]\gt ; [ 299 ] VII .
An Experimental Investigation into the Form of the Wave Surface of Quartz .
JAMES C. Mc CONNEL , Communicated by T. GLAZEBROOK , M.A. , F.R. Received November 9 , \mdash ; Read December 17 , 1885 .
I.\mdash ; INTRODUCTORY REMARKS .
ABOUT two years ago I read a paper before the Cambridge Philosophical ' Society describing some measurements of tho " " dark rings\ldquo ; in quartz .
The present paper contains an account of similar measurements made with greatly improved apparatus , and extending over a much larger field .
These " " dark rings\ldquo ; supply a delicate method of determining the retardation of the extraordinary wave behind the ordinary in the crystal and consequently the separation between the two sheets at various points of the wave-surface .
In quartz the wave-surface may be nearly represented by a prolate spheroid sur- rounded by a sphere Simg through the extremities of its axis .
The spheroid is slightly flattened at the extremity of its axis and the sphere slightly bulged , so that the two no longer touch .
The distance between the two at the extremity of the axis we know very accurately from observations on the rotatory power , while the two radii at the equator of the wave-surface are known from observations with the spectrometer .
In all the theories on the subject these three constants are assumed ; so the theoretical surfaces are made to coincide with the true one in the equatorial section and to give the true distance between the two sheets at the extremity of the axis .
At intermediate points their correspondence with the true surface lies open to the test of experiment .
My former observations were confined to the regions near the axis , as it is only there that the bulgin and flattening are at all conspicuous .
But the increased facilities of a new arrangement for measuring the rings led me to extend my observations over the whole surface , so that I have now determined the gradually increasing distance between the two sheets from the axis to the equator .
This was the more necessary because the known values of the radii of the equatorial section , or , in other words , of the principal wave-velocities , are not sufficiently accurate to give a good measure of their difference .
And it is upon this difference that the theoretical distance between the two sheets , even in the region near the axis , to a great extent depends .
sphere .
My determination of the distance between the sheets should not at any rate be in error by more than one part in 500 at from the axis , one part in 1000 at , and one part in 2000 at a considerable distance from the axis .
Of that , though by no means a complete investigation of the form of the wave-surface , these observations must be admitted to be a severe test .
For if one wave-velocity were to be changed while the other remained the same , the change would be indicated even though it should only amount to part of the whole velocity .
Indeed near the axis part would be sensible .
The theory of the formation of the rings requires some consideration .
Take a wave-front of the plane polarised incident light .
By the refraction at the first surface it is resolved into two wave-fronts which are polarised in a manner dependent on the constitution of the crystal , and which travel with different velocities .
Each of these on refraction at the second surface gives rise to a wave-front polarised in a manner somewhat different from its own .
Let the emergent wave-front due to the extraordinary wave be and the other A. Then and A are parallel .
If be in advance of A by an integral number of wave-lengths , A will be coincident with a wave-front exactly similar to , and in the same phase .
But we cannot say that , therefore , the combination of these two will be exactly similar to or even polarised in the same manner as the incident wave-front , for there has been loss of light by reflection at both surfaces , and this has affected A and differently .
The of the refractive effects is greatly simplified in our case by the circumstance that the optic axis lies in the plane of incidence .
Thus the principal axes of the ellipses of vibration of the waves in the quartz lie respectively in , and perpendicular to , the plane of incidence .
We will assume at present that the action at the surface produces no retardation of phase .
Let us first suppose that the incident light is plane polarised and the vibration lies in the plane of incidence .
This linear vibration on entering the crystal gives rise to two elliptic vibrations ; and we may , I thin , safely assume that , if the relative retardation amount to an integral number of wave-lengths these two elliptic vibrations will give rise on emergence to a single linear vibration also in the plane of incidence .
For we may consider the components of these two elliptic vibrations perpendicular to the plane of incidence as due to two equal and opposite linear vibrations of the incident light .
These two linear vibrations would suffer equal reductions at the first refraction , and equal reductions at the second refraction , so after both refractions they would again neutralise each other .
It is indeed probable that the difference in the angle of refraction of the two waves would affect the amount of light lost by reflection , but it would be to a negligibly smaJl extent .
Thus the emergent light will be sensibly plane polarised .
INTO THE FORM OF THE WAVE-SURFACE OF QUARTZ .
Similarly , if the vibration of the incident light be perpendicular to the plane of incidence , the emergent light will be also plane arised .
Consequently , if the incident light be plane polarised in any plane , the emergent light will also be plane polarised , though not necessarily in a parallel plane .
Thus , for angles of incidence giving a relative retardation of an integral number of wave-lengths , the refractive effects are equivalent to a small rotation of the plane of polarisation of the incident light .
And the same is no doubt very approximately true for neighbouring angles of incidence .
This small rotation can , as will appear later on , be eliminated by taking the mean of two sets of observations .
Up to this point , and indeed throughout , we assume that there is no perceptible retardation of phase at the surface .
Careful observations have been made on glass with regard to this point , and the only trace of retardation found was at very high angles of incidence , and even that was highly doubtful .
There seems to be no reason to expect such an effect in crystalline media more than in isotropic , at any rate when the rotatory effects are small .
In the present case too the rings are very close together at the high angles of incidence , so that only a small error would result ; e.g. , at an incidence of , which is the highest incidence I have reached , the elror in the diameter of a ring produced by a retardation of would be only 12 As to any surface retardation connected with the rotatory power , we know from measurements of the rotation in plates of different thickness that it is , at any rate , very small when the light is incident normally , and passes along the axis .
These urements have been made with great accuracy , and no such effect has been even suspected .
So we have a maximum limit for it , from which I deduce that it could only make itself felt slightly in the first rin , in the others , at all .
II.\mdash ; THE APPARATUS AND MEASUREMENTS .
My previous observations were made with one of GROTH'S polariscopes somewhat modified , .
as described in the paper , but since then I have found it a great improvement to use instead an ordinary spectrometer .
