]\gt ; [ 263 ] VII .
On the Dispersion Artificial Double Refraction .
L. N. G. FILON , M.A. , D. Sc. , Fellow and Lecturer athematics of University College , London .
by Professor F. T. TROUTON , F.R. Received January ) Read February 28 , 1907 .
SUMMARY 0F CONTENTS .
S1 .
Introduction . . . . .
264 PART I. 0F THE EXPERIMENT AND DISCUSSION 0F THE VARIOUS ERRORS .
S2 .
Simple theory of the experiment . . . . . .
265 S3 .
Description of the apparatus . . . . .
268 S4 .
Effect of introducing the polarizing Nicol . . . . . .
272 S5 .
Effect of finite breadth of the source . . . .
274 S6 .
Effect of relative rise and fall of the two beams and of elastic yielding of the bed-plate and knife-edges . . . . .
274 .
S7 .
Influence of obliquity 011 rolative rdation . . . . .
275 S8 .
Combined effect of flexur.e and obliquity . . . . .
276 S9 .
Imperfect adiustment of the inclination of the slit . .
, . .
) S10 .
Imperfect horizontal adjustment of the knife-edges . . .
, .
277 S11 .
Imperfect vertical adiustment of the knife-edges . . . . .
278 S12 .
Error due to weight of beams themselves . . . . . .
278 S13 .
Error due to } ) erfect annealing . . . . . . . . .
278 PART II .
EXPERIMENTAL RESULTS .
S14 .
Glasses observed . . . . .
279 S15 .
Linear law connecting and the stress . . . . . .
280 S16 .
Significance of this Linear law . . . . . .
284 S17 .
Methods of reduction . . . . . . . . . . . .
287 S18 .
Tables of results . . . . . . . . . .
289 S19 .
Discussion of the values of . . . . . . . .
293 S20 .
Systematic residuals . . . . . . . . . . . . . .
293 S21 .
Possible explanation by absorption bands . . . . . .
297 S22 .
Determination of absolute values of . . . . .
301 S23 .
Effect of chemical composition on stress-optical properties . . .
302 S24 .
Failure of HOOKE'S law for glass 2783 . . . . . .
303 S25 .
Conclusion . . . . . . . . . .
305 ( 419 .
) 26.7.07 S1 .
Introduction .
IT is well known that glass and other transparent isotropic substances , when compressed unequally in diflerent directions , behave like doubly-refi.acting substances and exhibit the colours of polarized light .
Attention was first called to this by FRESNEL ( ' AnIlales de Chimie et de Physique , ' vol. XX and by Sir DAVID BREWSTER ( ' Phil. Trans. , ' 1816 ) .
For further investigations in this field , reference may be made to F. E. NEUMANN ( ' Abhandlungen der .
Acad. .
Wissenschaften zu Berlin 1841 , .
; see also ' Pogg .
Ann vol. LIV .
) ; to CLERK MAXWELL ( ' Trans. Roy .
Soc. Edin .
, ' vol. , Part I. ; or 'Collected pers , ' vol. I. ) ; to G. WERTHEIM ( ' Annals de Chimie et de Physique ser. 3 , vol. XL .
, p. 156 ) ; to J. KERR ( ' Phil. Mag October , 1888 ) ; and to F. POCKELS ( ' ' Uber die nderung des optischen Verhaltens verschiedener iser durch el stische Deformation 'Anm . .
Physik , ' 1902 , ser. IV .
, vol. VTI .
, p. 745 ) .
Of these only WERTHEIM and POCKELS have considered how the effect varies with the nature of the light employed .
If honlogeneous parallel light is passed.perpendicularly through a plate of thickness which is sub.jected to principal stresses in its plane , these stresses being uniform throughout , then it is found that the light on traversing the plate is broken up into two rays in the directions of principal stress .
The relative retardation in centimetres of these on emergence is given by where the indices of refraction of the two rays .
Now experiments have shown that is very ) roximately proportional to the principal stress difference in the wave-front , P-Q .
Whether this is true for high values of P-Q is not certain , and some experiments to be described in the following pages ( see S19 ) will show that the proportionality of to P-Q in all cases must still be regarded as doubtful .
Assuming , however , this law , which is certainly very nearly true in most cases , at all events when are stresses of the same type ( tensions , or pressures ) , we have ( P-Q ) where is a coefficient depending only on the nature of the erial and on the wave-length of the light used .
This coefficient will be spoken of in what follows as the " " stress-optical coefficient , from observations of a uniformly compressed block of glass through which he passed successively sodium light , ( ii .
) white light , ( iii .
) white light filtered through a red glass , stated the following The relative retardation in air is constant for all colours .
In other words , the stress-optical coefficient is independent of the wave-length ; the difference of the refractive indices is therefore likewise independent of the wave-length , that is , the double refraction due to elastic strain exhibits no dispersion , DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
POCKELS , in his more recent investigation , observed the artificial double refraction in a number of Jena glasses .
His observations , though not primarily intended to test the effect of colour , nevertheless gave exceedingly valuable resuIts in connection , insomuch as POCKELS experimented with three different kinds of nearly homogeneous light , namely , those of sodium , lithium , and thallium .
The results would therefore be far more precise than those obtained with very mixed colours by WErTHEIM .
They show that , in certain glasses , the stress-optical coefficient does vary with the wavelength , being numerically greater in the green than in the red ; and in very heavy lead glasses this variation is more rapid as we approach the blue end of the spectrum .
Some years ago the present author , being at the time unaware of POCKELS ' experiments , devised a method of observing the variation of the coefficient continuously throughout the spectrum , the object being to test the exactness of law .
An account of this method , which was modified and improved from time to time , and of the experiments undertaken to carry it out , will be found in the ' Camb .
Phil. Soc. Proc vol. XI .
, Part VI .
; vol. XII .
, Part I. ; vol. XII .
, Part V. These experiments amply confirmed the results of POCKELS .
They also showed that the chief desideratum for obtaining accurate results was that the stress in the glass slab through which the light was passed should be sufficiently uniform .
Now the compression apparatus which was used by the author , and by previous experimenters , suffered from the defect that it was practically ossible to adjust it so as to obtain a uniform pressure in the slab of glass under observation .
Moreover , what adjustment could be made was long and difficult , and could be attained only by trial ; it appeared further that this adjustment was disturbed , in a way that could not be calculated and allowed for , when the load was aItered .
This greatly reduced the accuracy expected .
An apparatus was then devised , with a view to obtaining a system in which the stresses should be known exactly and in which the optical effects should be the same as those due to uniform pressure in a slab of constant thickness .
For the purposes of this research a Government Grant was kindly placed at the disposal of the author by the Royal Society , whereby the necessary apparatus could be constructed and the expensive glasses required for the research purchased .
The present paper is an account of the experiments carried out with the new apparatus and of the resuIts reached .
PART I. THEORY OF THE EXPERIMENT AND DISCUSSION OF THE VARIOUS ERRORS .
S2 .
Simple Theory of the Experiment .
Let ( fig. 1 ) be two rectangular slabs of glass , whose cross-sections are shown in the figure .
The slabs are bent in a vertical plane by couples without shear , whose axes are izontal and parallel to the plane of the cross-section .
How such couples are obtained will be explained subsequently .
VOL. CCVII.\mdash ; A. 2 NI The horizontal and vertical sides of the cross-sections of and are , respectively , and the centres of the two cross-sections are and Let be a point-source of light ; its image after passing through the points in which any ray through meets the vertical midplanes of respectively .
Fig. 1 .
Let be the image of , after a single refraction at the inner face of ( the one towards F ) , and the image of after a single refraction at the inner face of ( the one towards N ) .
