]\gt ; XI .
Double Refraction and Dispersion Spar : an t'estigation , voith comparison with HUYGHEN'S struction for the Wave .
R. T. GLAZEBBOOK , M.A. , Felloll ) of College , rnbridge .
ted by .
J. CLERK MAXWELL , M.A. , Received June 12\mdash ; Read June 19 , 1879 .
SECTION I. IN a paper read before the Royal Society , June 20 , 1878 , the results of an investigation into the truth of theory of double refi action in a biaxal crystal were stated .
The comparison between theory and experiment was made by a method suggested by Professor STOKES ( British Association Report , 1862 ) , to which the reciprocal of the velocity of wave propagation was determined by experiment and also on FRESNEL'S theory .
The greatest difference between the two amounted to 0009 , and there appeared to be some connexion between the differences and ) wave length of the used .
In the endeavour to follow up this connexion I undertook a series of similar experiments with light of different wave lengths , using three lines of the hydrogen spectrum and the sodium line .
The extreme smallness of the arragonite prisms I had previously worked with led me to use , at first at least , Iceland spar , which could be obtained in large pieces with ease , and for which the theoretical calculations were greatly more simple .
Professor STOKES had already made a series of experiments by the same method with this substance ( Proceedings of the Royal Society , vol. 20 , p. 443 ) and arrived at results construction .
The details of his experiments are as yet unpublished , I venture to think it might be useful to have arranged in tabular a series of results , to erve in the future as a test of any theory of double refraction might be proposed .
The method of the experiments , as suggested by Professor STOKES ( British Association Report , 1862 ) , is as follows : A prism is cut from a piece of spar , and the position of its faces with reference to the cleavage faces carefully determined .
The is mounted on a spectrometer , and the collimator adjusted so that the rays of a definite wave length falling on the prism are parallel , the edge of the prism parallel to the axis of revolution of the reading telescope .
The deviation of the light ssing through the prism in any position is observed , also the position of the image of the MR. R. T. BROOK ON DOUBLE REFRACTION slit formed by reflexion at the face of incidence .
From this and the known direction of the incident light we can calculate the angle of incidence .
Let this be .
Let the deviation be and the angle of the prism .
Let V be velocity of the light in air , in the crystal .
Let be the angle of emergence , the angles which the wave normal in the crystal makes with the faces of the prism .
Then we have ( 1 ) ( 2 ) .
( 3 ) whence we can find ( , and since is known , we can get at once and and then is given by either of the formulae ( 4 ) But since we know the position of the faces of the prism with reference to the optic axis , we can find the angle between the wave normal and the optic axis , and if be the reciprocals of the principal velocities , that of a velocity in a direction making angle with the optic axis , we have by HUYGHEN'S construction , ( 5 ) and this ) , we can find AN1 ) DlSPhl RSION ) SPAR .
II .
I. Description of crysta 1 .
II .
Account of with esults .
It was my object in out the work to secure a series of observations for values of from to , differing by about 1o 30 ' or rather less .
This I found could be obtained by the use of four prisms of or thel.eabouts , each having its edge perpendicular to the optic axis , which would therefore lie in the principal plane of each prism , the prisms being so cut that the optic axis made angles of , and , with the outward drawn normal to one of the faces ; the angles are considered positive when the optic axis falls on the same side of the normal as the edge of the prism .
Prisms cut in this malmer would , I found , enable me to work over a range extending from about on one side of the optic lxis to about lo on the other .
Iceland spar , is well known , cleaves readily so ) oblique hombohedron .
Fig. 1 .
Let A BCD , fig. 1 , represent a rhomb of spar , and let A be a solid angle , such that each of the three plane angles A A A is obtuse .
The optic axis is equally inclined to each of the faces A A A , the angle of inclination being about .
It is , therefore , perpendicular to the interior bisectors of the acute angles A. I procured large holnb of spal which was cut by A. , 196 , Tottenham Court Road , into .
prisms , the edge of each being nearly parallel to the interior bisectors of the acute amgle of the same hombic face .
