]\gt ; Radium emanation analogue of , , general ations for S.D. value of and link and links , serles serles source of 5648 separations .
value of the oun . . . .
series serles and Pserles value of the oun Analogue of 1864 . . .
Neon .
series , , parallel sets .
series , \ldquo ; link normal and lines . . .
, , value of the oun , WATSON'S separations . . .
ROSSI'S series Table of constants 4#$ . . .
4$7 . . . .
44$ . . .
46$ THE present communication has two objects .
Its subject matter is an attempt to obtain some knowledge of the series relations in the spectra of the group of the monatomic rare gases No to RaEm , whilst the methods employed will serve to illustrate the fundamental importance as instruments for further research of the new facts brought to light in the previous communications .
* The importance of the first object will be generally acknowledged , but it does not yet seem to be realised how definite and exact those new relationships are , even in their as yet undeveloped form , and how powerful an instrument is placed in our hands for the analysis of spectra .
It may be well therefore to commence by a brief rlsume ' of some of these laws as applied in the succeeding pages .
Further , as the establishment of the results obtained must by its nature depend on the numerical comparison of a very large number of lines in all five spectra , and as this evidence must be fully set out to * ' A Critical Study of Spectral Series Part I. , 'Phil .
Trans , p. 67 .
Part II .
, , , , 212 , p. \ldquo ; Part III .
, , , , 213 , p. 323 .
, , Part IV .
, , , , 217 , p. These will be referred to respectively as [ I. ] , [ II .
] , ] .
DR W. M. HICKS : A STUDY OF SPECTRAL SERIES .
enable a specialist judgment to be formed on it , the communication has unfortunately become very lengthy .
The mass of detail will perhaps be rather dreary to the general reader not specially interested in this line of study .
It is apt also to hide by its amount and complexity the general conclusions arrived at .
I propose therefore to give a slight general survey of these conclusions before giving the evidence .
As is well known the wave-numbers of series lines depend on four types of sequences , , , , and that .
in any one series they depend on the differences between one sequent of one type and the successive terms of the sequence of another type .
These sequences are all of the form , where is RYDBERG'S constant and is of the form fraction , the fraction being , as a rule , determinable as a decimal to six significant figures .
Our aim is to discover the properties of these functions .
The fractional part depends in some way on the order , although whether it can be considered a definite function of in the ordinary sense is doubtful .
* This fractional part will be referred to as the mantissa , and in dealing with it , it will be regarded as multiplied by , i.e. , as if the decimal point were removed .
The Oun.\mdash ; It is found that in each element a constant quantity particular.to each element plays a fundamental part in the constitution of the sequences .
This is called the oun .
The and sequences depend in definite ways on multiples of this quantity , whilst it also enters into the constitution of the and .
Its determination is therefore for each element a matter of the first importance .
Denoting its value by , the quantity is of such frequent recurrence that it is useful to treat it as one datum .
The oun is accurately proportional to the square of the atomic weight , and is given by , where denotes the atomic weight .
In the case of doublet or triplet series , the corresponding separations between them are due to different limits whose mantissae differ by amounts or ( say ) .
In all cases these are found to be integral multiples of the oun .
For triplets is always somewhat greater than 2 .
In the case of series where satellites occur , the separations of the latter are due to differences in their sequences .
The mantissae of these latter again differ by quantities which are multiples of the oun , and in the case of triplets they appear in normal types to be very close to the ratio 5 : 3 .
The \amp ; quence.\mdash ; In the normal type the sequent.of the extreme satellite has its mantissa a multiple of .
The only known exceptions are found in Sr , Cd which show the multiple law , Sr in and Cd in instead of in .
In both these cases also the Zeeman pattern is abnormal .
As the main lines ( and in triplets also ) have their mantissae greater than that of the outer satellite by multiples of the oun , it follows that all the sequences for the first order have mantissae multiples of the .
It is probable that this is true for all orders of , but the data are not * Astro .
44 , p. 229 , see also [ III .
, p. 339 ] .
, also ' Proc. R. 8 , 91 ( 1916 ) .
DR. W. M. HICKS : A CRITICAL DIE OF SPECTRAL sufficiently accurate to prove this , although they obey the rule [ III .
, p. 340 .
] The Sequence .
sequent of the ffist order also has the multiple of The material at disposal is not so comprehensive as in the case of the series , for , exoept in the second group of the periodic table , the lines occur chiefly in the uItra-red .
The proof of the above statement is perhaps , therefore , not so conclusive as in the case of the sequence .
It completely stands the test however in the rare gaaes .
There seems some evidence that series also show a satellite effect in a smaU degree \mdash ; of one or two .
In the second group it seems to be a general rule that in many of the low orders the sequents receive very iarge displacements from their normal value , so that a normal lin is much weaker or is altogether absent and replaced by others separated from it by considerable numbers .
This ako is found to be the rule in the present case .
Displacement.\mdash ; Regarding the ordinary doubtlet or triplet series we may consider the second ( or third ) as displaced from the first by the deduction of a oertain number of ouns from the mantissa of the limit ; or better perhaps regard the last satellite set as the fundamental one and the others as displaced by the addition of ouns .
When such displacements occur in the limit of one line the new one is indicated by writing the displacement on the left .
Thus or .
With satellites , on the other hand , the similar effect is produced in the sequence terms .
In this case it is entered on the right .
Thus or ; .
Displacements of both kinds are very common in spark spectra and put themselves specially in evidence in the succeeding pages .
A normal lin may not only show lines displaced from it , but often it appears to be replaced by them , and , in general , when it does not disappear its intensity is abnormally low .
This is practically what happens in the satellites .
The appear to be the normal lines in which we should expect descending order of intensity ; but most of the energy ( or the majority of the emitting centres ) appears carried over to the more intenss and displaced set ; and , again , most of what should be expected in is carried over to become the strongest line .
Frequently the line has disappeared and only the fragment of the triplet left .
In general , the lines of any element are the strongest of the series .
But in the present vacuum tube spectra ( spark type ) shall find very frequently that the line .
required for is comparatively weak , and in this case there appear other lines related to it by oun displacements chiefly in the limit .
As the real existence of these displaced series is a matter of some importance considerable space has been given in the discussion of the X spectrum ( p. 399 ) to its demonstration in the case of two series depending on the limits .
It seems a peculiarity of these displacement series that a term of one series may be absent but appear in another .
Thus may not be observed , but ) may be and vice versed : The presence of a similar effect DR W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
in the sequent terms of series has bnen referred to above .
One good illustration of double displacements fully established is found in the series , in which the indications are shown for , 2 , 8 .
A knowledge of the laws governing displacements is much to be desired .
Very little is known at present .
Lrnkages .
spectra are distinguished , as a rule , by the presence of well-defined series , depending on single groups of P.S.D.F. type .
In spark and vacuum tube spectra , however , these seem to be weakened , and a very large number of other lines appear which are related to one another by certain constant separations ( links ) to form congeries of linkages each connected to a series line .
These links can be oalculated when the values of , or of , are known .
The evidence for these was given in [ IV .
] .
There appear to be links of several types .
Those already discussed are of two types : ( 1 ) separations between successive double displacements of on either side of or displacements of on either side of or .
Of these , use is confined almost entirely in the present communication to one only of type ( 1 ) and both of type ( 2 ) .
They are , or , .
These links themselves may also be subject to small displacements by having their sources on , say , instead of .
For the present purpose , however , no use can be made of these .
In the prevalence of these separations in a spectrum in excsss of their occurrence from mere chance was exhibited in a series of curves with abscissae separation and ordinatesnumber of occurrences within a given smaU amount on each side .
Such occurrency curves are also given here for the links and for the of Kr in Plate 2 .
The links seem to be a normal accompaniment to series Jines ( often displaced , however , when directly attached to those of low order ) .
A further peculiarity of these linkages is the prevalence of the combination , or They are indicated by writing the letter denoting the link to the left of the line when deducted and to.the right when added .
Thus , in the example below , , or SOunding.\mdash ; In the following pages the unravelling of the complete series of linkages has not been touched upon , but the links have been used for testing the existence of lines outside the observed region , a method we may call sounding .
A link .
thus used may be referred to as a sounder .
In this way it is possible to obtain evidence of the existence , or of the exact value , of a calculated line which lies beyond the region observed .
It may even serve as evidence for the real existence of a line in the observed region too weak to have been observed , for it was shown in [ IV .
] that the link appeared to have a tendency to of the two lines to which it was attached .
The method may be .
example from .
In Kr , .
The value of as formula is , which is in the ultra-violet outside the .
This is within the observed region , and as a corresponding line is found at 4423761 with if is free from error .
an individual case this might be due to a coincidence , but when the with lin after line the cumulative effect becomes convincing .
To see necessary to get at a glance a survey of all the cases , and for this purpose they are exhibited in sets of diagrams in Plate 3 .
These diagrams ako include the observed region in order to show that where the method can be tested it holds .
It may be specially noted how the similar arrangement of sounders holds for the same order in the three lines of the same triplet , and how in certain cases the saem alternative .
Cf .
for example XS , or the main lines of the three paralld sets in X , viz. , or particularly the prevalence of the combinations in the unobserved lines for .
In RaEm these links are too large to be of wide application and in No too small to be of use .
In RaEm the link is 23678 and can reach from the unobserved ultra-red across to the unobserved ultra-violet .
In No the link is 196 , so that its reach is too small to be useful .
As this method of sounding is new and clearly of importance if substantiated , considerable attention has been given to its illustration , but as the details themselves are only necessary for a critical study they have been printed in smaller type and may be omitted on a first reading .
Abnormal triplet \mdash ; It has generally been held since discovery of the satellite systems that the triplet separations for the and series are the same .
The actual measures did not absolutely prove this , in fact , they indicated small differences , but the accuracy was not sufficient to establish a real difference especially as against a natural bias to expect equality .
MEGGERS , *however , has recently placed it beyond doubt that frequently the separations are really different .
In Group I. , for instance , the separations as measured from lines are less than those determined from lines .
In the rare gases also this difference appears quite decisively , but here ( group ) the separations as determined from the lines are , in general , larger than those from S. The key to the explanation is found in the fact that the difference between the two determinations diminishes with increasing order\mdash ; in other words , that the sequent in the same et of satellites is not the same , and that in a large number of cases the value of is the same in both and although themselves are different .
It is found to be completely explained by the displacement of one or two ouns between the sequents of the first or second members of a triplet .
Sometimes it occurs in the third member .
The moe plana .
tion accounts for the fact that the separations are frequently smaller than the 'Bur .
Stand Washington , No. 312 ( 1918 ) .
DR. W. M. IIICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
observed satellite separations .
It also acco nts for the appearance of satelites as shown here and in [ III .
, pp. 389-395 ] in connection with the lines in the alkaline earths .
The matter is considered in detail for Kr ( p. 363 ) and is found to hold for the cases of the other rare gases .
The Atomic Weight.\mdash ; It is clear that an accurate knowledge of a first sequent , or of a sequent which belongs to the satellite involving the multiple , gives the means of determimng the oun to in general a unit in the sixth significant figure .
For these mantissae are usually of the order of magnitude of and are known to six figures .
Hence , if the multiple is known , can itself be determined , and since is a known multiple of the oun ( determined by the multiplet separations ) , the oun is also known to the same degree of accuracy .
Further , as the souroes of determining it are often quite independent they serve as tests of the determinativeness of the oun itself to the same degree of accuracy .
When is considerable , its value is known sufficiently well for it to determine the multiple , and then this exact integer conversely gives the exact value of .
In the cases of A and No , however , the values of are too small to determine uniquely this multiple directly .
The difficulty , however , is surmounted by successively values with increasing accuracy from other considerations until the final test can be applied .
As a fact the No oun is amongst the most accurate found .
Its determination ( p. 461 ) is specially interesting , and indeed is only possible because the material at disposal depends on interferential measures and large accurately known separations .
That of X also is a good determination , and is interesting as depending on a number of quite independent data .
As the oun is proportional to the square of the atomic weight within the limits of error of determination of the latter , it is natural to assume that the relation is exact and that , where is a number between .
If this were sustained it would be possible to obtain with twice the degree of accuracy of the oun and therefore far in advance of any obtainable by chemical methods .
In fact the question is raised as to what is actually understood by the atomic weight .
Does it refer to the mass of the positive nucleus , or to that and all or a portion of the electrons ?
The hope might even be entertained of obtaining by this method some knowledge of the number of electrons partaking in the emission of a line if slight changes in the .
could be found .
For instance , we shall find in these spectra not a single group of , or series as in arc spectra , but several independent groups , viz. , and sequents , depending on different multiples of .
