]\gt ; Optical Dispersion : An Analysis of its Actual Dependence upon Physical Conditions .
By T. H. HAVELOCK , M.A. , D.Sc .
, Armstrong College , Newcastle-on-Tyne .
( Communicated by Prof. Sir Joseph Larmor , Sec. .
Received October 10 , \mdash ; Read November 24 , 1910 .
) CONTENTS .
PAGE 1 .
Introduction 492 2 .
Refractive Index and Physical State 494 3 .
Temperature Variation : Organic Liquids 496 4 .
Carbon Disulphide and Water 497 5 .
Organic Liquids ( Falk ) 500 6 .
Variation under Pressure 602 7 .
Artificial Double Refraction in Liquids 502 8 .
Temperature Variation in Solids 605 9 .
Change from Vapour to Liquid 508 10 .
Absolute Values of the Variable 509 11 .
Dispersion Formulae : Absorption Maxima 512 12 .
Displacement of Absorption Maxima by Pressure and Heat. .
617 13 .
Summary 522 1 .
Introduction .
The influence of a transparent medium upon light passing through it may be ascribed to two factors ; one is the existence , in each individual molecule or particle , of vibrating parts with certain natural frequencies , and the other is the physical condition of the aggregation of molecules composing the medium .
Suppose that a typical vibrating part is an electrified particle of charge and mass , whose free vibrations are given by equations of the type Let X be the corresponding component of electric force in the incident light wave .
The impressed force , to be supplied on the right-hand side of ( 1 ) for the actual motion , is not simply , but has an additional term to express the effect of the surrounding molecules ; the simplest conception of this effect indicates a term directly proportional to the ayerage polarisation of the surrounding molecules at each instant .
Hence , instead of ( 1 ) , we have an equation .
Optical : its upon Physical .
493 As regards the numerical value of the co-efficient , little can be said beforehand on theoretical grounds ; certain considerations , iven in the present connection by Lorentz and by Larmor , indicate that for gases and liquids should be approximately The i1lvesGigation is an attempt to extract from available experimental data , information about the numerical value of and its variation with changing physical conditions .
The problems concerned are those of optical dispersion in isotropic oeolotropic media , and the changes produced by varying density , pressure and temperature , or by the action of extel.nal electric or magnetic fields .
attempt is made to explain the intimate mechanism of these phenomena , that is , the physical process by which the value of is altered ; rather , the term is used as a possible means for the various effects under one formal scheme , at least for a first approximation .
The method is , in to the usual procedure of .
absorption effects by a term in into equation without thereby completely the physical process involved in absorption .
If we ignore the effect of molecules in equation ( 2 ) , that is , if we put zero , the dispersion formula which can be deduced is of the type where is the refractive index of the medium for wave-length is a constant associated with the ) rating e of type , and is pro- portional to the number of such ) unit volume , is the corresponding to the natural frequency of the same type , and the summation 2 extends over all the types .
Again , if we put to we obtain a dispersion formula In both cases , if we assume the quantities to be proportional to the density , and divide either of equations ( 3 ) ftnd ( 4 ) , we obtain on the -hand side a quantity dependent only upon the constitution of the individual molecule ; the expression on the other side also be independent of the physical conditions of regation .
In this two relations refractive index density arisen , namely constant ; constant .
( 5 ) These relations have been the subject of numerous .
The Dr. T. H. Havelock .
Optical : its [ Oct. 10 , second one gives generally a better reement with the facts , especially in cases of change of state from vapour to liquid ; however , when the refractive indices at different temperatures and under different pressures are compared , it is found that neither relation expresses all the facts ; these cannot be ascribed merely to change of density .
We then fall back upon simply asserting that the quantities and in ( 3 ) or ( 4 ) must vary with the physical conditions , or , in other words , the natural frequencies of the individual molecules must be affected by pressure , temperature , or external action on the medium ; for a small variation one would obtain from ( 3 ) a dispersion formula .
( 5 ) This in fact has generally been the procedure in previous investigations .
In colitrast to this method , we shall consider here to what extent the facts can be expressed in terms of two variables , the density and the quantity without changes in the inherent molecular frequencies .
It has been remarked that the cause of the reement between the second relation in ( 5 ) and experiment must be looked for partly in the fact that is not exactly equal to , partly also in changes that take place in the interior of the particles when a body is heated or compressed .
; these changes causing a variation in the value of the coefficients in equation ( 2 ) Without denying that both these are probably true contributing causes , it seems of interest to examine separately the former supposition , especially as it has been ignored in favour of the latter as sole cause .
2 .
' and Physiced Conditions .
If we retain the quantity we can deduce a dispersion formula like ( 4 ) but with on the left hand side .
The coefficients are proportional to the ) of vibrating parts of given type in unit volume ; if we make the usual assumption that they are proportional to the density and if we divide through by , we have .
( 6 ) On the hand we have now , supposition , a function of the wave-length and the constitution of the individual particle ; hence for a given wave-length and for all physical conditions of the aggregation of particles , we should have const .
* H. A. Lorentz , ' The Theory of Electrons , ' p. 147 .
1910 .
] Actnal Dependence Condition It may be noted that if we divide hout by we }vrite this relation as const .
This has been used as a generalised form of ( 5 ) ; but it clearly nores the dependence of upon the physical conditions , and has in fact been no more successful than the simplel relation .
The formula in ( 7 ) may be put more conveniently as const .
( 9 ) Now we have no means of independently , so this relation cannot be verified directly .
obtain numerical values of we shall have to fi-x its value in some condition ; ) so , we can deduce a diffel.ence- equation which can be put to the test of experiment .
If the suffixes 1 and 2 denote values in two different states of a medium composed of sinlilar individual particles we have Hence ) .