The angle of each prism was about the were cut so as to be inclined to the optic axis as stated above .
We proceed now to describe the experi1nents and give the results for each of these prisms numbered I. , Il .
, III .
, alJdIV .
each case let the faces of the prism , the angle between them , the angles which the wave normal in the prisnl makes with the normals to respectively , the corresponding angles in air ; is the angle of incidence or emergence according as the light is iucident or versed for The values of the angle of incidence on one face extend cideuce to the position of minimum deviation , forming an arithmetic progression of which the common difference is The prism was then reversed so that the face of incidence became that of emergence , and another set of results obtained , extending from minimum deviation to nearly grazing incidence on that face .
Each set of experiments was taken twice , and only in two or three cases were the differences between the results of the two measurements , usually made on different days , greater than In about 18 per cent. of the measurements the differences amounted to 20 in the rest it was less , so that in comparatively few cases is the difference between the mean and an extreme observation as great as The spectrometer was the same as that used in the experiments with arragonite , and was kindly lent me by Professor STOKES .
The method of taking the measurements and the means adopted to secure the parallelism of the edge of the prism and the axis of rotation of the telescope are described at length ( Phil. Trans. , 1879 ) .
The collimator and telescope were focused for parallel rays by means of a method suggested by Dr. SCHUSTER ( Phil. Mag. , February , 1879 ) .
The focusing was done once for each prism , and remained untouched during the expel.iments with that prism .
All the adjustments were made for the hydrogen line C. When the from this line were parallel no appreciable alteration was required to render the sodium rays parallel .
The other hydrogen rays and were very nearly parallel , but probably not quite so .
The experiments were performed in the spectroscope room at the Cavendish Laboratory , which was kindly placed at my disposal by Professor MAXWELL during February , March , and April of the present year .
The value given for the angle of the prism is in each case the mean of 10 measures , no two of which differed by more than 20 In the course of the preliminary work I found that variations in temperature of or C. , to which the room was subject during the months of February and March , produced a very appreciable effect in the value of the angles between some of the faces .
In making the final measurements , therefore , I was careful to keep the room at a nearly constant temperature of about C. by means of a gas stove .
For each position of the prism an observation of the deviation of each of the four was taken so that there are four values of deviation , corresponding respectively to these , to each angle of incidence .
bless I. , II .
, III .
, .
give the estllts of experiment for the red line of the hydrogen spectrum the four isms .
The error in the result , due to an error in one of the observed quantities , is greatest near the position of minimum deviation .
If we assume an error of in the values of the angle of incidence and the deviation taken so as to produce the maximum error in the result , that error amounts to about when a maximum .
The probable error of the experiments is considerably less than this .
AND DISPERSION IN ICELAND SPAR .
TABLE I.\mdash ; Prism I. , C. TABLE II.\mdash ; Prism II .
, Ray C. TABLE III.\mdash ; Prism III .
, C. TABLE IV .
Prisnl IV .
, Ray C. AND DTSPERSTON TN SPAR Column 1 gives the value of being the observed deviation , and the angle of the prism .
( In the calculations occurs only in the form , therefore is given in the tables instead of D. ) Column 2 the observed angle of incidence .
Column 3 the angle which the wave normal in the crystal makes with the normal to the faces of incidence calculated from the formulae already proved , and column 4 the values of or calculated from On comparing the results for the ray with theory I found so close an agreement that I thought it hardly requisite to work out all the for the rays and I therefore completed the calculations for only about a third of the observations , giving a series of values of in directions inclined at angles of about to each other , extending in an almost continuous arc from the optic axis to directions perpendicular to it .
These are contained in Tables .
and The middle column in each case gives the angle of incidence .
The columns on the right refer to the ray , those on the left to the ray F. For Table , Prism II .
