]\gt ; The of Electric by the what and complex subject , and to place certain finite features in theoretical relation to one another , that the discussion presented should be judged .
Some of the views expressed have been stated before in other contexts , in a more detailed and complete memoir I hope to refer to these and other theories on the subject .
In this account I mentioned , very iefly , only those pers which I have had actual occasion to use in the present discussion ; but it is hardly ecessary to statue how much such ation must owe to the labours of others who have previously studied the ided pealt It is a } ) duty to the ance which ] been placed at osal , in the execution of the computations necessary for this paper , the Government rant Cittee of the Society and by the by the Earth .
By G. N. , D.Sc .
, Assistant Professor of Pnre Mathenlatics at University College , ( Communicated ) Prof. J. W. Nicholson , F.R.S. Received , 1918 .
) During the last 11- ) year the of determining effect at a distant point of the 's sm face due to a IIertzian cillator e waves of a definite ] has ) )ject of nerous theoretical investigations .
When certain a sumptions of a physical character have been made , the roblem is of a definitely type ; it is in reduced to the problem of finding ) for the sum of a certain complicated of an oscillatory nature ; shall summarise the principal ethods which have been ' dealing with this The method of and is to replace the series by an and then to an approximate for the integral by of the calculus of residues .
The analysis employed by them , though ince wliting this ) , for example , I have noticed that the symmetry of the disturbance variation about the soln1 meridian plane ( S6 ) had been remarked by Bemmelen in 190.3 ( ' Terrestrial bfagnetism , ' vol. 8 , p. 153 ) , who also ( ibid. , vol. 5 , refers to a theory of " " current-vortices .
S11 ) by Sehmidt ( ' Met .
' 1889 , p. 385 ) .
'Palermo Rendicouti , ' vol. 29 , pp. 169-260 ( 1910 ) .
'Phil .
Mag 1910 , passir } Dr. G. N. Watson .
substantially sound , seems to be lacking in rigour in some points of detail ; it , moreover , exceedingly elaborate and their approximations become invalid in the neighbourllood of the antipodes of transmitter .
The method employed Macdonald* is to approximate to the terms of the oscillatory series , and then to replace the modified series by an integral .
rolL the purely lnathematical point of view this procedure seems opeu to question , as the sum of ) series is considerably smaller than the differences eCween the larger terms in the series and the corresponding terms of modified series .
It appears , however , that the sum of the differences is of same order of the of the series , and consequently , , in ysical 1 , the end justifies meatls .
It should be pointed the reason for the chief discrepancy between the form of result and the results obtained by Poin care and Nicholson is that Macdonald the approximation wheIl r is larcre : : but it is obvious from Laplace 's approximation for that this formula cease .
to be a yalid approximation when is large for to be ciable .
The problem has been treated in a directly arith1netical maunel ) whose lnemoir contains a complete raphy .
A completely diH'erent mode of is adopted MarchS follo by ybczynski ; these ators work ab with a definite integral in place of a series ; but it has been pointed out b ) Love that the whole of their analysis is fundamentally unsoun , since the expression which they assume for the netic force bas a line of singularities the atiye half of the axis of harnlonics , and such an assumption , of cour , , incorrect .
The initial .
seenrs to lie in an ' inversion formula\ldquo ; which somewhat resembles Fourier 's integral formula ; the formula in question , according to March , is proved in his dissertation , but I have not succeeded in obtaining a copy .
could also be made as to the manner in which they roximate to the integral ; for instance , while they profess to approximate to an integral involving a Legendre function of the first kind , it seems that they really obtair an .
approximation to an ral in which the function replaced ) the function ' Roy .
Soc. , vol. 90 , pp. , and earlier papers .
, p. 55 .
'Phil .
Trans , vol. 215 , pp. ( 1915 ) .
'Ann .
der Physik , ' vol. 37 , pp. 29-50 ( 1912 ) .
'Ann .
der Physik , ) vol. 41 , pp. ( 1913 ) .
The Diffraction of Electric Waves by thoe Earth .
In view of the discrepancies between the results obtained by different investigators , Dr. van der Pol has asked me to make a further examination of the problem , and to determine the luagnetic force at the { thtipodes of the transmitter .
This paper is the outcome of the analysis , and it ) } ) to show that the physical assumptions do not account for the observed amount of action .
