x. . functions of x by turning a/i. whilst the instead. way To put connected matters in one place. This substitute these g functions is a very curious change.2 &c.. we shall obtain the form . i .

324 ELECTROMAGNETIC THEORY. (179c) .r/v.(-a-) + . VII. That is also arranged in powers of o-. though only even powers. 367. CH.T) + (f -I. writing t for and x for <r. we have . whilst now we have the I functions of vt r or a-t <r(t .. Collecting identities conveniently for memory. But it is the ? functions of crt that occur there.. Compare with the expansion (99).i-/t>.r/r)..

by P.and in these the signs of x and interchanging t and . + x) + . (p <r 2 ) nI (<rt) when n is a Positive or Negative Structure of the Convergent Bessel Functions. (170B)^(t in terms of oscillating functions.H + t . . 2 Value of Integer.n. 9 (179F) t where the signs of x and may be changed... t may be changed.. Also.

. positive and integral. that r^-ra <r)(p= IoH). equation (36). Now in the last. 354. It may be asked how things work out when m is negative a nd It happens that a comparison between two of the integral ? and its result in P m (<rt). turn 2 oto . in 367 that and previously. so that it We found is worth while indicating the result in passing. we had occasion terms of to use the expression (yr -' 2 M1 ) I (<rt).377.. m being formulae previously obtained furnishes the answer. In the method of developing the complete wave of current from the formula f or the current at the origin followed in 367. 2 (181) ).

.2 o-' (l aP/vH Thus.

is Now . we see that the resultant Oi: functions are the same. explicit is course. ^_^/ by using (181) again. for exp because the resultant function expand. on the left side of (182). So o-. left side of Now we can easily make the operand on the (182) be I (cr). P f1 . jr ^Y!?\* \ ~V-<r*J (184. 825 Comparing the t last with (180).J where we put cosh even as regards .ELECTROMAGNETIC WAVES.. treated as a constant. as in (180). Thus.-.

comparing similar powers of . we obtain (p* O-PoOrt) = P H (<r*). w'. leads to *2 K P n being positive reciprocal is . (187) members Therefore. take n = l. The interesting thing in the above is the way it comes ou t from such very dissimilar operational solutions as (180).Then ^ So. For we know already that . To Then verif y our constants.r. Alt (188) when w Pn . (182). (185) we see that I (fft) ' (186) which shows the relation required. equating the right this being equivalent to (97). its 2 a positive integer. There are very likely easier . ways of getting (188). Thus.ilst (jf-o )" leads to in both cases. of the last two equ ations.

.This is and the the easily developed in tli3 way done before many times. on the right side integrations. only. r esult is (<rF/ On fflP^o-t). left side of (185) we have differentiations only. in accordance with (188).

VII.)* will still . whereas 2 2 (> -<r )P is really -la.cr)*P which seems to assert that a repetition of the operation 2 2 (p -o.X - . CH. As regards harmonisation with the vanishing of (p2 . that produces ( 2 P -o2 is easily What z (p -o*)*P of really represents is the impulse pi so a repetition the operation )l =^1 -o =. 0. . produce done.?!..326 ELECTROMAGNETIC THEORY.

(rt)-P -(192) It is now . let D mean d/dy. Thus. Then Therefore. if y stands for I 2 (Jo**) . It is worth pointing out here that these the readiest way of exhibiting the structure and connection of the Bessel functions.-giP i the third expression being got by algebrising the second functions furnish through the binomial theorem. we have 1 .

we do have n(rt for any value of ~Y n from n ~V oo ' .KDP = y-lD-P if (194) But n is not integral we do not have way exequivalence of differentiations and integrations in the Since. ^P n = 2. 338. however. generally. (193) integral.-Jp. Also !>*) = when n is integral. pressed in (194).easily to be seen that D-P = L. when n is D-P. . by (23).

. when n n provided that D~ l = /. as well as be treated separately.* m* to + oo is generally. 378."/ n This matter will integral. Another way V and C is founded upon the division of any of treating the general characteristic of initial state into- . Analysis Founded upon the Division of the Icstantaneous State into Positive and Negative Pure Waves.

positive wave only i f V = then U is W .LrC then a negative wave only. if R = = K. that is. 327 If two waves travelling in opposite directions.r) no waste of energy. = i(V -Lt-C (1) . and C be the initial states of V and C. That is. the state got by shifting U through the distance vt to the through the same distance to the left. and is //(. if t is and W U is/(. and the negative wave. V ). and the positive wave.ELECTROMAGNETIC WAVES. . If there is at fc'nie right.r . at the If Y = LrC the initial state re presents a initial moment.

so that U= i*f(x vt) . the same division into a positive and negative wave W occurs. (2) W=y(x+vt) become at time t later. W= e *o(x + vt) (3) express the waves at time t.is not zero. approximately. b ut balanced.= 0. Going further. the theory of the distortionless circuit with resistances and leaks inserted shows that (3) always represents the it phenomena remain will in the first stage. but there is. if R and K are finite. Their sum what U and express an d difference will show the real V and C.). If o. besides the translation. attenuation due to the waste of e nergy. then W U =/(&-!). long fairly true depends on the amount of dis tortion How in the interval due to . so that o.

ct and x + vt be the independent variables.the unbalanced wastes due to respectively. go back to -AV = (R + L/>)C. It is clear. (4) A as and p are the x and differentiators. . To show the connections of V and C. =XV + LrC)^ <r(V LrC)e^. that the quantities R and K U and W may themselves be made the objects of attention in a mathematical treatment differing from the preceding. and then add and subtract these equations to .i-A(V + LrC)/* form two new ones. These are the (5) (6) same Multiply the second of these by Lr. instead of x and t we should make x . where t AC = (K + S/>)V. At the same time. how to do this. then.

(7) .

(9) if U = l^(V + LrC). AU) consequently the equations (9) .A)W = crU.rA d _p + r A _-. . CH. w= la-(t-xjv} (10) = Jcr(* + #/<). (ill CT -OT . let W = e^(V-LrC). (11) be the new independent variables. VII. Going further. + (r(V + Li-C)^ (8) which are the same.:j^S ELECTROMAGNETIC THEORY. (p-i. (^ + rA)U = o-W. then d _p. V)e^ or.

Let a be the u differentiator and /3 the ndifferentiat n*. to satisfy The general solution is easily foun d.L+++ _ 0/a0 _/ x 1 1 '" \0 . Then -~i. except at simplified to the uttermost. It is easy to see that the mathemati cal expression is now All solutions have.become The third of these is the characteristic of U (or of W) obtained from the previous two connecting equations. so that (14) (a0-l)U=0. sources or places of imposition of external influence. the new characteristic.

So is one form of the general solution. In the same way we can see that the general solution of the partial differential equation where the d"s are any number of independent a*. may be written thus.. sum of any function of u and any function of w.. We can now see how it works out and examine so me special cases..x n . This can only be F(u) + G(i r). with respect to the variables xz .Here the O/a/2 is af3 is upon by made which when operated any function of u and to vanish. .. differentiators.

. But the treatment will be sufficiently (18).. and the F's .2 and so on. (16) the is But nothing can we apparently generate U out of nothing. . operand 2 F is simply 1. thus F! are arbitrary functions of all the variables save one If the a function of all save xlt F. the case of tw o variables. in fact. really come out of nothin g. we get the fundamental solution U = l+*4if ww -+-** + .2 of all save x. (19) of x is the product of all the variables.. as in (17). And. .ELECTROMAGNETIC WAVES. shown in discussing In (15). if 329 <l stands for the product of the d's. for the derivation of other solutions.

Simplest 379. and </ = .3G. and of 2-r/o-. Waves Length of Infinite Length. 1 the means the impulsive function Solutions. It can be seen to represent an Since there are two variables. a function which i s impulsive as regards both is $/ + cu/. in solving (17). not absolute zero. Inspection of (15). by (d = as above {/ in (18). simple solutions of the characteristic = <}>(u + ir). and // a function of . Coming to the consideration of the instantaneous electromagnetic waves U and W. f U So there are just two crt cases.impulsive function. then a/^U Suppose that there are a few specially which should be noted. where/ is n Now if/=aF function of u on ly. this impulsive function is afj(F + G). and therefore Simil arly.r only. . (16) wi ll show that this is the actual form assumed there. 1)U = 0.

=V e. U = V c~ because <j>" <rf . \V. actual C is zero. and oppositely identical as regards C. = it + n\ aU = Considering the first. This makes we derive W = U. the negative wave by is. The waves are identical as regards V.Kts This means that the uniform distribution of transverse voltage V may be regarded as the sum of two electromagnetic waves whose electric forces are similar and whose magnetic forces . and the actu al V is V = V e~P* That e"* the .U=V ' and = U shows that = <.

.are opposite.

. (20) where the constant & enables us Remember that u . there are two independent solutions. CH. additive as regards magnetic force.Rf L form current C is repre sented by two waves. As that U = <j>(u . U=V sin (o-. before. though far more developed in theory. and C = C e P c ^ = C t . to include the second solution. Suppose Then a/3U = U shows that <" = .r/r + 0). W=V cos (o-ar/c + 0). destructive as regards electric force. we have / . Co mbining them. VII.<. From these we can pass to a case which is equally simple in expr ession.330 ELECTROMAGNETIC THEORY.= a-x/v. . In a similar way the solution U = Li'C e r fft gives W= U.u. The uniwhich means V = 0.ic).

0). . Now U is a positive pure wave. and K length in question. W But the last is very important here. The voltage and cu rrent are -/>' V = e-^V Lt'C V= (sin (sin + cos) (o-jc/r + cos) (<rj/c 6). and Their translation in opposit e directions at speed v will give the later state of things. The waves are appaThis means that with the special waverently stationary. + The length. larger oWhen (and the distortion). of General Solution Functions. Development 380. except as regards t he distortion.is quite zero we have uniform V and C subsiding. the distortion due to u nbalanced R keeps the instantaneous waves unaltered in form and position.a negative wave. the smaller the wavea. in wl ?/ P m and (16).

1 directly. because the stands for occurs so often.(a/*)1 }.#2 /^) therefore {<r(t-*V* 2i ) U=I The new symbol quantity tion I it (z) is z :=Io(0. from which to develop we use the operator {1 . general solution Let But here uw = i<r 2 2 ( .<r wP m the operand Now take in hand the F + G be 1 then . We see that th e funcis others. unless the fundamental solution. (23) introduced for clearness. .

if 331 the operand is u m/m. Similarly. result is got by interchanging u and and is. the therefore. < 25 > From . is '"/ w or /5~m l.ELECTROMAGNETIC WAVES. then ^ i u-1 . u -j p . ~" And if the operand .

the above we see that a (2) (26) a comprehensive solution.)PoW = Po-HPi + . not itself perfectly general. let be seen by inspecting (24) operand be e then ?< . . but perhaps admitting of the derivation of perfectly general results. is The manipulation of such solutions : is made rather easy by these simple properties l3 mP (z) =l P t * (") = ormP (z) (23) the truth of which may To illustrate.

and to represent the waves resulting from actions . to represen t the effects due to given initial states.showing variations of method. (29) Derivation of C Wave from V Wave. with Condition at a Moving Bou ndary. namely. p.. of ^.. are two ways in which the solution (26) $ 381.. There occurs. Similarly.+ . Expansion Examples. and Conversely.

or the V wave when the C wa ve is known. CH. VII.a)U. . impressed at a certain spot.332 ELKCTHO. Here it is to be understood that the disturbances are impressed upo n an empty cable. (32) This is a known case. It is with the latter that we have been principally concerned.MAGNET1C THEORY. We by the have VP*= (1 + a)U.a V/ 1 + C wave when the a (31) The last enable us to find the V wave is known. and l (2). For example. let e be constant. LtfCV = (1 . so the treatment may be applied to these first. generated by impressed voltage at the origin. V^ = l_ LrC -a a 1 (30) P. LKV* = }. So definition of U and W.

rP + 1 )P (s) 2jp. e at put x = rt. (33) process is easy enough. It is is . What V + 2a + 2a 2 + 2a3 + + 2. but how make sure it is that the fractional operator in (31) might be right. . To is obvious. by (27). + 2'Jp 8 + . . seeing otherwise treated ? That the result satisfies the ch aracteristic But it must be correct at the limits as well. The test this. . V = <?e-p..wave ? Use the first the expression for the of (31) to find the answer. then iv = Q and P This indicates that the value of = .)(*).

. and is an alternative form of (115).)(o-0 (35) is an identity. was 0. (34) But . 367. the moment t = 0. and nothing more. t But put x = then w = %<rt and (33) gives . Therefore V ~ == ec pt t fft = et~ Kt/S Now we know . 3 69. See (101).)(<rt). Perhaps There are tables (33) is the best for numerical calculation.l. of the . so we obtain V at the origin also... So (88) is correct.. is the voltage impressed at the origin that does produce the current wave (32) a bove.

and the P H functions are closely related.I n functions at the end of Gray and Matthews' work on Bessel functions. that this .

The outer boundary. But there is no The disturbance there distortion at the very front of a wave. and is known in terms of the disturbance at the origin. (0 therefore. 33S have made use of the usual principle that the characmust be satisfied. is at the teristic We front of the wave. . in which the V solution is derived from that for C. with this extension.ELECTROMAGNETIC WAVES. and travels at speed v. x = vt. and also the boundary conditions. The easy way siderations. + 1)W. consists merely of what has been transmitted. that they have to be satisfied whe n a boundary is in motion. as above. might lead us to overlook it some other conFor might go thus.

because Now the factors u. not merely because the sign is wrong. use /? differentiations. But if we. &c. u does not vanish at the wave fr ont on the + side. as shown by ' ^i we shall p= r^i p =(1+2/3 " 1+ . if we make act by integrations. instead of as in (38). but . we know. instead of w. (36) V^ = ft j-~j LwCcf*. (37)Now. u 2 . but it .LrCeP* = (/8 . w 2 &c.)Po (88) come to the same result (33) precisely. as J3 integrat ions.1)W. we get This is wrong.. But in getThat is the same ting (33) we made a act by differentiations.

So (39) belongs on the left side of impressed there. practical note to or ?r l We make to a problem concerning the origin. First. as in (33) or (38). The is that we must use a differentiations integrations. due to a negative voltage are not concerned with that. which way of expanding the m fractional operators should be employed Answer.we have does on the affairs side. so as to get w factors1 * the operand to the I P m (~) functions. in order to introduce and not un when we are c oncerned with the wave sent from . a and /3.* There are two points in question. B the origin to the right side.1 are equivalent on P m(-) x u m or x ?r" Secondly. .

or /3 differenti ax and 3 got by using ft integrations W W . Nevertheless there '. interest in the matter. tions ? Thus. It is clearly What is given (0-l)W = P indeterminate in its expression. is CH.say But what is the connection between the two principal solutions. W? . VjJ.334 ELECTROMAGNETIC THEORY...Thus.)P0= _(p \ \\\ o + wP + 1 " 2 p.+ . without further information. some further . (40) W^-JjL = 1 p To see the The result is U+W -(l + /i + ^ + .

So . (43) functions is an important identity.2 = e <rt or. (35).. which is the same. + ^P. expand them. (41} difference between and W . we reduce this ident ity to the old one.. / 4Ct\ (42) and for the same without the 2 we obtain Wj . The argument of the P z.. We may .. *=P This 'is H + *)PI + ^P.+ also write 2 . When x = Q.. W W first term.

.. makes V eP* = ee~ . combined with p ^It from x to x . given (46) to represent the V wave. (43).}P =(2a )POJ (44) to receive all integral values See also (29) and (29A) above. which make. (45) effect of alternating signs.it thus )t ^=[l + a + ^ + a2 + /? where n has +. In connection with (33) we may take note of the Thus.

.fft at the origin.

be obtained by the second of (31).2 + .P (47) Also note that in we are not restricted to integral values of if m Pm solutions. which makes = . 1 . (48) shows the impressed voltage . .2a + 20.. .ELECTROMAGNETIC WAVES.2a + 2a + . V = e~ nt. Thus. . that 335 is.L is ing current wave may 1 The correspondthe impressed voltage.

/?. That is.p). 382. V and C Waves when V is Constant V = *e~ K'/3 Expansion of 4* in I Construction of the Wave of V due to any . From this the current wave when V = e. Referring to cP* that the operand e fft e there used has to be altered to t&e. Deduction of the from the Case Functions. when the ing impressed voltage is constant.V . the curr ent wave is . equations (106). constant. We know that = ec**. . then the V wave is (49) This with is it done by simply generalising (48) so as to harmonise and contain powers of w as i nitial factors. P-P First put (50) p in terms of a. we see or the same as operatoperator p/(p <r) in (107) top/(p />) on the old solution b y (p . This is the same a s changing the 368. (107). when V deduce Now pass on to a further extension. Impressed Voltage. the current wave is CeP * = eP Q (z).<r)/p .

will not do.Then powers of a + /3. Besides. the result p-p But This if we expand series of this in rising must be a Pm functions with um as well as wm factors. we found before that we should .

if c = 2/>/ofor . substituting or for Do so here. in (51).33G ELECTROMAGNETIC THEORY. namely. The result must now be a proper w M P m brevity. differentiations CH. series.p p/ that is.a 2 ). V CT 1 1 2/Hx/oa' 2 (58) (T \ (T a series in rising powers of (2/m/cr . VIU employ a or ft /? 1 therefore. then get We ) p. integrations.

(76). + c% - . the law (c is 11 obvious. to verify that it is correct at the origin. where the a) ." value of C at the origin is expressed in a special form. Then c(l = el + -c{l! + cl.i O ^ I O * (54) Though complicated. Put # = 0in(54). n I. since the coefficients follow the rule of There is only one thing needed now. This we can do by comparison with (77). 364.

find the V wave due to steady V at In the the origin. on it by Operate the operator (52) expanded as in (53). (76). 1-a p-p 1 ^ 2 The result is *(l + 2a + 2a' + 2a3 +. + 3cI 6 (55) In (54) it" is z. latter case we have the solution (33) above . +c I4 . ^. the argument of the I functions being art.. shall obtain the previous result (77). exactly. by d eriving it from the case V e/^ = e^..}.3rl. If we Therefore Continuing the above. (54) is correct..2cI 4 + 15 I 7 +. )PQ c a + 2a + 2a + 2a + : 2 ' ! .. now we write out th e first few terms of the I functions.

}P 8 5 ...) + c'-(a :1 4 ... ..) + 4 (a' + 2a + 2a + + 2a"+ .. (56) .. . ) .2e'(a + 2a' +..) + ." 4 fa2 + 2a + 2a + 4 4 3 . . . .(a + 2a + 2a"> +....

at x= it makes (58) It Vis l + 2i + 2 1 ll + .(<rt). 337 Arrange in powers of a. 2 Here an V= e..ELECTROMAGNETIC WAVES.. and then turn to w m Pm functions. so . To verify. We get (57) This is the complete wave.

Then 2 VcP^ZA^P^) \n is (59) series is the required result. will be Identical Expansions of Functions in I n Functions. for and also shows how t o obtain every possible variety instance . this is the process whereby the comp lete V wave can be developed when V is given as any function of the time. it and can be done specially in special cases. Expand V P* in the form AJn (<rt). a physically-minded man can at once conclude that every possible sort of m . Formula 383. in the light of the preceding. for (%x) n Electromagnetic A pplications. The process of expansion in the Bessel merely the equation of coefficients of powers of <r. provided it be consistent with the conditions imp osed. since in the real motion there is no limita tion to any particular motion. certain kind of function represents a normal vibration. Furthermore. identity. But more satisfactory now to give th e general formula.we have an expansion of t? 1 in I functions. to harmonise with the boundary Then. . It is a commonplace in mathematical physics to require the expansion of an arbitrary f unction in a series of The dynamics indi cates that a functions of a given type.

of course. And so it of this normal vibration so as conditions. but with somet hing quite .otion is included in the special normal motions. and therefore that an arbitrary function can be expanded in a series of normal functions. is But that it we have not with these expansions in normal functions to deal at present. always is.

But when we say that f(x) = form f(x) . or 1 = 1 or x = x. identity.g. it is with an absolute or true identity that we are concerned. VII. If it were so. and f (x) i* expressed by a power series. &c. CB. So we do not have identity. or if we 2 have f(x) identities. = 0. the function f(x) has to be turned to the in order to preserve the equivalence. and to satisfy some conditions at the limits . we should have an absolute This may sometimes occur. series assume i nfinite values. to hold go od between certain expanded limits. aided by mutually destructive additional terms. then f(x) it is not usually the case that th e coefficient of x on the left side is the same as on the right side. e.338 different. when f(x) = sin nx . but merely equivalence. By mere rearrangeme nt of parts. a function of x being in sine functions. the form of f(x) must be varied in different ranges of the variable . If we have f(x) = An sin nx. under limitations for example. the coefficien ts of powers of x in the normal but. ELECTROMAGNETIC THEORY.. under similar restrictions . x 2 where I n (x) is not a normal function (or is one only in a changed sense). 2 . or similar ly for any other power of x. = 2 A nJ (n#). A nln (#). it means no more than . although these equations are loosely called identities. they are not real An absolute identity is such that the expression upon one side of the equation e xpressing it can be converted to that on After cancelthe other side by mere rear rangement of parts. lations. in general. Expansions in normal functions are not of this sort say that = A n sin nx.

&c. It follows .. t. difficulty. without indicates plainly enough that the either possible in general.= A. As an easy preliminary example of this sort of expansion. s.en +i. coefficients (60) where the r. Our physical problem is expansion from which And so it works out.. consider the function where n has any value. and the alone.JX*) + AAW + AA(*) + . We have xn/[n = en . or else is the for m to derive other series in cases of primary failure... A are determined nature of f(x) without reference to by boundary conditions.

. coefficient of xn+ * be ze ro. This function is the /\7i term in I n (#). save for failures when n is a negative integer. it n (^x) first . with a proper factor. the Bessel expansions are by no means so simple as the last. this formula is valid for any oo to +00.. Then the addition of In+ 4 with a proper factor will make the And so on. Therefore introduce In+2 This will n2 make . and so on. or gives a useless I_n when n is a negative integer. But it fails. the coefficient of x + be zero. like In = construction. . This case. but the same principle is concerned precisely. Say Now is that is to be expanded. Thus. (62) 2[ +2 &c. whether the n's be integral or fractional. wh ich involve xn + 2 x n +*. the right side is redundant in the other terms of I H (^). .ELECTROMAGNETIC WAVES. 339 obviously that any power series can be identically expanded in a series of e n f unctions. so all the coefficients become known. The result is By the value of mode of n from identity. So if we equa te f(x) to I n(#).

) (x)..) (*).7I + 9I }aj=(I1 -8I 3 + 5I 5 7 9 -. as will the case of the logarithm.)(*)..)(). (64) (64A) (64B) (64c) .application to electromagnetic waves.2I6 + 2I 8 -..) a)/ 8 n (*).. =(^)V 2 (I2 4 I 4 + 9 16 16 18 + 25 1 10 -. will be done separately.5 15 + 14 1 7 . = (I3 ... 1 . .3 0 19 + 55 I ... (64D) z2 . Examples its : and = (I -2I 2 + 2I4 .

VIITo make no mistake about the the last formula shows that origin is if for x. then the wave of voltage generated ^ r 1 2 ^" * 14-'J n n cancellations bring us to x = x . . application. Thus. provided (ab)x is any index concerned. mean . more symmetrically. Since (63) is an ident ity. to show the st ructure.340 ELECTROMAGNETIC THEORY. The formula may be written. it does not matter what xn x n +%. For in any case the numbers. m jO '1 (66) ... m = axm hx m when a and b are &c. put a-t CH. then expressed by is V Q tpt = the impressed voltage at the e(<rt)*/ J. the function In (crt) being turned to (w n j\n) Pn ( 2 ). and .

. The most important case is useful formulas. from the beginning. xn Examples. tion with one another. the expansion or (66) repeated again and again. .1. we are in fullpobsession n There may be any n umber of the expansion of the series n# of terms in this series.2. terms and get further. j being Expansion of any Power Series in I H Functions. But should there be no connection in I functions will consist merely of (63) given.. let B + Bjftx) + B Given the 2 . with various values given to w. Having got the expansion of . &c. 2 Thus. and we see because carrying it backward will . and with initial coefficients B. [ only introduce zero terms. step 1. 384. as Then we can collect unit step or step 2 from one to the next. and we cannot simplify But practically the indi ces will follow some law. infinite.It is now it is perfectly regular that a complete series. with for first index. and the indices need have no connec2B .

This will the coefficients of the d ifferent powers of x.. find the A's... (67) There are two ways. . . equate then A 1} then A 2 and so on. . The other way is firs t A give. = (A T + AA + A 2 I2 + . . First B's.)(*).+ .

: . way we come JS B B T + BJ.8BJI.ELECTROMAGNETIC WAVES. this does not go far eno ugh to exhibit the law of the A Nevertheless. + (B -4B 3 + 2B )I4 + (B5 . to the result 841 Either and rearrange terms. and is exhibited in the following table of the numerical coefficients in the A's. . + (B 2 . to use the formula (63). it is very simple. .2B 4 )T2 + (B..5B S + 56^5+ (68) But the argument of the I functions being x throughout.. from A up to A 13 coefficients..

2I 9 I 10 + I u + 2I12 ) + table.)(&). . from which come e.29I + 76I9 . 381.. VII. where the specia l electromagnetic problem is con(35).199I U + . cerned. (72) These are constructed by the application. = (Ij . .. ...x shin x and cosh x. Compare with (33). CH.4I + 11I5 . sin \x smz = (I + I3 + I5 + I +. if As an example of the impressed e is is voltage . 1 7 (70) (71) 3 7 7 I8 . .842 ELECTROMAGNETIC THEORY..) (x). .

given by i3 P 8 + ^P6 + U> .+ because it .. to. and the characteristic everywhere between. may the development to any be referred extent 382. the formula give (58). according to (70).. In the above the step in the index was 1. This includes the B's. is given by Jsincrt then the wave of voltage..+8p.)(). ) (73) As another example.. steady impressed voltage. (75> satisfies is right at the origin. relating toThe table will required. and at the wave front. step . where constant. Another example is This will make 1+ 8P.=V pt/e.. (*).

For this purpose use (66). and other cases. . Thus.. by the vanishing of certain of it is now necessary to give the formula for a power series when the indices are not integers... but still so that there is unit But step from one to the next.2. 0 (1 (n + 2)I B+8 + (2 (n + 4)IB+4 .

collect the I n terms together.ELECTROMAGNETIC WAVES. and that the coefficient of I n is The result is A m This 10 m-2 II to -4 12 -?!^^ + ""/ \ n-6 \n 18 ~~ it may is written symmetrically to show the structure. 843 Now so on. and the I n+ 2 terms. be simplified to But (n " 6)(n 7)( .

and the coefficien ts A will in either case be finite series. on indefinitely.. and the first B i s B we produces A n + m obtain the previous results. series is endless in the negative direction. by exhaustion of the B's in des cending order. This formula applies to all the A's.. The series for A n must be continued until it stops. But if the B series in . going backward. in (68) and the table.t will give a Bessel series expansion inI_. whether th e series volving f(x) stops or not in the positive direction..^=2AJ. &c. save near the beginning.. or with decreasing n. . will make . then the This occurs when t he power series for An does not stop either. assumed Thus the power (80) f(x)=ax-* + bx-* + cx* + dx* + which begins with x~*. Ij. I_j. changing n to n + m When n is integral... But the series (81) which stops at the term .9)B U _ 10 + . It is not (79) does not stop.j. That is. . that the series B series begins with any particular n..r?.8)(n .<*). (78) (i which also shows the structure.. 2B.

But (79) will be a n when f(x) is identity for all that. And should f(x) be endless in both directions .the A's be infinite series endless in the negative direction. . we may expect that the series for the A's may be divergent.

2.. we may regard the B endless in the negative direction. A = -2-B_ -2B_ ~5B_ -. CH. I [ B_ 1} B_2 may have any Unite values without interfering with the value of the power series being 1... when the power series is endless backwards. The formula (78) makes A = 1 as before. let all the B's all values 0.. 1 3 5 . series as = 1. 2. VII. and st opping at being . are infinite. and then give n Or.. and therefore the term B x. This matter of the endlessness of the original An series. Let f(x) = 1. be zero except B to find the A's. &c.. we shall obtain endless series for the A's.344 ELECTROMAGNETIC THEORY. &c. then ? Test this. in (78). See (64) above. &c. examp le result already. Will th e results be wrong.. 1. But in using (79). turns up in a One curious form when integral powers are concerned. For 1. A!= -B_ -B_ -2B_ -5B_7 -. We know the proper will be sufficient. 2 . an d then . But if we do this.

(63) should Expansion of a Power Series in Jn Functions. .2 4 6 . Practically. and the true result already obtained is arrived at. Examples. But I_j is the same as I lf and I_ 2 the same as them together. Doing A_2 = . .2 . we shall require to include I-^x). find that A_1= and lf -A formula rejects all the redundancies. I_ 2 (a?). but of J . the second although the and I 2 (#). all the arbitrar y B's are elimiI2 So. we shall require the expansion of a function of x.A 2 this. be used. The explanation arbitrary B's below is that in virtue of the inclusion of the . 385. with n = 0. joining Our nated. to be made shortly. not in a series The connection is of I a functions. of course. The first may have first term is correct. These are the coefficients of I^ar) also. In another application. B we &c. and similarly in other cases. any value.

I n (xi) = i*J(a/'). and the quesnow how modify the preceding formulas to suit the changed circumstances. . and are to is. Say that (82) ZB-ZAJ^-ZEAM. The expansions tion still an identical nature.n functions. this makes 3 n (x) be an of oscillating function.

The right side becomes 2/B. then the 6rst equation 2ByM' = 2A ^J H n (z). (83) fe In this use (77) or (78). 1 (n 3)B Now for B iw H rt - . The &c.ELECTROMAGNETIC WAVES. Put xi for x. 345 We know becomes the A's in terms of the B's.+ 2 B n_ 2l since i -+ 2 = 2. 1. Now find the E's in terms of the B's.i. i* = 1.

] J n (z).+ write . B n_ i"2 writo B H_ &c..3)B._ + 4 . } J m (a). B tl _2 + 1 ( .. . result is " the required result. for (84) 2 Bn ._4 i. . En = BH + . (85) That is. .

when the indices are integers beginning with above becomes 2 B JM!1 = B The same Examples J + B!Jj 4 + (B2 + 2B ) ) J 2 + (B3 + 8BJ J 3 : . to 0.. . with all signs taken positively. (68) produce E M Thus... to alter all the .B_ + .t series. + (B + 4B 2 + 2B : J 4 + (B 5 + 5B 3 + 5B ) J 5 + .3)B n _ 4 + (* + ..signs in the f ormula for A n to + signs. therefore. To expand the function in J. (88) cos x 1 (n 4)(.5)6^ . (87) table is to be used. we have merely .

)(x) = J + 2J 2 + 2J + 2J6 + . . by the fourth row of figures. . (93) (94) -7I3 + 41I 5 -.)() (90) These are done by the table.)(*). ( 4 ... J -J + 7 .y*.= (J . ( 4 . the 34 in the last formula 2 is 2 4 x 1 -f 2 x 4 + 2.. 5 (91) (92) 2(J1 2(I1 -J + 3 41J 5 +..2J 2 + 2J . (89) .. But this is not general... Sometimes the expansions in I functions a nd J functions are quite similar. E.. 5 + I 7 -f. . Thus..2J e . . except in the signs.)(x) *= J + 2 Jj + 6J 2 + 14 J 3 + 34J + 4 .)(z).)(*).)(a?).

sions of the identical kind in J w functions. in general.346 ELECTROMAGNETIC THEORY. VII. the above method has the adva ntage of simplicity of reasoning and of working. a nd by the properties of Bessel functions. OH. V m Pm (z) is another way of obtaining the waves of and current generated by a source at th e origin in voltage the form of a series of waves of the type c~**P(*)trt. -**V 2 ( . we may operate upon the known fundamental solution P (z) in a suitable ma nner directly. process a special example was given in being constant. Th e fractional cases. In386. as already explained.. The Waves of and C due to any Y developed in w Functions from the Operational Form of V. . besides being comprehensive. There stead of. Perhaps this may be the best way sometimes but I think. that (95) V= where r stands for or 2 . In treatises on Bessel functions will be found various expanThey are usually derived from definite integrals by trigonometrical processes. for example. expanding V e/^ in I functions. relating to VQ. We know e.-. It can be generalised thus. Of this 382. and then general ising. and these might sometimes be very difficult by integrals and trigonometry. are done by the same formulae.ep*.

Now we know that .Vp = e-*V Now let the operator which generates a function out of 1 be denoted by enclosing the fun co)i/v. V^^VoC. tion in square brackets. then (98> But we also know that therefore (98) reduces to v the required result. for example. W-yT^r Put this at the = end of ^W 07) (96). Vrf*-. . *]!.-TV6 or ' i *]. Then (96) .

and [FJ 7** . Correspondingly. reducing ta G when x = 0. a function of x as well as of t. and [G This is ] their generating operators. It may be simp ly shown thus. But if we change G to G. t. 347 = [V ] P (*). . (101) by (100). Thus. in passing from (98) to (100). being mere mathematical jugglery. we can do so with operators. the n F will also become changed to a functionof x and if t). (102) obvious. we two steps forward and in passing to (101). Then 1 F = fFo][G ]G . The structure of (100) is worth notice.ELECTROMAGNETIC WAVES. Let F and G be two functions of t. The advantage of conversion of functions to operational form is that w hereas we cannot shift the order of conjoined functions and operators. if we work properly. we shifted eshifted the fractional o perator one step forward.

F = [F ][G ]-iG. equation (100). function of t. and it reducing to I Q (<rt) at th e origin. the special function of x and t. . complete But time differentiations on makes V is not 0.-. When x wave. and G is P (-s). = in (100). Put # F V e^. x and . by its c onstruction. (100) represents the . G a certain characteristic partial (variablesTherefore F is theso does F. satisfies t solution produced by F just as G is produced by is G . P*=[V ^] 1.Thus. which is any In our application. (103) Now.

P (2) are complicated. U se them. We can at once eliminate /?. This makes P (z). Put Jo-(a + /3) for p. (101) become . because 1 ftfor and a are equivalent with operand p in (100). The proper simply working differentiators are a and (3. So put Jo-(a + a-1 ) p +a p (105) Therefore (100).

Example. the first operator is to be expressed in terms of a. a nd Functions. we get this interesting application of Take x = in (107) or (105). of course. When this has been done.848 ELECTROMAGNETIC THEORY. provided to [F ]l. That is. to pure mathe matics. fe (108) Remarkable Formula for the Expansion of a Function in I n Modification of 386. if the combined operator acting on P (z) makes a power series. First turn (109) an ] is turned to I n (<rt) F of . VII. and Examples. V function of the time. [F in the result.^). physical ^ in I functions. l is its expansion in I functions. CH. where. s ay. -then we at once get the full development 3 V o( =2 A. Incidentally..'-! . F being any Then we expand $ 387.

Ry using (109). It is obviously true by the method employed. Then multiply by . and is the expansion of F in I functions. then put 2 1 . all the previous results of the kind got. an means The theorem is very striking. %<r(a + ar l ) for p. It is may . a simple but by pure rigorous mathematics there seems nothing whatever about (109) even s uggestive of Bessel property. let alone the result./?. being a function a2 ) (1 x (1 -t-a )The if result is 2A H an I n (<rt). based upon a special use of (103). functions.

. (110) ' = J^oV> (1 ( + ay+i' Expand in rising powers of a by division. with the variable vt here instead of x there. Say a ~~~7/T then (109) makes (i^) n~ F _(icrt) _/<ry_/ \2P ) VfT^V y.be only necessary to give one or two examples for the sake of explicit illustration. The result is the expans ion (63). or otherwise.

Expand by not wanted. Again. Thoug h will serve as a good example of the treatment of operators. by (109).p. will be enough.p)]. means [V ] with p . of (100) is just worth mentioning. And. We have = [V so (p we may put the last result for [V ep] in (100). {1 and then put . Say (116)__. A modification it The result is equation (71) before got. using a instead. 349 a2 +(l+a2 ) 2 Therefore. Now turn p to p. division. [V Qt? . from the former. (100) becomes ^ One example. different (us).ELECTROMAGNETIC WAVES.p put for p.

expanding by division just the beginning part.(2/>/<r) a + a2 } 2 + S2 (2a/of we get Using this in (115). we get the expanin I functions. as can be verified by the for mula It will be seen that this is not so good a way as the correct as far as a sin st tpt . <jr ^ J (119).Jo-(a + a. 1^ + o^/ + . with x = 0. or..) 1 for p. . This is sion of (68).>(*). For. V^'= ?? a/1 + *Ea aI o3 a'( \ 8. The result is a.

it is simpler to make V^ instead of V be the time function under treatment. as in the former method. equation (100).other.. That is. .

t (121) is . Thus. The result is . Although the development of t in I functions fails to give immediately. VII. be employed without change in these exceptional cases to obtain the waves genera ted by impulsive voltages. Impulsive Impressed Voltages. n 388. CH. Expand in I functions by the formula (63). the impulse the variation of does being not count in the no time of an impulse. let simple or multiple. the complete wave of voltage generated when the imp ressed voltage V n varies as e~Ph particular treatment and nis any negative integer. a^ the moment t = 0.850 ELECTROMAGNETIC THEORY. but requires a which will be given yet the formula . because ^ ^l = A(r(r_ 1 -I 1 ). b y the proper generalisation. and the Impulsive Waves and their Tails generated. may still (120) This means that e fl V is an impulsive function.

provided u difference is merely impulsive. By inspecting the formula for unP n . instead of the <rt of the I fu nctions. as now. and n. we have values of u and iv.Now Ix and I_j are identical when finite. pulses . here. In fact. and w are The _positive or negative. are actually in question. being the impulsive function i tself. we get but they are most important the argument of the P functions being z as usual. we see that -'i This finite is pi (123) It expresses the generalisation of (121). Generalising (121) in the usual way. The difference is not absolutely identical. = for all ^_ *for any integral value of finite. may in general.

be quite negligible But not when imUsing (123) in (122.. < 125 > <we get .

which is the same as i<r/?l. Behind this impu lsive wave is the continuous tail. falling from at the origin. Both are represented in (125) and previous formula for V. a t the place w = 0. and of a tail. in which V = Li-C = oo but with finite ti me totals.V starting at the moment t = T. at z = vt and its time integral is 1 y . increasing V we . and be followe d by V = 0. the factor c~P* expressing the effect of absorption in transit. But if we let V remain on only the very time T. attenuated by ence of the two waves. only exists part. rest expresses the tail. the same as the impulsive voltage at the origin.ELECTROMAGNETIC WAVES. and to an impulsive impressed voltIt generates an imp ulsive pure electromagnetic wave. . that is. including the head. Put on steady The result is a finite wave of V. finally come . of depth VT. and then dropping :short V suddenly to zero. ponding C formula is got by . which may. The and finite from the The meaning may be made plainer thus. then the difference of the two waves. continuous origin to the wave front. 351 This expresses the complete wave. in fact. the effect is the same as keeping on the steady V but followed by a second impressed . be the most important The first term. because u-iv = a-xjv.. age. resulting from the differShortening t indefinitely. consisting of the uncancelled part of the first wave. The real V is voltage . viz. V at the origin to V e ~ & at the wave front. at the wave front. and consists of a head.

