]\gt ; [ 323 ] VIII .
A Critical Study of Spectral Series.\mdash ; Part III .
The Atomic Weight Term and its Import in the Constitution of Spectra .
W. .
HICKS , F.R. June 7 , \mdash ; Read June 26 , 1913 .
CONTENTS .
Pags eviations . . . . .
323 The " " oun\ldquo ; . . . . .
324 Collaterals . . . . .
336 The diffuse series\mdash ; General . . . . .
338 Table of denominator and satellite differences . . . . .
341 Satellite separations Order differences 353 The term 372 The SandPseries .
377 The series . . . . .
379 Closcr approximation to the vnlue of the oun . . .
400 Ag and Au . . . . .
403 Snmmary . . . . .
407 Appendix I.\mdash ; Sc . . . . .
408 II.\mdash ; Wave-lengths and notes . . . . .
411 ABBREVIATIONS .
[ I. ] and [ II .
] denote the two previous parts of this discussion published respectively in the Trans , vol. 210 ( 1910 ) , and vol. 212 ( 1912 ) .
The formula for a lin is is the limit or vaIue when denotes the correction to be added to any limit adopted to give the true value .
is referred to as the part ( variable ) .
is referred to as the denominator of the line .
" " Separation\ldquo ; of two lines is the difference of their wave numbers .
" " Difference\ldquo ; of two lines is the difference of their denominators .
' ' Mantissa\ldquo ; is the decimal part of the denominator .
denotes the separation of two linss of a doublet .
is used for the denominator difference of the two which produces are ilar quantities for triplets .
2 Published separately , October 22 , 1913 .
values of for the various elements will also be roughly proportional to the squares of the atomic weights .
For this reason it is convenient to refer to them as the ' atomic weight terms .
We shall denote them by in the case of doublets and ] and in the case of triplets , using as before to denote the separations .
Two questions naturally arise .
On the one hand what is the real relation between them and the atomic weights , and on the other what relation have they to the constitution of the spectra themselves The present communication is an attempt to throw some light on both these problems .
The Dependence of the Atomic Weight Term on the Atomic Weight .
DR. W. M. HICKS : A CRITICAL STUDy OF SPECTRAL SERIES .
Hg 19 , and , further , that in many cases and were also themselves multiples of the same number .
As , however , Mg with a difference of 450 and of 543 could not possibly be brought into line with the others , this line of attack was given up .
But later the case of Zn , which at first had seemed to stand in the way of an explanation on these lines , gave cause for encouragement .
The series for Zn are well defined , the measures good , and the formulae reproduce the lines with great accuracy .
Great confidence can thus be put in the values for and , and it was noticed that they were both extremely exact multiples of the difference .
In fact , the values are and .
This relation could hardly be due to mere chance , especially when it was also noticed that is very close to 3/ 2 the former 360 , and , further , the 450 of Mg is about 5/ 4 the same .
In other words , with the rough values used , and .
This looked so promising that a systematic discussion of all the data at disposal with limits of possible variation was undertaken .
The theory to be tested then is that the of any element which give its doublet or triplet separations are multiples of a quantity proportional to the square of the atomic weight .
We will denote this by It will be convenient , in general , to deal with the 360 quantity , and will be used to denote this .
If other multiples are dealt with as units a subscript unit will be used giving the multiple of the 90 .
Thus denotes the smallest , .
and so on .
The results are given in Table I. below .
The value of is obtainable as the difference of two decimals with six significant figures .
It is convenient therefore to tabulate the values of 10 .
The exactness of the calculated value depends on ( 1 ) the correctness of the adopted value of , ( 2 ) the exactness with which is measured , and ( 3 ) , when expressed in terms of the exactness of the value of or the atomic weight .
In the case of the latter the value is used and the values of are abulated .
The method adopted may best be seen by taking an actual example , say that of calcium .
The values of as found by least squares from the -series are .
The value of as given in Table I. of [ II .
] is , and the correct value is supposed to be larger .
The numbers are then thrown into the form , and the denominators are , giving for differences , which are tabulated as 2791 , 1369 .
The corrections for the error are found to be and .
Moreover , the last digits of may be 5 wrong and the value of the be out .
In cases where the are known to three decimal places , the calculations are carried out with 9-figure logarithms , and the values of determined without this ambiguity .
The values of may be wrong by , i.e. , , \amp ; c. This will produce a variation in of \mdash ; in ge1leral is a fraction .
Thus * See Table I. of Part II .
Table I. gives the values for those elements in which the series have been established .
The second column contains the atomic weight as given by BRAUNER , except for the few belonging to volumes of ABEGG'S 'Handbuch ' not yet published , with estimated possible error beneath .
