]\gt ; Incidence of Light upon a Transparent Sphere of with the By LORD RAYLEIGH , O.M. , F.R.S. Received Read April 2 In a paper , now nearly thirty years old , applied Maxwell 's equations of the electro-magnetic field to the disturbance produced by an obstacle upon plane waves of light which travel through a medium other- wise uniform , particular attention to case where the properties of the obst , acle differ but little from those of its .
The difference may consist in a variation of the specific inductive capacity , or of the netic capacity , or of both ; but it was shown that the last supposition leads to results inconsistent with tion , and that the evidence favours the view that is to be treated as inval.iable .
electric displacements by , the primary wave was taken to be , ( 23 ) so that the direction of propagation is atively ) , and that of vibration parallel to being omitted , the electric displacements in scattered wave , so far as they depend upon tlJe first of , have at a great distance the values in hich .
( 36 ) In these equations denotes the distance between the point , where the disturbance is required to be estimated , and the element of volume of the obstacle cbcdydz .
It is evidently implied that in a direction parallel to that of primary vibration , the scattered vibration vanishes , whatever may be the size and shape of the obstacle , so that any which may appear in this direction must depend upon powers of higher than the first .
If be uniform the obstacle , it may be removed from under the integral sign in ( 36 ) , and if the value of for the external medium be taken as unity , we may write .
'Phil .
Mag vol. , p. 81 , 1881 ; 'Scientific Papers , ' vol. 1 , p. 518 .
In the second term of Equation ( 32 ) and following , should read 26 Lord Rayleigh .
Incidence of Light on [ Apr. 4 , If the obstacle is very small , and may be treated as constants in the integration , and , ( 40 ) if be the whole volume of the small obstacle .
Using this in ( 35 ) , etc. , we get being the wave-length .
These equations apply , whatever may be the shape of the small obstacle ; and , of course , they undergo no change if the shape be spherical .
In that case it was shown that the restriction upon the value of may be removed , provided we replace by We have seen that the scattered light vanishes to a first approximation in ( K\mdash ; 1 ) for the direction parallel to .
In the case of the sphere it was shown that for this direction we have to a second approximation for a small obstacle , ( 60 ) with .
In ( 60 ) denotes the radius of the sphere , and is measured from the centre .
" " Comparing ( 60 ) and ( 41 ) we see that the amplitude of the light scattered along is not only of higher order in but is also of the order in comparison with that scattered in other directions .
The incident light being white , the intensity of the component colours scattered along varies as the inverse eighth power of the wavelength , so that the resultant light is a rich blue It is obvious from ( 41 ) that in general the intensity of the scattered light varies as " " There is anothel point of importance to be noticed .
Although when the terms of the second order are included , the scattered light does not vanish along the axis of , the peculiarity is not lost but merely transferred to another direction .
Putting together the terms of the first and second orders we see that the scattered light will vanish in a direction in the plane of inclined to ( towards ) at a small angle , such that ( 61 ) Some experiments in illustration were then described .
Returning to the case where is neglected , we consider a spherical obstacle of any radius , whose centre is at the origin of co-ordinates .
From ( 23 ) , ( 36 ) 1910 .
] Sphere of Dimensions Cornparable with -Length .
27 being the distance between dxdydz , and the ( distant ) point at which are to be estimated .
It is evident that , so ' as the secondary ray is concerned , depends only upon the which this ray makes the primary ray .
We will suppose that in the direction vards along the primary ray , and that the primary ray continued .
The integral may then be found in the form ; ( 48 ) the distance of the point of observation from the centre of the sphere .
If we expand the Bessel 's function and drop the suffix in as no longer required , we get , ( 49 ) in which is written for .
From this follow at once by ( 35 ) , etc. Along the continuation of the primary ray , and reduces to as was to be expected .
It is to be observed that in this solution there is no limitation upon the value of if is ected absolutely .
Prof. Love , in a valuable paper on the ' Scattering of Electric Waves by a Dielectric Sphere treated this problem by a different method , limited on the one hand to the sphere , but on the other applicable may be the value of ( K\mdash ; 1 ) , and ( so far as the general analytical expressions are concerned ) whatever may be the size of the sphere .
In the 1881 paper I had treated in this way the problem of a .