'We have (128) .2P ). . and the conexpressing the impulsive C following it. this makes .y PI . the same results come out of th e operational solution (105).2 + a)P = 1 ^P_. + . Without using the preliminary expansion in I functions. J (127) -tinuous tail of C at the wave front.The corres= (a.2P (126) "Using (128)..

352 ELECTROMAGNETIC THEORY. CH. After this detailed notice of a simple impulse. which makes.. They are to be done in the same way as continuous functions. VII. If this is V^' (constant omitted). by (63). produces the multiply impulsive wave at the front being brackets. Thus.* shown in the The next case . expressing the same as (125). Similarly. by (105). if representing V p *. 3 (132) which. then This shows the multiple impulsive wave at the front. only the terms whic h vanish in treating the latter must be carefully retained. followed by the continuous tail. th e peculiarities connected with multiple impulses will be readily understood.

oU=W. the imrepresents the impulsive The other terms. and see that nections. Similarly wit h the other impulses in the text which. and consider that w~ /(-_3 wave multiplied by e^. for example. however. . we 3 appa rently ignore all except the highest. may needed in order that "W and L?-Ce^ may satisfy their mutual conConstruct the wav es U and W. (134) Since the terms in the brackets are of different orders of infinities. and /3W = U.is -4 * -I 2 ). It is. in fact. are pulse at the origin shifted in position. have more than one term in the expressio n of the impulsive wave.

in order to repres ent the impulsive wave at the front. involving infinite forces acting for infinite ly small periods of time. ~ substitute u n P_ n /[? for (n). n = 8. + 3(3) + 2(1) + 4(4) + 5(2) 1(6) 1(7) + 5(5) + 9(3) + 5(1) KB) + 6(6) + 14(4) + 14(2) 1(5) To develop \n these to Yd* waves. is to be able to represent with comparative simplicity effects which.wn P n Tendency of Distortion to vanish in rapid Fluctuations. . n = 6. &c. Say (n) stands Then the expansions of (^<rt)~ n /\n are n= 5. The object of using impulses. Let us first It generates an impulsive wave single impulse.but v astly more complicated in expression.ELECTROMAGNETIC WAVES. considered finitely. Effect of increased Resi stance in rounding off corners and distorting. we require a condensed notation. as the head decreases ac cording followed by a tail. Considered finitely. In time. w = 7. the beginning of a cable by an impressed voltage V . might be nearly the same in character. 389. After that for 353 I_ n In . send . and retain the terms which vanish when 10 is finite. the effect Let us ope rate on of a multiple impulse may be thus stated.

that tions. send a multi ple vanished.e a f* caused by absorption in transit. Then we see that the message need not be a . to ~ y The result is a nearly pure electroT.~P therefore. the different parts of the head producing cancelling tails. we could faithfully " " in its variasent by receive the complicated message V As for the tail. would depend upon t he time total of V principally. about zero. magnetic wave of depth vr travelling at speed v. in which all the variations in V are (nearly) faithfully copied. Let V vary anyhow with extreme rapidity in the small one. the head having practically But instead of a single impulse. it is the tail that is the significant pheno menon. and might But if V consists of fluctuations ther efore be serious. rapid variations of voltage or current. provided they were fast enough. the tail need not be of any importance compared wi th the head. and then cease. if sufficiently sensitive means At any distance. but interval t attenuated in transit according to f. existed of registering ti .

short multiple impulse. and to little no distortion at the near the front of the continuous wave due V By the steady. so there is little distortion if we cut it short. sufficiently rapid flu ctuations we use the heads and destroy tails. and is still travel nearly without distortion. VII. The important point here There is very front. CH. The tendency for all fluctuations. naturally includes the simply The variation of the attenuation p eriodic train of waves. if sufficiently rapid. to be transmitted with out distortion. Self-induction salvation. But the attenuator is e ~f>t is a serious factor in a long cable.354 ELECTROMAGNETIC THEORY. . extreme rapidity of fluctuation. but may be continuous.

in the theory at the place i. a distinction to be drawn between the ideal V perfection of a theory and the reality of which it . p rovided we keep only to rapid fluctuations. momentarily fixed by x vt t provided V commences at the moment t = Q. and likewise the variation of phase. Say that V is /(*). But there is the value of is/(0)e~P*. as the frequency increases.with the frequency tends to vanish. is . propounded. all along the wave from the o rigin to the wave front. at the wa ve front itself . the distortion tends to vanish..is not zero. under the And it circumstances stated as regards the variations of V to be borne in mind t hat (135) is always true. which is exactly true in a distortionless circuit.e. may still be approximately true in a circuit in which a . so that. then V=f(t-x/v)e-P*i\ (135.

duces a wave in which at the the impression of V V V V wave front (at distance x). as steadily at the origin pro~ x/v P falls from at the origi n to c circuit When constants of a imagined. in order to under given circumstances. owing to the ignored circu mstances assuming importance. Theories g enerally fail in application when pushed to extremes. to the reality. It is always desirable to develop a theory exactly when it can be do ne profitably (this has no reference to professedly rigorous to know what expect methods of working). without approximation to a definite theory. and then drops abruptly . the This has a distinct application here. by the usual process of ignoration. are truly constant.professes to be an approximate representation. confusion between an and the approximation of that theory is which another matter.

But they are no good. as the wave This means the rounding off of the two front passes a place. penetration of the wires by the magnetic force not being instantaneous.substitutes square corners for rounded ones. To get the rounded curve we must enlarge the theory. The reduction in L is relatively a small matter. because the conductivity i nvolved in K is so small. nearly to full value The relatively slow assumption of the rest of the full value. but L has a sensible variation. There are probably no real jumps like this in any physical phenomenon of wave ch aracter. to zero. corners in the abrupt jump. taneous assumption of the full value. will have the effect of removing the abruptness at the wave front. and augm enting the resistance in the way described. The imperfect penetration is a cause of such . lastly. K and S do not need to have any important change made in their values at different frequencies. and K may have a large one. and allow for the variation of the constants of the circuit. or of most of it. The wave front. perhaps not even in the ether What we should expect is the rapid. First of all will come a slow sharp and.KLKCTROMAONETJC WAVES. theory of constancy of p and o. then a ve ry rapid one. This is important precisely in this question of the state of things at conductivity involved in the The resistance increases as we pass from the wave front in the case of the conti nued wave due and may be largely increased at and near the to steady V wave fron t. Now. It is this 355 abruptness that is impossible in reality. but n ot instanitself. . origin to the . a rise. unless iron wires are in que stion. because the R is (relatively) so great.

being the same as R/2L -*. there is a cause o f distortion left uncompensated. Consequently.K/2S. does not tend to constancy. unless p should tend to ultimate constancy. it is represented by equation a (135). since it But it is not likely varies as the square root of the frequency. when calculated by the magnetic theory . which increases with the frequency. t aking count of the wire and As regards the state of things external conducting boundary. Now R. If we allow for the permittivity to follow this law for ever. but with changed value of p.a substantial character that other causes having a similar effect need not be co nsidered. tended to at a very high frequency. increased In the simply periodic case we can fully trace the effects of R and reduced L. .

containing a jump. . it CH. It tends then to ultimate The constancy. after first following the law just mentioned. because in the production of a wave when V is discontinuous. is no end to elaborations when we introduce minor involving other considerations than the four constants of a circuit. This or thereabouts. There effects. trillions per second. of the conductors does not. VTT. according to an old calculation of mine. harm onic analysis involves the consideration of infinite frequency consequently.356 ELECTROMAGNETIC THEORY. is excessive. so far as they can be determined.e. far more difficult but very complicated. it may be inquired whether it is feasible to tackle the due to V problem of finding the real shape of the V wave steady (a fundamental case) as m odified by imperIt is feasible. harmonic integrational method of calculation. we are bound to take count of the extreme tendencies. Keeping to the latter entirely. does not necessarily put the question entirely on the shelf. fect penetration. however. frequency needed. and their v ariation with the frequency.. by the . h owever. i.

wave formula. but for theory of this chapter. Z we must use the full resistance operator of un it length of their full expressions in the tive reduction to simplicity. derived from magnetic considerations. But then the type of the formula for a train of waves remains the same when R and L vary with the frequency as when they and since there is no necessity to write ou t are constants . whilst Z is R + Lp in the where q There is no need to alter Y. Consequently a definite solution in operational form. to allow for It is a known func tion of p. It is true that the expressions for the effective E and L at a given frequency are very complicated themselves in that case. But the work is so practical case of constancy of that I have not been able to bring it to a complicated R manage- . . may either treat the wires as mere cond uctors. on the other hand. and may be directly V algebrised by similar ways to those already employed in the and L. . or allow for = e~ qx VQ becomes their permittivity as well.It is than the simply periodic case. and we imperfect penetration. there is an effecBut the case is different in a continued wave lik e V due to steady V The calculating In the harmonic integrational method is impr acticable. we have V is (YZ)*. and Y is K + Sp. = e~ 5 *V operational method. the wires.

1. A difficulty arises at the very beginning. Say V c^x $-*. or $~V(jii ^ s impulsive. if V eP* v aries as t~ l with a finite factor. to with. so I say nothing the main question. begin We ^ know already that when V eP* varies asjpl. 857 now. Therefore. 390. but return to more about it Wave Due to Impressed Voltage varying as e"^ i. .1 . . Now t' 1 is an infinite multiple of t~ l /[ .ELECTR KVAGXKTJC WAVES. able form. Let us now tackle the postponed case of V eP* varying inversely as an integral power of the time. and generates an impulsive head wave followed by a finite and continu ous tail. the result front.

but this is not The practical value of an necessarily o f any importance. infinity may be zero. Why the V infinite wherever exists. .must be V = oo and zero beyond. The real reason why V is infinite all al ong. the problem is not it physically realisable. is from the origin up to the wave Put in this way. fact that V is initially infinite does not account for wave front. is is that the is. however ? The mere it. At must be infinite. of course. V although V is only momentarily infinite.

infinity. notice If is here three kinds of infinity. so But if fV dt is finite. of different is V always finite.momenthe time tary infiniteness is infinite. V contains momentary always does the V wave at corresponding points travelNow if fVQ dt is zero for the momentary ling at speed v. it has no importance. The infinity has nothing in i t. = before to just after just intensity. the V wave But if infinities. it if means a finite impulse. integral of V from So there is generated an infinitely impulsive el ectromagnetic wave as head. We may significance. and the mere dregs of it make a tail of infinite That infinitely impulsive. finite and varies in any way. generating .

fV V is kept on. . we may send an catch up the It can never impulse. between the two impulses. This will continue for ever. of course. infinite negative first To restore finiteness. so. if is i nfinite. and so is the tail. impulse. if of the right sort. a finite tail. V must be infinit e. though finite save initially.dt is infinite. the impulsive wave Finally. but it may be finite in the rear of the second impulse.

Cutting off the begin ning is the same l . enough now. except at the . and which allows us to use the old formula. and l . VII. There are also multiple. differ series is the result V . CH. In the case of V eP'x t~ to have finite results we must cut off the first part of V Say The by having V ^oc (t + r)-~ where T is small and positive. another wave solution having a different geneSimilarly. . There is another way wh ich is neater. sending an infinite negative impulse initially to cancel the infinite initial l impulsiveness of t~ head. as superposing the beginning infinite part negatively without In the limit this means cutting off the beginning positively. wave is finite now. These derived solutions from the original. The result is finite now. by further differentiations with respect solutions. inasmuch as they are due t .ELECTROMAGNETIC THEORY. If and the rating to n. on the right be differentiated with respect to n. but one is infinite impulses of higher orders. But a fresh formula will be wanted.

and negative. for our pur. Thus. as W3 may see by developing the wave. and g'(n) is its derivative. and varies inversely as a negatively integral power of t. a negative integer g(n) = Q. It doe s the infiniteness of the wave generated by so. so it may destroy l alone. poses. Say n = then we get (138) The initial impulse is infinite. = 0. f . The result is (137) if reciprocal of . only begins at t The 1 function. but ca n in series involving I n and its differential coefficients we generate other with respect to n. differentiate t n / n with respect to n.o impressed voltages which cannot be expanded in the form 2 A n I n (o-). Then (137) may be taken to represent a function which is finite. g(n) means the is When n together with an initial impulse. and g '(n) is finite.

I'_ 1 means eP { ... or + 4)l.ELECTROMAGNETIC WAVES. Thus. .. We get + (Jo*)..1\ -l l + I 3 . (140) X is -1 and the accent means differentiation dljdn. (139) X where and then put n= -1.l+4 Differentiate to rc..= lU . 359 Use formula (63).I5 + jU 7 the initi al auxiliary impulse given by . (141) to n.. 1 -. with 71= Consequently. the left side of (140) represents V . if 1 after the differentiation.

the wave it generates is (n= 1) by generalising the I functions in (140) in the usual way. T he terms on the left side must be on the right side. Like (139). differentiate Thus. the derived equation (140) must express an absolute identity. like life. and the rest on the right s ide repeated must come I n to 7i. to 0. To verify this. I'n .

. (144) l\ = I.. difference.= In log ^t + (Jrt)V (n) + (J<rt) w + . (145) Taking the !'_. log Jcrt + icr^(l) + (Jo-0 8 !_. + (Jcrt) . 5 + + . we obtain 3 ( - ... (1 43) from which I'_! = I_! log Jo-i + 1 ( Jo-0+ ( Jo-0^) + 3 ( Jo-0 ^ c.....

.F..X + (J^). .. (H6) .+ (1-0 + J + + .

by (68). be generalised to a eP* but that is wave. Equation (1 47) may. (140).860 ELECTIIOMAGNKTIO THEORY. Its truth may be n ow verified by (63) same formula from which we started. or. But (147) by itself is of no use. Vlt. a power series in I functions. Comparing with (140) we Here we have an expanded the identity of the original kind. of course. another question. more easily. The terms omitted from (140) are vital in the electromagnetic wave problem we started with. see that Ull. It is (142) that is the generalisation of differentiated part is . if the left side b e taken to represent V .

C wave is (148) so the full Vf^^iog^J+ffp. (149) front.. complete.and requires development.. and is finite everywhere save at the wave As regards the values of g'( n)..log* -1 wv e ( -^i-iaThis is . (150) .... The 111 .. they are r^z.

proof They may be found in treatises. is still. must be paid . 36 it 1 when n easier is a positive integer. 3.ELECTROMAGNETIC WAVES. on the negative side. and produce may 1-1 3 it* w v 5u.5 I explicitly have given a good deal of detail in this case. but the treatment of the subject in high mathematica l works is not of a nature to encourage physical students. we terms involving y. thus. may be similarly fully Particular attention developed. I must defer the of these side results. &c. in order to show how the cases n = -2. truly. What is necessary for physical purposes can be more simply Xot being able to given. collect Taking the constants for granted at present.. (151) do everything at once. Also.

and . and so have apparently a finite wave (except at the l All condit ions would front) arising from V^P* varying as t~ seem to be satisfied. otherwise the mathematics is useless. $391.to the impulsive electromagnetic waves are wanted. includin g an apparently identical expanl but the initial momentary sion of t~ in I and I ' functions failure of this expansion would be fatal to the electromagnetic terms if real . Another case of "transcendental" character occurs the impressed voltage va ries as ~P t logt. Wave Due when to V t varying as e~P logt. vitality. Nothing would be more natural at first than to omit the impulsive term from both sides o f the equation. . The logarithm of t cannot be expanded in a convergent power s eries.

CH. dn ^ = log* + y. we Or thus. therefore also not in a series of I functions. the wave generated is when where n is to be put =0 after the differentiation.362 ELECTROMAGNETIC THEORY. and come to a numerically calc ulable formula. So if V cP* = e log ^a-t. But log t is the value of the differential coefficient of tn with respect to n n = 0. Now shall develop. VII. using the values of g'(n) already given. By (137). .

which can be generalised to make an electromagnetic wave. log Joy(T 21. for example. So we get an expansion of log t in I and I' functions.) where the coefficient of 7 have V e^ = elog (r. we have ( (157) . we equivalent to 1. To develop the last line.\n whenw = 0.2I 6 + . The fundam ental expansion is functions. . (w 6 P 6 )' means w n P n )' with n = 6. the wave generated is constant.2 + 2I 4 . where e is is If then. where in the last line. (154) from which we see that logt + y is expansible in I and I' But 7 is a constant. a nd expansible in I functions itself..

Also observe that by differentiating the V wave with respect to * w e can obtain the wave due to the impressed voltage At first sight this looks as though a similarly differentiated. although the value of V at the wave front is infinite. Consequently (156) becomes 363 5w*p 20if p _ \ lal [..&c -> in ( 158 ) which everything It is to t.'(0) pjT ..ELECTROMAGNETIC WAVES. when V e^ varied as t~l since ^logi = t~1 by . is known.] + uwff(l)y(l) + uWg(2)g'(2) 4]. so no impulsive terms occur in the wave.. is be observed that there Its initial nothing impulsive about the integral is logarithm of momentary time zero.

p(logj<rt . in fact.finite wave arose But that is not true here. Effect of a Terminal Resistance as expected in found in 1896. . but it is necess ary for some one to do the work though. to break the monotony inseparably connected with r egular developments. 1887 and as 392. (159) it This is the true differential coefficient. Therefore by time differentiation be initial led to the wave generated by V l varying as t~ and the impulse. Now for a little change. It is grievous that they should be so dry. who may . there are cynics do not see the necessity and. since includes the initial impulse. l) = log J<rf . So f unction log Jo-t is zero before and existent after t = 0. of course. it sa y they . The the differential calculus.pl+tr we shall eP* l .

an attack of influenza. easy to become cynical oneself is after. itself unnecessary without question. which a demoralising disease.is . need not depart far from the preceding environment to We obtain the change of air and scene desirable for reinvigoration. . say.

But we need not go into th at now. we may sum up the action of terminal " apparatus " in the form of a was terminal condition. The proper differ materially according as the waves are long they are really long without q uestion. with extended meanings of the symbols and operations. and further still. of condensers. some degree what happens to them when the end of the guiding circuit. Leaving the subject let us consider in they arrive at treatment will or short. except to remark that long waves are essential. as in telegraphic and telephonic applications. The treatment is quite similar when self-induction and leakage are included. because such an equation as V = ZC. only does so on the hypothesis of instantaneous action and reaction betwe en the different parts of the apparatus. about which a good deal said in the last chapte r on diffusion. V = ZC. resistances and inductive It must . But it may well happen that this proced ure is insufficient. VII. say.864 ELECTROMAGNETIC THEORY. when constructed in the usual way to represent the action of a set coils. CH. If of the development and progress of waves.

the wave length to be only a fractional part of the size of when clearly there is no opportunity for anyan equilibrium theory (in a certain sense) to be established by mutual actions. still obtainable will be a under certain circumstances completely circuit were absorbed by a terminal resis tance. the circuit is where L and v is . A resultant terminal condition of the . which are really the resultant of wa ves transmitted to and fro at finite speed between all the apparatus. Take the very simplest case of all for initial illustration. If the circuit is distortionless. thing like parts of the apparatus. and the amount of the terminal resistan ce Lv. I showed in 1887 that electromagnetic waves sent along but the form of Z and much more complicated.certainly become insufficient when the waves are As an extreme case we may imagine shortened sufficiently. form V = ZC is quite different.

the . per unit length of then the absorpreflection. So. if tion is V = LfC. tion is complete. is The relation between B is the terminal resistance. or there no terminal The reason V and C in very simple. if E and Lv are equal.the inductance is the speed of light. the circuit in a wave of any sort going in the positive direcThe relation at the terminal is V = RC.

for instance. there is no back effect at circuit is not distortionless. There will be a different theory for every different arrangement of resistance. since distortion takes time. V requires to be generalised to a Z of complicated structure in Similar remarks or der to represent the course of events. that is. the cumulative. is The same thing may be approximately true when the and With very high frequency.. it is clear that the k ind of terminal resistance requires consideration. this is not altogether true as re gards the terminal resistance. R viz. must apply to the numerous experiments wi th waves along . where R is a constant. resistance behaves to the circuit itself in the 365 same way as a continuation of the circuit all. for the The waves must evidently be rather reason given above. shortened sufficiently. in the truly distortionless circuit itself thf But although consideration of the frequency is unnecessary.ELECTROMAGNETIC WAVES. If they are long waves as regards the terminal resistance. reflection may be nearly destroyed. In another form. even though in the long.w ave theory they would be all alike = RC.

336 to 1. and its use was abandoned. 1897) with waves 8 J metres long. E. But a resistance of intermediate va lue. A small coil wound non-inductively was unsatisfactory by It is quesfailure of insulation. apart showed that a fair approximation to (length of line 116m. Barton and Mr. Mag. Another of 1. produced a quite small reflection in com . to how short waves may be allowed to be before an assumed mere terminal resistance needs regards It is its to be studied in detail as reaction upon arriving disturbances.. when terminal effects are in It is difficult to make more than a rough guess as question.wires after Hertz and Lodge. however. on a circuit of a pair of parallel co pper wires 1-5 mm. the results were satisfactory. to observe that the experiH. Bry an (Phil.355 ohms produced still larger reflection of a different kind. B. ments of Dr. January. in diameter and 8cm.). complete extinction of the reflected waves could be obtained. But when the terminal resistance was constructed by pencil markin gs on glass. of 549 to 560 ohms. G. about the value of Lr. tionable whether it would behave like a mere resistance apart from the question of insulation. very interesting. A terminal resistance of 261 ohms pr oduced large reflection of one kind.

The area covered is not stated. though pro. we have to regard the wires as mere guides. if not more. Having a sort of kindly paternal interest in Dr. . but it could not have been large. casts It light on the subject. and distribute their resistance uniformly in the ether outside them. they too must be trans ferred to planes at the proper places. VII. though quite as much. viz. forming an extremely thin sheet. Having got truly plane waves in this mann er. Vol. or at any rate in the results expected and obtained. I have sought for and found the exact solution of a case of waves along a straight wire circuit terminated by a resistance. as regards the particular sort of terminal resista nce concerned. I. pencil markings. parison with either. if there are any intermediate or terminal influences.. not as electrical resistivity. fect extinction of the reflected bably not to the same degree. however. Barton's experiments. regards the transverse dimensions of the line itself and the experiments show th at they are also fairly long. CH.. At the same time there was far from perwaves. admitting and C theory of waves remembered that in order that the some V along wires may be an exact theory in plane waves. and is fortunately of a kind In the first place it should be of easy description.>bb ELECTROMAGNETIC THEORY. so as to act evenly on the plane waves. but as magnetic conductivity. as I have explained in Chapter IV. than I could have ventured to expect w ith cerThe waves of 8 metres are certainly fairly long as tainty.

or ratio of the reflected 1} V 2 to an incident dis' turbance V is V2 = _ ^V =_ Pv n\ See " Electrical Papers.. p. Similarly a terminal wire bridge must be replaced by a transverse plane sheet of the same effective conduc tance. The second which R . Now. an intermediate conducting bridge across the practical circuit mus t be transformed to an infinitely extended and infinitely thin plane sheet of un iform conductance. its effective conductance from wire to wire being made equal to that of the practical bridge. 142. in the case of an intermediate bridge. in is written he re for the bridge resistance." Vol. the reflection coefficient. for the first form. II.For instance.


The wave R = oo or r= x> . p. We cannot make V 2 vanish except by of unit cross-sec tion. show that. A Series of Spherical Waves. Papers. however short a distance.. We must consider the change of ty pe that occurs. however. goes right through. with a certain value of r.ELECTROMAGNETIC WAVES. the sudden stoppage of a charge moving in a straight line at the speed of light. and virtually lost from the wire. is a complete annihilation of the reflected wave' along the The reflected wave runs along the plane instead. 53 to 61. . 1 have described several cases in which the simplest kind of spherical electromagnetic wave. published by me in 1888 (" Elec." II. In the first volume of this work. For instance. Now. 393. on the other side of the The formula does not fully apply at a real terminal.e. accompanied by a p lane Vol. it is vital to the success of formula (1) that the bridge should be really intermediate i. 403). whic h is the resistance of unit area of the bridge In the second form we consider a tube of energy flux sheet. mere ether on the other side of the plane bridge wave cannot go on entirely as a plane wave when it has lost its linear guide. b ridge. Reflection at the Free Eiids of a Wire. and R becoming r. L becoming /x. the main circuit must be continued. 367 form occurs in the true plane wave theory. sible. because the transmitted there is wire itself. occurs in an instructive manner. It does not look probable that a cancellation of the reflected w ave is even posI shall. This is equivalent to no bridge. there is when resistance..

In fact. concerned when reflection occurs at the free end of a wire. it is easy to see that there must be something of the kind.electromagnetic wave. and is then suddenly stopped and sent back when it comes to a free end. it may be regarded as a cone of infinite ly small angle. An easy modification enables us to include the effect produced by a transverse terminal resisting pla ne plate. . and is therefore included in the investigations just referred to. As regards the wire itself. because the electrification is transferred along a wire at the speed of light. generates an expanding spherical wave joined on to the It is with this sort of phenomena that we are plane wave.

The represented by the two spherical waves first is nothing more than th e expanded original wave (a porThe election only. If the impulse acts from left to right. densities of D and B. of things at time t later is a positive charge at B and an equal negative charge at 0. placement put magnetic induction following the lines of latitude. .368 F. Then. is the relation between the intens ities of E and H or the . In the sheet of dissimplest way possible in the sheet. VII. Let there be a fine straight wire of no resistance.LECTROMAGNETIC THEORY. the poles o f a spherical surface of radius = AC. let after the a be the free end wave reaches the of the wire. possible way The arrow-heads show this symbolically. The displacement is joined from B to C in the r= AB is By symmetry. or D = rj'B. Now. the displacement goes along the secondary small spherical surface from E to D. no other than by following the lines of longitude evenly. at once reversed its has gone back to E. then the state say A. To generate a spherical wave upon it. and then diverges into The magnetic induction must be put in as before. That is. the state of things is and DE. on arriving at the end a. CH. The displacem ent in the E by the original spherical wave is joined on to the charge at wave.vH. and the comple te wave is represented E =//. with its core removed. a little while free end. it is merely necessary to produce a voltaic (or electromotive) impulse in the wire at any point. in the diagram).

paper at the top in the big wave. FDG motion. the superposition in the neighbourhood of D of the original and secondary waves has the result of practi- . direction right. It is up through the care to have its. but down The flux of energy VEH settles the direction of H. but not great depth. taking trification.the big one. we generate one of finite If. instead of an impu lsive wave. It in the small one.

Too many diagrams would be needed to show the various cases. Two more two at the be seen that any schoolboy. there seems no reason It will and so on. with a pair of compasses. Keeping to th e unaltered diagram. (In fact. waves are generated by the r eflection of the last right end of the wire . but there is no difficulty in the general idea. then a third spherical is generated directly the negative charge (at C in the A fourth spherical wave w ill case illustrated) reaches it. 369 The displacement cally cancelling the sharp corner at D. does not go as far as D . If the wire is infinitely long on the left side. be generated when the positive charge (at E in the second wave first case illustrated) reaches the left end of the wire. instruction in electromagnetic waves should not Locome " an elementary subject i n the " Board schools. the further off does the bending occur. save insensibly. The deeper the original wave.ELECTROMAGNETIC WAVES. nothing more happens. as they are why absurdly called. can follow up the subseque nt history to any extent.) . But if it is of finite length. the further course of events is got by expanding the spheres centred at A and a. but bends round in curves from one sphere into the other at a little distance from D.

So far there has been nothing that is included in the previous work referred to above. a little while after the first wave (the big one) reached the ends. The arrow-heads show the course of the displacement at the in question. howBB . for instance. 394. shows the state of things due to an impulsive voltag e at A. in the middle of the wire ba. It is moment unnecessary to multiply diagrams of this sort. Reflection at the End of a Circuit terminated by a Plane not virtually It is.This diagram. Resisting Sheet.

and let be struck flush by an incident plane sheet wave. CH. at the charge has gone back on the + . way from angles to is the wire terminating at a. it uniformly resisting plane sheet at the end of the wire. 3^wo parallel The one on the right is the planes are also represented. ba which is wave) the . at right the resisting plane sheet. VII. transmitted plane wave sheet (or part of very large spherical In the diagram.370 ELECTROMAGNETIC THEORY. These two sheets of displacesheet. at a point. moment when wire to the point marked ments are joined together by a spherical and the arrowt SL SH . necessary for preliminary purposes to make the following Put a relating to a terminal resistance easier to understand. ever. that on the of time left is the reflected plane wave sheet. that point should be a long the end of the wire. as before. If the source was.

the ratio Vg/Vj or Eg/Ej of the electric force 3 transE mitted to that incident.heads show the course of the displacement. is. It only remains to specify the intensities in different parts. It is inward tc the wire in the left plane sheet. That is. there are really two hemispherical waves. the other similarly to t he left. one at speed v to the right. and plane nega tively on the left side. that If s is the transmission coefficient at the resisting plane. The plane sheets separate. The resisting is positively electrified on the right side. then we have (2) if r is. as before. . then along the lines of longitude of the sphere to the right plane sheet. whilst the connecting spherical sheet expands so as to keep up with the m. the resistance of unit area of the plane.

When 2r = /xy. is the charge in the initial incident wave. E is persistence of the total induction. right side of the resisting plane (in t he complete ring). then. If g reflected negat ively. If. save insensibly. an d there is On the other hand. Observe that H is reflected positively. sE and sH are the values in the transmitted sheet on the . If the initial wave is not impulsive. not usually continuity on the left side. there is complete annihilation of the reflected charge. 371 being The size of is s is anything between and 1. but is of finite depth. when the a perfect conductor. the charge reflected is the sum spherical and plane waves of the displacements leaving the wire at the point + . and there is loss of displacement. that is. and (s sheet. is . Similarly on the right side. for the The displacement in the right plane wave is continuous with So sq is the total c harge on the that in the spherical wave. and sq is the total on the left side. I t may be positive or negative.ELECTROMAGNETIC WAVES. E and H and magnetic force in an incident very thin electromagnetic sheet. on account of the charge on the wire.1) E and (1 . and 1 when it is of zero conplane are the intensities of el ectric ductance.s) H are the values in the reflected right. then the displacem ent in the plane wave on the left will turn round into the spherical wave withou t going right up to the But there wire.

Instead oir = ^v (or R = Lv). as well as on the right. Then s = J. with a sheet of f inite depth. then. That is. the complicated its so far as the wire is concerned. which is the condition of complete absorption by a terminal resistance in the long wave theory applied to a condensed resistance. the energy is wasted in the terminal BBS . the displacement runs away from the . for the sphere expands. and wave moves away wave is lost altogether. the two plane waves wire altogether on the left side.are alike (except as regards the directions of the displacement and motion). when the terminal resistance is In the former case all spread over the complete plane wave. and there is perfect continuity at the point + Practically. we find that the condition of no refle ction along the wire is r = J/w. practical junction with the left plane from the wire.

wave is more complicated. through the plate. Practically. is fully A similar. ELECTROMAGNETIC THEORY. is Just at the beginning. pair of wires terminating upon the transverse terminal plane. the process of development of the resisting ing to its But with a reflected is true. from the above considerations.372 resistance. VII. then. it the process next diagram. and In the present case. It remains (in our this occurs in the act of transmission somewhat ideal state of things) in the spherical and two plane waves. as shown in the spherical wave of the above kind is gen erated at each wire terminal. but duplicated. assuming that there is . these being the extreme values. only one half is thus wasted. The other half is wasted in another way. CH. The course of the displacement shown by the arrow-heads. accordarrangement. we should expect the value of the terminal resistance which annulled the reflection most completely to lie bet ween Lv and %Lv.

t t positive reflected charge on the lower and a negative on the upper wire. I am not prepared to say. lead to minor disturbances. becaus e any disturbance from one wire comes into collision with the other. It is the same in all mechanics when we go into detai l. of course. There is nothing peculiar to electromagnetics in c omplications of this kind. A fresh kind of disturbance then begins by reflection from the wires themselves. waves. an infinite series of resultant of this complicated process of terminal can be a fully effective and c omplete annihilation of the reflection along the wires at a distance from the en ds.a t "WT Wire. originating on the wires. reflection Whether the . But thi s state of things only lasts until the spherical waves reach the opposite wires. not so important as the spherical and plane These disturbances. It is.

terminal arranement is a condenser and induction >nd. t hen the regular procedure for long waves may be applied. let the particle it is then c entred upon be dislodged. Similarly. Wire. Say. and can scarcely be at tacked at all in detail. the free end. no auxiliary spher ical wave will be generated. and carried forward at the end of a wire without alteration appears to For instance. Cond. Then the resistance operator of the arrangement . for Terminal Reflection. a plane or spherical The only way off the to make wave run clean be to plane sheet of displacement reaches the end. by simply letting its core impin ge on the The wire will then serve as guide. at least to some extent . Gap Wire. 373 Long Wave Formulae 395. Then it will keep up with the plane wave. when a thin carry forw ard the charge. coil. and speed of light. But if the waves can be really treated as long waves. In the Hertz-Lodge experiments the reflection at the transmitting end is a much more difficult matter. the plane wave with carried core can be slipped on to another wire without disturbance.ELECTROMAGNETIC WAVES.

the mutual inductand Bj + Ljp the resistance operator of the primary ance. co il by itself. add on the third part. then it acts as a shunt to the induction coil.is If the gap is short-circuited. we have only the condenser term. or Z = (Sop)" 1 where S This first l is condensers. but the primary i s not closed. the effective permittance of the two term in Z should be generalised to ($Q p)~ + r + lp. . then open use the second part as well. to allow for the resistance and inductance of If the gap is th e part between the condensers and the gap.2 + I^p being the resistance operator of the seco ndary. If. Say its resist- . involving M. and non-conductive. If the gap is conductive to a definite amount. the primary is closed. E. in addition.

the second with turned to Z.1 ). and V 2 the corresponding reflected distu rbance .~Lv the or '-Z .374 ance is ELECTROMAGNETIC THEORY. 1 2 and d + C^C-Y^?.1 is the resistance operator to be added to that of the con denser. CH. reflected wave train in the case of simply periodic waves. (transverse voltage). Liv . . All cases of this kind can be readily developed to show the r. The reflection coefficient for V is R . E first being with a terminal resistance R. We have V + V = V. the resis tance operator of the terminal arrangement. VII. be regarded as a condenser of permittance s and resistance r.Ly V. Vi is an incident. the Z in equation (4). s being what is added in equaThis r may be generalised to (r-1 + s p)~ l if the gap tion (4). whatever it may be for instance. then (r~ 1 + s.

f/L * s ). .e.-!^ ^ + Lt. the reflected disturbance in terms of the incident is got by putting p = ni. 1 (S^ 9V i ZV l _2 _ t/s ^v (rj] and V= = 2VX1 ." V. beginnin g when = 0. In the simply periodic ca se. Then. = (Sop). . showing the relation of V2 to Vj explicitly. (8) V rises from to 2V 1? not instantly.*. but very quickly.. if considering only the big condensers. where n = ZTT x frequency thus .(6) V and C are the actual (resultant) voltage and current. let the incident disturbance be Y! = constant. V. These equations give com plete information. and let the gap be short-circuited. Thus.

without limitation to approximately d istortionless transmission on the line itself. without further investigation. In general V The other "' 9V -1 . As regards the resultant Thus V at the terminal. But in general we must allow for a . i. only a few metr es long. (11} cases can be done in the that the inclusion of the induction coil same way but I fear and shunt at the gap will lead to results of very questionable validity when the short waves usual in this kind of experiment are employed.ELECTROMAGNETIC WAVES. 375 simpler.wave reflection. 396.e. wave of the same type. This occurs when the reflection o f an impulse at a terminal is also impulThen an incident wave of any sort genera tes a reflected sive. that is much explicitly. Reflection of Long Waves in General.. Passing now to long. it is to be remarked that there a re a few simple cases where we can develop the complete solutions out of the initial wave generated.

long. the structure of pj. cessive Knowing V 1? All the sucwaves are therefore developable.change of type. Z is and Z x the resistance operator of the line when infinitely that of the terminal arrange ment. p and p lt at x = Q and I respectively. Then (1) if V and C are the resultant voltage and current at the terminal. . or p enabl es us to find V 2 . let Vj. Thus. So are the reflection coefficients for the voltage. say. This may be done by means of the reflection coefficients. V 2 be corresponding elemen ts in an incident and the reflected wave at the terminal x = I.

CH. VII. mere resistances in the But in (3) so we cannot get clean reflection without change of type unless the terminal res istance operator is a constant multiple of Z. or else equivalent distortionless circuits. All the following waves may be obtained from 0. Let V be voltage impressed at x and V x the first . or else practical imitations. the theory of this chapter. In all cases of this kind. obvious case. The rewaves are then of the same type as the incident if Z also constant and Zj are that is. Z is the constant J-iv. it by changing the origin and argument suitably. it is sufficient to find the formula for the initial wave. which means that Z and Z x may be other lines with suitably chosen constants. flected In the distortionless case.376 ELECTROMAGNETIC THEORY.

To produce the reflected wave. On arrival at x l (4) = l. On its arrival at x = 0. Put = in V2 . multiply by .wave. reflection begins. the third wave /) . put x = in V.. multiply by p v and then introduce the e-^ 8 -*>V attenuator e-^-*'. Thus V2 = P is 1 (5) the second wave. Then V^r^Vo. begins.

which means the same.and then by e~qx .* + PoV<. producing V. Observe that to get the -*-* V . So the total V is 2 V = (1 + /w. . V3 same way V4 is got from from V. (6) 1 .* + 4 )(V 1 + V 2). (2l . we multiply by p p e~^ V 2 and any Vn from V w_2 } In . (7) or.

when.~x) v . except when the line is infinitely long. (4) above is sufficient. however. Also . which is the condensed form of solution for V in terms of V The latter is not th e real voltage at # = 0.

the relation between 877 V and e. e is impressed force producing V . may be developed to Fourier series by means of 282 and after. is V =Z (9) Operational solutions of the condensed kind. as (8). series Even to prove identity with by rigorous methods involves the evaluation of definite integrals of a formidable nature. It is unfortunate the expansion theorem. and its modifications. that the results. When .ELECTROMAGNETIC WAVES. save i n relatively simple cases. should be unmanageable their for numerical the wave calculation. when e . But that i s hardly the right way to work.

families of But results it Whole new may be better than that " " may tumble out of themBut this is selves. the practical way to evaluate them is to find another wa y of solving the same problems.. an apisode.. This may not be rigorous. whether the indi . say. altogether beyond rigorous treatment. not recog nisable as known forms. that our solutions involve certain series and integrals.in mathematical physics we find.. and then equate the results to the former ones. we may do so without ambiguity. If we write . by following Fourier's methods. (10) of although we know that only a limited number terms on the right side may be in existence.