Then it is evident that must be in one straight line .
Let height of above a fixed horizontal plane , , , , , the same plane , , , , , , , distance of from the nearer face of , , between midplanes of refractive indices of respectively .
We have and since are collinear ( 1 ) , or , writing .
( 2 ) , ( 1 ) becomes ( 3 ) , Suppose now monochromatic light proceeds from S. be the stress-optical coefficients of the two slabs for this kind of light , their bending moments reckoned positive when the slabs are bent concave downwards .
DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
The relative retardation after traversing the first slab for any ray which passes through at a small angle to the horizontal .
For , although the stress is not uniform along the path of the ray , the mean stress along the raystress at the middle point , since the stress varies linearly as the distance from the neutral axis .
Also the length of the ray differs from the breadth of the slab only by quantities of the second order .
Hence the result above .
Accordingly the total relative retardation , after passing through the two slabs , is given by . . .
, .
( 4 ) .
for from into ( 4 ) , ( 5 ) .
Now , if is to be independent of , we must have ( 6 ) .
This condition will not of course be accurately fulfilled for all colours at the same time ; in the first place , because and do not in general follow the same law of variation for the two slabs ; in the second place , because contains , and therefore involves the wave-length .
It is easy to see that the latter cause of error is quite negligible .
For if be large compared with or , which was the case in all the experiments , the error introduced in by a variation S in the refractive index ( taking , which is practically true ) is approximately Now , centims. in the experiments to be described ; centims. about ( or larger ) , S001 between the and lines of the spectrum which represent fairly well the extreme range of the observations .
Hence , bein about Accordingly the error introduced by this cause would correspond to an error in of less than 6 in a hundred thousand , an error which is absolutely negligible , since the errors of ordinary observations in the method to be described amount to or of C. Camb .
Phil. Proc The different variation of and with the wave-length would be far more important .
This , however , need not be considered , for the two slabs and are taken from the same cast , so far as possible , so that and should be identical .
In some cases it was found that and differed ; but , at the same time , the experiments showed conclusively that for slabs of the same material and remained very approximately proportional one to the other for all the values of examined .
In this way condition ( 6 ) is satisfied independently of the wave-length .
It follows from ( 5 ) , using ( 6 ) , that the relative retardation is given by . . . .
( 7 ) .
Accordingly the relative retardation , after the light from such a point-source has traversed the two is the same for all the rays from S. Two such slabs are therefore optically equivalent to a single slab which would be under an accurately uniform tension .
By properly adjusting the differences of height , , the amount of relative retardation may be adjusted within certain limits .
In general , will be chosen small with respect to .
Thus is a proper fraction , of order .
Equation ( 6 ) thus shows that and are to be chosen of opposite signs , and approximately equal in magnitude .
This gives at once the physical explanation of the result ( 6 ) .
The rays pass through approximately horizontally .
If we compare two rays passing through at different levels , the ray which passes through the regions of greater tension in passes through the regions of lesser tension ( or greater pressure ) in , and the two variations balance one another .
Further , since the amount of relative retardation as given by ( 7 ) involves only the relative heights the axes of the slabs and the source of light , the latter may be moved parallel to the axes of the slabs without affecting the relative retardation .
Hence a horizontal lin -source , parallel to the axes of the slabs , may be used instead of a poin -source .
This was , in fact , indispensable in order to obtain the required intensity .
S3 .
Description of the Apparatus .
Light from an arc lamp was passed through a condensing lens and through a thin horizontal slit ( fig. 2 ) , which was placed from to 3 metres away from the Fig. 2 .
glasses and straining apparatus .
It was polarized by a Nicol , whose polarizing plane was roughly at to the horizontal , and then passed through the two slabs and F. These were adjusted so that their levels differed very nearly by centim. DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
.The two slabs were cut from the same piece of glass , and every precaution was taken to ensure that they should be as nearly identical as possible .
The dimensions of the cross-sections were practically the same , namely , in the notation of the last section , centims. and centim. The length of each slab was about 13 centims. Bending moments of opposite sign , in a vertical plane , were applied to these , so that the light passed through parts of the glass either altogether under tension , or altogether under pressure , according to the manner in which the bending moment was applied .
Of the method of applying such bendin moment a fuller account will be given below .
After emerging from the two beams the pencil of light traversed a Nicol , which was crossed with the Nicol P. It was then focussed by a cylindrical lens ( which consisted in practice of a glass beaker filled with water ) upon the vertical slit of a spectroscope and the spectrum observed in the usual way .
The condensing lens was focussed upon the Nicol ; both and were introduced in order to improye the illumination .
It was found otherwise that so much light was lost that only a very faint spectrum could be obtained , and this was useless for the purpose of the observations .
The latter consisted in measuring accurately the position in the spectrum of the black bands corresponding to light completely quenched between the Nicols and A. Light of any colour will , of course , be quenched .
between crossed Nicols when the relative retardation of the two rays ( polarized in horizontal and vertical planes respectively ) in which it is split up by the strained glass amounts to an integral multiple of the wave-length .
Referring to formula ( 7 ) this occurs when . . . . .
( 8 ) , being an integer .
If were independent of the wave-length , as WERTHEIM'S law would require , then , for a band of a given order , is fixed and the wave-length of extinction is proportional to If , however , varies with , then is oportional to By observing the values of corresponding to a given , and varying , we obtain the relative magnitude of the coefficient for these varying values of The probable error of setting on the centre of a black band was calculated by the author in the 'Camb .
Phil. Soc. Proc vol. XII .
, Pt. V. , pp. 314-315 , and was found to be about 1 ' , so that the wave-length of extinction is determined with a propor- tional error smaller than .
The average error due to inaccuracy in setting the cross-wires in the eye-piece upon the centre of the black band is then about 8 to 10 tenth-metres , so that the wave-lengths may be considered known accurately to three figures .
The bending moments were applied to the slabs by means of the apparatus shown in fig. 3 .
The slab rested on two knife-edges and S. On it rested two other .
knife-edges and V , supporting a graduated steel bar I. Fixed to the top of I was a triangular knife-edge , from the pro .
ecting extremities of which hung two Fig. 3 .
symmetrical hangers A. These passed through holes cut in the bed-plate , which supported the whole apparatus , and by means of a cross-piece and another hanger a load was applied which acted on I vertically downwards at its middle point .
In order to ensure that the reactions between and the knife-edges should be vertical , the knife-edge rested on steel bicycle balls , so that it would readily move under horizontal friction ; V was made a double knife-edge , the plane containing the two edges being carefully adjusted to be vertical .
The knife-edge could slide along I and be clamped in any desired position .
When was clamped and the load applied , the apparatus was perfectly stable , the knife-edge being kept in its place by the friction of the bed-plate .
In order to prevent the knife-edges from cutting into the glass and breaking it under the large loads applied , four small slips of steel were inserted between the knife-edges and the glass .
These distributed the actual stress without altering the actual statical resultants , and at points near the centre of DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
the slab the effect of such local perturbations must be negligible .
( See ' PhiL Trams , vol. 201 , pp. 114 , 145 .
) When the apparatus is in perfect adjustment is exactly midway ( measured horizontally ) between and V , and the horizontal distances between the edges of V and and of and are equal .
If these be each , then the bending moment applied to the part of the slab between and is constant and equal to For , since the reaction at the upper edge of V is vertical and the load at is vertical , then the reaction at the lower edge of is shown to be vertical by considering as one system .
Thus the reactions at the lower edge of and V are each equal to .
Again , the reaction at the upper edge of is vertical and therefore the reaction at the upper edge of is also vertical .
Hence these reactions also are equal Also it is to be noted that , if the adjustment bs perfect , the bending moment applied to the beam or slab is a pure bending moment .