With some slight additions the spectrometer became a most efficient instrument for measuring the rings .
The general arrangement is indicated in fig. 1 .
The source of light is a flame in the edge of which is placed a bead of sodium carbonate .
The light is emerging from the quartz is brought to a focus in the focal plane of the objectglass D. So the needle-point has to be carefully adjusted to lie in this placie .
In this particular the spectrometer has a great advantage over the polariscope .
The focal length is much greater and so the adjustment can be made with greater accuracy , and the spherical aberration , which in the polariscope is very troublesome , is so small as to produce no sensible error .
is the eye-piece and the second NICOL by which the light is analysed .
I found that the ordinary eye-piece magnified too strongly , so I substituted a single lens and fitted it up with the NICOL in a large cork , so that the combination could be readily exchanged for the ordinary eye-piece without disturbin the needle-point .
Even with the low power the apparent size of the rings was vastly greater than with the polariscope .
This was mainly due to the greater focal length , which was about nine inches .
Though the definition of the rings suffered with the increased magnification , there was on the whole a decided gain in ccuracy , and the observations became far less trying to the eyes .
The spectrometer table too was furnished with a slow motion screw which was no small advantage .
The spectrometer was fitted with a circle graduated on silver , and both the telescope and table were provided with verniers reading to half minutes .
By replacing the slit and the ordinary eye-piece of the telescope the spectrometer could return to its ordinary use .
The normal to the faces of the quartz could thus be set at right angles to the axis of rotation of telescope and table and its position determined relatively to the axis of collimation of the telescope for any given readings of the telescope and table verniers .
For levelling purposes the qualtz was mounted on a tripod stand resting on three adjustable screws whose rounded ends rested in a conical hole , in a -shaped slot , and on a smooth surface , respectively on the table of the spectrometer .
Each screw carried a loose nut which , when the screw had been properly adjusted , could be screwed down tight onto the tripod .
This made the three screws a perfectly firm support and the tripod could be removed from the table replaced accurately in its original position\mdash ; a great convenience .
The distance A from the polarising NICOL to the object-glass of the collimator was over four feet , so the light which fell on the lens had passed through the NICOL nearly as a parallel pencil , and therefore was polarised approximately all in one plane .
The two NICOLS could therefore be crossed with greater accuracy .
In order to cross the NICOLS the quartz was removed and the cork turned till the light was quenched .
In my earlier observations I simply used the sodium light for this purpose , but I found that this led to errors of two or three minutes in the smaller rings , so the INTO THE FORM OF THE WAVE-SURFACE OF QUARTZ .
crossing for all the observations given below was performed with a gas-flame placed edgeways behind the polariser .
This gave ample accuracy .
[ Added -May 31 , 1886.\mdash ; Probably the lenses had some slight doubly lefracting power , so that each turned the plane of polarisation through a small angle depending on its azimuth .
In the present case however this effect was a matter of indifference , as I was measuring the angles of incidence , for which the emergent light from the quartz was polarised similarly to the incident light .
In the larger rings , it is true , the quartz produced a considerable lateral displacement of the light , so it no longer passed centrically through the two object glasses .
But in the larger rings the effect of any small error of crossing might , as will be seen shortly , be completely neglected .
] The plane of polarisation was inclined at to the plane of incidence .
If we look at the appearance presented in the polariscope when the two NICOLS are crossed , we see that all the rings are circular , and though the inner rings are tolerably uniform , the outer rings are crossed by dark brushes in directions parallel and perpendicular to the plane of polarisation .
These dark ushes have the effect of making the rings very ill-defined , so that on that account alone it would be impossible to get a satisfactory measure of the 15th ring if the plane of polarisation were parallel to the plane of incidence .
If now we rotate the analyser slightly we see the rings dilate , but not uniformly .
The dilatation is much more rapid in the direction of the brushes .
The rings cease to be circular and become four-cornered .
This phenomenon was observed and explained by Sir GEORGE AIRY ( Camb .
Phil. Trans. , vol. 4 , p. 85 ) .
It is an easy deduction from his formula for the intensity at any point of the field that the value of for the darkest part of the ring when the NICOLS are not accurately crossed is given by in the direction of the brushes , but by in directions at to the brushes .
Where is the relative retardation , S is the small angular error of crossing , and is the ratio of the axes of the ellipse of vibration of either wave , being less than unity .
The following figures will give some idea of the values of .
For sodium light when\mdash ; being ffie angle between the wave normal and the optic axis .
the ring in the directions , due to a small error in crossing , is almost nil , the displacement in the parallel and perpendicular directions is extremely large .
The latter , however , is masked by the general obscurity of the brushes .
From this it is clear that the best position for the plane of polarisation is at an inclination of to the plane of incidence .
It occurred to me , as an objection that might be of importance , that there would be some rotation of the plane of polarisation through refraction at the surface of the quartz and the surfaces of the lenses , and that putting the plane of polarisation in the position made this rotation as large as possible .
But it is easy to see that if the polariser and analyser be turned into the other position the rotation will be in the opposite direction .
Thus in one case the rings will be slightly too large , in the other just as much too small , and by taking the mean of the two the error will be eliminated .
I have , therefore , taken observations throughout in these two positions of the polariser , though no effect of the kind has made itself manifest .
The first set of observations recorded below were made on a piece of quartz , which I will call Plate 1 .
It has four polished faces , and the section perpendicular to these faces has the figure of an equilateral triangle described on a square .
The two sides of the square adjacent to the triangle were polished faces forming a parallel-sided plate .
The two other faces were convenient as supplying a.definite plane of reference passing through the normal to the two plate faces , though they were originally cut for quite a different purpose .
Each face was about an inch square .
The whole piece was perfectly clear , and the only flaw was a slight intrusion of contrary quartz into one corner .
The plate faces were very nearly parallel planes , the error of planitude not exceeding 1 ' , and the error of parallelism not exceedin .
Measurements of the rings were made in two different planes .
