]\gt ; Spectroscopic and By P. .
BEVAN , Sc. D. , Royal Holloway College .
Communicated by Sir J. J. Thomson , F.R.S. Received January 4 , \mdash ; Read February 8 , 1912 .
[ PLATE 9 .
] The present communication deals with a further study of the absorption spectra of the vapours of lithium and caesium .
These spectra were investi , ated , ) results communicated at the beginning of last to the Royal Society .
* The present account deals with an extension of the principal series lines for the two metals , and a more detailed examination of the expression of the wave-lengths by a formula , as well as an investigation of the broadening effect in a } ) of lines with increase of density of the vapoul of sium .
We shall deal first of all with the lithium spectrum .
As has remarked before , lithium vaponr presents some difficulties , because of the .
temperature required ) the action of the vapour in this condition on the material of the containing vessel , but this difficulty can be surmounted by the use of a double steel tube and a quantity of lithium .
With the tube at a bright red heat , a dense vapouu : is produced to show in transmitted light a large number of the principal series lines ; the number that can be observed depends , I convinced , only on the resolving power of the spectrograph used to investigate them .
With the instrumen6 at my disposal , I have been able to measure 41 of these lines , and I think in this case the resolution has been pushed as far as is practicable , the lasb few lines being separated only about ngstrom unit .
Fig. 1 , Plate 9 , shows the absorption lines up to No. 25 .
The finer lines after this do not appear in the reproduction .
The following table gives the results of measuremenbs .
In the column is the number of the line in the series .
The second column gives wave-length in Iriternational units founded on the absolute measurements the red cadmium line .
The next column contains an estimate of the errors in the individual .
The fourth column gives " " oscillation frequencies\ldquo ; in number wave-lengths per timetre ; and the last column the rences , observed\mdash ; calculated the calculated values being those obtained from a Hicks formula discussed later .
' Roy .
Soc. Proc , vol. 85 , p. 54 .
Spectroscopic : Lithium Ctesium .
Table I.\mdash ; Lithium .
The measurements for the lines 1\mdash ; 7 are by Kayser and Run , and 9 easured by Living and Dewar , but were remeasured ; 10\mdash ; 27 were given in my former paper aheady referred to ; new .
The errot .
S indicated as possible are probI ) in nearly all cases mnch outside the limits .
They have been increased for the lines 10\mdash ; 27 , owing to the fact -that the uncertainties in the measurements of lines and in ] ) tions of wave-lengths were ) Jated .
To begin with , there is some considerable uncertainty in the values of the wave-lengths of the cadmium lines .
For example , the values of the possil ) errors given by Kayser and unge for the lines , and are , and , and when the cadmium spark is used as the sonrce of light these are 1nuch -exposed in , so their positions calmot bo determined with any boreat raCy .
Then is some uncertainty in the Rowland on which the lleasurements of the lines are based .
At } ) resent , Kayser has not extended the third order stamdards into this region , so that we have only the secoud order stan dards of Fabry and Buisson to go , but there llay easily be ular errors of some hundredths of an strom uniu in the old Rowland standards .
As far as my actual 1lleasuremens are concerned there should not be , in the case of lines not very near to cadmium lines in wave-le errors of more than a hundredths of an ngstroln uuit for lives near the Dr. P. V. Bevan .
[ Jan. 4 , end of the series , even when possible unsymmetrical broadening and other effects are taken into account .
Taking all sources of error into consideration , I have put the possible errors high ; the diHerences are due to the uncertainty when an absorption line is near an emission line , so that the true centre of each is difficult to determine .
After the 25th line the measurements are expressed to hundredths of an Angstrom unit , because , although the possible error is much above this order , yet , as a rule , the errors will affect lines almost equally , in so far as they are produced by uncertainty in the measurements of the position of the cadmium reference lines .
The difference of between successive lines is thus much more accurately determined than the actual values .
In reducing the measurements to the International scale , the values of iron lines as determined by abry and Buisson were used ; a table of the necessary corrections has been published by Birge*which gives practically the corrections used for Table I. The reduction to vacuo for the oscillation frequency was obtained from the ] of corrections published in Watts ' Index and the temperature corrections in the same volume .