There is no total shear across any cross-section between and S. In such a case it is well known that the distribution of stress obeys accurately the Euler-Bernouilli laws and consists only of a tension parallel to the axis of the beam , where applied bending moment , height above neutral axis ( horizontal line drawn through the centroid of the cross-section in the plane of the cross-section ) , area of cross-section , radius of gyration about neutral axis .
The formulae ( 4 ) and ( 7 ) are therefore verified .
In fig. 3 the knife-edges and V are outside and S. The bending moment is therefore positive , with the convention of p. 266 .
For the second beam the arrange- meant is the same , except that now .
and V are inside and , so that the bending moment is negative .
The difference of height between the slabs was obtained by placing the knife-edges for one of the beams upon two steel blocks of height centim. instead of directly upon the bed-plate .
The bed-plate itself was a solid plate of steel , very strong and resting upon two heavy tables of the same height .
In the above description no account has been taken of a large number of small errors which must theoretically affect the method .
The principal are the following:\mdash ; ( l ) In the theory explained in S 2 modifications will be introduced owing to the fact that a polarizing Nicol is introduced in the path of the rays of light between the source and the slabs .
( 2 ) The source of light is not a line-source , but a slit of finite breadth .
( 3 ) When the load is applied , the middle part of one beam rises and the other sinks : thus the heights and the relative height in formula ( 7 ) are not fixed .
( 4 ) The bed-plate and the tables are not absolutely rigid .
This will alter and , but not .
( 5 ) The rays do not through the glass horizontally and at right angles to the axes of the slabs , and the assumption that the mean retardationretardation at mid-point of path is only an approximation .
( 6 ) The slit used as a source of light is not accurately horizontal .
( 7 ) The knife-edges are never quite accurately adjusted .
( 8 ) The weight of the beams themselves will affect the stresses .
( 9 ) The beanoe are not always perfectly annealed and the permanent stresses in the glass modify the appearances .
In the following sections the corrections.due to these errors ] be inyestigated .
S4 .
Effect of Introducing the Polarizing Nicol .
We shall now consider the effect of introducing the polarizing Nicol upon the inclination of the rays of light .
In order to estimate the magnitude of this effect , it will be sufficient to treat the Nicol as a singly-refracting substance .
If the larger index of refraction be adopted this should , in general , give us an upper limit to the error introduced .
If no sensible disturbance is found to be thus introduced , we may assume that this will be the in the actual experiment .
Let ( fig. 4 ) be the source of light , the image of a point of the mid-plane of the Fig. 4 .
nearer slab , viewed by refraction through the face nearest S. If the Nicol were not present the light would travel along the line .
In consequence of the introduction of the Nicol it travels along the broken path SCDP .
Let be the angles of incidence and refraction , the angle which SP makes with the normal to the faces of the Nicol .
Let the perpendicular distances of and from the nearer faces of the Nicol be and , and the thickness of the Nicol be Then or , writing ( 9 ) .
Using Hence ( 9 ) becomes , retaining only first powers of ( 10 ) .
DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
Now in the experiments and the maximum variation in for rays passin through the slabs amounts only to But from ( 10 ) ( 11 ) .
Accordingly the greatest variation in for the rays passing through the slabs is less than . . . . .
( 12 ) .
To compute the order of this expression take and , which last is an estimate .
( 12 ) gives S that is , S(x-(b ) about .
The change in relative level of the points of fig. 1 due to the above amounts to , and in the experiments centims. roughly .
Thus the change in relative level centim. The proportional change iri the total stress change of relative level -height of slab , and this will produce a negligible error in the stress-optical coefficient .
Accordingly for mono- chromatic light the effect is to increase the obliquity of all rays by a small constant amount , or to change the effective height of the slit .
As the absolute value of is not known , and will be found not to enter into the calculations , the presence of the Nicol will not affect the observations for monochromatic light .
For white light , however , it may do so if the quantity vary sensibly with the colour of the light used , other remaining the same .
From ( 10 ) S -S sin Now for calcite Taking the ordinary index as the basis of computation and as before , S tan nearly .
The proportional error in deduced from this is obtained by multiplying by , i.e. , by 27 .
It is therefore .
This error corresponds only to the VOL. CCVII.\mdash ; A. 2 dispersion between the lithium and thallium lines .
The error will be greater when we reach the violet end of the spectrum , but it will still be too small to affect the observations .
In experiments demanding great accuracy it might be desirable to polarize the light before it passes through slit .
The accuracy possible under the present circumstances did not seem to justify this additional complication in the apparatus .
S5 .
Effect of Finite Breadth of the So , urce .
It may be shown that if the slit have a finite breadth , the intensity of the light that gets through is proportional to where being the relative retardation corresponding to the middle of the slit .
In the actual case approximately .
The minima are given approximately by .
The proportional error in the wave-length of ext-inction is therefore , which is negligible .
Also the minimum no longer corresponds theoretically to perfect darkness , but with a slit between millim. and 1 ] .
wide the bands were very dark and quite definite .
S 6 .
Effe ct of Rise and Fall of the Two of Elastic Yielding of the Bed-plate and Knife-edges .
Owing to the elasticity of the glass , the middle parts of the beams will undergo a vertical shift owing to flexure , and the -plate and apparatus as a whole will sink .
In consequence we have variations , Sh depending on the applied load .
Thus the right-hand side of equation ( 8 ) is multiplied by a factor This may be allowed for by supposing ( or W ) multiplied by the same factor , equation ( 8 ) remaining otherwise unaltered .
The effect is then to add to the applied weight a correction Now the relative rise and fall of the beams themselves is an elastic effect and may be taken , in sucl ] a small correction , strictly proportional to the load .
DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
The sinking of the bed-plate was measured experimentally and found to be elastic in its nature , the recovel.y being complete .
Generally the experiments showed trace of permanent set , the readings being the same when unloading as when loading .
We may safely assulue that is p.ptional to , so that the correction to be applied to on account of these errors is of the form The value of is uncertain and depends very largely on the circumstances of each experiment .
Using and AMAGAT'S values of YOUNG'S modulus for glass ( i.e. , between 600,000 and 700,000 .-weight per square centimetre ) , the part due to relative rise and fall of two beams was calculated to be ) .
Thus for kilogs .
the proportional correction is as high as 2 per cent. The part of due to the sinking of the bed-plate was found experimentally to be of order .
Also the could be arranged in such a manner that the two corrections operated in different senses ; and this precaution always taken .
Their combined effect will give of order , and even for the highest loads used the correction will be small .
In practice this correction was determined from the observations themselves , in a manner explained in S 17 .
For most sets of observations it was found to be insensible .
S7 .
Influence of Obliquity on rdation .
We may consider the glass as optically made up of a series of horizontal homogeneous layers .
In passing from one of these layers to another , the refraction takes place approximately in a plane perpendicular to the optic axis .
It will be sufficient for our purpose to consider a ray passing through in a crosssection , that is , in a plane throughout perpendicular to optic axis .
If the curvature of such a ray be igible , we may take it that we can neglect the curvature of all oblique rays .
Now if be the index of refraction at a point in the glass distant from the neutral plane , being the index of refraction the unstrained glass .
It may then be that the curvature of a ray passing through nearly horizontally is approximately Now it has been shown by KERR ( ' Phil. Mag October , 1888 ) , and by POCKELS Physik , , that the absolute variation due to stress in the index of refraction for either ray is of the same order as the difference in the two indices due to the same cause .
In general , for the highest stress employed , the -gradient is of the order .