From these expressions Love deduces first and second approximations applicable when the radius of the sphere may be treated as small in comparison with the wave-length .
It will be of interest to compare these approximations with those already referred to .
In Prof. Love 's llotation the electric forces are represented by X , ; and , so far as relates to the medium outside the sphere , they may be identified with fg , used above .
The co-ordinates of the point of observation are denoted by instead of by , Love supposes the direction of ation to be along ( instead of ) , and the direction of 'Lond .
Math. Soc. Proc vol. 30 , p. 308 , 1899 .
28 Lord Rayleigh .
Incidence of Light on a Transparent [ Apr. vibration to be parallel to ( instead of ) , changes represented by taking one step backwards in the cycle .
Thus ( with omission of a constant multiplier which runs through ) the primary wave is represented by .
( X ) to this , the complete expressions , so far as terms of order for the forces in the scattered wave at a great distance are found .
to be XLII ) So far as the first approximation involving , this agrees rvutandis with , since and We may next consider the application of ( XLII ) to the secondary ray in the direction of primary vibration , for which the first approximation vanishes .
We have ; and this agrees with ( 60 ) when we introduce the supposition that ( K-1 ) is small , so that in the denomination may be identified with 5 .
I have already commented upon this agreement ; but recentl.y in the comparison I came upon a discrepancy .
It is evident from the theory already given that a great general simplification should attend the supposition that ( K\mdash ; 1 ) is very small .
Introducing this into ( XLII ) and omitting for the moment the factor we find for the three last sets of terms 'Scientific Papers , ' , p. 536 , 1899 .
1910 .
] Sphere of Dimensio -Length .
29 showing within the bracket the same dependence on , \ldquo ; as does the first set .
Hence ether , expanding the exponential , we may write in this case This correspond with , ( 49 ) of the former ation if in the latter we stop at the term in where so that It will be seen that in all respects except one the agreement is complete .
For example , has disappeared otherwise bhan as contained in .
But there remains a numerical discrepancy , since is not equal to For some time I was at a loss to account for this discrepancy , until I noticed that the omission of a term in Love 's equation ( XXXIV ) would produce harmony , and this led me to scrutinise equation ( XVIII ) , where ( I believe ) the error lies .
It appears to me that these equations should be ' the change in the substitution of for in the denominators of the first terms of .
We should then obtain simply , instead of ( XXXIV ) , which involves also the factor .
This is matter to which I shall presently return .
In view of numerical ca]culations , it may be well to complete the statement of Prof. Love 's , and to show that , especially when COl.lected as above , they admit of considerable simplification .
30 Lord Rayleigh .
Incidence of Light on [ Apr. 4 , For the functions which occur in ( XVIII ) we have , and for greater than unity , ( XIV ) denoting as usual Legendre 's function , so that is a solid harmonic of degree .
For we have and the functions satisfy certain sequence equations which may be written .
( VI ) We have also to deal with another function , , such that , ( VII ) and the functions satisfy the same sequence equations as the functions .
It will be seen that is the imaginary part of .
As included in , .
( XXXVI ) This being understood , the disturbance in the secondary wave , so far as it depends upon , may be written ( kr ) ( kr ) , ( XV ) with two similar equations for .
The complete expressions for X , involve a summation with respect to which takes the values 1 , 2 , 3 , etc. All is now defined except the constants , which depend upon the boundary conditions to be satisfied at the surface of the sphere .
For the present case of a dielectric sphere of refractive index , or Prof. Love finds the equivalent of , ( XXIII ) This series is the same as that which occurs in the expression for } 1910 .
] Sphere of Dir lsions Cwith .
31 ( XXIII bis ) which completes the analytical solution .
Returning now to we see that the terms are solid harmonics of .
Also is a solid harmonic of degree .
When differentiated respect to ' or , it becomes a solid harmonic of , and when further by , a solid harmonic of .
Hence both terms in are solid harmonics of degree / , and it will appear indeed , that they are of the same form .
Writing for , we have from .
( ) Again ( b ) Now ( Todhunter 's 'Laplace 's Fumctions , ' pp. 43 , 46 ) the quantities within in ( a ) and ( b ) are equal to one another and to , so that , ( ) and by .
We thus obtain the reatly simplified forms ( d ) in which does not appear at all .