, the results for the angle of incidence have been calculated the value of the angle of the prism instead of the value used for the results in which the angle of incidence is denoted by .
The reasons for this will be discussed in connexion with the theory .
This closes the experimental part of the work .
AND DISPERSION IN ICELAND SPAR .
TABLE .\mdash ; Results of the Experiments .
SECTION III .
I. of the position of the plane of the II .
Proposition proved.\mdash ; The principal plane of I. , IlI .
, be treated if it through the optic III .
Theoretical calculationsfor the reciprocal of the Our next step will be the determination of the position of the prismh with reference to the optic axis .
This was accurately determined for each prism by measuring the angles between them and two of the rhombic faces of the crystal .
The angle between these faces and also the angle between the cut faces of each of the prisms were accurately observed .
Let us take point within the crystal as origin , and from it draw to the optic axis nleet the sphere in Fig. 2 .
Let be the points in which the normals to the two faces , , of one of the prisms meet the sphere .
Let us take the plane as plane of , the internal and external bisectors of the angle as axes of and respectively , the axis of being perpendicular to the plane , .
Then and lie in the plane , Let are known from experiment .
Let be the direction angles of P. Then from triangles ( 1 ) From triangles AND RSION IN TCELAND These formulae give us the values of and or the angle between the normals to the rhombic was observed in three pieces of the crystal used .
The values were Mean of four measures .
Maximum difference , 10\ldquo ; Mean of five measures .
Mean of ten measures .
Maximum difference , We may therefore put with great accuracy The temperature indicated by a thermometer placed almost in contact with the crystal and shaded from the direct radiation of the light used to read the vernier , was C. to C. Each of tlJe angles was obseryed ten times for each face and the mean taken , the temperature being kept as nearly as possible at The greatest variation between any two observations never exceeded TABLE VII.\mdash ; The position of the normals to the faces of the crystal prism .
The next two columns give the values of , the last two those of The values of show that the principal plane of the prism which contains the normals to the faces and is nearly coincident with the plane 2 We proceed to find the position of the line of junction of these planes and the angle between them .
Fig. 3 .
Let ( fig. 3 ) meet in M. Draw arcs to Then from triangle P PM PML From triangle Q PML Whence and we have ( 3 ) If and are on opposite sides of , we get ( 4 ) In these formulae , being known from Table VII .
and the angle of the , we find M. AND DISPERSION IN ICELAND SPAR .
Then by substitution in the formula ( 5 ) we obtain L. If we call S , and the angle , we have for the four prisms respectively TABLE VIII .
We shall now prove that in the case of prisms I. , III .
, and .
we may neglect the inclination of the plane of the prism to the plane .
For } ) eing the optic axis , the point in which any wave normal meets the sphere , the intersection of and Let we have say neglecting This we may do , for is In neglecting the mutual inclination of these planes , i. e. , in putting , we omit a term in of the value and in of the value This is not greater than is not greater than Term neglected is not greater than In the three cases considered is less than 14\fnof ; m circnlar measure of 14 ' is 004 .
Using RUDBERG'S of we have is less than is less than is less than 5 Therefore , greatest difference is less than or 000004 Hence neglecting or supposing the plane of each of the prisms I. , III .
, and IV .
to coincide with the plane will never produce any change in the fifth decimal figure in the value of In the case of pl.ism II .
the value of is nearly , and we may have to take account of the obliquity .
For prisms I. , III .
, and IV .
the value of is given by formulae of the form ( having ) meaning attached to it in the results of experiment .
AND DISPERSION IN ICELAND SPAR .
To determine we require to know the position of the optic axis with reference to The optic axis is equally inclined to ( fig. 2 ) .
Hence ch of the angles subtended at by the arcs is Therefore is whence . . . . . . . . .
Again , from the right-angled triangle and see Substituting the value we have . . . . . . . .
( 7 ) refers tu the face P. The position of the wave normal is also given by where being the direction angle of the face Q. From these we get the following table of values to determine the angle between the optic axis and any wave normal .
