]\gt ; In examining the dispersive powers of a great varicty of lasses prepared by the late Rev. W. Vernon Harcourt , I had occasion to examine several prisms were too much iated to show clefrly en the boldest dark lines solar spectrum .
I found I able to get a fair measure of the of these by a method depending on the izing of one prism ftnot Lcr .
If method succeeded even with such prisms , it to reason it would be still more with prisms of For the construction of an objcctive we require one datum as ards the dispersions , namely , the of the d , or rather , the riitio which on being treated as if it the ratio of the dispersions gives the best results in practice .
If it were not for ] , the matter would bc comp simple .
The ratio of the dispersions would bc the for whatever interv al of the spectrum it were ; we should merely have to take two well-defined lines , ated f nearly as may be at the extrcmities of spectrun ] , of observation should be divided as ] , a ) as ) ) } to measul.e the efractive i two ] usscs { of ' two lines , and to take the ratio of the i in ) line to the other .
in conscqnencc of a different ratio to the ] oose ; are obliged , unless we adopt some , to vc more than two lines each glass ; Lave got the esults , Prof. Stokes on an accurate Method that is to say , the indices for several fixed lines in each glass , it still becomes a question what it is best to do with them so as to produce the best resuIt .
Fraunhofer proposed an empirical rule for combining the results , * but remaxked that the number thus obtained was not exactly that which answered best in practice .
I have elsewhere shown reason to think that the rule may be taken to be simply that the ratio of dispersions to be chosen is that for an infinitely small portion of the spectrum at its brightest part ; or in other words , that the ] of the combination of two glasses must be made a maximum or minimum ( it is practically a minimum ) at the brightest part of the spectrum .
The refractive index of one glass may be expressed in terms of that of another , or of some standard glass , by an interpolation formula with three , or at most four , terms .
The most accurate observations with isms are only just sufficient to show that a fourth term is needed , a1id for the practical purpose of the construction of objectglasses , where in consequence of irrationality the ends of the spectrum are sure to be a good deal out of focus , it will be amply sufficient to confine ourselves to three terms .
The three coefficients in the interpolation formula may be determined by observing in each glass the indices of l.efraction for three selected ines , thou it is well to observe more than three lines , and combine the results .
The angular extent of the spectrum being but small in practical cases , the necessity of determining three constants by bservations de in it requires great delicacy of measurement .
Small el'rors of observation would easily produce an error in the deduced ratio which would be sensible , or even material , in practice .
For the actual construction of an object-glass we require , as has been already remarked , the knowledge of only a constant relat- ing to the dispersion , not of two ; and if we can find some test whereby to know when the required condition is satisfied , we may dispense with buch extreme accuracy in the angular measurements .
Such a test is afforded by the secondary tints , which change with extrcme rapidity when the refractin , angle one of the mutually itcbromatizing prisms , or the position of one of tLem in azimutl ] , is altered .
If a moderately narrow object with vertical sides , black on 1 : ground , or white on a black ground , be viewed hrough opposed prisms , one of crown and one of fiint glass , with a small telesco } ) and the prisms be set to achromatize each other as as m.ay be , it will be found that the htest touch tering one of the prisms in azimuth aIters notably the secondary tints .
The secondary tints in an objective are readily shown by directin *riften der K. Akad .
Wiss .
zu ' '1 ol .
1 , p. 215 .
" " Rcport of the British Associ for 1855 , ' ' } ii , p. 11 .
1878 .
] of ining Ratio of ersions .
the telescope to a vertical line sepat.ating light from dark , such as the of a chimney seen in the shade ainst the sky , and covering half the object-glass with a screen a vertical .
So delic is this test that on testing differenb telescopes by well-known opticians , a difference in the 1node of achromatism may be detected .
The best results are said to be when the secondary green is intermediate between green and ellow .
This corresponds to making focal length a minimum for the brightest part of the spectrum .
To enable me to form a judgment as to the sharpness of the test furnished by the tint of } secondary green , as compared with the performance an , I tried the following experiment .