( 10 ) Thus , for two physical states , the difference in is constant for by supposition is independent of the , hence the left-hand side of ( 10 ) is a function only of physical conditions such as temperature , pressule , density , and so forth .
two relations given previously in ( 5 ) are special cases of ( 10 ) iven by ) and ; so that if either of the formul held , the constant in ( 10 ) would have the value zero or respectively .
Two special forms of ( 10 ) for small changes should ) noted , that is , when it may vritten deonstant .
First , if the ) ressure is constant and the temperature the independent variable , we have const .
, ( 11 ) where coefficient of cubical expansion .
Secondly , if the temperature is constant and the pressure varies , COIlSC .
where coefficient of compressibility .
Dr. T. H. Havelock .
Optical Dispersion : its [ Oct. 10 , From ( 11 ) we have .
( 13 ) Thus if is a positive quantity it is possible for the temperature variation to change from positive to negative ; this would occur for a wave-length at which the index of refraction equals .
We shall examine now the relations ( 10 ) , ( 11 ) , and ( 12 ) .
Where special reference is not made , the data used have been taken from Landolt and Bornstein ( ' Physikalisch-Chemische Tabellen ' ) or from Winkelmann ( ' Handbuch der Physik 3 .
Temperature : Organic Liquids .
A preliminary comparison may be made by using a set of liquid organic compounds which have been examined in a similar manner by Voigt* ; the data are taken from the collection in his paper .
Voigt adopts the hypothesis that thermal or mechanical deformation produces a change in the natural frequencies of the vibrations within the molecule in addition to the effect due merely to the change of density ; he gives expression to this change by tabulating for different wave-lengths the values of a quantity defined by .
( 14 ) Apart from a few exceptions , due probably to inaccurate data , Voigt finds .
that for liquids is positive and increases regularly with decreasing length .
Now if we compare ( 14 ) with ( 10 ) we see that is equal to ; it is possible then for to remain constant for different wave-lengths since , in general , and increase with decreasing wave-length .
Table I shows the results of the calculations for the heat expansion of nine liquids , namely , those used by Voigt as a random selection from available data ; in each case the mean value of , that is of , is compaled with the value of , so that one sees the divergence from the simpler Lorentz formula .
One may say that the general result of these calculations is favourable to the relation given in ( 10 ) ; such divergences as occur are ular in character , and are probably due in these cases to insufficient accuracy of the values of density and refractive index .
Without laying too much stress on the actual numerical values , it appears that may be considered constant within the range of wave-lengths concerned in each case .
* W. Voigt , ' Annalen der Physik ' ( 4 ) , 1901 , vol. 6 , p. 459 .
1910 .
] Actual Dependence Physical Conditions .
More recent experiments are available in the case of }vater and carbon disulphide and we shall examine these no Table I. Aniline . .
1Ie Acet.vlene dibroll .
Mean 005l6 ; Ethyl alcohol . .
Mean C. .
413 408 407 415 Ethylene bromide . .
Mean Benzene .
Iean C ; Iodobenzene . .
Mean Methylacou{te . .
Mean . .
67 60 61 Thiophene . .
Mean A. .
656 689 486 434 Vinyl tribromide .
) . . . .
589 486 434 4 .
Disulphide Flatow* has examined in detail the dispersion of water and carbon disulphide , the absolute refractive index at and for several different temperatures ; he concludes that the relation const .
is pproximately satisfied , and he has also calculated for each temperature a dispersion formula of the type in order to study ths variation of with the temperature .
We shall consider dispersion formulae in a later section ; at pzesent use 's experimental results to the relation ( 10 ) .
* E. Flatow , ' Annalen der Physik ' ( 4 ) , ) , vol. 12 , p. 498 Dr. T. H. Havelock .
Optical Dispersion : its [ Oct. disulphide.\mdash ; We have measurements of at nine wave-lengths for the temperatures , and .
The densities have been calcu- lated from the known expansion of , and with the value We are able in this way to form three sets of differences ; using the .
notation the results are shown in Table II .
Table II.\mdash ; Carbon Disulphide .
mean values of and the changes in density , we have the following : ; ; ; ; ; As regards the constancy of , the results may be regarded as satisfactory , considering the order of accuracy that may be expected .
Considerable deviations occur at the two smaller wave-lengths and .
It must be remembered that here we approach the dominating region of absorption in the ultra-violet , which is in the neighbourhood of ; and it is known that also a minor region of absorption occurs between these two values , near , which is sufficient to cause slight anomalies in the refraotive index.* In the relation ( 10 ) absorption has not been taken directly into account , so deviations may be expected as one approaches a region of absorption .
*W .
Fricke , ' Annalen der Physik , ' 1905 , vol. 16 , 1910 .
] ldence uPhysical ) ditions .
Another of expressing these results is by the alternative form ( 11 ) , the chances small and the pressure constant ; thus approximately Using the mean value of we filled the rate of of at iven by In this case we notice that remaius ative for all leno t Graphing respect to from we obtain practically the curveas that ooiven by Flatow as the esulC of his experiments ; in the visible region varies ouly slowly .
but in the it diminishes very rapidly algebraically with decreasing .
As oftrds variation with the temperature , it should be noted that both the numerical coethcients in would be slightly different for other nperatures .
Water.\mdash ; We have observations at five temperatures , , and ; using the same of with the same notation , we have the results in Table III .
Table III.\mdash ; Water .
From these we ] ve for mean values and difference of densities : ; ; ; ; ; ; Dr. T. H. Havelock .
Optical Dispersion : its It cannot be said that the results are very satisfactory when analysed in this ma1mer .
On the other hand there is nothing decisive against the present theory ; for if has a slight regular decrease , and increase a little with increasing wave-length , while is ular .
Further , if we make the calculations for the smallest wave-length used , we find small values for the four quantities , viz. , ; .
Flatow himself , in calculating dispersion formulae , found greater deviations for wave-lengths between and ; owing to these appearing at all the temperatures , he concludes that they are not due to experimental inaccuracies , but possibly to a ected r of absorption .