It therefore seems that further physical facts , such as the ionisation of the upper regions of the atmosphere , play a dominating I must not omit to acknowledge Natefully the help which Dr. der Pol has given me in connection with the qical aspects of the problem .
2 .
The essential advance in this paper is closely connected with the fundamental error of March and Rybczynski which was pointed out by Love .
In dealing with an oscillator on the positive half of the axis of harmonics , those writers express a Hertzian function by of , the integration being carried out with to the degree of the Legendl : function ; such an integral a line of singularities the line and is regular the line The fact is that , when harmonics of introduced , the appropriate function to use is not , but ; this fundamental point is SOllewhat obscured by the equation , which holds the functions whose rees are integors .
failure of ence of an integral involving along the line when an oscillator is placed on the positive half of the axis of is strictly analogous to the failure of oence of the series all round the circle on account of the single ularit of the function at the point A simple electrostatic example is afforded by the potential of a unit charge at distance the origin .
The potential near the cvin is where the contour starts from and to after cling the points , 1 , 2 , .
which are poles of the integrand .
On swinging round the contourso as to surround the othet poles of the Cf .
Barnes , 'Lond .
Math. Soc. Proc. ' ( 2 ) , vol. 6 , pp. 141-177 ( 1908 ) .
] ; formula for , valid when is large , has to be used to prove the convergence of the integral .
Dr. G. N Watson .
inf , rand , and evalnating the residues , we find } series for in powers of , valid ) 3 .
We give an outline of the analysis ) leads to the for the tic force , and with the of the calculus of residues we shall transl.orm the series into a apidly convergent series well adapted for lulnerical computation .
The analysis of Poincare leads to an approximation to the dolninant ternl in this series , while the more powerful contour of this paper give the complete series .
The analysis is made compact by using the Hertzian function and Bessel functions of the types employed by March rather than the functions } ) reviously used by analysts .
It is supposed that the ]arth is a homogeneous imperfectly conducting [ qphere of lius snrronnded ) eueous d sending pparalus is Hertzian oscillator at distance from centre of the phere , axis is along a iadius of the bpflere and is taken to be the axis of halmonics .
The of perioLl 2 and the electric and magnetic at any point are the real parts of the ectors Ee and are cted by JMnxwell 's equations are constants of the medium nitting the waves .
The spherical-polar co-ordinates are , the components of are components ; we also write Then the components of and , in the case of symmetry about the axis of harmonics , are expressible in terms of the single Hertzian function by the equations* ; satisfies } vavo ( atio where Now the function due to an oscillator in eneous infinite bpace is function difiers from March 's function ) the factor ; it is Love 's function in case of an oscillator ounded by infinite eneous d ; the presence of the factor in the equations is explained by the equivalence of the operators ) so far as functions of are concerned if eylindrical co-ordinates .
The actio of Electric Waves by the where denotes the distance of from ; and this function is expansible in the forms ( ) , ( ) , according as , where* , the synll ) the second of the two Hankel-Nie ] functions ( which are solnetimes called Bessel metions of the third kind ) .
We now have to take into account the fact that have not the values when as when ; we accordingly denote the of these constants for the interior of the arth by the symbols We assume that the disturbance in the Hertzian function outside the sphere is , and that the disturbed fuhction inside the sphere The riate series for and are , where the coefficients The boundary conditions effectively when .
On substitution we find the value of by .
two liueal equations ; the value and we dednce that the value of just outside the rface of the earth is iven by ( kb ) ' provided that we use of the aCion We first discuss the limiting case in which is ible c * It should be noticed that this function differs from the function which Love denotes by Dr. G. N. Watson .
with ; this assumption is equivalent supposing that the earth is a perfect conductor .
The value of to be considered is lto sinll ) To exalnine the convergence of this we observe that and , while remain fixed , so serieH converges , since Further , since* , it is easy to see that the series obtained by differentiating term-by-term is unil'ormly convergent with to , and , so no difficulties occur in and from so long as 4 .
If we notice that is the residue at of , are led to the of the taken round a snitable contotl Since it is obvions that is an even integral function of when any assigned value quch that from the equation it is evident that is function of when is positive , and the relation hows that is an eveu ctio of The integrand is therefore an odd function of only poles are at the zeros of the functions .
