]\gt ; Series of Lines in Spectra of Alkaline Ear.ths .
451 a point near the middle of the ]ength of the wall and at a distance up it which is enerally about three-quarters of the height .
The horizontal adient for a distance outwards not exceeding 1/ 10 the of the wall ill not differ by more than 2 per cent. from the normal vertical yradient over a large horizontal area .
On the need Series of Lines in Spectra of the Earths .
By W. M. HICKS , Sc. D. , ( Received IsIay 15 , 191.5 .
) The problem of the limits and numerical relations between the of the enhanced series of doublets in the alkaline earths has for been a difficulty to spectroscopists .
Ritz*in 1908 gave arrangements for the Sharp series from Mg to Ra inclusive , and proposed series formulae for Ca , Sr , Ba , in which alone he had three lines from which to calculate the co1lstants .
The absence of extra lines rendered it impossible to test his formulfe , but the values of the constants obtained for his formulae were quite out of line with those of the constants in other series , and produced an instinctive doubt as whether it gave the correct relation .
It is now possible to test his limits by considering whether the denominator differences which the observed separations have any relation to the oun or not .
The result of this consideration is definitely adverse .
In none of the three is it possible to make the multiples of the oun without supposing observation errors in the doublet separations which are quite inadmissible ; and even then in the cases of Ca and Ba by taking odd multiples of , which is never the case for doublets in any other known series .
There can be little doubt but that Fowler has at last settled this question by taking the Rydberg numerator constant to be in place of , thus combining in one set lines which on the old supposition would be in two series , depending on Sharp and Principal sequences .
The object of the present note is the determination of the connection of these series with certain laws which have been arrived at in previous to this Society 'Phys .
Zeitschr vol. 16 , p. 521 .
' Phil. Trans , vol. 214 , p. 225 1914 ) .
Phil. Trans , vol. 210 , p. 57 1909 ) ; , vol. 212 , p. 33 ( 1912 ) ; , vol. 213 , p. 323 ( 1913)\mdash ; referred to in the following as ( I ) , ( II ) , and ( III ) respectively .
, , and series and for the series of .
These values will be in what follows .
In order to get a conspectus for the whole group I dded a discussion for the corresponding lines in Ba and A few preliminary remarks on the series formulae may not be out of In ( II , p. 35 ) I have reasons that in the triplet series of this group the series depend on the sequence which in the alkalies give the , and rice ( .
If weassume the same result for the doublet series it will be found that the new formulae reproduce the other lines with much greater closeness han those used by Fowler\mdash ; who has very naturally taken a Sharp sequence for a Sharp series .
The modification consists in taking the constant as. .
fraction or , which comes to the sallle thing , writing the denominator .
The formulae on this basis , using the first three lines for data , are Ca and the values of , compared with Fowler 's , are Ca Sr In the case of Ba ( as allocated below ) the formulae calculated from the second , third , and fourth reproduce the first line with values of when fraction , and 159 when fraction .
It is probable that a formula with would reproduce the lines of each element with very fair accuracy .
But the residual errors show that these limits as well as those of Fowler are too large .
The cause is probably due to the fact that the first line is not the calculated but the observed .
They never exactly agree .
In the case of and series I feel some doubt as to whether the actual lines follow a given formula sequence , although it may be possibl : they are in some way related to such typical sequences .
There can be little doubt , I think , but that the values of the wave numbers are determined by an expression of the general type but the Series of Lines in Spect of the Alkaline ioertainly cannot itself be a mathematical function of the for both the and lines , or for both the and lines .
They are best discussed Wy calculating the actual values of from the observed wave nbers , then attempting to investigate the relationships between them .
This is the method adopted in ( III ) and it will be followed in the present note .
The values of for the elements in question are ( III , pp. 344-346 ) Mg .
Ca. Sr. Ba .
Ra .
The theory that the separations of the series depend on multiples of the oun , and that the satellite separations of the , and consequently the separations , also depend on multiples of the sauue quantity , can , I think , be considered as established , and ma .
be taken as a tesb for new cases .
This part of the discussion will then be taken first .
The exactness of the determination of the denominator differences which produce the observed separations depends almost wholly on the exactness of the separations as measured , and only htly on the true of the lilllit .
In fact the value of the difference is ected by the same percentage error as the value of the separation adopted .
Any error in the limit produces only a small effect , because it enters as the difference of two numbers each affected by nearly the same errors .
To illustrate this point the ca]culation is given for Ca , and the results only for the other elements .
Fowler quotes for Ca a doublet separation and limits for and for .
Take the true limits as being greater and that the value of is affected with an error .
Using the numerator the denominators for these limits are respectively and .