(10). In (8) all the (6) . So. more shortly. formulae or by operational formulae. it begins when t reaches the value x/v. we see that the terms come into existence one after another at the proper ..vidual waves are represented by algebraical The latter are (4). The derived wave V 1 has the same property. The function V is zero before and begin s when t = Q. But If non-existent they d o there is no metaphysics here. whether it be a lgebraical or operational. though they may not be in existence. nothing. modified by the qx It only exists for positive values of t which are operator c not less than x/ v. &c. Or. (5). and so on. Befor e these . Similarly V 2 begin s when t reaches (21 x)/v V 3 begins when t reaches (Zl + x)/v. the former are their algebrisations. considering the complete formula epochs they are zero. waves are added.

t ^ = o-(*2 _ x*l v *)\ (11) then the second wave will involve w 2n P n (z2 ). In all cases the w's must be positive. and the same as regards the quantities under the radical.378 ELECTROMAGNETIC THEORY. if the first wave involves ^ nPn(z 1 1 ). And so x= 2 on to the rest of the waves. we have V = Vi o nly. From t = l/v up to %l/v. where and so on. and 2/ x. between to vt. up to x = I. and its range is only from x = l/v. where (12) = HJ-(2/-)H will involve ^-^-(ai-aOW. Thus. It is only necessary to follow the wave front runni ng to and fro at speed v to see the extent of operation of the waves. but V = V + V 2 beyond. moments indicated by their structure. or at least zero. where . from t = up to t= we have V = Vi onl y. VII. Thus. CH. This consideration makes the formu la (10) explicitly correct all along. w^Wt-x/v). the positivity of the arguments limits the existence of special terms. Terminal Reflection without Loss. Wave . and the third w3nPn (23 ). In the algebrised formulae themselves. 397.

2 + a?.H. n (15) and the second by V 2 = Pl and the third by -^2 A. of (14) which I have given first explained. (16) ^2A and so on. let V be known. 4Z Pn (z3). as before V^-^ZA^P^i). wave is expressed by Then.it Solutions. . Now. introducing the changing x to 2Z a?. and p &c.^P n n (*2). the numerous examples. Let be expanded in the form Vo-c-P'SAJ.

1 + x. 4. .alternately. (17) factors p 1 and #.

bu t operators. then there is change of type. (16). They are similar as regar ds C.1. note that there are four practical case s in which the reflection coefficients are constants (besides the general case of terminal arrangements having resistance operators constant multiples tion of that of the line). pQ or />j complete positive reflection of voltage at a discon. If either there be no leakage. but opposite as regards V. all the derived waves are similar infinite series. i. but in the opposite s ense to the voltage. of a circuit closed . Then. the solution is fully completed when pQ and p l are When not constants. of course. The current is also completely pQ or p l = + 1. There is only one practical case where the waves are expressed by a single term. 379 Thus. that and there is =1 is. require the performance of the pQ and />! operations to fully How to do this wil l come later. which is is sufficient to build It usually thoroughly practical but V impressed be law. and V be steady. to right and left respectively. . tions for a line of finite length.. but constants.e. Equations (15). round and overlap each other and on e another so the resultant V and C are represented by infinite series of waves T he above description just as in cases of terminal reflection. up the complete formulae. At present algebrise them. reflected. Here an impressed force at any part of the circui t generates two waves. by . So.ELECTROMAGNETIC WAVES. we can construct the full soluletting p or pl b e + 1 or nection.] happens that the initial wave is expressed by an infi nite series of the form (15). reflection of voltage at a short-circuit. They are also the same as if they But they trav el round and entered infinitely long cables. namely terminal short-circuit or insulaThere is complete n egative at either or both ends. [In this connection the case upon itself should be noticed. (17) are still true. that is.

. It follows that if there both terminals. then. the complete impressed force V at x = is circuit at result be short due to the = 1 . if there be leakage. as before seen. is the first current wave.Kf /3 or else. to subside according to the leakage e Q where is steady . made V =e . next best.

. equation (33). CH. VII. same circumstances.380 because there the terminals. It follows that the second wave f f*\ is Va = e P (s2 ) + 2u> 2 P 1 (sa ) + 2-^P2 (s ) + . is ELECTROMAGNETIC THEORY. k 2 V I ( . complete positive reflection of the current at On the other hand. the first voltage wave by 381. under the is..

e -K/s } V C If. (22) and so on. . When a steady curve of V (and jf x = vt. is also) is applied to a distortionless circuit the shaped according to e. And wave s is V = +e e~ Q />* P (23) + 2^(33) + 2P V a (*8 ) + . .Ka:/Sw or .21 ) the because there third is negative reflection of the voltage. . .

we let decrease with the time according to the same law. we have p = 1./ 1 ~" j ow """ ov 8 / * . pl = +1. If the line is insulated at x = l. short-circuited is at # = 0. and V = e. extending from x = Q to v t. the resulting curve of V will be a straight line .therefore. No doubt this simplicity is the ultimate reason why the primary wave of whilst still C just treated is represented by one term only.K */s x constant will represent the wave. The complete C wave therefore ~ ov.

* ov*v ' *** i f25tt \ / If insulated at x =Q there e current. and do it . or make some other change before can work. either. take the show the connection with the Fourier series fundamental case of (19). Now. An impressed shift e e We must is negative reflection of the produces no C. to form of solution. Comparison with Fourier Series. and therefore no V away from the or igin. Solution of Definite Integrals. 398.


J. (26) and the corresponding V formula (<r .1 = ? and V = e ~ Kt 's = e/ ff ~ P). therefore the complete C 1 -. (24) C Here we must put result is Po = .ft is -*-"J. 381 by the expansion theorem for normal solutions. The shin ql R+ is -P\. second waves are The first and C.-.ELECTROMAGNETIC WAVES.

but that in the case of C But in the expansion there is an additio nal root p = . left.y = shmq(l-x) f shin g . is steady.p) see that the determinantal equation gives ql = mri in the case of V. Here it is a function of the This is easily put right. Shift the exponential time function to the whilst increasing p to p + v-p at the same time. We theorem the operand time. Then _ pt (26) becomes * *' where the operand turns it . excepting t he zero root.E/L.

. In a similar manner the expansion theorem turns (27) to *) u sn (31) which is equivalent to the sum of (20). This solution (29) is therefore equivalent to (19) above. (by the usual work. ranging from 1 to oo. presenting nothing spec ial) to < 29 > where A = {(ion/I)* ^ = (mV n . &c. (22). (30) and in the summation n is integral.The expansion theorem is now steady. (21).a-2 )*.

CH. then the step dm is ir/l. viz. equating (18) and (32). the integrals. Multiply both sides by cos (nrx/l) and i ntegrate to x from to I. we obtain Po (2l ) = ?! /"cos wia TTj dm. the rest having no chance to make a start. VII.382 ELECTROMAGNETIC THEORY. 7 LTTJ o A (32) where A there is is as in the second form of (30). A (33) Again. So. and (29) _n C = eQ e & t becomes 2 F cos mx sin Xt dm.. summations become definite Put wr/l = m. At the same time only one wave. equate the C's in (19) and (29). We obtain . (18) above. If the line is inBnitely long.

. J^. left S becomes r Use this in (33). (6b) The other formula (31). (35) Then follows in a-t shiner* o=2 TTJ o F sin mv sin At am. --_A m when Z= Wi \ oo.cosmxdx=v o . A (34) On the before side the function integrated ranges in the way When the line is infinitely long (34) explained. reduces to (37) 7T J A .

cos A* ./ and this is equivalent to the first wave. we obtain !(V* . with _ 2 I" sin/ og + gi ^ m\ A TTJ o A< / _ dm_ Z finite.\ / .A sin M\ m. So we obtain =^ But. equating the sum of the V waves to (31). (39) . equation (20). both sides by sin (nvx/l). an d integrating to x from multiplying to I.

taking the mean. third and following waves as being they come on.../-) = V^/'o = Po(~) + 2/rP^) + of . for instance. x is greater than all vt. when x = vt. as in (20).. or 1.1. But it when x is actually = vt because if it is the least bit . we true t ^ . If. WAVES. So we get m This is is + 5gi A v / dm not true less. &c. and then only the first F funclimits in (39) may be tion occurs. the and vt. If.. t is not grea ter than l/v. as no disturbance has reached x. the left side is zero. So.ELECTROMAGNETIC. to introduce the second. for the left side of the left side of (41) should be is P (0). the change argument from x to (2Z x). in (38). (38) when x is the least bit greater than vt. 383 (40) where F(. the rest being zero.

J = * v ["cos TT mvt S -^dm. In any case. This of disit integrals places continuity has to be very carefully remembered. Thus. it behaviour of definite at may very In the operational and wave formulas. value just before and behind the wave front to evaluate the integral. there is no puzzling change of formula involved at the is wave front. it may be noted that in (3 3) the value of the right member is zero when x>vt. (42) Change the sign peculiar of or to obtain another formula. From (42) the formula (36) may be derived.obtain . A J . Similarly. or be a dangero us source of error. This is plain by the manner of construcBut at the wave front itself we must take the mean tion. on the annoying.dm. other hand.

say e i tself. short circuits.e. on the whole. When the impressed force is steady. then the .(43) Perhaps.. under the same circumstances terminally. i. it is as well to keep away from the definite integrals. i f we can get formulae clear of them.

without the y = .Bhin(Z-a?)(RK)* shin Z(RK)* and C=cos-g " _ e where ? is mrjl as before. I have given the V and C wn ~P n funct ions for the first wave. VII. applied to (27) exponential operand. and have shown Now. gives at once theorem above. CH.384 expansion ELECTROMAGNETIC THEORY. 382. in solutions in how to carry on the formulae to any extent in 384 in another . equations (54) (57).

I = oo in (44) makes oos At +BinA( (46A) equivalent to (57) 382. first wave sent along a is circuit . we need not go the first far in elaborating them. Bat. as explained above. with General Way The its of Finding Second and Following Waves due to Terminal Reflection. Considering only wave.way. (45) to the wave formulae. and derive a fresh batch of solved defi nite integrals if required. 399. The succeeding waves are obtainable by change of arguTherefore we may equate the wave solutions to (44) and (45) respectively. as they are all included in the equivalence of (44). ment.

when voltage is independent of the nature of the arrangement at the far end. for a short period of time. It follows. have a known meaning viz. This follows from t he property of applied at beginning propagation of disturbances at finite speed. representing . that series will. again. so as t o suit given terminal arrangements. that if we const ruct.. by the expansion theorem for normal functions generalised. the series whic h represents the complete solution due to impressed force at the origin.

. But after Z and Z : symbols a re operative. these conclusions are perfectly safe reason and sound. in every term. the first 385 The series will contain. but only in the part of the In the remainder. the first wave vt<l. both the cannot change their values without affecting the numerical meaning of the formul a. n ext circuit occupied by the second wave. If. then we have the full interin all electrical of But. I. and we t= 2l/v. pretation of the Fourier series. the Z l symbols are still inoperative. we know the nature of the second and succeeding waves. but in such a manner that only those in Z wave have any effect is. Though far-reaching. obtain the formulae for the series of waves. from the hidden com plex roots involved.ELECTROMAGNETIC WAVES. and has no exceptions in its applicati on. obtain is problems involving waves when done in terms normal functions. because the fundamental sound. the origin. Similar conclusions. upon the numerical meaning in vt. of a more comprehensive nature. in more general cases. one after another. it may be of a pracI shall now explain how to tically unman ageable nature. as in a few ex amples lately given. only. for the electrical properties of t he terminal symbols standing arrangements. that But in the region from x = up to as soon as vt reaches the value provided then the symbols in Z x become operative.

Although the British Association blundered grateful to that b ody for sadly about the electrical units. GK even though ance. about which function (oscillating) are in preI know nothing.and illustrate by relatively easy examples.-. and are comprehensive . 336 and later. the wave formula e are harder to obtain. V 2 = Pl r-*"-"V ol (46) cc . They are readily obtainable. it be not. I am :ables of the I n(x) functions. There is this to be said for the sol utions in the Fourier series. Y. If tables of the K tt function (continuous) and the paration. Much trouble wil l be saved in the future. Against. excessively difficult in the interpretation. they may be On the other hand. but when got are easily calculable. a question of practical importThat the calculation of the second and succeeding waves is The first and second waves being feasible may be seen thus. like 4-. I venture to strongly recommend that they be standardised as done in this work.

then we may in th e Pl factor. every one from its proper impressed = 2Z + o?. ( 47 ) wave.z-y. (21 -x) and write V 2 = r^ViV Comparing now with the Similarly first .386 the second to ELECTROMAGNETIC THEORY. wave differs from the first in the change from x Put 2Z . VII. because it shows that all the waves may be obtained like the fi rst. CH. if z the wave sent in by the impressed This is not a bad way of looking voltage pQpiV Q at the matter. So V3 is ' And . we see that V 2 is the wave sent into an unlimited circuit by the impressed voltage p^Qm (48) .

But. voltage. if it be understood to mean the prior determination of /^Vo. pop^Q. this way not convenient in pract ice.so on. V c P* 2 A J7l (^) if makes V. H wave will be V 2 = ( -ft 2 Bj! I! ?(*). we write the voltage pjV Q thus /^o = the second ^2B I>0.. (50) and is so on. &c. and then There is a bette r way of carrythe generalisation to waves. as functions of the time. and their subsequent expansion in I n functions. though easy to follow in principle. =Thus. . ing out the work. = c : ^A^P^) (49) Therefore.

by us ing only This can now be followed up. by (105). instead of d/d x and d/dt. (53) the modification in/^a) required corresponding . differentiators a We Knowing that V 386. 1 (51) where. 1 ^=/ (a)P (). we shall now have V 2 ^=/ 2 (a)P where /2 (a) is ( s ).found that a great simplification arose by the use of the and /?. and practically one of them. with the operand P () throughout.

So. PI shift (54) being a l function of to p.{-/. To find it. the other side of p (allowing us to ~^)> an^ then put Let the result be ov Then (55) /2(a)= Similarly 3 (r 1 /- . turn p to p forward p 1 ia-(a + or ) for p.ELECTROMAGNETIC WAVES.-/>. to the passage from the note that first of (46) 387 to (47).(*)}. ?.()?.

The successive waves of current may be done in the same 886 for the first wave. ) o-(a being the function of a obtained from p by turning + a" 1 p. to o-j l obtain the following waves : Say.OTO as fac tors.1 (a). V e<"=/ (a)P (z). Thus. and so on.W is the fully developed formula. as w 2 and z.'P. Then introduce way. instead of + ar and + <r above. by (53). . See (106). It is only just at the last that we need think about the proper arguments. as then. 3 (56) (57) where O-Q /s ( a ) = ovr/^a).X-PoW = 2 B. w3 and ? 3 in the third. and . p to The matter regards V2 .2 in the second wave. is now reduced to plain Expand /2 () in powers of a algebra. . (53) V/ = 2 B.

= [V/]y.).Pi = Zrr = -L" I \p + p ^ / / ?n .()?(. . the Critical Full Solutions with Resistances. Second Wave with any Resistance.9l Lt-C/ -*. . out. . Zj. sayR x atz = L Then 400. The next thing is to show how the above plan works The simplest case is that of a terminal mere resistance. 2TZ + cc2 (61) . Application to Terminal Resistances. (59) (60) and so on.

Make the above-described changes in p (or use the middle 386). c .388 ELECTROMAGNETIC THEORY. CH. VII. __ We get 1 ^ _Lf(l+q)(l-a)C = --B 1 _ a Or (TI II f c = ^v ! * /Q\ 1 + Ca LV + R! The quantity R! = LV. formula in (104).

. vanishing when then expect a great simplification. At present it will destroy a lot of terms. So when the first wave is (64) the second wave is ). if (65) R = "Lv = -a is the similar terminal condition at x . This terminal condition destroys reflection in the distortionless case.varies We may from -1 to. we have oalso so that the third wave is . We have o= a. = 0.+1.

'/ Simi larly as regards the C waves. when V = e e. (69) (70) (71) C and so on. the simple case so often use d before. where eQ is constant.(66) The complete result is therefore (67) arising from any voltage impressed at x = producing (68) V. wh^n there is the critical resistance "Lv at both ends . we get . with terminal resistances of the critical amount.-. Ks Thus.-KSAJWrt).

the differentiator But a and 1 if derives any wave from the preceding one there be a short circuit at x = 0. and since there is no reflection at all of a pure plane wave when B a = Lr in the distortionless case. so th at .ELECTROMAGNETIC WAVES. there .<T = O 1 ./>. . all waves after the first are zero at first This is because the front of the wave travels without distortion. and +a therefore . the operators of derivation are alternately and we obtain Notice in the above that their fronts. of the line. 389 so The reflection coefficients are now .^ and .pQ = 1.

the second is . Bj has any value.). then c 2 1 + ca 1 + ca <r 1} )a(l-ctc + a2c2 -. wave front now. (73) This is the expression for in rising powers of So. when iV'LKV-Pofo).. Any departure of B : from must produce a reflected wave which is finite at When This will also appear in the formulae to follow.. is the first wave of current. a.is no reflection of the value L?its front. as in (69).

current is + at its front. say this being a sufficient expression in general. . and the reflected if wave of current is negative.B T >Lv.p3 For distinctness the arguments are omitted. cient to destroy the initial reflection. When = c 1 . . the terminal resistance being insuffiWhen c is .).2AXP. When c is + then Lr>B 1 and the second wave of sidered.. the first wave is of a complicated type. ..(). Similarly. . then 1 +ca !^. (74) . Use w 2 and zz we reduce to the cases of terminal earth or When c = we fall back on the case already coninsulation.

..ca+2 + CV*+3 is . CH. what we must do unless we can bring A n a n to a simpler form with advantage.)}P ( ). VII. one after another. Derivation of First Inversion of Operations. in fact. 2 (77) the meaning of which known. and apply it to every term in the summatio n by itself. Wave from the Second. operate on the second wave by-0-f 1 the 401. This in. which may be done 2 by specialising it. this question As an illustration. and ask Given the second wave. Here we may use the development (73). . : . we may reverse operations. The c2 ) part belonging to An is A B {ra + (1 (a+l .390 ELECTROMAGNETIC THEORY. wh at was the first one ? To answer this.

The C impressed at . then the first is such that c a c (79.2 ). (78) of the fundamental simplest type. c \ c c which makes (80) in is which w and % must l be used.reciprocal operator. if ^LrC or the second wave is 2 ^=P (. Thus.

this being an episode of curiosity.x= now obtainable. as exhibited in (76).e. It is -* i) (l. Now 1. c/\ That (81) c this is the proper answer may be verified by taking (80) as a given primary wave and calculating the second by the The result is the second operator <rlt i. But it does not follow that the first wave is a It might be that the given second physically possible one.. wave (78). < . wave is such that no finite first wave could produce it. but I think a (81) looks suspiciously divergent when c is further examination (which.

wave. We know already when R x == Li? the disturba nce must be zero at the front the second wave when the first wave is finite.ELECTROMAGNETIC WAVES. Returning to the main question. and determine the possibility byexamining the result. that of But now the reason is also evident. posed second ware the case. when the possibility or imposs ibility is not evident beforehand. the second wa ve being exhibited generally by (76). 391 need not be entered upon) will show that the primary wave There is. Save in this I see no reason for failure. the first wave must be very intense. take the Then case n = 0. to . (82) =A with s2 {c + (1 c*) ( a_ and w 2 in the second wave. But our profirst is not zero but finite at the wave front. Naturally. So wave must needs be infinitely inten se. when c differs from 1. one obtained is usually quite a fair one. (77). little Derivation of Third and Later Waves from the Second. In a similar manner we can work back from any wave any preceding wave. When c = we obtain an infinite first case of evident failure. 402. when . however.

B Br the type (78) or (82). But taking matters of the type (83). with literally as in (82). for is of . is C 3 WB 23 . and c . the terminal arrangement at the origin Find the third wave. with the change mention ed. Here <T O and c only differ from o^ and c in the substitution of The third wave comes from the second. in the same manner precisely as the second came from the Thus if the second wave first. therefore. . is a mere resistWe shall have .ance. the third .

O-Q = co + 1 ^oX 1 . t o find the third. Here C CT O is like alt and 2 2 is given by (85) . namely.) .-. then.co a + C o -. we require to expand <T CT QI in powers of a. that these equations represent the first and second waves.(83).

.c 2 )(l -c *)'< + . The result is. A [cc + a(c + c )(l cc ) + a3{cc 2(l c 2) + c oC2(l ..(c + C )(l . ( H.c ) . VI 1. may be applied to (83) directly. as far asP 3 .392 This ELECTROMAGNETIC THEOKY.

. just 2 as. grow complicated. = l. and simple in principle..]P n (86) Remember This.ois The work of will give the fifth wave so entirely mechanical II. being Algebra. and say o.1 "P (^). on the formation . the fourth multiplying the right plied wave may be obtained from (86) by member by -ov. It will be seen that the full expressions for the successive waves get mere and more complicated. and m known reflections at .. that it is unnecessary to elaborate the developments unless they are actually wanted. and the result multiand so on t o any extent. using u. with the proper w and z. . represents the third wave. in fact. by . the algebraical expansions of <r lt a-^.3 and %.'"o. Similarly. after n reflections at x o. or thereabouts. that aM P means (w n l\n}^ n (z) throughout. o-Qoy &c. Chapter products by multiplication. So it is best to use the condensed forms.

+ . Summarised Complete Solutions. out of V *.(1 +<rx + oooi + (T^ + oW oV . in the way already p After the development of the a. From the above it will be seen that the following is a convenient way of wr iting or describing the complete solution for V : Vc^ = (1+01 + oi + + . and the summation in the second line expresses its equivalent to be obtained. /> .proexplained. ducts fo r any wave the proper z's and w's are to be inserted. . along with the .x = Q.)2 -l (87) when "V^ is the first wave. The last equation does not strictly represent the operational solution in wave form. understanding by in o-j certain functions of a to be expressed power series. . 403.) V* A n a' P (2).

But their effect is allowed ./^ coefficients. That contains exponential terms.

derives any C wave from the corresponding V wave and conv ersely with the inverse operator. have the particular forms /> . or conversely. 893 o-'s.ELECTROMAGNETIC WAVES. obtain <r out of p. for. The C waves can be obtained fro m the V waves.and o^ after them. So same operator. 1 if Si is the permittance. Reflection 404. its When the reflector is resistance operator Therefore we must put in the previous results. by a Condenser. belonging to terminal resistances. in the final Similarly. but has general application One or two exampl es of other to arty terminal arrangements. But it expresses is not ne cessary to go through the whole work twice over. correct choice of along with the change from the p's to the w and 2 for any wave. as the case may be. the (89) This is for the first wave. (Si/?)" is . For we know that LwC^-Ll^V^. 386. The just described summary does not apply mere ly when and plt or o. thus 1 Putting . P (z) (88) the complete C in terms of the first wave. arrangements will now be examined. taken positively or negativel y. 1 +a by (101) and (104). a non-conductive condenser.

makes -cr.being the reflection only needs developmenc in l . (92) where is the time period SiLr. e/^. and arranging in powers of a. we +a 2o Clearing of fractions._ x <T " <T _.<r(a + a" ) . coefficient for the voltage This a.p 1 for p.

if the first wave constant. OH. as far voltage. VII. The easiest 3 as a . the secood is /5 V ++ 8 _8P l ** 4\ (zl ) t e represented by is V ^ = eP being wave following (93) on to P 3 if wanted. and the for illustration. . powers of a to show its effect upon a given incident result.2 with the P of course. way is by long division. functions.394 rising ELECTROMAGNETIC THEORY. and further still to any extent by carrying on the long division. is which is sufficient a-e _4 So. Use z.

We may correctly readily overify It is that the last solution reduces when = 0. impressed on the condenser by the the condenser itself first wave is So at y_ M^ 1-ZJZ y Vl _ 1 (0P )TTc^p . then the case of reflection at a terminal condenser in a distortionless circuit. The voltage ee~P t .

we turn p~* to t /\n. collecting terms.l/v.+ 2 __ 2_ 6(p. n But the initial moment Integrating this.P) \ '"J P(p-pf The work want the is developed in this particular way because we only result as far as p~ 3 . and so far as that all is shown. is l/v after th e incident wave left the beginning of the line. and we obtain the voltage of the reflected wave at the condenser. Or. instead of t write t . . So.

ELECTROMAGNETIC WAVES. just as resistance Lr.is . save that it is more general. Put x = l in w. since it gives the voliage all along the reflected wave . Reflection 405. as well as its terminal n. When the reflector a coil of inductance L : and no resistance. if it were a by an Inductance is Coil.2 to produce exact agreement. and t hat ic n contains the factor o-' The result is that all terms disappear save tho se in which the 1 . remember that cr = reduces all P functions to unity. divided by o-. 1 895 To compare with (94). ?r 2 by o-2 and so on. its resistance operator is L^. value. What is left is identical with (97). and therefore . The complete solution in the reduced case may be readily obtained by remembering tha t the distortionless circuit behaves towards disturbances coming from the condenser.

clearing of fractions and re -arranging.-. This somewhat resembles . is is (1001 ' the time period Lj/Lr. < afl+a2 i . (99) and. Turn p to p . this becomes --(*^M'*M)-where (92). (. p 1 V -. Lastly. . -2a .the reflection coefficient for the voltage. .p to obtain the reflec tion coefficient for the voltage multiplied is by for /*. put It is p in terms of a to obtain the expression rr r T l+a Ltc 1 :% T ij.

the first wave being the second is. By division we get So. by (101).but essentially different in detail. ^ = .o f "Vp u _ 1 "V^ Pn +1 _ JL( .

And it is remarkable how simply it terminal earth or insulation. that all disturbances travelled a t finite speed. Initial States.and o^ (see (87). ELECTROMAGNETIC THEORY.396 with % in the development. P A I omit fuller functions instead of z^ whole book would be wanted to carry it It is out in full for various practical terminal arrangements. excepting. at any time later on. however. and. Vll. an integration applied to the ini tial state coupled with some 406. readily by attending to the fundamen tal property of resistance operators. in the complication of the developof products of the powers of o. (88). combined with W. " may be regarded as the classical" way of expressing results. making Here the and (100) goes as far numerator and denominator. by the previous work. 1 managed to carry out this analysi s in the simple case of Another 10 years later. therefore. sufficient to giv e enough to show how the method works out. 403. of course. when I first recognised as a consequence of Maxwell's theory of self-induct ion. and is too lengthy Other combin ations may be made up to print in one line. I shall be glad to be corrected. A comprehensive case is that of a conducting condenser in sequence with a resist ing coil. I have extended the method to foll owing up any terminal arrangement. If so. I set myself the above problems 22 years ago. it is considered that the formula must involve When. to some other state . is now desirable to say something about initial To show that a given initial sta te is transformed when left to itself. Thomson's theory of the electric telegraph. Expression of Eesults by Definite Integrals. fraction corresponding to (92) 5 as a in the The working of the human mind is slow. that they combine like resistances. It .) Perhaps I may have done the work wro ngly. that the Fourier solutions could be broken up into an infinite series of But it was not u ntil 10 years later that distinct solutions. ment goes. CH.

other function. it is clear that the only essential part of the matter is to know how an left initial state confined to a single spot effects behaves when to itself.states. and that the execution of the integration may be impracticable. As regards the due to .

the state at time . special initial states. W&/ (4) where But here z' = <r{ 2 (x y)*/**} .(*>'</. usually be determined through the integral. or of Given. equations (121). I have shown (in 371. 2 have a distribution of Q.ELECTROMAGNETIC WAVES. they better by other ways than 397 may.y) for x in 2 o-( s. I think. V the initial transverse voltage. when we . (122)) that the and C at x due to the cha rge Q initially at the origin are V where point z y. . then. If Q is at the a?/t? ) Now Q = SV</fy. has the usual meaning then put (x . t is V as a function of y.

that is. Moreover. new limits that = Q. and requires special attention because it is different iated. (4) to . do not need any change. whilst d/d(vt) and + d/dx represent unit impulses at x . The result is that if we use the new limits we transform (3). Considering the space and tim e variations separately at the wave front of PO(^). V x-vt It is just at the V x z' y x + vt r and P (z ) drops from It results that when y is between the new limits we 1 to 0 . the finite speed of propagation makes the practical limits be x + vt (3') P is and x rt. the V there has had no effect at x.vt.a discontinuous function. at the distance vt on each side of the point x where V and C are to be found. but at the limits we want extra terms. showing the effect of the discontinuity in P (z'). If y is outside this range. the operators d/d(vt) and d/ dx with unit operands both represent unit impulses at y = x + vt.

but and of course the impulses At x= +vt. V has the same sign and value. from the purely mathematical point of view . x = vt. and (2) that C amount ( Qf) ~^. V 01 is the value of V at the lower. showing that there is a pure electromagnetic wave at the front. and V 02 its value at the upper limit. The impulsive wave goes the other way. where V . (2 ) not only represent continuous distributions of also impulsive V waves the at the limits must not be overlooked. the effect of an initial distribution of C is to be obtained by interchan ging arid C. The Special Initial States J (o-. impulsive to the amount ( Q/S)e~^. At the back. but C is reversed. R and K. V and C L and S. it may be as well to repeat that the fundamental formulae (1). producing As the reasoning about the limits is.398 ELECTROMAGNETIC THEORY. CH.The ratio of the first is to is Lt<. where x = vt. troublesome. Simi larly.r/r) . impulsive to the second and C between the limits vt. VII. (1) indic ates that V is .

initial state. according to a law which tends ultimately to that of variation The biggest inver sely as the square root of the distance. to exhi bit special cases it will be desirable to leave them alone. . or only existent We shall now remove this restriction. represent an It is an oscillating function. hump is in the middle (# = 0).r/v). within certain limi ts. the P (z) or functions have been discontin uous. of amplitude e at the origin. Having included the general integrals for the sake of completeness. and it is of extra length also. but decreas ing in amplitude as we pass away in either direction.obtained results. like the cosine. In all Pn (z) Let V= where e is *Jo(^). (9) constant. and make use of prev iously.and J n (o-. when generating waves. 407. the preceding.

side) we have the if original middle hump If x<vt (on either spread out. V-c-P*P is (*). 899 First. Besides 2 {o-(# /y 1' 21 ) }. 2 As time passes on. (10) what arises from the given initial state under certain circumstances to be considered presently. the function J (a-xlv) is converted to J which is the same as P (z) or I (z). without any 2 oscillation. There are two regions to be considered.ELECTROMAGNETIC WAVES. That is. two things happen. there is the -/>*. attenuation all over according to the time factor that. .

But and the function is oscillatory as As time goes on. that it does 2 = 0. being positive. and that it. and to the left on th e left side. z 22 *' i. As regards the speed of motion of the nodes. widely spread and flattened. Then or "= a + * and =__ initial (12) satisfy the characteristic. all the nodes and humps and at first. What has been omitted is the consideration of the i current. This is a rough general description of what occur s.It represents the ~P Q (z) solution. so indepen 2 . let . is that (10) is known to dently of the sign of expresses the initial nitial and subsequent ym be a root ofJ (//) = 0. The proof x 3 2 > 2 is negative. when state. hollows move out to the right on the right side. leaving behind only the middle hump attenuated and changed in shape .

or. hump in the middle that drives away In a similar manner if the state at time t is V--(*gp. in full.- .(. ys = 8-653. greater ymt the less the speed. v.. It is.The speed is zero.). are 2h The ultimate speed is v. no doubt. the big the smaller ones. The The values of the first three y's = 2-404. ?/2 = 5-520. (13) as regards w and z.

expressing the ( 378). all CH. V and C expansible in J n functions. since 2 2 + BX)? + . The States of V and C resulting from any . it is necessary to consider the state of current as well as that of voltage. VII. momentary pure waves. 408. values of x. the initial state may be (15) V = e3 n (o-x/v)x(-l)\ But for a better understanding. Suppose that ..400 holding good for ELECTROMAGNETIC THEORY. U = B P + (A^ + B^P! + (A Then. For this purpose the functions are convenient U and W. Initial States.

.BO) Ji + (A8 + BO J + (A4 ..(a . 2 then (A 8 B J + ./3") + B . we have .) J.. suppose that W = (initial state of W).B 2 J 3 + . Then all 8) 3 ... . (17) with argument In these put x t 0.. U = B J + ( A . ) J 2 + a ) the A's become known in terms of the B's. + (A 2 + B. and equations (16.B.. (16) W z. W = A^ + (Aa* + BoiiOPi + (AX + B^ 2 )?2 + If .2> . 17) are reduced to U = [B (l + p) + B. (18) = A! J + (A 2 . = all. (19) W Now.

As a further test.and of U and such that We now + (B + B 1 W W have a system U =B J +B 1 J 1 . note that /3W = U. JP . (20) W = [B (o + ft + B^a 9 .]P (21) (. .p) + 4 .j8) + B 2 (a 3 + P) + B 4 (a 4 /3 ) + . . . . .2 2 (a + (?) + B 8 (a .

do this for < U .. I have shown how to expand any power function Bessel functions. in 385 .. .J +C J1 + .. Now. (23) C's. say 1 U =2 CA(^A = C ) . (22) The argument of the J functions is always crx/v.. Comparison with (22) finds the B's in terms of reduces and U to ..+ (B + B 2 )J 2 8 )J 3 -}-.

/3) + (C + (C .C0(a /3 ) + ...C 8 )( c 6 + F) + (C 8 . f?) (24) .2 ) x U = [C(l + /3 + C (a .]P 7 .C 4 2 -C 2 )(a 4 2 )(a + /3 ) + (C 3 .

401 have now the complete solution for the initial state U for we may derive coupled with the initial state by ttU = W. (25) or U = (C + ad + \ + (c o a2 d+ a C 3 3 + . (28). W .)-.. + ft* .a" + a6 -. because it simplifies the work. may also rearrange (24) thus See (27).p> + p> -.. I use the convenient notation an P and /3' P as preWe W : . -/3C 1+ ^Ca -^C3 is .. .. 380.}-^-* /1+a2 . 1 viously.ELECTROMAGNETIC WAVES. We + d(a2 Or ...]P .

.... &c.)j^.. So... is (28) paired with this the solution which expands to It is + now it &c... taking t his form (30) as known (which was not at the beginning).. (27) of which the full expansion + To be ... We can at any rate .... we can at once derive to derive U = #W. (30) that W obvious at a glance from the meaning of u and w vanishes initially...+.

utilise this knowledge DD .

Thus + (C . + Remember states are (C .C 8 1 .) makes U = 0. The corresponding W + . CH. . without the troublesome work.C. ^ or P (31) is all.. W = (C.402 ELECTROMAGNETIC THEORY.da + C 2 a2 .. VII. the similar solutions for U and W so that U = 0.a + aft a.* ..C 3 u' +.

The fractional cases need not be entered into at present.. 409... Here are some illustrative examples : U = 0. W..)j.. (35) (36) In a similar manner we have . (34) 5 -. The (32) initial that = in these workings. U = 0. Some Fundamental Examples. We W = 2C J H H (a*/). (33) have now completed the solution of the problem of the states of V and C resultin g from any initial distributions which admit of expansion in Bessel integral fun ctions. = J (crjr/r).

(37) -^P + ^P+ + + [8 (88) *p i.. and LuCe** = U .W find V and C. 2 W 4 = 0.e . time . . a momentary pure At In the resultwave of the form 3 (<rx/v) gives rise to (38) (39).U =J U=P (</r).. -. Ye** = U + W. 9 (39) (6 In the latter case positive V = ~LvC initially . initial state former case the .

. ing in (35) (36).was a pure negative wave.

. we know that 403 ^=P by (43).to becomes . becomes ta.. '-^'Ps o-i. (41) with argument zi. Now . w H. zt (40) 381.-.. 2 crfx /^ (i *)*. or then we get 35/1".. Again. t + (u + if)?! + + .uw so . If ztz0 becomes wi and we change o..ELECTROMAGNETIC WAVES.

interchange < and U = sin (crx/v) = )?! -P. C of the expansions of sin (<rx/v) and cos (o-jc/i-) of although. (42) (43) with argument z as usual.. W forms a system of subsiding according to e~Pt without any shifting of the nodes.'v = u. Of this an easier proof was given before. + e. <rx. Here V. 379 the is an interesting way of exhibiting the indepen dence present . . of course. We see that U. t occ urs in every term. The General Solution for any Initial State...w. . the time.

DD2 (3) . Another way of getting solutions of the characteristic equation of V or C should not be passed over. 410. (2) s is where ir-i A and B are time functions.(^(VeP*). and _ 0-2^. 2 (1) by 5 ( )> ( 6 )> 378 we formerly used the form of solution > B. having interestThe characing conne ctions with the preceding methods.and some Simple Examples. teristic of VeP* being v 2 A2 (Ve^) = (p . But write (1) in the form (p2 jp2(V the operator p).

and (5). V (4) where A and B are now x functions.A' + o-'f. moment t = 0.A+^^B. VII. see that another CH. The form the (4) is suitable for the derivation of the results due to given initial states. and we form of solution is Ve^ = coshrt. given that V = V and C = C at data serve to determine A and B.401 ELECTROMAGNETIC THEORY. therefore V and C. with the assistance of . 2 2 and (5) i/=(t. These Thus. (6).

378. D V (6) (7) is derivable from the first and C. provided of x. have here the means of constructing any number of are given continuous functions V can be carried out by ordinary if For example. so that the operations rules. V / V. L and S. functions Also notice that the functions of that is. . The results are + TBhinV. is mentioned.-^tc. by We and C solutions. V and C R and K. whilst the V due to C is derivable by odd differentiations. 2 So the V they are functions of A due to V is derivable by even differentiation s with respect to x. immediately derivable by the interchanges vt are even (5). . physical meaning . If by interchanging one solution has been worked out. another having a different Notice that the second of these V .

but without change of type. and be positive or negative. is negative. The V = e-^A* = A*e. is The corresponding current 2IV The current ais the same everywhere.K Here '/ s . result is Then A 2# = 0. whether .taking v V = A#. (8) V subsides by leakage. so it is the same as = o-.

Similarly. we -*'*.R' /L .Kf/3> V = A* e K ^ + ^/e. When <r = 0. a positive wave. we combine the two initial states. we reduce to . obtain C--B* It is If v = -Af 21So-\ . and .ELECTROMAGNETIC WAVES.Kf/s Y / (io) now V that is negative for any value of a-.B^2o-\ is / ) (11) the resulting V. a nd make B = A/Lr. then V = Li-C initially. if 405 C = Bar.

(12) V = Li-C = A(a? -)"'* representing a positive wave. to a V = A#2 A^3 . which may be more con(avenient. provided . then (15) (16) are the simply periodic solutions.Ac . &c. . (7) represent . potence of A is /<. as the general solutions.. if A (raV <r )-. If V and C are simply periodic functions of x./"shin ^ h#*. Passing on complicated. the solutions explicitly. the potence of A is-m2 and (6). More important than the last case i s the simply periodic. where (14) v means the positive constant (W+f)^ by (5). Similarly. so the results are (18) = . provided v has the constant value -2 2= mV)*. may be They are much more complete power series. Or. say of sin mx 2 and cosw#. the results due to developed. Here the the simplest is V = Ae**.

. This we had before..2 defines A. They may also be regarded .X2 = 2 A 2 + o. or m = cr/v. t379. There is no change of type as the time advances.A curious case is A = 0.

A more practical sort would be V = Ae. that the extend continuously over the wh ole range of x. V 2 infinities it does of initial state both sides. Now since (7). we should obtain practical solutions showing the genesis and progression of wave s. &c. algebrise (7). (8) for disconti nuous V or C tinuous that . initial states if Moreover. 411.2 vanishing at infinity on and by an expedient admitting of concentration at But the working o ut of this case is rather elaborate. .7u. And although the simply periodic solution has no one side. Short Cut to Fourier's Theorem. the origin. CH. It is by the consideration of states which are initially discon. # a?. VII. is . Conversion to Definite Integrals. represented by x. All the above solutions labour under the defect that they are not practical in this respect. it increases infinitely as we Even V = e^ becomes i nfinite on pass away from the origin. not vanish at infinity..406 ELECTROMAGNETIC THEORY. . So we have to find how to There are two ways. (8) are the general operational solutions they are true for any V and C discontinuous as well as continuous.