In the third column the-top number gives and the second .
For triplets there are therefore two sets .
The fourth column gives , and the multiples of which it is composed .
In general the 360 ratio is taken , but in several cases it is necessary to take or 180 and .
The second line of columns 5 , 6 , 7 , 8 gives the coefficient of the error corrections to be applied to this number 360 , or \amp ; c. , as the case may be , and the upper line of figures gives the maximum errors estimated , which have , in general , been less than those permissible by the observations .
The last column gives the difference between and the factor given in the fourth column , except that when it is not the 4 term it is brought up to it by multiplying by 2 if it is 180 and 4 if 90 .
The maximum errors are also attached .
In many cases it will be seen that the number of significant figures in the calculated numbers is larger than in the data from which they are derived .
In these cases the number of significant figures in the data must be supposed to be .
made up to the proper number by the addition of zeros .
This enables new calculated values To be determined when the data are improved without the trouble of recalculation .
TABLE I.\mdash ; Evaluation of and of Notation .
atomic weight ; with atomic weight ) , error in , error in , error in on the fourth sigIlificant figure .
series , require in several cases multiples of or .
It is curious that the groups which ffist indicated this relation do not show it so markedly and with so little doubt as the doublet series , in which by themselves it would probably never have been noticed .
There seems to be some kind of displacement with the middle lines of the triplets .
If , for consideration , the values of be taken , this irregularity disappears , and , moreover , with the larger observed quantities , the proportional errors will be less .
If we agree to look upon the 861 as the normal type , and for numerical comparison multiply the 90 by 4 and the 180 by 2 , and , if further , the results are supposed to be weighted by the estimated limits of variation assigned in the last column of the table , the method of least squares gives for the value of Group I. II .
, III .
, All three groups DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
possibility it may be well to determine the amount of such error required to bring , say , the value to , and to , as it is probable the true value of the ratio lies between and .
The former requires an increase in atomic weight of , and the latter a decrease of of the accepted values .
The following would be the changes in atomic weight required:\mdash ; Ca. .
Ba .
Mg .
Zn .
Cd .
Hg .
; According to the estimates of accuracy given by BRAUNER the changes for Mg and Ca are quite impossible , for Zn just possible , and for the others possible .
In the case of Mg and Ca , however , small errors in are considerable proportional errors and the deviations may be caused by these .
It is necessary to have these estimates before us .
Notwithstanding them , the close agreement of the numbers in each set , and the difference between the two sets must produce the conviction that the differences are real , and are not due to errors either in the spectroscopic measurements or the atomic weight determinations .
In the table the multiples given are those which give the oun most closely .
An inspection , however , shows that in each element there is some disturbing influence affecting the and in opposite directions .
Moreover , the sum of the multiples chosen are in certain cases not the multiple taken for , and this should clearly be so .
This happens in Cd , Eu , and Hg .
There is apparent a general rule that is too small and is too large , the deviation increasing with the atomic weight .
The discrepancy is equivalent to a transference from the true to the true Evidently the transfer in Cd , Eu : and Hg has been so large as to increase by more than , so that the closest multiple now appears to be too large by unity .
If the multiples in be diminished by unity , the sum is equal to that for , and the discrepancy between the ouns from to increases in a regular order .
similar change has occurred in Sr , only here while the multiple of has apparently increased , that of has apparently decreased .
If the ratio be taken as 79 : 171 in place of 80 : 170 the discrepancy again falls into order with the others .
With these changed ratios the values become Sr . . . . .
Cd . . . . .
Eu . . . . .
Hg . . . . .
This transference must take place in the term .
The values are given in the following table in which the first column gives ths value of , the second the value of the transference , the fourth the transference in to , the fifth the new value of , and the third and sixth are as explained later :\mdash ; What is the nature of the modification Perhaps the simplest explanation to test is that a fraction of is transferred .
The third column gives the fraction of which is equal to the transfer .
It is noticed at once that the two groups fall into two separate sets .
With the exception of Sr , the fraction in the first is about17 .
Mg and Ca can both fit in with this , for the values are so small that they depend on decimals in the value of , and therefore beyond our significant figures .
In Ca , indeed , evidence is given later that is somewhat higher and would bring the ratio close to 18 .
But Sr is quite out of step with the others .
has no transfer , Cd and Eu are equal , but Hg is 5030 .
If the ratio in Hg be taken to be 82 : 243 however , the fraction agrees with those of Cd and Eu .
The oun is then in place of and closer to the mean value , and as will be shown later there is evidence for the new value of ( see p. 397 ) .