The plane which is perpendicular to all four faces I will call plane A. The plane at right angles to plane A I call plane B. The four polished faces were very nearly in the same zone .
The observations in plane A were made in the following order .
The quartz was fixed on the tripod and all four faces set parallel to the axis of rotation of the spectrometer in the usual way by means of reflections from the faces .
Readings were also taken from which I could find the position of the normal to one of the plate faces relative to the axis of the telescope for any given readings of the telescope and table verniers .
tripod and quartz were then lifted off the table , the slit removed from the collimator , the ordinary eye-piece of the telescope replaced by the cork containing the low power eye-piece and analyser , the gas- flame placed behind the polariser , and the NICOLS crossed .
The tripod having been returned to its old position and the gas-flame exchanged for the sodium-flame every* Added Nay 31 , 1886 .
INTO THE FORM OF THE WAVE-SURFACE QUARTZ .
thin was ready for observation .
The temperature was taken , sually at the beginmn and end of a set of observations , from a thermometer lying on the table by the side of the speotrometer .
The appearance seen in the telescope was the needle-point standing out against an irregular patch of light crossed by one or more dark bands .
The bright patch was really the aperture of the analysing NICOL , and as the telescope was moved round one or two degrees the patch moved across the .
This was due to the light which reached the eye coming from a different part of the object .
glass and thus passing in a different direction through the Thus it was really equivalent to a slight rotation of the analyser round its axis .
A similar effect was produced by turning the table owing to the lateral displacement of the beam of light by the quartz .
To avoid the latter evil I at first always turned ) telescope till the needle-point was exactly in the centre of the patch before I took a reading .
This was done in most of the observations on Plate 1 .
But for the small rings in Plate 2 I simply clamped the telescope in the position where the NICOLS were crossed , and kept it there .
By trying exaggerated cases I found the error was very small and it tended to increase the radius on one side while diminishing it on the other .
The rings of course were fixed relatively to the quartz , so that the difference of the readings of the table vernier for the two sides of the ring being on the gay the diameter of the ring in air .
For the first half dozen rings it may be shown , that the mean of these two readings should coincide , or in other words that the rings are practically circular and have a common centre , provided that the deviation of the axis from the normal do not exceed two or three degrees .
This common centre may be called the axis in air , and is related to the real optic axis by the ordinary law of refraction .
From this it is easy to see how the deviation of the axis from the normal was determined from the measurements of the rings .
I found that the error was 1o 45 ' measured parallel to plane A and 18 ' measured perpendicular to plane A. It is clear since the axiS lay 18 ' out of plane A that instead of measuring the diameter of a rin I was only measuring a chord .
The small correction for this was however to apply .
But for the measurements at right angles to plane A this chord would have been perhaps only half the diameter for the first ring if I had proceeded as before .
It was necessary to tilt the plate forwards till the axis in air came nearly into the plane of measurement .
I was at a loss for some time how to produce this large tilt on the spectrometer with any accuracy .
At last I hit * When the direction of ight incident on a NICOL'S prism is inclined to the principal plane of the prism at an angle , then , in order that the light may be completely quenched by the prism , its plane of polarisation must be inclined at an angle nearly , to the plane through the ray most neurly parallel to the principal plane of the prism .
See a paper by the author , Phil. Mag. , vol. 19 , p. 320.\mdash ; [ June 15 , 1886 .
] MR. J. C. Mc CONNEL ON AN EXPERIMENTAL INVESTIGATION upon the following expedient .
I fixed a small piece of silvered glass on to one of the plate so as to be approximately in the same zone as the four faces and tely perpendicular to the axis in air .
I first fixed it in its place with soft wax , adjusting it with my eye at the telescope , and then secured it firmly by ) in electrical cement behind it .
Aftsr this was attached , it was necessary to bring its normal into the plane of measurement to secure that the diametral chord was being measured .
Plane then may be defined as passing through this normal and being approximately perpendicular to plane A. The first ring was considerably broader and less sharply defined than the others ; yet in obtaining the diameters recorded in the tables below I have relied on a reading of each side even of the first ring .
For during a long course of preliminary measurements I had found , that two readings of the first ring would seldom differ more than , though I could not be confident that the true reading lay within , of either .
It seemed in fact to be easier to bring the needle back to the same point in the band than to make sure of that point being the middle of the band .
To obtain real accuracy I relied on varying the conditions as much as possible .
Thus the four sets of measurements recorded in Table I. were made on four differenb days with different positions of the polariser , analyser , and plate of qnartz .
I also made a number of measurements of the larger rings in Plate 1 , but these , I regret to say , have proved to be useless .
The formulae , which are compact enough when the plate is cut truly perpendicular to axis , become hopelessly complicated when the deviation of the axis has to be introduced .
Since the deviation of the axis was small I tried approximations , carrying them as far as the squares of the deviations in the two directions .
But the discrepancy between the results for plane A and for plane seemed to indicate that the cubes also should have been included .
determined to obtain a new plate in which the deviation of the axis should be negligibly small .
I selected a good piece at HILGER'S and got him to cut it roughly at right angles to the axis .
I then set it up on the spectrometer , determined the and direction of the error , and sent it back to HILGER to be recut .
By repeating this process , getting it recut no less than four times , I at length attained my object .
This plate I will call Plate 2 .
It was perfectly clear and free from maccling .
The two polished faoes were good planes about 80 mm. square , and inclined at an angle of about .
The two planes through the normal and parallel to the sides of the square I will call plane A and plane B. The error of the axis was about 6 ' parallel to plane A and about the same parallel to plane B. This I found was small enough to be negligible throughout .
The thickness was about 20 mm. only .
The first ring had nearly the same diameter as in Plate 1 , but it was produced by a retardation of three wave-lengths instead of four .
Owing to the smaller thickness -he ) ands were broader and further apart than before ; but I do not think this INTO THE FORM OF THE WAVE-SURFACE OF QUARTZ .
impaired the accuracy of measurement except perhaps slightly in the first two I had the plate cut down to this thickness in order to secure good observations at the high ] of incidence .