A set of parallel lines of increasing fineness was ruled with ink on a sheet white paper , and broader black as on it as well , parallel to lines .
} ) was placed , with ) the ck lines vertical , at a in a , and was viewed two opposed prisms , one of crown glass and when other of flint , of angles as nearly to achromatize each other in ) positions of minimum deviation , and then a small telescope .
The achromatism was effected , and ried in aracter , by moving one of the prisms htly in azimuth , and after each iteration the teGscopc focused afresh , to get the sharpest vision that could be had .
I that the azimuth of the prism was within cided l rower limits by the condition that the secondary green should be of such or such a tint , even no attempt was made to determine the tint erwise than by memory , than by the condition that the vision of the fine lines should be as sharp as possible .
Now a small element of a doubie lass may be regarded , so far as chromRtic compensation is concerned , as a pair of opposed prisms ; and therefore we may .
that the tint of the secondary green ought to be at the very least as sharp ; test of the of ths matic compensation : ) the actual of the telescope .
And @lch , to whom I mentioned the test , found to be actually case in the progress of ) construction of the ctor for the Royal Society , the instrument at prcseut in the hands of It follows ereforc t at two opposed prisms , representing glasses to be employed in an objective , clrC to be deemed to each other when one of the two secondary is about between yellow and green .
The condition of aehromatism of two opposed ) is given in the ordinary on optics , , so far as I have seen , ratber an optical curiosity than as a matter practical utility .
Brewster in his treatise on " " Ncw Inst r menis ) !
alludes to it as furnishing a ivable of ) siye powers , but mentions it only to it .
I cannot why , for at least with the modifications which I ve been to Prof Stokes on an accurate Method [ June 20 , troduce in putting it in practice , I find it to be no less excellent than easy .
The ' are a good deal simplified by the prisms to be compared achti'successively one and the same prism clIosen arily , and retained in a fixed , instead of them dircctly achromatize each other .
I will first describe the method as I used it , ' that as the prisms for which it was primarily were not of good glass , I was content with less perfect arrangements than would have been desirable for really good glasses .
Nevertheless the apparatus when used with good prisms gave very results .
The object observed was a vertical slit , so wide when viewed through the opposed prisms a broad white stripe was seen , mercly fring.ed at the two with the secondary green and purple .
The slit was at one end of a horizontal plank , near the other end of which was fixed a verlical lens of about four focus .
The slit was in the principal focus of the lens .
The pIallk rested on ttlrec knobs , one under the slit , the other two near the other end .
This mode of support prevented torsion of the plank , which }vould have produced a lateral derangement of the slit .
The rays from the slit , rendered parallel by the lens , fell upon a prism , which I will call the primary prism .
This prism is fixed during the observation .
Its ange and persive power , and the azimuth in which it is set , are within wide limits .
The beam emerging from the primary prism falls , at a little distance , on a second ' on a stand movable along with the verniers of a protractor reading to minutes ( for the loan of which I am indebted to Professor Millcr ) , which latter rests on the pIank .
As a matter of convenicnce , the prism has an independent motion in azimuth , but when once ] aced in a conve1lient position it is not afterwards moved except with the verniers .
The beam refracted and dispersed in contrary directions by the two prisnls is received on the of an achromatic telescope unconnected with the aduation , with the of the slit , if so wide an aperture may so be called , is viewed .
In general the of the aperture arc seen coloured blue and red , which colours , on turning the prism its azimuth of achromatism , are into red bluc by secondary and purple .
That ge of the slit at which the ' is seen is alollc attended to , and is treated as the fiducial edge .
isms are deemed to omatize each other when this colour is between yellow and Should the observcr wish to aid his meant b observing the purple as well , it would bc propcr to use an apcrture of the form represented in the figurc , so that ) grcen purple should be seen and left , a common , and therefore at the same ] of incidcnce , or rather at angles of incidence the same hori1878 .
] of determining Ratio of Dispersions .
zontal projection .
