III .
" Note on Internal Radiation .
" By GEORGE G. STOKES , M.A. , Sec. R.S. , Lucasian Professor of Mathematics in the University of Cambridge .
Received December 28 , 1861 .
In the eleventh volume of the 'Proceedings of the Royal Society , ' p. 193 , is the abstract of a paper by Mr. Balfour Stewart , in which he deduces an expression for the internal radiation in any direction within a uniaxal crystal from an equation between the radiations incident upon and emerging from a unit of area of a plane surface , having an arbitrary direction , by which the crystal is supposed to be bounded .
With reference to this determination he remarks ( p. 196 ) , " But the internal radiation , if the law of exchanges be true , is clearly independent of the position of this surface , which is indeed merely employed as an expedient .
This is equivalent to saying that the constants which define the position of the bounding surface must ultimately disappear from the expression for the internal radiation .
" This anticipation he shows is verified in the case of the expression deduced , according to his principles , for the internal radiation within a uniaxal crystal , on the assumption that the wave-surface* is the sphere and spheroid of Huygens , In the case of an uncrystallized medium , the following is the equation obtained by Mr. Stewart in the first instance .
Let 1 , R ' be the external and internal radiations in directions OP , OP ' , which are connected as being those of an incident and refracted ray , the medium being supposed to be bounded by a plane surface passing through 0 .
Let OP describe an elementary conical circuit enclosing the solid angle aq , and let 8q ' be the elementary solid angle enclosed by the circuit described by OP ' .
Let i , i ' be the angles of incidence and refraction .
Of a radiation proceeding along PO , let the fraction A be reflected and the rest transmitted ; and of a radiation proceeding internally along P'O let the fraction A ' be reflected , and the rest transmitted .
Then by equating the radiation incident externally on a unit of surface , in the directions of lines lying within the conical circuit described by OP , with the radiation proceeding in a contrary direction , and made up partly of a refracted and partly of an externally reflected radiation , we obtain R cos i= ( 1 -A ' ) t ' cos i ' 0 ' + ARE cos i 1 , or ( l-A)Rcosia-=(l-A')R'cosi'l p ' ... .
( 1 ) In the case of a crystal there are two internal directions of refraction , OP .
, OP2 , corresponding to a given direction PO of incidence , the rays along OP1 , OP2 being each polarized in a particular manner .
Conversely , there are two directions , P , O , P , O , in which a ray may be incident internally so as to furnish a ray refracted along OP , and in each case no second refracted ray will be produced , provided the incident ray be polarized in the same manner as the refracted ray OP1 or OP2 .
In the case of a crystal , then , equation ( 1 ) must be replaced by ( I -A ) R cos i =(=(1 -A , ) R , cos i + , q+(1 -A , ) R2 cos i , 0y ( 2 ) In the most general case it does not appear in what manner , if at all , equation ( 2 ) would split into two equations , involving respectively I1 and R2 .
For if an incident ray PO were so polarized as to furnish only one refracted ray , say OP1 , a ray incident along P1O and polarized in the same manner as OP1 would furnish indeed only one refracted ray , in the direction OP , but that would be polarized differently from PO ; so that the two systems are mixed up together .
But if the plane of incidence be a principal plane , and if we may assume that such a plane is a plane of symmetry as regards the optical properties of the medium* , the system of rays polarized in and the system polarized perpendicularly to the plane of incidence will be quite independent of each other , and the equality between the radiation incident externally and that proceeding in the contrary direction , and made up partly of a refracted and partly of an externally reflected radiation , must hold good for each system separately .
In this case , then , ( 2 ) will split into two equations , each of the form ( 1 ) , R now standing for half the whole radiation , and R ' , A ' , &amp;c. standing for RJ , A1 , &amp;c. , or R2 , A2 , &amp;c. , as the case may be .
It need hardly be remarked that the value of A is different in the two cases , and that R ' has a value which is no longer , as in the case of an isotropic medium , alike in all directions .
In determining according to Mr. Stewart 's principles the internal radiation in any given direction within a uniaxal crystal , no limitation is introduced by the restriction of equation ( 1 ) to a principal plane , since we are at liberty to imagine the crystal bounded by a plane perpendicular to that containing the direction in question and the axis of the crystal .
Mr. Stewart further reduces equation ( 1 ) by remarking that in an isotropic medium , as we have reason to believe , A'=A , and that the same law probably holds good in a crystal also , so that the equal factors I-A , 1--A ' may be struck out .