Referring to ( XIV ) and introducing the value of , we find , ( XXVI ) Thus in the case of we have without any approximation depending on the smallness of R. 32 Lord Rayleigh .
Incidence of Light on a Transparent [ Apr. 4 , For further reductions it will be convenient to introduce a modified form of , viz. , defined by We have now to consider the solution expressed in ( XV ) , and ma at a once introduce the simplifying condition that is very great .
In this case from ( VII ) ; ( XXVIII ) and we find from ( XV ) , ( d ) ; ( e ) and in the cyclical changes by which and are deduced from are to stand unaltered .
For systematic calculation it is best to express in terms of P. As in ( a ) ; ( ) so that we may write The differential coefficients which appear in ( e ) are now readily formed .
We find , ( h ) ( i ) The in ( j ) may alternatively be written , in virtue of the differential equation satisfied by It should be observed that the above values satisfy , as of courSe they ought to do .
1910 .
] Sphere -Length .
33 At this it may be convenient to write down the values of for the earlier values of Thus from , ( i ) , ( j ) , ; , ; I3 ; Our expressions naturally nplify when we limit ourselves to secondary rays which lie in the principal planes .
No real loss of enerality is thereby incurred , since whatever may be the direction of primary vibration , that vibration may be resolved into two , in and perpendicular to the plane which contains the primary and secondary rays .
Thus , if , that XXXIV .
34 Lord Rayleigh .
Incidence of Light [ Apr. 4 , is for secondary rays whose direction is perpendicular to that of primary vibration , we have generally , and .
( k ) Again , if , that is for secondary rays in the plane which contains the directions of primary vibration and , and while retains the form ( j ) .
The direction of vibration is in the plane , and is , of course , perpendicular to the secondary ray .
Its nitude is given by .
( t ) It will be seen that and ( l ) differ only by an interchange of It may be remarked that when and ( l ) should become identical , except as to sign .
If , the signs are the same ; if , they are opposite .
Thus when , the in ( l ) becomes simply .
And in this case by the differential equation , When is even , , and then when is odd .
, and then .
Accordingly , ) gives in the case of , the upper sign in the anbiguities relating to Arranged in powers of the general expression as far as sive is Mo , 1910 .
] Sphere of Dimensio -Length .
35 The corresponding expression for from ( k ) is derivable by interchange of and N. When , we have in consider In this when / is odd , and takes in succession the values 1 , , , - , etc. , while when is , and takes the values [ 1 ] , - , etc. Using these in ( l ) , we find for , ( ) and the corresponding value of in ( k ) is obtained by interchange of and N. We pass on to consider the values of as given in ( XXIII ) .
By ( VII ) , when is small , is of the order , so that is of the order is two orders higher , , since the nuulerator is itself a small quantity of the order .
When is small , the most important terms in X , depend upon and are of order or .
If we are satisfied with an approximation as far as inclusive , we have only to consider , and .
For we need , and so on .
Let us begin by supposing that ( K-1 ) is small , and retain my the first power of this quantity .
For example , in , we may write .
R. We find , omitting for brevity the argument in , etc. , , ( 1 ) ( K-1 ) The denominator in , ( q ) appears at first sight complex , since is 36 Lord Rayleigh .
Incidence of Light on a Transparent [ Apr. 4 , If , however , we separate real and imaginary parts , writing in accordance with the definitions , ( VII ) , ( ) we see that , ( ) from which the imaginary part has disappeared .
In this case accordingly are wholly real .
The denominator ( s ) admits of simple evaluation .
By ( VI ) .
Now , if be a solid harmonic , the same in both cases , and both satisfy the equation , and hence by Green 's theorem Theory Sound , as applied to the space included between two concentric spheres , we find that , where is independent of .
To determine we may suppose very small , when Thus and .
( t ) It will be understood that ( t ) is true without approximation as regards if ( K-1)2 is neglected .
It is now a simple matter to calculate to a moderately high power of .
So far as inclusiye , we have , , Thus 1910 .
] Sphere of Comparable toith Wave-Length .
37 in , so that becomes a factor of the whole expression as we know it ought to do .
Also In X the multiplier of be the same as the multiplier of in Z. In order that this 1nay be so , we must have Ns , a relation satisfied by the tabular values .
In the expression for , the part which does not contain the factor , or , vanishes in virtue of the relations already mentioned .