Another point in which water is peculiar in the present connection is the occurrence of a minimum density in the region considered ; as the constancy of , we see from the comparison of and that this fact makes a considerable divergence in this region .
Taking the observations and calculations for water as they stand , one concludes that , although the results allow of the present interpretation , the evidence is more doubtful in character than for other substances examined .
5 .
Organic Liquids .
Falk* has ated in great detail the variation of refractive index with temperature for a series of liquid organic compounds , using the sodium line and the three hydrogen lines ; the density of each liquid was also determined at various temperatures .
It was found that within the limits of experimental error , both the refractiye index for given wave-length and the density were linear functions of the temperature .
For five temperatures from to , Falk calculated , for the four wave-lengths , the quantities , and ; each of these quantities shows given wave-length a small , but regular and continuous , change with the temperature , either increase or decrease .
The quantity was also calculated , so as to give the best average results for the relation constant , for each liquid .
For the present purpose , the data for the two extreme temperatures , and , have been used to calculate , t'ne difference in vaJue of for the two temperatures for each wave-length ; the results are shown in Table only substances which show regular variation in the values of are nitrobenzene and monomethylaniline .
It is stated that owing to the oolour of the liquid the live ( 434 ) was too indistinct for measurements G. Falk , ' Journ. Amer .
Chem. Soc 1909 , vol. 31 , p. 86 and p. 806 .
1910 .
] Actual Table Wave-length 656 .
Dimethylaniline -Heptyl alcohol Diisoamyl Benzvl alcohol .
138 acid Isobutyl acetate lacetone Ethyl -butyrate Isoamyl acetate 141 ketone 97 itrobenzene .
122 Monomethylani ] in . .
Benzyl cyanide. .
93 Benzaldehyde 99 to be made in these two cases ; thus the regular increase seelns clearly to be connected with exceptional absorption .
For all the other substa1lcGs it can at least be said that is no increase or decrease of with decreasing wave e , so that it is permissible to the variations as accidental .
In the previous section we noticed that has meitHured the effect of temperatul.e ) the chnnge in , and fouud this increased ] decreasing in general .
The same remark applies to the present collection of with one 01 two doubtful cases ; however , in about six the inctease is by the quantity ) for the line than for the 1 ) line .
of this can be noticed in the column in the table ) , if we calculate the differencc in value of at and for alcohol ior the fonr , we the quence 0 ) , and ) .
This tJht possibly indicate some eperimental e for observations with the line ; if not , it seemt ; a ious effect .
The mean values in column of Table are eqnal to the lnea]lues of from relation ( 10 ) .
We cannot as yet cnlate vnl for , but the numbers measure the rate of increase of temp erature .
The VOL LXXXIV.\mdash ; A. 2 Dr. T. H. Havelock .
Optical : its [ Oct. large value of for acetyl acetone may be noticed ; according to Falk , this substance is exceptional in the above collection , owing to its being a tautomeric mixture of two forms , the relative proportions changin with the temperature .
Anotber point is that the isomeric substances , isobutyl acetate and ethyl -butyrate have practically the same rate of change for the product 6 .
Prjssure Variation .
As regards the change of refractive index with pressure , experimental data are not so numerous for liquids ; at least not a form suitable for the present analysis , for which obseryations are needed at various wave-lengths .
In order to illustrate the alternative formula ( 12 ) , a set of observations on carbon disulphide is taken from a paper by and Zehnder .
* The data are the values of for the , and lines , the , and the index for the sodium line .
From the values for the sodium line we calculate the value of in relation ( 12 ) and obtain thus .
( 16 ) Estimating the values of for the other two lines we can calculate from this equation the values of .
The results are shown in the Table Table Carbon Disulphide .
7 .
Artificial Double Befraction in Liquids .
In previous the present scheme was developed in a simple form in connection with artificial double refraction produced by mechanical strain or by the action of an electric or netic field .
The formal basis of the theory is the supposition that the effects can be represented by an oeolotropic in the quantity .
If the density remains constant and if the * W. C. Rontgen and L. Zehnder , ' Annalen der Physik , ' 1891 , vol. 44 , p. 24 .
'Boy .
Soc. Proc , vol. 80 , p. 28 , 1907 ; also ' Phys. Rev 1909 , vol. 28 , p. 136 .
1910 .
] Actual Dependence upon Conditions .
medium becomes doubly like a uniaxal crystal with and for principal refractive indices , the relation ( 10 ) becomes constant .
( 17 ) If the changes are small , with the refractive index for the isotropic medium at same density , we have -\mdash ; constant .
This relation has ) confirmed by recent researches , at least as a first approximation for the double refraction induced in liquids by an electric or netic field ; reference may be made to the experimental and theoretical investigations of McComb , *Skinner , Cotton and Mouton , and atanson .
the present aim is rather to include various phenomena a single fo1mal scheme than to analyse in detail the physical mechanism of each effect , a few remarks may be made in the present instance .
All the relations such as ( 10 ) , ( 17 ) , and ( 18 ) have been developed by considering the medium as effectively all of optically isotropic particles ; all the changes which occur in the dispersion , whether isotropic or oeolotropic , are expressed in terms of two variables : the density and the quantity .
In the previous work quoted the physical process was conceived of as a rearrangement of the particles in space .
Cotton and Mouton prefer to regard it as an orientalion of anisotropic molecules , and they give a theoretical deduction of relations ( 17 ) and ( 18 ) which is said to rest on this hypothesis Now this assumption is possibly preferable for certain physical reasons , but it is not clear where it enters into the proof referred to above .
In fact the proof is formally the same as ours , in that it ascribes the birefraction to a directional variation in the quantity ; the hypothesis of orientation enters in the assertion that it is the cause of these in and of course if the particles are to orientate under a ficld of force they must be in some way anisotropic .
