]\gt ; [ 113 ] IV .
The Di.ffraction of Electric Waves Round a fectly Reflecting Obstacte .
By H. M. MACDONALD , F.R.S. , Professor of Mthemtics in the University of A berdeen .
Received August 13 , \mdash ; Read November 4 , 1909 .
IN a previous communication*it was verified that the effect at a point on a perfectly conducting sphere due to a Hertzian oscillator near to its surface was negligible in comparison with the effect that would have been produced at that point but for the presence of the sphere , when the point is at some distance from the oscillator and the radius of the sphere is large compared with the wave length of the oscillations .
what follows it is proposed to find the effect at all points produced by a Hertzian oscillator placed outside a conducting sphere whose radius is compared with the wave length of the oscillations .
For simplicity the axis of the oscillator will be assumed to pass through the centre of the sphere , but this assumption will not affect the generality of most of the results .
An Appendix is added in which the more tant mathematical relations required are established .
1 .
Let be the cenre of the conducting sphere of radius , and let the oscillator be at a point , the direction of the axis of the oscillator being , and the distance being .
In this case the lines of magnetic force are circles which have the line for common axis .
If denotes the magnetic force at any point the distance of from , and the distance of from the plane through perpendicular to satisfies the differential equation where is the wave length of the oscillations .
Transforming to polar co-ordinates , where is the distance OP and the angle and ; hence , writing , the differential equation becomes ( 1 ) .
' Roy .
Soc. Proc vol. 72 ( 1904 ) , p. 59 .
VOL. CCX .
PROF .
H. .
hIACDONALD ON THI ACTION OF The general tion of this equation is ?
where is BESSEL'S function of order , and is the zonal harmonic of order If is the force at the point due to the oscillator , is the real part of , where V is the velocity of radiation and is the distance ; this has to be ressed in the form given above for the solution of equation ( 1 ) .
Writing where it follows that ' where , and \ldquo ; when , where Now therefore , that is or ( 2 ) .
CTRIC WAVES ROUND RFECTLY RTING OBSTACLE .
Similarly , when .
The solution of equation ( 1 ) that is required is that solution which becomes infinite in the same way as at the point and for which vanishes when ; the real part of will be the required magnetic force at any point , for .
will vanish when , that is , the electric force tangential to the sphere will vanish .
From the expression ( 2 ) for it follows that the required solution is of the form where , and the constants are determined so that when , hence when ( 4 ) , and . . . . . .
Writing \ldquo ; , , \amp ; c. , it follows that that is Again PROF .
H. M. MACDONALD ON DIFFRACTION OF now , that is therefore and writing , this becomes Similarly , and therefore Hence the relations ( 4 ) and ( 5 ) become when , that is .
( 6 ) , when , and . . .
( 7 ) , when 2 .
The value of at any poin will now be compared with the value of at the same point , and , the radius of the sphere being supposed to be great compared with the wave-length of the oscillations , it will be sufficient to compare the principal parts of and .
Now that is ; hence the principal part of is given by In calculating the principal part of it is convenient to consider first the contribution of the terms for which is greater the least of the two quantities ROUNI ) FFCTLY PCTlNG OBSTACLE .
and .
In this case the approximations to be used for and are those given by relations of the Appendix ; from these that is where is a large negative quantity .
Hence that is and Now therefore and ; hence the order of the corresponding term in , compared with the principal part of , is , and is multiplied by an exponential with negative exponent .
Further , the portion of the series containing these terms is simply oscillatory on account of and therefore the sum of any number of these terms is not of higher order than by the appendix ; hence the part contributed by the ternls of the series , for which is of the same or higher order than , is negligible compared with Again , the terms for which is of the same or higher order than , but is of lower order than , are at most of the order , and therefore , as above , their sum is negligible in comparison with .
Similar results hold when .
Hence the part of , if any , which is of the same order as , is contributed terms for which is less than or exceeds by a quantity of lower order than when , and by terms for which is less than or exceeds by a quantity oflower order than when .
Different treatment is necessary according to the form of approximate value of that is appropriate .
3 .
When is small , the approximate value of is give by whence to the same order and the series for is approximately when PROF .
if .