We assume for the moment ( see S5 ) that , function of , has no zeros on the positive half of the real axis or on the imaginary axis , and we take the contour of integration to be formed by a semicircle of large radius , whose centre is at the origin which lies on the right of the imaginary axis , ether with that part of the imaginary axis which joins the * This follows without difficulty by induction from the equation .
The action of Electric by the Earth .
points .
It is goiug to be supposed that through such values that no poles of the rand ever lie on the contour .
Since the integrand is an odd function of , the integral the part of the imaginary axis vanishes .
Next , consider the integral along the semicircle ; so as Laplace 's approximation which is valid when is large ; also the dominant terms of are of which the latter part is igible except near the ends of the semicircle : it follows thal , since the path of integration avoids the infinities of , the value of this fraction on the contour has its modulus comparable with and so , by an application of Jordan 's , the round the senlicircle tends to zero as the adius tends to infinity .
Hence the sum of the residues of the yrand its poles on the the imaginary is is .
If we assume ( see S 6 ) that the zeros of are all mple , and if we call them , we get and so we lind that when is neglected and equation expresses the transformation of the series for the HerCzian function which the object of this paper is to obtain .
It will be found later ( S7 ) that this series converges very rapidly except when is quite small , and thaG term-by-term differentiation with respect to is permissible , since the test for uniformity of convergence is satisfied .
Further , it will be shown in S7 that convergence is uniform with respect to when ; and so , by Abel 's theorem , we may write for in the series when we wish to find the effect due to an oscillator placed on the Earth 's sm face .
The investigation which lies immediately before us consists of an intensive study of the zeros of ( ka ) qua function of The investigation is simplified by the fact that is real but is comllplicate by the fact that is large , so that asymptotic expansions for Bessel functions of large variables have to be employed .
It may be stated here Dr. G. N. Watson .
that for waves is 5 ilom .
is 8000 ; if the wave-length is 10 kilom .
( about Doreatest is practically realisable ) , is 4000 . .
We first prove two ated in S4 nely , does not vanish is ) ositive or when is inary .
To prove them , we write , is wliltell for ) in of , ( use the relation lVhen and positive , ) and ) are conjugate uplex numbers , and so if one vanishes the other also vanishes ; and this is impossible in of the ove relation .
If is a pure , the } ) to is , and this is equal to ; hence if vanishes , so does , and this also impossible .
Hence never vanishes for positive values of ' ; or for purely inary v .
We shall examine of the functions ( kn ) , by means of the method of steepest cents .
On Somnlerfeld 's rals for Bessel of we see that we shall have to study contour integrals in which the integrand is , the contours through stationary points of the electrical problem under consideration , is large and , while is an plex riable .
We write* where are real .
In view of the that asymptotic expansions fail when , we supposc , while assumes all real values ; the requisite cuts in the -plane are effected by this hypothesis .
The stationary points of are the points where assumes all integral values ; we therefore study the contours formed by parts of the curves I .
The curve 011 which I * There is no risk of confusing these variables ( 3 , with the physical constants introduced in S3 .
of Electric Waves by the has a double point at ) and the inclination of one its branches there .
branch will be called , and the other branch the be called 1Ve shall write : it known*that , when , the functions and equal that passes from - to lf is defined by the tion then increases from to as in either direction from Similarly the curve a double point at .
the branch of the curve which has inclination *Cf .
' Camb .
Phil. Soc. Proc vol. 19 , p. 104(1917 ) .
Dr. G. N. Watson .
at that point will be called , and the other branch through this double point will be called We shall write .
Now , if rary continuously from a value for which , the curve nues to from to , until curve changes its form by a rocess of bifurcation ; such a change can , only occur when the curve has a second double point ; and so bifurcation can occur ways*according as the second double point on ( or ) is the point or the point .
We consider the two types of bifurcation in turn .
Fig. 2 The first ) occurs when I ; and it can be shown that if the branch bifurcates , while if the branch bifurfig .
1 the contouls are shown just before bifurcation takes place ; the various cation studied Debye , Miinchen Sitz .
The Diffraction of Electric by the Earth .
cates .
In the latter case the bifurcation does not affect the form of , but in the case , after bifurcation , the curve starts at and , after passing through , ends at , so that the value of has changed abruptly to .
These bifurcations occur when is on the curves shown in fig. 2 by continuous lines starting from the point The second type of bifurcation occurs when I i.e. when is on the curves shown in fig. 2 by continuous lines from .