If these gave slightly different values of the oun it could be explained by a transference of electrons .
There is little evidence of such variation , but it might occur , for instance , in the oun as deduced respectively from and .
As depends alone on the measurement of the separation of a triplet it is not susceptible of such exact determination as , and , as a fact , a suspicion sometimes arises that such a slight difference may exist , and that from is somewhat less than from [ III .
, p. 333 ] as also here .
DR. W. M. HICKS : A CRITICAL STUDY OF spmTRAL Suppose the atomic weight is and the number of electrons Then the oun is given by .
If another value depends on electrons , whence which gives the transference .
At present these considerations are only of speculative interest , but a numerical illustration is given below ( p. 381 ) in connection with Kr .
The results obtained in this investigation have given the oun with much greater exactness than any value obtained in [ IIL ] , even than that of Ag .
The value of has been determined [ III .
, p. 404 ] as near with Ag .
I now believe from later work that the true value is closer to this than I thought at that time , but in any case it is far less accurate than the ouns themselves .
While , therefore , we can use the ouns to give extremely accurate values of the ratios of the atomic weight of the gases , the actual values in terms of Ag are not so exact , although more accurate than those obtained by chemical means .
This statement of course depends on the supposition of the exact proportionality of oun and square of atomic weight .
The values of as obtained later are here collected and the atomic weight deduced from them by taking No .
A. Kr .
X. RaEm W. . .
Chemical .
222 to It will be seen that in all cases the spectral determinations are much closer to integral values than the chemical , except in the case of RaEm as estimated from HoNIGSCHMIDT'S value for Ra .
In this case , however , the spectral material is defective .
It is shown from one of the criteria that a value of the oun is just possible but improbable , or almost impossible .
These would give respectively .
It is curious also that from the defective observational work for Ra [ III .
, p. 327 ] the value of from whence is also greater than HoNIGSCHKDT'S and more in accordance with the value obtained by Mme .
CURUE .
The value for the Emanation is , however , much more reliable than the above for Ra .
If , regarded as a whole , the deviations ffim the chemical values ( RaEm excepted ) are greater than chemists allow possible , it DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
would seem that in this case we are not dealing with precisely the same entity in the two cases .
\amp ; ries.\mdash ; There appears to be a remarkably stable triplet series of the apparent in most of the gases , but more especially evident in X , in which element it was ffist noticed .
Not only are the lines strong and present in a large number of orders , they appear , at least in X , to be little susceptible to displacements such as are common in other types .
The separations are 1864 , 829 .
The oocurrency curve for 1864 is shown in Plate 2 , fig. 3 .
In this , in strong contrast to other such curves , it rises to a very high single peak and is practically symmetrical on both sides of the peak .
The similar curve for A is shown in Plate 2 , fig. 5 .
Summation Series.\mdash ; In the investigation of this XF series a quite new type of series was brought to light .
The hitherto recognised series appear as the differences of two terms .
The new one has its wave-numbers of the form .
In other words , where the old series are difference frequencies the new ones are the corre- sponding summation frequencies .
The notation adopted is to write the corresponding terms in Clarendon type .
Thus .
The list of the lines in X is given on p. 385 up to .
For low orders , the lines are in the ultra-violet and have to be sounded for .
Similar summation series coupled with other series also common .
It probably explains also the crowding of separations in spectra like that of Cu in short wave regions far beyond the limit which has always appealed to me as a difficulty .
It is possible that summation series may also exist for the P.S.D. series in all elements , but , as a rule , the limits of these are far larger than the , with the consequence that any P.S.D. lines must lie very far in the ultra-violet , a fact which explains why such types if existing have not hitherto been recognised .
The existence of these summation series is thoroughly established and their importance as bearing on theories of the origin of spectral lines is evident .
They would seem difficult to explain on any of the current theories .
But apart from this the existence of the type is of great value for quantitative determinations .
This is fully dealt with on p. 884 and it need not be recapitulated here .
Its importance for this purpose may be realised when it is seen that it forms the starting point in the analysis of the RaEm spectrum , that it settles in a quite definite way a difficulty arising in the evaluation of the oun in Kr , and that it fixes a very accurate value for the limit of the 1864 series in X , thus simultaneously fixing a particular sequent subject only to observation error in one line .
Groups of and Series.\mdash ; Not only do we meet with different groups of series dependin.g on different multiples of but in the case of Kr there appear to be two * As an example , see p. 403 , in X with groups depending on and VOL. COXX.\mdash ; A. sets of lines suitable for \mdash ; in other words , there are sets give different separations with , and consequently different give the same oun , and in connection with them appear two groups satellitss depend one on a multiple of , and the other on a multipk of be suspected that this is only one example of what may be a spark spectra .
The order of presentation is generally that in which the investigation The key was found in the system , a result first rendered the publication of wave-lengths in the ultra-violet by LEWIS ( 1915 ) .
Amongsl them the triplet was found .
XS , AS come next in order of definiteness .
The spectra of RaEm and No are more difficult to deal with , the first because of its defectiveness in range , number of lines , and accuracy , and the latter because of the smallness of its oun and its triplet separations .
After the series of Kr , X , A come the and series of Kr , X , the spectrum of RaEm , the and for , and , lastly , the whole sets for No .
Led by possibly a fake analogy to He [ I. , p. 105 ] , in which doublet series appear in the blue spectrum , the blue spectra were chosen for investi gation , and the family group being of even order triplets were looked for .
In No , with a single spectrum of composite character , the results obtained may have some reference to the red type as well as the blue , especially in connection with certain remarkable constant separations found by WATSON and analogous to the RYDBERG constant separations in the red'spectrum of A. One is inclined to think that these red spectra consist mainly of lines of the .
But the red or first spectra are outside the scope of the present communication .
Although it is a very lengthy one as it stands only the beginning of an analysis has been made .
The im has been to lay the foundation for the series framework of this family of elements , and little beyond has been done .
The linkages , as a whole , have not been isolated , the red spectra not touched upon , and many interesting effects which will require clearing up are passed over without reference .
A great field for investigation is open in these and other spectra for any who are willing give the necessary time and patienoe .
In some few cases the presentation might have been slightly .
shortened by merely stating the final result and showing how the necessary conditions are satisfied .
But not only would this have passed over certain phenomena of special interest , but one of the objects of the present.communication would have been missed , viz. , to illustrate the power of the new facts to guide a search even when the details are most bewildering .
Moreover , the evidence itself is the more striking when developed from step to step than when the result is directly presented as a finished product .
Kwpton.\mdash ; Krypton shows two spectra , without and with capacity , the former in the red region and the latter further towards the blue .
We have measures of some of the stronger lines by RUNG , and a considerable number of weaker lines , not observed by others , by LIVING and DBWAB , although the latter only given to DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES the nearest ngstrom .
The most complete and reliable sets of measurements are by ( red spectrum ; blue 5871\mdash ; 2418 ) , and LEWIS ( blue 2416\mdash ; 2145 ) , both of about the same degree of accuracy with probable error in the neighbourhood of 03 A. Of exact measures there are only two by FABRY and BUISSON for lines at 5870.9172 , 5570.2908 I.A. In the red spectrum RUNG has pointed out constant separations of 945 , to which PAULSONS has added three others .
The observations of LEWIS gave me the first clue to the set of lines and thus formed the starting point for the present communication , although a great deal of preliminary discussion of material for this group of elements had been previously done , especially in connection with the separations for certain linkages .
In the case of Kr a very large number of separations in the neighbourhood of 786 to 788 had been found , connected also with others of 309 , indicating groups of triplets having these values for .
Amongst LEWIS ' lines a set was found with separations in the reverse order , clearly pointing to a first set of or lines and corresponding sets for other orders were then easily found .
It would seem that there are always a considerable number of separations governed by oun displacements in the limits of the first order , and that of these , three seem to be of a more stable value and correspond to normal triplets .
For instance , in all these gases.we find a large number of cases where a or line is followed by a line with a separation very close to They force themselves on attention on account of their value bein so close to the half of a number being sought for , and others may be present although they have not been looked for .
In the present case two alternative sets of lines for the series , one with and the other with appear .
In the original search the former was taken because it is reproduced in the series as well .
But later certain difficulties in the determination of the oun , combined with the fact that the corresponding oun multiple in , although quite definite , is out of step with the march of their values in the other gases , led me to include the second .
This gives a multiple quite in step with the others , and also affords the means of obtainin a good approximation to the oun .
The lines are given in the folowing table , which also embrace a few obtained by sounding , both wave-lengths and wave-numbers are given:\mdash ; 'Trans .
Roy .
Soc , vol. 202 , p. 183 ( 1903 ) .
' Astro .
Journ 43 , p. 67 ( 1915 ) .
'C.R .
, ' March 25 ( 1913 ) .
S 'Kong .
Fys .
Salls .
Hand N.F. 25 , Nr .
12 .
The first three lines give for the formula The calculated wave-numbers for are in order .
The first gives ; the others are outside the observed region but are reached by sounding and give O-C values of , 12 , 22 , the errors including those of the sounders .
Quantities relating to the will be denoted by dashed letters .
With the limit and separations the values of are found to be , , .
The value of the oun is thus given by , with the uncertainty usually found from triplet separations [ see IIL , p. 332 ] .
From the limit and the values of the links are at once found .
The result for is , see displaced sets below .
Is , see p. 376 .
In region at very end of list ; 41$41 is his last .
S ( 6 ) 2291 26 .
Obtained as least square value from , 2 , 3 , supposing equally probable .
DR. .
W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES . .
From this the , can be calculated .
The results are with The occurrency curves for these are shown in Plate 2 ( fig. 1 ) .
It will be seen at once that the maximum occurrences appear with sharp peaks in close agreement with the calculated values .
The calculated lines for are outside the observation region , but their existence can be indicated by using the above links as sounders .
At the same time , in order to give more confidence in the application of the method , the similar linkages are examined for the lines which are observed .
It is , of course , understood that the existence of a single link is no conclusive evidence of the existence of the unseen line , as it may be a coincidence .
The evidence is cumulative and is seen at a glance in the diagram in Plate 3 ( fig. 1 ) .
The actual data are given here .
In the application of this the frequent lin modification must be allowed for , as well as the effect of observation errors .
Consequently separations deviating by not more than two or three units from the values determined above are admitted .
As a fact , this will exclude a number of real link connections as well as inelude a certain number of pseudo ones .
But by limiting the deviations to these small amounts the conclusions drawn will be more reliable .
SOUNDING Data . .
( 2 ) $7340 ( l ) 39799 1044.49 $9284 $18449 38497 ( 3 ) 41166 ( 1 ) Here the links form a series inequality in with 39284 and a corresponding parallel inequality with .
The link with is probably spurious , since 38497 will later be shown to be ) ( p. 371 ) , and the link as shown is excessive .
It is omitted in the diagram .
No links are seen with .
1 ) 21332 29130 ( 5 ) 31726 1994 .
17 ( 2 ) 23276 ( 4 ) 24062 ( 1 ) 30430 ( 3 ) 33613 1940.93 ( 5 ) 24678 26003 2 35654 Here is present a chain of links in so frequent in and Au [ IV .
] .
Also , as in the case of are found the same modification effect oi about 2 in the links .
It may be noted that in addition to DB .
W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
It is important to get evidence which can carry conviction as to the reality of displaced lines , especially where the displaoements occur simultaneously on both limit and sequent .
A numerical coincidence in this case can have little weight by itself In fact , it can have weight only in each particular case provided it is known from other considerations that such displacements are a common and universal rule .
For this purpose it is instructive to adduce here some striking evidence afforded by certain sets of lines connected with the series .
The lines in question are arranged in the two following schemes , expressed in wave-numbers:\mdash ; 41496 41935 41625 26407 ( 1 ) 26442 ( 2 ) 27175 ( 10 ) 26424 ( 1 ) 27207 ( 1 ) 27520 ( 2 ) 26724 ?
( 5 ) 26759 27506 These lines are numerically displaced with reference to the by amounts represented by the following parallel schemes:\mdash ; , , , , , , , , ( 2)(-6 ) , , ( 2)(-6 ) , , , .
In addition , for , we have seen that in place of normal the displaced is seen .
The parallelism in spits of lacunae show that the set are definitely related , and the fact that the same displacement on the sequents for , 2 , 3 are required to represent the observed separations is specially striking , it being .
remembered that a displacement on the limit gives constant separation for different orders , whilst one on a sequent gives different for different orders .
Here , for instance , 252 in , 35 in , and 16 in , all depend on the same oun multiple , , displacement in the sequences .