I 1 when x So there is is between a ..the general solutions to Fourier integrals. Here first do the integrals.~ _N 7T \ (17) fc I 6 being + I. It is readily proved in wo rks on plane trigonometry that . The third way was the ind irect proto an impressed farce. / . second way was a similar process applied this problem..). My cess of going via Fourier integrals. \ . 1 STTX = 4 / TTX + Jsm QTTX +ism-_ + -(sin . and then evaluate them. The other wa y is to find a transformation of One is to convert the general operational solutions which will allow of their direct The direct wa y was the first way I attacked algebrisation. n . to get it over. corroborating the results got by the previo us ways.

. is converted to 2 + l = _jf^smmx dm.and I and 1 when between is and jump at the origin. This important. Denote the coefficient of x in any sine term by m. Make infinite. so (17) . is 27T/Z. then the step dm . TTj o m (18) being + 1 when x is + and - .

The jump is 2 .1 when x is .

. at the origin. where the total f(y)dy. The by2)> Al i. except at x = y. 407 result is (dividing Next differentiate to x. zero. represents Finally.y) dm. ''(*) =7T. condensed at the origin.e.ELECTROMAGNETIC WAVES. ir = 1 cos mx dm. for physical purposes. 1 (19) TTJ o This an impulsive function of is the effective meaning total 1. So (20) is 7T JO fcos m(x ./ ("VW*JO ("cos m(x -z .

. (21) = oo to + oo by integrating (20) from y The above is a short cut to Fourier's th eorem. J Then y) dm.ni2 and the result is tion. hm vt cr)g V ^L ("cos m(x TT (22) In this form the v operator can act on the discontinuous operand by direct differentiations applied to the cosine funcThe potence of A 2 is . (19) or (20) initial state we want be here. V= P< .y) dm. . (7). . Go back to (6). But it is and let the V <7y at the point x = y.

oc to + oo will give V and C due to any integration from y C initial distributions V So the general solutions (6). .. = ._ )rf by doing the same work for the part of (7) depending on V These being the V and C arising from V dy at the point y = x. .mV)*. . ( 7) . and V. (23) The corresponding C Ll A ginm ( .(p + o-) Xo^ ~ is IsiI /" J Q * L^cos m(x . are equivalently expressed by where v = (<r2 . C are expressed as functions of v.y)dm.

let us f ind the space-integral of V due to V. (7). can evaluate these integrals in any case in which determined. and zero on the right side of the origin.408 ELECTROMAGNETIC THEORY. not help us to evaluate for a discontinuity. The Space Integrals of 412. V and 0. V Q dy and therefore l JO TT J n . A where A = (m 2 v 2 . thx &c.//// at the We have origin. When this is done we can easily derive the V and C due to elements of V and C anywhere. say by the operati onal soluFor example. But these special evaluations do We V and C have been . CH. V = constant. The full realisation then depends upon eva luating this integral. and without k nowing the result. which is the fundamental one. we can get some information out of (23). Before solving (27). Thus. (24). Say V is constant on the left.a 2 )*. we know the results due to tions (6). however. o r a?. due to Elements at the Origin. VII.

+ shin ) vt .dm. \ v } (29) This can be evaluated when x = . dm. V v r (28) -=V (18).cos mx (cosh +' shin) rt. is p TT /x / / v EE^f dy x is Jo m ( cosh oo .

So put ??i = We make v = <r. if the initial state . factor shows that half the . a charge initially at one spot subsides in amount Kt s according to e~ / however it may be redistributed.Pt is C dy. cancel. the effective meaning concentrated at the origin of m.Look back When to equation made infinitely great. which we The are not supposed to know. . Exactly similarly. that the 1 is and get . we have r.Ke s / . It is the only place where the infinitely rapid oscillations do not effectively in the t function. charge only goes to the right. Z = !_ _ !T fft = i e . (so) That is.

r -*< -Je-Bf/i-. (31) .

we may give x in (29) and still have the result (30). .ELECTROMAGNETIC WAVES. 409 That the total total momentum 3 charge subsides according to e-^/ and the KL according to e. and = if 2 TTJ Q f siD ""Ycosh + \ V vt. (32) is vt. Since there s no V vt. beyond x = vt.*/ may be proved in an . ))l shinV / dm. any value exceeding therefore f* iix>vt. But . elementary manner through the properties of the distortionless circuit.

r in general. This means that the is value of the right number is under vt.x is less than the value of this integral is a function of . (30). amount JV f/ye~P*. the to wave going from / the in right. When . Except when x of just under We this then exclude the head of the disturbance. namely. Deduct \dx therefore.

due to Elements at the Origin. (42). and say 398. C. for another arrival at these of (32) vt.u + (p/X)si n *(> An* Jo Jo A L -2b cos A + + BK) cos Xt .1actually ^-J. First.See (41). when a is just take the mean. c^-l. we may prove by ordinary i ntegrations that / _ pt sinA* =T _ _ pt cos. . The Time Integral of 413. Next we may find the time-integral of C at any point.-sin At I eft ^ j _T where . results.

2 (m2 JLl-w 2 f "* + VQ' ^ "" rf + EK RK) It . we obtain. CA = *>' 7T 1^ J KS -' LC JO This is Li. and by the companion /" by (24) for the result due to to (23) for the result due to C . (35) a well-known form of integral. . Therefore. c T V .pt 2 (>?rr = K/2S.6 . if initially we have V dy and C dy at the origin.

makes (36) .

formula makes All the charge goes to the right But we get only -JQ returns. and never when there is no leakage. when L/R = K/S. then the last If . the time integral of the current exceeds Q. becau se R/L is always greater than K/S But in the application to electromagnetic wave s practically. is CH. and least (zero) when C . VII. in general. all of is It Q goes to the right at first. showing separately how much due to V and how much to The latter vanishes when there is no leakage. increa ses indefinitely with K at the origin whereas the former is greatest when K = 0. V = LvC so that we start with a positive wave to the right and Q = SV <fy is the initial charge. Whe n K/S exceeds R/L. K = oo. both cases may have . Though again. . and which is more striking.410 ELECTROMAGNETIC THEORY. This is abnormal. Q comes back the resistance that causes this.

by Taylor's . before progress For example.. the (sinA)/A in a series in various ways. employed in the same manner. a short cut whereby the complications due to using a full algebraical expansion of (sinAtf)/A may be There is. we can. avoided. Evaluation of the Fundamental Integral. effect the integration Or other expansions may be But such work is very commight not be satisfactory.to be considered. A is a function of 2 o2 . Theorem. We may expand the function 414. concerned in u We may then. Thus. equation can be made to complete knowledge by the Fo urier integral method. Next. without guidance. may be employed. we have (27). /(<r ) . So. wit h special expansion (166). plicated. 375. to find e /v. however. to evaluate the integral u. if term by term. and c jv and theret fore (sin Az)/A. ~ vt vt meanings of the symbols. . and..

/'(0) + ^V(O) + 2 t (38) But it may easily be proved by differentiating r' e**/v that (39) = i*T- .=/(0) + o.

2 t/o-' 2 -= v J " f \t v *~dt. (40) and \p~ l t are therefore equivalent. to avoid misunderstanding meaning p~ 2 l t. because of the o= reduces X to mv. Thus.ELECTROMAGNETIC WAVES. of Or. and we may employ the changed operator to obtain the expansion 3 of vt /v by integrations to t instead of differentiations to a The same is true of the function Ar 1 sin \t. 411 Therefore -f-. The operators d/da. sinA* = sin met A + ^P smmrt .

w?r 22 dt + (jcr ) ?. .}* [2 foosii*^. The process here indicated.. but the operator which Thus... mv /Wo (43) by shifting the operator outside the sign of integration to w?.. But the po int at present is not to employ the developed expansion. vides an alternative = (l \ + JoV ' + 1 ^^> + a . proalready referred to. 375. develops if out of (smmvt)/mv.. (42) this gives a correct expansion of the function concerned be verified by comparis on with the formula (166).ii> yo 2/o V^ ^ sinm^^ mv Jo + That . may method of finding the expansion of c"*/^ and the connected functions. by (27) and (41). in fact.

Now. So. Therefore cos nix sin mvt J sin m(vt we can t+ 1 + . since + x) + J sin m(vt x) t see. must be from xjv (45) x/v x/v . that the value of the integral in the last equation multiplied by 2/. (44) That the operand is when t<x/v has the result that if we use the operand 1 only. the limits of integration to t instead of from to t.. 2 / (0 or 1)..r is v"1 when vt>x. and zero whe n vt<x. by comparison with (18) above.

ELECTROMAGNETIC THEORY. VII. u= 2 " Try o f cos mx . the third is () ( x/v ^)*-^^V-*W./^V/t 47 ) the fourth l(l>? I' / [ t(P iC A/^V .aW O : (48) and so on. the process being to multiply any term by t and then integrate to obt ain the next term. The first term in the brackets is 1 . . CH.a. the second is -*/*'). We can see at a It is glance now what the complete formula is .

398.S ^ t dm = 1 1 [<r(P aW)*J . Generalisation of 371 for instance. if < x/v. Compare with (33). . 415. And u = 0. (49) provided t>x/v. we can now if we like proceed to obtain the formulae for the waves of V and C due to V rfy and C Qdy. and so need the Integral. as in not be repeated here. They agree fully with the previous results. Both kinds of Bessel Functions. where this result w as arrived at in another way. Having got it out of the Fourier integral.

But if n/v < m.. dA.. then the value of the integral in . The second is its co mpanion.The following generalisation of the result is of importance in some other electromagnetic problems. Using the notation for Bessel functions explained towards the end of the last chapter. provided n/v > m. /KAN (50) f Wo ? The first of these is equivalent to (49). we have i o . concerning the other Bessel function. and may therefore be convenie ntly put here for future reference.

.>V<' Also.. ig>2-. we have this if rj depend upon x 2 (r + A 2 )* (53) . and instead 2 G (.ELECTKOMAGNETIC WAVES.) on the left side of (51) we must write )V].. whether result : (52) m be greater or less than n/v.). is 413 of (50) zero instead of J (.

by (7). [Here p is the time time differentiator as usual. It may be skipped over without any loss. initial The C due to V . a full development of the functions concerned. 416. consideration will show that the Fourier integral plays an unnecessary part in the work. Returning to the investigation in 414. there are it worth attention in view of understanding conthe inner workings of the mechanism of analysis. Now. On the othe r hand. and Modification. is C-V^T^Al. Operational Method. we employed an operational process to avoid the complicati on of directly. given initially V=V . 842]. not seeing how to evaluate the integral simply by commonly-emplo yed orthodox methods. constant on the left side. and the special previous results (56). Thus.This formula includes the preceding. . the current that results. there will be a gain in simplicity by making the whole of the work opera tional. some things about We verted the original operational solution to a Fourier integral because we assume d that we did not see how to algebrise it But then again. and zero on the right side of the origin.

because (54) AV = V which Al here. (56) . is Now use the operational expansion (41). or.l. the same. shin ^ V = e lo*P -*t shinty V& (55) Then we get C^-ei^-^shin^A.

vt to + vt. ~ rtA l VII. t > < the . and is 1 from 276. Now e^ A l is 1 from x= -vt to x= + rtto oo by Taylor's theorem.414 ELECTROMAGNETIC THEORY. . but since the operand is when x vt or Now the and vt. So their difference Therefore where the indicated. r epresents 1 from x = . . 1 means a function of a? existent between the limits limits for the time integration are as usual. oo CH.

result (58) suggests a modification which did not preIf two operators are strict ly sent itself naturally previously.effective limits are x/v and t on the right and . the rest is plain enough. (58) 367. The rest side of the origin. Granting that. 'oM* as before found. making 2 a. It will be K> 5 i. of the work is as 414. observed that the only troublesome part of this investigation is the initial rec ognition of the expansion (55). . It is true on both sides of the origin. identical.x/v and in t on the left side. for the right side.

but 2 shin^A. 1 The original vt. if substitution may also be made when same the operators are not identical. Thus. But the general. a positively existent function of x. one may be substituted the other in worked rightly. if their effects are the under the circumstance s considered. perhaps universally.The however for different in appearance. and the (They are not oper ator turns it to a function of x and f. as above. functions at all in the limited sense of function theorists.l means 1 is between x = +vt and operand that does not matter. slave Physics to it is above mathematics. and the to suit must be trained 2 .

= and # /v is 2 2 C= ? I aj. function of 2 The 1 is and the operator destroys the part between .) Now compare 2 . we may write as before. -(*W D l. therefore. where D means now a positively existent The resulting function is therefore 1 provided 2 So it is the same function of x and t greater than x*/v Instead of (56).work the master's convenience. with the 2 t differentiator. t . (59) 2Li> .

415 The significance of the therefore .ELECTROMAGNETIC WAVES. T hey corresponding may therefore be interchanged. "1 l-I (crO. change will now appear. and the variable does not appear in the operators. (62) . For since and (59) is the same as -1 (60) differentiator in question. making There is now only one C-^'/'VfrWD 2Ly Now we know that ^D "1 1. (61) ^D as in (19).

and is acted upon by the correspondin g differentiator in the first. . Or we may verify (62) a.on the . It is not necessary to interchange the operators as above done but it h as the result of considerably . Then the other operator in (61) turns t? to & . We have.377. because the var iable t is present in the second operator. V is The corresponding treatment more difficult.2/* 2 and brings us to the result (58 ) again. The 417. We could not interchange in the form simplifying the work. (56). by (6) of the V due to initial (68) if initially V=V . spot. V due to initial V .

It is only positively existent. (p + cr)e l . (c -e )x. We TT get V= -^_ for ft . V = Vo " P *(p + a)^P- . (64) where the x stands Or.constant on the + side of the origin. Now use (55). A~ l.

<{ [x + rt] [x vf] }. (65) .

where the functions in the []'s are only positively existent. when x is is V=V and . we know the result by elementary reasoning. The disturbance has not reached x. ' = *~ l sllin . And when x between the operand is zero. and V is zero. We should verify this. Finally.t. Now. the operand is 2>. The operand is 2vt.vt. That is. when x is t han vt. CH. (x being on the negative side). VII. This triple function of greater x and t is fully re presented either in (64) or in (65) under the limitation to positivity mentioned. o \2t <K + oo . when x is less than . It so V there is the original V attenuated by the leakance. -vt and +vt.. the operand is x + vt.K '/S .. .416 ELECTROMAGNETIC THEORY. greater than vt.

although doubtless.This makes V c-<V' <i ) (67) which is the required result. The side away from V the negative side here. by another method. When x is + or in the reg ion of V as the time goes from to t. . it can be This is horrid. We have to find U. is . and the operand is x + vt. . (68) the limits in the integrations being x/v and t. 2 of the resulting series going only as far as_pThe beginning . the operand is %vt from t = to x/v. (69) and need not be pursued further. It is not necessar y to do both sides. . given by . t hat an infinite series of Bessel functions was needed. and is then x + vt till the full t is reached. When x is . is the easier t o manage. by extension and proper arr angement. . But when x lies between rt. and is then x + vt from t = xjv to the full t. We found before. the work is considerably complicated.the until the time reaches o perand is x/v. U = (l + JoV 1 *+ ) (x+vt).

then we calculate For distinctness let V be on the negative V on the positive side. 417 more conveniently side. If the initial V is is on the left corresponding U on the right side obtained by changing the sign of a*. the exhibited. 386. versely. for the side. the C wave. \Ye have already found a concise and convenient way of developing the C wave when the V wave is known.ELEOTBOMAGNETICT WAVES. and therefore V-Vf-fU*). and conIn the present case we ha ve (58) for 382. so as not to b e troubled with the subsiding tors. Now . (70) V wave. V .

an impressed force at the origin. Z (73) See also 381. l-a' as explained in n 382.. /? differentiaand we have . The on the right distinction. This makes (72) . +2 P 2 + . ( ).use the a. whilst negative of V at .x are identical. then C at x and at .. side of the origin If the source is investigations are practically the same but on the left side there is a .

K *^ to the negative of the at x. initial V We and C now come to the very last way I shall give of 418. Final Investigation of the V and C due to . V V at -x. algebrising the operational and had especial value in being the some his torical interest. It has solutions for initial states. 1 . But when the source is the distribution V above ass umed.-x is the assuming that x is positive. first way in which I was able to obtain full mathematical confirmation of the results relating to the generation of positive and negative tails obtained by a consideration of the distortionless circuit. then we must in (73) to obtain V add on V c.

CH. YIT. if we express v~ l t vt in the form e r* A/(vA).418 ELECTROMAGNETIC THEORY. Then.. explicitness. (7). then we separate the distortion from the translation. by the resulting states of V . say that (6). without any C. which is a transktional operator. For V=V + <r) x initially. ) AV Now vt in the distortionless case c reduces to e^A. this is useful depends Whether . shin vt^ r P^P -V . V=e . and C are (Y 4 /r~ . But since when t here is distortion the property of propagation of elementary disturbances at speed v remains true. a nd enables the algebrisation merely to be immediately effected.

and the r's are fanctions of (vA)" 1 thus. . it is clear that the solutions (74). The **] 7 rt A I 2 1 +M\* + \/2\2/3 + . ~ Since changing the sign of t give s us e vt /v. > Wi /(3 I (76) where s = o-/vA. (75) are expanded as required . . and^? to -yA. then the work is done. /(v A)V The required expansion may be effected by 375. 375. s2 to o2 .upon the form assumed by/(i'A). only with yp there changed to vtk here. equation (166). If can be algebrised.Turn y to result is t. and since .

1 This is to be don e by to rearrange (76) in powers of (vA)The result is inspection of the r functi ons. . the expansion suitable for the t reatment of discontinuities. is In the practical execution it is perhaps as well.as described in /(vA) contains A inversely in s and in the r's. namely . V (78) where the #'s are time functions. or better.

being the same as in up in another way. equation (179).4.8..6.4. or A vt (p + or) /v. _ 419 The next 376.8. 10' 15 16 Similarly 16 2.12* we can develop by using the tiating (78).WAVES.6.).. where they turned three are 8 IT _ 2. either series in 375. by differenWe get (81 ) .8 2^6 2. where the/'s are also time functions. of which the first six are ~ .8* 5o-t5 _ '""leF 4 2 2.ELECTROMAGNETIC.4.4.

)^.4. Jo.cr <r*\ 16\ 8/ 42"T6 the sign of o-..10 V also that of By changing and we get (83) ^-o-) and so.). 2. -j (84) .s y3 + . + ^ 1 / q-A 12/' t. l^ = e-^(l-5/ + sy2 ..6 8. (83).. (75) are therefore brought to the forms + ^-^-^(1-^+^-^ + . Our original equations (74).. (74) by taking half the sum of (81) and becomes expanded.

V . s Suppose. .L .) Vo. . for example. . . are the integrations sV 2 is constant. except the trans. b ut exists &c.)V (85) The only operations to be performed now.. and we want V outside the V EE2 . V only from the point y to +x . .C = pt^(i + sgi + ig + fffi + _ 2 . lations to right and left respectively.

This diagram will illustrate : If it be Vo The dotted line has attenuated much shows the beginning of the V wave before it at its front.V is V a (a -y) /\n. where y negative. V is zero.420 region. Then the second line in (84) is zero. and zero on t he other side. grams will show the approximate shape under different . all these integrals exist Then c^ A turns x to x + vt. without any allowance for . The curve of V is not But later diaa straight line. VII. ELECTROMAGNETIC THEORY. the attenuation of V if there be any. and in the first line. as mistak enly represented. A. = x + vt-y. because V is n W zero. a nd makes (84) from y to oo become . Here yx must be positive. when x>y. That is. CH.

to = y + vt and y-vt. If the initial V is on the left side of the origin. In show how the wave fromoo to x = 0. Similarly. that is. they are easier than the corresponding Bessel formulae.circumstances. the other formula (85) makes 1*0It gives -K^JI+^+^OVC^T?^ \v ( v \v/ la / [o J } < 87 > the current between the limits x fact.iV. when t It is pretty easy to calculate these formulae big. then the proper change in the meaning of y 1 makes V.-"l + otf express . is not too begins.

l -* + /l '+.. And the corresponding C is beyond given by (89) . from x = V is zero. to vt\ (88) which.. V at a on the positive side.

without leaves <rtglt &c. change LvC to St>V. V to C .ELECTROMAGNETIC WAVES. other change. and V to C . oin (89).. 421 the sign To obtain of athe formula for the C due to C change and dne in (88). collect Comparing different . because reversing the sign of the same. explicit and in the /'s. of course also writing To obtain the formula f or the C instead of V and C V .

" Boussinesq. I have gone Another way of attacking over pretty wel l all the ground. is and as space And . more.E ""'*""' 8 (m V . due to lately published Electricity and Magnetism. There are also combi nation formulae.~\'dm o2 (92) )* (93) There are several other forms.!). but these are the principal. (91) 2V r IT I n o cosw. the mathematics of the subject will be found in Webster's " 25 5. I have had enough of it for the present. Naturally. But they do not involve fresh running short. we may some of the results thus pl t* */. I prefer m y own ways. I now bring the development of thes e formulae to a conclusion.: ways of working. and methods b y Poincare and Picard are also In these the reader may revel if he wants referre d to there. because I can understand them better. its tail behind. pure wave casts investigation. when there is initially for example. to show how an initial exi stent both V and C .

2. 427.best of all I like the distortionless circuit. and permit particular people to a dmit the truth of the conclusions. thing essential to a general understanding can be foreseen by its means without any of the complicated mathematics. pp. ..). which however is of course necessary to complete the matter. 119 to 155. and to allow of close calculations being made if desired ("El." Vol. and pp. 475. & c . Pa. 381. because every.

What follo ws is supplementary. the subsequent history is got by moving the positive wave to the right a nd the negative to the left. Undistorted Waves without and with attenuation. in which V = ~LvG. I. in its application to conducting circuits as well as to plane waves in a un iform medium with two conductivities. at the same . the case of no resistance and no leakage. In it V = ~LvC. who may and draw the take curves according desire to go into numerical to the preceding formulae. Vol. Vil. described pretty fully the meaning of the the ory. IV. by the ordinate erected on the base li ne which stands for the circuit.422 ELECTROMAGNETIC THEORY. particularly for the assistance of those readers detail. CH. wave. Having in Chap.. Then let there be two equal waves The one on th e left is the travelling towards one another. 419. it is unnecessary to repeat the information here. First Eepresent V moment. On the right is the negative positive This being the state at a certain wave..

Draw it it. since V = Li-C) as the wave passes any point #. B shows the curve of voltage at a certain b will show what it becomes a little later. are . The variations of voltage (and of current. b FIG. and then and will correctly represent the transit of the wave as modified by the resistance and the leakage. time interval being the distance Aa (or B6) divided by v. y To illustrate. travel at the s ame speed. along the circuit let it 423 is therefore y = c~^x Lv ^ . say a positive wave. Note that this Similar ly a circle will shrink to an ellipse. then . an elementary plane wave slide along at speed v. shrinkage does not count as distortion at all. 1. All the elements of a given arbitrary wave. the moment. and attenuate at the same rate. if A.ELECTROMAGNETIC WAVES.

then later point x.Lt-C and then moving the firs t to the right and the second to the . _. the other with equal negative current making V = . 2 What results is to be two equal parts. found by dividing the initial state into one with positive current. only reduced in size. without any C. left. making V = L t-C . according to /().ayo side of the origin. it makes f(t * -x/v) x e-**/ 1 at any Now side consider the initial state V = 2V constant on the left and zero on the right .identically repeated when it passes y further If the voltage is impressed at the on. left end. at . FIG.

speed r. the upper dotted line shows the . In the figure.

and c-Kt/s the leakage It follows that. Next suppose B/L K/S remaining the same. V on one side of the Origin. but at the same rate as the V and C in the wave. age only.*/ = f. Then or B/2L + K/2S. to be increased a little. K/S. is made P t is . The shaded region shows The t he electromagnetic wave V = L^C spreading both ways.424 initial state. dotted part of the curve of voltage has no magnetic force The voltage is there attenuating by leakassociate d with it. New ~ attenuator. ELECTROMAGNETIC THEORY. and the lower dotted Effects of Resistance and Leakance on an Initial State of Constant 420. greater than the wave attenuator. the later time line what it attenuates to at Ks making e. with t he same initial circump. CH. VII.

at every moment is equal to half V the this is moment in question additional attenuation along the wave to the right. 3. V . at the particular and there is . instead of increasing B/L. The case is quite altered if. in the last case.is stances as in the last case. and -?*=*.Kt / s = f. The thick line shows V the curve of voltage at the moment making e. reduce taking the distortionless case as initial standard.. making V at the front be -V The general characteristic as the wave spreads and V attenua tes all over behind the wave fV . instead of th e greater than at the origin of the wave. the attenuation at the wave front So. For we shall now make p less than K/S. is the greater attenuation at the wave front than at the origin. At the orig in. result is that the attenuation at the forward . the curve of V horizontal straight line of must fall to the ri ght and rise behind. FIG. we it.

and in the forward wave generally. 4. the same attenuation by leakage alone . and in the region behind the wave oc cupied by without any C. V In the diagram. is less than at the origin. Fig.The remarkable wave front.

FIG. but at the FIG. and the At the other extreme we have is . and t he thick line shows the curve of voltage. The constancy of V at the origin enabl es us to regard the . FIG. and the same value wave front V is . Two moments of time after the first moment are considered. is 425 forward shown. B = 0. the remains steadily at V = V value of ~P t is J. One If K = 0. W ith the same initial state there is no attenuation behind the wave region. p = K/2S = .is negative. at the origin. the other R = 0. and t he voltage at the origin Next pass to the is K = 0.er that is. there . 4. and o. wave proceeding voltage to the right as being impressed by a steady V . attenuation by leakage only.V . we have /> = E/2L = a as well.ELECTROMAGNETIC WAVES. 5. In the case illustrated. two extremes partially indicated. 6.

to . . is easily seen .V . CH. t he place of zero voltage moves towards the origin as the wave spreads and attenuates.426 ELECTROMAGNETIC TttEOBY. and . and at the forward wave front JV On the left side of the origin V has become The ultimate limit of this negative in a portion of the wave. the initial 2V has attenuated . corresponding two stages of the voltage curve are shown. when this the forward wave front f V wave front has gon e twice* as far. at the origin we have iV . . Vll. And at the later moment. When 2V has fallen to -V the voltage at the origin is -V and at .

we superimpose the state V= and a. This is Enlarge the time . The difference of the two curves is the cas e of no leakage. a FIG. scale to suit in the second curve. V from . 7. wrong. If chosen infinitesimal the result will show how 421.Division of Charge Initially at the Origin into Two Waves with Positive or Negative Charge between them. 8 arises.x to x= +. From the previous curves may readily be derived those representing how a charge initially confined to a limited region oo For if upon the state V = V from x = s plits and spreads. and take the difference of the new and old curve It may be to represent the V arising from V in the length a. to 0. any l ength. . They are not Not twice. On the other hand . if a be chosen large. 7. shows what we want. Bringing it down to the base line. Fig. one another. we magnification. This lapping. So shift the the resultant is simply V between x = curve of V arising from the f ull + V through the distance a to the right. Bat then the curve requires a point charge splits and spreads. do not get clear representation of the waves. * It shows the two waves running away from quite pure waves. as in Fig. owing to overSo take a of a modera te size. because they are but 2'7 times as far.

Between them is the diffused electrification cast behind by the two heads as they travel . that is. it is the double tail formed by the superpo sition of the tails of the two waves. is The complete area of two heads and the double tail the same as the area of the original rectangle.ELECTROMAGNETIC WAVES. 427 of finite depth. shown dotted. is not HL^*^ . At any rate. it ou ght to be so. and by the look of it.

The Current due to Initial Charge on one side of the Origin. if the distributions V and C are similar and equal. So the V/V curve positive. 9 shows the correspondin g C/C curve when o. also represents the making owhen R is negative. 3. CH. changing their values. in excess. reversed. We may make whereas asimilar interchanges of factor remains the is L and S.is positive. V . VII. or o. 4 s hows the C/C curve And Fig.is C/C curve when K is in excess. then V/V in the first case is the same as C/C in the seco nd.428 ELECTROMAGNETIC THEORY. . The c~P* same when R/L and K/S are transposed. Similarly Fig. 8 shows the C/C curve arising in Fig.negative). with persistence of mementum now instead of electrification. whilst Fig. from a central distribution of C when R is zero (o. where R is in excess.

10. Then is the formula for C. . general type is shown in Fig. but they are similar when opositive or equally negative. Let initially there be 2V on the left side of the origin.422. positive. There remain and of V due to C for consideration the curves of C due to . These are is alike. 10. Not only that. The FIG.

and 2. take.we Say <rt = lj good value to Here p =. require to give p a greater value . K . taking ais zero. in the figure are shown the curves .cr ( 1 + K/S<r) so.1 . unless So -J is a then &~P t must not be greater than e. a-t = l and The the corresponding values of e~P t being Kt/s voltage at the origin is V e~ . of C when J and i.

by taking differences as before. 11. FIG. This the negative part of the intermediate current that represents (in a great measur e) the tail of the positive wave. have a positive tail the negative a negative It appears anomalous. in the one going to the them to is positive left. as will be more plainly seen presently. tail. current is Two stages are shown positive in the wave going The in Fig. and the diffused current between and negative also. 1 1 to the right. The positive wave appe ars . but it is a m ixed up case. .ELECTROMAGNETIC WAVES. whilst the posiis tive part represents that of the negative wave. and negative . we can showing the current due to initial Y through a certain distance. 429 From derive the curves these curves.

. To obtain the formulae showing the tail cast behind = LfC we have merely by an initially pure wave in which V add together the V resulting from V to that r esulting from to LyC to find the V due to both. and similarly as regards the . Negative Tails.The After-effects of an Initially Pure Wave. . It would take up too much of t he rapidly-decreasing space remaining in this volume to enter into detail concer ning this method. The reader who will study the short account given elsewhere an d will take it in. The expansions immediate But the following Fig. namely.xjvt) rough numerical calculation. are convenient for in powers of (1 . be able to construct tail The results are rough. II. 12 has been obtained by an entirely different p rocess. Positive and 423. In this method you go behind the scenes. by making numerical interchanges.. will easily (" Elec. 318). show all the main characteristics in size and shape. p. Pa." Vol. C due to both causes. as well as a variety of others. but curves.

and the maximum is a long way from the In the later curve 3 the head does not count for much. advancing to the right. The remains of the unit of time being arbitrary here. 1 . There are signs of the maximum having been rea ched in at J. its depth as wel l as height are shown at 2 and 3. and the tion . VII. the curve has become mo re symmetrical with respect to the middle point. later. and 3 . and see the inner workings of the formulae when time and space vary. Bo th the voltage and the current curves are obtained from the same series of numbe rs. the middle of the head on the right merely indicate the position of on the other hand. and the curve 1. CH. be too big to get into the diagram without increasing its The vertical lines at dimensions inconveniently. on the right side. 1 2. becaus e they would vertical 0. Fig. the head are shown for the times t = 2 and 3. Supposing there is initially a pure wave of small depth at the place 0. it were. In the later curve 2 nearly all the electrificatail. but no heads are shown previously. at moments t = J. the maximum has moved back a little.430 as ELECTROMAGNETIC THEORY. 12 shows the tail of V when there is n o leakage.

and perhaps a little on the right side as well. on the same scale. is represented the developtail of current. the tail (widely spread) continues to tend to become symmetrical with respect to 0. and at the same time. change of curvature shows itself markedly on the left side. thickest at the tip. J. In the next diagram. when the head disappears. and 1 are left out as before. 12. 13. The heads 2 and 3 are shown the same as This m eans that the curve is that of LvC. not C. ment of the moments of The heads the for t = 0. and decreases towards the head. Fig.is in the head. Note in passing t hat the sloping lines at tail is entirely . At first negative. Later on. in Fig.

Going further. main reason for this is on account of hat can be made to telegraphic. and the After that.ELECTROMAGNETIC WAVES. At the moment f = 3. The instead of E magnetic forces). telephonic and Hertzian or plane electromagnetic waves in general in a conducting consider E medium. 431 finite the tips in Figs. it is positive next the head. the FIG. Remember that the head is greatly reduced from the origin al head at 0. 424. At the m oment = 2. negativity of the tail has nearly ceased at its junction with the head. nearly all equations and results erms of V and C (voltage and current). there is positive part of the tail has a maximum. 12 and 13 are merely consequences of the depth chosen for the head. more natural to and H. and think of what goes on withi n and along have been expressed in t and H (the electric and the direct application t waves along wires. and a long way behind. These sloping lines are vertical when the head i s of infinitesimal thickness. In the preceding. But f . 13. at the moment t = l. further progress has been made towards symmet ry. 1 to 13 described in terms of Electromagnetic Waves in a doubly Conducting Medium.* Figs. it is. a continuation of the progress towards symmetry. which causes overlapping. of course.

. We must have recourse to a scienticulistic exTrie spy who drives the cart informed me that the retardation planation. " making the speed of the current be less cipient action of Bub the tanner confi rmed this. KR oath. Both are scienticulistically impregnated. of the top of the wave with respect to the bottom was indicative of in" the law.* I am unable to explain by my theory why the waves at 2 and 3 should lean backw ards. with an at the top than at the bottom.

It is sufficient a single tube of flux of energy of unit section. OH. Turn V and C to E and H. Also. here merely to restate crete to the specific. how to transla te solutions from the conIt would scarcely be necessary were it not for the singular part played by R whe n we pass from real wire waves to pure electromagnetic waves. The peculiarities connected with the last should be noted in the following. 1 to 13. Fig. the inductance per unit length becomes /A the inductivity. Similarly.l. of (fictitious) of the medium. the electric conductivity. in Fig.432 ELECTROMAGNETIC THEORY. K the leakance pe r unit length becomes k. becomes c the permitt ivity of the medium through which the waves are passing. and S the permittance per u. 2.l. 1 E and H of arising may stand for either . VII. uniform field E on the left side of the plane x = 0. The wave Referring to Figs. and E the resistance per the wires becomes g. where the initial state is a f//fj. The former are the line integrals of the latter. the magnetic conductivity u. Then L. since if they are not understood the subject is rather puzzling sometimes. shows how plane waves of would be transmitted in a medium any type possessing bo th conductivities of such amounts as to make = k/c.

the wave of the conductivity is wholly electric . when the electric conwhen k is in excess. here it is the magnetic conductivity that in excess. Fig. 5 we have the wave of E due to E when the conductivity In Fig. 4 the curve shows the wave of E arising from E due to wholly magnetic. E arising from a similar initial state ductivity is in excess. 3. Or else. the wave of due to du ctivity is wholly electric. the wave of H arising from H when gis in excess. and the wave of arising from the initial state It equally well represents the wave of H. is represented. I n Fig. H . is H H when H H when is it is . 6 shows the wave of E due to E when the conOr else. Or else.E or H. since E =/^'H. is E Passing to Fig.

or turned to a real cas e by considering fictitious E H instead of E. if one case owing to the non-existence of g. it may be instead of H. In these examples. which separate and ft .wholly magnetic. 8 shows how an initial state of E in a plane slab divides into two equal pl ane slabs. Fig.

when & = 0. 433 and leave part of their contents behind is and between them.ELECTROMAGNETIC WAVES. 9 shows the initial state of E and spreads when when . It is the induction that divides now persists. and g total displacement. Fig. and how is finite. attenuate. when g = 0. finite. it Or k else persistence of shows the division of initial is There H .

Or else how a state of H divides H similar H . due to initial uniform E on 10 shows the wave of Fig. the same initial state when the conductivity is wholly electric. 13 shows the tail of H in different stages du e to the same initial state as last when there is magnetic conOr else the tail o f E due to the same when ductivity alone.g is finite. shows the wave of H due to a slab of E at the Or else the wave of E due to a sim ilar slab of H origin. 12 shows the tail of E in different stages of growth = /*vH when the conarising from the initial pure wave E Or else the tail of H d ue to ductivity is wholly magnetic. 11 . there I is electric conductivity alone. Or else the w ave of E due to the left side of the origin. and k <7 k is finite. Fig. Fig. . and = Q. Finally. = 0. Fig. am aware of the deficiencies of the above diagrams. Perhaps some possesses the patient laboriousness sometimes found associated with early manhoo d .

fe-v curves. in Vol.* It should be a labour of love. and perhaps to a great extent the maker of other works of a more sellable . and the most eminent publishers will not keep a book in stock if it do es not pay. of cours e for although if done thoroughly there would be enough to make a book. but not enough for what go further. II.electrical student who may find it worth his while to calculate the waves thoroughly and give tables of res ults. nature. * Webster's "Electricity and Magnetism" contains a I refer to. Gray's work.. may perhaps . Storage room is too valuable. and several curves in every case. even though it be a book that is well recognised to be a valuable wo rk. it would not pay.

It is only turns up incidentally. not the main subject for one thing. this 425. A Formula for wobtained by Harmonic Analysis. GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. I must say a few word s on the subject of generalised differenI have put it off to the last for tiatio n and divergent series.CHAPTER VIII. Therefore 1 come to the last matter to be considered in Omitting some other investigations concerning plane waves. it. how can they be good enough for Electromagnetic Theory it ? I really cannot answer such a poser. if investigations of this matter are not good enough for Another Place. several reasons. I am reminded that volume is growing fat. not easy to get up any enthusiasm after it has been Neverartificially cooled by the wet blankets of rigorists. and Again. I have asked myself. Then again. is .

I hope the prize-winner will have somethi ng substantial to I wish him success. . 1893. I have been informed that I have been the means of stimulating some interest in the subject in certain places. whoever he may be.theless. But this would take about 120 pages. 1894. Dec. say. way out of my difficulty would be to just shove three papers " On Operators in P hysical Mathemy matics. Perhaps not in Engla nd to any extent worth speaking of. that analysis. II. a lesson that may or of the part played is better than the by divergent series in wet blanket." Royal Soc. June 8. down for reading June 21.. 15.. and Part III. 1892 Part The easiest Well. and sets in here . and not read).. Proceedings (Part I. but certainly in Paris it is a fact that a b ig prize has been offered lately on the subject may not be followed in other quarters.

Let A be a differentiator. The case because our operational solution s involve them. is in the equation what ]n (1) subject to (n A-l= . passing over with large extensions. come at once to the n function i n its relation to fractional differentiations and integrations. and no doubt other We want to find values turn up in other problems. n = ^ also occurs in cylindrical problems.P. is (2) the last datum being necessary it.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. say J or J. say with respect to x. the papers will be quite sufficient for the present work. and can be given in a limited space.. to fix the size of the function. but also when it is fractional. especially because a good deal which has been given already in one form or another.. 1 to 14. then we have had to n consider not o nly A~ l when n is integral. We are obliged to consider such cases. Q. already given So. 435 An account of certain parts of is out of the question. = (n-l. or to standardise as is . (1 = 1. Part I.M.

many .We know that there such a function defined by (2) given in all treatises on the Integral Calculus (equivalently) un der the name of Euler's Gamma 1 n= most books only treat of it from and have nothing to say about the region from x to 1. A formula for \n. and con sider it in the light of practical work in function. operational mathematics. With a differen t zero it is also generalised called Gauss's II. Use harmonic analysis. and still less about the application to differentiation. based upon (1) may be readily found. We _ I have * Al = if TTj cos mx dm. though I think to + oo . But put aside from me mory for the present what may be said about the Gamma function in books. function.