Arago long ago showed experimentally that light is reflected in the same proportion externally and internally from a plate of glass bounded by parallel surfaces ; and the formulae which Fresnel has given to express , for the case of an isotropic medium , the intensity of reflected light , whether polarized in a plane parallel or perpendicular to the plane of incidence , are consistent with this law .
In a paper published in the fourth volume of the Cambridge and Dublin Mathematical Journal ( p. 1 ) , I have given a very simple demonstration of Arago 's law , based on the sole hypothesis that the forces acting depend only on the positions of the particles .
This demonstration , I may here remark , applies without change to the case of a crystal whenever the plane of incidence is a plane of optical symmetry .
It may be rendered still more general by supposing that the forces acting depend , not solely on the positions of the particles , but also on any differential coefficients of the coordinates which are of an even order with respect to the time , -a generalization which appears not unimportant , as it is applicable to that view of the mutual relation of the ether and ponderable matter , according to which the ether is compared to a fluid in which a number of solids are immersed , and which in moving as a whole is obliged to undergo local dislocations to make way for the solids .
On striking out the factors 1-A and I -A ' , equation ( 1 ) is reduced to R ' cos i B. = ... a ... ..(3 ) -Cos , f o &amp; In the case of an isotropic medium , R and R ' are alike in all directions , and therefore the ratio of cos i c9 to cos i ' cq ' ought to be independent of i , as it is very easily proved to be .
The same applies to a uniaxal crystal , so far as regards the ordinary ray .
But as regards the extraordinary , it is by no means obvious that the ratio should be expressible in the form indicated-as a quantity depending only on the direction OP ' .
Mr. Stewart has , however , proved that this is the case , independently of any restriction as to the plane of incidence being a principal plane , on the assumption that the wave-surface has the form assigned to it by Huygens .
It might seem at first sight that this verification was fairly adducible in confirmation of the truth of the whole theory , including the assumed form of the wave-surface .
But a little consideration will show that such a view cannot be maintained .
IIuygens 's construction links together the law of refraction and the form of the wave-surface , na manner depending for its validity only on the most fundamental principles of the theory of undulations .
The construction which tIuygens applied to the ellipsoid is equally applicable to any other surface ; it was a mere guess on his part that the extraordinary wavesurface in Iceland spar was an ellipsoid ; and although the ellipsoidal form results from the imperfect dynamical theory of Fresnel , it is certain that rigorous dynamical theories lead to different forms of the wave-surface , according to the suppositions made as to the existing state of things .
For every such possible form the ratio expressed by the right-hand member of equation ( 3 ) ought to come out in the form indicated by the left-hand member , and not to involve explicitly the direction of the refracting plane : and as it seemed evident that it could not be possible , merely by such general considerations as those adduced by Mr. Stewart , to distinguish between those surfaces which were and those which were not dynamically possible forms of the wave-surface , I was led to anticipate that the possibility of expressing the ratio in question under the form indicated was a general property of surfaces .
The object of the present Note is to give a demonstration of the truth of this anticipation , and thereby remove from the verification the really irrelevant consideration of a particular form of wave-surface ; but it was necessary in the first instance to supply some steps of Mr. Stewart 's investigation which are omitted in the published abstract .
The proposition to be proved may be somewhat generalized , in a manner suggested by the consideration of internal reflexion within a crystal , or refraction out of one crystallibed medium into another in optical contact with it .
Thus generalized it stands as follows : Imagine any two surfaces whatsoever , and also a fixed point 0 ; imagine likewise a plane II passing through 0 .
Let two points P , P ' , situated on the two surfaces respectively , and so related that the tangent planes at those points intersect each other in the plane IT , be called corresponding points with respect to the plane II .
Let P describe , on the surface on which it lies , an infinitesimal closed circuit , and P ' the " corresponding " circuit ; let 6q , 6q ' be the solid angles subtended at O by these circuits respectively , and i , i ' the inclinations of OP , OP ' to the normal to II .
Then shall the ratio of cos iaq to cos i'aq ' be of the form [ P ] : [ P ' ] , where P depends only on the first surface and the position of P , and [ P ' ] only on the second surface and the position of P ' .
Moreover , if either surface be a sphere having its centre at 0 , the corresponding quantity [ P ] or [ P ' ] shall be constant .
It may be remarked that the two surfaces may be merely two sheets of the same surface , or even two different parts of the same sheet .
Instead of comparing the surfaces directly with each other , it will be sufficient to compare them both with the same third surface ; for it is evident that if the points P , P ' correspond to the same point P1 , on the third surface , they will also correspond to each other .