Takin , the curvature is of order Hence the greatest deviation from the straight line ( curvature ) ( thickness of slab ) divergence which cannot possibly affect the ] Thus we are justified in treating the paths of the rays as linear .
Moreover , the divergence of the ordinary and extraordinary rays after refraction at entrance ( of incidence ) very nearly , and it is easily verified that the effect of this is also entirely negligible .
Therefore we may treat the two rays as geometrically coincident .
The paths of the rays being linear , the planes of polarization are fixed throughout .
For these can be proved to be the plane through the ray , and the line of strain and the plane through the ray perpendicular to the first plane .
And the line of strain is always parallel to the axis of the slab .
Also if angle between ray and line of strain , the relative retardation introduced by an element of path is CT Hence the total relative retardation is where tension at of path , 2thickness of slab , inclination of ray to the horizontal perpendicular to the axis .
In practice the limits for are , and for are .
It follows that the factor introduces a proportional error less than in the relative retardation .
It may therefore be altogether neglected .
S8 .
Ytbined Effect of Flexure and Obliquity .
The relative vertical displacement of the two slabs due to flexure varies with the cross-section taken .
Now the pencil of rays used passes through a comparatively large region of the beams , extending to about 1 centim. on either side of the central cross-section .
It is readily shown that the changes in , due to flexure as we pass from the central cross-section to sections distant from central one are given by the slab being bent concave downwards and concave upwards .
greatest possible change in , due to this cause , is numerically equal to or taking kilogs.weight square centimetre ) The proportional correction in the stress amounts to nearly , i.e. , to DISPERSION IN ARTIFICIAL ) OUBLE REFRACTION . .
Thus for the extreme load of 50 kilogs .
it is only , and may be disregarded .
S9 .
Imperfect Adjustment of the Inclination of the Slit .
If the slit be not horizontal its different parts will act as different sources of light at different heights It is clear that if the inclination be too great the different parts of the slit will give different dark bands in the spectrum , all overlapping .
The integral band will be diffuse in consequence and not readily measureable .
It is quite easy , however , to make this adjustment to a nicety , as ] Let AB ( fig. 5 ) be the slit , the image of AB in the cylindrical lens ( of fig. 2 ) for rays proceeding in a horizontal plane .
Then each element of the slit gives a vertical line of light through .
Let be the opening of the slit of the spectroscope .
The latter is a good deal smaller than the image , so that in practice only a moderate of the luminous slit is used .
If now the cylindrical lens be moved to one side or the other , so that travels from one end of to the , then , if the luminous slit be not horizontal , the band shift in the spectrum in consequence .
When no such shift occurs , we know that the adjustment is very exact .
There is very little difficulty in making this adjustment , and accordingly there is no reason for anticipating any sensible error from this S10 .
Imperfect ustment of the Knife-edges .
In practice it is impossible to ensure that the two pairs of knife-edges shall be exactly symmetrical with regard to the vertical through the load .
Failure to satisfy this condition introduces shearing stresses in the beams , so that the axes of polarization are no longer horizontal and vertical and further the bendin moment varies from cross-section to cross-section .
The complete analytical investigation of the correction in this case is long and difficult , but the results may he summed up as follows , for the simplest case , when only one of the slabs is supposed imperfectly adjusted .
In general there is no longer perfect extinction , so that the band is not quite black .
Assuming that the " " overlap\ldquo ; of the two slabs is half their height , the position of the band for rays passing through the edge of either slab is unaltered .
The position of the band for a ray passing exactly at mid-level is shifted towards the red end of the spectrum by an amount not exceeding of a tenth-metre .
Thus , remembering that rays which have passed through the glass at different levels correspond to parts of the spectrum also at different leyels , we see that the band is no longer straight and vertical , but curved , the convexity bein towards the red .
This convexity is , however , so small that it would not in any case be observable .
If the condition ( 6 ) of S2 is not exactly satisfied the band will still be straight , to a first approximation , but no yertical .
Thus when the bending moment varies from cross-section to cross-section for light passing to the right of the mid-section the band is tilted one way , for light passing to the left it is tilted the reverse way .
The integral effect will be that the thickness of the band will no longer be uniform , but the band is still symmetrical with regard to a vertical line , corresponding to light going through the mid-section .
The settings which are made on the middle of the band are therefore unaffected .
S 11 .
Imperfect Adjustment of the Knife-edges .
It will also happen that the knife-edges will not all be at exactly the same height , so that the axes of the two slabs are not exactly horizontal and parallel .
The effect will be that for rays passing through in a plane distant from the central section instead of being exactly constant , being a small coefficient .
This will broaden the band and render it more diffuse , but will not shift its centre .
Observation shows that this effect must be very small , as , in general , the band is very well defined .
S 12 .
Error due to Weight of Beams themselves .
In computing the stresses no heed has , so far , been paid to the fact that the weights of the glasses themselves will introduce certain stresses in the slabs .
The weight of each slab is on the average 120 grammes .
This , although very small compared with ] the total load in most cases , may introduce a small error in the case of the band of the first order , which corresponds to a smaller load .
For the beam the weight of the glass was found to introduce practically no bendin moment in the centre , as the supports were very nearly at the quarter and three-quarter span points .
For the beam the moment introduced is the same as if the weight on this slab were increased by exactly its own weight .
It is quite easy in practice to eliminate this by adding a small counterpoise to the weight on N. S13 .
Error due to Imperfect We now come to the only error\mdash ; with the exception of that due to rise and fall\mdash ; which is sufficiently important to be allowed for in the reduction of the observations .
DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
The annealing of the glasses used , which were supplied by Messrs. of Jena , was found to be by no means perfect .
In some cases was revealed even by a cursory inspection between crossed Nicols .
In other cases , the glasses being unloaded , a one-wave plate of ] enite was introduced between the Nicols , its axes being horizontal and vertical .
This showed a black band , on the same principle that the strained glass shows such a band .
Now if the glasses had no residual stress the relative retardations should be the same when the azimuth of the axes of polarization of the selenite plate is altered by If there be residual stress , however , it will affect the light difierently in these two cases and the band will be shifted .
In most cases the existence of such a residual stress was exhibited very plainly by this method .
As a rule the band due to the selenite plate was straight and vertical , showing that the residual stress was fairly constant .
If be the three components of residual stress in a vel.tical plane parallel to the axis of the beam , then the axes of polarization make an angle with the horizontal , where and the principal stress difference If we neglect squares of it is easy to calculate that the retardation has to be increased by , or , taking in these corrective terms , the retardation must be increased by and this is equivalent to putting in a constant correction to W. PART II .
EXPERIMENTAL RESULTS .
S14 .
Observed .
The glasses used in this research were made for me by the firm of ZEISS in Jena .
The makers being unable to communicate to me the chemical composition of the glasses , the latter were analysed for me by Mr. W. J. REES , on the staff of Messrs. CHANCE BROS .
To Mr. ' skill I am indebted for the following results:\mdash ; The majority of these glasses belong to the borosilicate variety , excepting 2783 is a flint glass , and was stated by the makers to be identical in composition with another glass , , the composition of which ( see 'Camb .
Phil. Soc. Proc vol. , Part V. , p. 314 ) was stated by Messrs. ZEISS to include and It seems probable that the composition of the later glass is a little different to that of S 15 .
Linear connectinq the Stress .
Since it was known beforehand that corrections to of the type would have to be applied , being due to the imperfect annealing , and to relative and absolute rise and fall , instead of calculating the stress-optical coefficient directly , as was done in previous experiments , the relation between and was first studied , with a view to disengaging the corrections .
In practice , readings for and were taken for both first and second orders of the band , and even , where possible , for third orders\mdash ; for both tension and pressure .