I do not know that anything would be gained by his ; I have found it sufficient to attend to the green .
The employed had best be of the sky , reflected horizontally by a -glass .
It should be fairly bright , but not approaching to dazzling .
Thus the light reflected from the sky near the sun vould be too bright .
The room need not be darkened ; in fact , it is better that the eye should be kept fresh for the appreciation of colonr by ally looking about on ordinary objects .
A simple collimating lens such as I have desc.ribed is sufficicnt , though doubdess an achromatic would theoretically be an improvement .
It should bc of focus , at least in the case of a simple lens , lcst displacement right or left of the middle of tho incident beam should troduce t minute dispersion due to the lens cting as a prism .
If the middle of the beam be not quite central , that does not signify , provided eccentricity be constant ; for then the dispcrsion is mcrely added to that of the pl.imary prism , which is arbitrary .
To cnsure constancy of incidctlce on the lens , it is well to limit it by an with vertical sidcs , to that beam employed is wide enough to fill ] it is well to take care the beam falls centric , or pretty fairly so , on the achrom } telescope .
But what is of much more cons ) is thnt the ray passing through the optical centre of the should centrically , as otherwise the dispersion of the eye-piece for eccelltricil pctlcils would altcr the tint .
It is wcll , thercforc , that tclescol ) should be cross wires .
telcscopc is movod a littlc till tidncial is on the cross wires the tint is ) The tion of the of achromatism is { hc capital observation on tho of tl1c rebult depends ; in oomparison with it , the rost of the onts r be decllc exact .
a number of vatioJtS of zimuth shonld be taken , and the mcan adopted .
The will with circumstances , but it be tnken ordiuarily as a few miuutes . .
xxv11 .
Prof. Stokes on an accurate Method [ June The mean reading for achromatism by itseIf alone gives nothing , and must be combined with another determination in order to be available .
The second determination which I found it most convenient to make was of the angle of incidence of a particular part of the spectrum on the second prism for a known azimuth of this prism .
If the brightest part of the spectrum were marked by a definite line , we should choose that .
As it is , the line , though not exactly at the brightest part , lies sufficiently near it for our purpose , as will be better understood presently .
No dark lines can be seen in the reflected light , because we are using an aperture , and not a mere slit .
But if the slit be illuminated by a soda flame instead of , a well-defined yellow image.of the aperture will be seen , the fiducial edge of which can be pointed at with precision .
To determine the angle of incidence I have employed two methods , both of which I think it will be worth while to describe .
The first is the less direct , and involves a little subsequent calculation , but has the advantage of not requiring any additional apparatus beyond what is wanted for the determination of the azimuth of chromatism .
In carrying out the first , and in applying either , I suppose the angle of the prism and its index for the line known from a determination in the usual way .
First Method.\mdash ; The azimuth of minimum deviation for the line I ) could be determined pretty fairly by bringing the wire of the viewing telescope to the fiducial edge when the aperture is illuminated by a soda flame , the ed ge is in or near its stationary position , and taking the mean several determinations of the azimuth at which the edge is stationary .
The angle of incidence for minimum deviation is virtually 1neasured in the process of finding the index , and is therefore known ; and by applying with its proper si gn the difference of mean readings for achromatism with daylight and for minimum deviation with a soda flame , we get the angle of incidence for achromatism .
This method is mentioned merely as naturally leading up to that actually employed .
It must be rejected as too slovenly , since the uncertainty of the determination of the azimuth of minimum deviation is liable to be greater than that of the azimuth of achromatism .
If we place the wire of the viewing telescope some way o# stationary position of the fiducial edge , there are two azirnuths of the prisrn at which the edge will be on the wire , across which it will move with a finite velocity as the prism is } in azimuth .
Hence either azixnut could be determined with accuracy , and if they were equidistant from the azimuth of mimmum deviation , the latter could be determined at once .
, however , is not the case , but nevertheless the azimuth of minimum deviation can be determined from the result .
Since the course of light may be reversed in refraction , it readily follows that in passing from one to the other of two azimuths which 1878 .