The object glasses of the telescope and collimator were barely 25 mm. in diameter , yet I was able to obtain good readings for the 150th ring at an angle of incidence of over , which would have been quite impossible with a plate 25 mm. thick .
The observations on the small rings in Plate 2 were similar to those te 1 , except that three readings were taken throughout for each of the first ring and two readings for each side of the second ring .
The greatest difference between the readings was , in the first ring and 1 ' in the second .
As will be seen below first ring gave results which were at first sight rather discordant with the others .
So I made a few special observations on the first ring with a very bright sodium-flame .
This was obtained , without the use of oxygen , with the aid of a thin platinum wire carrying a very small bead of carbonate of sodium .
The thin wire was raised to an intense heat in the hottest part of the flame , while sufficient liquid oozed down the wire from the molten bead to replace the sodium given off The flame was nearly if not quite as bright as if oxygen had been blown in , ] the definition of the first ring was decidedly improved ; though traces of other colours appeal.ed which were a little troublesome .
The measurements of the larger rings deserve a few remarks .
The rings were narrow and ] defined , so the needle-point , could be placed with great and another important source of error was renloved , for turning the polariser or analyser slightly would not appreciably displace the ring .
I believe the discrepancies between different measurements that appear in the tables arose chiefly from errors of , in reading the verniers .
Owing to the lateral displacement of the light when the plate was placed obliquely , the telescope had to be moved various distances up to about 1o to either side of its mean position .
To make allowance for this motion the telescope vernier had to be read .
Thus each diameter given depends on four vernier readings .
It is not then very surprising that two diameters should occasionally differ by ; while the mean of the four diameters is probably very accurate .
The patch of light alluded to previously as forming the field of view was , as the angle of incidence increased , gradually encroached on by the edge of the of the telescope .
This was avoided by moving the telescope .
But soon the edges of the quarbz plate began to encroach , and this was avoidable .
When the 150 th ring was on the needle-point the field of view was very narrow , so the readings were not quite as accurate as for the other rings .
Still I do not think the error of setting the needle-point was ever as much as 1 ' .
Notwithstanding the movement of the telescope the light at the high angles of incidence entered near the edge of the objectglass ; so I carefully tested the telescope for spherical aberration and found that no appreciable error could arise from this cause .
As a matter of curiosity 1 found it possible to obtain a reading of both sides of the 1 rd MR. J. C. Mc NEL AN EXPERIMENTAL TIGATION above observations relate to values of ranging from to .
For larger values I required a plate cut parallel to the axis .
So I had two new faces cut on PJate 2 , at right angles to the former ones and therefore nearly parallel to the axis , forming in fact a new ] , which I will call Plate 3 .
The two new faces were good planes and made an angle of about , with one another , and not more than 9 ' with .
the axis .
In the observations on Plate 3 , it was to set all four faces nearly parallel to the axis of rotation of the spectrometer table so as to secure that the axis should lie in the plane of measurement .
The bands were shadowy and compared with those seen in Plate 2 .
This was only to be expected , for we really have two distinct sets of bands due to the two sodium lines .
For a given value of the retardation is very nearly the same for either line , but if it amounts to 300 for it will be about gths for .
Thus the 300th band for will divide the distance between the and bands for in the ratio of 3 to 7 , now in the bands observed the retardation ranged from 253 wave-lengths to 312 .
So it was no wonder the bands were indistinct , the wonder was rather that they were measurable at all .
Fortunately however a large angular error in the measurement of a band only produced a small error in the retardation deduced .
To take an extreme example , in the first band an error of 10 ' in the diameter only produced an error of paxt in retardation deduced therefrom .
I took two sets of measurements of these bands with the analyser in one position and two sets with it in the rectangular position .
The 2nd , 5th , 15th , 30th , and 50th bands were measured in all four sets , the first only in the first set , and the 60th only in the two Iast .
On either side of the first band I took three readings , the extreme difference between two being 14 ' .
For the second I usually took two readings , the extreme difference in one case being 9\fnof ; mCthe next greatest being .
For the fiftl ] in four cases I took two readings , the extreme difference being .
But , as will be seen below , these differences are quite eclipsed by those due to small differences of temperature on different days .
Indeed if it was desired to get a very accurate measure of the change of with temperature a most powerful method would be to take similar observations to these , using a plate of quartz about 70 mm. thick , so that the bands for the two lines would nearly overlap , and thus narrow and sharply-defined bands would be presented in the field of view .
The Measurement of the Th , ickness .
The thickness of the plates was measured with a pair of oallipers made by ELLIOTr BROS .
These were fitted with a vernier reading to ousan dths of an inch , and the graduation was so good that it was easy to divide the vernier divisions into halves or even quarters .
To avoid any error arising from the jaw faces of the callipers not being strictly plane , I clamped two pins on to the jaws of the ] lipers with their heads opposed to one another .
The direct contact of the two pins ' heads was observed INTO THE FORM OF THE OF QUARTZ .
with a magnifying glass , and when the quartz was interposed each pin 's head was brought up into contaot with its ima in the surface of the quartz .
Care was of course taken to set the plate of at right angles to the length of the callipers .
This gave the thickness of Plate 1 , inch or mm. , of Plate 2 , 7865 inch or mm. These are probably correct within inch .
If I had measured Plate 3 in the same way I might have got an error of part in the ratio of its thickness to that of Plate 2 .
Now it was the ratio that I was anxious to get accurately , for , as will appear below , I was more interested to see whether would come out from my calculation a constant , than to find out what might be the value of that constant .
I had arranged that Plate 3 should be of nearly the same thickness as Plate 2 .
If then I could get an accurate value of the difference between them , it would supply an accurate value of the ratio .
This difference was measured with a spherometer reading to 001 mm. The plate was laid on three steel points firmly fixed into a table .
The three of the spherometer rested on a mahogany block cJamped to the table .
Each leg had been pressed gently into the wood to make a small pit , and found that the spherometer could be lifted off and replaced several times , and et give the same reading .