If the bifurcation does not affect , but if the value of becomes and this is equal to .
Similar arguments apply to , but it seems unnecessary to give them in detail .
The following Tables give the expressions for , in terms of , , for each of the regions of the -plane numbered 1-7 in fig. 2:\mdash ; .
1 , 3 , 4 2 , 6 6 , 7 Throughout the plane the following are the dominant terms of the asymptotic expansions given by Debye when is large and is not small:\mdash ; sin ; and term-by-term differentiations of these expansions are permissible , whence we find that , if and are regarded as the independent variables , , , , , , .
I Dr. G. N. Watson .
It follows that the large zeros of and their derivates are near the curve on which The portions of this curve near which zeros of ( ffj ) and its derivates lie are shown in fig. 2 by broken lines .
Since , we see that the large zeros of ( x ) are given approximately by the equation i.e. where is an integer large compared with ; and the corresponding value of which makes have a simple zero is [ A closer approximation shows that ] .
7 .
We can now discuss the nature of the convergence of the series for obtained in S4 .
It is obvious from the asymptotic expansion of given in S6 that does not vanish when lies in any of the regions of fig. 2 in which is represented by a single series except when is quite small ; and so all the terms of high rank in the series for are comprised in the expression where Also , at the large zeros of , it is found that the real parts of the exponents in the asymp otic expansion of , ( ka ) do not differ appreciably whenl is nearly equal to unity ; and so the general term in the series for is roughly equal to and this is comparable with except when is small .
It is obvious from the last expression that the convergence properties stated at the end of S4 are deducible from the test of Weierstrass for uniformity of convergence .
8 .
In order to obtain results of a numerical nature it is necessary to investigate those terms in the series for for which I ) is comparatively small ; and to do this we shall require approximate formulae for when is small .
The Diffraction of Electric Waves by the Earth .
It has been shown* that , when , if , and if , then is a bounded function of when and is positive .
Bub the fact that this function is bounded is obviously not connected with the fact of being real , but must depend on the fact that , with the exception of , there is no stationary point of , qua function either of or of , which is in the neighbourhood of the curves in the -plane , and in the -plane on which is positive .
Suppose now that we confine our attention to those values of for which .
The values of for which has a second double point are represented by the curves which separate the region 3 from the regions 1 and 2 in fig. 2 .
The values of for which the corresponding curve in the T-plane has a second double point are represented by the real axis on the right of , and the curve passing from to the origin ( shown by the dotted line in .
2 .
The latter curve touches the boundary of the region 1 at , and so , if we exclude from consideration the sectors which are shaded in fig. 2 , then is a bounded function of and so long as Hence , when and the shaded sectors are excluded from consideration , we find that where A is a constant depending only on the angles of the sectors .
Hence we where , the phase of vanishing with ; the value of is when lies in the regions 1 and 2 , and it is when lies in the region 3 , the transition taking place as crosses the dotted curve ( the equation of the dotted curve may be written and , finally , In a similar manner so long as , in addition , lies in the regions 1 and 3 , and is not too close to the boundary which separates them from the region 2 ; and To discuss the approximate value of we examine the function .
'Camb .
Phil. Soc. Proc vol. 19 , pp. 103-110 ( 1917 ) ; in stating the following results , the notation has been modified by taking and ( instead of and ) as independent variables .
Cf .
' Camb .
Phil. Soc. Proc vol. 19 , p. 110 ( 1917 ) .
I 2 Dr. G. N. Watson .
Consider the value of on the boundary of the shaded sector which separates the region 1 from the region 3 .
On the arm of the sector which lies in the region 1 , we have On the arm of the sector which lies in the region 3 , we have Now , on this arm of the sector , , and hence , using the appropriate approximation when is small or ( say less than S or more than 8 ) for and using the fact that it is bounded for intermediate values , we find that , where is an absolute constant .
On the part of the arc which forms part of the boundary of the shaded region , by using the asymptotic expansions of , , we find the same inequality for .
Now is analytic throughout the interior of the shaded region and so its real and imaginary parts have no true maxima or minima in the shaded region .