Also 123 in and 17 in on the same , whilst the constant displacement 800 is explained by on the limit .
DR. W. M. HICKS : A CRITICAL STUDy OF SPECTRAL SERIES .
Xenon.\mdash ; Xenon also shows two spectra , without capacity in the red region and with it , extending far into the ultra-violet .
Practically the only material at disposal is contained in the extensive lists of BALY* ( red spectrum 6198\mdash ; 2536 , blue 6097\mdash ; 2414 ) with an accuracy of about A. This is supplemented by observations by LIVING and DEWAR , especia .
by longer wave-lengths up to 6596 , but unfortunately only measured to the , nearest unit .
BALY draws attention to the large number of lines apparently common to both Kr and X. The number of lines in the whole spectrum is very large .
BALY gives 1376 in the blue spectrum , but perhaps the most noticeable point for our present purpose is the great variability with change in the conditions of excitation .
This is very clearly indicated by a comparison of intensities of corresponding lines as observed by L. , D. and B. The following are a few examples out of a large number illustrating this .
The numbers following a wave-length give the intensities as estimated respectively by BALY and by LIVING and DEWAR:\mdash ; 5068 not seen , 5 10 , 10 As between the two spectra also , a fact noticed by L. and D. is of importance .
They say " " there is one very remarkable change in the xenon spectrum produced by the introduction of a jar into the circuit .
Without the jar the xenon gives twobright green rays at about and , but on putting a jar into the circuit they are replaced by a single , still stronger , line at about 4922 .
In no other case have we noticed a change so striking They also state that changes occur with the same kind of discharge as between different tubes .
These are clear cases of oun displacements .
PAULSON again loc. gives some constant separations in the first spectrum .
The triplet separations observed are about , in due order of magnitude with those for Kr .
No line suitable for comes within the observed region , but there are two lines with W.N. separated by , which would serve for and and are in a similar position to the lines .
They clearly suggest that the lin is at , using as found from the set .
This is further substantiated by employing the value of the lin , found below to be 7314 , as a sounder .
It requires a line at about 34839 , and such a line is found at ( but see discussion under * Phil. Trans , vol. 202 , p. 183 ( 1903 ) .
'Roy .
Soc. Proc vol. 68 , p. 389 ; 'Coll .
Papers , ' p. 494 .
DR. W. M. mCKS : A CRmCAL STUDY OF of examining the lines of the blue spectrum for shown in Plate 2 , fig. 2 .
It is distinguished from those of elements hitherto by showing one definite maximum alone at about , although there are indications of the appearance of another peak beyond 7317 .
The displaced link shows a difference of , so that a second peak might be expected at .
If the actual value is at it would require , or about 30 , distributed between the two , 2 ( 2 ) lines .
This is possible with on each line , but probably excessive for an error on one of them .
Both values are tested as link.s below for the observed lines , and the results show that with the exception of the value of , calculated from the original , is extraordinarily exact .
For this reason , and because exact position of the peak of the frequency curve depends on several disturbing conditions , the original value will be used for sounding purposes on lines outside observed regions .
Again the first line and the limit 51025 gives 93178 as the value of or ) .
From this the values of the ljnks are found .
The complete set are Links .
Changes per displaoement .
The results obtained by sounding are shown at a glance in diagram Plate 3 , which embraces orders up to .
The cumulative weight of the evidence is overwhelming in port of the general application of this method .
The existence of a series paralel to the normal at a distance is proved , whilst the presence of other linked lines is rendered extremely probable by succession of similar linking in the same set , and in neighbouring orders .
Compare , for instance , the triplets in , 3 , 4 , 7 , 8 and the sets for , 9 , 10 .
Detailed Discussion.\mdash ; In the following discussion the starting point for the consideration of each triplet set is\mdash ; after 3\mdash ; the value of calculated from the series formula obtained above .
The sounders are indicated to the left of each observed line and the values O-C in are given on the right , the observed or line being regarded as the observed sounded linethe link as given above .
The value entered in the table of lines ab.ove is , however , not the line as calculated -from the formula , but the most ble value as deduced from sounding .
They are indicated below by asterisks .
For the first three orders the values of obtained by using are placed to the right of those depending on DR. W. M. mCKS : A CRITICAL STUDY OF SPECTRAL SERIES .
the modified with eu .
These and give .
[ 60276 68 ] [ 52053 58 ] * .
[ 50385 80 ] [ 52163 70 ] These higher orders are necessarily close to high orders of the series , and many are therefore apparently common to both .
E.g. , 34216 has been adduced as ) and is also connected with a lin. Also their numbers are now so high that it requires two sounders in series to just reach the limits of the observed region .
The later identifications are therefore all doubtful .
the red or non-condensed spark spectrum about lines between 8015 and 2476 together with 16 lines in the ultra-red have been observed .
For the blue or condensed spark spectrum the number amounts to about 780 between 6682 and 2050 together with another 40lines in the extreme ultra-violet between 1886 and 1333 .
The measures in the red are due to ( ultra-red ) , KAYSER , RUNG and PASCHHlN , and EDER and VALENTA , S and in the blue to EDER and VALENTA , S and ( extreme ultra-violet ) .
In addition we have exact interferential measures in I.A. by for some red lines and measures by BALY for a few extra lines in the blue spectrum .
The red spectrum is noted for the existence of the sets of constant separations discovered by RYDBERG .
The present communication , however , deals chiefly with the blue spectrum .
The search for the series in A is more difficult than in the cases of Kr and X. There are an extremely large number of separations of about the same value but clearly distinct .
They range round 179 to 188 , and , as will be seen later , ments are very common .
The clue is given from the analogous lines for Kr and X. The only strong triplet lines in the corresponding positions are those given in the following list : 'Ann . .
Phys vol. 27 , p. 637 ( 1908 ) .
'Berl .
Ber .
' ( 1896 ) .
p. 651 ; 'Astro .
Journ vol. 4 , p. 1 ( 1896 ) .
'Astro .
Journ vol. 8 .
S 'Denks .
Wien .
Akad vol. 64 , p. 216 ( 1896 ) .
'Astro .
Journ vol. 33 , p. 107 ( 1911 ) .
'Ann . .
Phys vol. 51 , p. 95 ( 1916 ) .
Phil. Trans , vol. 202 , p. 188 ( 1904 ) .
' Astro .
Journ vol. 6 , p. 338 ( 1897 ) .
DE .
W. M. HICKS : A CRITICAL STUDY OF AS .
1 .
423870.7 2 .
( 5 ) ( 2 ) 373988 ( 9 ) 3729450 3 .
( 1 ) ( 2 ) ( 1 ) 75 .
( 4049874 ) 4 .
( 4517930 ) 179.50 5 .
6 .
[ 2049.5 ] [ 4877674 ] ( 47681 ) [ 4903194 ] The observation errors for and 3 .
are very considerable , since the measures are only given to .
and 05 A. produces about .
Consequently it is possible only to obtain approximate values for .
On the other hand , for where we have very accurate measures , there must be some doubt about the allocation of because its intensity , 9 , is so excessive in comparison with the 5 , 2 of and , and the separation of is greater than observation errors on the lines for , 8 allow .
The latter objection , however , can be set aside as it corresponds to the excess observed in Kr and X diffuse sets and , as will be found later , in Net .
In these cases comes out to be normal .
Here , however , the sum is about too large , and with the separations the typical would be at or , probably of intensity 1 , and so overshadowed by the strong line in the list .
As will be seen immediately , the linkages will show that this value is preferable .
The linkages will also show the probability of a line at 40498 for .
The three first lines give the formula For the determination of , the oun and the links , reliable values of and are required .
We have seen that the values obtained from the observed sets of lines are subject to large observational errors .
NevertheJess that .
the true value of is A CRITICAL STUDY OF SPECTRAL SERIES . .
357 shown by is indicated by the fact that there are several of about ( 6 ) 20621 ( 8 ) 20800 ( 4 ) 20980 ( 3 ) 26893 17935 ( 1 ) 27072 ( 1 ) 31359 ( 4 ) 31538 of which the first is part of a linkage .
With the limit and , the values of are , , or The oun is thus given by with some uncertainty owing to inexactness in the observed triplet separations .
Its value calculated direct from the atomic weight should be , the uncertainty being due to the uncertainty in the atomic weight .
The value of from which the links may be calculated .
The links are found to be The examination of the spectrum gives the occurrency curve shown in Plate 2 , fig. 4 .
It shows a very distinct maximum in the region around 720 but little to show the exact position .
The values appear somewhat irregular .
If , e.g. , the ordinate for be drawn , it would be the same as for those at and 721 .
It may be noted that if the maximum is taken at is larger and the oun cakulated from becomes the same as from .
But this is rarely the case in triplets .
As , moreover , the link when applied to the observed lines gives .
better agreement than a link we shall use it for the purpose of sounding .
The results of are exhibited in diagram form in Plate 3 .
The details foUow : .
[ 41922 39 ] 43355 41460 We find in the three sets clear indications of displacements producing separations of 12 or 13 .
As they appear in different such displacements can only arise in the limit .
Two cases occur in , viz. , and .
They land between the lines indicated above , 13 76 and , or the same within error limits .
Now on lines in question are or , 41231 .
That is , repeated from .
09 24665 26098 42 25804 31 ( 3 ) 26248 11 Several interesting points emerge from the above .
( 1 ) The links to 26806 as are all bad .
On the other hand , if the links are regarded as correct , the last three point back respectively to 26803 97 , 06 34 , 03 35 , or a mean .
If is correct and as assumed should be , the same within the various error limits .
This is therefore supporting evidence in favour of the supposition made above that the very strong line ( 9)26806 is not itself , but that it overshadows the true one , which ought to be a very weak one .
( 2 ) The existence of separations in the neighbourhood of 12 is also very marked .
There are lines ( 4 ) at 12 22 above , ( 2 ) 26744 15 at 12 75 above , and ( 3 ) 26098 42 at 12 above the linked line .
Further , 12 , 2 , appear in connection with the different linked systems , showing the existence of sets of lines depending on displacements in .
( 3 ) In the allotted line is 181 64 ahead of instead of the normal .
The majority of the links from are to displaced lines , in which gives a separation of about , so that no direct evidence as to normal value of is directly available , beyond the fact itself that displacement is very prevalent .
The two only direct links , , would refer back to lines at 26730 34 , 27 77 , or a mean of , which is 179 3 ahead of and clearly points to the normal .
In other words , the normal is displaced by , but the linked lines refer back to the undisplaced normal lines .
With the development of our knowledge of the laws of spectral construction , such facts as these may be expected to be of the greatest importance in settling questions of internal constitution of individual vibrating systems . .
39536 ( 1 ) 41612 39420 16 ( 2 ) 40926 $7 02 DB W. M. HICK8 : A CRITICAL STUDY OF SPECTRAL SERIES .
again the links to the line observed near normal are bad , whereas the links treated as good refer back to a lin exactly ahead of .
Further the 12 or 6 separation is again in evidence .
Also the links refer back to lines differing by about 6 from .
[ 4517930 ] [ 45358 ] [ 45434 40 ] Here the calculated lines agree remarkably with the sounded .
The calculated are therefore adopted . .
[ 47501 68 ] [ 47681 [ 47756 78 ] The only links apparent are for , again with the 6 displacement .
These lines are all close to the limit of observed region . .
[ 48776 74 ] [ 48956 34 ] [ 49031 94 ] 47522 The lines are now so far out of the observed region that the links are too small to refer back to well observed regions even if such lines are really existent .
The foregoing discussion of the series affords evidence of the existence of displacements of about 12 ( or 6 ) .
This requires further consideration as affording material on which additional knowledge may be obtained regarding the laws which such displacements follow .
The presence of the same displacements in successive terms of the series points to a modification on the limit\mdash ; either a pure displacement , a linkage effect , or , as the separation is small , possibly the difference of two links .
The further evidence to follow points decisively to the existence of displacement .
Whether they are due to displacements by multiples of the oun on the limit is , of course , not so convincing .
The numerical relations are very closely represented on this hypothesis , but in the case of argon the is so small that it produces in separations of only .
In lines whose wave-numbers lie about 40,000 or greater , this produces changes in of 06 and therefore comparable with observation errors .
In the case of only\mdash ; wave-numbers of order 26700\mdash ; does it produce The measurements here are by KAYSER , whose errors are probably , almost certainly , and a close agreement between calculated and observed lines will give evidence of some weight .
CCXX . .
In cases of the displacement associated with the linked lines have been given above .
They also exist in connection with the lines lves .
Near are and , both of intensity 1 ; the first above and the second below , or a difference of .