M. MACDONALD ON THE DIFFRACTION OF Writing the appropriate approximations for , and are , where , where hence approximately that the oscillations of depend on , the principal part of , arises from the terms in the neighbourhood of the term for which vanishes ; for this is small since is small , and therefore , unless is nearly equal to , which means that the point is close to the oscillator , is small compared with for the terlns that contribute the principal part of .
Hence approximately and the principal part of is equal to the principal part of ; that is , the principal part of is given by ; tppendix , Relations ( ii ) .
ELECTRIC WAVES ROUND A PERFECTL CTING OBSTACLE .
Now the distance from the oscillator to the point is giyen by whence retaining only the most important terms and therefore The same result follows when .
Again writing the principal part of arises from the terms in the neighbourhood of the term for which vanishes .
When is less than , this requires that is small , since is ; when is greater than , it requires that is small , that is and are each nearly a right angle , which means that the point is close to the oscillator , thus the case of small need only be considered .
In this case the part of is equal to the principal part of that is whence or This represents waves reflected from the neighbourhood of the points on the sphere , which are such that makes an angle with the radius OQ , and if is the distance the distance PQ , where is the point , the above becomes , after some reduction , the result that would have been obtained by elementary lnethods .
Hence the effect at a point for which is small is the sum of the direct waves the waves reflected at the surface of the sphere .
4 .
When is nearly equal to the appropriate approximation for is given by or , writing and ; hence sin when , where to the order the terms for which exceeds by a quantity of the same or of higher order than can be ected .
Writing the terms of this series , for which is less than , and is of higher order than , are at most of the order , and therefore their sum will be negligible in comparison with unless vanishes or is a multiple of for some value of in the series .
Using the appropriate approximations for and , this requires that is small or nearly equal to a multiple of is small , and this is impossible , for and are each less than .
The terms of the series , for which is of lower order than , are at most of the order , and their sum will be negligible in comparison with unless is very small or nearly equal to a multiple of for a value of in this series ; substituting the appropriate approximations for and this requires that is small , which is impossible , for is less than Hence is negligible in comparison with .
The same result follows when Again , writing the terms of this series , for which is less than , and is of the same or of higher order than , are at most of the order , and therefore their sum will be negligible in comparison with unless vanishes , or is nearly equal to a multiple of for some value of in this series .
ELECTRIC WAVES ROUND A PERFECTLY REFLECTIN OBSTACLE .
Substituting for , \amp ; c. , their approximate values , this requires that is small , since is small , and this is impossible for ; hence the sum of these terms is negligible in comparison with .
The terms of the series for which is of lower order than are of the order , and their sum is of highest order when is small , or nearly equal to a multiple of , for a value of in this series .
In this case the approximation for is [ Appendix ] , and the aboye condition becomes that is small , or nearly a multiple of , which is satisfied if both and are small .
The sum of these then is sin retainin only the most important part , and the sum is therefore of the order , which is of lower rder than , and therefore negligible in comparison with it .
When is greater than and is of the same or of higher order than , but is less than and is of the same or of higher order than , when , the terms are of the order , and their sum will be of the same order as , if is small , or nearly a multiple of , for a value of in this series .
This condition is satisfied if there is a value of in the series for which is small , or nearly a multiple of , which requires that both and should be small .
For the other values of which have to be taken account of , the order of the terms is , and the condition that their sum should be of the same order as is , that for some of these values of is small , which is impossible .
Hence the part of the series which is of the order of arises from the terms beginning with a term for which is equal to , where A is a positive quantity of the order of unity , and the principal part of is therefore equal to the principal part of Now that is Appendix ] where is a negative quantity and increases rapidly with .
Also , when VOL. CCX.\mdash ; A. and are so great compared with that and can be replaced by and the principal part of equal to the principal part of which is equal to When is very small the value of the above integral is approximately which is equal to the value of at the point .
These results correspond exactly with those obtained by the usual treatment of the effect of an opaque circular screen interposed between a source of light and the point of observation , when the source and the point of observation are both at distances from the screen great compared with the radius of the screen .
It has been assumed in the above approximation that is a small quantity whose square is negligible ; when this condition is not satisfied the value of the integral involved depends on quantity where , and diminishes rapidly as this quantity increases .
5 .
When is not small or nearly equal to can be replaced by its approximate value whence and therefore , when approximately , that is which may be written ELECTRIC WAVES ROUND A RFECTLY R OBSTACLL .
where Since , the formulae ( ii .