The tension and pressure observations were obtained by altering the relative heights of the two slabs by interchanging two steel slips which raised the supports of one of the slabs .
The bending moments were not altered .
A typical set of results is embodied in Table I. below .
DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
TABLE I.\mdash ; Observations of Glass 1809 .
VOL. CCVII.\mdash ; A. 2 The refer to observations for tension : to observations for pressure .
In the columns headed are placed the observed weights , the suffix indicating the order of the band observed , bands in pressure observations being taken of a order .
The same notation will be kept throughout .
The colurnns eaded .
contain the observed value of the wave-length of the light quenched , in etres .
They are deduced from circle readings of the spectroscope , the law connecting these circle readings with the wave-lengths being obtained from tions of a known comparison spectrum .
The spectrum of the arc between rbons soaked in calcium salt was used this purpose .
Now , if we look at Table I. , A. , under the heading , we see that the differences of observed for unit differerlces of have a fairly constant average value , as is well ] on taking diffel'ences corresponding to differences of of three units .
This is done in the column headed .
of Table I. , A. It wonld seem , therefore , that the relation between and is approximately linear .
This impression is found to be confirmed when differences are taken in Table I. , .
In each case the differences are sensibly constant , especially if we bear in mind that an experimental error of 10 tenth-metres is to be expected .
There are some local inequalities , ] of which will be shown later to be probably significant , but as a first approximation it seems we may assume a linear relation between and W. Fig. 6 shows the observed plotted to for the set of observations of Table I. , The observations obviously lie very close to the straight line given by the equation This equation was obtained by assuming a formula . . . . .
( 13 ) .
was obtained by taking the mean value of the differences in the columns headed .
of Table I. , , and was then determined from the condition that the best straight line must pass through the centre of gravity of the observations .
The equation ( 13 ) clearly leads to the relation or , being the relative retardation for a band of order , , where effective tension , thickness , a constant independent of the wave-length .
DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
Thus , the stress-optical coefficient ( unit stress ) is given by the approximate formula . . . . .
( 14 ) , or . . . . . .
( 15 ) .
The curve connecting and is therefore a hyperbola .
When , and as increases without limit , decreases to its limiting value Best fitting line , Observations .
Line , Fig. 6 .
Typical diagram showing relation of to W. Thus , if the law continued to hold accurately for small , then for light of the critical wave-length the stress-optical effect would become actually infinite .
No doubt this law is purely empirical so far , and will very probably not hold for very small wave-lengths .
It is , however , cient to indicate that , as we approach critical values in the ultra-violet , the stress-optical effect will very likely be largely increased .
In order to show the accuracy with which the observations above determine the value of , the straight line passing through the origin and through the centre of gravity of the observations has also been plotted on fig. 6 .
Its equation is and on looking at the diagram it is obvious that no such straight line can fit the observations .
Further , it will be shown that all the observations , merely of the ] 1809 , but of the six other glasses examined , conform to a first approximation to the linear law .
S16 .
iqnifccm c , of Linear Law .
We have now to enquire how far this linear law has a physical meaning otherwise than as the expression of the trivial result that within a certain range of values all continuous variation is approximately represented by a straight line .
In previous papers , when and were the quantities plotted , the relation was not well expressed by a straight line , the observations lying , in some cases , on a very decided curve ( see ' Camb .
Phil. Soc. Proc. , ' vol. XII .
, Part V The obseryations which led to such curves were therefore re-reduced .
The glasses selected which showed the effect most strongly were the Jena lead glasses and Both of these were very closely fitted by a linear relation between and W. How close the fit was may be inferred from Table II .
below , which shows the observed and calculated values of for no set of experiments with , which is a very heavy flint glass , containing 80 per cent. of In the table , denotes pressure in kilogrammes applied by means of a compressing apparatus described in the paper referred to , and the entries in the column headed .
are computed from the formula The other sets of observations of and , which have been re-reduced , show equally good agreement between the observed values of and those calculated from a formula of type ( 13 ) .
Now the mean residual in Table II .
is less than 5 tenth-metres , whereas the probable error of determination of the centre of a band is about 10 tenth-metres .
Thus the law appears to fit the observations as closely as is possible within the limits of experimental error .
It is worth noting that with these glasses , which contain a high percentage of lead , no deviations from the law , such as will be shown later to take place in some borosilicates , appear to exist .
Nevertheless one important experimental fact throws doubt on the universal validity of the linear law , even for lead glasses .
POCKELS has shown ( ' Ann. DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
ns of ( re-reduced ) .
Physik , ' 1902 , p. 745 ) that for a glass containing between 60 and 70 per cent. of the stress-optical coefficient changes sign , and an experiment made by him with such a glass pointed to the fact that the stress-optical coefficient did not vanish simultaneously for all colours , a result which has been independently confirmed by the present author from considerations of curves showing and plotted to percentage of lead Camb .
Phil. Proc. .
Now if the law held universally , the vanishing of would imply the vanishing of for every wave-length .
Moreover , it seems impossible to find theoretical justification for such a formula .
It is well known Theory Optics , that the index of refraction is given by the formula , where -length in of light belonging to one of the natural periods of the glass .
added July 3rd , 1907.\mdash ; Throughout the paper I have followed DRUDE .
But if adopt LORENTZ'S formula , .
:\mdash ; or similar formulae , the most essential part of the reasoning remains in most cases practically unaItered .
] It seems , at first sight , highly probable that the effect of stress will be , not to introduce different free periods of the atoms for differently polarized rays\mdash ; that is , not to alter \mdash ; but to change the coefficient , which depends on the number and arrangement of the molecules .
This will lead the result S2 or ( 16 ) .
Now itself , when expanded in powers of , will involve terms in , etc. Therefore will involve only such even terms .
Hence no formula involving to odd powers can be theoretically acceptable .
If we suppose that one term , corresponding to wave-length , is active in producing the dispersion , both in ordinary refraction and artificial double-refraction , we have . . . . .
( 17 ) , . . . . .
.(18 ) .
The formula ( 17 ) is open to the same objection as , namely , that it does not satisfy the case of a glass where the double-refraction vanishes for one wave-length without the dispersion vanishing at the same time .
It is clear that in this case other free periods , whose effect is usually negligible , become important .
For other glasses , however , the formulae ( 17 ) and ( 18 ) might be good approximations .
To get from ( 18 ) remember that for wave-lengths greater than 4300 the dispersion terms are of the whole .
Then , using the Binomial Theorem , we find that , to an accuracy of nearly , Hence ( 19 ) , where ( 20 ) .
A formula of the type ( 19 ) for would lead to a curve connecting the wave-length of extinction and the load of the type In general , when and are small , it will be found that either formula , represents the observations almost equally wel ] .
DISPERSION IN ARTIFICIAL UBLE REFRACTION .
But in the case of the results of Table II .
a hyperbola was fitted to the observations , equaticn being The differences between as calculated from the above formula and observed are yiven in Table II .
under the heading .
The mean value of the residuals baken without regard to sign is between 13 and 14 , or nearly three times the value the mean residual from the straight line .
Even this mean residual hardly exceeds the probable error of observation , so that this would not be conclusive against the hyperbola .
But examination of the individual numbers in the last column of Table II .
shows strong systematic positive esiduals in the middle and negative esiduals at the ends , and these systematic divergences certainly suggest that the hyperbola is not the most suitable curve .
The index of refraction of this particular glass is tolerably represented for the visible spectrum by the formula Thus From this of formula ( 20 ) comes out to be .
This differs entirely from the value obtained from the experiments , namely 3460 .
We are thus led to the interesting conclusion that in this glass at least the free periods which produce the ordinary dispersion are probably not active in producing the dispersion of artificial double refraction .