Again ( 3 ) 4240145 is above .
Now displacement on produces a separation of , and of one of .
If , therefore , the observation errors are very small the lines in question are respectively , , and .
There is no corresponding , but ( 1 ) is with .
In we find direct displacements on the lines with a whole set of linked lines .
They are indicated in the following table :\mdash ; ( 9 ) 26806 14 ( 4 ) ?
( 3 ) 7500 1 ) 26833 37 ( 3)26123 26285 00 24845 00 The line 26562 has been already adduced .
It is with The lines 26022 , 26098 under and differ by or a normal , and refer back to lines which would give the true triplet separations with the first line 26542 . .
From the two lines ( 1 ) and ( 4 ) are displaoed successively by and , the same separation as in the case of .
They are and , with O-C given by in both casea The first has a modified link and the second another of to observed lines ( 4 ) 409$67 .
The and Series .
It will be found that the rare gases show a large preponderance of sets of lines which have all the characteristics of belonging to satellites of series , or to parallel multiplet series .
other words where , as in other elements , the sequences show two displacements from the main sequence when triplets are in question , the rare gases show a large number of such displacements , both in the and the sequences , many of these being due to large multiples of .
No attempt has been made to determine the whole system of these satellite sets , a problem belonging to an intensive study of individual elements , but a sufficient number have been adduced to prove their existence and to exhibit some of their characteristica DR. W. M. HICKS : A CRITJCAL STUDY OF SPECTRAL SERIES .
\mdash ; The following table contains lines allocated to the series for krypton and discussed in the present communication:\mdash ; It will be clearer to consider first the lines by themselves , and then the satellite lines in each order .
From the first three lines the calculated formula is The limit is within errors of the value found for , but in view of what happens in the cases of X , and RaEm to be considered later , it may be noted that the difference is very close to a displacement in , which would produce a separation of .
[ Note.\mdash ; If , here DR. W. M. HJCKS : A CRITJCAL STUDY OF of .
If , here of from the formula for the succeeding lines are all outside the the are respectively in wave-numbers The detailed discussion immediately following will show the evidence for their existence by sounding .
The existence of a parallel set at a distance is brought to light which show for the above calculated lines values of O-C , 00 , 14 , The agreement is remarkably close and would seem to show that the value of used is very close to its true value , about one or two units in excess .
The results of the discussion are exhibited as a whole in Plate 3 , fig. 4 .
As in the previons cases , the linkings with the observed lines , 2 , 3 are given , in order to show that the method is justified where it can be tested .
1 .
2 .
3 .
( 6 ) 20976 38560 ( 3 ) 44426 4 .
5 .
6 .
[ 47079 52 ] [ 48500 56 ] [ 49349 39 ] 43894 [ 46166 ] 7 .
[ 49896 60 ] 41588 05 In negative links lead to in the red where observation is defective .
LIVING and DEWAR give a line to the nearest unit hich may possibly be a link The line 22914 is far out to be an exact , but its difference is 4 47 , which corresponds to an exact displacement on .
the limit .
In other words , it belongs to the limit as calculated from the series .
In the sounders all show lines in evidence .
The sounder for , 6 would show lines just within the boundary of observation , and the absence of corresponding observed lines is therefore explicable .
the lino 41951 is part of an apparent triplet .
( 1 ) The diminished values of indicate that the seoond and third.lines The value of however sufficiently doee to make the difference due to DR. W. M. : A CRITICAL STUDY OF SPECTRAL SERIES .
41961 would correspond to a , leaving the calculated 47079 as a true , whilst is 47077 38 .
That 47079 exists is also shown by the sounder .
In the table this arrangement is adopted .
The line entered for is the calculated , as it is so close to the deduced .
For , 7 they are the deduced from the sounder .
The sounders for , are indicated in Plate 3 , fig. 4 .
The Abnormal .\mdash ; The separations of the lines suggested for the satellite sets show abnormal values in that they are roughly about 2 greater than for the triplets .
The difference is real and not due to errors of observation , and we shall find a corresponding abnormality in the other elements of the group .
Taking BALY'S maximum error to be , the maximum error in for the lines will range from 21 to 24 , or , say , 45 on a difference of two lines .
All the readings for can therefore be the same within observation errors , but cannot possibly agree with that for the set .
Those for however , , cannot be the same without allowing errors larger than .
If they are to be the same the excessive error is probably in , which is nebulous and would require .
Further , in addition to the lines assigned here to the series , there are a very large number of other lines showing separations of 788 .
The question arises , therefore , as to the origin of this abnormality , and it is important to discuss the various possible sources .
The formula gives so closely the values of the lines for from to that there can be little doubt as to the essential correctness of the allocation .
The limit of the series cannot then be very different from .
( 1 ) Is 788 a real separation\mdash ; i.e. , is it produced by a displacement on by a larger oun multiple\mdash ; in this case of in place of ?
If so the separation would be greater , or 791 instead of 788 , and such an explanation is therefore quite inadmissible .
( 2 ) Is it a link modified by displacement If be as found , i. e. , , will be increased by 09 or to inappreciable change .
To produce a change of 2 in the value of or the limit would have to be , which gives a value above .
But , as a fact , the Iimit found is quite close to .
This explanation is therefore excluded .
( 3 ) Are these displacements on the sequences ?
In the normal case the sequences for a given satellite triplet are the same , but are displaced from one satellite set to another .
Is it possible that the sequences suffer displacement in the same triplet also ?
Take , for example , the sat.ellite set whose first line is 20669 .
The sequent is .
A displacement of on the denominator increases by .
If the displacement is on first line of the triplet it must be , if on the second , and both give practically the same value for the apparent separation in general agreement with the observed value .
There is nothing , however , to show whether the displacement should be on the or on both , as the observed is within our assumed error limit , but it is interesting to note that if the oun for be the same as for the true value of would be very close to the observed .
Iu this connection it is to be DR. W. M. HICKS : A CRITICAL STUDY OF rRAL remembered that in the sequences the oun is not affected by the peculiar tripk .
modification shown by all elements .
In the 20842 set the observed is , but the value of the observed is very nearly normal .
This means that the third line does not suffer displacement , but only the middle one 21631 .
If this explanation is correct , the modifications must diminish with increasing order .
For instance , in , produces a change in separation of 56 , and the new as against observed .
For produces23 , but the possible observation errors .
in n\mdash ; maximum \mdash ; are now so great that the observed separation of is well within the limits of .
So far then as merely numerical agreement goes this explanation would seem very satisfactory , but the changes required are so small that by themselves they can give little confidence .
We shall , however , see later how it explains certain effects in the which depend on the sequents\mdash ; and , how it also explains similar modifications in the other elements of this group .
Meanwhile further evidence in its favour may be obtained from linkage considerations .
Some examples follow .
( Note.\mdash ; The observation errors in the separations should not exceed about ) The mesh ( 1 ) 20714 ( 1)19928 ( 2)21026 ( 4)21813 ( 1)22123 ?
Here with 20714 is normal , abnormal , but .
Our explanation gives .
Thus on the upper set the first and second have the same sequent , whilst the third has .
In the lower set , on the other hand , 20718 has , the same as for the third .
So also in .
In the first set has in second and third lines , in second set it has in second line .
( b ) This is a striking example of persistence of the displacement in linked Jines .
Of the Jines in the second column , the first has the same sequence term as 22603 , the next two have the links , but the sequent displaced .
The line 24178 keeps the same displaced sequent as these last , and therefore has a normal to 23392 , but the abnornlal to 23390 .
So 24487 keeps this same displaced sequent and so shows a normal to 24178 and to 22914 .
In other words , if the first line is denoted by X , the above sets may be denoted by the following scheme:\mdash ; X X DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
( c ) This example involves the links whose values are respectively .
It belongs to the system .
The lines enclosed in are hypothetical and introduced to indicate the transitions .
The relations are indicated by the accompanying schematic arrangement , starting with It may be noted that in this fragment of a linkage all the links involved are .
A similar preponderance of these in linkages connected with lines is a marked feature also in Ag and Au [ IV .
, p. 389 ] .
The conclusion is to be drawn that there is very considerable evidence in favour of ( 3 ) as explainin the origin of these modified separations .
It does not , of course , follow that the effe , ct indicated in ( 1 ) does not exist amongst the lines of a spectrum .
We know also that the effect indicated in ( 2 ) is existent , and we shall find clear evidence of it in the existence of lines depending on limits which are displaced ) , e.g. , in the case of 19928 considered below .
Further support to ( 3 ) is given by the separations ( p. 370 ) .
The \mdash ; Any allocation of satellites which may be regarded as firmly established is a matter of some difficulty on account of the large number of separations which enter indirectly as links , and the prevalence of sets depending on displaced limits and displaced sequences .
The lines in the list for are placed there provisionally and for special discussion , although other lines certainly belong to the system .
Consequently , the denotation is to be regarded simply as a means of referring to the different sets considered in this communication only , the giving the ordinal position starting from 20976 as The criteria that these lines should be possible satellites with the same are that the differences of their mantissae from that of should be multiples of the oun .
Should any belong to another group with limit , their mantissae will be modified , bu the qualification test will still hold with reference to the difference between their mantissae thus modified and that of the original , for they must all belong to sequents and so be oun multiples .
Nor will the test be affected by the existence of the supposed cause of the abnormal triplet of given by both the has been seen to be with .
The maximum observation errors will be taken as A and } actual O-C error be written .
The mantissae for the various lines are then 1 .
20976 2 .
20875 12$-193 .
20871 , 3229 \mdash ; 4 .
20842 , 4109 \mdash ; 5 .
[ 20763 ] , 6542 \mdash ; 6 .
20669 , 9399 \mdash ; 7 .
19928 8 .
19116 The test requires that the expressions to , the right of the first term in the last column should vanish .
The effect of any possible value of is small .
Further , as is probably known within , it is only in the two lines , 7 , 8 , that it can be effective towards satisfying these conditions .
It is clear then that the conditions can be satisfied within possible observation errors by Nos. 4 , 5 , 7 , 8 , and not by Nos. 2 , 3 , 6 , nor by 8 , if requires diminution by as much as 1 .
Considering 1 , 4 , 5 , it may be noticed that the separations of 4 , 5 from 1 are due to and .
Since and , these separations are in the normal triplet satellite separation ratio .
Moreover , the mantissa of the extreme satellite within our present degree of approximation to .
This again satisfies the normal rule .
It is clear , therefore , that these three lines form a normal triplet set .
As to Nos. 7 , 8 , the ratio of separations is not that oftwo satellites to 20976 as , although near it .
The mantissa of ( 8 ) , however , is within our present approximation to , and suggests that it is the extreme satellite of another group .
Returning to the other lines it remains to see if they satisfy a test with a displaced .
Now on produces a change of and consequent changes in the mantissae as follows : in ( 2 ) , ( 3 ) , in ( 6 ) , in ( 8 ) .
As it follows that the test for ( 8 ) is unaltered and those for 2 , 3 , 6 become ; ; ; whereas in ( 8 ) the corresponding condition is not changed , but the mantissa becomes .
In this case a displacement of one oun in the limit produces the same effect as that of two ouns in the sequent .
The test for the others is very closely satisfied by , or the supposition that they belong to a .
W. M. HICXS : A CRITICAL STUDY OF SPECTRAL SERIES .
group with a limit or with a limit , decreasing by the mantissae of the sequents and requiring ; 5- ; ; which again are easily satisfied within observation errors .
With regard , however , to ( 2 ) , ( 3 ) , it is not common to find two satellites so close together , and as they both belong to doublet sets we should suspect that they belong-to different groups , and in fact the condition is satisfied by arding the limit of 20875 as displaced by on that of 20871 .
We shall consider the conditions more fully later in connection with the more accurate determination of the oun .
The preceding results are sufficient to show that ( 1 ) , ( 4 ) , ( 5 ) are a definite normal satellite on the basis of .
That ( 7 ) ( 8 ) belong to a series probably based on , and 2 , 3 , 6 to a parallel system based on the limit or .
I omit details about the satellite links , but the following points are interesting .
The line 19116 affords an example of the two displaced lines analogous to that given in as explaining the source of the abnormal separations .
In addition the linked line shows the same effect with two lines ( 2 ) , ( 1 ) are curious as showing successive abnormal links .
Thus ( 1 ) 18780 ( 2)20669 ( 4)22557 ( 1 ) ( 3 ) ( 3 ) ( 1 ) 2365 29 The two lines which have been shown to have a relative displacement of show links to a displaced from each .
Thus ( 3 ) is 194215 or below their mean ; whilst ( 1 ) is or below the mean of and The satellites for are more doubtful on account of the large changes in the mantissae produced by observation errors .