The difference of these digits , for the in is , therefore , In this will be less , and unless the value of the limit is very much in error the true difference cannot differ from 3952 by more than 2 units in the last place .
Now for Ca is , and .
The possible error in is probably much less than one per .
These numbers agree easily within any possible observation errors , and may say that the separation is due to a denominator difference .
In ( III ) , , have been used to denote the corresponding values in the triplet series , and for .
In order to have a consistent notation we might denote the multiplicity of the series by a corresponding dash and write respectively for triplets or two separations ) and for doublets .
The same notation might be extended to the series themselves , thus would VOL. XCL\mdash ; A. 2 and the other elements as showing no line corresponding to negative values in the Diffuse .
A similar difference is shown also in the triplet series ( III , p. 356 ) , and seems to show that Mg is more analogous to the low melting 1 point elements of this group than to the earths .
This difference is exemplified also when the atomic volume term is considered as is done below ( p. 462 ) .
$$13 Passing on to the sec.ond law , the separations of the satellites in the Diffuse series are caused by denominator differences in and .
That for the first set is always greater than for succeeding sets , for which latter they are in many cases , though not always , the same .
Moreover they are generally multipIes of .
The calculated results are here affected by observed errors in the single lines , but again almost independent of the limit error .
The results are of course independent of any formulae beyond the fact that the wave number is of the form The values of the denominators are given in I. The figures in brackets denote sible errors , and as the effect of the change of limit is practically the same for and it is only entered for the first .
Mg shows no satellites , as is the case also in its triplet series .
The corresponding numbers for Ba and Sr are also placed in this Table from the discussion given below .
In order to obtain the corresponding quantities in Ba and Ra it will be necessary to make a digression to discuss the enhauced series in general in these elements .
gave for the series the Iines ; to which Saunders*has since added 2287 , 2202 .
The Zeeman for the first two pairs have been observed and the ' Phys. Bev vol. 28 , p. 152 .
Series of Lines in Spectra the Alkaline Earths .
Table I. Ca. 858 880 lculated from The satellite differences are therefore 59 , although , as third line probably has large observatiou errors , the 55 , in spite of its exact agreement with the observed separations , , 56 , or even the same as for second line .
The denonlin{tor difference for and 6 is so excessive as to throw some doubt on the allocation of the last to Sr. 4587 2429940 ( 16 ) 4284 3 4 .
4197 The satellite therefore 66 .
The values for the second and third not so cIose as in Ca , but are both within error )nits .
( so ) 8087 85662\mdash ; 26376 \mdash ; Zeeman patterns are those for a doublet series .
The doublet separation is .
There can be no doubt but that the allocation is correct and also that they form an enbanced series .
The formula calctllated from the 2nd , 3rd .
and 4th is This reproduces the first with an error\mdash ; 143 .
The limit may be taken as probably correct within .
The two limits are then and additional or .
These give a denominator difference .
Dr. W. .
Hicks .
On the Enhanced Now , for with an uncertainty of about unity in fourth Hence For Ilifluse series take the set\mdash ; 887 .
1690.59 ( 4 ) 416624 205.41 1690.97 ( 4 ) 37845 97 94.90 ( s ) 3794087 169066 ( 5 ) 2528 In the first set , the line ( 614S ) is taken from an observation by Moore* in the course of bis investigation of the Zeeman effect in the Ba spectrum , and it was strong enongh to enable him to determine its Zeeman pattern .
There can be little doubt as to the allocation of the first two sets , as they both show the Zeeman patterns for diffuse doublets .
It is curious , ever , the of the first old be so luch weaker than the , which is abnormally strong .
It would seem that only a small proportion of the urations which give tlJenorn ) line split up and laterally displaced to the set .
As a ltlle they appear to be l\amp ; ecaron ; ss stable than those giving .
It will be seen later that a similar effect is shown by Ha .
In the third set ] ( is a line observed ) Exner and Haschek .
The other lines in the first three sets are measures of Kayser and Bunge , whose estimates of ) ossible errols , lated to vave n are inserted in brackets after the measures .
The remaining lines are by Saunders .
cit Using the limit and the second and third lines , the formula for is This gives for , which is Saunders ' 2234 , and for , which establishes the set .
denominators are given in Table I. It is seen that the denominator difference for the first satellite is close on , but the diHerence between and the observed is too large to give so satisfactory a decision as in other cases .
This , however , may well be due to the nown observation error in .
This line was not observed 'Ann . .
Phys vol. 25 , p. 314 .
kaline Emake teparation 8outhan 5rror bthrown oation ohenonuenon Trror iconfirm Ra.\mdash ; The case of 1s more difficult as the spectrum has not been measured above 6640 nor below .