Perhaps one might reconcile the hypothesis of orientation and the theoretical of relation ) by the particles as in some way physically anisotropic but isotropic , just as one for instance the particles of a crystallinle substance which is optically isotropic ; then one cvht have the possibility * H. E. McComb , ' Phys. Rev 1909 , vo ] .
29 , C. Skinner , ' Phys. Rev 1909 , vol. 29 , A. Cotton and H. Mouton , ' Ann. de ChinL et de Phys 1910 , vol. 19 , p. 153 .
S L. Natanson , ' Bull .
de l'Acad .
des Sciences , Cracovie , ' June , 1910 .
Cotton and Mouton , toc .
cit. ante , p. 217 .
Dr. T. H. Havelock .
Optical Dispersion : its [ Oct. 10 , of orientation without interfering with the simplicity of isotropy in the optical equations .
If we wish to consider the parcicles as optically anisotropic , and the isotropy of the medium to be the result of averaging due to all possible orientations , we should have the following scheme .
The equations of motion of a typical electron take , instead of ( 2 ) , the form , .
( 19 ) Suppose now the extreme case when all the particles are orientated the same way ; then we should have a crystalline medium with three principal refractive indices , and given by the equations .
( 20 ) On the other hand , if the particles are arranged at random in all possible directions , we should have an isotropic medium of index , for which we could write .
( 21 ) In general , if we adnlit optical anisotropy of the particles , this in itself would contribute directly to the observed double refraction , as well as indirectly through the alteration in the -quantities .
One hypothesis would be to ignore the quantities , or make them all equal ; this would ascribe the doubly-refracting properties to the differences between .
But the relation between these quantities does not seem obvious , except that the right-hand side of ( 21 ) is in some way an average of those in ( 20 ) .
Another hypothesis , one by which the simple relations ( 17 ) and ( 18 ) are obtained , is to regard the right-hand sides of equations ( 20 ) and ( 21 ) as all the same function of the wave-length ; that is , we do , in effect , treat the molecules as if they were isotropic .
Cotton and Mouton have also investigated variation of the induced 1910 .
] Actual Dependence upon Physical Conditions .
double refraction with the temperature ; they find that in general the diminution is reater with increase of temperature than could be accounted for by the mere of density .
This result was involved in one case in the calculation given by the present writer for carbon disulphide ; expressing the results of Blackwell in the form it was found that at and at So that , although the in ?
is allowed for , there is in addition a decrease in C. With the present notation is , and if we take the average decrease veen the two temperatures we have It might be possible to connect this with the rate of of with the temperature for the isotropic medium ; but it would doubtless be necessary to examine in more detail the physical nature of the changes which occur .
[ Since the above remarks were written , a note has been published ) angevin , dealing with the theory of electric and magnetic double refi.action from the point of view of molecular orientation ; the note records the results ined .
Apparently the lnolecule is supposed to be anisotropic , but accou is taken of differences between electric , magnetic , and optical asymmetry ; action of the external field is to modify the distribution of the molecular axes in space .
As ards t dispersion , it is stated that the relation ( 1S ) above is obtained under certain conditions .
] 8 .
Vari(tion Sofi ds .
Although one expects more complexity in the case of solid substances , it seems of interest to cxpress the erimental results by the same method as for liquids .
It been well established that the eHect of temperatule on the disf ) ersion of solids is not merely due to the of density , and it is here especially that lecourse has been had to dispersion formulae showing in the natural resonance-frequencies .
Glass .
extensive researches of are available , and we choose three examples which have already been exalnine by Voigt .
The Table shows the results:\mdash ; .
cit. , p. .
Langeviu , ' Comptes Rendus , ' August 16 , 1910 .
C. Pulfrich , ' Annalen der Physik , ' 1892 , vol. 45 , p. 609 .
Dr. T. H. Havelock .
Optical Dispersion : its [ Oct. 10 , .
Heavy Flint Glass , 2.810 O527 .
Light Flint Glass , S40 .
Crown Glass , In the above table , is the quantity calculated by Voigt in the manner referred to previously ( S3 ) ; for small changes the relation is , where is the cubical coefficient of expansion .
Also the quantity of the present theory is given by .
It appears that is negative and increases numerically with decreasing wave-length .
For the heavy glass , the increase in the factor is not sufficient to ) teract the increase in , snd the quantity shows a regular increase ; in the other two glasses the effect is the other way and shows a ular decrease in the same direction .
One must remember that glass is not a simple substance but is a complex mixture , a fact which is illustrated by its optical behaviour under mechanical stress .
Further , Pulfrich considered that the observed changes in refractive index were connected with a variation of absorption in the ultra-violet region ; .
Konigsberger*has also observed in solid bodies a displacement of this absorption region towards the longer wave-lengths with temperature .
The direct influence of absorption has not been taken into * J. Konigsberger , ' Annalen der Physik , ' 1901 , vol. 4 , p. 796 .
1910 .
] Actual Dependence upon account in the present theory , so that one cannot say how the results would be modified thereby .
Meantime , it appears that the variation of the -quantity must play a considerable part in a complete theory of the phenomena .
Similar eneral remarks apply to the following calculations , which are tabulated here for comparison .
Amorphous Qnartz .
Observations on this substance have been made by Martens ; the from to .
With the usual notation the following Table shows the results for some of the wavelengths .
If the substance were the same as fused quartz or silica it would have a small coefficient of expansion ; whether this was the case or not , it appears that if we take to have the mean value for quartz , the values of show a remarkable constancy in this range .
If we calculate the values of as before , we find it varies ularly from at 508 to at .
We notice that is positive in this range .
If we take the mean value of to be , and if we assume to remain constant , we can calculate when would become ative ; it would be zero when , that is , when approximately .
As ?
is at and at 656 , the wavelength in question would be large ; judging from the refractive index of ordinary quartz , it be in the neighbourhood of 5 , approaching a region of absorption in the infra-red .