The line 38531 has a mantissa difference from that of of , exactly the same as in .
It should , however , be one of a doublet with the second line stronger .
The others show mantissae differences of , which , allowing for the usual changes in the second order , clearly point to the sets as indicated in the table .
There appears , however , no analogue in for 38324 .
In this case it is important to notice that the L. CCXX .
separation corresponds to the displaced sequent for .
It thus the explanation adopted for the triplet modification .
For reference the multiples of the oun giving the satellite separations are collected in the following table , where refer to the first and second line in set .
.\mdash ; The lines form parallel series in which the constant separations depend on the satellite separations of the series .
In other words the limits are the sequents which form the satellites .
Now it has been shown above that the abnormal triplet separations which the series exhibit is probably due to the fact that the sequent for a given satellite triplet is not the same for each of the three lines , but that they are subject to a displacement of one or more ouns .
For instance where is 788 in place of 786 ( in round numbers ) the difference 2 is due to the fact that is not equal to but is .
If the strongest line is to be .
taken as normal , we should expect the , or to be normal rather than as is vays weaker than the other lines of a triplet satellite .
In this case and the separation which is less than the observed satellite separation by about 2 .
As a fact we do find these diminished separations .
In searching for lines therefore we have to examine the spectrum for wave-lengths longer than and showing as multiplets with separations the same as the satelite separations or less .
In the particular case of Kr these are 100 , 105 , 133 , 213 , 307 , 1048 , and we are to expect series which we will denote by with from 1 to 8 .
From this point of view it is unfortunate that allowance has to be made for the rule as to the excessive displacements occurring in the lower orders and that a complete multiplet , showing all the above separations , is not to be expected .
With and .
The mantissa of is in general large , say , between 7 and 99 .
Consequently the region in which is to be found is where the wave-number is less than * There may , of course , be others depending on lines than those nsidered in the text .
DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
or say , , or .
This region is examined for the separations in question and lists made .
It is then found that some depend on the same first lin , in which case , they clearly refer to or related lines .
Several sets are found connected by one of the ordinary links , which excludes at least one of them as a direct line .
It is now easy to select a few sets from the lists which seem suitable for the line .
This with the given gives , and then RYDBERG'S tables give a rough approximation to .
It is then only necessary to examine the lines near these for th separations , in order to find the actual .
The result of this examination is to show that for the first three orders , 3 , 4 , the only sets which exist without displacements in any correspond to the separation 301 , with lines ( 1 ) , ( 3 ) .
The formula calculated from these gives as the limit .
This is only29 less than and is thus in very satisfactory agreement with the rule .
Using the value of as found from the series with the actual formula is given by \mdash ; In this if , -27 , but if , The mantissa of the first line is within error limits .
The lines have been selected as showing the given separations .
Quite independently they give a formula with the proper limit and with the first sequence mantissa a multiple of .
The evidence therefore for the correctness of the allocation is incontrovertible .
In the consideration of the notation for the various parallel series it will be necessary to determine what is to be understood by the normal separations .
If the latter are to be decided directly from the lines , the separations must be those given above .
But a glance at the table lines will show that there is a considerable variation from these , and indeed from one another\mdash ; due as we have seen to the great variability in the oun displacements .
Especially is this lloticeable in the separations given by the satellites as and 30727 .
In place of these we find values about 2 , 4 or 6 less , corresponding to 1 , 2 and oun displacements .
The separation 301 is the most frequent displacement from 30727 .
Now alters the separation by , and therefore alters to .
That these deviations from normal values correspond to real separations can be seen by their frequent repetition in connection with lines .
Sse for instance the maps for , especially , in Plate 4 .
Here there is an example of or of appearing as links , omitting as a doubtful connection .
The is the exact satellite separation .
Moreover , the mantissa of 17594 is , differing from that of by .
This is a clear displacement by , for cannot be more than a few units and is a fraction .
It is clear , therefore , that 17594 is and 17695 is .
The next two must therefore be only the linked lines .
There is a lin separated by from 17695 , which might suggest the displacement in the limit .
But is too large , and the line itself is really a linked line to , viz. , ( e ) .
( 2 ) ( 2 ) ( 1 ) ; ( 1 ) ; ( 1 ) ( 1 ) This is a specially interesting displaced set in that it contains all the seven separations , three directly depending on the line 17972 and the remainder on a line linked to it by the link .
The mantissa of 17972 is , which is above that of .
This is very close to .
The lin may be written .
Call this .
Then remembering that produces in the limit and in the sequent , we see at once that in the others , for is exact ; for is ; for is .
The others depend on .
The separation is the mean of which depend on two sequents differing by .
It may therefore be written as either or , we will write it .
For the displacement is in the sequent , or the line is .
For and the line is ( 3 ) ( 1 ) ( 1 ) ( 1 ) ( 1 ) With the mantissa of 19134 is above that of .
It is in the neighbourhood of .
The whole set of Iines representable as follows ( putting as before ) . .
DR. W. M. HICKS : .A CRITICAL STUDY OF SPECTRAL SERIES .
The numerical agreement is very close , and 19956 is a series inequality in which a change of 1 makes the two separations 822 , 308 exact and links .
On the other hand , as we have already pointed out , 20991 may be a line depending on the limit .
The following points should be noticed in the foregoing allocations:\mdash ; ( 1 ) The presence of the large displacements in the sequence term by quantities differing from multiples of by one or two ouns\mdash ; in this respect quite analogous to corresponding effect in the alkaline earths .
It seems to point to a kin of satellite effect in the series analogous to that shown in the , where the main strong line is displaced from the normal satellite depending on a multiple of by the addition of a few ouns .
In this case the satellite is in general too weak to be observed , except possibly the linked lin .
depending on .
( 2 ) That where a multiplet line is absent , it frequently appears as a linked line , but that the linked line never appears directly linked to the ( 3 ) The lin 20991 occurs twice as and as , also its possible existence as a kin of independent line has already been referred to .
It cannot of course be all , and at least two of the suppositions must be due to chance .
It is also probable that such coincidences may occur in some of the other allocations .
The evidence for the general effect is cumulative and not dependent on a single numerical agreement . .
The system of the first order have been considered in rather considerable detail in order to establish what appears to be a very general rule that in many groups of elements the configurations producing the normal lines appear to have been subjected to a sort of explosive effect whereby other configurations producin sequents displaced by large multiples of are produced .
As a natural result the intensities of the normal lines in the spectrum are diminished since the observed intensities must depend on the number of emitting centres as well as the energy emitted by each .
We have seen that they are displaced in pairs or sets containing one displaced line , but no attempt has been made to search for sets not containing the .
As we shall see later these displaced sets in the lowest order give a means of obtainin very accurate data for the determination of the value of the oun .
In dealing with subsequent orders such a detailed discussion is not called for for this purpose . .
The normal lines are observed for , and as , the correctness of the explanation given above for the diminished separation 301 in place of 307 it will be noticed that the normal value is shown by line 23349 .
The other lines of the set do not directly appear , but as in the case of , they are in evidence as linked lines .
Some of the lines linked to , are represented in a map in Plate 5 .
The denotations of the lines are entered in place of the wave-numbers , which can be reproduced by adding the given separations , and each can be referred to by the column and order in the column in which it occurs .
Again we have several examples where small changes ( errors or displacements ) in lines make all the links take their practically exact values , and so give evidence for the reality of each .
Thus , on al give on a2 give and the mean of ( or the lin or , say ) ; on gives on c5 gives .
These new sets give us representatives of all the lines missing from the direct normal lines .
It is noticeable how the sequent persists .
Certain displaced sets are given in the tables .
If 23418 is less ( the become exact and it is .
The numerical proofs of these allocations are not given , as these displacements have no importance at present beyond the fact that they exist . .
Direct lines are found for , 2 , 4 , 6 , whilst 5 appears displaced , to observed lines 26265 , 26275 .
There is also a line 26067 ahead of 26065 by .
If this be due to some displacement it is probably on the limit and some ouns on the sequence , or all by ouns on the sequence .
The order is so high that it is not possible to decide , and it is shown in the tables as a difference .
Linked lines are shown in the map ( Plate 5 ) .
Again note that on makes the links exact , and that here again the appears .
Symmetry would seem to indicate that the true or 26057 should be about less .
This would diminish the calculated limit of the series to a value nearer that given by the calculated .
For the values for calculated from the formula are .
With later , we should expect values less than these from about for the first varying to for the last .
None of these appear but they have linked lines whilst other of the parallel sets also appear directly . .
No line has been observed at 27498 , but there are lines with it for and others for by a lin. A value of is and reduces it as just suggested .
The connections are exhibited in the map ( Plate 4 ) .
From this order and beyond there appears to be a parallel set at a distance 16 units less .
For , this starts from 2748272 as .
As is seen in the map ( c9 ) it has a very large linkage to lines with similar sets to those connected with the calculated .
We may explain its source as a displaced , as the difference of two -links , , or as the direct congeries of lines depending on 20991 as an independent line .
In the map the notation depending on the second is adopted .
In the list 27482 is written as leaving the question of the origin open .
The -links are particularly prevalent .
This was found to be the case also in Ag and Au , the only elements in which the linkages have been examined with any thoroughness .
In particular the series of successive and \mdash ; links from 27497 recalls a similar * This has been given as a bad .
The suggested change makes the link worse , which increases the improbability of its belonging to the system .
The calculated is retained however in the map , as the links show the repetitions more clearly .
DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
arrangement in the linkage shown in the columns of the map for AgPiu [ IV .
] .
The series is in fact continued further than is shown in the present map .
Starting from 26727 we find ( and ) , the actual separations being .
Further , it should be noted that each successive pair is a parallel inequality , one in excess and the other in deficit of normal value .
It means an increased displacement in each alternate line .
But if the observation errors are small , there appear to be indications of simultaneous displacements in the sequences well as on the limit .
In fact a similar phenomenon is indicated in the two next orders though naturally some elements are wanting .
A precisely similar connection is shown by , [ IV .
, p. 382 ] , in a still more striking and regular series of changes .
The elucidation of the laws governing displacements is of the first importance and should be one of the immediate objects of investigation .
For this purpose examples of continuous series of simultaneous and like displacements will be of the utmost value .
For this reason maps of certain near lines ( Plates 4 , 5 ) , are given for all the orders from 3 up to , but no attempt has been made to indicate exact displacements involving unity .
The parallel series about 16 below exist for , 6 , 7 .
The sets connected with all show the displacement unity .
In the lists the true lines are entered as 1 less for and less for 7 , 8 , 9 , 10 ( i.e. , about ) than the calculated values .
As is seen it makes the observed separations more normal and in so far supports the putting of the limit about less .
Later the actual change in the limit is found to be .
During the work of examinin the X spectrum a new type of series , associated with the series , came to light .
Whilst the known type depends on the differences of two sequences , the new type has a series of lines whose frequencies are given by .
We shall denote the lines of these series by so that will denote a difference frequency and a summation .
We have already referred to the general properties of these series in the introduction .
Some of the material from the Kr spectrum bearing on the subject are here collected .
In the following lists each order is considered by itself .
The examination has not been exhaustive so as to involve displaced values , but it is believed all the direct observed lines have been included .
A few abnormal ones , with considerable displacement in the sequence , have also been entered , as they raised questions which require future investigation .
The and lines are arranged in parallel columns .
The mean of the two corresponding lines is entered in thick type between them .
That for the first corresponds to the fundamental limit .
The succeeding ones are given in the form mean of the ffistdifference , and the difference only ( which settles the denomination of the set ) is entered .
Thus for the first mean is that for is and is entered .
Also over each line the difference from or is entered .
Notes on detail are appended below the lists .
The evidence is clear as to the existence of a ssries of the form .
If VOL. CCXX .
DR W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES . .
To the only two direct there appear two direct with the same mean and .
Also there is a direct line to .
To the linked corresponds a linked .
In this connection it must be that all the are large , over 44000 , and an link would reach to lines outside the region of observation .
Three other sets are included in the list , , whioh involve digplaced s6quents .
The second pair give a mean 30674 7710210 and belong therefore to the hmit midway between and .
But the sequent half the difference l339995 in place of 13353 26 , on the supposition of a common limit .
So also the pairs for show sequents 13362 41 and on the supposition .
If the observed 44028 is really , where has the same limit as , and may ) called the normal , the mean of the observed and of is 30677 80 , and should give the true value of the limit subject only to observation errors on the two lines , i.e. , within maximum error of 1 1 with and within a probable error much less .
corresponds to with ] mean 30677 some cases of displaced as in .