The formula on which the last column of Table I is based is that due to Hicks ' where is the oscillation frequency , universal Rydberg constant , and takes the values 1 , 2 , 3 , etc. , for the different lines of the series , and being constants for the particular substance .
Using the oscillation frequencies based on the International scale , we have to use a different value of N. The value has been calculated by Birge in the paper referred to from the wavelengths of five hydrogen lines .
This value is instead of the value 109675 .
Using the first three lines of the series to calculate we obtain and these values are used for the last column of Table I. The possible errors work out to be practically the same as those given by Hicks for for , for If we take a higher member of the series to determine with approximate values of and , the method I have used before and adopted by Birge , we obtain a higher value of ; for example , from the lines 25 , 30 , 35 , 40 , we obtain the numbers , but in each case a * B. T. Birge , ' Astrophys .
Jour vol. 32 , p. 113 .
'Phil .
Trans , vol. 210 , p. 96 .
'Roy .
Soc. Proc , vol. 83 , p. 425 .
1912 .
] Spectroscopic Observations : Lithium nd .
323 possible error .
This method does not give us such close limits as the method of taking the first three lines of the series if the limits of error are reliable .
A slightly higher value of A would make the numbers in the colmnn look better , but the agreement would not be so good in the earlier members of the series , and in view of the larger possible errors in my own measurements , which quite possibly may all be too small , it is preferable to rely on the Kayser and Rung measnrements for the earlier lines .
The only line for which there is not agreement between the calculated and observed values is 4 , for which the limit of error is givetl by Kayser and Rung as and the difference appears as .
This , probably , is to be explained by an error in the determined wave-length owing to errors in the old standards .
It is quite possible that an error of may be introduced owing to this .
The agreement in the other lines is fairly complete .
In former measurements of the absorption lines of caesinm the absorption at1nosphere was obtained by , caesium chloride with metallic sodium or potassium .
Sufficient was set free to give a vapour which showed 24 of the absorption lines belonging to the principal series , but owing to the comparatively easy volatilisation of the sodium or potassium the spectrum was complicated by the presence of the sodium or potassium absorption spectrum .
A method of obtaining caesium by heating the anhydrous chloride in with metallic calcium was given by Hackspill .
* This method I found not very satisfactory for my purpose , as it was difficult to control the rate of formation of caesium .
A more suitable method was found in heating the anhydrous chloride with metallic lithium .
The lithium volatilises with difficulty , and a steady atmosphere of caesium vapotlr of sufficient dellsity can easily be maintained .
With this method the sel'ies lines were photographed up to the thirtyfirst of the principal series , and again the limit is only imposed by the resolving power of the instrument .
The principal series for all the alkali metals , with the exception of lithium , consist of pairs of lines , the separation of corresponding pairs being as we go from sodium to caesium .
The lines are not equally intense in all cases , the line of greater wave-length in a pair being the less intense .
There are , in fact , bwo series , , in the notation of Rydberg .
It is only the cases of rubidium and caesium that we could expect to be able to distinguish the pairs for more than a few members of the series , as , on the assumption that the ' Comptes Rendus , ' 1905 , vol. 141 , p. 106 .
Dr. P. V. Bevan .
two series have the same limit , the pairs become so close together they cannot be separated .
But in the case of caesium especially separation is considerable , and so the pairs of lines can be more distinguished .
The later members in all these series belong to -this becomes clear from the cases of idium and caesium , where intensity of the P2 lines falls oif with higher order in the series until line cannot be observed , although its position assuming an formula should be quite easily distinguishable from the corresponding .
The lines of P2 cannot be out by increase of density of the vapour , as this produces .
of the lines , which spread over the position of the P2 lines in the photograph .
This disappearance of P2 lines has led to some errors in applying formulae to the series .
has calculated constants for the Ritz formula for the alkali metals and has found that the formula is inapplicable in the case of caesium , but the explanation is that the first four membels he ] ) used lines\mdash ; those of the pairs with greater wave-length\mdash ; while the rest of the wave-lengths are of lines .
The wave-lengths used by him for the higher members have not much accuracy , but the formula would give very much better results if the four lines of less wave-length were used .