Table \mdash ; Amorphous Quartz .
Fluor Spar Salt .
observed the ariation of for eral substances , with rangiug from about to 589 * F. F. Martens , ' Berichte der Phys. Ges 1904 , p. 308 .
F. J. heli , ' Annalen der Physik , ' 1902 , vol. , p. 772 .
Dr. T. H. Havelock .
Optical Dispersion : its [ Oct. 10 , All the curves obtained by graphing erainst show a towards smaller values of .
The most striking result is that for rock salt , where varies from at 202 to at 643 .
If we calculate the quantity in this case , we find that it varies only slowly in the visible region , having a value of about , but increases rapidly in the ultraviolet .
If we use the value ' to calculate where sign , and estimate the wave-length from a dispersion formula , we obtain the position as ; the experimental curve cuts the axis at about 220 .
Similar remarks apply to the calculations for fluor spar , where has a value of about in the visible region , but rises rapidly in the ultra-violet to at In the case of solids such as those examined above , we find that if we use the value of in the visible region to calculate a curve for , the curve so obtained falls below the experimental curve when the latter begins to vary rapidly in the ultra-violet .
It is unnecessary to add further calculations of the same type , or one might examine data available from experiments on elastic deformation of solids , or from the double refraction of natural crystals and its variation with temperature and pressure\mdash ; probably with similar results .
The general conclusion is that the variation of the quantity needs to be considered in each case , but the changes in solids are too complex to be brought under a simple scheme with only two or three variables .
9 .
Changj Vapour to Liquid .
In this region the relation has had its greatest success compared with the other simple formulae .
Calculations for many substances have been given by from his data we calculate the quantity the present theory for six typical substances , including those for which the Lorentz formula shows most disagreement .
Two wave-lengths are given , the Li and Na lines ; the data for the vapours are for a temperature of about , while the liquids vary etween and .
Table VIII shows the values of , the differe1lces between values of for vapour and liquid , for the two wave-lengths ; the agreement between the second and third columns is as good as could be expected .
The values of are given for comparison .
* Briihl , ' Zeitschrift fur Phys. Chem 1891 , vol. 7 , 1910 .
] Dependence Physical Conditions .
Table VlII.\mdash ; Values of 10 .
A bsolnt of We have so far been concerned only with a difference equation , the quantity being equal to thus we ) tain a ition between the values of in two different states .
If we wish to assign numerical values to , we must fix its value in some state .
At first it might be advisable to follow the ordinary theory , and make equal to in the gaseous state .
If we try this for carbon disulphide we can calculate from Table VIII the value of for liquid at : knowing the densities of the vapour and liquid , and putting , we obtain thus .
This is less than the value of in the state ; but represents the additional force on a particle due to the molecules , so it seems preferable that should be greater for a substance in the liquid state th when gaseous .
This result can be obtained by making very smadl for gases ; this is allowable , as , to the small values , there is no evidence to decide between the of constancy of the various forms or To make the matter definite , we decide provisionally to zero for gases , and for simplicity we fix the standard state as the gaseous condition supposed to be at and a pressure of 760 mm. With this assumption , the values of in the last column of Table VIII have been calculated .
One cannot deduce much from the values , as the data for gases are possibly ]lot very accurate and the liquids are at various temperatures ; otherwise it might be of interest to trace , for instance , a connection between the values of and the chemical constitution of the substance .
It may be lloticed that the " " specific refraction\ldquo ; is ; thus , for example , the value of for propyl iodide would give a corresponding divergence in the specific refraction ordinarily used .
On the present scheme , anomalies in Dr. T. H. Havelock .
Optical Dispersion : its [ Oct. 10 , molecular refraction might be ascribed , in part , at least , to the indirect effect of intra-molecular action specified by the -quantity .
Air.\mdash ; Magri* has made observations on the refractive index of air at high densities ; he found that gave the best approach to constancy .
In Table IX some of his results have been used to calculate ; in the column under are the values under the assumption that is zero in the standard state , while under are the values if is in that state .
The column shows the value which should have in each case in order that should have always its value for unit density ; the last column gives the actual value of this quantity referred to unit density at C. The temperature for the first row is , while for the other it varies and .
Assuming the experimental data to be sufficiently acucrate , the variations both in and are curious , rising to a minimum with increasing density , and then falling to approximately .
It may be noticed how an apparently difference between and makes little in the in the last column .
The resuJts given here are referred to later in S 12 .
Table IX.\mdash ; Air .
Carbon Disutphidc.\mdash ; By combininrro the observations of Flatow , referred to previously , with values of for the vapour , we can see how the value of varies with temperature and density .
For the vapour , we know for one wave-length of the series used for the liquid , viz. , the lins 589 .
If we use the data , and for the vapour , together with and for the liquid at , we find * L. Magri , ' Physikalische Zeitschrift , ' 1905 , vol. 6 , p. 629 .
1910 .
] Actual Dependence upon Physical Conditions .
this value for all wave-lengths at , we can calculate the values for various wave-lengths at and ; the results for four cases are contained in Table X. There is a small regular increase in with rise in temperature , the density diminishes .
Table X.\mdash ; Values of for CS ; the vapour we use , and for sodium light .
Carrying out similar culations , we find the following series of mean values of for the liquid:\mdash ; It is of interest to compare with these some values for ice , which is a uniaxal crystal ; Pulfrich has measured the two indices ?
and for various wave-lengths .
Let and be the corresponding principal values of for ice at ; with the density ) , the table shows the values of for certain .
absolute values of the indices:\mdash ; Ice .
Now for water at , and wave-length 589 , we ; thus , if is the value for water at , we have the equations From these , with , we find and ) actual numbers would probably be modified with with which to compare ; however , it appears in the to the solid state is increased .