The set appear also in , and as they contain several examples they are placed in the list .
There are two lines 37836 36 and 38050 16 ( separation 213 79 ) whioh as and give a mean 30671 71 .
The lines in the list show an unobserved line for , which is the basis for the others , its actual value is taken as displacement on 37836 .
The mean is 30674 76 which , on the supposition adopted abovo , corresponds to .
Since , the line 37946 may be written as normal , giving mean limit .
In the set is a line 39799 , giving with a mean 1860 193067878 . .
There appear no direct to the lines .
But they occur in the parallel set ; but as .
The mean is 30677 16 . .
Here and appear , but as the mean depends only on calculated it is not reliable .
If be taken from the observed line by the link , the line would be , giving mean 30677 25 .
There are also lipes connected with the parallel series which has a limit 16 below and , gives mean 30660 40 , which is about the proper amount below .
With this goes 3414682 as with ation 3 .
The unubserved lines supposed for the first pair are calculated respectively from the observed , and .
The line 33297 is .
Corresponding to 33076 as , the mean limit with is 30658 in place of 30660 .
With this might possibly go 34018 93 as gives a mean limit 30674 71 .
Also with gives mean limit .
Also 33511 22 an exact with mean 30677 34 .
givps mean 30676 36 , but is uncertain . .
I have not found , but 32015 37 as gives , which gives limit .
No found , but 31726 38 as and 31841 05 as gives the same value 31625 5 .
This gives a mean limit The evidence seems therefore clear for the existence of this type of series .
The Value of the Oun.\mdash ; For the evaluation of the oun there are at disposal:\mdash ; ( 1 ) The as determined from the separations .
These have given ( p. 346 ) for a first approximation to , the value from and from the two alternative .
The are so ill-determined that these might possibly refer to values giving the same .
But the fact that the value of calculated from agrees so closely with the maximum ordinate in the corresponding occurrency curve ( Plate 2 , fig. 1 ) shows that must be exceedingly close to the true value , in which case it is 378 DR. W. M. mCKS : improbable that the difference in could be attributed to a single oUrvation error on each of .
As it has been shown in [ III .
, p. 832 ] that the triplet rabo .
always shows a slight difference in the from ; it is probable that the same occurs here also .
The evidence there given to show that the value obtained ffim is always closer to the true value .
We should expect , therefore , a value between the two values above .
( 2 ) The evidence obtained from the qualification test .
( 3 ) The satellites whose mantissae depend on multiples of , viz. , 19116 on , and 20763 on ( 4 ) The mantissa of Before however conditions ( 3 ) , ( 4 ) can be applied it is necessary to obtain if possible a closer approximation to from ( 2 ) .
The material for discussion is that given on ( p. 366 ) .
We shall discuss it on the two bases of and where is certainly not greater than3 .
The complete conditions are , using the displaced values for ( 2 ) , ( 6 ) and omitting ( 3 ) as parallel to ( 2 ) , ( 2 ) ( 4 ) ( 5 ) ( 6 ) ( 7 ) ( 8 ) It is quite clear that the conditions in the first column cannot be satisfied without assuming very large observation errors unless is negative , nor on the right hand column unless is .
positive .
In other words , must be and .
The first four equations , however , give no indications of amount , as the multiples of are not sufficient to make the term in more important than the error terms .
In conditions may be written with } ( 6 ) \mdash ; ( 7 ) ( 8 ) .
Nos. require to be about equal and opposite in the two cases , say , This would make ( S ) give .
This case offers some difficulties which we will consider later .
For a further approximation we will therefore put which give , and .
Then , ( p. 366 ) DR W. M. mCKS : A CRITICAL STUDY OF SPECTRAL SERIES .
Henoe Therefore A. Here depends on an extrapolated value from a line , taken because the showed a dimimished value corresponding to the usual satellite modification of .
If this is correct is the observation error on 21858 and is with .
This is certainly the most a priori probable supposition .
If on the contrary hypothesis is normal is and .
This gives Taking now the case of we have already seen that a change of on the limit , or is the sequent produce the same change , so that is the same as , and the qualification test remains unchanged , although the mantissa is altered by Supposing this displacement to take place the mantissa of is Here again , as in the case of , the question arises whether the value of the normal sequent be taken as that of or which latter is larger .
The condition then for the exact multiple of is The two cases give B. Neither can be easily satisfied with , but with the conditions can be satisfied with , for the two cases respectively .
* The remainders in the two cases are for and for The list of lines also give the f.oUowing which can be used as tests .
( 1 ) ( 6 ) , i.e. , mantissa ( 2 ) ( 5 ) , ( 3 ) ( 2 ) .
* See , however , final order below .
380 DR. W. M. mCKS : The mantissa of ( 1 ) , using is This requires , or combined with the condition for , and can be satisfied within error limits .
The mantissa of ( 2 ) which differs from that of by .
This requires easily satisfied for both cases within observation errors .
The mantissa of ( 3 ) is , differing from that of by 72518 .
With and this becomes and 72582 , or and on their respective values of The amount 23 is perhaps excessive to be covered by the various possible errors but it just comes within .
It may be noted that .
These three data do not decisively distinguish between the two cases .
This , however , is not to be unexpected because the two arise from a displacement in , the sequents in this neighbourhood are such that on the limit and on the sequent are nearly equivalent , and the multiples involved 185 , 189 , 195 are too close to produce contrasts .
Incidentally , also , the discussion strengthens the allocation of the lines to the displacements given .
The only further test with our present knowledge is to obtain some independent evidence as to the exact value of the limit , and naturally we turn for , this to the mean of the and series .
The series however in Kr is not nearly so well developed as in X. As has been already seen there are only three sets of observed pairs and these give for respectively values of , , .
Since a displacement of produces change of in the first may be due to the fact that the ] taken for is really , when the true mean would be .
It is natural to seek further as to the existence of summation lines corresponding to our last three examples .
The result shows a most remarkable agreement .
The sets are shown in the following list together with those obtained from the normal and P. .
F. .
P. ( 5 ) ( 1 ) ( 1 ) ( 2 ) 3067770 ( 1 ) These are remarkably concordant , especially when it is noted that the are diffuse lines and not so susceptible of exact measurement as the others .
The mean DR W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
limit is and may be taken as practically correct .
That calculated from the series , and used in the preceding discussion is .
This , therefore , requires a ction o. The equations A would be satisfied by AB however this value of is probably correct within 1 the best value of is obtainable from , viz. , If the difference between as found from be real and depend on electronic changes as hinted at in the introduction , the changes calculate to electrons In other words , the would refer to mass of nucleus electrons and to mass of nucleus electrons .
Is it merely a curious coincidence that the atomic number of Kr is 38 , that of being 1 ; all the electrons acting in one way for and in the opposite for it being remembered that when lines are emittted one electron at least is absent .
Xenon.\mdash ; The X diffise system appears to be a most complicated one .
AB we shall see later there appears to be a congery of series rging to limits which are coUaterak of , and connected with these there are again congeries of series converging to limits collateral to the various sequents or , say , the normal .
These series further show the existence of other words the sequence is also subject to slight collateral displacements .
This renders their disentanglement a very intricate problem not only in itself , but because it renders the region of the spectrum involved very crowded , with lines close together , with the consequence that coincidences occur which may not refer to real relationship .
In fact there are cases where the calculated values of supposed lines of different series are the same within observation errors .
* This crowding is also increased by the existence of the allied series referred to above .
The complete discussion of all these related should afford valuable material for arriving at a knowledge of displacement laws .
Here it will be sufficient to indicate the nature of the problem and to deal with the material so far as to give confidence in the results as to the assignment of series and especially as to accurate determinations of the oun and the various links .
As vacuum tube spectra approximate to the spark type , the difficulty of drawing definite conclusion from the existence of a triplet separation is again enormously increased by the presence of the lin relations which these spectra show .
In arc spectra the appearance of a or separation may always wiffi some certainty be ascribed to the fact that the ] in question are directly connected with series terms .
A case in point is and in the series next considered ; also and .
Here however no such certain conclusions can be drawn .
They 1nay enbr as linh Although their true connection may ultimately be definitely settled the doubt as to whether they give true or false scents renders the task of unravelhng most bewildering .
IJJ about the region to be.expected we find the set ( 1 ) ( 5 ) ( 10 ) showing from their separations and order of intensities an indubitable satellite set of type .
Below these come a number of indubitable doublets of type , and then a number of strong lines of type .
There are also in this region an extremely large number of ]ines with separations between 1780\mdash ; 1785 corresponding to the enlarged links already found in the spectrum .
This portion of the spectrum is set out together with their separations in the following table:\mdash ; ( 3 ) ( 7 ) ( 1 ) 19785 84 ( 3 ) 19815 84 ( 1 ) 19829 47 ( 1 ) ( 1 ) 19942:53 ( 1 ) ( 1 ) ( 1 ) ( 1 ) ( 7 ) 18266 88 2002912 ( 2 ) ( 4 ) 20080 93 ( 1 ) 20107 ( 3 ) 21413 28 ( 1 ) 21457 48 ( 5 ) 815$0 ( 10 ) 21717 98 ( 6 ) 21769 99 ( 2 ) 23650 23 Before however considering these lines in detail it will be desirable to take here a preliminary discussion which involves a new fact in series , and at .
the .
W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
same time will give some reliable data which have a bearing on the present problem of the actual and series .
The line 20312 and those in its neighbourhood show the following sets of separations , viz. :\mdash ; 1 .
( 1 ) ( 10 ) ( 6 ) 2 .
( 6 ) 17638 55 ( 1 ) 20333 22 3 .
( 4 ) ( 1 ) 1859 .
79 4 .
( 8 ) ( 6 ) 5 .
( 3 ) In the general survey at the commencement of the investigation a large number of separations of an amount near 1864 was noticed .
Suspecting that it indicated the existence of a second type of sharp series , a second smaller separation ( for triplets ) was looked for , analogous to the of the 1778 set .
A further separation was found for a value about The whole , spectrum was searched within the limits .
The result is shown in the occurrency curve of Plate 2 , fig. 3 .
This curve is unique amongst those hitherto observed in its great height above the pure chance line and also in the steepness of its rise and culmination to a single definite peak .
The search brought to light also a very considerable number of long successive chains and of meshe .
( see e.g. , Nos. 2 and 5 above ) of the same amount , proving that 1864 enters not only on its original source , as due to a displacement on some fundamental sequence , but also as a link .
Now the lin of the normal is 187263 and , thus suggesting a possible origin , also a corresponding link for , i.e. , a separation produced by a second displacement is , or with a modified .
These suggest triplets formed by the same on instead of on , in other words , series whose limits are , 546756 , .
But the way in which the separations enter with the suggested lin 20312 indicates that they stand in fundamental relation to it and neither in a linkage relation , nor with the limits named .
For in the latter cases it would throw out of gear the whole relation of 20312 with the set , which some provisional work had seemed to establish .
In this work they were considered as part of the system through 20312 regarded as a line with 18447 , 17615 as sateuites .
In this case the sequence is of the order 51025 .
As against this idea is , of course , the greater intensity of 18447 , the supposed satellite over that of the , and also the fact that no normal * As a fact , however , there are several others depending on oun multiples also present , but which at present we need not deal with .
VOL. CCXX.\mdash ; A. 3 separation occurs with 18447 or with 17615 .
But for the moment .
If they represented a special set , the reappear in a triplet series of the type , and in the reverse order .
From the already excerpted the lines ( 8 ) , ( 10 ) , ( 5 ) appeared to have all the signs ( RyDBERG'S tables ) of belonging to'one The formula calculated from them brought to light a whole long series of observed lines .
The limit found was 30724 , close to the value already found ( 30713 ) as of the order of magnitude to be expected .
This so far supported the supposition of the relation , but there also came to light another result of evident importance in general theory\mdash ; viz. , the series already referred to .
The ordinary form of a series is one in which successive lines obey a formula of the type .
In this case we find series associated with it whose successive lines are given by .
This holds for each of the triplet sets , so that the complete series are given by , , where .
Quite apart from the importance of this fact in the theory of spectral series the phenomenon is of special use in calculating the various constants on which the series depends .
For instance the sum of the wave-numbers of two corresponding lines gives thus determining the values of the limits quite independently of the nature of the series formula used .
Moreover , the displacements which so frequently occur in the and series in the sequence term introduces uncertainties .
This happens in two ways .
First through the modified values in which it is not always possible to say whether the displacement is produced in the or the line .
Secondly because the typical line in any order is often wanting and only appears with a very large displacement of multiples of on the sequence term .
Th.is effect , however , provided it occurs for both sets , does not influence the values of thus determined .