) of the Appendix can used for the terms of and , for which is less than and is of the same or higher ordel than and the corresponding parts of and are given by where and A is of the order of unity .
Now and cannot be a small qua1ltity for , since ; hence is at most of the order , and therefore negligible in with Again and will vanish for a value of between 1 and provided the point lies inside the sphere on as diameter , but not close to its boundary .
If and now correspond to the value of for which is the angle and the angle in the figure , * , and the principal part of is given by where is large , that is * The corresponding value of is given by , where is the perpendiculnr from on If , then and therefore that is for the points defined above The same formulae ( ii .
) of the Appendix are applicable to the terms of the series and , for which is less than and is of the same or higher order than .
The corresponding parts of the series and are given by a where .
Now and for for , hence is always finite , and therefore , as above , the sum of the terms is order than .
Also and will vanish for a value of between 1 and , provided the point satisfies conditions to he determined .
For let , where , and let be a straight line through at a distance from , meeting the sphere in ( fig. 2 ) , then the angle is , where and the angle is , where .
Hence , drawing QP in the plane making the angle TQP equal to , and taking any point on , the angle is equal to , where , and the angle which is , is equal to .
For any position of on the line QP the terms of the series in the neighFig .
2 .
bourhood of terms for which contribute a sum which is of higher order than any of them , and the principal part of is given by ELECTRIC WAVES ROUND A PERFECTLY REFLECTING OBSTACLE .
whence that is ' , which after some reduction can be put in the form and this represents the effect at of the waves reflected from the sphere at Q. * From the diagram it appears that the points for which the condition can be satisfied are the points outside the tangent cone drawn from to the sphere ( that is the points outside the geometrical shadow ) , and as approaches the surface of the cone the value of approaches ; hence the above result holds for all points outside the tangent cone and not close to its boundary .
omitting for the present the discussion of the terms of the series and for * This resuIt can be verified by elementary methods as follows : If is the amplitude of the magnetic force at the amplitude of the magnetic force at in the waves reflected from , the intensity at in a beam of rays from of small cross section is and the intensity in the same beam at is , and these intensities are equal , therefore ; now whence The amplitude of the of the waves from at is Ma , and of the reflected waves at is , therefore the of the reflected waves at is which is the result obtained from the series .
which is of lower order than the terms for which is of the same or of higher order than will now be considered .
Denoting these terms of the series and by and , and remembering that , \amp ; .
, where is a negative real quantity increasing rapidly with and are to the required order of approximation given by a The value of is , which , since and are less than cannot be very small , and therefore the sum of the terms is of lower order than .
Again the value of is , and , if is a point outside the geometrical shadow for which the angle is less than a right angle , there is a value of in the series for which vanishes , , that given by where is the perpendicular from on , and the principal part of is given by where that is by or where is the distance .
Hence The terms of the series for which is oflower order than and is of the same or higher order than are of the order , and , that their It will be proved below that these terms are important only in the neighbourhood of the geometrical shadow .
ELECTRIC ways ROUND A PERFECTLy REFLECTING OBSTACLE .
sum should be of the order of must be small , that is , the point must be on or close to the boundary of the sphere described on as diameter .
In these terms of the series and can be replaced by , and the sum of the terms of the series required is the same as the sum of the corresponding terms of the series Now , for the values of specified above , can only be small when is a right angle or differs from a right angle by a small quantity , hence the principal part of the sum of the terms of the series is the same as the principal part of , that is , it is equal to 6 .
When and are interchanged , the value of may be written and the discussion of the series is identical with that of the previous case .
If a plane be drawn through perpendicular to , then for points on the side of this plane remote from the sphere and not close to the plane the principal part of is ; for points outside the geometrical shadow and not close to its boundary the terms of the series , for which is less than , have a sum of the same order as , which represents the effect of reflected waves ; for points outside the geometrical shadow between the plane through and the sphere centre , radius , the terms of the series , for which is greater than , have a sum whose principal part is and for points in the neighbourhood of the plane through the terms of , for which is of lower order than , have a sum whose principal part is Thus for all points outside the geometrical shadow and not close to its boundary the principal part of the value of obtained from the series is the same as that which would be given by applying the methods of geometrical optics .
7 .
The effect of the terms of the series and , beginning with those for which is of the same or of lower order than , has now to be obtained .