I found the difference of thickness to be 276 mm. within 002 mm. Thus I obtained the ratio correct to part .
.\mdash ; CALCULATION DISCUSSION 0F RESULTS , It is convenient to discuss the observations on small values of separately from the others .
Near the optic axis the additional separation of the sheets of the wavesurface , due to the rotatory property , has to be treated as of magnitude comparable with the separation given by HUYGHENS ' construction , both being very small .
While at a little distance from the axis the Huyghenian separation is greatly increased , and the additional portion sinks to the nitude of a small correction .
Thus for each case a special formula of approximation is appropriate .
The boundary may be drawn at , though either formula will do for some distance on either side of this boundary .
The Small Rings .
In my former paper I calculated the radii of the various small rings on MACtheory and compared their values with observation .
In order to compare the observations with other theories the results will now be presented in a different form , and it is convehient to introduce a quantity to denote the number of wavelengths by which one wave lags behind the other in air after the light has traversed normally a plate of quartz 1 mm. thick , the normal to whose faces makes angle with the optic axis .
Let denote the retardation actually observed when the light .
J. C. Mc CONNEL ON AN EXPERIMENTAL INVESTIGATION passes obliquely through the plate .
We have to determine the relation between and D. If we were free to assume that the two wave-fronts in the crystal might be treated as parallel this ] be very simple , and in my former paper I made this assumption ; but further consideration has convinced me that the assumption inadmissible , at any rate without a good deal more explanation .
It is true that the two wave-fronts are very nearly parallel , but it is also true that the retardation to be measured depen on the small difference between two nearly equal velocities .
I therefore give the following more rigorous investigation , which leads , as will be seen , to the same result as the questionable assumption .
Fig. 2 .
Let FGC in fig. 2 represent a plate of quartz .
Let ABCD be a wave-front ; of which the part AB has not reached the quartz , and the part BC is an " " ordinary " " wave-front in the crystal .
Let be the angle of incidence , the two angles of refraction .
Then the retardation of the ordinary wave in passing through the plate is the perpendicular distance between CD and ABE .
Similarly the retardation of the extraordinary wave is .
So the relative retardation is . . .
This is a well-known result , the above proof being due to Dr. ROUTH .
I inserted it for the sake of completeness , and to show that so far no approximations have been made .
Let be the two wave-velocities in the quartz , then LNTO THE FORM OF THE WAVE-SURFACE OF QUARTZ .
sm Let We shall see that is always than , so that although contains as a factor the squares of may be neglected , is smaller still .
From ( 2 ) we obtain sill Then Whenroe Again I have examined the effect of retaining the squares of and in finding relation , and find that the above equation is true within one part in ten thousand .
For the successive rings we have , where is an integer .
So ( 3 ) where is measured in millimetres .
All the nine theoretical formulae I have examined for the relative retardation the axis in take one or other of the following forms:\mdash ; or ( 4 ) where and are constants which have values assigned to them .
is the value of when , and is known from the rotation .
If be the rotation , measured in degrees , of a millimetre plate of quartz If the plate be cut strictly at right angles to the optic axis and the mete of each ring is .
Usiug the equation , where is the ordinary index , and eliminating between ( :3 ) and ( 4 ) we have MR. J. C. Mc CONNEL ON AN EXPERIMENTAL Thus , from the observed values of , we can obtain values of and for each ring .
According to one set of theories , according to the other should be always the same .
The observations on Plate 1 are exhibited iu Table I. The first line gives the value of ; the next four the observed values of the diameters taken in plane A or in plane with the principal plane of the analyser in the position indicated ; the sixth line the result of taking the mean of the two diameters in plane A and correcting it for the deviation of the axis from the normal ; the seventh the same for ; the eighth the mean of the two last .
The ninth shows the diminution of or for an increase of 1 ' in the diameter , so as to give an indication of the probable accuracy .
The next line gives the approximate value of , and the las two the values of and deduced from rin TABLE I. ( Plate 1 ) .
The observations on the small rings of Plate 2 are exhibited in Ta.ble II .
TABLE .
( Plate 2 ) , On looking at these tables we see that gives values steadily increasing with while } very nearly the same throughout .
I think that the final values of the of the four larger rings in each table should be correct within or while the first three should not be in error more than 1 ' .
In Table I. there is a line INTO THE FORM OF THE WAVE-SURFACE OF QUARTZ .
indicating the change in or corresponding to 1 ' in the diameter .
So we see that comes out as a constant within the limits of experirnental error except for the first ring in Plate 2 .
It was for this reason that I made a special set of measurements of the first ring in Plate 2 with a very light .
The values obtained in three different positions of the analyser and plate were , and , mean , mean temperature .
This gives precisely the same value , Plate 1 also gives a decided drop in for the first ring which is barely within the limits of error above assigned .
Now a change of in the value of employed in the calculation would alter by 0275 for the first ring in Plate 2 , and by 0344 for the first ring in Plate 1 .
Thus a decrease of two parts in the thousand in the value of the rotation would brin the deviation of from the constant value well within the limits of error .
The other rings are much less dependent on ; thus the same change in would in the second ring only increase by 0086 .
The most extended observations on the ation are those of SORET and SARASIN ( ' Comptes Rendus , ' tome 95 ( 1882 ) , pp. 635-638 , \amp ; c By the method of FIZEAU and FOUCAnLT they obtain for , for , mean ; while by direct observation with sodium light they obtain .
I have calculated from the mean of these , applying the temperature correction which is an ddition of 00324 for each degree rise of temperature .
The deduction of two parts in a thousand will bring the down to Another , and , in my opinion , better explanation of the peculiar result from the first ring will be given later on .
Omitting the first ring my observations give a constanlt value of We have to examine how agrees with the values assigned theories .
CADCHY gives where and are the wave-velocities at right angles to the axis .
LOIVIMEL gives ; KETTLER gives SARRAU gives The discrepancy between CA value and the observed value is outside any probable errors of observation .