TABLE IX .
Tables X. , XI .
, XII .
, and XIII .
give the values of calculated from these formulae from the values of given in Tables I. , II .
, III .
, IV .
, V. , and VI .
For prism II .
we must calculate the value of from the formula SM SM NM\mdash ; tall ( S ) if ( 9 ) Now we have from the triangle M ( fig. 3 ) QK PML We know that PML [ Tables VII .
and VIII .
] . . . . .
( 10 ) AK KQ . . . .
( 11 ) But [ From Section III .
( 7 ) .
] . . . .
( 12 ) AND DISPERSION IN ICELAND SPAR .
Also from Table VIII .
Nbl Hence [ Section III .
] . . . . .
( 1.3 ) From these values we obtain the values of corresponding to the angles of incidence in Tables I. , TI .
, III .
, and IV .
TABLE X. Theol.etical esults f the line C. CCCLX X. 3 ABLE X.\mdash ; continued .
having been ined is given by @ ( 14 ) AND DISPERSION IN SPAR .
If ( 15 ) . . . . .
( 1 and we require to find is the maximum radius vector of the spheroidal sheet of the surface of wave slowness .
This is given by From Table X. we have , considering at present the line , when is also the refractive index of the ordinary wave .
Its value was determined by observations on the angle of incidence and deviation of the ordinary ray in prisms I. , IIT .
, and The values were Prism I. Prism III .
Prism IV .
We take then as the value of 1 .
Observations of the minimum deviation were made to determine from the usual formula being the minimum deviation .
The mean of these was 1.65441 but the error of this last method is much greater than that in the former , and as ally error in the observed value of would probably increase , through the prism not being exactly the position of minimum deviation , we should expect to get a value for rather in excess of the true .
The values given by MASCART and RUDBERG are respectively 1 .
and 1.65446 To determine we must consider the minimum radius vector of the spheroidal sheet ; this is given by MR. R. T. BROOK ON DOUBLE RHiFRACTION Now when we see from Table X. But Substituting the values of , and we get The values given by MASCART and RUDBERG are 1.48455 1.48474 The middle column of Table X. gives the values of in the directions given by the first column for the values The Roman numerals I. , II .
, \amp ; c. , in the first column refer to the tables of experimental results from which the values of in the fourth column are taken .
The fifth column gives the excess of experiment over theory .
These differences it will be seen are much greater in the case of prism II .
than any of the others .
They are also greater for the first part of the results in Table II .
, in which the face of incidence was , than for the latter , when the light was incident on the face Q. Postponing , then , for the present the consideration of this point , let us compare the differences between theory and experiment for prisms IIII .
, and IV .
We notice at once their extreme smallness\mdash ; the greatest is only , and only in eight out of the sixty measurements taken do they amount to as much as ooo1 .
Ths mean irrespective of sign is 000055 .
The differences are , the whole , negative near the major axis They tend to become least at about away from either axis .
From that point they are positive and reach a maximum value at from to away from the major axis .
So that the curve given by experiment would , though very nearly coincident with an ellipse , lie inside the ellipse near the major axis , cut it at about from that axis , and lie outside for the of its course .
The difference , however , between the radii vectors to the two curves drawn in the same direction would never be greater than part of either .
My first inference from these results was that HUYGHEN'S construction represented ur to a degree of exactness comparabJe with the probable error of experiment , AND RSION IN ICELAND SPAR .
Before considering the results for the rays and we must return to the experiments with prism II .
The large differences it gives , overlapping as they do values given by experiments on prism I. , in which the differences are small , pointed clearly to errors of experiment .
On referring to my note-book containing the direct results of experiment , I found that the observation of deviation and incidence for the face had been made on March 29 , while the observations for the face and the angle between the faces were made on April 1 .
It seemed possible that the temperature of the prism had been different on the two occasions , and that this was the cause of the error .