Finally , in powers of , or X , in agreement with ( 35 ) , ( 49 ) so far as the present solution extends .
The former method is incomparably the more appropriate in this particular case , giving the result for the sphere to any degree of approximation in , as well as admitting of application to obstacles of other forms .
The general conclusion that ( o ) vanishes when ( K-1)2 is neglected would probably be difficult to establish by the present nethod .
Abandoning the restriction as to the of ( K-1 ) , we find from ( XXIII ) as far as inclusive ; where it will be seen that to this order are still real .
The values of X , are iven by X , .
( ) This agrees with Prof. Love 's solution , except the first set of terms , as already mentioned .
The special case where the secondary ray lies in one or other principal plane is given by ( n ) and its derivative , when the approximate values of are introduced .
38 Lord Rayleigh .
Incidence of Light on nsparent [ Apr. 4 , When , or , can no longer be treated as small , we must calculate and from the eneral formulae ( XXIII ) and ( XXIII bis ) .
Separating into its real and imaginary parts , we see that the denominator of is equal to numerator .
Also that the numerator for exceeds the numerator for by ( K-1 ) ; and that the real part of the deno1ninator of exceeds the real part of the denominator of by ( K-1 ) , while the imaginary part of the denominator of is numerator of .
When the numeratolB and denominators have been computed , the values of musb be reduced to the standard complex form .
In selecting special cases for calculation we notice that and and are both arbitrary .
With the view of distinctive , it seemed desirable to choose a refractive index not too small .
For this purpose seemed suitable , as large enough , and yet not outside the experimental range .
Tbis corresponds to .
By further choosing make the tabulation of serve a double purpose .
The most obvious procedure for , equivalent to , is to use the sequence formula ( VI ) , , ( VI ) starting from and , as given by ( XXXVI ) .
So far as regards the real part , represented by , this course is satisfactory ; but the same cannot be said of the application to .
As increases , the successive 's tend to equality , and any errors are multiplied at the next step by the factor .
So much is this the case that although use seven-figure logarithms throughout , even the first figure in or becomes doubtful .
It is true that these higher ordel .
S in do not need to be known very accurately , for they contribute but little to the final result ; nevertheless , an improvement in the calculation is called for .
An easy escape from the difficulty is provided by the series in ( IV ) .
When is moderate , as here , and somewhat large , the converce is good , and we may calculate and by a pel.fectly straight1910 .
] Sphere of )Wave-Length .
39 forward process to a degree of accuracy .
obtained and we may then use the sequence formula ( VI ) iu the reverse direction to deduce in turn , without loss of accuracy .
values of and , calculated for the various cases , are here tabulated .
It is possible that they may prove useful in kindred enquiry From them are deduced the complex values of and as already explained .
In this case the tables the logarithms of the real and ary parts .
Tables of and 40 Lord Rayleigh .
Incidence of Light on Transparent [ Apr. 4 , Tables of and N. For a systematic calculation it appears preferable to employ the forms ( l ) rather than the expansions in powers of .
The functions involving , once calculated for the various values of , are then available for any value of .
The following table exhibits the ebraic forms:\mdash ; 1910 .
] Sphere of ensions C. 41 The particular cases have already been considered .
It will be observed that both functions are either entirely odd or entirely even .
The tables give arithmic values for , 12 , 34 , 1 .
Thus , in the first table for , we have the number whose arithm is and this is to be taken atively when foreover , since this function of is odd , the sign is to be reversed , i.e. , made positiye , when Table for Table for The way is now clear for the last step , the calculation of and by with respect to of the values given in ( l ) and .
As to this there is little to call for special remark .
Of course the odd and even terms are added separately .
42 Lord Rayleigh .
Incidence of Light on Transparent [ Apr. \mdash ; Cur ( dotte reprt 1910 .
] Sphere of Dimensions tvith -Length .
4 The tables exhibit the complex values and the nlodulus , subject to multi- plication by the factor .
Thus for , we have , where influences the phase only .
As the values of , or , or , the numbers 1 , ) , 9/ 4 were at first chosen .
Subsequently it appeared desirable to fill up the at the more interesting , and so calculations for vere sdded .
The values of the modutus are plotted in the curves , , as functions of for the various values of .