The same remark applies to of Birge 's calculations\mdash ; he has used the lines where they are observed , while after the lirst few members it is the lines which are measured .
This makes very little difference in the case of sodium , where the pairs are very close together after the first three\mdash ; the only ones which have been observed .
In the case of potassium also five pairs have been observed and for later pairs the differences would be hardly greater than the errors in the experimental numbers .
For rubidium the diflerences are greater , and again this accounts for the bad fit for the numbers calculated by from the Ritz formula .
In Table II are given the wave-lengths on the International scale .
The first column gives number in the series , the second the wave-lengths , the third the possible error , the fourth the oscillation frequency in vacuo , and the the values of observed-calculated wave-lengths , to be dealt with later .
In Table II the wave-lengths for the first pair are measured by Lehmann .
There are no measurements which give a check on his values , but judging from his measurements of other lines in this infra-red region we can accept them as being reliable .
The measurements for the second pair are by Kayser and Rung , and the values agree with those of Ramage and Exner and Haschek .
The rest of the lines are my own For my value agrees with Kayser and Runge 's ( reduced to InIernational units ) \mdash ; but their valne for the other one of the .
is 387659 , with a possible error of .
Ramage gives , which agrees better with my own value , and would give better yreentent with the calculated values .
I have kept my own value , as it seems certain that the Kayser and value is too .
For yser and 's values are ; corrected to international units gives ; ayser and 's limits of error are given as and respectively .
For my own measurements the line of greater wave-length was very no , the breadth being about ; the other line was broader , about unit .
It seems probable that Kayser and 's value for P2 ( 4 ) is too low .
Ramage for gives ) , and for lines , .3314 , .
The line has no error attached to it , as it is practically coincident with a rubidium line which ppeared in the photographq , the two lines appearing together sinlply as one .
It is doubtful therefore what this measurement represents .
For the determination of the constants , of the Hicks mula ' the method adopted was to take a fairly high order of line for , 15 , which with nate values of and determines A. Then from early lines and can be redetermined .
Successive approxilnations give Dr. P. .
Bevan .
very quickly the constants required .
P2 the sixth line was used the first two .
The work of approximation is , of course , slower than but not much arithmetic is involved .
The possible errors in the are determined as by Hicks , and it easily appears that the best are obtained by usin , the first two lines .
We obtain , for the 1 , 2 , and 15 , 3 , and 15 we obtain .
For we obtain , using 1 , 2 , and 6 , The two values of A for , come out the same , which is in with the assumption that the two series have the same limit .
I think , the best evidence for this relation between the pairs that we at present .
The column O-C in Table Il gives the differences observed and calculated values of the wave-lengths derived from these of the constants .
The agreen ) is everywhere within the limits of error except for .
A permissible change in A would this inside the limits , but in any case it is not a very reliable it is the last of the series visible , and so a very faint linle , difficult to out and .
Perhaps a greater error should have been allowed in case , but , as is hardly necessary to say , the possible errors were allotted the calculated values were obtained .
The values of the constants and enable us to make more complete investigation with regard to the relations they exhibit the series the alkali metals .
For this purpose I have recalculated the constants of series for sodium , potassium , and rubidium , the wavethe International scale and to .
The reduction is not very as for the infra-red we have to extrapolate , but it seemed worth make the aGtempt to get as near the right values as possible .
The table gives the values of and with limits of error for the four metals , Rb , Cs , for the Na Bb Cs , 1912 .
] Spectroscopic Observations : Lithium Coesium .
327 These numbers were calculated from the lines 1 , 2 , and 35 for sodium , with a slight increase in possible errors owing to uncertainty in the old Rowland standards\mdash ; an increase of this kind has been made in all cases . .
potassium the lines 1 , 2 , and 22 were used , and for ) idium 1 , 2 , and 25 .
The hmits are on the whole narrower than those given by Hicks .
In the case of sodium the limits are wider owing to the fact that the publishPd possible errors have been increased for the reason mentioned above .
It follows at once from these numbers that the relations*for , and are not we have in fact the following:\mdash ; Na 2 Bb 5 Na 2 Rb .
5 Cs We see in the values of a regular decrease with increasing atomic weight , but clearly the fraction is not constant .