Hence in the shaded region Hence in the region 1 and in that part of the region 3 for which , by using equations already given for , we find thnt while , in the rest*of the region ( 3 ) , the function has to be replaced by , the constant implied in the symbol being independent of so long as It may be shown by purely formal analysis ( involving a use of Cauchy 's theorem ) that it is ible to differentiate this approximate formula with regard to or ; and we find that oosh sech cosech The function on the right vanishes when vanishes .
The three smallest zeros of this function have been calculated by Macdonald their values are ; and from what is known concerning the zeros of Bessel functions of the first kind of high order , it is in the highest degree improbable that , when has any The shaded area near the real axis has to be omitted , but it can easily be discussed in a similar manner .
'Roy .
Soc. Proc , vol. 90 , p. 54 ( 1914 ) .
The of Electric Waves by the Earth .
assigned value exceeding 4000 ( this is the smallest value of which is of practical importance ) , the zeros of , qua function of , differ from these values by more than about 1 per cent. at most .
The three most important values of are therefore where , 2577 , 383 Also , to a first approximation , Hence each of the dominant terms in the series for the Hertzian function is given by the formula If be the wave-length ( measured in kilometres ) these terms may be written approximately ; and takes the values Hence by Laplace 's approximation when is not nearly equal to or and by Mehler 's approximation*when is neariy equal to , we find that the order of magnitude of the dominant term is , and , compared with this , the other terms are negligible .
This result is in substantial agreement with the approximations obtained by Nicholson and Macdonald .
It would be easy to construct a table of values of the dominant term for various wave-lengths .
It is to be observed that in this theoly there is not a " " focus\ldquo ; at the antipodes of the oscillator ; in fact the magnetic force vanishes at For purposes of comparison it should be noticed that if the arth were replaced by dielectric the Hertzian function at any point of the Earth 's surface would be : 9 .
We shall now discuss the value of the Hertzian function on the hypothesis that the earth is not a perfect conductor .
*Cf .
a forthcoming paper by the present writer in the 'Messenger of Mathematics .
' The Diffraction of Electric by the Earth .
The contour integral to be discussed is The sum of the residues at the poles , , is ; the integral along the large semicircle vanishes as before ; and so we find that ( .
where denotes , and the summation applies to the zeros of for which Now the values of the constants and are where is the permeability , the dielectric constant , the conductivity , and is the velocity of light , all measured in rational The values of the constants are as follows:\mdash ; For air , For sea-water , For dry earth , It follows that for sea-water at least , and that for dry earth varies between and , while is These values indicate that differs little from unity in the case of sea-water when is nearly equal to , while , even in the case of dry earth , the difference is not great .
It appears that the series for is not much affected .
We have , however , to consider the integral , which may be written in the form Now Now when is a pure imaginary and less than , this is approximately Note on Effect of Wind Pressure on Pitch of Organ Pipes .
99 The approximate value of when is and the exponent is by using the complete system of asymptotic expansions when , we find that the integral is convergent and is of order of magnitude .
It follows by easy arithmetic that the integral is negligible compared with the series .
The effect of the waves which pass through the Earth is therefore trivial compared with that of the waves which travel round the Earth .
10 .
If we take the series for and expand into the series 2 .
and multiply the result by , which are the exponential factors in Laplace 's approximation for , the separate terms in the product denote the disturbances produced by waves travelling from the oscillator in either direction round the Earth after passing completely round , 1 , 2 , .
times .
There seems to be no similar significance for the separate terms corresponding to different values of Note on the of Wind Pressure on the Pitch of Organ Pipes .
By A. MALLOCK , F.B.S. ( Received May 30 , 1918 .
) In books on acoustics the pitch of organ pipes is treated as being dependent solely on the length of the pipe , while at the same time it is recognised that , as a consequence of the conditions existing at the open end or ends , the length in question is somewhat greater than that of the pipe itself .
The " " correction for open ends\ldquo ; has been calcuJatsd for particular cases , and for the open end of a pipe of circular section amounts to an addition to its length of about times the radius .
The conditions at the base , or speaking end , are more complex , and for this ( as far as I am aware ) no similar correction has been worked out .
For these reasons the natural pitch of an organ pipe cannot be accurately determined from its material dimensions , but only by experiment ; such , for instance , as noting the frequency of an exterior source of sound which produces the maximum resonance in the pipe .
When an organ pipe is made to ' speak\ldquo ; in the ordinary way , it is a matter of common knowledge that the pitch of note produced is to some