Cases in point are the Kr sets , , given on p. 380 .
In consequence it is possible to determine the separations B-A , C-B independently of satellite or other displacements .
That such sequence displacements occur in these 1864 series is shown by separations which deviate from the no1mal by more than observation errors .
But , further , the difference of two corresponding and .
lines , say , should each give , if as is the normal rule the sequence term is the same for each line of a triplet .
When however\mdash ; as we have seen in Kr , and shall find even more markedly in X\mdash ; there are displacements in for successive lines in a triplet , these differences will not be the same , and the observed separations will vary from the normal values .
For instance , suppose becomes for the second set , and for the third .
The lines are , The values of calculated from the sums are not affected , and the real values of the separations given by , C-B are not affected although the observed separations are and .
In some cases we shall find evidence from close lines with different or y\mdash ; but the results are quite definite .
If , however , in DR W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
the corresponding terms of the and lines the are different , then the value calculated from their sum shows a change from the normal limit .
The effect shows itself at once and the interpretation is less certain .
It is possible that where such an affect appears it may not be real , but due to the existence of the two close lines just referred to , of which one in each set is too faint to have been observed .
Thus if the displacements are instead of finding , the 2nd and 3rd , or the 1st and 4th may not have been observed and we should be led to a wrong conclusion by taking , say , the 1st and 4th as correspondi lines .
There are cases of this kind and also where one only is absent\mdash ; i.e. , we find one close doublet for one of the or lines .
The lines composing the series are given in the table below .
The limit calculated from the first three lines was found to be , the uncertainty being due to supposed maximum observation errors of in each line .
The later discussion of the rule will show that the limit should be very close to with an error probably .
The formula was recalculated with this limit by supposing the three standard lines to be in error by , i.e. , by half their supposed maximum possible errors .
The formulae for the and series then become LIST of and F Lines .
In each order the first line of numbers gives the set , the second the set .
Between these are entered the mean values of the and which give the corresponding limits .
When the values are deduced by methods explained in the notes they are enclosed in , when calculated and not observed in 830 ( 5705 00 ) ?
( 58441 20 ) 1864 50 ( 60305 70 ) 830 ( 8 ) ( 6 ) ( 4 ) 31102 1863 17 32965 33 .
( 10 ) ( 1 ) ( 2 ) ( 1 ) ( 5 ) ( 35083 1866 16 ( 3695O 13 ) 831 54 ( 37781 67 ) ( 6 ) ( 1 ) ( 1 ) ( 1 ) probably .
Or .
DR. W. M. HICKS : A CRITJCAL STUDY OF SPEOT The lines of the series seem to be exceptionally numerous .
The of the examination up to are given in the table and the notes thereto .
There are certain lacunae\mdash ; especially for .
In cases however nding disphced sets are in general observed , and naturally with large values of this effect is more frequent .
In certain cases where a set is absent a parallel set is observed linked to the normal type .
This is the case for instance in The question naturally arises whether ] exist for .
If so the formula gives a triplet with the first line at , far in the ultra-red .
In other spectra these values extrapolated for differ considerably , often by several hundreds , from the correct ones .
We can only conclude that if there are sets based on they must be such that must be in the neighbourhood of 3100 .
he matter can only be settled therefore by other considerations which must depend\mdash ; with our present knowledge at least\mdash ; either on sounding or on the presence of combination lines in the observed region .
The evidence for such a triplet is given below in the notes to the list of lines .
The value of found is corresponding to a wave-length in vacuo of .
The mantissa of with the limit is .
The uncertainty in as found from is too large to settle the exact value of this with so large a multiple as 90 , but the fact as it stands that the mantissa differs from a multiple of by only a few ouns is what is to be expected if the series belongs to the type , and so far certainly supports the more direct evidence given below for the existence of the set depending on .
With the value of found below A glance at the list will show that the separations observed in the second and third orders of are less than the normal values .
This points to a satellite effect .
The values of are which show deficits of , 58 from the true value as indicated by the occurrency curve .
Now a displacement by one oun produces a change of in , and 50 in .
The deviation is then completely explained by the supposition of the existence of the satellite effect depending on The show similar deficits , which may possibly be due to observation errors .
We should expect to find a similar effect ( not necessarily the same multiple ) in the order .
In this order the oun produces a change of For sounders and for link evidence the data have been restricted to Ijnks only .
If we may judge from sxamples of and Au , the and linkages show a preponderance of the lin , and no doubt fuller evidence might have been adduced by using them , but it was necessary to set limits to the work , as as to this communication which is long as it stands .
But as examples we may give some links belonging to the orders 2 , 3 of .
The value of is .
For the lines ( 1 ) , are , above and the real The two lines ( 5 ) , ( 2 ) are respectively displacements of which is ( or c ) above the observed or 187315 above the normal value For we find DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES . ?
318 .
17 ( 4 ) Here , 25891 , 27754 form a series inequality with forms a parallel inequality with 28341 and 28047 with ; 18466 , , 27754 a series inequality with ; and a parallel inequality with As a rule in the list only displaced lines are dealt with and only occasiol)ally linked ones .
The evidence is strong for the existence of possible series with limits 30725 \mdash ; especially 2\mdash ; and that for each order most of energy goes to a line with one of these limits , not necessarily the same for different orders nor for the same triplet set , nor for the same or type .
Notes.\mdash ; A displacement produces on , on , and on . .
The shortest sounder to reach from the calculated to the observed region is , giving values in the neighbour.hood of 17700 .
In this neighbourhood we find a triplet ( No. 2 of the sets on p. 383 ) ( 6 ) ( 3 ) ( 1 ) which in separation and order of intensity corresponds precisely to the values required .
With the normal value of they indicate lines with wave-numbers at .
In connection with the question of satellite effect it may be worth noting that lines in the neighbourhood of the above may be arranged with them thus ( 6 ) 176.38 ( 3 ) 19503 ( 1)20333 12.52 ( 3 ) 25.53 ( 5 ) To complete the set there should be a line at , but it was not observed by BALY .
A sequence displacement in of produces a change of and of of .
These are practically exact and correspond to the diffuse satellite arrangement with this difference , that the triplet appears as a main set in which the first line is strongest .
They depend on displacements , so that the mantissa of the doublet set , , that of 19515 , is exceedingly close to With should go .
This is outside the observed region on the other side , in the ultra-violet , and requires even larger sounders than .
The treble link requires a line at .
There is no line here but ( 1 ) is less by 5 , .
We can aITange lines to go with it as a satellite set , viz. , ( 1 ) ( 2 ) ( 2 ) 25 .
( 3 ) ( 1 ) Here we find the same for and , whilst in we had it only for .
The set with separation 17 show no , and again is numerically an exact oun multiple displacement , viz. , which gives .
The line found close to the above\mdash ; forms part of a chain ( 1 ) ( 3 ) ( 1 ) It may be the representative corresponding to the satellite displacement observed in for , 3 .
Using links and we should find lines at and with the corresponding lines to beyond the observed region .
As a fact , we find lines at and ( 1 ) .
The first corresponding to a displacement should give ) .
The second would differ from the same displacement by3and is therefore inadmissible .
It should be noticed that both the and sounders give lines larger by unity than the expected value .
This may , of course , be due to combined errors in the sounders and observation errors , but the are entered in the list as if the sounders are correct .
As none of the or lines can be observed , their means can have no weight for an accurate determination of the limit . .
The lines are in the ultra-violet , and should be , with the limit chosen , at .
With sounder we find set with a smadl separation corresponding to that for .
It gives and But the set appears to be really . .
A displacement in producss .
The line is not observed but ( 2 ) is practically an exact .
In further illustration of this we find ( 2 ) hnked to it by and 5 61 ( i.e. another ) ahead in the remainder of the triplet ( 1 ) 818 13 .
The lines in question ar6 ) is ( 1 ) 32362 which , as referred to above , is also ( 1 ) close to link .
For or or 29076 .
l6 .
Take the mean 29059 47 as .
Several other lines in this neighbourhood show indications of displacement by multiples of , i.e. hat energy proper to hue gone into a number of colhterals .
None of the direct are but there appears a parallel set pkod ( 1 ) ( 1 ) with ( 1 ) due to a further displaoement .
The lines are entered as due to these .
DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
Both separations are about 1 6 too large and further about the same amount too small .
Now on the sequence term produces a change of .
This would indicate that with the displacement on the limit concomitant displacements of occur in the sequence so that the observed lines are , .
The whole set again is remarkable connection with parallel sets separated by the normal .
There are observed lines for is entered as depending on a link .
F2 is too large by about one unit , but is normal .
This order affords good evidence for the existence of the displaced sets .
Consider the following system of lines:\mdash ; ( 1 ) 2767652 30370 68 ( 5 ) ( 1 ) ( 1 ) ( 3 ) ( 5 ) ( 2 ) Here 1 84 is an exact displacement in the sequent .
The mesh shows series inequalities with the normal values .
The two lines in question are clearly displacements on a normal line .
On , gives , and ; on gives and .
These show how closely all the condilions of the allocations are satisfied .
Further , it shows how 29561 has its excess value and that the normal sequent should be the same as in ( 8 ) lsss , thus making its separation with l86461 .
Th ; same sequent change is shown by .
Both give it as .
Again with an even order the lines do not appear , but there are apparent also a congery of displaced lines analogous to that in .
The lines given in the list give wavs-numbers 28498 61 , , 31191 93 .
On the contrary , lines are observed although has probably been displaced .
These also show evidence of displaced sets , e.g. , 32972 50 1864 28 34836 78 is 20 53 ahead of , and on the limit gives 19 88 . .
The values of , as deduced from the means are clearly too small .
is ve close to the calculated value , so that if any error has been made it is probably due to the which should be about 1 , and suggests a close doublet , i.e. a small sequence displacement as in the preceding sets .
As supporting this there is a line ( 1 ) 32426 .
15 which as would give 32431 12 making with the limit 30725 22 .
This corresponds clearly to the normal value .
A similar displacement is also found in in the line ( 4 ) 29024 43 , which as gives 29019 46 for .
It should be noted that the energy of has passed chiefly to the displaced line , whilst in most of its energy remains with it and a fraction passes to the displaced line .
This probably means that only a small number of the normal configurations are broken up , whilst most of the are .
as ( 6 ) 31717 13 gives too large .
This line and ( 4 ) 31705 47 are separated by ll66 or a displacement , so that there is a concomitant sequence displacement .
A similar effect is shown in with two lines , ( 2 ) 35126 05 .
The lines entered appear correct for they give the normal limit , but their half difference shows a displacement in the sequent .
The normal line would appear to be given by making with .
No line is found for the calculated , or 77 23 if we allow the same O-C as for .
The lines ( 3 ) 2936841 as ) and ( 1 ) 29403.29 as give respectively 29378 34 and , which are larger than should be expected .
The calculated value has been taken as correct .
Also the lines ( 1 ) 32048 92 and ( 4 ) 32098 20 as give respectively 32073 76 and 73 36 or VOL. CCXX .
a mean of 73 56 .
But 32048 is also .
This is not a mere list coincidence .
are very nearly equal , and if both existed would show as a double line too resolved .
The second line has a separation 1864 23 to ( 2 ) , and its doduced calculated the limit 30725 18 very close to the definitive value found below .
supported by F2 33948 71 giving F2 , but the F2 is large and 31242 hows a separation with of 1865 39 also large .
Also ( as gives which makes and the separation from l86464 both improved .
On the other hand gives precisely the small differences depend partly on observation errors and sequence or satellite displacements .
In the case of and the equally and oppositely displaced lines give the same mean as the lines calculated from them . .
There seem considerable displacements in the sequences here .
The calculated values for the first lines are 29632 80 and 31817 72 .
They are not observed , but the corresponding lines are .
There are two near observed lines ( 1 ) 29629 10 and ( 2 ) 31820 18 which give the mean 30724 64 , which is small , but ( 1 ) 31810 would give 31820 79 and the limit 30724 94 , close to the normal value .
31820 18 is then .
With ( 5 ) 31483 37 as and as we get respectively and 94 63 .
The mean entered , and a similar displacement gives as entered .
The normal third lines are observed .
Probably the having the same sequent as the adopted should be that given by 34527 64 or with .
The allocation seems satisfactory .
The limits also are very close to the correct , but the different triplet separations show that the successive sequents suffer displacement , but the same in each , F. .
The calculated is 29966 21 .
With ( 1 ) 29982 .
Moreover the last has links to ( 3 ) 22652 88 and to in very striking agreement .
The value as calculated with normal is entered .