The corresponding part of may be written now for values of greater than it follows , from the corresponding expression for [ Appendix ( yi .
) ] , that the factor ) does not oscillate but approaches the value as increases , hence for these values of it may be written .
Writing and remembering that it follows that ' where differs from unity by a small quantity for all values of greater than .
and is a negative quantity increasing numerically with ; whence and In the series writing , since , the coefficient of in is of the order ; hence , if the value of where is of higher order than , the sum of this double series is of lower order than its first term multiplied by , and therefore to the order required negligible .
Again , in the series the factor oscillates and , writing [ see Appendix ( vi .
) ] , where and therefore the sum of this series is of lower order than its last term multiplied by when the value of ELECTRIC ways ROUND -A PERFECTLY REFLECTING OBSTACLE .
for is of higher order than .
Hence the principal part of is equal to the principal part of when the value of for is of higher order than Now ; the principal part of the second series has been found above , and arises from the terms on both sides of the terun for which is the corresponding value of , and is given by hence , writing , the principal part of the above series is Therefore , if denotes the distance from to the point , and making the substitution * If is an angle of higher order than the semivertical angle of the tangent cone to the sphere from , and the tangent cones with vertices on whose semivertical angles are are constructed , the space excluded is that between these two tangent cones , which are both close to the boundary of the geometrical shadow .
VOL. CCX .
the above expression becomes where hence In the corresponding part of has the opposite sign and therefore the sum is negligible in comparison with If is the perpendicular from on , and when is greater than the point is outside the boundary of the geometrical shadow , is negative and increases rapidly as increases , and the above value tends to the value found in the case where ttJe point was supposed not close to the boundary .
When is less than the point is inside the boundary of the geometrical shadow , and the above value of is the principal part of , that is where has the value given abov This result expresses the ratio of the magnetic force at a point inside the boundary of the geometrical shadow to the magnetic force due to the oscillator alone in terms of one quantity .
For values of greater than 5 integration by parts will give a sufficient approximation , for values of less than 5 it is convenient to write the above in the form where and use tables for these .
An important particular case is that for which and are both close to the surface of the sphere , as in the case of wireless raphy .
The tables given below show how the amplitude of the oscillations diminishes as the distance along the earth 's surface from the transmitter is increased .
calculating these tables the transmitter and receiver have been taken to be vertical antennae , the wave-length is five times the height , and the results are given in two cases : for oscillatiollS of wave-length one-fifth of a mile corresponding to antennae of height 211 feet , and for oscillations of wave-length one-quarter of a mile corresponding to antennae of height 264 feet .
The first column gives the angular distance of the receiver from the transmitter , the second column * The investigation umes thab the series in obtained by the substitution converge yalues of up to , and this will be so if is the boundary of th cELECTRIC WAVES ROUND A PERFECTLy REFLECTIN OBSTACLL .
1.31 the distance of the receivel from the transmitter in miles , the third column gives the corresponding value of calculated from the formula the last column the ratio of the amplitude of the oscillations at the receiver to the amplitude of the oscillations due to the transmitter alone .
APPENDIX.\mdash ; INVESTIGATION 0F MATHEMATICAL RESULTS REQUIRED .
1 .
of Series .
If denotes the general term of a series , and the sum of terms , the first of which is which can be written symbolically where is and is put equal to zero after the operations are performed .
This is equivalent to or which can be expressed by The present object is to replace this integral by expressions suitable for calculation in different cases .
The first case is that in which contains an exponential factor , where ( an integer ) does not vanish or is not very small .
Writing where contains no other exponential factor of the form that is or hence that is .
Therefore whence The sum is thus expressed as the difference of two series which , in general , are divergin series , but they can be used for approximate calculation if the last term taken is within that part of the series which converges .
When the series *Cf .
BOOLE , ' Finite Differences , ' ELECTRIC WAVES ROUND A PERFECTLy TING OBSTACLE .
1.33 converges to a limit when 7 ' increases without limit , the limit to which it convel.ges will be given by , it being understood that only the terlns of this series where it converges are to 1 taken .
The above clearly fails it , for in that case vanishes ; it is also inapplicable if is very small , for then is very small , and the terms of diyerge at once .