512 Dr. T. H. Havelock .
Optical Dispersion : its [ Oct. 10 , 11 .
Dispersion Formuloe : Absorption Maxima .
It would seem possible to obtain values of by constructing dispersion formulae of a suitable type ; before examining a few cases it is necessary to consider certain points in regard to the wave-lengths for which absorption is a maximum .
The argument may be given first for a simple formula , such as .
( 22 ) Consider the values of in the hbourhood of a " " resonance\ldquo ; wavelength , say ; suppose that this is sufficiently isolated so that we may take the other terms in ( 22 ) as constant , that is , we write .
( 23 ) From this equation is ative , and consequently imaginary , for the range of wave-length given by .
If in ( 23 ) we regard as complex and equal to we should have zero , e.xcept within the above range , where it increases from zero at the lower limit to infinity at the upper limit .
This effect is , of course , not a true absorption effect , but would mean that the medium refused to admit certain radiation within a definite of wave-length .
If we now insert a term in the equations of the medium to represent absorption , we have instead of ( 23 ) , an equation .
( 24 ) Following the method used by Lorentz*for a similar equation with equal to unity , we write the last term of ( 24 ) as ; then we have Solving these , we obtain .
( 26 ) If we assume , further , the usual simplification , viz. , that is , we obtain finally the approximate result .
( 27 ) * H. A. Lorentz , ' The Theory of Electrons , ' p. 310 .
1910 .
] Dependence upon Thus the retention of does not affect the form of the result , and the deduction follows that the maximum of is in the hbourhood of that is , of the wave-length Returning to the dispersion formula ( 22 ) , the previous uluent is of the type which has been ] on when the quantities have been identified with absorption maxima experimental]y ; as far as the simple formula is concerned , these quantities are the upper limits of the ranges which is imaginal.y and are -lengths for which is infinite .
If we turn now to the type of formula used in the previous worl we have .
( 28 ) In the neighbourhood of we now .
( 2 !
) If we solve this for we find ( 30 ) with ; .
The lower limit for which is inary is equal to while the upper limit is Thus both the limits of the of imaginary values il are ) the introduction of the quantity , a fact which was noticed by in optical dispersion .
It is clear if we introduce term true absorption , the argument would follow the same lines from equation ( 30 ) as ; consequently , we led to lentif the absorption maxinlun with upper limit of the above in ( 31 ) .
Witb a dispersion formula of type , the quantities sent true\ldquo ; resonance\ldquo ; wave-lengths to the individual atonn or molecule ; it not permissible to identify them with maxima of , or refleclion , which have been determined ) erinlenta ] the nlediuln as a whole .
To J. rmor , ' Phil. Trans , 18 vol. 190 , p. 240 .
Dr. T. H. Havelock .
Optical Dispersion : its [ Oct. 10 , determine the latter , in any case , we must write the dispersion formula as in ( 29 ) , and the wave-length required is , instead of Numerical illustrations may be taken from two cases , where a dispersion formula of the type ( 28 ) has been used without taking account of this change .
Cccrbon Disulphide.\mdash ; In the paper quoted , Flatow has calculated dispersion formula of the type for at various temperatures ; he finds to be about , varying slightly with temperature .
There.is a region of absorption ranging from to with a maximum at approximately ; there is also a minor of absorption near .
By a method of continued reflection , Flatow found a maximum at and concludes that the value of for agrees well with this for a natural frequency .
He also states that a dispersion formula of the Lorentz-Planck type gives a value of for and consequently gives a worse agreement than the older form .
In the latter calculation Flatow ignores the effect considered above , and also uses a specially simple type of formula , .
But for carbon ulphide we must use at least a three-constant formula .
Suppose , for comparison , we assume that is , and make the calculations with determining the constants from the data With sufficient approximation for the purpose we find the values Thus we find is .
But this is not to be compared with the experimental data ; according to the previous argument , the quantity to be so used is in the present notation .
Working this out for the above values we find the value , giving practically the same reement as the older formula .
Rock Salt luorite .
Larmor 's development of the present type of dispersion formula , to which reference was made above , Maclaurin*has deduced and illustrated the form .
( 32 ) R. C. Maclaurin , ' Roy .
Soc. Proc , 1908 , vol. 81 , p. 367 .
1910 .
] Actual Dependence upon Conditions .
For a given physical state of the medium this form is equivalent to above , with equal to ; however , the previous form is preferable when we wish to consider changes of dispersion for physical conditions , and it also enables us to separate out a quantity independent of these conditions .
Maclaurin gives numerical calculations for two substances , rock salt and fluorite ; identifying with the dielectric constant , he uses ( 32 ) in the form A very close eement is shown between calculated and observed values over a range of ; in addition Maclaurin claims that and agree with the mean values of the best experimental data .
Howevel the quantities and are identified directly with the observed of absorption or reflexion , : we calculate now what difference is by the argument of the present section , using Maclaurin 's values of the constants for the dispersion formula ( 33 ) .
For rock salt we have\mdash ; Putting for in the last term of , we have , in the vicinity of , with Hence the required wave-length is given by , that .
is , .
The lower limit of the range for which is imaginary is or .
We have seen it is the upper limit which , consistently with the similar ument for the simpler dispersion formula , should be identified with the experimental data .
For the neighbourhood of find and result is that is imaginary between and ; bence if we use the erical values given above it is that should be compared with experiment instead of We may compare with this Paschen 's formula for rock salt of Sellmeier type ; this gives a natural wave-length of 60 in the infra-red and in its neighbourhood we have 5 .
We find then that is imaginary between approximately and 60 ; the upper limit * F. Paschen , ' Annalen der Physik , ' , vol. 26 , p. 130 .
Dr. T. H. Havelock .
Optical Dispersion : its [ Oct. 10 , being compared with the absorption maximum , although Paschen regards as an open question how far the observed " " Reststralhen\ldquo ; maxima can be identified with the wave-lengths of the natural vibrations which dominate the dispersion in the infra-red .