J (3) a remarkable formula already used times. the potence FF2 . Now with the simply periodic operand cosmx. So (4) A "1 = A -<n+DAl " = A (n+1) 7T I "cos mx dm.

A = wi. so (4) A~ n l = e I r cos nix dm mn + . VIII. we assume fractional. the final form in (5) involves only the operator of Taylor's theorem. is CH. So A2 -i 2 . 2 mn +! That is. __ _.436 of ELECTROMAGNETIC THEORY. not identity in general. the algebraic imaginary i is equivalent t o A/i. Or. the sign of equality merely meaning equivalence under the circu mstances. that the equivalence of when n becomes is On A to mi persists this understanding. That is.

and denoting the reciprocal of d (n we have A Formula between It ^ = Jo ' mn+l its (8) Thus it is not \n.l n This. i \n \n TTJ o mn + l (7) and so. (8) is . is a formula evaluating A~ l. According to (1). obtained without any reference to the \n function. we have > fco i i . observe. then. but gated for our purposes. putting x = 1.

as before ob tained in another way.P. which must be worr ied to display its .reciprocal. that should be investiquite correct and is intelligible when n lies -1 and 0. The integral then convergent. be readily checked that [ = TT*.M. Although the above way (not 'given in O. n = Q. The interpretation outside the limits named it fails. the result is a definite integral. because equation it is known between .) of getting a formula for g(n) is very direct and short. may when need not be considered. (2) finds n all over Algebraical Construction of g(n). Value of g(n)g( n). But just at the limit giving only half the real value. 426.1 and 0.

at once The curve of g(ri) can be drawn by plotting its known values when n is and it. l 2. 1. 4. %r ^T &c -> a* n 0. f 1. > for every negative integer. By the look of we is see that the curve oscillatory has a maximum between and 1. mately viz. J. approxiintegral. 437 effective meaning.. 3. &c. and on the .GENERALISED DIFFERENTIATION AND DIVERGENT SERIES.

. for n = J.M. whilst continuously positive on the positive side. If we turn left. The range As n of the oscillations is now increases negatively. 17. we shall obtain the similar curve of g(-n). 19. Form the curve . This is confir med by plotting the midway values. 18. The two curves are remarkably connected.P. f. increases.negative side. also known. &c. right to approxithe range O. and continuously positive on the left side. with a maximum between and 1. oscillatory on the right side of the origin. mately indicated.

. . Now (9) is the same 9(n] 9( ~ n) = (1 2 ) (1 Jn*) (1 n*) .. and has the value 1 at the origi n. since the zeros of g(n] are known. (10) therefore. .. follow at once. That is. This defines the function (sin mr)/mr..* g (n)g( for all values of n. &c. It of the product g(n)g( -n). has a zero for ev ery integral value of n. positive and negative. (9) The values as 7i of #(J).It is oscillatory on both sides.. -) = 8 .

.and that X = l.Jn) where X is -i some function which does not introduce fresh zeros. * and There might be an extra factor. &c.r.= X(l+n)(l + i. This is an oscillatory curve crossing the ?iaxis at the points .i)(l+Jn) .n).1.. to determine the factor to be ... is not.2. th ought when n is + as may be easily seen by drawing the curve .J)(l.. 0(n) (11) (12) ^-fO-X-'Cl-nXl. But there . it might be Since w = But it is oo instead when n is . with a limited number of products. to ^) as harmonise g(n) known. up to . . makes #(0) = X apparently in (11). . Or we may regard the value of g( g( 1.

When r is oo . and the vanishing between these values. the range of the oscillations gets to become infinitely s mall when r= oo. and tends Now as r is increased. &c.438 ELECTROMAGNETIC THEOEY. VIII. r + l terms in the expansion of (1 . confirmed by the fact that (13) is smaller. a radical distinction betwee n the vanishing of y at 2.. This is identically the same as viz. The There X is multiplier must therefore be oo when n is negative. n= the 1. CH. y is zero for all negative the first and oo for all positive values of n.l)~ n by the binomial theorem. ..

In first The case the zeros are absolute. makes or \r_ X[n +r For example. and cannot be magnified. then 1. if n is 10 and r is 100. . 110' . so as to lift up the curve of y from the zero line to the g(n) curve. 3T4 As .7109 . Thus.. choose X so that equation (11) shall This be true when the re are r products and n is an integer. 110 101 . r is indefinitely increased this becomes r~n so The limit that .. But the vanishing of y between can be cancelled by infinite magnification. . curve crosses the axis at the points named.2.

.. . logr f (16) S w =l + (16) * + * + A d =^ = 0-5772.S2 > 2 S3 + S 4 -. By logarising (15) we get logp= where -?iC + .is tended to when r is made infinite is the value of g(n).

(17) From of n.6C 2 S 2 + 8CS 3 + 3S 22 .6S 4 )^ + ---(18) . may be derived a formula for g(n) in rising powers namely + (C 4 .

Passing on to O. a long succession of O's. C has been calculated to a large number though I do not know that anything is So mething wonderful might however turn up.P. 427. 439 Like TT. The cosine function may similarly be and (n split into two. 21. gained by it.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. for instance. Generalisation of Exponential Function. 20. J)/(n) =f(n 1).. making (19) f(n}f( -n) = cos.M. the value of of places of decimals. the usual formula and expressing Euler's inte gral is \n_ equivalent to defining Gamma . But this f(n) is not useful like the g(ri) function.

when x= which oo. is the same limits. the other from x tooo. we may divide the integral (20) into two. [M?J + w 2] b etween any two limits. .(20) provided n is over " 1.1. because if n is > . one going from For the first use (21). and for the second use (22). since w 2 = w l = when x = . zero. since to x. But does its value depend upon w? No. There are two ways of developing the indefinite integral by parts." say ( +2 m+3 + (2 rn ( + rn-l i So [icj between any two limits is equal to [ .wj between That is. We get I = w1 + w 2 tfn. (23) . Therefore w 1 + iv 2 does not c hange its value with x.

1 in (24) makes no that n> . xn-l xn+l '+IEI (F t*l s xn ' (24) The assumption was the series to be continued both ways. . is the same. But the change of n to n .or.1.

VIII. W =A expanding by the binomial theorem. It is always e*. and now . change the value. the periodicity is only apparent.440 difference. The series on a periodic function of n in appearance. to be given later on. The function w^ may be developed in a rising powe r series operationally thus. since the But as regards of n to n + l r eproduces the series. ELECTROMAGNETIC THEORY. So (24) is true for all values of n. convergent on its though all on the When Note that # is a positive quantity all through. CH. negative another formula is needed. as usual. side. In the right is numerical calculation only the series initial convergent part of the divergent side is to be counted.

But a similar operational in (22) in a power series ~x the usual formula for The using w2 .algebrising. (28) This is the result . we get This may be confirmed by (21). treatment of result is w 2 fails. (27) But MJ + t# 2 = l. So expand by direct multiplication. so 1 =2 -G ^ (n xn .

By (1 . turns up. A (28) will occur later.we shall get it power series. . more general formula including The apparent numerical unintelligibility is no necessary bar to the It often use of (28) as a working and transforming formula.l) n is to be unde rstood the expansion according to the binomial theorem. multiplying by developing (24) to a x by the usual c~ expansion. in powers of the second 1.

which is a special case of the generalised series.. denoting d/dy by D.P. but the is not to be expected in general to be equivalent to the stopping series. to the Binomial Theorem. Substitute the generalised exponential. 441 Application of the Generalised Exponential to a Bessel Function. Yet there may be apparent equivalence.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. Thus (O. 22) the solution of in rising powers of as is 3242 by putting 2 y = J# and . and to Taylor's Theorem.r) ? Take M = . and see what we get.M. It which n result is easy to generalise a common stopping series in is integral to another in which n is fractional. The result is Now is u the same as I (. 428.

not n.n-7(33) . but is only apparent ly equivalent otherwise. algebrising. 23) is to the a^-c* = e*A-l-S^ (32) where we use the generalised exponential. with only rough values of \n. without fair When n not J.P. I also found found the explanation. then By numerical find. Later on I found that the assumed equivalence led to wrong results. This will come later at present it is enough to say that u is equivalent to I (x) for integral n and midway between. Another application of the generalised exponential Thus (O. and r is the general exponent. any fault to agreement at first. which is constant.M. . binomial theorem. So. and in Part 3 calculation this is is I (z).J. we get y.

(35) These are doubly-stopping series when n is integral. VIII.. we have the form If r = n we have the form ra (l+ft) \n =j [TO x [l[w _!_ ^!_ + [2[wj-j2 .442 If r = ELECTROMAGNETIC THEORY. and become identical. In the case r = \ we get ( ^ => -+ + iilEf this a& + E^rT+ . When n is fractional they stop only one way. In general the series (33) does not stop either way. or any integer ** CH.

(88) Similarly. without going to the general formula or to tables.r = lin (33). /(* + A) -... 4*(^) #(J). (39) 81 . take r = J.a.. w = J. Applying the generalised exponential to Taylor's theorem. So we know and g(%)... Then [n."/(*) if =2 . formulae for (37) In special cases (O.P. r = . makes +-. #(J)."1 (86) If = 1.M. 25) we can get fully convergent Thus. = 1 3 in (33) makes -. w = J. .. .

we should expect the generalised n exponential to apply as well as to the special function x /\n.M. We get (O.(40) *j/(). 24) _ 1 A (x)..P. (41) . f(x) is a power series.

= 1. 443 The extent If f(x) of application may be left open. then we get Put h/x = c and use (9) above in every term. and the equiva . 429. This is another formula of the to be positive.M. Algebraical Connection of the Convergent and Divergent Series for the Zeroth Bes sel Function. GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. 27.P. Another application (O. 28) of the generexponential is to show the algebraical connection alised between the zeroth Bessel function in the usual ascending series.. then sinnr ) 1 =2 TTT } where c is nature of (28) above. In fact (28) may be derived from (43).TJ .

*. as in 353 ab ove. and C is an equivalent form. They also act equivalently as operators. certainly someBut aside from the extent of equivalence. But use the generalised that is *-2). They satisfy the same differential equation. what is the How turn one to the algebraic connection between A and C ? other by mere algebra ? It cannot be done with the usual form of e* because then we have integral eve n powers in A and nothing but fractional powers in C. Both come out same operator. <"> is I (#). (47) . and are numerically equivalent as Here A of the well.lent descending series. times.

with the special value the others being contains all 1 = J. &c. Thus C . coefficients. it the representative term. that is. in J. &c. = J(A + B).444 ELECTROMAGNETIC THEORY. VIII. 1J. integral powers of positive and negative. r CH. J. Then C becomes it an integral power Evaluating But does not agree with A. series. we find that (48) x . for x.. 2J.

. Later on this to be equivalent to the same. B and C.. (49) which is the characteristic of A.. in general. Remem ber. we substitute . . . that x is to be positive.. It is noteworthy that ( if in the differential equation A 2 + a~1 A)M =*. and of another form already found 428 above. + Z^or 1 + Zyr 2 + (50) we shall obtain the result u = aA+b'B. however. u = a + a^x + a 2 # 2 + . on the understanding mentioned about half of the usual I Q (x). where a and not b are . The common theory of i does not hold in general in these transformations. in sum process of algebraical conversion will be extended to any Bessel function.the is.

M.(51) independent constants.P.. but are independent forms of one solution only. when r = and-J. There are only t wo equivalent independent convergent solutions of (49). The result (O. generalises to () This is meaning equivalent to I Q (x) is considered later. but any number of the divergent. but its full . 29 and 30) (52) by using the generalised binomial expansion. solutions of But A and B are characteristic.

. 430.. l). (55) On the left side there is the vanishing factor g( So.P.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. 31. We get 3 -#-*+. 32.M. on right side .)j. In (If). 445 Limiting Form of Generalised Binomial Expansion when Index is -1.) When the exponent is a negative integer the generalised b inomial expansion becomes ambiguous by the existence of vanishing factors. (O."=2 -*pln-r (54) put n= 1.

) + c(x. as it consists of the two forms in (56) combined in any ratio. To make plainer. the quantity c having any value. (1 + A- . (1 Using true when r and s are infinitesimal.s -. the limiting form of (54) when n is 1 is indeterminate.we should have This asserts the equivalence of the two forms of expansion of + x)~ l in integral power series. But if r = as well. l r/s = c.. (43) becomes r+l " s-r Dividing by s... first put n= -1+s in (54). we have (59) = (l-c)(l-x + x*-.. and then let both r an d s be infinitely small. (l + x)-i and r = That is. s-r+s-r r (58) and then putting = 0.). (55) takes the form x (1 + x}~ 1 = 0x0. s = 0. Remarks on the Operator 431. but without any connection.-x-* + x.

we using the two extreme forms of expansion of are led to two equivalent formulas f or the .1 )". By (1+A1 )"*.

the examination of the operator (1 This suggested partially carried + A1 )". equating the two forms of the exp ression . 33 to 42. as done before... VIII.). It is like This process could be carried on indefinitely. function c~ ia: I (J.446 ELECTROMAGNETIC THEORY. we shall obtain two new functio ns which might are. 353. Say (60) (61) with unit operand. be equivalent. out in O.}(n 2) q (62) (12 ' <"> The suggestion is that u and v may be equivalent.M. 9 n(n l.. Ctt. If they and we put A-1 for x i n the last equations and use unit operand again.P. making l+nx+ n(n-l) T V + J---JL-V+..

Now v are equivalent when n is . (63).S (\r] * m (wj-r (64) which result by putting r = and n in turn. 3. as the case -f J differs by a whole and may be expected l. It does. The So . results of numerical comparison of The functions u and already considered. There is identity when n = l. 2. differentiation. (62). with x = to give excellent results.J. 9 Increasing n to f and T ^ also produces apparent equivalence. are briefly The these.

It is much worse. a nd T^ . t. when 65 > < which show no sort of agreement.T9^ to be worse. made probable by further examination. J. o. For i f we go down to n= f the function v for . then. We may expect n= n= n = . . . and not merely at n J. (or last convergent term) must be adde d to make the value of u. along.1. far. we have Going from finally. 0.9 to -1 accentuates the difference.J also show an apparent equivalence.cases of n = J and . it is suggested that there is real equivalence all But this is not l. oc = 1 and 2 is rather too small. and a part of the term following the 1. v being zero.

GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. 3 v=(A-A + A -. Now if we apply harmonic obvious..)l. 447 Yet even in this case of way of reconciling the extreme apparent failure there two forms of operator. 2 (66) (67) That u is e~ x is analysis to the other form. we shall get e~*. 1+A 1= ~ l-AMo ..)1...1 + A. Thus is a M = (1-A.2 ~. cosmxdm when is x is . Thus.

P. there will be some notes later. becomes indeterminate in form when so we may expect anomalous results. Two suggestions aris e First. some other cases of apparent equivalence. This change from to e~x also true of u. that as the generalised binomial n is 1. . say u = v + iv. Remarks on the Use 432. As regards the extension to (64). of Divergent Series. 1 the function iv becomes with the values of n from f to At any rate I know that this is what happens in important. w here w is quite zero for the special values of n but that when n has other values producing apparent equivalence. I have stated the growth of views about divergent series up to that time: I have avoided defining the meaning of equivalence. of n. In O. 43 to 48. since the operand 1 only begins when x = 0.M. u and v are not really equivalent. That has to b e . except with special values interpretation..+ and . the value of w is too small (with the values of x t ested) to influence the result whilst finally. but that an auxiliary series is wanted. when x : is . requiring special theorem The other is that.

my found out by experience and experiment. has an infinite value. o f Solutions of physical problems course. my first notion of a series was that to have a finite value it must be convergent. The definitions will make themselves in time. should always be in finite terms or in convergent series. otherwise nonsense is made. Starting from complete ignorance. . of course. A divergent series also.

according to a known negative. It is a bad arrangement of parts. CH. but the last + and -. it may be practical. Then came a partial removal of ignorant blindness. p. In some physical problems divergent series are actually used. But by the same reasoning. Aft er initial convergence. It cannot represent the solution of a physical problem involving finite This seems to be w hat Boole maintained in his values. Now here the terms are alternately term included. a finite quantity divided into parts in this way. VIII. is infinite in value. of course. referri ng to the divergent formula for the oscillating He showed that the error was les s than the function J B (z). the terms do get bigger and bigger. and alternately positive and In the latter c ase we are. It is possible to imagine alternation of sign is significant. but as the initial convergence guides one to the value.. terms + . a continuously divergent with all its series. notably by Stokes. "It is known Differential Equations" (3rd Ed. 475): that in the employment o f divergent series an important distinction exists between the cases in which th e terms of the series are ultimately all positive. " . This seems to give a clue.448 ELECTROMAGNETIC THEORY.

is But.use A. later on. . divergent series presented themselves to in a different me operators may as differentiating operators. The be the same functions of A. The plausibility of this argument obtained in this way." He illustrated this by integrals ascribed to P etzval. but not A. permitted to employ that portion of the series which is convergent for the calculation of the entire value. The equation to this. when x is + may So there is only one solution of the characteristic b ut not B.law. According to Boole. use B. (tf is satisfied It is equivalent + x. or dfdx t as would way viz. or else series which would be either . evident. H . that is. as evident as t hat A is oo when x is + .) But if a? is .. we multiples of (a.&)u = u 1 (69) by and K (). make convergent series.

was analytical only.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. one convergent. use I made of equivalent series. and the connected generalised There are certainly three kinds of equivalence. and to This and other con siderations show a divergent by another. the solution of a problem arises. mean The . The se ries. A B above are really independent solutions. was numerical. connection is given by C = J(A + B). The third was algebraical. is the secret of the continuously divergent series. between the two c ases asserted by Boole disappears. Introducing an operand and algebrising. Examination into the reason why two series. and we seem to have something like a distinct theory. using a particular does not identity. For an operator may lead to a convergent solution by one way. Equivalence 429 above investigation in The identical illustrates several points in this conn ection. But this view is soon found to be imperfect. if the tically. then. They are n umerically meaningless. 449 alternatingly or continuously divergent if x took the place of A. as well as algebraically and analyIn the analytical u se. must be considered numerically. even when contin uously divergent. using x\ the proper use is as differentiating and So the series operators to obtain convergent solutions. connectin g the first series concerned by means of generalised series. examination shows t hat the initial convergence determines the value of a continuously divergent series in the same sense The supposed distinction as an alternatingly divergent series. every term must be used. But we cannot use all terms in And numerical the numerical case. the other d ivergent. and in a convergent form. Here. and neither should be rejected when used as operators. result is a convergent series. one of which is conThe second one tinuously div ergent. that divergent series. are equivalent. leads us to consider generalised differentiation. because there is no limit.

But it seems to effects the reconciliation. viz. convergence er of cases.formula for e*. This further explains some other things. The series B ng of divergent series still initial obscure. the different behaviour of equivalent when x is +) and C (which are numerically on making x imaginary. as will be A shown later on. But the numerical meani remains The property of estimation of value by the is a very valuable one. and is true in a large numb fail in a marked manner GO ..

when mathematicians of could not see the validity of investigatio ns highest repute The results reached. (See the examples of CH. considered. The limit of (d/dn)x* when w = is logx. and the step (O. There was a corroboration. say. or. In the analyti cal use. 433. they involving the algebraic imaginary. is not numerical purposes. There will Logarithmic Formula derived from Binomial.) merely for above and suggested exseries.P. as In any case operators. planation. VIII. 49. The error fa llen into by Boole was striking The use divergent the time. Using (72) where r has any leading value. and required independent But there is now a th eory of the imaginary. too. ELECTROMAGNETIC THEORY. 50) . however. have to be a theory of divergent series. were only suggestive. including convergent and divergent series in one harmonious who le.M. a larger theory of functions t han the present. we must not be misled by apparent unintelligibility to ignore the subject.450 sometimes. the question of value does not arise.

to S *r -l. But here </(<>) = 1. to n.is unity. (76) by (17) and so we reduce log (1 + x) = . (77) . (28) . = = 0-5772. (75) by equation 0(0) (9).C + Z aty (r) g'( r). we obtain '-^"-"' <7S > where the accent indicates the derivative with respect That is.


we get (O. (80) This reduces to l-x + x* .1). 51).. when r = 0.M. (1 + x). we ^-l) = l. and . If (78) we differentiate (75) to x. (l+^-Zaf'rtr-lfcX-r). or./(-3) = (2. 451 The common formula must use is the case r = 0./(-4)^-^&c..GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. To obtain it.^(-2)=-l.P. increasing r (79) by unity.= 2 *g(r)g( 1 r .

But this formula turned out to be i ncorrect later. with respect to r.M. the left side will make e~*. Similarly. 54. O.it may be readily tested to be numerically right when r = J. wanted in the theory of Bessel functions.M.P. and we introduce unit op erand.. and only Logarithmic Formulae derived from Generalised Exponential. Thus. . 52) as true when A' 1 is put for x. may be skipped. erroneous formula.P. being based on the partially valid. so need not be given.P. (81) We get (82) . proper one. * = ? *</(>').. and so the right side should be a generalised formula for r*. 53. If the last result (80) can b e regarded (O. to Its failure led to finding the be given later. 434. differentiate the generalised exponential we may derive 56 to 62) some formulas involving the logarithm. From (O.M.

0-*loga? + 2sy(r). we (84) may write . In another form. A second and a third differentiation give 0= -e*(log*) 2 + 2*V'W. (83) 0==+e*(log^ + 2*y''(7-). and so on.

If .. Thus a?* = e*{l + log x + = + .452 ELECTROMAGNETIC THEORY.. (86) by the use of (85) becomes 2^-/+ --. (r).)W = 2 afy(r. (87) (88) we use (82) thus.1) = &2 arg(r) = xtx by using Taylor's theorem. These various formulae to may be combined to make others.}. CH. (89) to obtain a formula for ment of terms we . show a fit. VIII.

also and brings us back to x~\ 1 We get . if not usable... (43) in (82) to specialise. positive integers we have -(l*i + J+. There is only poor initial convergency for calculation with The second line of ( 93) may also be small values of x. comes power series for x.].+). written . A The equation take r = Now in. (78).log x in powers of are led to x. by rearrangewhich fails.. The values For of g' for negative integers were given before. is striking. These bring (91) to (92). (91).

2.. + ?-logr. 45S A different way of reaching (93) is thus : Integrate the identity with respect to x.. giving x the values 1.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. with r increased to oo . We get . .. in turn we approximate pretty quickly to the v alue of C. (98) Connections of the Zeroth Bessel Functions.. 3. 435.. given by the limit of (97) when x = oo . much faster than by By C-l + i+J + . (96) where C is the constant of integration.. &c.

equivalently.q. tigation later. at the sa me time turning prefactor. by differentiating the function u = 2f[g(r)]* with respect to logarithm. though. zeroth Bessel function as in and its companion with the But the results 428. 64 to 67) were considered on the idea tha t u repre(O. sents IQ(X). if. A to A . Then put A (A + q). which is only true specially.P. Introduce 5Z as and e"5* as postfactor. y stands for \y?. there the special cases. (99) r. The brise.M.In a similar way. So apparent numerical equivalence beyond this is superseded by the proper invesin d ivergent series The companion zeroth Bessel functions are got by algebrising was done in 353. The result is \7rqxJ ( 8qx (101) (8qt) [2 . we obtain various formulae involving the I Q (x)..1 for the final e~ x and algefirst . as before is mentioned.


CH. and to show 68 to 71). vin. Thus.454 ELECTROMAGNETIC THEORY.M.qx . was got (O. (101) only differs from the last in all signs The equivalent convergent series is in . Now it is striking that the second operator in (100) similarly treated leads to the companion function. for distinctness. by introducing the pref actor .P. how (101) A" 4? [2 1 + 1232 " I J ' (102) Sqx (8qx)*\2 The function being positive.

But q = si .j4 q = si and be real. Thus Z'l JoW = 1 is '+ r4 . (105) in (103) the original Fourier-Bessel function. Now letting the oscillating functions s come out by putting .which I Q (qx) is given by The functions (104) and (103) are.. as usual. being conversolutions of rigorous mathematici ans.. gent.. in which respects there is no settled practice.6 r6 + . the companion in the standardisation of (103) by the two constants 2/ ?r save and C.

makes K (^) = G (^)-a o (^). (106) where the new function G (sx) is given by A It is Lt A U J . G that is the proper companion (oscillating) to J .

(110) (111) G Q (sx) = (-L) \7rS#/ (S'cos Rsin)(* JTT). 455 Similarly q = si in H and K makes (^).GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. J W(108) < 109 ) GO(*) # o(**) the bars indicating that divergent series are used.' where R and S' are the real functions of sx given by ~ 1232 + 12325272 ~'--' 1 232527 2 92 . and (Rcos + S'sin)(8*-i*r).

iG }(sx) from the first and 2J (sr<) from crepant. form. Here we come H We get (J . the second.q . 1st Ed) in a split and not identified. 154. But I have shown that the identical connection is In this put q = si._ 1 1 2 3252 ~to~(8^ (107). which is an G (s#) and G (s. and there results where So we produce harmony by the function G (s#). or rather equivalised with G Q (sx). equivalent form of We have J ()AG (p) G (s*)AJ M= . Th e series (115) occurs " in Lord Eayleigh's " Sound (Vol. (2#) and 2I (^) are But the transformation q = si makes them disequivalent.r). L. + 711 7^]5~' Equations (110) and (105) are equivalent. and so are (111) and to another matter explained by the generalised The functions exponential series. p.

(111).( 2/. (117) But in the transition from (117) . or else the pair (110). (107). HoMAKofe*) . (116) when using the And similarly pair (105).K (^)AH using the pair (101). 7 r) = 4/^. (102).

M.iG Q (sx) H (qx). Then we have rp-iA? i (A^-^o algebrise. (116) by q or J Q (sx) si it ctt. . r and vt. vin. and should be studied in immediate connection thereLet there be two with. . for which there is no space here.P. is indifferent for whether we substitute 23 (sx) . The H .456 to ELECTROMAGNETIC THEORY. and K These the principal mutual re lations of compactly formulae occur naturally in the treatment of columnar elast ic 436. whose differentiators are A and q.. Q Operational Properties of the Zeroth Bessel Functions. variables. say. 72 to 76) brings together following (O. waves.

or [b] or [<?]. The result [d] may be derived from from [b] to [d].w (118) 7 ^\4 . and ~ vtA to [P]. and algebrise. [a]. w L J This makes one set. and A and q throughout. To get [a] from [P]. Thus. and expand introduce the prefactor the properly transformed operat or in descending powers of A. To get [c]. and expand the properly transformed operator in descending powers of q. Another set is obtained by interchanging r and y. . and alg ebrise. To get [6]. introduce the prefactor t qr to [P]. expand [P] in descending powers of A.

To get [d] from tions. use harmonic analysis. 457 Similarly from to [d]. In the theory of pure diffusion there are analogous results.t o outward-going waves. of the simplest is this. (121) Try o known integral. equivalent to [d]. is obvious by executing the differentiaSimilar remarks apply to the second set above referred [e] to the the change from one to the other set being usually related change from inward. to. /"* V</r)ol a i Io(ar)cossi'f if <&=-! J Try o 00 (sr) cossv* ds. d/d(vt) (1 22 ) One . thus.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. Change the meaning of q from On this understanding we s hall have to (d/d(vt)}*. [c] To get [tZ] from [a].

and has since received it. and therefore to be evaded. which must be boiled down to 20 or 30 by omission of details. When algebra reached a certain stage of development. Coming toO.3. however. But it would not submit to be ignored.Part.P. Remarks on Common and Generalized Mathematics..60pp. the It was exceptional. if possible.As regards the proper use. + bi. and imaginary turned up. 437. It demanded consideration. and great extensi ons of mathematical knowledge have arisen out of the investigation of this once impossible and . now say a a specialisation of the algebra of the complex quantity.S.M. The algebra of real quanti ty is unintelligible. = about 80 or 90 here.R.of Proc. that ca nnot be understood save in concrete applications. sometimes of one method sometimes of another.

Vlli. and is the imaginary. and i n the same way. Some done. writers on double algebra appear to think that when they multiply together compl exes they are multiplying vectors In an illustrative sense. vectorial work is be ing together. generis. may be questioned whether it is entitled to be called a quantity. It has no essential connection with vectors or with quaternions or with circular functions. But it is versorial rather than vectorial. It turns up by itself in these subjects It is sui just as in scalar algebra. bi) Say . non-existent quantity. though it may be used illustratively. It CH. but there is no question as to its usefulness.458 ELECTROMAGNETIC THEORY. and the algebra of real quantity would be imperfect without it.

The point is that the multivectors at all. plication does not refer to prevalent (though of late years has been disappearing fast) that vectors ought to possess the associative property in multiIt is not in thei r nature to possess it. when these are re presented by lines in a plane. and + bi and c + di are also versors with a stretching faculty as well. and their product is another operato r of the same sort." . This is strictly scalar algebra.(a + x = (ac (c + di) bd) + (ad + bc)i. Again. The idea is too it a proper vectorial equation. Thus.bd) + (ad + lc)i\ X The x may be a unit vector. in Quaternionics. But even illustratively. all directions in space were real. and may be omitted altogether. then (a is + bi)(c + di)JL = [(ac . it is not the multipli cation of the vector c + di by the vector a + bi that is done. to a form a new vector. Versors plication. possess this property naturally. Keally i is a quadrantal versor. if x is a real \~ector in the plane.

Yet quaternionists profess to be doing vector work.made this "equally imaginary. It is not a new subject by any means. The real imaginary enters into quaternio ns just as usual. a puzzling subject. as at present expounded. and go on confounding versors with vectors. I think there is much room for improvement in the way of expounding the science of quaternions to those who are hardy enough to attack what is. . and therefore equally But imaginary foundation gave way long ago.

the 7j/2. it is only a spurious that C = Ye determines C from e. t he time differentiator. containing p. Then if we specialise e to be simply in the other way. One is to assume a complex form of solution the end. For that I prefer to disc ard the quaternionic idea completely. field of its more desirable because there is no doubt that Quaterand should have a useful own.T. and use a simple algebra b ased upon vectorial ideas. It comes out complex at two parts may be selected for a The algebra is that of t he real imaginary. Another use of i is in the treatment of physical -differential equations whose s olutions are required to be simply periodic f unctions. This is quite independent of the question of the suitability of the present Quaternionics for the purposes of physical inquirers. But at the beginning. periodic. if the . 450 It is the nionics has been unduly neglected.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. and e also C.ri2 . 2 power of p in Y is . through the operator Y. Then real solution. either of its if i be used at all. for the usual purposes. There are two ways. Say function of the time. a gi ven imaginary.

specialised form of it.frequency is fall of be This reduces C to the form C are real functions . as the imaginary was the source of a large branch of mathematics. may be done illustratively by lines in a plane as before alluded to. tial one. and asserts the necessity of a prop er development Besides. and that suggests its avoidance in favour of more roundabout but understandable methods. This oc curs when we know unique solutions to exist. Now just as the imaginary first presented itself in algebra as an unintelligible anomaly. of the subject. so I think it must be with Ordinary analysis is a generalised analysis and series. methods. we come to the result C = (Y. + Y 2m') e. for the i means p/n. But it is Either of these not. we come in practice quite naturally to fractional ones and combinations. when Y! and and of n 2 The solution But if we say p = ni. the algebraical complex method or the differenexplicit. There is a universe of mathematics lying in between the complete differentiations and integrations. which looks imaginary. a differentiator. so does fractional differentiation turn up in physical mathematics. Y2 = (Yj + Y^)?. It seems meaningless. But it refuses to be ignored. Starting from the ide as associated with complete differentiations. .

when found. As regards some of the problems worked out later. The question of physical application raises another. But it is unnecessary to appeal to utility. to a mathematician. Generalisations are to some extent arbitrary.4. It is equally true that we cannot fully understand algebra. what is useful will find its way out. In illustration of this I may refer to the equivalences eralised analysis. whatever why the series great extension follows. without genI do not say that it is fully understandable without mor e light. for instance.460 ELECTROMAGNETIC THEORY. with treated of in Ordinary algebra furnishes no reason A and C should be equivalent. of it may not be useful. Compare. Vllt. The same can be said of the physical The bulk imaginary lore. It is true that we cannot fully understand the usual algebra of convergent series without the imagi nary. I do not know any way of doing them by ordinary analysis. darkness. whether real o r imaginary. and to furnish the most ready way of getting results simply. generalised analysis does. But a little light is better than it. It may be possible to . The This is n ot an isolated example . according to the direction they take and the nature of the co ntrolling ideas. There is another analogy to be drawn. CH. with the complication o f the subject when done by ordinary analysis. the simplicity of the processes used in 353. As the s ubject may be developed. But some of it I have found to be very useful. 429.

simple themselves. It is these general characterist ics that seem to give reality to the mathematics. along safe paths to useful results. and serve to guide one Every one who has gone . and their variation in time and space according to de finite laws. In physical mathematics But the trolled quantities concerned by the special are not arbitrary. but are conin relations involved certain laws. So we have a definite march of events from one state to another. without that complica ted multiplicity so common in pure mathematics. for instance. or it may be complicated functions of other quantities. the necessary positiveness and singleness of cert ain quantities.elaborate generalised analysis in different to be useful in physical applications. and their continuity of existence in time or space. should be developed to suit them . it ways. or both. involving.

That would mean to stan d still. Prof. having started on a physical foundation in t he treatment of irrational operators. a physicist wants is a good view of the physics itself in its mathematical relations. beginning with numerical g roping. This is groping. The Eucl idean logical way of development is out of the question. more and exceptional peculiarities are useful than exact mathematics. something more rudimentary must take its place. in any way possible. and obtain the im portant assistance of physical guidance in the actual work how important for This being the case generally. The mutual consistency of interval between Parts 2 results is more It is satisfying.M. which was successful. the physical guidance becomes of getting solutions. in seeking for expla nation of some results. ignored. O. more important still. 461 seriously into the mathematical theory of a physical subject (though it may be p rofessedly only an ideal theory) knows it is not to look upon the symbols as standing mere quantities (which might have any meaning). it is clear t hat when one is led to ideas and processes which are not understood. and when on e has to find ways of attack. the formalists or algo rithmists. and it is experimental work. when perfo rming some very complicated and laborious calculations. we are left nearly in the dark. Klein the distingu ishes three main classes of mathematicians intuitionists. I got beyond the physics altogether. Now.. So I made a detailed and close exam ination of some of the obscurities before alluded to. with of course some induction and deduction going along with it. For what apart from the narrowness of special mathematics. But when intuition breaks down. and the logiNow it is intuition that is most useful in physical cians . for that means taking a broad view of a question. If it be wanting.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. mathematics. First get on. The result was . but to bear in mind the physics in a broad way. and it is quite a secondary ma tter to have logical demonstrations. and let the logic be left for later work.P. These remarks are caused by certain experiences in the and 3. and was left withou t any guidance save that of untrustworthy intuition in the region of But success may c ome by the study of failures. when the mathematics is well known. pure quantity.

to clear up most of the obscurities. correct .

it is only compreherisive. series we cannot do this. varying according to certain laws. A physically-minded man need have of no difficulty in conceiving the existence of a function and time. and I take the view that the whole s eries is significant functionally. may b e. No. if found. we may go as near as we like. they would be much too position application requiring various It is characteristic of rigorous mathematicians. and its mode of expression in symbols standing for numbers. CH. and with any number o f discontinuities in it. and consider that it is the fun ction. that they think too much of the formula. the errors involved. what would be simply a single function in its physical meaning. . without any power to find formulae which will have the necessary properties. One form of expr ession of a function is a divergent series. It is generally the dress. the value of th e function by adding on terms at a uniform rate. some importance to distinguish between a function the physical sense. It is of in some account of the results will follow. I think. and by their revision to obtain correct formulae and extend the results considerably. though we cannot reach numerically. so on But in a divergent to include the whole in a finite time. and not merely the few terms that may be util ised In a convergent series. several dresses are needed. Leaving out the details of the groping. and need not be a convenient fit. the physical limitations and reservations. too large or. VIII. disconnected. a nd it is easy to imagine practically taking the later terms in larger and larger groups. though the initial convergence . Perhaps.462 ELECTROMAGNETIC THEORY. for instance.

even when showing very small difference. the principle has a very wide application. How far does this property extend ? It can be demonstrated to be true in certain Yet it can cases. to test their equivalence to other convergent series. It has enabled me to distinguish between true and appa rent equivBut I alences. of divergent series. for we might make hardly up series arbitrarily. to continuously divergent as well as alternatingly divergent I have employed it in the examination of a large series. having no particular characteristics. Nevertheless. and h ave found it very useful. but something more g eneral is wanted. number do not think that the size of the smallest term does always . be considered to be generally true.for may guide us to the value approximately.

Now it converges to infinity instead suppose we slightly alter the law of the successive terms in a suitable manner so that the converseries gence ceases at a finite distance.. the terms preceding and following the smallest influence the value. is called a series because it sums up to infinity. that it is convergent. Yet. only that of to a finite value. ad inf. We may say convergence. after which the becomes Then the point of convergence finds the value. 468 govern the size of the error. But I do not say that all formulae admit of a continuity of c alculation if they pass from a state of finiteness through meaning of the formula may change infinity to divergency. at the same time. This simple illustration will serve to illustrate the nature of This series 1 + J + J + J-f.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. term s get smaller and smaller continuously. May be. as divergent.. because the glance. The The Generalised Zeroth Eessel Function Analysed. . near as the smalles t term will allow. a t first divergent it might be thought to be convergent.

. in this c ase we On first have had separate verification by algebraical use of the *=i I also tested u for r = J. 2J. Besides. Then. f. There was apparently rapid way. though not minutely verified. Consider the generalised function *. havin g no tables of g(r). This includes 1J. Thus. But an anomaly presented itself. u is I (#). (124) (125) if G(r)=y'(r)/g(r). equivalence. Diff erentiate U to r. though only in a case. a rather important function.438. Is then u a function of both when r is not integral ? experience it would seem not. tion. &c. but that U remains This is closely true when r = J by calculaIQ(X) as r varies. if r is zero or any integer. and y1^. I (#) and K (a. In any case it satis fies the characteristic of I (#).). (123) Let y = \y?. But that has two solutions.

So must involve K (o. and must vanish when r is integral. (127) where b is independent of x. as in (103) above. in spite of its apparent equivalence I (a).464 ELECTROMAGNETIC THEORY. and must satisfy (126) whe n r = 0.. and when r is J. in the case r 0. It must be a periodic function of r. really I (#). (r = 0) U'=-7rK (126) where K (#) is the other solution. U to Say U = I + 6K . VIII. U=I . viz. These considerations suggest a sine function. is Now U U f But get it is not zero We (z). if CH. = according to (125).). is zero. with period 1.

For. = 2. rapidly oo at the origin. with the not small values o f x which are needed to produce marked initial convergence in U. The function I is 1 at the origin. is term like a small satellite. if (128) is true. the extra I rt now U always. . (128) -7rcos2irr. 2-279 when y=l.sin27rr.E (129) seemed to be easy to see why.U'= It is . but decreases so fast that it is only . 4-252 when y On the other hand. increasing to oo. and K is so on.K .

when (r = ). = J) (r = f) (r U = I -JK U = I + JK U = I .7rrK .O072 when and I when y = 2. so on. (r We and require . 0-009 when y = 3. (130) . the distinction between U may be invisible numerically. 0-027 unless specially So. = 0) U'= -7rK . The next thing itself in is to see whether the satellite really shows the numerical results. and sought for. y = l. U'= +7rK andU' = 0.

