]\gt ; Spectra of of the By P. ( Communicated by Prof. Sir J. J. Thomson , .
Received nuary 6 Read , 1910 .
) [ PLATE 4 .
] .
W. Wood , of Johns Hopkins University , has ated the absorption spectrum of sodium pour and has shown that its mosC characteristic feature is the appearance of lines of the principal series of Rydberg and lCayser and He was able , by a of considerable dispersion , to ) serve { -lengths of lines of the principal series , thus giving by far the mosl complete yet been ) served .
Wood stated , in a of his on " " The Dispersion of Sodium } ) \ldquo ; that sodium vapotl behaved in special manner .
It appeared a cloud of sodium vapour could its boundaries in a closed evcn wheu the rest of the tube was sted to a very acuulno , and that in is way sodium was peculiar , other lnetals no corresponding conclusion is , , not correct .
The author of the present communication has shown* the odium cloud does not exist in a vacuum , but that it behaves like ordinar vapour , in a vacuous space instantaueously from a to a colder part of the containing vessel .
The ellormons optical deusity of the light in the region of the ives the sodiunn cloud an } ) ) of density which is very deceptive .
It been shown potassium and lithium behave very much like sodium , so that similal effects of ) dispersion have been observed ab some of the principal series lines in the case of these two metals .
In the i of potassinnl vapour for anolnalous dispersion , the author found that the ) ) spectrum of the vapour was of great interest in itself , as it showed a number of the lines of the principal be the with some channelled spectra , but not lines of the series or lines hich are not connected with any series .
The present colnmunication is to boive the csulls the found with .
Phil. Soc. Proc vol. 13 , p. 129 .
' Phil. Mag 1909 , vol. 18 , p. 407 .
' Phil. Mag 1910 , vol. 19 , p. 200 . .
P. V. Bevan .
regard to the spectra obtained in the same way with rubidium and The results for potassitun have already been communicated to the Physical Society , so that the actual measurements ofneed only be given here to complete the series for the alkali as far as they have as yet been ated .
The method has resulted in the discovery of 15 new lines in the potassium spectrum , 20 in the rtlbidium spectrum , and 11 in the caesium spectrum , so that we now have the wave-lengths of 24 lines for potassium , 25 for rubidium , and 21 for caesium .
These with Wood 's 48lines for sodium give us a pretty complete set of principal series lines for the alkali metals .
The case of lithiunl has not yet been satisfactorily treated : this will be ref.erred to later .
The method for potassium was to pass light from a source through a steel tube closed at the ends by quartz plates , enclosing some potassium nletal which was heated by a row of gas jets .
For rubidium and caesium the metals themselves were not available , but if the chlorides of these metals are heated with potassium or sodium , quite sufficient decomposition takes place give a vapour which is dense enough in rubidium or caesium to give the absorption spectrum .
Undoubtedly modifications are introduced into the spectra by the vapour being a mixture ; some of these will be referred to later .
The source of used was either a Nernst lamp or an arc lamp , and the iron arc was used for lines to serve for measurement .
In the following tables are iven the wave-lengths and " " oscillation frequencies\ldquo ; for the metals potassium , rubidium , and caesium:\mdash ; Potassium ( fig. 1 , Plate 4 ) .
The lines to 24 are new .
of Rubidium ( fig. 2 , Plate 4 ) .
The lines numbered 1 to 4 are iven b is ) lines 6 and 7 are not so accurately deterruined as the others .
hesc lines are doubles , but appear in the absorption spectra as dened lines , so that the separate components could not be measured .
The are , as near as could be , those of the lnole ) onents .
After 7 the are so close that they ) ) a fairly riuite line .
The nunlbers are to ?
or 4 nnits in esimn ( .
The lines nullbered 1 to ) are ) ivell l Knybe :S 'lfandbnch der Spectroscopie , ) .
, ' .
Soc. Proc , p. ) .
'Handbuch der ) .
S Raulage , .
Soc. Proc , 1 ) .
304 .
424 Mr. P. V. Bevan .
8 and 9 are iven a , as Ramage 's measurement of these two lines is not exact .
The lines 10 to 21 are new .
The accuracy is about the same as for the rubidium lines .
These values of the , together with Wood 's values for sodium , give us means of testing the various formulae which have been suggested for the waye-lengths of the series lines .
The formula of Kayser and Rung , , represents the observed wave-lengtha well for up to about , being 3 for the first member\mdash ; with the exception of the early members .
Kayser remarks that a term is wanted to make the reement better , but even with this term the formula cannot be made to fit the observations well .