It corresponds to a change of 4 ' in the diameter the 7th and 11th rings .
Such an error seems out of the question .
}IMEL and Thus SARRAU alone succeeds in explaining the observations satisfactorily , though VOIGT does put in a proviso that his may vary with In obtaining the above figures the difficulty arose that was known only roughly from spectrometer observations .
I have , therefore , had to use the value of deduced on SARRAU'S or MAcCuLLAGH'S theories from the observations on the very large values of , viz. , It might perhaps be objected that it is not fair to test one theory with the value of constant obtained from another .
But it should be remembered that the separation between the two sheets changes slowly with when is nearly , and that I have obtained a measurement for .
Thus on any theory I should have obtained much the same value of from this measurement .
Of the above writers CAUCHY , LOMMEL , and LANG give the actual expressions I have quoted .
The others only give equations to the wave-surface , from which I have had to deduce the expressions given above .
MAcCULLAGH , CLEBSCH , LANG , BOUSSINESQ , and VOIGT all give the same wave-surface , viz. , where is a constant depending on the rotatory power .
SARRAU'S wave-surface , neglecting certain constants , and , which he says are probably small , is Thus these six theories agree with the Huyghenian construction when the rotatory term is neglected .
This is a priori a great point in their favour , since we know from STOKES'S and GLAXEBROOK'S observations that in Spar the Huyghenian construction is a very close approximation to the truth .
The more complete form of SARRAU'S wave-surface will be examined towards the end of the paper .
The references are as follows:\mdash ; .
Roy .
Ir .
Acad. , vol. 17 ( 1837 ) , p. 463 .
CAUCHy , Ann. de Chim .
et de Phys. , ber .
3 , tome 30 ( , p. 68 ; a paper by JAMIN .
better value is , which gives for SARRAU .
Thus , though SARRAU'S formula agrees with observation much better thau any of the others , there is a small outstanding discrepancy in ths three larger rings in Table II .
, which is as hard to account for by errors of observation , as the great drop in the first ring .
This is discussed , and an explanation given near the end of the paper .
INTO THE FORM OF THE -SUIIFACE OF QUARTZ .
CLEBSOH , ' CRELLE'S Journal , ' band 57 ( 1859 ) , p. 356 .
SARRAU , Liouville , ser. 2 , tome 13 ( 1868 ) , p. 101 .
BOUSSINESQ , Liouville , ser. 2 , tome 3 ( 1868 ) , p. LANG , Pogg .
Ann. Ergbd .
, band 8 , p. LOMMEL , Wied .
Ann. , band 15 ( 1882 ) , p. 389 .
KETTLER , Wied .
Ann. , band 16 ( I.882 ) , p. 109 .
VOIGT , Wie\amp ; Ann. , band 19 ( 1883 ) , p. 898 .
For some account of most of these theories see GLAZEBROOK'S ' Report on Physical Optics , ' Brit. Assoc. , 1885 .
Corrections for error of in Plate 1 .
The calculations were made as though the optic axis were coincident with the normal to the faces .
To compensate for this certain small corrections have to be applied to the observed diameters .
The error parallel to plane A was 1o 45 ' , parallel to plane .
First consider the measurements plane , and take the error by itself It is clear that we may calculate what the error would be , on the assumption that MAcCuLLAGfi 's theory were true , for any value of which is near the true value , and take this with sufficient accuracy to be the actual error on any one of the theories .
Let giye a band on one side .
The angle of refraction is then , so we have Solving we get , and from the equation we obtain , the angle of incidence on that side .
Similarly , putting for we obtain the angle of incidence on the other side .
The observed diameter is then , and we take to be the value of which would giye the dark band in a piece correctly cut , where .
But the true value is where By calculatin and we obtain the proper ection to apply to , and therefore to the observed diameter .
In practice I found that the correction could be obtained with sufficient accuracy on the assumption that On this score I had to subtract from the observed diameters of the successive rings had to be added 2 ' On the whole I had to add Secondly , for the measurements in plane B. Now only , so the corrections on this score are ible .
In thig case the plate was tilted till the apparent position of the axis in .
lay the plane of measurement .
So was no longer the radius of the rlngs and consequently had to be altered .
To allow for I had to add to the successive diameters The corresponding corrections for Plate 2 would be infinitesimal .
The Large Rings .
Since MAcCuLLAGH'S theory will be found to be in tolerable agreement with the measurements obtained for the larger values of , it is convenient to take it as a basis and discuss later what modifications would make the agreement more MAcCuLLAGH starts by assuming the equations of motion ( 5 ) where are the displacements parallel to at any time , and the axis of is the wave normal .
Also ( 6 ) is a constant .
From these equations we get the wave-surface ( 7 ) which has three sheets ; the third sheet , however , corresponds to an index of refraction of about 40,000 ; purposes we need not consider it .
INTO THE FORM OF THE WAVE-SURFACE OF QUARTZ .
Now we shall find the whole effect of is so small , not eeding part of the chief terms in the final formula , that we may not merely reject its squares but also put in the small terms , except , of course , in the factor itself Resuming the rotation used for the smaller rings we have ( 8 ) and By equation ( 1 ) the retardation is Assuming that the nornrl to the plate coincides with the opti we have for the ordinary wave So , neglecting C. For the extraordinary wave with sufficient accuracy .
Hence where To obtain the part of not involving we have ; therefore Similarly 318 MR. J. C. Mc ON AN EXPERIMENTAL INVESTIGATION therefore term involving C. This term where is to be put zero in each differential coefficient after differentiation .
and therefore Similarly therefore ( 9 ) This equation is not true for small values of or , for we have assumed in the course of the proof that is small .
We cannot , therefore , put and obtain from ( 9 ) , but by putting in ( 7 ) we obtain Equation ( 9 ) is not convenient for calculation , for each of the radicles in the bracket is very large compared with their difference .
And it is on their rence that depends .