] of determining the Ra of Dispersions .
give equal deviations , the angles of incidence and ence are simply interchanged .
Hence when the prism is turned from one to the other of two azimnths for which the image of the fiducial edge is on the wire of the telescope , the angle moved through is equal to the difference between the anges of incidence and ence in either position .
Both , and therefore their difference , can be determined with accuracy , provided the azimuths be sufficiently remote from that of minimum deviation .
The process of observation , then , is this .
Set the prism a good way , such as or , from the azimuth of minimum deviation , and read the graduation .
Turn the viewing telescope till the fiducial edge is on its wire .
care to keep the telescope fixed , turn the prism through the position of minimum deviation till the is on the wire , and read again .
Let ' be the anges of incidence emergcnce or emergence and incidence , of which let be .
Let 2 be the , of the prism , 2 the sum of 2 and the deviation , the meastlred angle through which the prism has been turned , or .
Since the sum of the , we represcnt them by and .
Let be denoted by 2 .
Then we shall have\mdash ; Eliminating from these two equations , we find\mdash ; .
( 1 ) , which gives / , and then , i. e. , or , is known .
The an gle of incidence for a known reading of the graduated circle known , we have onJy to apply the difference between this reading and that for achromatism to that of incidence in order to get the angle of incidence for the azimuth of romatism .
ilTcthod.\mdash ; In this a little telescope is used , which is attached to the stand of the prism , so as to move with the verniers of the circle .
The telescope need not be achromatic , but has cross wires in its principal focus .
After determining the azimuth of achromatism , slit is illumiby soda flame , and the prism with the verniers turned till the fiducial edge is on the wires of the measuring telescopc , when the circle is read .
The prism is then removed , and the telescope urned till the fiducial edge is seen directly , and the circle is read .
Half the supplement of the diffcrence of the gives the angle of incidence whon the reflected image was on the cross wires ; and by applying the difference of readings for reflection and Prof. Stokes on an accurate fethod [ June 420 , nchromatism , we get the angle of incidence in the position of achromatism .
This angle , 1/ / , having been determined by either of the above methods , we have , by the known formula\mdash ; ( 2 ) , the ]etters with the suffix , refel .
to the second prism .
For the prisms would achromatize each other , as is supposed in the deduction of the above formula in treatises on optics , under the same circumstances in which they would achromatize , in succession , the same spectrum .
In the applicatiOn of the , it is to be remembered that , of the angles , the former is that which ] on the side of the white light , and is , therefore , the angle of incidence for the first , but of emergence for the second , of two prisms which mutually achromatize each other .
In the application of the formula ( 2 ) the angles belong , strictly speaking , the brightest part of the spectrum , which for shortness I will call , for which the value of the differential coefficient , : is supposed to be sought .
But the distance of from is so small that it will hardly make any sensible error if we use the values of the belonging to , for not only is the correction to the product for either prism very small , but the two corrections are in the same direction , and therefore tend to neutralise each other in the ratio of the products , with which alone we are concerned .
If , however , we care to introduce the correction , it can be done at the expense of a little additional calculation .
In a crown glass the index for may be taken as greater by than that for , and in a flint glass as greater by multiplied by a rough value of .
The angle of emergence may be taken to be the sa , me for as for D. For the deviation , regarded as the function of the index , is a maximum or minimum for ; and bein so near , the difference of deviations for and will be quite insensible compared with the errors of observation of the azimuth of achromatism , with which it is associated .
Let the letters , \amp ; c. , refer to , while , \amp ; c. , refer D. is obtained by observation , and must , in any case , be calculated from thence by the known values of and .
We have now merely to calculate the and for each glass from the formulae substitute these values in the equa , ion ( 2 ) instead of those belongto the line D. primary prism had best be made of very low flint glass ( or else 1878 .
] of determining Ratio of Dispersions .
be a compound prism formed of two prisms of crown and flint glass respectively , with their angles turned the same way ) , so as to fall about midway between the glasses to be compared , and thereby divide between them the irrationality which has to be encountered in the observation .