This removes theoretical justification in this case for the formula even if it had not been shown inferior as a purely empirical fit .
We may then provisionally accept the law , and the results in what follows will be reduced with reference to it .
At the same time it must be remembered that the physically significant formula is probably of type ( 16 ) .
It will be shown in S 21 that even in the visible spectrum there are local divergences from the linear law .
S17 .
Methods of Reduction .
In fitting a linear law to a set of observations , the corrections due to the sinking and permanent stress had to be taken into account .
The correct value for the load is given by It is therefore a formula which has to be fitted .
Assumin for the present that for monochromatic light the relative retardation is strictly proportional to the load , we have , for the band of the order , , where is the same constant for observations of all orders .
Thus Suppose we have observations , the first step is to take a number of differences corresponding to differences sufficiently large to minimise the effects of accidental irregularities and to form the fraction the suffix denoting that the band observed is of the order .
If our observations correspond to values of differing by a constant increment , and if we take differences of corresponding to differences , we obtain equations\mdash ; , , .
If these be added up , we have , where mean value of W. We have then ( 21 ) .
A comparison of the values of then enables us at once to discover whether a correction is needed for the obseryations or not .
For most of the glasses examined the values of do not indicate a correction of this type of sensible amount .
In doing the reductions for such glasses has been taken equal to zero .
For one glass has a sensible value .
In this case a suitable value of having been DISPERSION IN ARTIFICIAL DOUBLE ACTION .
formed , corrections were applied to the obseryed load and the observations re-reduced with From this point onward , therefore , may be taken zero in the reductions .
is then equal to In practice the values of vary slightly with different .
In most cases , however , a sufficiently good fit is obtained by taking for the mean value of and reducing the observations of different orders by means of this single value .
In one glass HOOKE'S law did not seem to hold quite exactly , and the observations of different orders were reduced independently .
having been determined , is found from the condition that the best fitting straight line .
( 22 ) must be satisfied by the mean values We thus obtain equations etc. Two of these equations are theoretically sufficient to determine and .
In practice three are often obtained .
The three equations are then solved by least squares .
The solution is given by From the values of so determined has been computed from the formula ( 22 ) and compared with the observed value .
S 18 .
Tables of Results .
The following gives a table of the constants for the various sets of observations .
Observations corresponding to tension and pressure are distinguished by the letters respectively .
When are known , the wave-length of the band of order is computed from the load by the formula ( 22 ) .
The average discrepancy in tenth-metres between thus calculated and observed , for each set of observations , is entered in the column headed ( O-C ) .
VOL. CCVII . .
2 TABLE III .
For the glass 935 a correction was applied to being taken for pressure and for tension .
In ( 22 ) we have then to substitute for W. This glass is badly annealed and does not seem well fitted by the formula .
The glass had to be reduced differently .
This is a lead glass , a specimen of which had been examined under simple pressure and whose behaviour had appeared peculiar ( see 'Camb .
Phil. Soc. Proc vol. XII .
, Part V. , p. 323 , where the glass in question is described as ) .
There are two sets of tension observations , denoted by in Table IV .
, and two sets of pressure observations denoted by , D. The values of ( see S 17 ) are given in Table IV .
DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
In this case there seems to be a progressive increase of with the order , for both tension and pressure .
This excludes a correction for sinking , since the latter must act opposite ways for tension and pressure .
It is here probably due to a failure of HOOKE'S law , which the observations have shown otherwise , and which will be discussed in a later section .
Under the circumstances no real advantage could be derived by attempting to reduce the various sets by means of a single formula .
The sets have , therefore , been independently reduced , using the formula .
The actual observations of all glasses are given in Table V. for purposes of reference .
Each column corresponds to a single set of observations .
As a rule the order of the band observed will be clear from the place of the observation in the series .
Wherever this is not so , or where observations of different orders correspond to the same load , Roman numerals have been added to indicate the order of the band .
TABLE V. DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
S 19 .
Discussion of the Values of The first thing which strikes the eye on looking through the results of the last section is , that although tension observations of different orders and pressure observations of different orders are fairly well fitted by the same , and the same does not hold of tension and pressure observations taken together .
The differences in are only what should have been expected , since depends on the adjustments .
With regard to the values for pressure and tension should theoretically be equal and opposite .
For if light traverse a thickness of glass in which a residual tension exists , a term is added to the relative retardation when external tension is applied and subtracted from it when external pressure is applied .
Again should be a constant the glass : and therefore the same for tension and pressure , if the stress-optical coefficient be independent of the nature and magnitude of the load applied .
Now Table III .
shows clearly that , although the values of differ in sign , they are only very roughly of the same order of magnitude .
Possibly this might be accounted for by the fact that in different experiments the light did not pass through the same parts of the glass , so that the value of the residual stress might have been different .
The divergences in are considerable ; appears systematically larger for pressure than for tension .
* S20 .
Systematic The residuals have in all cases been plotted on a large scale against Three of these diagrams are shown in figs. 7 to 9 .
The pressure and tension observations have been plotted to a different base in each case , to avoid the diagrams overlapping , so that two zero-marks appear on each scale of residuals .
Of these the upper zero mark refers to tension observations .
Most of them are irly irregular , which is not surprising when we bear in mind that an error of 1 division in the ordinate ( 10 tenth-metres ) is the probable error of the observations .
Two glasses , however , 3296 and 3453 , figs. 7 and 8 , appear to show very strongly systematic residuals between 4200 and 5500 .
If we look at fig. 7 we notice that the curves rise from 4200 to a peak about 4700 , after which there is a sharp fall with a trough about 5050 .
After this the curves run fairly horizontal , with indications of another peak at 6300 .
Note 3rd , 1907.\mdash ; Later expeliments do not confirm the systematic difference betwee11 the values of for tension and pressure .
Very probably the divergence previously noted was due to a change in the adjustments which had to be made when passing from tension to pressure , and rendered the observations of the two kinds not strictly comparable .
The true value of appears to be the mean of the values obtained for tension and pressure .
] The course of the diagrams in fig. 8 is very similar .
There is a well marked peak about 4700 , followed by a depression in the neighbourhood of 5000 .
There are also slight indications of a depression at 6000 , with a subsequent rise .
Ist obs .
IInd , , \mdash ; IIIrd Tension obs .
Pressure , , Fig. 7 .
Glasses 3296 .
Diagram of residuals .
These systematic residuals , which are in most cases quite large , and which are shown in the same place by all the tension and pressure observations of these glasses , cannot be chance effects , Neither can they be affected , denoting , as they do , comparatively rapid in , by any of the slowly varying corrections which have been discussed .
They can be accounted for only in the following ways : ( 1 ) Possible erroneous identification of a spectrum reference lin in the neighbourhood and consequent wrong determination of observed ; ( 2 ) Bad division errors of the spectroscope circle ; ( 3 ) Bad error in some of the weights employed at these points ; ( 4 ) Systematic change of personality of observer in this neighbourhood , due to change of colour ; ( 5 ) Actual variation of the ] of stress-optical effect in this neighbourhood .
( 1 ) is ruled out by the fact that in the glass 3453 , where the effect was first DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
noticed , it was discovered , not from any such curve , but from the actual circle readings of the spectroscope , which usually increased by steadily increasing ences ; in this glass , just after the readings corresponding to , the difference decreased instead of increasing .
Readings were taken several times with great care , and the effect was confirmed in each case .
This demonstrated that the cause was not idental .
4500 5000 Ist order ) Tension obs .
IInd , , , , Pressure , , Fig. 8 .
Glasses 3453 .
Diagram of residuaIs .
Clearly a wrong deternunation of a reference line is out of the question ; this could not cause irregularity in the differences of circle reading .