Remembering that , we may put ( 9 ) into the form ( 11 ) Of the quantities involved in this equation , and are known with great accuracy , certainly within one part in ten thousand .
also is known with considerable accuracy , but the value of is uncertain , to the extent of perhaps 1 per The third term is small compared with the others , and so requires only approximate INTO THE FORM OF THE WAVE-SURFACE OF QUARTZ .
values of and .
We may therefore use ( 11 ) as an equation to find inserting the observed value of for each of ths rings measured .
If the value of does not come out the same for each ring , then theory is inconsistent with the facts , unless indeed the discrepancies may be attributed to errors of observations .
This mode of presenting the results of the measurements possesses the great advantage that it shows at once how far the observed surface departs from the theoretical one .
For instance , if for a certain value of comes out part too large , this indicates that the real separation of the two sheets is very approximately part greater than that given by the theory .
This is true throughout , except for the first two rings .
The results from the large rings in Plate 2 are exhibited in Table III .
The first line gives the value of .
The next four lines give the observed diameters , two sets of observations having been taken in plane A and two in plane B. The oblique strokes indicate the position of the principal plane of the a1lalysing NICOL .
To each set is attached the mean temperature at which it was taken .
The sixth hne gives the mean observed diameters .
The seventh the values of deduced from MACCULLAGH'S theory .
The eighth the change in , which would be produced by an error of 1 ' in the measured diameter ; in this , to save space , the first five ciphers after the decimal point are omitted .
The ninth line gives the value of TABLE III .
( Plate 2 ) .
The eighth line is an important one since the accuracy with which is determined varies so much .
Thus , while an error of in the diameter of the 150th ring gives a change of one part in eleven thousand , the same error in the 20th ring gives a change of one part in fifteen hundred .
To bring up the numbers in the seventh line to one constant value we shall have to suppose errors of 2 ' , , and 1 ' for 32 , 52 , and 77 respectively .
But it is very improbable that the mean observed value of the diameter of any ring , except , perhaps , the 150th , is more than 1 ' in error ; so theory must be condemned .
Now , the explained the observations on the ] rings satispoint in discussing the results from Plate 3 .
The Bands in As increases the correction due to the rotatory term becomes rapidly s1naller .
Thus for the 150th ring in Plate 2 the change in the deduced value of , which would be made by simply calculating from the Huyghenian construction , would be only one part in five thousand , and it may be readily shown that in Plate 3 it would never exceed one part in ten thousand .
These figures refer to theory , and in SARRAU'S expressions the rotatory term is smaller still .
By neglectin this term the formulae for Plate 3 are greatly simplified .
We have now ' Since the plate is cut parallel to the axis we have for the extraordinary wave ; therefore and ; while But and therefore , or In calculating from the observations was taken to be half the angular distance from any band on one side of the normal to the esponding band on the other side .
For convenience I will call this distance the diameter .
Then was deduced from by INTO THE FORM OF THE WAVE-SURFACE OF QUARTZ .
the relation , and determined from the formula ( 12 ) .
But now the question occurs , what value must be given to .
In Plate 2 was easily determined for the first ring , and therefore for all the rest , by a rough calculation from the known rotatory power .
But in Plate 3 is so large that tolerably accurate values of the constants are required to distinguish between one integer and the next .
Indeed , without the observations on Plate 2 , I should have been unable to decide between five six different integers .
However with the aid of these observations the difficulty disappears .
Using the value in equation ( 12 ) I obtained for the second band , reckoning from the normal , after correcting for temperature , , so I concluded that the true value was 311 .
Fig. 3 .
erhaps the best evidence on this point is supplied by the diagram , fig. 3 .
In this diagram the ordinates represent , while the abscissae are the values of .
The small circles are got from SARRAU'S theory , the crosses from MAcCuLLAGH'S .
where the two theories agree there are two ]ines of small circles .
The upper line is obtained on the supposition that for the second band , the lower line on the supposition that The diameters obtained in four sets of measurements are given in Table .
The first line gives the number of the band reckoned from the normal .
The seventh the resulting values of , omitting the decimal point and two ciphers , and the the change in corresponding to 10 ' in the diameter , omitting the decimal point NDCCCLXXXVI .
322- MR. J. C. Mc CONNEL ON AN EXPERIMENTAL INVESTIGATION and five ciphers .
This line shows that the large discrepancies between different measurements mean only very slight variations in .
They are no doubt .
due mainly to changes of temperature .
The temperature of course varied one or two degrees during each set of observations , and no doubt the temperature of the quartz was slightly different from that.of the thermometer .
The ninth line gives the values of , and the tenth the values of corrected to the temperature .
This was the mean temperature of the measurements of the large rings in Plate The correction to per degree of temperature .
The last line gives the values of , taken on the supposition that for the second band and corrected for temperature .
We notice that the greatest difference between the numbers in the tenth line is 0000006 , being about one part in six thousand .
This is a strong confirmation of the truth of the Huyghenian construction for the region far removed from the axis , extending as it does from .
is TABLE IV .
( Plate$ ) , Referring to Table .
we see that the numbers given there deduoed from SARRAU'S theory are uniformly lower .
This may be partly due to a wrong value of but , as I have said above , I do not think the error in the ratio of values of for the two plates exceeded 1 in 10,000 .
eover , SARRAU'S diminishes steadily with throughout .
This is very apparent in the following figures , which are obtained by grouping three or four rings together , allowing as a possible erro from the mean the cornbined effect of in temperature and in diameter .
In the case of the three rings , the 2nd , 3rd , and 4th , , is allowed .
In the case of the 1st 1 ' .
bands to lies between 0037941 and 0037949 4 rings .
to , , , , .
0037928 , , 0037940 3 .
rings to , , , , 0037910 , , 3 rings to , , , , , , 0037916 INTO THE FORM OF THE WAVE-SURFACE OF QUARTZ .
3 rings to .
lies between and 0037954 rin , , , , 0037620 , , 0037808 Taking into consideration the probable accuracy of the observations the discrepancy is about equally strongly marked in the case of the first ring and the three rings to , and to a minor degree in the case of some of the others .