In the values of , however , there is no apparent relation to the atomic weights .
It also appears that it is impossible to make the fraction when is different from 1 a constant multiplied by the rn mbers 2 , 4 , 5 , 6 .
An expression of the form is giveu by Hicks which applies to Na , , and Bb , and to Cs if the term is dropped .
This expression also is not exact within the limits ibed .
It rees for the independent of for Na , , and ) but the part reptesenting is outside the limits for Bb , and the numbers when the term in is dropped .
be made to agree with and for caesiu1n .
These relations then cannot be considered as exact , they are very and provide fairly good approximations to the actual lines .
We can therefore only regard the formula based on them as approxima .
It would doubtless not be difficult to add other ternls to the Hicks expressions on the atomic weights or other properties of the * Hicks , , p. .
so .
Dr. P. V. Bevan .
metals and thus obtain expressions correct within the limits prescribed , but there seems little point in attempting this at the present stage , as Hicks ' own expressions are so near the actual values that new terms would be so small and the possible errors so that we could conclude nothing as to their real correspondence with the actual state of affairs and they would be mere interpolation terms .
dening of with of Density of Vapour.\mdash ; Absorption lines are never quite as sharp as good emission lines , though in some cases their apparent breadth can be certainly less than .
U. , and therefore there is some doubt in measurements of wave-lengths by means of absorption , as errors may be introduced owing to the broadening of the line not being symmetrical .
It seemed worth while therefore to examine a typical case with some care to see the effect firstly on the determination of ] cvth , and secondly for the interest of the asymmetry of the broadening itself .
It is observed in all cases these irlcipal series lines the broad line obtained by using dense vapour behaves in the same way .
The line broadens unsymmetrically , extending much further on the more yible side than on the other .
As all the lines I have observed behave in a similar way a careful examination of one pair was made , and for this purpose the second pair in the caesium series was investigated .
The advantages of taking this pair were that it occurs in a visible part of the spectrum , so that ordinary glass for lenses and prisms could be used , and also that the effect on each line could be studied , as the two lines are separated by a nsiderable interval .
To get more dispersion for this case a small Thorpe grating mounted on a prism to form a direct vision spectrograph was used .
The effective aperture was somewhat less than 2 cm .
This was mounted on an ordinary laboratory spectrometer and the telescope aCed by a camera .
The mounting of the apparatus left a reat deal to be desired in the way of rigidity , but some satisfactory raphs were obtained .
Fig. 2 of the plate shows enlargements of the photographs obtained : ( 1 ) shows the absorption using a white source of Nernst lamp ; ( 2 ) has the iron arc photographed in addition to the absorption spectrum .
With the apparatus used it was very difficult to keep the adjustment when changing from the absorption spectrum to the comparison spectrum . .
However , as the series of photographs shows , with a certain number of failures there were obtained some good results .
a glance at the figure the asymmetry of the effect can at once be seen .
Measurements were made of the edges of the lines\mdash ; the edge of the line is of course somewhat indefinite as there is no abrupt change , but measurements of the two edges are probably fairly comparable , as , 1912 .
] Spectroscopic Observations : Lithium and Coesium .
at any rate until the line becomes very broad , the aspect of the two is practically the same .
The following table gives a set of measurements of the two edges of these lines:\mdash ; 4555.26 .
4593.16 .
4555 .
4555.78 4593 .
57 .
92 .
92 .
94.20 59 .
92 .
53 .
95.74 95.59 The wave ] were determined from the iron lines The second of these was not a good line , and there may easily be in some of these an error of A-U .
from this ) in any case it is clear that the broadening is very lnuchulOl'e on the side of greater wave .
The narrowest lines of ave were considerably broader than ] used in -length determinations .
dening effect is eater in the earlier members of the series , and the lines after the first few of the early members relnain very fine , except in the case of very dense vapour .
For example , in the one of the raphs from which was determined , the breadth of the line was less than .
In all of lines used for eluent the eadth of the lines was not greater than and in most cases not rreater than .
From the above table it is clear centle 1Jf the will give too tlreat a , but the is ] for account to be taken of it with the accuracy available for ments .