With regard to ( 2 ) and ( S ) , the weights and the divisions of the circle were tested with great care and found correct .
As to personality , the jump of 40 tenth-metres between 5000 and 4800 would require an error of 6 ' in locating the centre of the band , and a change of personality of this amount , in a fairly bright region of the spectrum , is unthinkable .
Besides , if the effect is due to such a change , it should appear in all the glasses , which is not the case .
Under these conditions it seems safe to assert that between wave-lengths 4500 and 5500 there exists a definite deviation from the straight line law , which deviation takes the shape of an undulation , with a crest at about 4700 and a trough between 5000 and 5100 .
To trace this effect more exactly , a new set of readings were taken with the glasses 3453 .
The readings were taken with special care at intervals ofload ofhalfa amme .
The residuals from the best straight line were computed as before , and they are Ist order obs .
Tension obs .
IInd \ldquo ; Pressure , , -.-.-.-.-.-lIIrd \ldquo ; Fig. 9 .
Glasses 3413 .
Diagram of residuals .
shown on fig. 10 .
Here , the observations being much more numerous , the residuals indicate a distinct curve .
The dotted curve in the figure has been drawn freehand through the points to give some idea of the general shape of this curve .
It is seen that this curve amply confirms the previous set of results , although the observations were take1l at several months ' interval , with different adjustments , and probably different personal equations .
On examination , only 3296 shows anything like so marked an effect .
1809 and 2783 show the effect in much the same place , but weakly , and some sets of observations do not confirm it .
935 and 3749 are hopelessly irregular .
Nothing definite can be asserted about them .
As to 3413 , the pressure observations do not show this effect at all , and the tension observations show it only very doubtfully .
DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
There are indications , however , in this glass , of a systematic dip at 5500 , and a subsequent rise .
Various glasses also show signs of a peak in the red , between 6200 and 6500 .
None of these , however , are more than mere indications , and it is only the curves for 3453 , and in particular fig. 10 , on which any safe deductions and measurements can be based .
If we refer to the table of S14 we see that 3453 and 3296 are very much alike in chemical composition .
Apart from this , no relation between this effect and chemical .
composition can be predicated .
It seems almost certain that boric acid has nothing to do with it .
The glasses richest in do not show the effect .
can hardly be the explanation , or 3749 should show the effect more strongly .
It seems not at all unlikely that a small impurity , such as magnesium or zinc oxide , .
may be the cause of the result .
It is noteworthy that the only glasses which seem to show the effect at all definitely , are precisely those which contain and , and that the one which really shows the effect in a measurable manner contains quite a respectable percentage of S21 .
Possible by Absorption Bands .
The shape of the curve of residuals resembles the curve of index of refraction plotted to either period or wave-length when we pass through an absorption band .
This suggests that the effect may be due to some faint or latent absorption band of the glass in the visible spectrum , which band corresponds to a period active in producing the artificial double-refraction .
Following , as before , DRUDE ( ' Theory of Optics , ' cap. V we have , being the index of refraction , and the co-efficient of absorption , terms not depending upon the absorption band . . . . .
( 23 ) , where is the wave-length of the absorption band , is a coefficient which increases with the absorption , and is a coefficient depending on the arrangement and number of the electrons .
This leads to Now suppose the stress to ] eave and unaltered , and to alter Q. whence , eliminating ( 24 ) .
VOL. CCVII.\mdash ; A. 2 To simplify the calculation , we may suppose that in the initial state of things , so that .
As a matter of fact cannot be , but the assumption that is good enough to give the characteristics of the phenomenon sufficiently well for our purpose .
We have then In all the above denotes rate of divergence with regard to of a quantity for the two oppositely polarized rays .
Thus value of if there were no absorption band .
Accordingly , gives the deyiation produced in the stress-optical coefficient by the absorption band .
Calling this SC , we have .
: .
CSC : C. Fig. 10 .
Diagram showing curve of residuals from straight line for glass 3453 .
Thus the deviation .
which is given by fig. 10 is are all comparatively slowly varying : the factor which causes the oscillation is . . . .
( 25 ) .
This factor starts with the value when , decreases to a negative minimum when , and then increases to a positive maximum when .
It then decreases down to 1 when DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
being here small with regard to , the wave-lengths of minimum and maximum are approximately .
Neglecting the variations of the other factors , this result gives us an easy means of obtaining from fig. 10 .
is the horizontal distance between the maximum and the following minimum .
is therefore between 300 and 400 tenth-metres\mdash ; say 350 ; from the same diagram is about 4900 .
This phenomenon also gives us experimental evidence in favour of a non-alteration of the period .
If we refer to the physical meaning of , we find it to be where period of the light corresponding to the absorption band , number of electrons in unit volume which vibrate in this particular mode , charge on such an electron , mass .
It follows that when we suppose , that is , to vary being initially zero , as before ) , what we have called SC is given by ( 26 ) .
Now , in the neighbourhood of being small , the rapidly varying terms which determine the shape of the curve are those involving .
If we call this quantity , and put in the other terms , we obtain some conception of the shape of the curve .
Taking . . . . .
( 27 ) , has one minimum value when .
It has two maximum values when or .
The general shape of the curve is shown in fig. 11 .
It is at once obvious that it does not show the alternate large maximum and minimum required to fit the curve of fig. 10 .
So far , then , the experiments bear out the hypothesis that the free periods are not altered .
Before we can proceed further we have to settle finally the convention about the sign of C. It has been usual to call positive for ordinary glass , such as that investigated by BREWSTER and KERR , and negative for heavy flint like This convention has been adopted by the author in previous papers .
We will now rigidly define the stress-optical coefficient as . . . . . .
( 28 ) , where denotes the index of refraction of the ray vibrating along the direction in which the principal stress ( tension being considered positive , as usual ) is T. Since the direction of vibration\mdash ; if we take the electric force to be the light-vector \mdash ; is perpendicular to the direction of polarization , and the index of refraction is inversely proportional to the velocity , this means that a positive implies a higher * Fig. 11 .
Diagram of the curve velocity in the ray polarized in the direction of greatest stress .
Now this ray corresponds to the ordinary ray if the glass be compared with a uniaxial crystal whose optic axis is parallel to the direction of greatest stress .
The glass , therefore , produces the same effect as a positive uniaxial so placed .
This is , in fact , what does occur .
This definition , therefore , agrees with the earlier one .
Now , since ( for this glass ) are all positive , it follows that if we are to have a maximum followed by a minimum , as in fig. 10 , Now , in , since does not change , the only quantity which can change with the polarization is , the number of electrons per unit volume .
is then , where are the number of electrons of period which respond to vibrations in the directions of and Our result will , therefore , signify that a tensional stress the number of electrons which respond to vibrations in the direction of the stress , relatively to the number responding to vibrations in a perpendicular direction .
In other words , tension appears to tend to set the electrons vibrating in a plane at right angles to the line of stress , pressure having the reverse effect .
It seems probable that if this effect were due to an absorption band , a glass DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
showing in its unstressed condition strong absorption bands in the visible spectrum would exhibit this effect in a most marked malmer .
A small slab of didymium glass was obtained and examined under direct pressure with the apparatus described in the previous paper in ' Camb .
Phil. Soc. Proc already referred to times .
The observations were not very precise , and it is hoped to repeat them with the flexure apparatus , when larger slabs can be obtained.ae Such as they were , however , they gave a negative result .
The band due to stress passed through the double absorption band in the orange without showing any marked irregularity .
It seems , therefore , that the didymium electrons which produce the absorption band are not affected by the stress in the way described .