I therefore proceeded to observe afresh the angle between the faces of the prism .
The result differed by from that found on April 1 .
I therefore recalculated the experimental results for the prism II .
so far as the face was concerned .
Table X. ( A ) gives the results of the calculations .
In Table X. ( B ) these are compared with the theory .
TABLE X. \mdash ; Prism II .
Ray C. TABLE X. ( B).\mdash ; Theory for same .
Thus this iation has tended to decrease the differences between observation and theory , and hatit educed them to almost the same magnitude as those given by the face of the .
They now agree more nearly with the results of prisms I. , III .
, and IV .
, though even yet the differences observed are greater than in any of the other prisms .
Prism II .
, however , was at first cut wrongly from the crystal , and when recut it was so small that I formed the intention of not using it at all , and leaving a gap in my series of observations between the values and .
I found , however , on a second and more careful trial , that the images formed by it were clearer and brighter than I had thought , and so determined to take a series of observations with it .
When I observed a second time the angle of prism II .
, I took a series of measure .
ments of deviation , \amp ; c. , which lead to results in agreement with Tables X. ( A ) , X. ( B ) , so that on the whole the results given by this prism are in accordance with those already arrived at in prisms I. , III .
, and IV .
Our next step is to consider the theory for the rays and The position of the plane containing the two normals to the faces of the prism is of course the same .
and therefore so are the formulae which give the relations between and and The values of the axes of spheroid on HUYGHEN'S theory are , however , different .
Let us take the green line , , first .
is , as before , the value of the ordinary refractive index .
We have as for the line the four values Prism I. Prism III .
Prism IV .
We take is the vaIue of when in able X Now for experiment gives we take this as the value of Hence for we have MASCART and RUDBERG give respectively and RSION ICELAND For we have , as before , .
Prism III .
Prism IV .
Whence and for wheu Table XII .
we have Whence Thus for Tables XI .
and XII .
give the esults of the calculations .
TABLE XI.\mdash ; Results of Theol.y for F. .\mdash ; Results of Theory for IIIIIIIII .
AND DISPERSION IN TCELAND SPAR .
TABLE XIII.\mdash ; Results of Theory for C. JIII 1jfII 1From TFrom Experiment .
Table XTII .
gives the results for the ray for the same values of the angle of incidence as those given in Tables XI .
and XII .
for and .
This enables a comparison of the results to be more easily made for the three rays than if it were requisite to refer to X. In each case the results are similar .
The differences are least near the axes , being negative for near the minor axis , and for near both major and minor .
For the errors are positive throughout , so that a small increase of the axes of the curve given by theory would , on the whole , bring theory and experiment into closer agreement .
For the differences near the minor axis being negative , we should require to decrease the minor axis of the ellipse .
This would increase slightly the positive errors , render , on the whole , the variation from FRESNEL'S spheroid more marked , and greater than the variation of the red ray .
variation from FRESNEL'S theory near the middle of the arc more marked than in the case of the red or green rays .
In fact , while for the red , supposing the variations in contemplated above to have been adopted , the greatest difference between theory and experiment would be about for the green ray it would rise to and for the violet , , to SECTION IV .
I. Comparison with previous ts .
II .
of variation of constants .
As an additional proof of the accuracy of the experiments it may be worth while giving the results of a series of measurements covering the same ground as the second part , Table I. , made some months previously .
Since the prism did not occupy exactly the same position relative to the instrument as it did during the experiments in Table I. , the values of the angle of incidence , and therefore of , were slightly different to those in Table I. In making the comparison , therefore , the results of calculation had to be altered by interpolation to give the values of corresponding to the values of in Table I. The result is contained in Table XIV .
TABLE XIV .
AND DISPERSION IN ICELAND SPAR .
The agreement between the two results is striking , and seems to show that we may assume the experimental results to be true with an error which is not greater than .
In my paper on a biaxal crystal I was able to show that the assumption of certain errors in the determination of the position of the plane of the prism with reference to the crystallographic axes led to results rather more in agreement with experiment than those obtained at ffist .