The result shows that the sum , however great is , is not of an order higher than that of the terms that compose it when does not vanish or is not very small ; this may be compared with the known case of integral which vanishes unless the range of integration encloses the origin .
The second case to be considered is that in which ( to which the case is always reducible ) and contains an exponential factor ; writing and remembering that where are BERNOULLI'S numbers , it follows that When the real part of is not greater than zero the important part of the is that contributed by small values of , and writing it follows that \amp ; .
Now that is * When only involves as a polynolnial the series has only a finite number of terms which represent the exact sum .
For the determination of the important part of the integred it is sufficient that should be expressible in powers of for small values of and therefore , not being small , the important part of is ; similarly , the important part of is - ; hence , the important part of is , provided is of lower order than with corre- sponding conditions for , \amp ; c. Therefore , when is of an order higher than unity , the most important part of is , in this case the sum of an order higher than that of the terms that compose it .
The third case is when is small and contains an exponential factor of the form ; as in the immediately preceding case , the result depends on the value of the integral If the real part of is negative , or if and are both pure imaginaries , the important part of this integral is , with the same restrictions as in the previous case .
The fourth case is that when is small , and contains an exponential factor ; as above , the result depends on the value of the integral and , if the real part of is negative , or if both and are pure imaginaries , the important part of this integral is , with restrictions similar to those of the two preceding cases .
The integrals in the third and fourth cases have been fully investigated and tabulated .
2 .
Approximate Expressions for the Bessel Functions .
Most of the expressions to be investigated in what follows have been given by L. LORpNZ , obtained them from expressions for the sum of squares of the two solutions and the product of the two solutions .
The investigation that follows derives them directly from the fundamental expression a solution of BESSEL'S equation and the passage from the periodic form of the expressions to the form involving real exponentials , which is insufficiently treated by LORENZ , is traced .
Writing and using SCHLAFLI'S formula for , it follows that ' ( Euvres Scientifiques , ' tome I. , p. Math. Ann. Band III , In what follows will be taken to be a positive integer , and ) will be written for and ELECTRIC WAVES ROUND A PERFECTLy REFLECTING OBSTACLE .
Hence It is required to find approximations for the integrals above when is large .
Taking first the case where is less than , the important part of arises from the first integral on the right-hand side , and the most important part of that integral is contributed by the values of for which the exponent is nearly stationary ; writing is stationary when , and putting , the correspondin value of is .
Hence substituting \amp ; c. , and ; that is Now unless is small , the important part of the integral on the right-hand side is and unless is of the same or of higher order than , the second and third integrals on the right are negligible in comparison with the first , that is , if is of an order higher than nd therefore , with the same restriction as to the magnitude .
of the term of highest order only retained .
Hence The parts contributed by the second and third integrals in the first expression for are of order and thereforo gligible .
when is of an order higher than , and this is equivalent to where and ( ii .
) .
When is of the same or of lower order than , the part of the first integral in the expression for that is most important is that contributed by small values of ; the second integral now becomes of the same order , and the most important part of it is contributed by values of near to zero ; hence the remaining parts being oflower order , and therefore writing it follows that where and are large quantities , the first being proportional to and the second to , and as only the most important part of is required , and the parts contributed by large values of are negligible in comparison , and may be replaced by .
Hence all but the parts of highest order being ected .
( iii ) .
* The values of and in this case have now to be obtained , and it is convenient to take first the case where is small compared with .
The expression in ( iii ) can for this purpose be expanded in ascending powers of whiffi gives that is For the ) oses required it is sufficient to know values of and to the second powel ' of , hence neglecting and higher powers of ; It be verified that this integral is a solution of the differential equation , which is approximately 's equation for when is of the same or lower order than ELECTRIC WAVES ROUND A PERFECTLy REFLECTING OBSTACLE .
therefore Now , TI , whence , similarly , ; hence .
Again , therefore , where is given by ' whence where , and The value of , where is also required ; it is , however , more convenient to calculate .
From the relations it follows that Now therefore , that is , ; to obtain to the second power of , the third power of must be retained in and , hence Theso expressions for and are to the order given , the same as those given by VOL. CCX.\mdash ; A. that is therefore where that is whence .
When is of the same order as , the series for ) in ascending powers of is not suitable for obtaining approximate values , and must therefore be replaced by a series involving inverse powers of .