(131) (132) . (133) all these. Then = 1'005706.r = TJ?r . Test (133). r is very small. I have tested calculated. /(T ^) = 1-011443. . and find them true when carefully and with values of x and r suitable for allowing the satellite to show itself. . It is not necessary to give details. Say. so I will quote one or two results.

itself (134) is Now though small far larger than is permissible according to the size of the l. these make 2-27958. and that is all that we ca n do. j-UU which agrees with (134) up to the fourth 6gure. for th e error is now brought to be smaller than the l.c.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES.t.c. But the value of I -U = 0-00220. So I this difference U = 2-27728..t. 465 is and when y = 1. But (135) ^K = 0-00227. Io -U = 0-0139. With y = TG> tfv* results are I = 1-1025. This makes matters right. (136) U = 1-0886. .

using (125).0-0147. Generalisation of e~x 439. We establish not merely that U may be the function of I and K given above. r = -^KO. perhaps arising from errors in calculation. the result get . (129). and make note of the size of this half-term . Nearly the whole of the evidence supports the tr uth of the formulae (128). in the series U.t. Putting H Two .The smallness but in other cases. I usually count J the l. we J in turn in (129) and subtracting. when y is The other cases may be skipped. for instance.c.t. but also that the error may be reg arded as limited by the size of the smallest term in the divergent failure. and usually employ smallish values of y to avoid the very lengthy calculations involved large. the case is I only refer to these tests because a typical one. lUU of r allows small values of y to be used here.c. Algebraical proof will follow. K (z) in Terms of Expression of the Divergent (a:) and G eneralised Bessel Funct ions. assumed equivalence would have been rejecte d at once. larger values of y are required for clearly showing that the satellite serves to bring the error within the limits permitted by the size of the l. the If there had been distinct series U. J. and the little that does not is of a dubious n atui^.

r=0 and 7rK W = (2-2V^r)y(r) r= r=0 (137) .

What then is the function ? identically in terms of the ascending series for To answer this question.466 ELECTROMAGNETIC THEORY. H and the r = and J cases of the preIs there a similar connection between the usual descending K and (137) ? If so. we had This expresses the function K without the logarithm. CH. we should obtain the two oscillating functions J and G out of the right side of (137) by putting zi. But this fails on trial. not a mer e equivalence. we r equire a generalised formula K for e~ . and (137) expressed an identity. Now before a remarkable con nection between the usual descending series for sent function U. VIII.

t (140) tests of this formula are by no means so satisfying For instance. Suppose we alternate the signs the generalised e*.it x . The simplest way of doing it is to say . we require v = 0. . of the terms in ^ . To obtain the correct grope again. xr ^ 189 j Then v satisfies the differential equation of e~ x It is e~x and is . and w hen r is odd. This means that the gene ralised e* can . r = J. I first used a formula before referred to it ( 433). but would not work. t> = e-*cosnr = 2^. Say xr + generalised formula. and on examination was easily seen to be incorrect. with as those concerning the generalised c*. so it must generally be Ae-*.e~ x when r is or any even integer itself when r = an odd integer. where A is a -1 periodic function of r which is + 1 when r is even.

the rule regarding the error limits is satisfied. we can only conclude that v is a small quantity. and in the cases r = J and . Now we have identified + U^ . Express O (. so (140) may very well be true.Numerical then be divided into two equal parts. This is satisfied as far as the initial convergence allows. the suffixes (#) with meaning the values of r.-B) similarly in te rms of H U K . in this. However. but since v is the diffe rence of the two series. sum of even terms = sum of odd terms.

5 \ + / (g) 4. 467 wo x U's. therefore. only necessary to size up the coefficients of the leading terms. In equation (138).2. We have V2. /L3y 2 2 i2 2 /L3. thus.3. y .2. To identify K of series of the type U.2.5X 2 V2. When this makes a pair it is arranged in powers of x or preferably in powers of y. (i)* 21 2 2*33 2 L3\ 1 7T /1. x "u/ 9i9i1 9^1^19 "^ ^~ 'u a "a "a / . put for e~* the ordinary t~ series.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES.

involving the other powers of x. the value of [g( . (143) In the same way. t I i + terms The to 1-0782. (142) coefficient of (2/7r#)* comes to 1-1795. . and that of (2a?/ir)* The former is We get 0'665 and 1-217. all . Divide these by TT*.9 3 l4i'-i ^ 4. by taking (142). We therefore prove that K W = U. -U i i .J)] 2 and the la tter of [#(i)] 2 .

signs positive in (141) and we show that H W = U_.Sn^. The Divergent H n (#) and in Terms of Two Special Generalised Bessel Functions. any r. 440.+^-Ko. (145) (146) (147) Ur + U for m=H K n (z) Ur-U. + U we i . corroborate the formula (128).rr. ). .K . then. (144) So for it far. or U = J(H makes -sm27rr.

any order. . (147) with r left arbitrary by e* employing the generalised and the proposed "* formula . We can now extend our results to Bessel functions of We may verify (146).

VIII.n. and co-ordinate it with Hn and Kn the div ergent series also satisfying the charac. We have to find out what U represents. r-1. the series to be made complete both ways. though only one function )t in terms of I n and I_n The factors will is to be expected. namely. 2 (A + ar^A)U = (1 + ^> )U. teristic.468 but this ELECTKOMAGNETIC THEORY.. got by changing n to . . &c. CH. as usual. because be functions of the characteristic (148). 2 (148) if A is dfdx. because the work may be done upon The characteristic is the gene ral formulae for n and K n H . only the leading term is written. All the rest follow by changi ng the constant r to r + 1. The function U n must be expressible it satisfies r. is unnecessary. and a convergent solution is The companion solution is !_(#). Both these known func tions are included in the generalised formula where.

1'Pick out the terms involving of each of the two H -y-*{l + O^y? f i W (l'-4n')(y-4 " ' (2(8)2(2 12 4->J^ .First of e* all. at and and ~x see the effect of using the ordinary series for in these formulae.+ IT 2 "" / . ar*. because series of the Un we want only one term Put x = 2^*.y PI We series l-4-i"_. then type.

r=^ and J are involved .i ll(8)(2 [2(8) [3 -J (153) + other see that terms. and since in the .

we mus cancel the in by inclusion in the i Thus n + r=-\ -^ therefore -2 + r= t makes makes r= r= jl-n -. jj^-n j in first case.\ CIK~\ = = ./ m %. 469 we have the exponent r + \n in ihe leading term. r. and -l 1 . ^ ^.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES.

is (_*_.i).) + !!. by alternating signs in that Hn . Hn = U n (153). and that of y is In the same way. we can show (158) K n = U. Comparing with that we equation g( J see that the coefficient of y~* in l ln)g( \ + -. ..(}_. So H n is the sum of two Un . with leading terms as just (157) found. thus. (-J-Jn) ~ U n .series second case.

(-J-{) = K H n+KH).) in Terms of any Generalised Bessel Function of the same order.(i+Jn).+ is )} (160) the general relation. The Divergent H n (?) and Kn (a. Changing n to -n. by (154). makes no difference on either side of the equation. that the exponant r becomes So it is suggested that from U 0ir t o U n>r U. . U 0> r = }(H -K . Una-n) =i( H -K n (158). true if (159) is true. we found. subject to verification. (159) We also in passing know. and. increasing r by n. at the same time. that sin27rr). 441. r = J{Hn . Spe cial cases of (160) are U n. Now.K n sin 2ir(>.

(163) .K n sin 27r(r i n}\ . by negativing n in (160). U_ n>r = i{H n . (161) (162) These are equivalent to (157) and Also.).

We may also express Hn and K n in r's differ terms of two U n functions (168) (169) . .Kn cos TTH sin 2?rr. (166) (167) K = ?-""" whose sin Trn cos 2?rr V .rr.^ n>) n>r (164) (165) B>r n>r . HH = U re H>r + U_ n r + (U_nir . cot tan 2. ELECTROMAGNETIC THEORY.U_ = . CH. U + U_ = Hn . and from these again.470 so.U n>r ) . U . VIII.K n sin irn cos 2.

case r=J is worth notice.r+i.by J.. . It is easy to see .. because it is not the same for n When n is not in . by (160) (170) U Wii = U _ n>1 = J(H n .U n > .Kn cos iro) the U function is the same for n as for . It looks unlikely at first that the (171) expansion of Un according to (154) should be valid here. = J(Hn + Kncos7r). The that that is. Similarly UW) = U .r + Un. sin 7r(n + 2r)K H = U n r+ . Thus Hn^UH.n when the value of r is J.

we require 4r = odd. then 4r is odd. then n is an integer. unless it Un>r = U_n>r be integral. Further. of to reduce these equal series to JH requires 2n to be integral. To make n. and 2r is integral. (174) K n sin . the Kn formula. for any value when any r will do. If odd.tegral as for n. save when n is integral. there can only be equivalence. as before. Going back to (167). Or U^. To examine this. but not yet verified algebraically.. we find that equivaComparing lence depends upon (172) These agree with the r = J case of the generalised formula (139) suggested above . If even. take n = J. results by the two m ethods.

.r.n cos 2?rr = U_ n r .

Putting I n for Hn we have +i . as usual.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES.M) step of r is 1 ' (176) In the summations only the leading terms are written.) log y% + 2 *"^"-G(n + .^ log y* 2 j^G(n) . Then differentiate Kn 7r cos irn cos 27T?= (Un> r + U _. is integral... (175) In the case r we should reduce to known formulae. .) + /-'" G(.+ r). The now. 471 Both members vanish when r to r to find K n The result is .

an d in the second the generalised e~ x or Multiply up. Hn. and we have (178) briefly in In the first general way. powers of x or y.r) ) (179) (180) where <(w. Hn = 7T 2>-^(.The result may are stopping series easily be expanded to the known formula The two summations where S r =l + J + J+. go back to (151) and (152). oscillating functions. . '-i*(n.r) is given by . and we obtain.r). by arrangeme nt in (140) above. K n and Un U_ n . To a more show the connections of Kn (*) = i-(Gn -a n)(*). use the generalised e*. put To obtain the x= zi.

Take half the sum and J(H. + K. Then (182) =L 2. cos nr) difference.r). . >-^(72.

VIII. to harmonise with . \ or r or r ~ 2s +w+J in (182). Un r. not 1. . J0H. cos nr) = . CH.472 ELECTEOMAGNETIC THEORY. Put Also put \r = * + JM. + l = 2s-?i + iin in . + !) ( 188 ) is But it In these last summations the step of r is 2. r + J = s.K. better to alter so th at the step shall be 1. 2 * + *(.Jw.

(187) This involves 1 I2 .-^ r +r 7? v . will make summations which the step of is 1. we see that (W-U^-Z.s. (183). (184) (185) K n sin 7r(2r= 0(-). s These Then. put r for The )} results are 6l( -K n sinT(2r + W where = W ). )} I2 Comparing with the previous investigation. after the change.4n 2 A + 32 - .

of the generalised series for t* and cmakes the power series t . alon g with x the generalised e formula.x to 442. Possible Transition from e. the product of the same when generali sed came to 2 or to x we could be sure there was something wrong. (188) which is an identity. then. The product. say Q.. makes it practically certain that An algebraical p rocess leading to it it is the correct formula. will follow. But we need no t expect that the product should reduce to 1 in a plain manner. yet the fact that its use. If..+n (^(r .. The product of e r and c~x is 1. Product of the Series for e x and t~x cos nr. the Although the generalised convergent and the divergent. e*. e~ x formula has so far only been used speculatively. enable* us to connect algebraically the two sorts of expression of the two ?ith Bessel functions.

(189) (190) where +(r) = -^*(l-. . and . .unless when r is 1. .r + r+l\ r + 2\ to which may be transformed So <(r) integral. + o* r <(r) .. is oo is 1 when r is . .. Jl-*(l-^(l. on. . But when r is fractional. is Q is when r is + and 1 when r is 0... 473 Q= . and so .a^+tyfr + 1) + x^+*<f>(r + 2) r (Y)1 .GENERALISED DIFFERENTIATION AND DIVERGENT SERIES.

Q assumes the indefinite form of a succession of infinities.--COSrrr r (192) r. not plainly reducible to cosrvr. When x is + = . Differentiate to then (193) w' (log x **ir cosrTrV l l + G(r) + TrtanrA / This is zero when x is e considered. ic is e~ z .* Write (140) thus. representing x . w = ?.

This being 1 wh en x is -H if it is different . The present formula does figures the same are differently arranged. also satisfy .. it is made a function of x.c.1) u = Q.+ Compare with u and w' before and 0. x is . which makes the l. (194) Compare with approximation not. so does But u' t u". when So there seems to be an irreconcilableness requiring some modification of ideas. for That. function of Since u satisfies (A . large. where A is a r.t. perhaps. it must be Ae*. to the value of C. makes a rapid (93) above. by (97). &c. The r = case of (193) is .

when x is positive.s. expre. * \vill X = u + Ay + AX + A + the value of X is simply u. Similarly. Moreover. since the square of the expression for e~-x cos nr. and '" .the characteristic. (195) *. or 3 . the characteristic (A In + l)w? = is satisfied by 447 later. the improbability of Q coming to cos rir algebraically for e~ x "cosr Tr is the appear. .siuu ..

(197) Put x = e* z = zi". and the value A's Y = + BX + B 2 W* + B 3 /" + (196) x when x is positive. one. or t~ Iff . negative. or complex.. + Bio'. . . . Say *-* = .. VIII. As the and B's may be various.i74 ELECTROMAGNETIC THEORY. we may have transit ions from Here is X to Y. CH. we do not know what they are. Then we tan nr) get = 2 -(l+i (r . When x is of Y is w.

* Power 443. We may derive a power series for the logarithm of x . fore x negative. both equations are true. equations For with z positive an d there(197) and (199) are consistent. Series for log x. (200) Whether this be a proper transformation or not. be written (198) which splits into equations which may (199) = \ U& rfr/ 2 [r tan r:r. if (197) is true for x But further investigation is negative as well as positiv e.+ B2 (1 + i tan nr) (log z + G(r) + wr + TT tan nr). reserved.

from the binomial theorem .

. In the case r = J. (207) The two formulae (205) and (207) both lead to log i = JiV.-} (x*-x. -y-*}~'.. + x and (1 + or1 ). 2 (205) that is. Differentiating (204) to x. without the zeroth term for r ^= 0.)+^(x -x. 7T T or..1 . 475 When r = this makes -Iogz = 2-cosr7r.)-.. (206) putting z = y 2 . That 2 3 3 1 logx = x-x. loga-ls^ainrar. we .GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. the sum of the logarithms of 1 according to the common formula. (204) is.

. Differentiating (209) to x gives But it is = Zz Equation (206) r sinr7r. l (209) l = case is = (1 + x)~ .get (208) to x Multiply by x and use the result (43) .x~ (l + or )1 1 . that the more general formula (203) goes vanishing when x = 1. (210) 1 Also (208) seems to give a power series for x" . only a particular form of (20 9). (211) I have also verified used numerically. r* we then reduce 0= Z The r p cosrjr.

putting x t zi in (203) we (213) which.may be when r = J.. the error Put y xi in (207). (214) ...s. (212) = Again. when r = 0.. makes the well-known sm3. . We get the well-known --get ..

then we get (216) Do the same again. . CH. 444. (215) Put A1 for x and algebrise with unit operand e-*\\ V*f[ff(r)]*g'(-r-l). In 433 this formula. for the theory of generalised functions is in nubibus as yet. and Rectification. they are only experimental. VIII.476 ELECTROMAGNETIC THEORY. Although these substitutions of imaginary for real values are successful in the special cases chosen. was arrived from the binomial theorem v id the logarithmic function : at l (1 4x)~ =2 *rg(r)<j( -r-1). and Jo(*) we get II . equation (80). Examination of some Apparent Equivalences.

and sometimes gives true equivalen ces.-V(. Multiply by 1+x.r-1). The result is (218) which is . (217) Here it 11 is used to indicate a possible or apparent equivalence it . may. a very speedy way of generating new formulae. or The method is Now then? tested. may not be one. (217) is easily seen to be incorrect. at other times only partial ones. And is not (215) to be suspected? What about It may be (216).

and often turns up as a connecting formula. Numerically can only be used when x is near to 1. therefore u it is considered. series as When x = l.e. however. i. constant as regards x. we have such ('=i) -x-i-i+ -*+~( 221 ) . (220) Aw = 0. equation (43).identically the same as (219) which formula was obtained before. It asserts that e A l = 1.. It satisfies the characteristic 0.

. Now # = 2 makes v= 1-009 with. then x=5 makes v = 0-9993 by a closely. so far. But this x m ay not be small long enough to show an auxiliary satellitic function. 1 x sin rir 112^-^r because it is . if it exis ts.t. . and 1-0009 without the l. to doubt (215). (222) There is DO reason. see on what the satisfac tion of the characteristic of e~ x by the right The result is If the can ba replaced by = r . Again with x = 1. .c. v seems to be 1 very Thus take r=J... sum of about 20 terms. I get 1-063 with and 1-0093 without the l. then (216) will go. Integrating to x.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. (224) though the left side might be numerical testing.t. Passing to (216). 477 (r = J) ^2 = l+^-^ + T ^-^ + member depends.c. a function of r. 1 for a special case.7.

Applying the results are a lit tle less these to (224). nearly per cent. it would s eem that the error rule. However. can say than 1 throughout. passing to x = J.t. = = by the formula (188) above with n 0.t. ^. although now the er ror itself has become large. We when the . would allow. that v is certainly closely equal to 1 save when x is small. by also with which the l. finds the error.c. all through. and x = J. Try with r = J.On The is error rule seems to fail here 1 . with #=2.c. Testing (223) in the same way. y\l) Here = 1-21 690. with the same values of r and x. . the last result size of the l. therefore. and within t he error limits. more than the for example. 1. the value comes out right (within the error limits) Also with # = 3 . is not a complete rule. the rule is satisfied. and r J. ^(J) = 1-10313. Unless. errors in calculation are involved.

t. is But that (224) differentiating it too big lor a certain conclusion. is not a real equivalence is to be seen by r.)r 1 sin . to Thus rS cos + 2.c. u ^-(rr.l.

So v 11 1 is only an apparent equivalence. VIII. this reduces to 2 CH.} 13 Here t.178 ELECTROMAGNETIC THEORY. The characteristic of v is (A . When + -ij 12 (226) jJ-.' is zero only when x = oo ...\v = a + 1)^0 = 0. r = 0. (228) (229) .

] = when x= CD.Yl where C. and a To harmonise with (226).f dx = k + ar\ogx + C-(x-~+. and is nearly any a is over 1. or G(0).. A (280) So v = 1 for 1 when r.. if r is integral and midway between. means the s ame as utilised thus.. The sine function thus reached may also be Equation (228). by (224). . . and a are functions ofr. From the preceding A is 1. we may write is p eriodic. is brought in to make [. (231) Divide by 1 Ti sin rir cos rir.

Numerically tested.Then 2 . we see that the present investigation x leads to the gener alised t~ formula which harmonised the Bessel functions. Passing now to the generalised formula . Determination of the Meaning of a Generalised Bessel Function in terms of and K H . J. 2 ?^L_ = IT cos ?'7r e-* = e-*. 445. sin 2vrr (232) Comparing with (140). the formula (230) goes with 0=1. .

which is 1 when r is = afc + 6B.GENEEALISED DIFFERENTIATION AND DIVERGENT SERIES. (45). in an integral power series was shown to split into two sc ries (even and odd terms). characteristic of I (. moreover. 479 before obtained. It therefore involves I But can K come in at all ? is the convergent I when r = 0. and is the equivalent divergent JH when r-^J. Now when r is arbitrary. equivalent. but is I (#) x periodic and J. equation (53). Or. by the use of the corr esx to an integral power ponding generalised e series. where A is IQ (x) and B the even and odd series. But the series to find to test this numerically. so it was no good function of W . I tried is the equivalent series in odd powers as in (44). we may. to fix a and b. K . W W that W does not involve K at all. and are.r) First test that W satisfies the or identically. which satisfy the characteristi c sepaThis strongly suggests rately. a and b are divergent with wide error limits. in terms of r. or both. reduce But the solution o f the characteristic equation of I series.

Then tak e r = . U W result is then W= Zyi #r). + or . The form of of W is therefore probably I .this The result when x is big. as in the case and c has to be found. increases as x satellite.as the case may be. before. Finding the result of to an integral power series difficu lt to manage. is sensibly I ft W . but falls (numerically) is reduced to 1. r . This difference. slightly below I as x now rises slightly above I As x is reduced and test. decreases.cK . is way. In order to detect the that W take r = J.J. f \. The try another way. is 1 . . reducing Use the ordiuary e* series in (233).

(237) .(234) where // = J. and is convergent. With r = integer + J. a stopping series. so can be closely With r integral.r and 2 the step of r whilst (235) This is calculated. we get I (#). we get a result reducible to W-l(U4 + U_ In general. (234) makes 4) = iH (*). (236) .

W where This makes 4 = 0-8525 U4 + 0-1460 2nr . K ) (239) (240) . VITT.sin Wj = JH . Understand as in with the previous M>r function.JK0> Note that (238) x *. (x). U_!. where that A and B is are functions of r to be found. (238) U r = K H o . 439.480 ELECTROMAGNETIC THEORY. CH. In the case of r = J the result is U U 7i = 0.

subject to the series whilst (240) is a conveniently substituted equivalence. J. the result it namely. &.U_ lr sin 2 JTIT. use of the ordinary is an identity. Using (239) in it. is The satellite for The value of </>(r) W W . numerically for eight values of &c.J(H r P J sin 2nrKo) (*). up to I-.very closely. (242) has only half the mass of that for is U r. (241) have verified 0. The general ident ical formula corresponding to (238) is . W I r = U ir cos" 2 JTTT. in (235) . r..

when operated upon by (DyD r 2 ~D*2y [g(r)] with . The characteristic is (DyD-l)K = 0. implies that the seri es.(r)-[0(Jr) cos **]. if (244) D = d/dy. which satisfies the characteristic as a whole. pendent ways. before referred to. Both the even and the odd powers have to be included to of r The reduction W K to satisfy the characteristic. does not do so in two indeThe expansion (137) above illustrates this. (243) an integral power (involving ) series. On and this is satisfied because the 1) gives odd power r series .

which separately satisfy the split But it seems that this is the exceptiona l case. satellitic The terms in both cases. r = 0. as in (2 41). with is expanded in the form (234). (244) is + (the same). and that there is no double satisfaction when the genera lised * W obeyed. characteristic.= i. become very But then the series . is used. the origin. U important when x nears and W. when into two series. it does Of course. therefore. whilst the even power series gives addition.

446. practically. If we change y to A" obtain .GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. the other two solutions are likely to Similar remarks enter satellitically in the general solution. . r=0 makes the primary. where D = d/dy. but there is some work to be done may apply to investigate fully. and It will be separable therefrom by gi ving special values to r. (245) Some Apparent Equivalences. thus. The apparent equivalence (246) arises from "1 c 1 1 . and y = \x*. =I (a) . both U and and the small correction may be eliminate d by using rently I W . 481 cannot be used for calculation. its three solutions will be al l included in one generalised solution. to higher orders. are appaSo. on account of the largeness of the possible erro r. If the characteristic be of the third order. two values of r .

is One WQ with . but (247) is not. we The characteristic is now (DyDyD-l)w = 0. the left member of (217). w . 10$ \\ w r is an apparent equivalence. All three are included in u. the ratio being I'OOl when y = S.r . the original e quation (246) is a true equiva= lence. Taking r= J. has three independent solutions.and algebrise again. Nevertheless. (Besides r 0. It (248) of them. Going a step farther in the same way to the fo urth order. Increasing y improves the = 3 and above making w\ a little greater than equivalence y . . ^ and f a re probably y= For u* is a little greater than w% when the important values. with unit operand. say <> and comparing ic .) 1 the ratio being about 0-967. r = 0.

the numerical equivalence is so close ii .with Wj.

One more case to show further falling off. CH. 1 .482 as to Still ELECTROMAGNETIC THEORY. make one think w% is it may be an exact one here specially. VIII. Let (251) The . when y = v= though a trifle greater from y = 2 up to Passing on to 8> 2-J(250) there is a falling off in agreement . for as y increases from 1 to 36. the smaller 9. the ratio W$/WQ falls continuously from 1-37 to 0-907.

Various Properties of these and other Divergent Series. Part it . shown to be excellent. There is equality at about 49. Anyhow the tendency shown in the last case is now more marked. that it is suggested to examine the formula which arise fro m the equation (219). The first _. are found by a cursory examination not to fu rnish apparent equivalences. 16 of the coefficient of y^ to that of very large ratio. whereas the succeeding ratios are quite i/-i This may affect the determinat ion of the error in a small. fo r w/w falls from 3-3 to 0'75 as y goes from 1 to 100. Cotangent Formula and Derived Formula for Logarithm. So far 0. in a similar x to A and algebrising. of the types equivalence l sin = 2^r (^~1 m. 447. IV 2X1-rr-) / \ r /sinnr\ n wv = ^ of/sin nr\ n 2r\ / ( rr TTT )> /OKO\ (252) with n integral and greater than 1.. turning way. others. ap parent equivalence (224) above. is the one. But. M. The above formulae produce such good apparent equivalences at the beginning. complete rule. P. at least as regards r = 0and J.peculiarity here (and in the last case less markedly) is the . namely 2 in the series 10$. tho ugh so bad later on.

The original will not be published in Another Place before it In the meantime I have only space left here for is wanted. . I regret that the condensation should tend to reduce to the dry bones of mere formulae.3.

The formula (219) may be written (253) r The companion formula TT is TTT cot A=2 = v. v Just as u comes generalised come from the Thus. so does -*. interest. as there are plenty of nuts to crack. A1 =21 = 2 ! *"" [r I r -r .GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. 483 a few notes in connection with the preceding. (254) and possesses several points of from the generalised V. They may be useful to investigator s. .2 g smrar r TT .

oo This . integration produces a jump.(255) jr_cos?-. We get = 2 of. the second is discontinuous. So in the latter case and e~ A Al an impulse at x +1. . as found below. That there 1. may be understood by c onsidering that e A Al is an impulse at x = . at x= 1.r r (256) TT Now not 1 But when x (254) is exact when z=l. without A. above below is a jump up through the amount A when x is just and a jump down to the same ext ent when x is just A being about 1-895. Why there should be a discontinuity in r.1. Or we may differentiate v to x. is is continuous through is. jc = l. The first the functions u and v d iffer in behaviour. 1. though not in u.

admit that it had. Note convergent functions the r z being complex. by itself. I think.a? means that 2# 2 . the v formula makes . plays the part of an impulz exp ression sive function at the point 2 = 1. any particular meaning. By differentiation to r. functiontheory of theorists also that in would not. of course. r is an impulsive function. although. the It is at = l. (257) and other formulae may .

is Perhaps the most r interesting particular case of v when = 0.be derived. Then there is n2 .

from the left side.=A.. 0<1).. The result is Deduct it -. 8 (#>1). VIII. (258) Remembering the potency of A. (259) or = log (1 -#). this leads to w=#+ K + i# +.logO-1). CH.. (260) This w function is infinite at from -log(l-ar) ..484 ELECTROMAGNETIC THEORY.. an infinite term o/0 on the right side of (254). .

. convergent than the primitive but this does not help as regards close estimation of A. with x between 1 and ..). (261) formulae all lead to about the special care Numerical examination of various cases of the cot and log same value of A when d one from 1-8 to 2. \ = 2(x + JaJ + fe* + 8.ar 1 Jar Jar* -. usually must be taken. To find A pretty clo sely. roughly. But to find it more closely.. < To show A by itself. Some of the derived formulas .. When x 4 [JL(X a=1 1)" .to log It is (about 6 653)..1 it is alternatingly divergent. because the error becomes multiplied in another way. is beyond that it changes where /* is a constant then continuously divergent. are I used the much more w formula (259). we have .

2.. Thus. when regarded as a continuous curve. we have to find the lowest point. say 1*1. The results are very points.c. The regular when attention is paid to certain error. would be large unless a value of x so little greater than 1 were used But there are ce rtain that the labour became prohibitive. If quite ....t. Representing the series dia grammatically thus. considerations allowing of closer work with less labour. such that there are two equal bottom terms. in the series w = x + Ja? 2 + . give x a value between 1 and 2. if estimated by the size of the l.

way. so A is not quite. symmetrical. to get the .. the results will be staggery half of it. or always oneshould plainly count a portion only of of such size as harmonises w ith the case of two if we always count &s. In this right.c. Use this rule requires whether m be integral or fra ctional. the lowest point would be at A. the l.t..t. count three terms and ^ of the next.t. if z = l-3.c. We nih term to equal the n = m. though nearly. excellent results are obtainable by st opping at the end of the first of the two. th when (tt + l) Now l . For example. n = 3J. But there never is symmetry. and the shaded area would represent the value of the series. l.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. choosing valu es of x producing two equal bottom terms. T hen regular equal bottom terms.c. if But x is such that there is but one the full l.

x=l+m~ results arise. as may be seen by the following : X .

The generalised Thus. a similar conflict in the cot formula.486 There is ELECTROMAGNETIC THEORY. then we harmonise the algebra at the ex pense of making the series v be complex The formula for u is satisfied. v. containing A. e* and t~ x formulas . by squaring. r (263) (264) v\ = TT cot + v + j). when x is real. it CH. VIII. By squaring the series u and can be shown that cot rv(v r rir(v r u\ = TT iv+i). but that for v. will not do. we put A = in. If. however.

. find them. but take things as I . But it is not true. The I effective meaning of the series is not t~ x but e^cosrTr. we can show that i. do not explain. when .show a conflict also. ^X + (266) r *. ( ) x2 = 2 series.e. with the generalised manner. r is integral.* In a similar (266) This would be satisfied by e~* = save 2 # r#M.

It is .1. again But there is not necessarily a change o f this kind in passing The w through infinity. and then alternating!^ divergent and calculable.series above runs up to oo at x = 1. For instance. (267) when #<1. U = 1 + Ja?(l represents (!-#)"* . with a constant multiplier introduced. At #=1 it is infinite. </(2r) . convergent down to x= . But it is directly divergent and calculable * The and #(2r + generalised binomial theorem furnishes the expansions of l) which appear above. and fchen runs down on the other side. + }*(! + fc(l + Ja?(.

where which represents (1 a:)*. and when U is A reason can also be given why the 1)"*.1)~*. (x = (1 .#)'U is numerically violated. But ft is not 1. we should say it then represents c(xl)* c is is a real constant. as in the last case. Here there is with the algebra of series But the x<-l. and then beco mes passes through zero when x Considered as a function satisfying a differdirec tly divergent. ential equation. and seems conflict to represent (js. For algebraical relation V though U and #U have the same point of convergence when z>l. i. The negativity can be V = (1 . and is alternatingly divergent when = l.a?)U by algebra. 487 when #>!. for . but J. by rough calculation. V is negative. the easily explained.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. yet when they ar e united to make the V series.

than elementary theorem. 1 . if that be the true It is remarkable that the divergent Bes sel functions value.But point of convergence is shifted to quite another place. and an impressed voltage e act at their formulas of attention. should be more understandable. cases algebraically and of Newton's binomial Three Electrical Examples of Equivalent Convergent and Divergent Series. and the voltage V = Z 2 C= . junction. The following for electrical is method of constructing If two worthy comparison electrical combinations whose resistance operators are Z x and Z 2 be put in sequence. the current there on the Z 2 side is is G = e (Z x + ' 1 Zg)" . both numerically. 448. I do not see why the factor should be J.

result is much more remarkable.+ ^1/^9 (269) Taking various forms of Z l and Z 2 we may obtain interestFor instance. It will do so if the result is That the other way should give an equivalent convergent./Zg)"1 by division. This occurs sometimes. ing resu lts by different ways of algebrisation. That either of the series of operations shoul d give V correctly is remarkable. there are two ways of expansion of the o perator (l + Z. .

and the divergent by Then . series connect them... v. u = a~ lp ~ i . and is convergent.a~^p~ 2 + This It is we and is right initially. though know it should be 1. Say Z l j7i 2 = ap^.a^ + a 2 /) 1 cfip } . But it usually happens that the divergent result is not the full Interesting general ised equivalent of the convergent result. CB VIII. starting at = 0.1 + a~ 3p ~ l * .cr^p. and e = l. The other way makes satisfies (269). Denote the convergent result by w. not immediately evident what V rises to finally.488 ELECTROMAGNETIC THEORY. v = 1 .

b and the solution then. not evident what means but 1 when t = x Here u and v are equivalent when a is +. Let u become w..+ a^p 2 . Equating them. . <"> it satisfies it is (269) . . we obtain the generalised 2 */ But u and v are not e quivalent when a is V = w is To see this put a = . but not V = v. u. let b be + Then use the generalised e*/& 2 in the first part of .. It is This is the solution when a we . ='-(+3*4 This also initially. it is .

see that in passing is negative. even of infinitesimal . When b positi ve is reduced to zero. a is reduced to zero. if we had a negative resistance. Comparing with v. from positive to negative a.oo takes place instantly.oo when or v to 1. and e be inserted there. the extra term 2 y / a2 is added. Then u ning of the cable from is positive. Let a (having resistance and permittance only) of infinite length be earthed at x = Q through a resistance. and u or show the rise of V at the beginwhen the terminal resistance w show the rise of V from to . the to 1 takes place instantly. the rise from to . When from That is. the resistance rise is negative. See The cable electrical interpretation is interesting. 242.


a. instead of a from 1 to oo . by equating them.b. vi z. See 243. Then 1 infinite voltage Here u and v are equivalent again when a is + as we see .. .3 + arSp-4i . v is + or . Thirdly. let Z /Z = a-J.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. V now It is a terminal condens er. (275) rises This comes from u by the generalised exponential.a~ is V is 6 ' ( 2 76) the convergent solution. But when a is and b is + the result is . dividing the other way. let 1 Zj/Z 2 = a? *. that is concerned. resistance. true whether a other result. V falls from 1 to 0. Then u = a~ lp-^ . As a second case. .2^. 489 amount. the finite voltage e would generate an instantly under the circumstances stated.

But the = 1 _ ap 1 + a 2/) 3 .a ' V 1 + a *P* ~ > (277) not equivalent to u whether a be + or . is this. . (278) is being complete both ways. . The reason ....... Let X= the series .. What its value? (279) Now a*.+l+j*+9+pii+. + 1 + ftp'* + by + fcyi + .

(6-ip-i* u' + &-2p~ 3 + b-3 r *l +.. (281) ... .. Therefore X = e c if c = fcl. (280) -^ + 1 + /^i + &y 4. and b is + and the equivalent solutions are .(/ and I find that the sum of every third term beginning anywhere equals J of the t otal.)... u = .

But when a + . when + . (a~ p2 3 + or 2 V+ 6 ) (282) .490 ELECTROMAGNETIC THEORY. Here by p~ n is is to be understood "/[. to save useless work. b is CH.)2(<i^.) .. Vlli.+ a~V" + 3 6 .

. . . when a + 244. relating to Sketch of Theory of Algebrisation of ] (1 -fy/)l..(288) Here 1 + a-y s + a-y 6 + is . The above three cases illustrate the general theory. l (b~ r p. f ). = J{c/ + 2e-'/2< Cos (*/c) (})*}.+ fr-y^ + fc-3p-s' +. . which may be just sket ched here.* is .). (284) so % transformed to (285) This explains the result (36). a terminal inductance. Let U= This is =^-JL. 449.

say n~ l . sary to consider r to be when b is + .(286) the convergent solution . (287) the series to be complete. when r is + ..because the theory is quite similar. X = ~/ where * . . . le t to the equivalent To transform U W -f . (288) the required result. When r is the reciprocal of any integer.. . Then W= -X+ is 1 + bp" + b*p* + by + . It is not neces.

positive to negative. pp. It is to be noted that the definite integral in that method corresponds to the divergent series in The extra term wh ich comes in when a changes from the present one. is. . L. Vol. are discussed problems equivalent to the a bove by the Fourier method with extensions. 153 to 169. But In ELcc. because of the generalised formula. accounted for by an extra results by the two methods are harmony. in full The method. in the Fourier root of the determinant*! equation. Pa../C (289) e* c = 61 /r .

.X. when a is+ both the solutions. -* odd. whilst when n is .J given above. -In any case the matter c is resolved into the evaluation of X. And when n is expressions are cancelled bjr the other %n.COS I-TT + COS TT .CO * TT -f COS ^ir) = 0. Thus.. omitting the constant. there is an extra term. \n generalised b is negative.GENERALISED DIFFERENTIATION AND DIVERGENT SERIES. For when n is even. . (290) It follows that in the physical case. 491 when c* = 0. as in the cases r = J and . X X= ~' (1 . . convergent and divergent. are obtained by division. the generalised expressions occurring n times also annihilate the result. n = 5.

.. c &c... [Prove by the where 1.Putting =1 for ease. The generaldoubles.. are the . thus. Expansion Theorem.. (291) but also for their not only true for r = l. save be special peculiarities it is when when r=lj. then e< f But X=l+. ( X=. So X-7V true it is not true for r integral. and there may and fractional. an integer. thou jh But continuity would seem to show The f ollowing to r=] generally. &c.=r-> is .=n-i(' + + c<*+. ^. + 1+// + p 2r +....r"+p-*+.. true from r = that is 1. c. ised -* when r is formula shows the validity of this doubling. J. exc ept when the doubling makes r = 2. J.) (292) 2 nn th roots of 1.

r ~* = <r~* cos rir. (294) we If X = sum of five correct real formulse.r> I refers to r m/n. Say r = J. * . and this sum comes as required. . case will be easier to follow. m being a smaller integer than n. then A special (293) Hers so 0V * get to e f . p r cos i = cos (t + fair) .

. the = imaginary may be employed t hroughout. r $. Thus.imaginary parts cancel one another at the end.

and I have no doubt a good deal more will be said when when r = m/n. proper mathematicians will thoroughly explore divergent series for physical purposes. (296) as required. are not true. real pa rt thereof. E. CH. Then p e =c e + i + i* + *)" + J(l + * + i + *c r ct r ct .g. 2 4 y . But this volume is . Numerical results are good.. p ^ is not P c-\ but only the There is more to be said on this subject. VIII. Now take X = *t* + 1(1 which is e*. This But the individual results. when process goes r r H complex.492 where c = 1 ELECTROMAGNETIC THEORY i *.

.now full up.