The ) of the azimuth of achromatism is most , accurate when there is little or no irrationality ; and , if preferred , the crown glasses might be compared with a standard crown , and the flint glasses with a flint , the primary prism being in one case any crown glass prism thati happens to be at hand , and in the other case a flint glass prism .
The crown glass to be measured being compared with the standard crown by using them in succession to achromatize the same primary crown in the same position , and similarly for the flint glass to be measured and the standard flint , we can deduce the ratio of the dispersions of the crown and flint glasses to be measured if we know , once for all , the ratio of the dispersions of the standard crown and flint glasses .
This may be determined by a specially carefill series of Qervations of the kind abovs described , made once for all , or , if preferred , by the method of indices .
The direct comparison of a crown and flint glass is , however , ) accurate , especially if a glass of intermediate quality be used for the primary prism , that I feel satisfied it will suffice for practical purposes .
It is hardly necessary to observe that , if the primary prism be of intermediate quality to the two compared , the hand the aperture will be the fiducial edge in the one case , and the left hand edge in the other .
In saying this , I assume that we have not inclinations or differences of angles to deal with , for there is a -irrationality observed even when two prisms of the glass , but different angles , achromatize each other , which is , , ordinarily so small that it may be neglected in comparison with the real tionality of the By turning the primary prism into ; different azimuth , or substituting a primary prism , and repeating the observation , estimate may be formed of degree of reliance thnt may be placed on the results .
To an idea of the of of which the results are SCePtible , I subjoin a fcw numbers cted from my note-book .
The prisms tcd , were experimental prisms of glass of different compositions .
They were lore or striated , were good to ) the principal fixed lines of spectrum .
In tho differc1lt doterminations of th ratio of dispersions , the primary prism was set at difIercnt .
In the indices ) nces tlxe are exhibited .
to 9@ , mcall , ; dilfercnccs , Profs .
Living and Dewar [ June 20 , ; mean , differences , \mdash ; It will be seen that , even with prisms such as these , by taking the mean of different determinations , the uncertainty can hardly be great as ome half per cent. Extremely small prisms are quite sufficient for the determination of the ratio of the dispersions of the glasses by the above method .
It may , however , happen that an opticia ; n cannot afford to remove even so small a piece of glass from a disk intended for an objective , and has not a specimen of glass on the identity of which with the glass of his disk he can thoroughly rely .
In snch a case it is necessary to determine the constants of the disk by means of facets cut on the disk itsclf .
A heavy and costly disk be treated like a small prism , and mounted on a small graduated instrument in the manner I have supposed a small prism treated .
To compare the ratio of he dispersions of two such disks , one of crown glass and the other of flint , the most convenient way would seem to be to leave the disk fixed , let the light pass through it first , and then achromatize it by a small prism of very low flint glass , mounted on a small graduated instl.ument in the manner already explained .
The dispersions of the disks vould be compared with each other by comparing them in succession with the same intermediate prism .
This , however , demands an additio1ial determination beyond what was required in the otlJcr process , since the prism through which the first passes is not the same in the two cases .
The element which best lends itself to measurement is the angle of incidence on the first surface .
The most convenient mode of this must depend on the eneral disposition of the apparatus adopted to take the measurements for the angle of the disk and the deviation of some one line , must be made in any case ; it is accordingly best left to the choice of the observer .
XI .
" " On the Reversal of the Lines of Metallic Vapours By G. D. , Professor of ChcmistIy , and J. DEWAR , M.A. , F.B.S. , Professor , University ofCambridge .
No. III .
Received June 19 , 1878 .
In our last communication to the Royal Society we described ccr{ain absorption lines , which we had observed to be produced by the vapour 0 magnesium in the presence of hydrogen , and cerfain other lines which were observed when potassium , and others when sodiuln , present , in addition to nesium and hydrogen .
lines correspond to no known lines of those elements ; but , inasmuch as they appeared to be regularly produced by the mixtures