Moreover , with increasing atomic weight the .
appear to get more unstable and to break up into numerous others , with the consequent disappearance of regularities .
By far the best measurements are those of the spark spectrum by Rung and Precht .
* Their plates were not sensitive above 6500 .
Exner and Haschek observed a few above this .
Ritz ( loc. cit. ) for the series the two sets \mdash ; ( 100 ) , -(50 ) , and ( 15 ) , ( 10 ) These givo doublet separations respectively of and .
To these may be added for , ( 20 ) , ( 50 ) , ( 50 ) These selections are probably correct\mdash ; although we have not their Zeeman patterns as in Ba .
They are not sufficient , however , to determine the limits unless a Rydberg formula were exact , in which case the two SliItes would be sufficient .
But in the other elements limits as thus found are ab 1800 too high , and thus quite inadmissible .
We are , therefore , driven to apply the ethod which is used in ( III , p. 405 ) in the spectrum of Au .
The Rydberg limit lrom the two lines is about .
Fronl with the other elements real limit shotlld be about 1800 less , in the neighbourhood of 56000 .
It should be that the value of for the observed separation is a mnltiple of the oun\mdash ; of from analogy with the others\mdash ; and that the difference for the separation of should also be a multiple .
If the first condition is applied , a series of values for the limits are obtained .
Using these , they are tested for the second condition .
It is found that one alone satisfies it , and this is then taken as the real limit .
In order to apply this method it is necessary to know beforehand the degree of accuracy of } the data are susceptible .
These data are respectively the values of , and the separation .
For Ra , where is probably less than , or say 1 in 3700 .
the other data it is necessary to know the possible errors of observation , which unfortunately and Precht do not give .
But for the special lines in question to the third decimal place ( excepted ) .
'Ann . .
Phys vol. 14 , p. 419 .
Sharp separations cannot be exactly the same if our error limits are correctly uned .
Those of and may be equal , and as is possibly really its is in apparent reement with indications in other series .
In any case they do not differ by an amount sufficient to affect the determination of the satellite difference .
] will , therefore , be detel.mined by The satellite separation , with the same supposed observation errors , is , or say Between and 55000 nine limits were found which satisfied the first colldition .
The values of the denominator difference ran from to .
These were tested on .
They all distinctly failed to satisfy the condition except that due to .
This belonged to limits and the actual limits to be ]arger , the denominator difference is .
With the assumed maximum values of above for @ , can vary only by .
This limit value is now used for , only will be slightly different if the for is .
The denominators for and are now and .
The difference is , where are the errors in observed and .
Now .
If hese are to be equal .
This equality can be met by values of well their maximum possible errors assunled above .
We are justified , therefore , in taking and the limit The set of lilles should be } reverse ordel in the red .
As no separation 4857 occurs there observed lines , at least is outside the observed , but it may still happel ) that the lines and are within .
If so , their denominator rence should be a multiple of and mantissae considerably less than the corresponding ones for .
Also the lines should be strong , and fro1n analogy wirh the other elements the satellite separation should be rather greater than four times that for .
There are * There are several other lines with } separation , but it will be safer to deal only with lineh .
and differ very closely by multiples of result by all rong .
denominator with a mantissa smaller than for , and latter one with a value much higher than with other } ) ectra would lcad us to expect .
Exner and Haschek give a spark line at with Tllis gives a denominator , with mantissa above that for which again is 146708 above that for ? .
It is .
therefore , in fair order of nitude .
The difference 58814 should be a multiple of , but it is affected with an observation errol of .
With , this is 7 Now , and as , even the large possible error of could not alter the multiple .
It is quite possible , therefore , that is ; the satellite is too faint to have been observed and is outside the region of observation .
If , as in Ba , the satellite differences for and 4 are the same , these unobserved lines would be at ) and and the satellite separation would be The first three sets of as thus arranged are therefore [ -l280351 ] [ -l280351 ] [ -l280351 ] [ -l280351 ] [ -l280351 ] [ -l280351 ] [ -l280351 ] [ -l280351 ] [ -l280351 ] [ -l280351 ] [ -l280351 ] [ -l280351 ] [ -l280351 ] [ -l280351 ] [ -l280351 ] 496.55 4857.34 [ 27627 ] 2428 [ 4117478 ] Series.\mdash ; In the series as given by Fowler the separations are both than in the satellites , for instance as against in Ca , and 4 ; as against in Sr. The differences are too large and too systematic to be ascribed to observation errors .
They seem to be constant for the lines oairs exist withth aration.iscussion oseries hhown thatdifferent owhich would point tisplacements iimits i uggested values forrect , eparation oshould bseparation s andthatinRa , expected .