It appears possible that the didymium particles really float , about , as it were , in the glass , in a free state , like particles in suspension in a fluid , and that they cannot be influenced to any great extent by stress applied to the glass .
Further research in this direction is in progress .
S22 .
Determination of Absolute of C. Although the experiments were primarily undertaken to show the dispersion effects , it is desirable to know also the absolute values of the stress-optical coefficients .
These are not given by the experiments as described in S2 , because the differences of level and cannot be measured with sufficient accuracy .
To determine absolute values a second slit is used , of which the height is The two slits were cut in the same diaphragm , so that is easily measured once for all .
Referring to formula ( 7 ) , the relative retardation due to the second slit , . . . . .
( 29 ) , where -optical coefficient of beam for the wave-length for which the retardation is Also .
( 30 ) .
Hence ( 31 ) .
If therefore we know the ratio we can find the absolute values of either .
If correspond to bands of order , Therefore added , 1907.\mdash ; Since writing the above the experiment has beon repeated under the more accurate conditions , and the negative result has been confirmed .
If the effect exists in the didymium glass it is certainly small .
] Assuming the law connecting and to be given by ( 14 ) , we have , where refer to the slab F. Hence Therefore ( 32 ) , and from equation ( 5 ) , .
( 33 ) , which give the absolute values of for both slabs .
A great many errors enter into the determination of these absolute values .
It is very difficult to measure the spans with sufficient exactness , and the differences are not large enough to allow of very accurate determinatiolL S23 .
Effect of Ch. Composition on Stress-optical Properties .
The mean values obtained in this way for each beam shown below in Table VI .
denote the two individual .
slabs of each pair , and is expressed in a unit equal to per kilogramme weight .
TABLE VI.\mdash ; Dependence of on Chemical Composition .
has been taken , and not the mean of the two values , because here certainly differs for slabs A and , and A was the slab analysed .
Lstimated from the values of on p. 291 .
DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
If we refer to the diagram published in a paper previously referred to ( ' Camb .
Phil. Soc. Proc. , ' vol. , p. 335 ) showing dependence of upon percentage of we see that until this reaches about 40 , has little influence on the stress-optical effect .
We are therefore to look at the two remaining principal constituents , and , for the cause of variations in Looking at Table VI .
, in which the glasses are arranged according to descending order of magmtude of , we see that the four glasses with the high all contain a percentage of which is considerable .
On the other hand , the three glasses with a low all contain a comparatively high percentage of .
We conclude that either raises , or lowers it , or both .
Also , looking at 2783 , which is the lowest of the seven , we notice that it contains the least percentage of , but not the highest percentage of .
This suggests that is more efficient in raising than in lowering it .
That the effect of must be predominant is otherwise evident from the fact that the order of the percentages of is the same as the order of magnitude of , with one exception ( 1809 ) .
The last column of Table VI .
shows the values of ( percentage of ) ( percentage of .
This places 1809 in the sequence , but throws out 3749 .
Also the marked difference between the four first and the three last is shown very clearly .
The glasses do not form a sufficiently regular series to enable us to go further and to determine exactly the law of dependence of upon the percentages of and of in the glass .
But the increase of certainly increases , and the increase of probably decreases it ; and seems to be at least twice as efficient as With regard to the mean values of , they all appear to be of the same order of magmtude .
At all events no definite dependence of upon the composition can be traced .
For percentages of not exceeding 35 and of not exceeding 25 it would seem that is independent of the composition of the glass or that the dispersion is , in every case , proportional to the double refraction .
S24 .
Failure of HOOKE'S Law for Glass 2783 .
Before concluding , the peculiar phenomena shown by glass 2783 require explanation .
This glass showed a ressive i with increase of load in for both tension and pressure .
Also when the load was increased the band , which was straight and vertical for moderate loads , became curved , being convex towards the red as shown in fig. 12 .
The load was eventually increased to about 59 kilogrammes when one of the glasses broke .
The band was obseryed when the glass was on the point of rupture and it exhibited a decided V shape as drawn .
This showed that the relative retardation was greater for light passing through at mid-level than for light passing through the edges of the beams .
Such an effect can be explained only in two ways .
Firstly , by supposing that the law of stress across the section does indeed remain linear , but that the relative Fig. 12 .
Appearances of band for glasses 2783 .
retardation is not proportional to the stress , increasing less rapidly for large stresses than for small , since here ( the overlapping being assumed for simplicity to be height of beams ) produces more effect than T. Such an effect , however , would imply an increasing falling off of the values of from the values which would be obtained if the linear law held .
The result would be a progressive , instead of a progressive , in Secondly , we may suppose that HOOKE'S law fails or stress is not proportional to strain .
Consider a beam bent under constant bending moment .
The axis will take the form of a circular arc .
If we imagine the circle completed , then the symmetry shows that the cross-sections must remain plane and all pass through the centre of the circle .
It follows easily that , whatever be the law of stress , if we can neglect end effects , the extension follows a strictly linear law .
Fig. 13 .
Thus if AB ( fig. 13 ) represent the vertical axis of such a section , and the stress at any point in AB be set off as RP perpendicular to AB , the locus BPC of is a DISPERSION IN ARTIFICIAL DOUBLE REFRACTION .
,305 genuine stress-strain diagram .
It will therefore , as is well known , take the shape in 13 , the stress falling offf rapidly as the strain approaches.and passes what is known as the yield-point .
The corresponding curve for the overlapping part of the other bean ] is shown by AQD , and the stress effective in producing the optical effect is the sum of the two ordinates .
It is obvious from the figure that PQ is a maximum in the middle .
Thus the peculiar shape of the band is likewise accounted for on this hypothesis .
Now let the straight line BI give the stress-diagram which would be obtained for the same moment if HOOKE'S law held .
This straight line and the curve must then be so related that the first moment about BX of the areas APCB and AIB are equal .
Let be the extremity of the ordinate through the of AB .
Draw BL AC at and the tangent at cutting AC at , BX at U. The curve is always convex inwards , therefore CPLB always lies on one side of TU as shown .
Now the triangles BLU , TLJ are clearly equal .
Hence area And the mean distance from BX of the area CPLJ mean distance BX of the .
Therefore first moment of CPLJ first moment of BSL , or first moment of ABJiirst moment of ABLC .
Therefore , if ABI and ABLC have the same first moment , ABI , or BI must lie to the left of BJ .
If BI cut ML in , then .
That is , the actual measured stress is greater than the computed stress .
Therefore the observed values of exceed those that would be obtained if HOOKE'S law held by a difference rapidly with the stress .
The result is a progressive iucrease in the lue of such as is actually observed .
The discrepancies which have appeared are therefore explained .
Incidentally this confirms the conclusion ( which indeed seems highly probable on theoretical grounds ) that the stress-optical effect is dependent upon the stress\mdash ; that is , the molecular strain\mdash ; and not upon the molar strain .
The latter , which is the sum of both plastic and elastic effects , is the quantity measured in most extension experiments , and is usually denoted by " " strain\ldquo ; simply .
S25 .
This completes the account of the results reached so tar .
The next step would be to obtain glasses of suitable chemical composition to show the effects discovered in a much stronger degree and thus allow of more precise determinations .
Research in this direction is being undertaken , and it is hoped that the results will form the subject of a future communication .
VOL. CCVII .
In conclusion the author wishes to express his most grateful thanks for the assistance which has been rendered him by the Royal Society , without which this could not have been undertaken .
He desires also to place on record his indebtedness to Professor F. T. TROUTON , F.R.S. , who has most generously placed the resources of the Physical Laboratory of University College , London , at his disposal ; and also to Mr. A. W. PORTER , B.Sc. , whose interest in the work and kindly criticisms and suggestions have been invaluable .
I