It will be remembered that all cases the refractive index is supposed to be 44 Lord Rayleigh .
idence o. Light on [ Apr. 4 , \mdash ; \mdash ; Modulus . .
Modulus .
3.0608 0.7844 0 .
0.3925 0.0834 0.2751 0.5429 When , when the circumference of the sphere equals the wavein the general medium , there is comparatively little departure from the conditions appropriate to the infinitely small sphere .
When the secondary ray is in a direction perpendicular to that of primary vibration , viz. , when , the vibration ( Y ) varies but little with .
It is otherwise when the seco1ldary ray is in the plane containing the of the primary ray and of primary vibration .
The second column shows that in that case the secondary vibration varies greatly with , and that in the direction the intensity is very small .
This is the direction in which would be no effect , however large the sphere , if the square of the diHerence of optical quality be neglected absolutely .
The lrture from the law of the infinitely small sphere becomes more apparent when .
The principal feature here is the displacement of the maximum polarisation from to the neighbourhood of , i.e. , to a direction inclining along the course of the incident When the incident is unpolarised , the diffracted light is polarised at all angles ( other than ) , and in the same sense as from infinitely small particles .
The latter law ceases to hold when rises to about .
A third neutral point ( direction of no polarisation ) enters at and moves inwards .
1910 .
] Sphere of nensions Comparable -Length .
45 When the change has proceeded so far that the neutral point is in the neighbourhood of , the polarisation for positive being the of that found for infinitely small particles .
At the reversed polarisation is very pronounced .
When attains , the neutral point is still further on ( near , and the maximum reverse polarisation is near .
It will be seen that these changes occur very rapidly in the neighbourhood of vould be possible to follow these calculations to greater values of , and such an extension would not be without interest , but the arithmetical work would soon become heavy .
Also , without , the refractive index be varied .
Some data for such calculations are included in the tables already given .
So far the primary light has been supposed to be either umpolarised , or , if poJarised , to be plane polarised in one or other plane .
In the more general case recourse must be had to the complex expressions , which allow the effects of the two principal components of the light to be combined , with the necessary allowance for any phase-difference .
But it is scarcely necessary to pursue the treatment of this part of the subject .
So far as I know , the only experimental work at all closely upon the preceding calculations is that of Govi and Tyndall ( ' Proc. January , 1869 ) , who observed the light scattered from smoke of various kinds .
Tyndall established the existence of neutral points at angles dependent upon the condition of the smoke , but his indications are not very precise .
For example , he not between density due to increased number of particles in a given space and that due to size , of which the effects lfiy ) ected to be very different .
Nor , again , do I see any suggestion that the position of the neutral point with the colour the .
Nevertheless , Tyn dall 's description is estive , ond it formed the starting-point of my work upon the subject .
In continuation of the experiments already referred to at the colnmence- meant of this paper , I have lately repeated the observation on the nascent precipitate formed when a little acid is added to a dilute and well-filtered solution of hyposulphite .
The liquid was contained in a small beaker placed in a dark room and exposed to a beam of sunlight .
The observation of the light dispersed at various may made with a nicol , but a double image prism is better .
This may be mounted at one end of a tube of which the other carries a small aperture of sucl ] dimensions that two oppositely polarised rectangular fields itre seen close juxtaposition .
In the early Incidence of Light on Transparent Spher.e , etc. the double-image prism , suitably rotated , shows complete polarisation of the scaf , tered , at right angles to the primary rays AB .
Somewhat later the minimum light seen along BC shows Tyndall 's " " residual blue indicating that for this colour the polarisation is no longer complete .
But if the direction of obseryation be changed towards BD , the polarisation is improyed .
As the particles grow , a stage is reached in which the effects for blue and red light are well contrasted .
If a blue glass be inserted in the course either of the primary or of the secondary , that polarised ponent which originally vanished ( along ) is found along BD to be the brighter of the two , showing that the polarisation is .
The substitution of a rerl for the blue glass restores the original character of the polarisation , the increase in the wave-length compensating the growth of the particles .
It is not pretended that the observed contrast is as great as the theoretical curves gest ; nor could this reasonably be expected , seeing that the particles are not all of the same size .
have recently learned from Prof. Love that he had become aware of the error in equation XVIII , and that a correction was published in 1 Math. Soc. Proc ) , p. 489 .