Taking Wood 's values for sodium lines and the formula\mdash ; 8 and 9 are iven a , as Ramage 's measurement of these two lines is not exact .
The lines 10 to 21 are new .
The accuracy is about the same as for the rubidium lines .
These values of the , together with Wood 's values for sodium , give us means of testing the various formulae which have been gested for the waye-lengths of the series lines .
The formula of Kayser and unge , repl.esents the observed wave-lengtha well for up to about , being 3 for the firsG member\mdash ; with the ception of the early members .
Kayser remarks that a term is wanted to make the reement better , but even with this term the formula cannot be made to fit the observations well .
Taking Wood 's values for sodium lines and the formula\mdash ; and calculating from the values of , 10 , 20 , 40 , we obtain the following:\mdash ; .
These values give , which is very far from the value , the term being very much too large .
If we takc values at , 6 , 12 , 24 , we find This set of constants gives instead of 302749 .
In the members of the series we get errors which are small , but .
correspond to errors of 2 or 3 in the value of There is no doubt that by adjusting the constants , a fairly good approximation could be obtained , but the remaining errors will still be more than the uncertainty in the measurements .
The particular sets of values chosen above for were taken , as they considerably simplified the arithmetic necessary to solve the equations .
If we Pydberg 's equation we find that this will not suit the measurements .
In this formula 7 is 2 for the member of the principal series , and , according to Rydberg , is an absolute constant 109675 .
It is to calculate the constants for 's equation ; but as this is only a particular case of Ritz 's equation , 1910 .
] ' of Alkali Metals .
which we shall proceed to consider , it is not worth while dealing with the form .
Ritz 's equation is where is the " " oscillation freque cy the absolute constant of and constants for each substance , takes the values 4 , etc. This formula undoubtedly gives closer esults for the for series lines than the formula of and or that .
It has the great of a connection between the principal and the associated series .
But if we use it for the whole sel.ies of sodium wave-lengths , it does not well enough with the observations for us to regard it as more than an empirical formla .
Using 's values corrected to vacuum values , we find the following values of the constants:\mdash ; Other calculations for the values For the accuracy of these numbers we can easily see that A is not appreciably affected by considerable errors in the other numbers , if we use a high value of ; so that A the convergence frequency can be regarded as accurate to the first decimal place .
The values of the other constants depend on the order chosen for calculation .
The two last values iveu for have possible errors and respectively .
Now this constant should , on 's theory , be the same as the constant in the Balmer series for hydrogen , which is .
This is sufficient to show that the formula does not give a scheme sufficiently accurate for the large number of lines in the complete sodium series .
We find the same results for the other metals .
For the rubidium sel.ies we DOet The accuracy of these numbers is not so reat as in the case of , as the wave-length determinations not so good ; but the departure of from the hydrogen value is more than is to be accounted by errors of measurement .
1otassium gives a better value for , 109667 ; but the difference ] ) is still than is to be accounted for by elTors of measurelnent .
The values of the llation frequency for sodium have been chieti on in the above a1gument , as the determinations by Wood have greatel accuracy than the admit .
It is clear that the formula can only be regardea ' as an approxima- .
tion , and that it must be modified it is to express with the required accuracy the values of oscillation frequency observed .
The required 1nodification can be brought about in several ways : we can regard No no longer as an absolute constant for all substances , and introduce a constant in place of it for each different substance .
Tbis would certainly give a formula of the same type , which would express the results better ; but the general relations between different series spectra would be lost , and it seems better to look to the kind of modification suggested by Ritz 's argument\mdash ; the introduction of another term in the denominator of the fraction of which is the numerator .
This would us a formula This , another constant , would enable us to fit the results Juore accurately .
But , of course , other ways of modification are possible\mdash ; the addition of another whole term to the -hand side of the equation\mdash ; change of into in the term , and so on ; and it not yet seemed vorth while to go the enormous calculations to see which of the large ' of modifications would produce the best result .
The conclusion is that the Ritz formula , like other formulae , is only an empirical formula .
Its value lies in the connection it indicates veen the associated series and the principal series ; and it is not worth while trying to modify it until we have a knowledge of the associated series which is comparable with the we now possess of the p1incipal series .
Hitherto we have discussed the question of the actual lines and their measurements .
There are some interesting points in the nature of the lines which llow need attention .
In Plate 4 , figs. 1 , 2 , 3 , are represented the spectra of potassium , rubidium , and caesium respectively .
Owing to the method of experiment the rubidium and esium spectra are superimposed on the potassium spectrum .
If the vapour be not very dense the lines appear fairly sharp and well defined ; as the vapour density is increased the lines in the case of potassium broaden and satellites appear .
These can be easily seen in { igs .