A smaller value of the constant would probably give a better agreement in this respect .
In fact Maclaurin states that he tried at first equal to 2 ; he found then a fair agreement between calculated and observed values of but with the values of and as much below the observed values as they vere above them with an ordinary Sellmeier dispersion formula .
As regards dispersion formulae in general , it should be remembered that we replace the actual substance by an ideal simplified medium with only two or three natural frequencies ; it may happen that.the actual substance has two or three very predominant regions of resonance , in which case there should be good agreement between calculated and observed positions for these .
On the other hand it may be that there are two or three erions of equal value not far apart , and the calculated value would be a mean position .
For instance , recent observations on rock salt show maxima at and It appears that the quality of the agreement between observed and calculated values of refractive index even over a considerable range is not sufficient in itself to decide between different hypotheses , and for instance to give reliable values of or In respect to the absorption regions in the infra-red , the dispersion often gives no more than a term , which we could make up from any number of places of selective absorption in the absence of further information ; numerical values are obtained only when we assume the definite simplification of a small number of natural frequencies and use the value of the dielectric constant .
For fluorite Maclaurin gives the ] owing values : With the same notation as before , we find in the neighbourhood of that ; hence .
Also near we have , and .
The relative displacements are larger here owing to the larger values of than in the former case .
If we used the value of obtained from these formulae we could calculate * Rubens and Hollnagel , ' Sitzungsber Berlin , 1910 .
1910 .
] Actual Dependence Conditions .
equal to ; this wo11ld give for rock salt , and for fluorite .
But the preceding argument indicates that the calculations need considerable revision .
12 .
cement of We shall consider now the displacement of absorption maxima under varying physical conditions , illustrating this by calculations for three cases in which suitable data are available .
Air.\mdash ; The first example shows the effect of of pressure and density in a , and we use Magri 's nents on air already examined in S10 .
For a dispersion formula under ordillary standard conditions we have one obtained by C. .
Cuthbertson*in the form ' ( 34 ) lere p is frequency .
the density of air as we put this into a form suitable for present calcu1-ations and obtain ( 35 ) with .
To make the calculations definite assume , as in S10 , that is zero in the standard condition , so that the quantity on the right of is invariant as regards physical conditions and depends only on the atomic or molecular constitution .
In any other physical state the quantity on the left of ( 35 ) is altered so that we have The argument of the previous section shows that the observed 1naximum of absorption in any condition should be near the wave-length iyen by We obtain the following series of values of to the values of the relative density ) C. and M. Cuthbertson , ' .
Soc. Proc 1909 .
, vol. 83 , p. 170 .
VOL. LXXXIV.\mdash ; A. 2 Dr. T. H. Havelock .
Optical Dispersion : its [ Oct. 10 , Table XI.\mdash ; Air .
variation of and of with the Ons may notice the slight initial increase of , then a sharp rise between the values of 15 and 40 for , followed by an approx imately steady rise .
We see also that decreases steadily with increasing density , falling sharply from an infinite value in the initial } .
No doubt the peculiarities af the lower values of are due largely to the assumption that is zero initially .
Still it does not seem possible to obliterate them entirely by assigning some small positive value initially ; it is conceivable thaG there may be a stage at which the inter-molecular action specified by becomes suddenly appreciable .
One may notice a possible analogy between the variation of in the present case and the pressure displacement of lines in an emission spectrum .
1910 .
] Actual upon Physical Conditions .
Carbon Disulphide.\mdash ; We can make calculations for with the observations of Flatow and the values of refractive index obtained by Rubens for longer wave-lengths .
If we know for a given wave-length for a physical state of the medium defined by and , the quantity is independent of physical conditions ; thus , by observations at various temperatures , we can form an average set of vahles showing this molecular dispersion .
With the convention adopted for assigning numerical values to , this would in fact , the values of for the substance as a gas at and 760 mm. For reasons which have been indicated , it has not been thought advisable to try for a complete dispersion formula for this substance .
However , four have been chosen , covering the range available , so as to form a four-constant formula to illustrate the present scheme .
the values of given in S10 and the observations referred to above , we find 0.274 .
0.508 .
1.037 .
1.998 .
Specific refraction Before proceeding , we may notice the corres o values of for the vapour ; using , we obtain for these four tYths in order the values .
The values of were , of course , calculated with the help of the observed value of for the wavelength .
The values of ?
so calculated would be values for the vapour , on the assumption that the change from liquid to vapour involves only physical changes which can be expressed optically in terms of and ; for instance , if some of the regions of absorption or resouance in the liquid were due entirely to molecular ates which were dissociated in the aseous condition , then from this point of view the vapour would be optically a different substance .
However , in most cases there does not appear to be an actual disappearance or creation of the principal dominating resonances , but only displacement and tion with changing conditions .
use the data above to determine the constants of a formuIa .
With sufficient approximation for our purpose , we find ' This gives a value of for .
If we take the dielectric constant to ] ) given by ) for liquid CS we ) ffiin a mean value of for the left-hand side for vice ; this , we may replaco the term by a single equivalent bsorption region in the -red .
We obtain in this a term ) Dr. T. H. Havelock .
Optical Dispersion : its [ Oct. 10 , now , and .
Experiment has shown that carbon disulphide has specially strong absorption in the region between and In any given state the maxima of absorption occur at wavelengths \fnof ; m and , given , as in the previous sections , by expressions like , when the dispersion formula has been put into a form suitable for the vicinity of the wave-length in question .
Table XII shows the results for the as and the liquid at three temperatures .
Table XII.\mdash ; Carbon Disulphide .
9.161 It is seen that is displaced towards the shorter wave-lengths with increasing temperature .
This appears to be the general rule for liquids : in contrast apparently to the behaviour of solids .