With the lines as entered it is seen that the msans of the corresponding separations for the two series are both normal , although the individuals are abnormal .
This shows that both corresponding lines have the same limit , and the same sequent , but that the latter shows a displacement from the normal value for the set .
This is supported also by the fact that there are a number of close lines to .
For instance , 33332 22 and ( 2 ) as give respectively 33348 51 and 6 29 for .
They are probably all lines showing sequence displace ments .
The first gives the triplet separations ) , 828 42 and limit 32589 77 , the second 1862 92 , 830 64 and limit 32588 66 .
In other words , the first gives , with same sequent as in and , the second as in .
Note the good agreement\mdash ; the same displaced limit for and same for and .
The two displaced sets give .
The calculated to 30 .
It is remarkable how the series seems to persist to high order .
It may be said that this is only apparently so , because in this region the spectrum is so crowded with lines that it is necessarily possible to select sets near the calculated values .
But in truth the reason of the crowding is because of the series .
The and lines crowd up together on either side of the three limits , and at the same tims there are different sets of limits depending on the displacements .
The spectrum has not been examined beyond , and from 14 to 30 the list indicates an allocation without further specffication .
There is , however , much evidence not adduced to actual cases where sequence displacement occurs .
The calculated values for from to 30 , 30345 73 , , 30421 l7 , 3045068 , 3047620 , 30498 14 , 3051736 , 3053418 , 30549 11 , 30562 32 , 3067408 , 3058461 , 3059407 , 3060260 , 3061032 .
The deviation from the calculated values for ) and , which , however , gives the correct limit , shows that the sequent receives a large meant value , so large indeed as to totally alter its mantissa .
The must be doubtful .
The whole set DR. W. M. mCKS : A CRITICAL TUDY OF SPECTRAL SERIES .
are more likely to be and .
In fact , in this neighbourhood , the difference of two successive orders in a series is comparable with the change produced by a displacement in the limit , and so introduces some uncertainty in allocation .
It will be noticed also that , in a few cases , the same line is adduced to fit two cases , which can only happen if a line happens to be a close doublet , an unlikely supposition to happen often .
We are now in a position to determine the limits with considerable accuracy .
Taking the average means where they are deduced from actual o- displaced actual values , we find the limits come to .
In the first attack on the problem values of unobserved lines were deduced from observed linked lines .
The corresponding mean values for the limits then found had for the last digits , very close to those determined from the displacements .
The individual deviations from the mean are quite small for , considerably smaller than for the others .
It is , therefore , the more reliable .
The mean deviation in magnitude is 28 and the maximum is in .
We may take it therefore that the .
true value of the limit is within a few decimals .
The separations given by the deduced limits above are and These very accurate values afford a means of testing as to their source .
If the limit were known to be a single number there could be no doubt as to its belonging to the sequences , or as to the series being of the type .
But there is just the possibility that it may be a composite number , comprising one or more links\mdash ; say , or terms\mdash ; and that the separations may be due to oun displacements in one of them .
The suspicion that this may be the case is aroused by the fact that the triplet 17615 , 18447 , 20312 , which would be the origin of the or term , and in which therefore the first two lines should behave as satellites do not show complete sets with the separation 1778 , 815 , as they should do if normal satellites .
Moreover , the intensity order with the middle line much more intense than the other is not normal .
There is no test for the composite nature of 30725 , but if it be really so , the most probable source would be , or some near collateral of this .
We will therefore test this as , where may be considerable , so as to include near collaterals , and also test 30725 as a sequent .
We will take the latter first .
At the start it may be noted that it is an argument in favour of 30725 being directly the source , that displacements by small multiples of the oun have fitted in so remarkably closely and frequently in the formation of the list of lines above .
Taking then the limits as the denominators are found to be In these cannot be ter than a few decimals and will produce no effect on the DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
the sets depending on .
In case , however , of being small and the mantissa involving the term in , we might expect still to find lines depending on the , as the presence of the suggests satellites .
To get , requires a displacement which decreases by .
In other words lines with wave-numbers 1702larger for and less for P. We do not find this completely , but the following sets are observed , already given in the notes to the list under ( 6 ) ( 3 ) ( 1 ) 12.52 ( 3 ) 25.53 ( 5 ) ( 1 ) 5.08 ( 1 ) 17 .
17.66 ( 2 ) ( 2 ) 25 .
26.31 ( 3 ) ( 1 ) 38337.09 In which permissible observation errors are .
As has been seen the 36493 corresponds to a displacement in the limit .
The others to , and in the sequent .
The lines 38345 , 39175 consequently have their sequent mantissa exactly Further it was found that is .
The next preceding line to this is ( 1 ) or behind it , again showing the required displacement and having the mantissa .
If it be granted that the series is of the type , the limit must be a -sequent .
Consequently the mantissa of must be a multiple of the , oun .
Its mantissa is with great exactness .
Let the true value of be .
Then if the relation is exact or .
Now we know that must be a small fraction , certainly .
Hence must lie between and .
We should , therefore , expect this value for except possibly where electronic changes of atomic weight came in , as has been suggested above .
If then the 1864 separations depend on exact and we get as closer approximations or and or , in other words when the hmit is 30725 .
When this limit is displaced by these change by 396 DR. W. M. mCKS : A CRITICAL STUDY OF Further , the conditions for require .
, or If the sounder was exactly normal this must be due to observation errors in 17688 of , but the value of is also subject to some small uncertainty .
In any case the result shows that the reference line does not depend on a displaced 30725 , for if so would be at least Returning now to the discussion of the series let us consider first this second group of clearly analogous series of lines:\mdash ; 1 .
( 1 ) ( 10 ) .
( 10 ) 2 .
( 1 ) ? .
( 2 ) ( 10 ) ( 10 ) They all , with the doubtful exception of 4 , 7 have the appearance of belonging to first , or doublet , satellite sets , in which the second line is always the stronger .
The 1780 separations are clearly associated with the now well recognised mid-triplet abnormality .
That it is not itself a normal separation is indicated by No. 3 in which the also occurs .
In ( 1 ) the separation is .
It differs from the displaced by 29 which is within error limits .
In this case the limit would be which is less than and .
With this limit the mantissa of 19942 comes to or within error limits .
This is the typical form for the second satellite set of a triplet series , but modified by the displacement , so common in this group of elements , though here it appears in an apparently ffist satellite set instead of the second .
We note at present that taking account of the small corrections , and writing as before its true value is .
The observation error is in this region and is probably In ( 2 ) 19989 is above ( 1 ) .
The change due to the displacement in the limit is , whilst in the sequent gives suggesting that the limit of ( 2 ) is the normal , with sequent .
With this limit the mantissa is found to be , or with small corrections Ox. In ( 3 ) we have the modified 178371 with the clearly real separation or .
Now the displacement due to on the is which practically exact .
This gives a limit 3 88larger or 5105717 , and the mantissa becomes 879855 , the same as for ( 2 ) and .
The line is behind the other .
A displacement of in the sequent term produces .
Thus 20017 is very close to a line with mantissa , but the difference 85 , corresponding to , is too great to render the relation exact .
Nos. 4 to 7 , although much further towards the violet should not be put aside .
They all show the exceptional separations .
DR W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
In ( 4 ) the separation is .
If it corresponds to a real the limit will be which reduces by and makes the limit about 212 less .
The mantissa would be .
This difference ( 71 ) from an exact multiple of shows this to be impossible .
The discussion of the 1864 series has definitely shown it to be of the type , and has given the limit within very small errors .
This limit is one of the sequents of the diffuse series .
Its mantissa was found to be or .
It gives one firm starting point for the discovery of the series .
The results just obtained indicate the lines which through their dependence on multiples of give the origin term of the diffuse sequence .
They point , as we have seen , to the existence of several displaced , or parallel , sets of diffuse series .
It is possible to show definitely that these exist , even if there be some uncertainty as to the lines occupying the position of .
In the normal case with a single diffise series , the is always the strongest line of the series .
Also in the normal type the limit is the same as that of S\mdash ; here .
When however displacement occurs , the energy of a single lin is dispersed amongst several others , and aline corresponding to the normal may be weak , or even too faint to have been observed .
As a matter of experience also it is found that the lines of low order are subject to these displacements in a much greater degree than those for higher orders of .
Now there are a number oflines , which by their position and absence of separations to stronger lines have the appearance of being lines .
If they are , their mantissae must differ from multiples of ( in the present case ) by multiples of the oun .
The fact that they may do so does not of course prove that they are lines .
If they do not do so it proves that they are not .
They may however in the latter case belong to a displaced series , satisfying the multiple law when the proper displaced limit is employed .
This gives us a method of testing as to what displacement a given line may correspond .
If our calculus were already fully established the next step would be to apply this test to the above lines .
But in reality we are testing our calculus to see if it can be rmly established , and our immediate aim must be to obtain independent evidence for the existence of parallel series .
For this immediate purpose it will only be necessary to apply the test to two lines , the general question being postponed for the present .
In the first attempt at arranging the series the strong lines ( 8 ) and ( 10 ) were taken .
, 2 , and the formula calculated with the limit .
As be seen immediately , this gave satisfactory agreement with sounded observed ] up to ; and as a matter of fact this series was used.to test for the parallel sets displaced on either side of it .
Now the formula constants for a set only vary slightly if the wave-number of the lin chosen for is changed considerably .
This therefore did not prove definitely that 20559 is the correct , and as a fact it does not satisfy the multiple test .
Its mantissa is 897337-3 14 .
That of the line 19989 , which is shown above to be the origin of the normal set is 879853-3 29 .
The difference is .
This is as far .998 DR. W. M. mCKS : A CRITICAL STUDY OF as it can be from being a true multiple , and it is quite any observation errors in the two lines .
For instance our can only change the mantissa by 6 .
If the test is valid therefore 20559 .
Yet it has all the appearance of such a line .
Is it a displaced one it corresponds to .
Now a displacement of on the limit changes its value by , so that .
If denote the observation error to the maximum permissible , i.e. , 20559 with the new limit is 19989 is .
Also 20559 is in which is small ( about ) .
Also from the consideration of 19989 above Eliminating in which and cannot exceed about 2 .
It is not possible to satisfy this with or With If , then , 20559 be a lin it belongs to one of , and is definitely excluded as a possible normal The next lin of higher frequency is the weak line ( 1 ) .
Its mantissa is within error limits .
This therefore passes the test .
If it is the actual its weak intensity is due to the numerous displacements for .
If it bs taken as with the previously mentioned ( 10 ) , and the limit , the series formula is found to be The lines after 2 lie in the violet outside the observed region .
To test them therefore recourse must be had to sounding , only the e.u.v. links have been used for this purpose .
The results are given in the middle column of the subjoined table and exhibited in diagram ( Plate 3 ) .
Details are given in the notes following the table .
Lines were calculated down to and tested .
The result may be regarded conclusive in establishing the series , as well as increasing confidence in the method of sounding\mdash ; a confidence which reposes not on a single coincidence , but on the recurrence of a large number of sucoessive ones .
be seen the agreement 406 DR. W. M. HICK8 A CRITICAL TUDf OF The mean of the two first separations is or parallel inequality .
The inequality is due to two 37159 is exactly extra on the calculated 87174 , The first two orders of the two are which the satellite separations are given as The group .
( 2 ) 21400 96 22210 48 [ 1601$ 46 ] ( 1 ) 17791 36 ( 8 ) 18607 47S 19Sg ( 1 ) 21769 99 ( S ) ( 1 ) ( 1 ) ( 1 ) 38129 37 3990758 ( 1 ) 40717 94 ( 1 ) 37169 39 44S 44S ( 1 ) $8940 39754 48 ( 1 ) 38199 58 ( 1 ) 19386 98 183S ( 10 ) ( 1 ) Without dealing with the whole of the material at disposal we will iUustrate its application by considering in more detail the portion of the spectrum given on p. 382 in which the majority of the lines undoubtedly belong to systems .
is to noticed that the effectiveness of the method in the present case depends on the faets , ( 1 ) that the observation errors do not exceed , and ( 2 ) that with it is consequently possible to determine the values of the mantissae to within 6 units in the sixth significant figures , whilst a displacement of one oun in the sequent produces a change in of the order , or twenty-four times the maximum observation error .
The limit 51025 being supposed displaced by becomes .
The mantissae of the sequences are then .
calculated with this limit , and expressed in terms of and where , and is the ratio of the observation error to the maximum .
The series more fully discussed above definitely taken as depending on the limit .
In other words the mantissa of 19989 is exactly which condition requires , writing for its and gives a relation between and term in in each mantissa is then eliminated by means of it .
There can be little the allooation of 19889 ,