For the third surface it will be convenient to take a sphere described round 0 as centre with an arbitrary radius , which we may take for the unit of length .
The letters P , , i , , q , will be used with reference to the sphere .
Let the surface and sphere be referred to rectangular coordinates , O being the origin , and II the plane of xy .
Let x , y , z be the coordinates of P ; X , 7r , ' those of P1 .
Then x , y , z will be connected by the equation of the surface , and , , , by the equation 422+ 2= 1 .
According to the usual notation , let dz dz d , z d2z d2z d=I die .2-r , die d 2t .
The equations of the tangent planes at P , P1 , X , Y , Z being the current coordinates , are z-z=p ( X)+q ( Y-y ) , sX+nY tZ= 1 ; ' and those of their traces on the plane of xy are pX+ qY=x+ qyz , 4X+ ?
Y=1 ; and in order that these may represent the same line , we must have = I , ... ( 4 ) px+qy-z px+qy-z ' To the element dxdy of the projection on the plane of xy of a superficial element at P , belongs the superficial element dS= V1 +p2+ q2 dxdy , and to this again belongs the elementary cos vdS solid angle -- , where p=OP , and v is the angle between the normal at P and the radius vector .
Hence the total solid angle Cos vAa within a small contour is - ' / Ip2"+ dxdy , the double integral being taken within the projection of that small contour .
Also cos i== ?
Hence Pz COs v cozs i cos= a/ 1 +P2q2+ dxdy ; cos i 3~ ?
=--~-V ' 1 &lt; L and applying this formula to the sphere by replacing z V/ .lt .
+2 .
by 1 , v by 0 , and p by 1 , we have cos z1=fff4dd/ , l the double integral being taken over the projection of the corresponding small area of the sphere .
Now by the well-known formula for the transformation of multiple integrals we have t ?
; ( l= / d~ dn d ddxdyJdy dyx &gt; 4 and therefore cos io zcos vV1 +pj+(/ cos8 ifl 3dd Ci dr t drl\ P\x die die dx/ But the first of equations ( 4 ) gives ( px + qiy-z)dp-p(xdp +ydq ) d ' ( px+qy-z)2 _ { ( qy-z)r-pys}dx+ { ( qy-z)s-pyt } die .
( px+ -qy_-)2 Similarly , _ ( px-s)t--qs } die +{ ( px-z)s-qxr } d ( px +qy-y ) Hence clt d tr ld qcl dVx die die dx~(px + qyz ) ' where V= { ( qy--)r-pys4 { ( px-z)ts { ( qy--z)s-pytJ } ( px-z 8-qxr = { ( qy--)(px-z)--pqxy ( rt--s ) Z=(z-px qy)(rt s2 ) .
Hence cos iS _Z COS , / / 1 - ?
p2 q2 ( z-x--qy)3 cos i1qp p3 z(rt-s2 ) But if are be the perpendicular let fall from O on the tangent plane at P , S-px--qy =V 1 -+2 +q2 .
r , and therefore Cos i_ cos v. Wa ( +-p2 + q2)2 Coszlb1 p3 rt-_s2 But -==p cos v. Also the quadratic determining the principal radii of curvature at P is ( rts2)V2 + ( &amp; C.)v ( 1 +p2 2)2-O ; and therefore if v , , v , denote the principal radii of curvature , w ( 1 +pl + q2)2 ( 1 ?
-2p2+ rt--s " Hence cos i/ 4 =cos4 .
v v , : , ... ... ... ( 5 ) and COS i ' ' COS 1C COS zi C , 'COS Y ' .
V1V , c,..(6 ) cos iY'( cos iz1 Cos iji cos4 S6 Cosi~q0-cos iC'O &amp; p cosi Oj/ )coss~r~.v~va~ which proves the proposition enunciated .
In the particular case of an ellipsoid of revolution of which n is the axial and m the equatorial semi-axis , compared with a sphere of radius unity , both having their centres at 0 ' , one of the principal radii of curvature is the normal of the elliptic section , which by the properties of the ellipse is equal to m m , m denoting the semi-conjugate diameter ; and the other is the radius of curvature of the elliptic section , onr - ' Also w is the perpendicular let fall from the centre on the tangent line of the section .
Hence from ( 5 ) or ( 6 ) cos im s4 mm ms w4m .
4n mn cos '0-~'-p4 n mnn2p4 p4 since zm ' = mn .
This agrees with Mr. Stewart 's result(p .
197 ) , since the Re and -R of Mr. Stewart are the same as the R ' and R of equation ( 3 ) .