Water.\mdash ; We examine now in the same manner Flatow 's observations on water at five different temperatures ; the values of were calculated in S10 , from a certain value of for water vapour .
Using these values we calculate the specific refraction for four wave-lengths ; the following table shows the values at the various temperatures and the mean for each wave-length .
Table XIII.\mdash ; Values of for Water .
Using the mean values and a density of for water vapour , with zero , we find the following values of for these four wave-lengths in order : ; the last value is of course the one which was used in order to calculate the values of for the liquid .
*Cf .
, ' Handbuch , ' vol. 3 , p. 350 .
1910 .
] Actual Physical Conditions .
Using the meau values above , we calculate the constants of a dispersion formula of the type ( 37 ) and find ; ; ; The value of is ; we calculate now the values of near which occur the maxima of selective absorption in the ultra-violet at the various temperatures .
The following table shows the data and the results for the liquid , usin , 0.33489 0.33361 0.3,3364 0.99233 0.98331 0.97191 0.1294 0 .
0 .
0 .
In the formula for it is the product which occurs as the variable ; this decreases continuously in this case and we find a corresponding regular but very slight decrease in with temperature .
result is in contradiction with that deduced by Flatow ; he calculated the constants of a Sellmeier dispersion formula for each temperature and a fairly increase in with rising temperature , the values anging from to approximately .
There is , of course , no direct experimental evidence available in this case .
In general the experimental data for the variation of absorption maxima with temperature are rather in charactel ; the results appear to depend on circumstances such as the homogeneity of the medium and the charactel , selective or general , of the absorption .
It has been stated that in several cases , at least in the ultra-violet , nlaxima are displaced towards the shorter wave-lengths with increase of temperature for liquids , while for solids they move to the longer If we wish to study the in the we must Gplnce the term of the dispersion by the effect of one or more equivalent terms ; we try first a term .
If we for the calculation the values of the dielectric constant iven by , namely at and 799 at , and if we use the colTcsponding values of iven above , we obtain a mean value of for the specific refraction for infinite -length ; hence and are given by .
These give the values ) and .
If , with the same notation as before , we form in order to find the approximate position of the absorption maximum , we find that ranges from about at to 55 at .
This lalge variation is due to the large values of and , and these come directly 522 Dispersion : its Dependence upon Physical Conditions .
from the normal value of the dielectric constant ; the only inference is that if we insist on bringing the dielectric constant in this case under the same scheme as the ordinary dispersion , one resonance in the infra-red is not sufficient .
It is well known that Sellmeier dispersion formulae with the same assumption indicate as the approximate position , and the subject has been experimentally recently .
Using ice , Trowbridge and Spence* find no metallic leflection between and ; with water , Rubens and Hollnagel obtain a similar result and also deduce that at the refractive index is of the same order of magnitude as in the visible spectrum .
We may use this information in attempting to make up the term from two resonance terms , .
To make the matter definite , we assume that is at 80 ; then , with this and the above data for the dispersion , we have the equations We have only three equations , with four unknown quantities ; but to illustrate the calculations we assume a value for , say 10 .
Then , solving the equations approximately , we find Calculating now , as before , the positions of selective absorption corresponding to at the various temperatures , we find they range from at to at .
The principal absorption corresponding to is far removed in the infra-red ; the results appear to 'be fairly consistent with the absorption of water , and they confirm at least the necessity for more than one selective absorption in the infra-red .
13 .
Summary .
The chief results of the previous investigation may be summarised briefly in the following manner:\mdash ; 1 .
The formal scheme of the theory is the representation of the effect of the physical state of the medium upon its dispersion by means of two variables : the density , and a quantity expreseing an effect of surrounding molecules .
2 .
A relation independent of is deduced , and is used as a test of the theory , namely , the difference of for any two physical states is independent of the wave-length .
3 .
As the result of an analysis of available experimental data , it appears that the scheme is sufficiently adequate for gases , for liquids , and for changes of state from gas to liquid .
For solid substances , the two variables and A. Trowbridge and B. J. Spence , ' Phys. Rev July , 1910 .
Rubens and Hollnagel , .
cit. The of Halley 's Comet .
are not sufficient to express all the facts , but it seems that the term must at least be taken into account in a complete theory .
4 .
It is shown that it is possible to include , either wholly or in part , the dispersion of artificial or natural double refraction in simple cases .
5 .
By assigning a standard state for the medium , gas at and 760 mm. , for which is zero , numerical values of are obtained for various substances in different states .
Anomalies in molecular refractivity are ascl'ibed , in part , to the variations in 6 .
Using a type of dispersion formula involving and , it is shown that the positions of the absorption maxima depend upon these quantities as well as upon the frequencies of the natural vibrations of the molecule .
The consequent displacement of the maxima for varying conditions of pressure temperature , and density are illustrated by analysing certain expel.imental data .
of Halley 's Comet .
By CHARLES P. , A.R.C Sc. , F.B.P.S. , F.R.A.S. ( Communicated by Sir Norman Lockyer , K.C.B. , F.R.S. Received November Read November 24 , 1910 .
) [ PLATE 7 .
] Owing to the interference it was inlpossible to for observations of the comet sp ectrum lfay from the Solar Physics Observatory at South Kensington .
Sir Norman therefore obtained permission to occupy the site on Fosterdown , Caterhall , which already been selected as the future position of the Observabory .
No observations of value were possible during the first three weeks of May on account of exceptionally bad weather ; but on May 23 and 26 photographs and visual observations of the spectrum were seeured .
The spectrum was seen as a small nebulous clou to in diameter ; with a binocular the central htening due to the intense tellar nucleus became clearly visible .
a whole the conlet was ) as conspicuous as a second nitude star , so that it was inter than had been expected .
The curvature of the was not so that seen in Comet 1910 iving the appearance of an to the tail extension seen with moderate power .
Photographs of comet were obtained with a lncycr camera of