In the present instance this is impossible , for any change in the position of the plane of the prism would produce effects of almost exactly the same amount in the values of for the lines and ; but the error we wish to correct in is only half as great as that in , and hence no change in the position of the plane can produce the required effect .
But , again , the telescopes used were not perfectly achromatic for the red and violet rays .
I found usually little or no difference in the position of the focus for the lines and , but there was an appreciable difference between and If the collimator be focused so that the rays from the line emerge parallel , those from will be divergent .
This may produce a variation in the angle of incidence between the waves and For the prisms were so placed that by turning the table on which they rested without altering the position of the collimator , either face of the prism could be made a face of incidence .
To secure this the edge of the prism passed nearly through the axis of the collimator , and in most positions of the prism the light from only about half the collimator lens reached it .
A figure will make this clearer .
( fig. 5 ) is the axis of the collimator , A the prism , A being the face of incidence .
Fig. 5 .
Ahnost all the light incident on A passes through the upper part , , of the of the collimator .
If the prism be turned round , a point in the axis S so that becomes the face of incidence , then only the lower portion , , of the collimator lens will be used .
Again , since the collimator is focused for red rays , they will be incident on the face A in direction , and if be the normal to A the angle of incidence will be N. the angle of incidence for the violet too great .
Again , if be any emergent ray , we have assumed the deviation to be measured by R. In the case of the violet rays this be too great , and too great by the same amount as the angle of incidence .
We must therefore consider the effect of decreasing the angle of incidence and the deviation by the same amount .
If be the angle of incidence , the angle of emergence , @SD\mdash ; S0 S ' is negative since S is so .
S ' is positive .
Hence is unchanged , is increased .
The value of will therefore in all cases be decreased .
Now the experimental values of are already too great .
Hence this alteration will tend to bring them more nearly into agreement with theory .
The amount of error introduced depends on the angle of incidence .
To find a general expression for it would be a work of difficulty owing to the complicated nature of the formulae involved .
Let us therefore consider the effect of decreasing the angle of incidence and the deviation by 1 ' , near minimum deviation , ( b ) near grazing incidence for prism I. The effects will be much the same for all the prisms .
We have from Table I. By decreasing and each by 1 ' we have the following values:\mdash ; : AND DISPERSION IN ICELAND SPAR .
Thus near minimum deviation the change produced in ulOUllts to about 0003 while at grazing incidence it is only about Of course an errol of the same kind occurs in the values of .
They , however , were determined from observations at nearly grazing incidence .
They may then be slightly too great .
To correct them completely for this we should have to reduce the theoretical values by a small quantity nearly the same for all ; while the experimental values require reducing by quantities which are greatest neal deviation , and decrease as we approach grazing incidence until they reach abont the values of the corrections applied to and The greatest error for the ray does not exceed , so that bhe results of theory and experiment for would be brought into very close agreement by supposing the violet rays of the light emerging from the collimator to be inclined to the red at angles not greater than 45 Thus , allowing for this probable divergency of the green and violet rays , it appears that HUYGHEN'S construction represents the result of experiment for the three rays of the hydrogen spectrum to a degree of approximation comparable with the probable error of the experiments .
*In the abstract pl'inted in the ' Proceedings of the Royal Society , ' I assumed that the violet rays issuing from an achromatic lens , focused so as to makc the orange rays rallel , were collvclgent .
this it followed that the correction for want of parallelism tended to increase the difference between and experiment , and led me to the inference that HUYGHEN'S construction might be true for the red rays and yet differ appreciably from the truth for light of shorter wave length .
STOKES has since pointed out to me that the violet rays are in reality divergent , and that the error introduced by assuming them to be parallel really tends correct the differences observed betiween theory and experiment , and so leads to the inference in the text that HUYGIIEN'S construction is true for the three hydrogen within the limits of