The observations on Plate 1 are not here included , as we do not know the ratio of the thickness of Plate 1 to of Plate 3 with sufficient accuracy . ?
] Up to this point we have been treating certain consta1lts , and , which appear in SARRAU'S expressions as negligibly small .
We will now examine whether by the aid of these constants we can obtain a more satisfactory agreement with observation .
SARRAU'S wave-surface is given by Putting , we obtain for the equatorial radii and We might treat this equation to the wave-surface exactly as we have treated the simpler one , and obtain values of from the observations on the different values of .
But it may be shown without difficulty that if we equate the values of previously found to where we shall attain the same result .
In this it is , of course , assumed that and are small .
I have found that a suitable value of is .
Using this value we obtain the following figures:\mdash ; to lies between0037941 and 0037949 to , , , , 0037931 , , 0037943 to , , , , 0037919 , , to , , , , 0037914 , , 0037956 to , , , , , , 0038037 * Added May 31 , 1886 .
MR. J. C. Mc CONNEL ON AN INVESTIGATION Thus the value0037943 falls within each pair of limits .
The last line of each list gives the results from the first rin obtained with the use of SORET and SARASIN'S value of the rotation .
Thus the sudden drop of for the first ring , -which was the most striking eccentricity of the observations on the small rings , is explained at once by the more general form of SARRAU'S theory , without the necessity of assuming an error in the rotational coefficient .
Of course we have had to assign an arbitrary value to .
But the form of the correction , viz. , , was fixed by the theory , and was exactly the form that was required , giving , as it does , the correction for some thirty times as great as that for We have now compared SARRAU'S theory and the MAcCuLLAGH group of theories with observation for all our values of .
With regard to the remaining three , has given , so far as I know , no general expression for the wave-surface , while LOMMEL and KETTLER are so widely at variance with the Huyghenian construction that I have not thought it worth while to make a detailed comparison .
I find , however , that the values of deduced on KETTLER'S theory from my observations would differ by about 3 parts in 1000 for and , whereas according to the Huyghenian construction they agreed within 1 part in 6000 .
We conclude , then , that out of the nine theories we have examined the only one which agrees with obselvation is that of SARRAU , and that to make the agreement as perfect as possible we must choose his constants , so that or IV.\mdash ; VALUES 0F CONSTANTS EMPL0YED .
I have used RUDBERG'S values for and , viz. , MASCART gives I have taken mm. This is the mean of ANGSTROM'S values for and FIZEAU has given coefficients of the expansion of quartz with temperature along the axis and at right angles to it .
He also measured the change of and with temperature .
These are so small as to be negligible in the present case except for the INTO THE FORM OF THE WAVE .
SURFACE OF QUARTZ .
change of .
This has been specially investigated by DUFET who finds the change per degree The rotatory constant I have discussed at length above , in connexion with the small rings .
The thickness of Plate 1 was found to be , of Plate mms .
, while the ratio of Plate 3 to ] was Plate 2 was cut from the same crystal of quartz as ] The rotation of the plane of polarisation produced by either is related to the direction of the light as the rotation to the translation of a right-handed screw .
Such is usually called left-handed quartz .
The observations were taken in the Cavendish Laboratory , Cambridge , during the months of March and June , 1885 .
V.\mdash ; PREVIOUS MEASUREMENTS .
The only previous measurements of the retardation in quartz with which I am acquainted are those of JAMIN in 1850 Chim .
, those of HECHT ( WIED .
Ann. , band 20 [ 1883 ] , p. 426 ) , and my own ( Proc. Camb .
Phil. Soc. , vol. 5 [ 1883 ] , p. 53 ) .
JAMIN rotated the plate of quartz between two crossed NICOLS as I have done , but he used parallel light , taking the reading when the field of view appeared darkest .
Most of the measurements agreed with CAUCHY'S theory within one part in thirty , though here were a few wider discrepancies .
This he considered satisfactory agreement .
He obtained similar results from another series of experiments in which the axis ratios were also involved .
HECHT'S method was not very dissimilar to that of JAMIN .
There is an important oversight in his paper .
In estimating the probable error of he makes the tacit assumption that a certain quantity he calls is known accurately ; while in fact the main part of the error results from the inaccuracy of this .
Correcting this I obtain -using my own notation\mdash ; for whence I d.educe for the former , for the latter The results obtained in my own earlier paper were vitiated , partly by the addition of some corrections founded on an erroneous argument , and partly by the bad faces of the plate examined .
The argument was as follows .
The observations were made by bringing the cross wire up to sensibly dark part of the band , first on one side , then on the other .
Then I argued that I knew the position and breadth of the sensibly dark region of the band , but I did not know that by taking the middle of this I should obtain the darkest point .
The distance from 326 MR. J. C. Mc ON THE FORM OF THE WAVE-SURFACE OF QUARTZ .
band to band becomes smaller as we go further from the axis , and I took as an approximation that the darkest point would divide the sensibly dark region in the same ratio as the band itself divides the distance the two neighbouring bands .
Here was the mistake .
For the sensibly dark region is that part of the band where the intensity of the light falls below a certain fixed value , and the intensity of the light obviously varies approximately as the square of the small angular distance from the point of total extinction .
It might then be expected that this darkest point would lie practically at the middle of the sensibly dark region .
This opinion was confirmed by a closer examination , with the aid of Sir GEORGE AIRY'S formula ; for the intensity of any point of the field .
By the removal of those corrections the discrepancies I obtained between theory and observation would be reduced one-half , and the remainder may be easily accounted for by the imperfection of the polished faces .
I knew that the rings would not be much altered by curvature of the faces ; for when the incidence is nearly normaL a slight deviation of the path through the quartz does not materially increase the length of the path , so I contented myself with noticing in the measurements with the callipers that the faces appeared tolerably parallel and plane .
Lately , however , I examined the crystal again , and optical tests showed that the curvature of the faces was quite sufficient to explain the discrepancy of some 5 ' between theory and observation .