4 and 5 on the plate .
These two photographs are taken with eased density of vapour\mdash ; the tube at a temperature .
The broadenin , of these lines is not symmetrical .
The less refrangible side of the line\mdash ; the -hand side in the \mdash ; shows more than the other .
This can easily be seen by comparing the lines at 3447 in the photographs .
The satellite line on the other more refrangible side remains .
Soc. Proc. A. vol. 83 , Pl 6 1910 .
] Absorption Spectra of Vapours of distinctly separated from the heavy absorption at the main line ; 4 is the photograph taken with greater density of vapour .
In the case of rubidium and caesium there is also broadening , which is again symmetrical .
Fig. 6 represents absorption due to caesium vapour .
The central part of the photograph is due to the light from the crater of the arc ; this shows the lines fairly fine in their normal position ; it can easily be seen that the broadened line in the other less intense part is not symmetrical with respect to the fine line .
This appears especially at the line 3611 .
More work , however , requires to be done on this effect of increased density .
Some effects of the vapours being mixtures appear .
represents the absorption due to caesium set free from the chloride by with sodium .
The caesium lines do not appear clearly , but the interest of this photograph lies in the banded spectrum in the region of the sodium line 3303 and the caesium lines Nos. 5 , 6 , 7 , 8 , 9 and upwards near it .
( The numbers in the figure refer to the numbers of the caesium lines in the table , p. 423 .
) With the density o sodium vapour present for this photograph no banded spectrum appears in this position with sodium alone , and no banded spectrum appears in fig. 3 at this position , which is the absorption of caesium and potassium .
We appear to have , therefore , a spectrum due to the mixture of the two vapours .
We have a similar phenomenon in .
The right-hand patch on the photograph shows two ions of absorption in the green .
These do not appear in the spectrum due to potassium alone , or in the spectrum due to sodium and caesium .
In potassium spectrum there is a band region in the yellow-green ( just visible in fig. 1 on the right-hand side ) .
This paru has , however , disappeared in fig. 6 , as for this photograph there was dense vapour , and the channelled space potassium spectrum has spread past the lines well into the orange-yellow .
With the small dispersion that has been used , these spectra have not been resolved into lines as yet .
Some experiments were made with lithium , but so far without success .
Attempts were made to get lithium vapour by a lithium salt with sodium , but no trace of the red lithium line could be observed .
This is not very prising , as lithium vaporises at so much a higher temperature than sodium .
But rubidium and caesium are produced so .
easily that the experiment was worth attempting , and , further , there is the observation of and Dewar that the lithium red line only appears as an absorption line in the presence of sodium .
Metallic lithium was also used , but it is difficult to get a high enough temperature .
The lithium acts on steel and also on porcelain , so that tubes of these materials cannot be used .
An attempt was made with a tube of hard Jena glass , containing the lithium , the whole being enclosed in a steel tube with quartz ends .
Dense enough vapour VOL. LXXXIII .
Mr. R. B. Sangster .
Novel Phenomenon in the could be obtained to show absorption at three of the principal series lines , but the glass tube collapsed before any further result could be obtained .
It is hoped that more success may be obtained with a quartz or silica tube .
The author feels much pleasure in expressing his thanks to the Government Grant Committee of the Royal Society for a grant to purchase the quartz which was used for the purposes of this research .
On Novel Phenomenon in the Inequality of gnetism at certain tations .
By ROBERT B. SANGSTER .
( Communicated by Dr. C. Chree , F.R.S. Revised MS .
received January 11 , \mdash ; Read February 10 , 1910 .
) In this research the author dealt with data from the Magnetic Observatories at Greenwich , Falmouth , and Pawlowsk ( Bussia ) .
The declination , the horizontal , and the vertical , forces of terrestrial magnetism are denoted by , and respectively , and the horizontal force in the astronomical meridian by X. Diurnal inequality increments are designated by the prefix .
The formula for obtaining is and the unit of force employed is C.G.S. .
The factor for conversion of from minutes of arc to units of is ; hours are counted continuously from midnight to midnight , and in the attached diagrams the scale value employed is mm. The research ultimately proceeded on the principle of the vector diagram .
The latter has usually been employed to display variations of the horizontal component , but in this case the author uses values of and , and so obtains a representation of the diurnal force changes as projected on the vertical plane containing the astronomical meridian .
Curves showing the force variations in this plane may be designated diagrams .
The X axis being taken as positive to the right and an increase of as downwards , the resulting curve has to be regarded as being viewed from the east .
The feature in these diagrams to which special attention is drawn is the agreement in the trend during the early afternoon hours with a line drawn at an