" waves in Maxwell's theory There are no " longitudinal to sound waves. Maxwell to ok good care that there analogous should not be any. yet. if such should be superadded to Ma xwell's theory. should he spoil his work by introducing longi' tudinal waves ? Although there does not.ON COMPRESSIONAL ELECTRIc'oR MAGNETIC WAVES. . indicated the absence of longitudinal waves difficulty in elastic solid theories to get rid of Maxwell's theory went of itself in the directions required. ELECTRIC OR MAGNETIC WAVES. by his invention of the electric current in non-conductors and his electrostatics and was good. seem to exist at present any distinct ev idence of longitudinal waves in reality. and having done a first-class piece of work in producing harmony between kinetics in a philosophical manner. the phenomena of light . he saw that it let it be. Why. is con- . Now. 493 APPENDICES. More over. then. APPENDIX ON COMPRESSIONAL D. and doctrine of the circuitality of the true current. in my opinion. He knew what he was about. them was a they could not even account satisfactorily for the elementary laws of reflection and refraction at the interface of transparent media. care should be taken that the modified or extended theory structed so as to harmonise with Maxwell.

known to furnish a formal analogy which is useful so far as The moving force per it can be followed with advantage. It seems to me to be out of court. unable to see that tha t theory harmonises with Maxwell's. Let us. in the first place. There are many ways in which compressional waves can be introduced. see how the rotational ether The rotational et her is will furnish compressional waves. then.Now. But in a primary theory we are natu rally limited to the simplest ways possible. and I am not aware that anyone has shown tha t it does. there are But I am quite compressional waves in Helmholtz's theory. others complicated. some simple. .

and n. APP. if c is t he density of the ether. if (1) G is the spacial displacement. I. for reasons explained ( 153). without any impressed force. connected with finite instead.. Then F = AV div G is v curl2 G = cG (2) the equation of motion for small motions." Vol.494 unit ELECTROMAGNETIC THEORY. The most convenient form of the analogy for present . to have compressional waves. and. A. v the elastic conshear.v curl2 G. retain A.) ("Electromagnetic Theory. equation (313). Here. D. volume arising from the stress in a medium which opposes finite elastic resistance to shear. put w = 0. compression and rotation is F = w(V2G + JV div G) + AV div G . compression and rotation 145. and keep v stants.

and putting v~ = /*. curl E = /^H.E. -vcurPG l (4) is the second circuital law. expressing that the rotation is elastically resisted.AV div p. the velocity of the ether. so (2) above makes curl H = cpE . first circuital (5) which -1 is the modified . with G. curl Also the same as H. as in Then (3) -vcurlG = H. the 159. becomes by l time differen tiation. electric force.purposes is to compare E.

divj9 E.vVNcurlG. disThe regarding any constant pressure which is inoperative. (6) . Equations (4) and (5) will be as well to see how A affects the flux of energy. curl H. is it the torque. if Acon v/>~ E be considered to be the pressure. other moving force. and Avdiv^E The expansion is means the moving force due to the law. or of the l pressure. space variation of the expansion (or compression). The mere The stress on N PN = NAdivG . convective flux of stored energy the plane being disregarded. arises from the space variation of being the working equations.

J/xH 2 the of compression. The ( (9) where &E . turn (5) to curl H = (k + cp)E . energy of rotation. This makes. (See also 145). multiplie d by the size of E (see 145).div W = EcurlH . after using (3).) the flux of energy representing its ( plane activity is th e negative of the conjugate stress on the (because E is the velocity).HcurlE + div(EA^1 divE) = ErpE + H/^H + A divp^E divE (8) 2 which is correct. equation (310).APP. Carrying this out. D.AV divjp^E. 495 145. the time rate of i ncrease of energy per unit volume. (7) Its convergence should represent represent the flux required. and A( v^E) 2 the energy To show the effect of electric conductivity. since <?E is the kinetic energy. and using the circuital la ws (4). ON COMPRESSIONS ELECTRIC OR MAGNETIC WAVES. (5) we get .

(5) it is to be are propagated identically as observed that circuital E and H in Maxwell's theory. by the introduction of k. frictionally as well as inertially resisted W is unchanged in (7). . produces an ext ra term E/rE in the right side of (8). Further.ether is is the conduction current. then (5) makes But if E which. = or is polar. if c is treated as a constant. is it polar and H remains steady in time and place. if H is polar. translation of the 159). expressing the waste of energy. or divergent without curl. but its convergence. As regard s the general interpretation of (4).

polar E. . That is.means the same as if E= . This mak es longitudinal electric waves. without magnetic force. and u2 = \/c. and the divergence of E.VP. and the associ ated potential. are propagated at speed (A/c)*.

There are the mechanical forces to be Even the existence of electrification explained. and the impossibility is that we need to have E b oth if . E is the velocity of the medium. D constancy of c is impossible in the fluid case.496 ELECTROMAGNETIC THEORY. for an impossibility is involved. So the analogy may be The useful only for small oscillating motions. the cas e is far worse. Sooner or later they have to be given up. Then we have simple longitudinal wave s according to (11). equation (10). is velocity. then a charged body has all to be directions. by a time differentiation. as an ether constant. equivalent. and then the ether gets out of shape. imagined to be continuously emitting fluid in That is. strict it APP. A serious difficulty with all analogies which re present electrical phenomena by bodily motions of an ether is that the motions h ave to be continuous to represent certain steady states. isolated in a die lectric involves the stretching of a If. That is the worst of analogies. for instance. for instance. The equation of conbecause tinuity is The div(cE) + = 0. we have to imagine an imp ressed source of fluid or something On the other hand. The electric force becomes rot ation or proportional thereto. point. But electrica lly there is no reason visible why c (the permittivity of the ether) should not be treated as usual. But there are many other difficulties . takes the form (13) first term must be negligible to reduce to the standard form (11). then stands for the density of t he fluid (with some solid properties imitated by the rotational elasticity). c(12) using which. or waves of compression and expansion which are propagated without change of type if they are plane waves. and we have allowed it to be compressible. We are not bound to follow up th e analogy when we find that it fails.

" A Dynamical Theory of the Luminiferous Medium. Larmor's determined attemp t* to make the rotational ether go.. U. H as the velocity.S. . with J.H circuital charge * ! and polar at the same time roundabout an isolated Dr." Trans. 1885-86. labours under this Larmor.

or (4) ductor. Given any initial state with H What happens in an unbounded nonconducting circuital. 497 have His electronic investigations apparently incurable defect. any specialised mechanical hypothesis about the Let us. the But according to (4). It does not its place. electrificatio n persists in left behind. have done with the analogy and its embarrassments. for keep ence one thing. There are If electrification electric energy shown in (8). which also go out to infinity. (5 ) be the two circuital laws without and (9) in a conspecialised mechanical repre sentation. . whilst the polar part of makes Nothing is longit udinal waves. no resistance to two kinds of is still to be measured by the divergence of the displacement. and let (4). then. th e rotational ether. without ether. but not in terms of Of course Maxwell's theory is dynamical. then we cannot have stationary e lectrification. (5) if electrification has same meaning. Besides that. In Maxwell's scheme. it only persists in the total. to be understood elec tromagnetically. time and place.\ ON COMPBESSIONAL ELECTRIC OB MAGNETIC WAVES. We reduce to Maxwell by A = (or compression in the fluid analogy). uniform medium is that the circuital E and H make Maxwelli an waves which go out to infinity.

+ VQ + . we may as well re member that two fluxes are concerned in Maxwell's scheme.curlE = fipB. The persistof total amount in the whole dielectric. Thu s. re quires us to modify equation (5). in order to show electrification stationary. if we have any at all. there is no present reaso n why we should not have both kinds at the same time.it may increase at some places and decrease is merely one fact that at others equivalently. Moreover. Here f and g are introduced to they are perfectly ar bitrary vectors. and that we can have l ongitudinal waves either of divE or of divH. make We H Maxwellian waves. divD does not persist at any place unless A = 0. and represent sources see at once that circuital E and are impr essed. (15) be the circuital equations. f. whilst we are about it. Thus. To amount a source is required. a nd still have longitudinal waves. let alter this The (14) . if T HH . But. along with circuital f and g.

(20) in (18). using we obtain (21) . Then.A div jr^Ej. Q= -y Q and div /r 'H^ (20) E2 H 2 . APP. . f E2 are the circuital g. Now let P = . show the connection between P. (19).498 ELECTROMAGNETIC THEORY. and H. and similarly for and we have (16) (17) for the circuital parts. D. and polar parts of E. as in Maxwell's system. and (18) (19) for the polar parts.

These H Given any initial state of E and H.(22) showing that polar E when free by longitudinal waves. if fj = curl e. then e may be regarded as impressed electric . we g l = curlh. It produces. may be regarded as the If To have remanent E and H. H waves have no E with them. density of the intrinsic magnetisation. Its effect is known. polar similarly proand w. Intermediately. Nothing is t left behind. g. and no f. we see by (16) that fda. and that . it produces both. a circ uital state of magnetic induction without electric displacement. when steady. is propagated at speed (A/c)* free polar H is 1 Call these speeds u pagated at speed (y//*) kinds of longitudinal waves are qu ite independent of one two The polar E waves have no with them. require f and g respectively. Similarly. the polar E at speed u the polar H at speed w. In an unbounded medium the resu lt is automatic division into three sorts of waves. and the another. the circuital or Maxwellian at speed v = (jj)~*. Or we may regard gl as impressed electric current.

force. when steady. . a state of displacement wit hout magnetic induction. or netic current. It circuital electric as impressed magfj produces ultimately. though both intermediately. or ce as intrinsic electrisation.

AVdiv/>. where E is polar.ON COMPKESSIONAL ELECTEIC OR MAGNETIC WAVES. (23) (18) Bo). stationary plainly.(E2 which is . say = VP . let electrification and its and now we can represent To show this analogue. 499 we coine to f2 and g 2 . o. Lastly. Then 1 becomes (24) = cP E 2 or.

or becomes steady after varying shows that E 2 = E is the final state as sumed. AV^-E.equivalent. (28) and if H is . . H =H 2 a-. r V (H2 2 (27) we have -H ). = P.. anyhow. if = yV divp-'Ho. steady. f2 Similarly. (25) P is the electrostatic potential.). it is E that is to be regarded as impressed arbitrarily. That is. if (26) p is the electrification. so that if it is (25) Here Now.

the final result is H indicating stationary magnetification a polar state of according to . is included. (80) 2 . But in our present system p and o. trification. in which E.steady. turn cp to k + cp. which is specialised to be zero in actual fact. -V It is to 2 Qo/i = o-. and pp to g + pp and (15). is and also of its analogue. lently. H.must be indepen dently given. This shows a striking difference from Maxwell's scheme. or P and Q equivaneed not be steady. f15 g l settle the s tate A knowledge of the elecof induction and displacement.A Vdiv/j-^B . and may be varied. for that div is B a special experimental datum. So we in (14) may write curl(H .E ). if k is t he electric and g the magnetic conductivity. if H =-VQ (29) be observed also that p and a-.h) = (k + q>)E . zero in reality As regards conductivity.

HH .

(31) or.e) = (g + /*p)H . ). . H p. where p and a(33) To illustrate. are arbitrary as well as e and h.y Vdiv^H . if it is e.curl (E . given that there is no E or o-. initially.y V div^H + y Vp-V//x. . in curl (H h) = (k + cp) E . equivalently.A Vdivjp^E + \Vp~lp/c. D. or at any time q. APP.e) = (g + w] H .curl (E . and no or h.H terms of p and o-. (32) .500 ELECTROMAGNETIC THEORY.

from time t = to ^ and then ceases. in fact. it is only valid within the Outside this sphere there is no disturbsphere of radius ut. be imagined to be veloci ty. If. steady point source. so that Maxwell's theory can also be imitate d in this respect. although this is t he solution still. But although we have a scheme which is really Maxwell's. On the boundary is a state analogous to condensation. being proportional to qtlt Between is the steady state according to (34). The price to be paid is To begin with.later. amount then is the displacement at distance r from q finally. yet I hasten to add that I have not the least faith in the ph ysical possibility of the extensions. the impressed term suspicion. The total condensaa longitudinal or normal wav e. temporarily. Similar remarks apply to a magnetic source o-. but their difference remains const ant. The total condensation and the total tion in this If q acts rarefaction increase uniformly with the time. But at time t. and have further added the means of exhi biting stationary and steady electrification if required. wave increases uniformly with the time. and on its inner boundary a rarefactional wave. . boundar y is a condensational wave. with longitudinal waves of E and of H (or either alone) added. ance. then fluid is being steadily transferred from the inner to the outer boundar y of the expanding the waves E shell. but merely then the If it is a solution will express the result of introducing p. the result is On the outer a shell of depth ut19 which goes out to infinity.

too great. and open t o Its only justification would be that it worked out . may representation of electrification by an be regarded as objectionable.

. (17). Let (k + cp) E be denoted by J. First put It is then the A = 0. well. We may by G. (16). y = same as in (32). tively.ON COMPRESSIONAL ELECTRIC OR MAGNETIC WAVES. exhibit the equation of activity in two strikingly different ways. not immediately evident. Now. if there is no conductivity. reducing to Maxwell. examination shows th at the representation of electrification by an impressed term in To the above way c annot be admitted for energetic reasons. 501 and did not lead it is though to embarrassing difficulties. (33). and (g + fip) H They are the electric and magnet ic currents respecThen one form of activity equation is (gE + H) is = Q +i>U + j?T 4 sum div Wr (35) On and the left side exhibited the of the activities of f g. we must examine the activity products. see this.

U. but W = V(E-e)(H-h). But it cannot be proved from these equations that they are the sources of energy.On sources of energy. This is concluded from experimental knowledge. U = cE2 . the time rate of increase of stored e nergy. according to Q = &E2 + #H 2 . and T are the same. and the divergence of the flux of energy. But in electromagnet ics. it is e and h that are the sources of energy. is The appropriate form eJ + hG = Q + P V + P T + div W. T he right side expresses the sum of the waste. (87) where Q. The above form of activity equation suits the rotational ether. the assumption that these vectors indicate the -f and -g are impressed forces. W^VEH. instead of (35). (86) Now it is a consequence of the circuital equations that f and g are the sources of disturbances. of activity equation. though f and g are the sources of disturbances. T = J/*H 2 . and (38) .

but differ Equations (35) in the sources of energ y and its flux. p and <r. and observe the anomalie s pointed out in the latter place. The sources of disturbance are g . .) Now pass to the more general equations (32) (33). varying disturbances E and H. But we cannot con- . fj. for the same entirely (Compare with 70 and 159.(37) are mutually convertible.

for of course the equations are too general for For example. D.JA(diviriE) given by T! = Jyfdivir^H)'. increase of momentum ordinary impressed Newtonian . whilst there is fresh stored 2 . energy W 2. t hen c/?E is the rate of time it.(gE + fH) = Q +pV +pT +p\J 1 + J>T! + div where Q. U and T t are as before. l (41) This form (41) suits the rotational ether. so far at least as it can be applied. We (89) . is APP. if E = velocity.502 ELECTROMAGNETIC THEORY. (40) and W 2 is W 2 = VEH .EX div p~ E . U .Hy div^H. elude from the equations where the energy have the activity equation taken in.

g2E = . EO ultimately. Similar remarks apply to f2 and cr. Thus. ! (42) Apply this to the case of a sphere The This is accounted activity increases uniformly with the time for by the longitud inal wave (here a spherical wave. by (23) and (26). but it will not do in respect t o stationary electrification.). But perhaps the other way will work out better. but practically only E in a sphere of finite radius. Guided by Maxwell's electromagne tics.(per unit vol.gE.). and therefore -g is force (per unit vol. fx and cr are the . although gj and p. sources of disturbances.EAV div^Eo = . with constant charge. This means the rate at which energy is being activity is taken in on the s pot. of activity takes the form (eJj On this understanding. . The above is all very well for consistency. and which ca n never be anything but finite. e and E h and H should be the sources of energy.E\Vp~lp/c. We cannot possibly admit a scheme which requ ires a supply of energy to keep up an electrostatic field. of course) at the boundary of the region occupied by the E set up by p. the equation + hGj) + (E J2 + HoG 2) = (Q + ? U + pT) + P U a + />T2 + div W . that is. and its .

s (43) where Q. whilst the new quantities are Jj and Gj are the circuital p arts of the true . and T are as before. U thus defined.

E ).A.H. W 3 = V(E e) (H . right parts of the true currents.. (g + /^>)H. G . 2 (44) U 2 and T a are the new stored energies given by U = ir!vp. of (30) and (31).)!*. W 3 is given by ) A.ON COMPRESSIONAL ELECTRIC OR MAGNETIC WAVES.(E-E )P 1 a > T^Wdivp-'tH. (45) and.V divjr^H .H ). J\ and Q 1 are the J 2 and G 2 are the polar J2 = Ay div^E . finally. that is 503 currents (k + cp]E and members That is.

y = 0. difficulty. The cure is to a bolish the = Then we come back to longitudinal waves. apparently needed to keep up. Maxwell and his purely . is (46) we have now it got rid of the energetic activities of EO If a stationary electric polar field exists. and the continuance and increase of At present. The no work and H But. by A 0. as conductor. are violated. the former difficulty turns up in a new shape. Superficially considered. but older theories. activity is existent are zero when their fields are established. o. Formerly. however far the region of E may extend. But the the expanding spherical bo undary of the radial E. Thomson's. EO = 0.E div^ 1 (E EO) rXH-Ho). looking closer.. W.(E . consid ering a charged sphere. Thus.= 0.h) . the s eat of activity is the activity were obvious. states by impressed terms in the above way. or /> = 0. = 0. the place of activity was the surface of the sphere. must therefore give up the representatio n of static That is. But when we do this we are no Ho We longer able to have stationary electrification in a nonNot only Maxwell's theory .

inasmuch as leaves the sources. Also. they do nothing. it given by divD. unless if electrification is exists. there conduction or . Then. Good old Maxwell ! When transition takes place is worth notice. persists. even if we do have the above p and <r in The way the action.transverse waves. his electrification cannot be introduced in a dielectric without conduction or convection. convection. for no disturbance This agrees with Max well. A and y are reduced to zero the wav e speeds u and w are also reduced to zero. varying anyhow.

J. as the recent work of the latter testifies. Prof. but an u nusual display of constants. Lastly. that is. had in a go at it. Curry have faith in it. employing the simple methods which are coming Elimin ate the potentials rationalise the into general use. and concluded that it would not do. J. which are very complicated. and it. as a possible extension of certainly Prof. if we look at And . There is not only the usual cartesian complication. not being able to harmon ise it with Maxwell. and put in vector ial language and seek the circuital laws. But Hertz appears to have believed Maxwell's theory . Then we shall see what we shall see. too. and the usual 4ir anomalies. Thomson seemed to be of the same opinion. essentially quite simple. formulae. without success. Larmor. D.* I think this state of dubiety is principally due to the extraordinary complicati on of Helmholtz's investigations. the use of several potential functions. and from another point of view. APP. Boltzmann and Dr. and worst of all. T ransformed in the desired way. Dr. in his Report on Electrical Theories. I make Helmholtz's it . a few remarks about Helmholtz's theory must be added.504 ELECTROMAGNETIC THEORY. I made acquaintanc e with it about 1886. the exhibition of results in the form of equations of electric and magnetic force. being fundamentally in conflict with Maxwell's theory.

p. There is no impress ed magnetic force in the equation.yet the matter is equation of electric force. also to be interpreted. to be explained presen tly. 55. * Curry's "Theory of Electricity and Magnetism. I also omit impressed electric force as unessential. but I have put it in terms of <. and /*. be 2 E = . 643. may be consulted for Helm holtz's investigation. Nov. in Hel mholtz's equation. jl and K is a numeric.pop pot (k + ctf) E -f WQ(! /c)Vpot/> </> -p curl pot /^H. Boltzmann's remarks theieoq. Sept. on the right. See also my review of that work in Tkt Electrician. 10. as in the third term. 1897. . Here understand that pQ and (47) C belong to the ether.V< . 5. 1897. and Dr.. p. in my own notation." 1887. c Q x are the extra parts due to matter. so that /^ + / i = At > and cQ + c l = c. Also < is a potential. There is another potential ^.

but is Maxwell's current in a non-conductor is pcE. notice that instead of cp"E. or a wider theory. It is supposed that so as to give Max well's theory. K in (47). viz. by varying a certain constant contained in <. (47). from Helmholtz's equation of magnetic force. CjE being pancy. by vectorial work we may construct Conversely. pD. except as regards the Similarly. This is a fatal discrenot an inadvertent mistake.pH. or intentional. by curling. without further information. obtain. as in Maxwell. we have c^E. however. the time variation of the polarisation only. the time variation of the displacement. Vol.c two remarks QpV<f>. It is can be adjusted the electric polarisation. (49) First. Next. But Helmhol tz's is c^E.ON COMPRESSIONAL ELECTEIC OR MAGNETIC WAVES. initially. 505 and we get (48) -curlE=>. to be made. that the There are. I. merely allows the electric current to have divergence. < .) from (48). terms. < (Compare 65.. Operate on (47) by curl. anot her circuital law in the form curl we H = (k + c lP) E . potential term. the second circuital law. simply.

div < We get K) pot E + V 2 = . The first . by the use of (50) reduces to (52) 01 ********/W<^ 5 (53) and therefore This gives therefore.frp pot div (k + c^)E + /^c (l V2p2 <f>. Take the divergence (50) c pV 2 <t> = div (k +. The next thing Then to do is to eliminate <.c lP)E. Vp<j> cQVp<{> = l2Ly div E.of (49). (51) which. take the divergence of (47). Again.

--Vdiv^E Wo V div~2E. (55) Now put clfi Q cK = X.A rdiv^-'EV divj^E. law is by (49). (k curlH = + c^)E . Then curlH = (k + Clp) E .(54) circuital in terms of E. (56) .

Th e nearest approach is by K = oo . propagation is (ftci)" 1 . and is ! It is infinite in the ether The speed of only finite in polarisable media. or A = 0. the elastic we constant connected with compression.506 This is ELECTROMAGNETIC THEORY. Take the case of K finite. perhaps the simplest form. but no conductivity. where A. It is K = c//x c A. APP. Then (58) . But no possible legitimate manipulation of K can reduce Helmholtz to Maxwell. D. is . in the rotational analogy.p)E. Comparing with (9) see the interpretation of Helmholtz's numeric K. This comes out of Helmholtz's conception of the current being the time variation of the polarisation instead o f the displacement. Then we make curlH-(fc + c. (57) inconsistent for the reason before mentioned.

being reducible to (48) and (55). (59) These indicate transverse waves at speed (fWi)-* and normal waves at speed (A/c) . then = 0. there would be. Note also that Helmholtz's theory. </> We do not need to have the electrification difficulties before described. theory. See (9). (10). But this is nonsense. When the conductivity difficult. < given to specify the state of the field. is not necessarily a distance action has gone out. would reduce (49) to Maxwell in appearance.curl H = (k + c^)E . in its unreduced state . is finite as in (56). (11) above. remember that even if Helmholt z's theory could be reduced to Maxwell's. Finally. c It is important to note that If it could be. still but of course c the interpretation is more incapable of reduction to Maxwell. making c = c^ cannot be equated to zero. .AV divp^E.

if desired. extended Let to formally include convec tion currents. be the displacement in Maxwell's theory. ledge about atoms and molecules is quite nebulous. Then pu i s the convection current. DISPERSION. The celerity was I wish the paper itself could be as quickly wonder ful. 507 APPENDIX A E. or 20 theories in an hour. a nd vary from one body to another. matical theory which shall agree with the facts. is exceedingly important and exceedingly In one respect only is it easy. July 17. J. or of electromagnetic and material uncertain. short time since Prof. understood as translated. But it is not easy to make a matheThe subject difficult.* u the velocity of p. Thomson in his Rede Lecture and then (The Electrician.DISPERSION. * It is . J. which are rather complicated. 1896) Prof. the case is worse still. Howard's translation of that paper. even The subject At present I only desire to direct attention to some of view. p the density of electrification. obscur ities and inconsistencies 1 find in Helmholtz's theory. Lodge directed attention to Helmholtz's (1893) electromagnetic This was followed next week by theory of dispersion. It is p erfectly easy to make a mathematical theory of dispersion. And Our knowas for a physical theory. The objections are made entirely in an enquiring spirit. and an hypothesis concerning the mutual action of ether and matter. Dr.

added December. did not appear in the original. (The footn otes to this paper. the curl of the magnetic is not implied that there hypothesis.) In any ionic and . because there is no sign of fication.charges volume electri- .we have demands and if vibrations. 1897. or electronic we must assume that the + balance one another in general. which is represented by the text. deserves study fr om several points D and D + />u the true current. any volume electrification. must be highly a fairly good mathematical theory .

If 6 is constant. and the it produces a grea t simplification. and the second is (1) -curlE = fiH if l (2) D = /cE. R) the electric force. If now d can be defined in terms of D.508 ELECTROMAGNETIC THEOEY. APP. Q. Then the first -circuital law curlH = D + pu = D + d. g. E. and it is given that D are constants. we shall have the essentials of a complete system. Now D is the (/. E is = *E and (P. and that K and [Equation (2) and the one preceding (4)]. Let the time integral of pu be is d. h) of Helmholtz's paper (translation). and d is the same in a thicker type. But 1 desire to use my reader will find that usual vector notation (and rational units of course). in a dielectric. we cannot have a theory . force. d = 0E.* Here is the first inconsistency.

(4) when 6 is regarded The (5)] . I cannot clearly understand this. say. later on in the paper.of dispersion. is A homogeneous must be a differential operator of some sort. made to be Ui= fD! +i d*_Dd K (/ /C (3) and if D and d are parallel. where B is th . The density of the electric energy. this reduces to T^-KK-ejB". say U 1 [equation (4)] is dielectric of the common kind the result. in my density of the magnetic energy is as usual [equation 2 if B = //H. although the supposed constant 6 is retained. or in the form (3). either as a constant. And it is so. which makes a secon d inconsistency.

if is way [equation (8)]. July 24. But by a wellsame as known transformation. Howard's translation of Helmholtz 's Paper in The Electrician.e induction. then T = J/xH notation. this is the 2 HB. and the true electric current in the old-fashioned But Helmholtz introduces another in my * notation. sort of energy. called the electromagnetic. say the true current. 1896. which is wholly incomprehensible to me. . 2 AC. since C is These references in square brackets are to Dr. It is defined in terms of the vector potential A.

] and So the 'electromagnetic energy represented by the product J mass of ion x square of convection current. far as I Now. That is a small matter. Finally.DISPERSION. [Equations (7). is the Principle of Least Action. Least Action has no more to do with the matter than the man . and the one between density is What can it be ? twice the magnetic energy density. Obvious ly this makes the kinetic energy x p*. There is also the friction assumed is to act on disturbed there ions. and B the curl of A. (8). which (13). There is also the kinetic energy of disturbed ions. ' may be seen in the " paper (12) itself. as (6). 509 the curl of H.

equivalent .d gets in. (2) above. to begin with. H = I) + d. Next. I cannot understand (6). it obscures and complicates the matter. remark advisedly. I make so much so a s sometimes to lead to serious error. is (5) in the sense before explained.to (6) I understand (5). Lastly.in the moon. we come to the circuital equations [equations One is equivalent to (14)] . It is quite unnecessary. The other H. remembering previous applications of the Principle of Least Ac tion to electromagnetics. curl (13). which is this much clearer without it. I do not see how . so can see. Compare with The Principle of .

sistencies alluded to. (7) (where I change k to h on account of another Jc).Least Action may do it. the relation between D and d is [equations (17)] . with no dispersion. so that 6 is an inverse operat or. Note that if the in d = OE is really constant. as is required in practical theories of dispersion. (6) with d eliminated m ake a simple homogeneous dielectric. But as a matter of fact . the matter is hopeless at present. and there are th e incon. . or else the reckoning of the electric energy as in (4) above. then (5). or the astonishing electromagnetic energy but as no details are given.

Now (7) can be understood broadly without going through the previous obscurities . reducing i t to (2). For (7) is simply an ordinary " eq uation of motion " introduced as an auxiliary to the circuital laws. would give an intelligible dispersion theory. the real force on p is tance. It might work.* Besides that. because of its unintelligibility. given below. take out the d from equation (6). but If we that the auxiliary hypothesis (7). Further examination. and frictional resistance. But I could not say the same if d is to be retained in equation (6). To proceed thus. APP. I should be inclined to assert not E/o.510 ELECTROMAGNETIC THEORY. This is Pass this correction by. E/a + VOB. E. without necessarily entering into details reg arding the nature of the frictional force. though when applied to atomic charges some p rocess of averaging must be gone through. then since (5) is the same as (1). &c. and that there is opposing elastic resistance. raises o r it might not.. It says that the ordinary mechanical force E/o on p produces motion of /o. . doubts as to its physical possibility. construct electromagnetic theories of dispersion rapidly. and inertial resisa reasonable elementary hypothesis. Write the two circuita l laws in this way : curl H = YE.

if fill we But if they can separate. and pu is the additional current. (8) where * Y and Z are operators. we have curl H : the space concerned with electrification. density \ p positive and the same amou nt negative. then E if will do it. making is K.curl E = ZH. Now this equ ation will still hold good ether. the time-differPerhaps the following transition from the continuity of the primary theory to th e discontinuity of the secondary theory may help.. provided these electrifications cannot separate. functions of p. In pure = /cE =D. D+d permittivity .

" and Dr. sho uld be consulted concerning iona and electrons in relation to electromagnetics. Lorentz's " Versuch einer Theorie der elecprocess of averaging. but to a constant elastically amount the two electrifications are merely connected according to the simplest law. if we concentrate the electrifications in numerous detached pairs separated by ether we come to a secondary theory. inasmuch as there is a (7). Larmor's work already referred to. . But . continuous structure assumed. A.the true current as in (1) above. trischen und optischen Er scheinungen in bewegten Korpern. then we have the auxiliary equation So far. The etherial It will be increased by d to a va riable amount in general. and th e equations (1) and (7) can only be understood to result by some H. If the connection involves inertia and friction as well. the theory remains a primary theory.

being only suitable for slow working. simply periodic states. . (9) Electromagnetic Theory. . To make a large scale model. IV. use the lin e integrals of E and H. whose theory I have described in Chap. and magnetic conductivity. If the four constants w ere independent of the frequency.curl E = (B + Lp)H. Z = pp. They may be direct or inverse. the case would be that of a homogeneous dielec tric of constant inductivity. electric conductivity. 511 dielectric. entiator. when we treat circuital laws to the form curlH = (K + Sp)E. . As is well known. permittivity. " But in are f unctions use an effective magnetic conductivity. Y = (& + */>). this fails to account for the facts of dispersion." general they And note that even though there be no magn etic conn2 ductivity in reality.DISPERSION. -2 by the assumption p*= n where n/2ir is the frequency. provided they represent possible electrical arrangements of coils and condensers (elementary). Say. Vol. In a common homogeneous after Maxwell. But Y and Z may be an y operators we choose. we reduce the In any case.. I. we are usually obliged in the above way to of . rational or irrational.

circuital laws. be quite certain beforeha nd that the resulting theory of dispersion will be a consistent one in all parts . I think. containing no electrical impossibilities. and regard the circuit as made up of a number of Z"s in series.V and C. K' the conductance between th em. 222. (10) dx ax express the resultant forms. L' the inductance. with R' the resistance of wires. and all being the frequency concerned. We can concentrate the f our quantities in lumps. with leaks represented S' the permittance. If and the we assume Y' and Z' to represent any real electrical arrangements of coils and condensers. That . They may be interpreted in terms of a generalised t elegraph circuit. for this generalisation of the resulting formulae). 221. we can. per unit length." Vol. for plane waves. L. all effective values at the by the Y"s (see "Electromagnetic Theory. Then.

is one thing. that if we . Another is.

E. -*K = Kp (l + ^\ * 2 For plane waves we get -\E. and do it properly. the telegraph will not . it if work properly. * (11) = ni then _ rfE = . APP. really re presents a possible arrangement. or to the equi valent theory for plane waves in a doubly conducting medium. Eliminate Test Helmholtz's theory in the w ay described. it raises a very serious question whether the theor y of dispersion is an admissible one. neither will the homogeneous medium and. d from (5) and (6) by means dx of (7). we can see whether If it does not. finally.51 2 ELECTROMAGKETIC THEORY. translate a theory of dispersion to the telegraph circuit theory.

2 2 The rate of waste per unit volume is .a + mn*)\ z ) X = (a2 . we see that positive.where Or. if a f /x7m X + //ffX(X .wrc ) we expand the 22 x ftV. the effective may be positive or negative. we have.) The nega tivity of S and and the KE BH L is have . more simply. (15) reciprocal of Z. Comparing with dx dx (9) for is (16) which are the special forms of permittivity the effective electric conductivity plane waves. + (Divide by 2 if the amplitudes are in question. inductivity. likewise the effective effective magnetic conductivity is positive.

It is impossible to be negative B when we construct Y and Z out of really possible combinations of coils and condensers. or in its analogue in a medium.for admissible at certain frequencies. either in the telegraph th eory. because they arise out of the positive waste in Joules' law . Effective resistances and conductances are always positive. but it is not allowable to E negative under a ny real circumstances.

Yis a conductance operator Kp stands for t he conductance operator of the first . and the denominator of the big fraction in Y is the resistance operat or of the other condenser a?/Kp in sequence with the coil JI/K it is here. as in (10) above.DISPERSION. when of . It is the resistanc e operator of the short 1 piece of the circuit.(J-p 2 + mp) is not electrically intelligible. a stupid mistake of my own made R come out negative.e. and Z" in (17) is the correspond ing conductance operator. 513 and its magnetic analogue. V portion of the telegraph circuit. and C. (12). It is unnecessary constants to suit If we consider a short finite to introduce fresh symbols.pp z becomes /m2 it is equivalent to a periodic states. . first we have Say . Helmholtz's system is not impossible.) operator no resistance and negative inductance. V and Now imagine the proper changes made in the in (11).. So fa r as the above goes. So we see that there are two arrangements in parallel. to which Y and Z belong. Then . or from the positivity of the real ultimate electric conductivity and its analogue. whilst the rest. although in (h simply . But let us see what a rrangements the Translate to operators in Helmholfcz's theory lead to. (Dimensions are of no importance Thus only the structure that is in ques tion. we see that the leak Y consists of a condenser without shunt. since + mp/K. . makes a proper and intelligible arrangement. But Z does not. C. I make these remarks because in my first translat ion into the telegraph theory. condenser. H . instead of E. in parallel with another condenser without shunt itself in which is sequence with a resisting coil. I.

. that does not matter. But in general.* * Similar equations of Dr. unlike Y. are not open to this objection to Helmholtz's system made in this Paper. and if the effective inductance is negative.a zp. Larmor. Here. the operator Z~ x is electrically unintelligible. and the other has the resistance + hp + mp*). LL . then. we do get positive resistance. so far as I have examined them. coil of positive resistance and inductance.p represents a coil Y np^ For this reason.one of which is a coil of no resistance (resistance operator pp). in simply periodic states.

= + Sp. and the question what . The type of a L= is ju. This brings us back to Helmholtz's second circuital equation again.d is doing there ? If we altogether.514 ELECTKOMAGNETIC THEORY. APP. as before suggested. we find R = 0. (17). whilst Z is pp. Z = + Lp. E. we do away with the is as before in unintelligi bility. in the simply periodic Then putt ing omit it Y YK R case. X = (a -ww 2 2 wave E-E P or -p .

equation (16). and Q .*sm(^-Q#).LSra 2 ) 2 ) 2 (K + 2 Sra 2 ). L. S refer to uni t volume. The same symbols ar e used here. 221. but cannot . S.(J)*{(R2 + L2 /* 2 )*(K + S>i 2 )* 2 (RK . whether L is positive or negative. (21) as in " Electromagnetic Theory. for all values For the square of (RK .) H Note that is less P and Q -f are always real of the quantities R. K. take the positive value of (L 7i )*. 2 (20) when N/YZ=P + Qi. L. only it is now E and that are concerned. so R. K. than (R 2 L% and positive.LSrc )}*. When R is zero." Vol. L. So the wave s peed n/Q may be zero or infinite.

Now if K dispersion but rently . But that does not seem to me to be the right way or to look at it.2 be negative or imaginary. Having the square of the wave speed negative at certai n frequencies is not desirable. 2 then (22) (23) P Q = (J)*{Ln(K + S 2 + LSn }* )* + S}*. Here L is + radical causes in the case under consideration. It is true that S may be negative. and the square of the wave speed seems to be negat ive. we shall have no trouble . so the outside = also. Say 2 fl 2 R= )* first. and then \/LS te imaginary when L is positive.

it P = 0. and Q = (LS^2 )* appaapparently. Q is imaginary if S is .. We should take the + value of the radical then. this wrong. then negative. . But . no waste and therefore. as is may be within certain frequencies.

. a stationary vibration. the type of the w ave is now c-pz sinn. is always + and In the case (18). (19). is constant. K P = 0. = In fact. if V is the . we by and when S goes on further the zero. and negative. and v infinite when . decrease of n. 515 when S is Q zero.S' say. not the correct interpretation. and by (13). Q remains zero. then if same way and becomes negative. on further decrease of n. It is easier to follow zero and have Q If quite is not quite zero. S become positive then Q becomes positive again. S becomes zero.DISPERSION. L K is Also. I shall be glad to be corrected myself. by (23). proportional to the square of the frequency. it is P that is (LS'n 2 )*. If this is again. When. but P is finite. and P zero.

as compared . one of the vanishings if K S on decrease Then P rises to of takes place by the route . . oo as n goes from the higher to the low er n zero. V.2 = where y = Sn/K. of positive to the same extent as = 0. (25) Or (26) Here X is always + When so. S was now negative. 5=g{(l + !/*) 4 + ?}. The two values . whilst y twice. . 0.wave speed.oo. then the critical frequencies with their values when it is is generally + but may vanish P = Q but in the region between P and Q change pla ces. + oo . In the limit.