To effect this it is necessary to obtain the principal part of the integrals in ( iii .
) ; as in this case is negative , the principal part is contributed by the second integral .
Writing the important part of the integral arises from values of in the neighbourhood of the value of that makes stationary ; this value is given by , and substituting it follows that when is not small the principal part of the integral on the right-hand side is the same as that of the integral whose lower limit is ' , and therefore the principal part only being retained , Hence , the principal part only being retained , therefore and substituting for its value , this becomes . . .
( vii .
) .
These are the forms that the expressions for and , when is of higher order than , take when is near to ; for , writing , it follows that ELECTRIC WAVES ROUND A PERFECTLY REFLECTING OBSTACLE .
becomes ; now therefore and ; that is , retaining the principal part only , The leading term in being known , a further approximaticn can be obtained from the differential equation the result is The same remark applies to the approximation for when is of higher order than , the ordinary differential equation for BESSEL'S functions being made use of .
When is of the same order as the corresponding series when is positive is required .
The principal part of now arises from the integral , the principal part of which is contributed by values in the neighbourhood of the value of that makes stationary ; that is , in the neighbourhood of , and writing it follows that therefore the principal part is This result gives the leading term in the value of , to obtain the leading ternl in the value of it is necessary to calculate the principal part of the imaginary terms , that is , the principal part of *Cf .
STOKES , Caml ) .
Phil. Trarls vol. , p. 106 .
The principal part of the first of these integrals arises from values of in the neighbourhood of , and of the second from values of in the neighbourhood of , hence writing in the first , and in the second , the expression becomes and the principal part of this expression is equal to the principal part of which , writing in the first , and in the second , is .
equal to that is , equal to now the principal parts of the first and second integrals in this expression are equal , but with opposite signs , and therefore the principal part is Hence ; and . . . . .
( viii .
) .
The leading terms having been determined , the approximation can be carried further by using the differential equation , and the ] is * It should be observed that the constant of the imaginary part is half the value that would have been given by STOKES ' , p. 112 of the )aper referred to above ; the explanation of this is that the value of the in STOKES ' investigation that corresponds to this solution is one that belongs to a boundary for the intervaIs of this case is not discussed by STOKES , but it is not difficult to prove that for such values the constant takes half the it has in the intervaL ELECTRIC WAVES ROUND A PERFECTLY RIlFLECTING OBSTACLE .
The value of is given by the relations , , . . .
, the leading term only being retained .
When is great compared with , the necessary oximations are obtained from the original formula for , viz. :\mdash ; The principal part of this expression arises from the second integral , and is contributed by the values of near to the value that makes stationary ; if this value of is S , then and writing , it follows that sinh -\amp ; sinh the principal part of which is equal to the part which depends on the integral being of lower order when is of higher order than ; hence where S is given by Sn To obtain it is necessary to calculate the leading terms of the imaginary part of the expression for , and this arises from the first integral ; the part required is the principal part of The exponent in the first of these is stationary when , and the exponent in the second when , writing in the first and in the second , the expression becomes the ncipal part of which is equal to the principal part of that is to the principal part of and this is equal the principal part of which is and therefore Hence , writing S- where , we have . . . . . . . . .
It remains to prove that as S becomes small these expressions become identical with those obtained for the case when is of the same order as .
When S is small S- S-S S , that is , Now is when S is small , and therefore as S diminishes , the form of the expression in becomes that in ( viii .
) .
The approximations in can , as before , be ca1ried ELECTRIC ways ROUND A PERFECTLY CTING O further by using the differential equation satisfied by , which is BESSEL'S equation .
* The various approximations are collected below .
convenience of reference :\mdash ; When is reat compared with . . . . .
( .
) , where When is of the same order as .
When is oflower order than . . .
, , , where .
When is of the same order as . . . . .
( viii .
) , . . . . .
( ix .
) .
When is great compared with , where @ , When is not an integer the corresponding results can be obtained by writing * Another method of approximating to the value of is to make use of the reIation which is not difficult to cstablish , and then deduce from the resnlt ; the method given is more dircct , and avoids the difficuIties that arise in determining the constants for the different forms of in this other method .
' for in the original expression , but , with a view to the special case where is an integer , it is preferable to use the solutions and ; the expressions to be approximated to are and .
With the same notation as when is an integer where and have the same values as when is an integer .