These are found , and logous sets can be found in all four 5 .
spectra wseparations ahich tZeeman pattern one of excepted)iven boor ( Of these one is probably negative corresponds to , the other to The laws relating to the dependence of the separations of doublets , triplets , and satellites on the oun may be considered so well established that they can be used as tests for ) theories .
The foregoing thus affor strong testimony in your of 's explanation .
We may use these new series as tests for certain other relations , which , although perhaps they cannot be considered as proved , yet have a considerable body of evidence in their favour .
( 1 ) We will take as the first of these , the relation that the inator of the first term of the -sequence\mdash ; that which in the alkalies ives the series\mdash ; is proportional to a multiple of the atomic volume .
It lvould appear that there is in connection with each atom a fundamental quantity , of the nature of a volume .
It is roughly proportional ( probably to about 2 per cent. ) to the atomic volume as measured the ratio of the atonnic weight to the density of the substan ce in .
the solid state , a measure which is dependent to some extent on temperature and other effects .
The law would state that the mantissa of the denominator of the first line should be ) , where is the atomic volume , a whole number , and a number in the hbourhoodo of , this number being probably correct to about ) per cent. The difficulty in determining its value more accurately depends 011 the facts that the true value of is not known within that percentage and to the uncertainty as to whether the fractional part of the denominator itself follows the law , or whether there , is some group constant to the term is added .
But , if so , this constant is so near unity , hat it is quite sible to determine whether the multiple exists\mdash ; and in that case its \mdash ; without uity , allowing factor .
lies the sequence colves the series and 462 Dr. W. .
Hicks .
On the Diffuse sites ivhich tuter sromsequence more ftype tasiest , therefore , which the more intense lines are formed by lateral displacement .
Here S is formed from b.y adding .
In certain ofher cases it would appear that the Sharp series depends on the sequence and the Principal on $ sequence .
In this case we look for the presence of the law in , which is .
is apparently the case in the triplets of Group 2 .
The values of the multiples for the elements already tested are given in ( III , p. 47 ) .
In the present case .
then , both and the must be tested .
The result is that Mg follows ) in and not in .
On the contrary , in Ca , , Ba the law is followed in and not for .
The densities used are the same as in ( II ) , , , , Ba .
The results are given in the following table:\mdash ; of ( 2 ) Ca. .
, , , , Ba .
, , , , .
, , With density of greatest deviation is shown by Sr , in which the density determinations different observers vary greatly .
It is clear that these enhanced series support the truth of the law at as a first approximation .
If is to ehave like Ca , Sr , Ba , its density must ) , where is an integer .
From it would be 6 or 7 , oo.iving a density of or A theory applied triplet serie .
, in which the numbers are less defillite , gave with or 6 .
( 2 ) Apparently the Diffuse sequence depends on multiples of the oun in a similar manner to that in which the sequence depends on the atomic volume .
In a very large number of cases the denominator of the outermost satellites of the first set is a multiple of the , and in all other cases the difference between the denominator and nearest multiple of is itself a nultiple of the oun .
Ill the latter case the nature of the evidence for low atomic weights is not decisiye as in them the oun is so small that it is easily smothered by observation errors .
In the present discussion , then , the question to be tested is whether the first satellite denominator is a multiple of , or differs froru such multiple by a multiple of the oun .
In the case of Mg the is too.small to settle the question no way or the other .
Series of Lines in Spectra of the Alkaline .
463 Ca , using Paschen 's value of For , mantissa with Sr , using Randall 's value of For , mantissa , is not a multiple of For , , , If , however , the limit be determined from the second , third , and fourth lines\mdash ; which is generally to be preferred\mdash ; it co1nes out 49lower .
To make the mantissa an exact multiple , the limit should be lower by , and probably also more correct than .
It is curious that Sr shows precisely the same kind of displacement in the triplet system , the exact multiple being there found in instead of in Ba.\mdash ; Here neither nor give multiples of , but they differ from such by multiples of the oun .
For , mantissa Now it was seen above in considering the satellite ) arations that with a large possible error makes the denominator too by 58 .
This corrected value would give for above Iantissa Ra\mdash ; For , lnantissa With the relations hold , within small observation errors , for both .
As to the changes in the limits equired by the above values of , it is to be noted that for Ba and Ra they should not be , because if the theory used to determine the Ra limit is correct the value obtained must be close to the real value\mdash ; and is quite allowable .
For Ba the limit is obtained from the second , third , and fourth lines and is a better value erefore than that obtained from the first or fifth .
Fowler 's limits for Ca and Sr are on the face too large , as the for the later lines .
Hence , here again the evidence of the enhanced series gives support to relation obtained in ( III ) from a study of other series .