=-N/r+^r m (29) iL2 . making (28) the two critical frequencies coincide. when But if possible. V = m(l + 2a w 2 2 ) .2mo(l + a2 )*.1 of ri* making S = are Between these frequencies. S is negative.

why in Helmholtz's theory. or in the form I have substituted for it. The general much the same as in Helmholtz's theory. that is. much or little as manages to get through at all at that The practical speed It m ay be infinitesimally little. It takes time to move them. we imagine a plane electromagnetic sheet travelling through the ether to enter a material dielectric permeated by the ether. frict ional resistances causing waste should be positive. It is far more complicated in reality when we come to simply periodic . E. and the bulk of the disturbance is left behind the first front.516 Therefore. for there is internal reflection and scattering to be considered. as of the sheet in the material dielectric is different. the matter has no influence when the frequency is infinite. is simply the way with t he ions or electrons. when W exceeds the value given in (28). of negative If ELECTROMAGNETIC THEORY. The thin sheet becomes widened by retardation.f To allow of this. so they do not move when no ti . waves (though so much simpler in the mathematics) the values of P and Q are dete rmined by actions proceeding both ways. ultimately tending to establish a stably progressive condition under the i nfluence of a continued simply periodic impressed wave. the region characteristics are S does not exist.* speed. being less. The reason as well as true absorption by the interposed storers of energy. * It seems to me very unlikely that the front of the wave should be unattenuated. it is a matter of first principles that the very front of the disturbance goes on at th e same speed as before. APP.

inertia is associated of light through material bodies. incon ceivable in molecular theories. exception. always distinguish tion of an impulsive wave a nd of a train of waves. comes about from the The distortionl ess circuit is the unique actions proceeding both ways. . The finally resulting wave speed. and do not disturb the passage of a wave train of infinitely rapi d oscillations. Although an impulse will begin by traveling at the speed of light. t According to the above. when a simp ly periodic train of waves has been set up. which is. it w ill be thrown back and redistributed in various ways. A simply periodic train of waves cannot be set u p until all parts of a body are well under influence We can also illustrate this effectively by the telegraph theory in its generalis ed form. according to the structure of Y and Z. save when the action of the matter is such as to produ ce merely an increased permittivity of constant amount. however.me is allowed. we must. the generalised and reacting. in the consideration of the passage between t he propagaThey will in general behave quite differently. for there is then no back action. leakance and resistance operators.

and (7). p. Then the equation of activity is conv VEH = E curl H .H curl E (30) Here ED = U. (2). Take first the simple system cons isting of equations (1). it is Bpu or Fu.DISPERSION. and seeing if it accounts for the work done. F . the rate of increase of the electric energy U = JED rate according to is my if reckoning. HB = T. The remaining That term Ed is. Similarly. the of increase of the magnetic energy. 517 Another way of testing Helmholtz's theory is by forming the equation of activity .

in (32) it to interpret the additional term.is is. and (7). is (33) The curl of for the increase of electrical .d/*)H. is assumed that is the flux of energy. (6). (32) and I am unable to Of course. and by (7) accounted for thus Kp showing that the rate of waste is 7m /3 //c. see how VEH It may be objected that this is not the case. conv VEH = U + T + Fw-H curl d/*. the moving force on the activity of F. just as in (30). but that the flux of energy should be V(E . we obtain l . the potential energy 2 2 is J(aV/*) (p~ u) 2 and the kinetic energy is J(H/o /K)w The activity is therefore fully a ccounted for. But in Helmholtz's system equations (5). But this does not help much. 2 2 . The convergence -5-^(1) + d) + HB.

or for the strange " electromagnetic energy. but (33) does not account energy as reckoned by Helmholtz ." d does not now appear. .

or it in the simplest manner. associated with potential and kinetic energ y. It be something equivalent In Maxwell's electroto eolotropy as regards the density. say the permittivity and the inductivity. and the magnetic force may with the induction. or dyadic. SURFACES BY HOMOGENEOUS STRAIN.* Simplex Eolotropy. or c and /*. magnetic theory the two properties are those connecting the electric force with the displacement. as Willard In either case the optical . Then either c or ju. These are. ON THE TRANSFORMATION OF OPTICAL WAVE. All explanations of double refraction (proximate. in the simplest case.518 ELECTROMAGNETIC THEORY. 1. constants corresponding to isotropy. APPENDIX P. The existence of eolotropy as regards either of them will cause double refraction. is Gibbs a symmetrical linear operator. with which the ether is endowed in order it to account for the transmission of waves through may be elastic eolotropy. not ultimate) rest upon the hypothesis that the medium in which it occurs is so structured as to impart eol otropy to one of the two properties.

In either case the fluxes displacement and induction are perpe ndicular to one another and in a wavefront..wave-surface is of calls it. And in th e other extreme case of electric isotropy and magnetic eolotropy. 20. the electric force being then out of t he wave-front. coiccident with the displacement. Dec. * Proceedings R. and magnetic forces are But it is the magnetic also perpenforce that is in the wave-front. in case of magnetic isotropy and electric eolotropy. though in the plane of the normal and the displacement. coincident with the induction. the Fresnel type. . whilst the electric dicular to one another. 1893. the electric force is in the w ave-front.S.

inasmuch a s the positions of the fluxes in the wave-front are such that the angle through which the plane containing the normal and the displacement (in either wave) must be turned. if existent. we obtain regulating plane waves in media possessing duplex eolotropy. wave-front conditioned by the effective components in that plane of the forces b eing made to coincide with the fluxes. is admita more general kind of wave-sur face. crystals may be strongly eolotropic electrically. to reach the electric force. because the inves2. possesses many points of interest. whilst their magnetic eolotropy. or the flux of energy. a sort of balance of skewness. is equal (though in the opposite sense) to the angle t hrough which the plane containing the normal and the induction must be turned to reach the magnetic force. This. althou gh in the wave-front. There are two waves with a given normal. In general. Properties connected with Duplex Eolotropy. round the normal as axis. and the wave-surfac e associated The chief therewith. It is almost a pity that magnet ic eolotropy should be insensible. 1). which are always in the plane perpendicula r to the ray. the fluxes displacement and induction. p. 2. justifies Maxwell's ascription of double refraction to electric eolotropy. as a matter of fact. When ted. including the former two as extreme cases. when there is a onesided and imperfect development of t he analysis concerned.surface " (Phil. are not coperpendicular. and it would be But there is impossible to satisfy this requirement for both. Corresponding to the two forces electric and magnetic.TRANSFORMATION OF OPTICAL WAVE -SURFACES. Now. is insignificant. " Electrical or Papers. Nor are the positi ons of the fluxes in the this. 519 whilst the magnetic force is out of the wave-front. electric and magnetic. duplex eolotropy. This brings out clearly properties which are not always easily visible in the ca se of simplex eolotropy. as regards the two eolotropies. These are merely rudimentary properties. I have investigated the wave-surface an d associated matters in my paper " On the Wave . Electromagnetic . of course." vol. attraction lies in the perfectly symmetrical manner in which tigation of the conditions the subject may be displayed. are not coperpendicular. though in the plane of the n ormal and the induction.

t June.Mag. 1885. .

of course. Larmor in his paper "On the Singulari ties of the OpticalWavesurface" (Proceedings London Math. be homologous with c. or strain operator. given by c* x constan t. Soc. state the connection . we convert is it to a simplex wave-surface whose one eolotropy . Then. though not the simplest. applied to the simplex wave-surface. let there be electric eolotropy. eolotropies are not merely homologous.. therefore. 1893). AtP. of a specially simplified That occurs wh en the two kind. incident ally. He points out. the principal axes are parallel. surface corresponding to homologous permittivity and inducand homogeneously stra in it. The connection between the simplex and duplex types of wave-surface has been interestingly illustrated lately by Dr. if the s trainer. the result is to turn it into a duplex wave-surface whose two eolotropies are also homologous with the original c . say c. This duplex wave-surface is. the strainer being proportional to c~i. that is to say.520 ELECTROMAGNETIC THEORY. J. becomes To more precisely a duplex wave -surface of a special kind. 3. Conversely. ratio. when subjected to a particular sort of homoge neous strain. that a simplex wave-surface. if we start with the dupl ex wavetivity. p. but are in constant The wave-surface then reduces to a single ellipsoid. with magnetic isotropy. Effects of straining a Duplex Wave Surface. vol. 24.

In another . strainer proportional to results are briefly these Any duplex wave. transformed surface i s of the same type.surface.surface (irrespective of hornology of eolotropies). and some more general transformations pr esented themselves. ft~*. usually remains a duplex wave-surface.homologous with the former two. that the equation of the duplex wave-surface is symmetrical with respect to the two eolotropies. the pure). so that they may be interchanged without altering the surface. though with different inductivity and permittivity operators. In one way the strainer is c~*/[c~i]. But in special cases it becomes a simple x wave. where the square brackets indicate the det erminant of the enclosed operator. Larmor's remarks that a similar reduction to a simplex wave-surface could be effected by a This was verified on examination. when subjected to homogeneous strain (not necessarily : The That is. it Eemembering struck me on reading Dr.

but is surface. first strain the duplex wave-surface to a s implex surface. to Now. The . is practically the generalisation for the duplex wave-surface of Pliicker's theo rem relating to the Fresnel transformation surface. formation. The strainer is complicated. the corresponding indexThe strain is pure. l l This c~ (cfjr )l. strain the latter Thirdly. These indicate the strain operator to be applied to the vector of the old surface to produce that of the new one. these simplex wave-surfaces may be strained anew their reciprocals with respect to the unit sphere. as it involves both c and /z. Furthermore. divided by the determinant of the same. of the same dup lex type.surface to a simplex one. 521 the strainer is /*-*/[/&-*]. Thus. So we have at least four ways of straining any duplex wave.TRANSFORMATION OF OPTICAL WAVE-SURFACES. which are surfaces of the same type. There are at least two sets of three successive strains which effect the desired transto its reciprocal. strain the last to the reciprocal of the original duplex wave-surface. for that also involves straining the wave-surface to its reciprocal. we may employ three successive pure strains. Instead of the single strain above mentioned. any duplex wave-surface may be homogeneously strained to its recipr ocal. Secondly. or the corresponding index-su rfaces.

and Wave-Surface Equations. and apply these equations. then D = cE.investigation follows. to be symmetrical linear operators in general. (1) where c is the permittivity and //. (2) the existence of a plane wave. B = /*H. the inductivity. circuital laws curl if We have also the H = cE. we find that D and B are in we assume .curl E = //H. whose Now. propagated at speed v without change of type. and the Properties of Inversion and Interchangeabiliby of Operators. the displacement and induction. . be D and B. and the corresponding fluxes . 4. Forms cf the Index. unit normal is N. Let the electric and magnetic forces be E and H.

are led directly to the velocity equation. Papers. . and others. and that there are two waves possible. We a quadratic in v z given N. 1 1 Sufficient to say here that c" and /r. Next. APP. 2. 11. if index-surface. I e mploy the vector algebra and notation of the paper referred to." v ol. of course. (" given for a reason that will appear later.522 the wave-front. F. equations (41)). 2 giving the two values of v belongin g to a we put its s = N/v. is then s is the vector of the and equation (3) which are.are the reciprocals of c and p . E and H are out of it. ELECTROMAGNETIC THEORY. equivalent to the velocity equation Two forms are El. p.

[c" ] denotes the determinant* of c"1 that . . and cs . for example. (4) v c 2 . cz 1 be the principal c's positivity of the energy). if referred to the principal axes of scs if c = c^f + c and s lt 2 22 .. 1 . and that scs means the scalar product of s c. to ensure s 2 ss be the components of s. Also.s + c 3 s 32 . (positive scalars.

The expansion of (3) to Cartesian form may be done immediately if c and /z reference axes i. for then we may take the parallel to those of c and /*. (e^Cg)" . j. for our purpose. as linear operators themselves.^s are perpendicular. But it is their For example. * . The operators be written ^jTs^J =v" asserting that the vectors s and [. k are homologous. is in conflict with the ordinary use of square brackets.. which is plainer to read in combination with other symbols than c .. as in (5) and some equati ons near the end. But there will be no confusion on this account in the present paper. and at o nce produce It occurs to me in reading the proof that the use of [c] to denote the determina nt of c.is. in the denominators of (3) may be treated. the first form of (3) may reciprocals t hat occur.

For consider the circuital equations (2). But of the inverse operators in the electromagnetic theory the equivalence can be seen to be true and predict ed beforehand. (8) tively in a dielectric These are the characteristic equations of E and H respecwith duplex eolotropy.TRANSFORMATION OF OPTICAL WAVE-SURFACES. It represents the same surface. Similarly as regards the second form of (3). we apply one of them. 523 where s/*s is as in (4). As regards the second form of (3).1 curl H = juH. with out the use of vectors and linear operators. The transformation from one form to the other. a nd we see that When therethey only differ in the interchange of c and /A. with /* written for c. is very troublesome in the general case. if done by ordinary algebra. (7) we eliminate E. l we obtain -curl pr whilst if curl E = cE. it is obtained from the first form by interchanging /* and c. If we eliminate H. we obtain curl c. When the operators are not homologous. . the complication of the form of the constituents makes the expansion less easy.

a second is got by intercircuital The same property changing the eolotropies. They therefore represent the same H equation.say that of E.surface and from the equation derive the equation of of s and its connection with r. if the E provided p and c be interchanged. The index equation being what we are naturally led to from the characteristic equation. H from the applies to any equation obtained equations with the electrical variables elimina ted. so that rs if = l. in which process E is eliminated. the equation of the wave-surface. to a plane wave to make the velocity equation. surface. we can see that a precisely similar investigation applies to th e fore So. the lead to the second form.surface. it is merely a matter of mathematical work to derive the corresponding wave. equation must equation leads to t he first form in (3). If we have obtained one spec ial form. . (9) r is the vector of the wave. we may . For s is the reciprocal of the p erpendicular upon the tangent plane to the wave-surface. for example.

F. Nevertheless. 2.524 r itself. . two forms are given. I have APP. the final equivalence of this transi tion from the index to wave -equation to mere inversion result that of the two eolotropic operators is such a simple one would think there should be a very simple way of exhibiting how the transition comes about. as before. shown (loc. comparing this with the analogous property sD = = sB. the equation of the wave-surface where. A and caution c. ELECTROMAGNETIC THEORY. symmetrical relations between the forces and fluxes in par. and constructing a process for leading fr om the former to the wave-equation analogous to that leading from the latter to the index equation. It then goes easily. in fact. ticular proving. Thus. pp. first. we are not concerned with these details here. cit. found the transition rather difficult. and. that rE = = rH. vol. Now.. or that the ray is perpendicular to the electric and magnetic forces. and by no means obvious I effected the tr ansformation by taking advantage of at first. 12-16) that the is result is expressed by simply inverting the operators in the index equation. However. I am not aware of an y simple investigation.

let be constant in (10). For example. at relative simplicity of the cost of generality. So we have gained a The . second. because it is more form general. If one of them be a constant initially.is necessary regarding the interchangeability of p They should be fully operative as linear operators.!) The first 'form is what we are naturally led to by initial assumption of constancy of p. We //have now 1 /or* [cjtrc. remembering that [//] is now /* 3 the same interchange in the first form does not produce t he . and therefore all through. Now observe that the interchange of fi and c in th e second form gives us the first form. we may not then interchange them in the s implified equations which result. But after a little reduction.

as we saw for the reason following (7) and This suggests th at some other transtrue.-= is pure. General Transformation of Wave-Surface by Strain. 4. Homogeneous Now apply a homogeneous strain to the wave-surface. 5.TRANSFORMATION OF OPTICAL WAVE-SURFACES. as done in Tait's work. the operations are rather more powerful. the much more general equivalence (10) is. On the other hand.-iq We need not suppose that the strain It becomes first of (10). Use . Though not difficu lt to carry out. may recondite. obviously formations involving the general cubic may be made plainer by generalising it. 525 extra complication of the duplex wave-surface is accompanied by general analytic al extensions which make the working operations in (11) The equivalence of the two forms be established by the use of Hamilton's general cubic equation of a linear operator. (8). Let =. . e mploying a pair of linear operators.

practised worker results may elaboration of coordinates. by mere inspection of equation I give.(12) in the 0. believing that the spread of vector analysis is not encouraged by the quaternionist's practice of leaving out too many of the steps. or with little more. however. <J>~ l q. (13) c Now the use of vectors and linear operators produces such a essentially concise exhibition of the freed from the artificial significant properties. In the first place. much of the detailed work that would then be done silently. . that a readily see his way to the following (13).

^. Also. the first 1 ^-'q.is the same as ti^'"*. in written q</>'~S and the postfactor <f>'may (13) may be then be transferred . if </>' is the conjugate of <. So <#>-vv-'a = a^'-V.*! 1 1 (U) in the denominator.

F. and then brought in as a postfactor. c are any linear operators. Now introduce some simplifications of form. So (13) becomes (16) = 0. of the denominator. APP. To do this. from the second. We be done by the elementary formula (15) where a. may see why this is to and then bringing it in as a prefactor. It follows <&"<' = A. it must be inverted. and by . Similarly the in the numerator may be merged in the denominato r by ^ inversion first. b.526 to ELECTROMAGNETIC THEORY. Let (17) <H' = 6. course.

reduce (16) to ^ (qA-q) =o=a " [b] 5L . which is p ermissible on account of the interchangeability of /A and c. that 1 ^-V. by the elementary formula (abc)' . provided they are self-conjugate.^-^^^')-^ 1 . They are. (17) to a [A] (19). Comparing (20) with (10).1 (15). Consequently (20) represents a duplex wave-surface whose operators are b and A. for. (20) (q^'q) where the second form is got from the first by interchanging A and b. (19) These three. (18) We also have [A] = [>] 2 [</>] . we see that there is identity of form.

(21) follows that (hc(p = (<pc<p ) . we may take the form (following Gibbs) as (28) .it = c'b'a f . is In case the strain a pure rotation. (22) and similarly of </> for the other one.

j. (24) Special Cases of Reduction to a Simplex Wave-Surface. obvi ously.ir + J.TRANSFORMATION OF OPTICAL WAVE-SURFACES. k is one. see. and I. 527 where another set of coperpeni. < = /*-*. J. \b] .-. this makes K <r=I. (25) Thus. Now take some tion of (17). A = l. Then (20) reduces to = = *-A_. that we can reduce special forms of <. by inspecto a constant. dicular unit vectors. 6.kr = Ie + J2/ + K.jr + K. first. & = /iH c/r-J. either of b or We A. For.

surfaces (26). (28) may now be strained to their Thus. a nd put reciprocals. The new (29) This makes 6*p ' ? PIE. a second = c-i < (q^q) showing that the original duplex wave-surface is reduced to a simplex one involv ing eolotropy &. A = c-fyc-^ (27) which reduces (20) to the simplex wave-surface (28) ' involving the eolotropy A. (30) Here the initial and final 6*'s . = 0. way & is = l. take the first of (26). given by (25)..(26) ~tf Similarly.

are identical. since we also have we bring the first of (26) to ^r Now They compare this = 0.may be removed to the denominator. (32) with the second form of the same is (26). except that b now inverted. and.- . Conse.

and therefore by the first. Transformation from Duplex Wave. is < What = c~1 1 1 (cfjT )*. we get the first. Quaternions. p. a pplied to the second form. 1 1 (34) Then we obtain c~ (op . 7. APP." 3rd Ed. quently (32) represents the index-surface corresponding to the wave-surface repr esented by the second of (26). 342). F.to Index-Surface by a Pure Strain.528 ELECTROMAGNETIC THEORY. In a similar manner the strain (29) applied to the second of (26) leads to the reciprocal of the fir st form. with A inverted. by homogeneous strain are equivalent to PJiicker's theorem snowing that the Fres nel wave-surface is its own reciprocal " with respect to a certain ellipsoid (Ta it. These inversions o f simplex wave-surfaces . since they are the sa me. we get the second form with A inverted and.. of greater interest here is the generalisation of Take this property for the dup lex wave-surface itself. In like reciprocal manner the simplex by surface (28) is strained to its Applied to the first form of (28).

c 1 l )*cc~ 1 (c//.c~ ) = fjT~ = c"1 i J (35) (36) . the first of which is obvious.~ )'//. whiJst in the second we make (37) use of iur l = 1 1 (cfjr )< . There are other ways in which this may .~ )* 1 (c/x1 (c//.

. But the first of (40) is the same as the seco nd of (10) with ^ and c inverted.be expressed. by (17) and (35). . (39) self-conjugate. use in (20) brings us to If this is we see. the strain converts the first of (10) to the first of (40). all of which lead <f> to /z<c</> = 1. and the sec ond of (10) to the second of (40). viz. that its * -i ^ c = = q -i ?-^ fi (40) That is.

. Thus.surface to its corresponding index. But we may change the expression for </> to this strain is pure. we have CfjT = C* . which is the reason I have employed two forms. nor in any of the similar forms in (38). In other words.TRANSFORMATION OF OPTICAL WAVE-SURFACES.surface. Observe that the c rossing over from first to second form is an essential part of the demonstration . because we have such a form as will explicitly show 1 its purity. In full. the strain has converted the duplex wave. 529 and the second of (40) is the same as the first of (10) with the same inversions . . in (40) ) is XVO* But to complete the () demonstration it should be shown that = <' in just assumed < to obtain (40). the strainer to be applied to r of the wave-su rface to produce the vector s of the index-surface (or q. Now the purity of this strain equation (2 0) is not obvious in the form (41). C*fJT C* l C~*.

by obvious cancellations. = c-1 (cfjT1 )* = c"* (cVWc" 1 (42) Its This is of the form fafafa. But is pure. where ^ is pure. the right member reducing to the Therefore left by taking the square <j> root. finally. So. is . < conjugate is therefore < 2 ' 62 This reduces to <j> itself if 2 is pure. . and may = c* (cVV) cji-i this be expanded to W (cVWc"*. because it is also of the form 0^0^ where 6l and So our single strain depending on are both pure.identically.

</>i<' 2<i. Substitution of three successive Pure Strains for one. First. This is dry mathematics. Two ways. and < r as equivalent to the three successive strains <h. 8. the strain (43) . if But it is at once endowed with interest the strain < we consider the meaning of the expression oi <f> 2 . < pure.

the strain -* This whose vector converts the simplex surface q to another simplex surface is p. the ~ where A = c"} /xc~i V* 1 -0-p-^-y. in fact.530 ELECTROMAGNETIC THEORY. (45) [A' ] (pAp) Finally. F. APP. and the result This strain (44) is is. same as (33). (46) converts the simplex surface p to a duplex surface s. equation (28). Next. converts the duplex wave-surface to a simplex surface. was done before. the result The strain from r interchangeability of // and c shows that we may also to s by a second set of th ree successive pure . the strain =p being (40). and which is the index-surface corresponding to the wave-surface q. which is the reciprocal of the original duplex wave-surface.

surface (26) the simplex surface (32) duplex surface s. thus. which brings u s to first . ^/^(/xVVftrt This is (47) the same as . (50) Now employ Hamilton's formula . straining the surface r to the simplex then inverting the latter. V = /" Y7 E^/^E'. 9. or and let r = /r'. 1 / (48) 1 (49) so that (48) becomes .strains. In connection with the above transformations it may be worth while to show ho w they work out when applied to the characteristic equation itself of E or H. Th us. take the form (7). and finally straining the last to the Transformation of Characteristic Equation by Strain. -rE = VV a" VVE.cf-*E = V/^vy-i V/-V/ ~IE'.

(51) . .

more conveniently written. and we transform (50) to . [/] .conjugate 531 < operator. 1 2 .TRANSFORMATION OF OPTICAL WAVE -SUB FACES. = f and we (54) ' = /VV(/>-V)WE'x Or. being here any self. Take 1 <j>=f-*. = VrVf-\fr 'I/) WE' obtain x [/-1] <f> (53) 1 .cf -* = VrV^/ WE' x [/] (52) . with . In this use Hamilton's formula again.

we can now make various any pure strainer For example. (57) which should be compared with the other characteristic. ". of H. to get rid of p~l from th e right and substitute c. we have an inversion of operators and also a crossing over from one process . As then. (58) analogous to our transformation from the duplex wave-surface to its reciprocal. / is specialisations. (55. or . Take ^y = c.-< So far./x-1!!' = VVcW'E'. side of (48)./*H = We-1 WH. = VV'VV'E . then f rf-i-\ = J*"" 1 (56) which brings (55) to the form . that which is (8).

when done in terms of y up to the last moment.is The above form to another. (60) VaVbc = b(ca) [c] /xH = cV(VfH) (VcV)eH. 10. in conclusion. c(ab). (61) MM2 . The elementary formula transforms (59) to in vector algebra. Start from the last equ ation (58). Derivation of Index Equation from Characteristic. exhibit how the indexsurface arises from the charac teristic. We Hamilton's formula (51) makes [c]/xH it become (59) = WVcVcH. may also.

let cP/dt instead of is Here VcH Now . It is the same as (cV)H. (63) So far we have merely a changed form of the characteristic. . But the induction /* H is circuital. or.velocity. J ).532 ELECTROMAGNETIC THEORY.H = cV(VeH) ) (62) from which ~ H). or (VoV) -[c] (! / . Therefore.[c>V(VcH). F. taking the divergence of (63). (65) the divergence of cH. (65) only differs from the veloci ty equation (for plane waves) in containing V instead of the unit normal N and a z i> v being the wave. we obtain (64) = V(VcV)c . APP. which is the same. Thus.

t>Vn = is a _ H. however. therefore put v V8 a and NV 3 for V 2 or d?/dzz V in equation 3 .then we shall have where. dt* z V2 specialised. being only for d*/dt We (65). thus making = NV 3 r(NV3cNV8)/x -1 [c] c-1 y 2 V 32 T NV (NV 1 8 3C H). (66) We may now =N cancel out all the 1 .

usually easier and si mpler to work upon the component equations upon which the characteristic is foun ded.V3 's except the last. and the index equation (3) then comes by s = NV-1 But. and we get the and simple. the process cannot be always recom mended on the score of simplicity. Now velocity equation pure . making (67) ["(NcN)/*[c] c-V~ "^(NVgcH). . on the contrary. although the above manipulation of the characte ristic equation has some analytical interest. It is. throw away the operand NV8 cH.

October. J. in fact. whether equal to or exceeding that of light. with finite energy when moving at the speed of light. and since this energy must be d erived from an external source. J. an infinite amount of work must be done. J Dr. I. one positively and the other negatively charged to the same degree. Th omson's conclusion. and a ssert that a single charged body may be moved at any speed. NOTE ON THE MOTION OF A CHARGED BODY AT A SPEED EQUAL TO OE GREATER THAN THAT OF LIGHT. Then the inf inity disappears. F. One easy way of disproving the argument from infinity is to use not one. t . but in a different connection. 1897. a charged body may be moved about anyhow. J in the cases of Thomson and Searle seems to be that since the calculated energy of a charged body infinite when in steady rectilinear motion at the speed of light. Gerald did so also. The argument implied is long ago. Searle remarks at the end of his paper* that it would seem to be impossible to make a charged body move at greater speed than that of light. 1897 ." Chap. Rememb er that it is not a question of whether the mechanical construction of the ether will permit it no thing is known about that but merely one of electromagnetic laws. Prof. Fitzsimilar remark in the same connection. Fleming has lately repeated Prof. an infinite resistance will be experienced. " Recent Researches in Elect ricity and Magnetism. that is . * If they are valid at any Physical Society. Mr. without any infiniteness of the energy or infinite resi stance to motion that. 533 APPENDIX G. G. and there you are. but two bodies. J.. There is a fallacy here. Phil. Mag. But I go much further than that. J. Tho mson! has made a Prof.MOTION OF A CHARGED BODY.

unless there should be infinite collections of energy at a finite distance from the body. To illustrate. . is APP. G. tribution of displacement travelled second position. finally bringing it to steady rectilinear motion Wait a little while.534 speed. there light in the time occupied in shifting it to a is no disturbance of the original distribution of displacement. nothing to prevent speeds of motion greater than light. Where does the energy its . at speed u. This holds for all time. Start with a stationary charged body then move it anyhow. then there ELECTKOMAGNETIC THEOKY. and there is no need for that. so the by energy is always finite. But it will not extend to infinity. and the special disround the instantaneous position of the body will be assumed. even when the speed u is that of light or more. For beyond the distanc e from the first position of the body.

applied moving force. 494). 2. whether there is one body or two. when u = v. But we may get over this difficulty. and the energy is being cone. by means of two charges. I later corroborated it by * But. together with a supplementary distribut ion inside the The pulling back is obvious. the pull-back becomes very pr ominent. p. .) There is a pull-back on the body exe rted by the deformation of the tubes of force their pull on the body is greater behind than in front until the regular steady distribution is fully assumed. and as I suspected in 1888* (" El. the constant exertio n of force is required to maintain the motion. and in any case. I see that I described the electric current as being toward s the charge inside the cone. does not This force tends to vanish when u<v.come from ? resistance as speed is increased. When u>v. This is Papers. as exactly when u before. But it (A single charged body is in question now." Vol. on reference. This should be th e electric displacement. which is fully accounted for by the activity of the . The body meets with Not only that. wasted at a steady rate by the constant growth of the cone at its apex. Thi s cannot be approximated to = v. and away from it on the outside. The displacement for a point char ge is a conical sheet behind the charge. but it requires force to maintain its speed steady until the steady distribution of dis placement appropriate to the speed is assumed.


and the constant pull-back exerted on the body by the displacement. The solution is described in my It is not the same f ormula as "El. 2. statically. when u<v." Vol. Make its being a point charge that is a surface distribution and it will all come plain. and must be postponed. p.MOTION OP A CHARGED BODY. 535 mathematical investigation. [Fully worke d out examples of this theory would be too lengthy here. They arise it from in question. The infinities that are concer ned in it are not essential. 516. as it were . the cone (or other surface) behind the charge. .] The tendency of the displacement to be left behind as the body moves is also the cause of the apparent increase We then calculate of i nertia of the body at slow speeds. in the approximate result it appears that when u J. Papers.

are concerned. 26.varies anyhow resistance to motion (J. is of considerable theoretical also importance. [Feb. It may . superficially considered at the Physical Society too late for The matter . And . the is equivalent to an increase of inertia Thomson). Thomson and othe rs have lately conX-Rays. and this case is no exception. yet there is the electrical tension but that is no apparent pull-back by because of the erroneous assumption made.] (t Cathode Rays " and importance in connection with " for J. that the change of displacement is instantly assumed as the bod y moves. (u/v being always very small). comparable with the speed of light. The moral is don't b e afraid of infinity ! [The above arrived the discussion. Allow for finite v." cluded from experiment that immense speeds of the cha rged particles. 1898. J. but in question is quite visionary.

. we may well believe that increased voltage will produce spee ds exceeding that of light. if they do not exist already. when the speeds are the same (or about the 54 to 59.become of some practical be fully confirmed.] same) as that of light. see vol . ReIf this garding the nature of the waves generated by starting or stopping charges. and so bring in the co nical theory. 1.

II. The plane waves. 422.* by electrical waves in sea water through thing whether. APP. NOTE ON ELECTEICAL WAVES IN SEA WATER. H. or whether both must be counted. the conductance is paramount. . due to Maxwell . enough p.. It is not necessary to investigate the conductance. If its validity it is problem for any parfrom which the waves proceed. at the frequency proposed. 29). APPENDIX To find the attenuation suffered its H. or the p ermittance.536 ELECTROMAGNETIC THEORY. Vol. is suffibe questioned for circuits in general. the first is to ascertain ticular form of circuit attenuating factor for cient. then to take the case of a simply periodic point source in a conducting dielectric (E lectrical Papers.

with the best spread. and is the frequency. But with long waves it is values of k and c pr ocurable.). cit. i is the same. discussion on Mr. (equation (199) where n/'2ir is mittivity. 1897 . " Whitehead's paper The Effect of Sea Water on Induction Telegraphy. where /* is the in ductivity.The attenuating constant loc. disregarding vari ations due to natural spreading. C.] . pretty certain that large. It is thus proved for any circuit of moderate s ize compared attenuator with the wave length from which simply periodic waves This formula must be used in general. c the perfl = The (/zc)~i. as in plane waves. then e~ ni r at distance r from the source.then . k the conductivity. en * f [Read at Physical Society. S." Mr. Whitehead worked out the case of diffusion of waves from a circular simply periodic source in a pure conductor. the conductance is sufficient to make 11 1 Say with common salt solution k = (SO ). June 11. viz.

None is given. Therefore.2*0*0*. whether the specific c/c to have the very large value of 80 or the smaller value effectively concerned with light waves. If. he having inquired regard ing the matter. then. 1316 The 50 metres is reduced to 13-16 metres. we assume We then reduce n^ to ^-(ftr/ifcfi)*. The doubtful point was the conductivity. and there is no reason why the conductivity of sea water should interfere its value is like that assumed above. k~ 1 = 3Gil and f=300. Now Mr. on behalf of Mr. measured in common salt solution by P. we obta in authority for this datum. Zeeman. if . 50 metres is the distance in which the attenuation due to conductivit y is in the ratio 2-718 to 1. I believe. Ayrton at the beginning of last year. These formula and results were communicate d by me to Prof.NOTE ON ELECTRICAL WAVES IN SEA WATER. Zeeman. I had no data. Hertzian waves. / is the f reqneney. ( A) This is practically true. we get . but took the above k from a paper which had just r eached me 10 l from Mr. is But a considerably can be accepted that . of which the attenuation has been even with as in a pure conductor. perhaps. if 537 This is large unless / is large. Whitehead uses k~ = 20 which I presume t here is good is no less than 15 times as great. Evershed. Using it.

greater conductivity the statements which have appeared in the press. therefore there is no means of finding the cause of the failure. that the f ailure of the experiments endeavouring to establish telerequired before it graphic communication with a lightship from the sea bottom was due to the conduc tance of the sea. It seems unlikely theoretically. are correct. . Stevenson has contradicted it So far as I (in Xa ture) from the practical point of view. know no account has been published of th ese experiments. and Mr.

e~ Px sin (nt Q#). 452. Using the notation of " Electromagnetic Theory. P or Q= i =>'"{ ('*)'('+)' so that the solution is V=e Now ~at is.* The connection between the case investigated by Mr. q to express the wave-train simply periodic.538 ELECTROMAGNETIC THEORY. say = e given by due to V at x = 0. . we have V= V -**V . APP. (1) (2) where = (R + Lp)^(K + Sjt?)*. APPENDIX I. I.. Morton of a wave." Vol. then p = ni reduces q to P + Qi. p. When V is sin nt. NOTE ON THE ATTENUATION OF HERTZIAN WAVES ALONG WIRES. I.train aris ing from a damped source and the standard case of an undamped source may be conc isely exhibited thus.

a in the .La. given by 30 that the R = R . the effect of shifting ~ qx .(4) if V be damped.La. change p p sin nt. say is = eQ to ~at to the left is to q. is K = K . that operator in This the same as changing R to R and K to K . (nt (5) wave-train Q .

[Read at Physical Society.V = e <rc*'* sin Q'a?).] . 1898 " Morton' s paper The Propagation of Damped Electrical Oscillations Along Prof. is in the first quadrant." . Nov. but the new R and K may be posi* discussion on Prof. not distortionless did not alter the results to an appreciable extent. and q instead of R and K. 11. Morton sho wed that the fact that the circuit was Parallel Wires. (6) where P' and Q' are the same as P and Q with R and K Of course R and K are posit ive.

the second quadrant. [To be in the second qua drant E /Lw + K /Sw must be negative. I think it must be because the real circumstances do not correspond closely enough to those in the ideal theory. v and F= (* 2v\L I + )-?.**/). (8) As regards the cause of the attenuating coefficient R/2Lv coming out by Dr. under the circumstances of t he experiments Q'-?.] Practically. Bart on's calculations* from his experiments twice as great as when E is calculated b y Lord Eayleigh's formula. and q may be in. for . and E*/2L * sin (nt V~e e ~ .at ax/ t "t . y/ v (7) where K is negligible. 539 live or negative.ATTENUATION OF HERTZIAN WAVES ALONG WIKES. The external resistance of unknown amount is ignored.

tions to Perhaps. 1899 " The Equivalent Resistance and I nductdiscu. * [See Dr.one thing. the circumstances implied The magnetic vibrain it are not those in the experiments. 1899. June 10. . Barton has calculated ADDITION. 1898. viz. its surface is subjec ted to an impulsive vibration lasting only a very minute fraction of a second. moreover. also. that it furnishes an approximation to the real resistance. second. the change made in the f ormula for the resistance of the wires in his = experiment. it is not to be certainly expected that the formula in q uestion is true for millions of vibrations per We can conclude from the experime nts. which is very rapidly damped. again. though.] f [Expansion of a shor t note read at Physical Society. and the resistance is greater than according to Lord Eayleigh's formula.. 27. Nov. When a wavetrain passes any point on a wire. 1898. ance of a Wire TO an Ocillatory Discharge. So there is no definite resi stance. even if it were rigorously true. the terminal reflections involved in Dr. Then. R' (JB/tfi)*. E being the steady and //. Jan. by supposing the impressed resistance. as assumed in the formula. 8. which the wires are subjected are not long-continued and undamped. Barton's paper " Attenuation of Electric Waves Along a Line of Negligible Leakage.ssion on Dr. the inductivity."] . a vibration. But. Barton's paper. Barton's calculations ma y introduce error. Dr." Physical Society. f January 20.

there is only 5J per cent. APP. vibrations to be damped. instead of R. R" as the waves afc But it does not seem likely attenuate by the c. at the wave-front is great enough to cause some The following. and decreases the wave-front. an under-estimate of the effect of the resistance. and finds that if K" is the corrected value. then R" = 1-054 R'. I. because the formula is not valid right up to The resistance i s greater there.damping. of radius b. is equivalent to Dr. It is. impossible that the conductivity of copper is less to vibrations 35 millions per second than to steady currents and that the . inductivity /*. it is not . This R" is to be used in the attenuator c-**/^ increase. In addition. however. . ar e operative. some of which the wires will be enough have been suggested. tha t the total attenuative effect of the variable resistance of as the wave-train advances.540 ELECTROMAGNETIC THEORY. That is. Barton's investigation brought into harmony with the above and simThe resistance operator of a straight round wire plified. voltage leakage. down to (16). tending towards and other causes.

is where h = Jiirpkp. 187). and the tangential electric force at the boundary is E. Vol.. Now. when n(i K) is big. and ni similarly. It is such that in the case of a pair of parallel wires in which the current is C. and C varies = a/n. Pa. t he divergent series for I and Ij may be used. Whe n E = ^-"'sin nt. or p. 63. the potence of p or d/dt is n(i K). II. and k is the conductivity. the equation connecting them is E = (L p + 2Z)C. if K is understood not to be the complete time differentiator. and I =I lt Then where foxg-{(l .and steady resistance R per unit length. where (10) is the inductance of the dielectric (Elec.. p. but L only as regards the simply periodic part of operands.

</ Z = -R'(f+gi). So /=(2/c)-*. 541 Here/ and g K is big tend towards (12) where and R f = L'w = (JRjm)* fl R*-(/+K f)R' f U'n^K'g. Passing to the complete pair with dielectric between.ATTENUATION OF HKKT/IAN WAVES ALONG WIRES. where R" and I/ take the place of R' and L'. (13. If n/27T (16) = 350. and when = (2K)*. and the resultant effect is . /+/<# = 1-054. are 1 when K = 0. we have E = (R" . This is fo r one wire. which obtain when a = 0. and K = about 0-09545.L"a + L"m)C = (R" + L'C where (14) (15) R" = 2(/+^)(JR^) L"-L + 2(^R/rc)V. # = 1-049. then /= 0-953.

Both R" and L" go up to oo with infinite increase of K. and L' and we reduce Z to the form R" + ~L"p. the magnetic energy. and the further multiplication when the currents subside according to ~"# . But only Q a= can we say that the mean Q equals the mean an d the mean T equals the mean JI/C 2 When a is . T not zero. anel practi cally none in the inductance. they differ from R' If V = ZC is the equation of voltage of a combination . 5= 1-054. the corresponding property . (17) and also is VC = Q + T. we have the equation . But practically there is little increase in the resistance under the circumstances supposed. These express the multiplication of the resistance to periodic currents of const ant amplitude.I' = 81-6. " of activity VC = R"C + jL"C 2 2 . the waste and (18) where when R"C 2 . on the understanding that V and C are of the type e~ a sin nt. . which is sensibly L As regards the meaning of R" a nd L".

is that the mea~ Q x ?"* .

I. E at Here h = s(f+gi). (nt sgy) shows the penetration. Since the conduction is superficial. Ey at distance y from the surface in terms of . if s = (2*p&n}* = ? -a -s/y sin ' so (20) E. mean Txc 2*' equals the 2 *. the APP.542 equals the ELECTBOMAGNETIC THEORY.-e-*'E expresses the surface. Thus may (19) E. . the penetration be represented by a plane w ave. mean B"C 2 Xe2a and 2a mean L"C X *.


SEP 1 : 3 K 1 I MllUliC AUG 3 ii>QO LOAN INTFRLIBRARV JIJN 4-5UNIV.14 DAY USE RETURN TO DESK FROM WHICH BORROWED LOAN This book is DEPT. of on the date to which renewed. due on the last date stamped below. Renewed boo ks are subject to immediate recall. OF CALIK. HER LD 21A-60m-7 (G4427slO)476B General Library University of California Berkeley .