If this explanation is valid it must be possible to bring Sr into the scheme with a transfer of , but it is difficult to see how this can be done .
639 is about four times too great , in other words , where the others are modified by a fraction of , Sr is modified by the same fraction of .
The above arrangement brings Mg into the Ca group and upsets the law whereby its first should be a multiple of .
As this law seems to have a considerable weight of evidence in its favour , and moreover , as will be seen shortly , Mg tends to go spectroscopically with the group , it may be well to see the result of keeping and the ratio .
This will require a transfer of about with a considerable uncertainty owing to the small values of and , and .
With this the fraction of is .
To bring to the same fraction as in Cd the transfer should be about 4 , which the uncertainty in is not great enough to permit .
As the fraction77 is of the order 1-215 it suggests that the modification is produced by adding to the atomic volume term in the sequence of the series , viz. ( atomic weight term ) .
The question must be left open at present .
It has been noted that the arrangement which gives for Mg throws it out of the rule that the first members of the different groups are successive multiples of .
When the calculations were first made , the values of the DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
international atomic weights were used , and for Mg it is in place of BRAUNER'S .
This clearly gave and with the uncertainty 65 not including that of atomic weight and bein chiefly due to uncertainty in .
The transference required now is , and the fraction of is 502 , again clearly not that of the Ca group , but when account is taken of the uncertainty in quite possibly agreeing with that of the group .
The assumption that the international value of is more correct than BRAUNER'S certainly gets over the difficulties mentioned above .
But we are not justified in choosing the values from the particular systems which best suit our theories .
The discrepancy between the international and BRAUNER'S is very from 1 to 15 times BRAUNER'S indication of his possible error .
Another suggestion as to a possible explanation may be given .
There have been various indications in [ I. and II .
] that small variations in may occur .
If so it is possible to produce the changes observed by a small change in the middle lin of the triplet .
The necessary changes to do this are given in the sixth column .
The changes clearly depend on the squares of the atomic weight , for if they are expressed in the form they are Sr . . .
Cd .
Ba . . .
Eu . .
Ra . . .
in which it may be noticed that .
Again multiples of a quantity depending on the square of the atomic weight enter , and it is especially interesting to note that the group are affected with the multiple 6 .
If Ca and show similar displacements , Ca would require in place of 16 and is clearly , i.e. , is unaffected , but considering the small numbers involved in Ca and consequently large proportional errors , Ca might well show instead of 16 .
The question naturally arises , do these quantities depend in any way on the oun Now any change in may be supposed to arise either as a real change in itself or an apparent change due to the introduction of a factor in connection with the .
In other words , the quantity is N. or N. Looked at from this point of view , requires or .
Now would give , but if the present explanation is the true one , this is not a likely value since it not include the alkaline earths .
A value would be expected .
The Ba value would give , or practically .
It rather looks as if this explanation is a part of the truth .
If more exact measures were at disposal it might be well to assume these results as holding , recalculate for the denominators and discuss the rearrangement now required .
It may be noted , however , that in the Zn sub-group a factor in would reduce the calculated oun below and raise it in the Ca sub-group above and at the same time increase the factors in the numbers above towards and .
The factors may of course enter either as or Collaterals .
The first set in doublet or triplet or series is always the stronger .
The others may be considered as receiving a sort of lateral displacement , by the atomic weight term , in the recognised way , and may be called collaterals .
This kind of displacement is , however , not confined to the series generally recognised , but is of very common occurrence , and , indeed , depends not only on the but also on other multiples of In fact , the doublet and triplet series are only special cases of a law of very wide application .
Some evidence of its existence will be given below .
It will be sufficient now only to refer to certain points connected with the law , and to a convenient notation to represent it .
This kind of relation was first noted in the spectra of the alkaline earths , as the lines are both numerous and at the same time strong and well defined , and , therefore , with very small observation errors , any arguments based on them must have special weight .
Moreover , there are long series of step by step displacements involving large multiples of between initial and final lines , so that we may feel some certainty that these large multiples are real and not mere coincidences .
As a compact notation is desirable the following has been adopted .
In general the wave number of a line is determined by a formula of the form , and lateral displacements may be produced by the addition ( or subtraction ) of multiples of , say or , to or .
This is indicated by writing to the left of the symbol of the original line when it is added to , and to the right when added to Thus is .
So far as numerical agreement goes is a collateral of this represented by .
This means that whereas , see [ II .
] , Wave number of Ca , Wave number of * A note on this relationship was given at the Portsmouth meeting of the British Association , see ' Report , , p. 342 .
Though not always , as I hope to show in a future communication .
DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
3.37 Before going further it is desirable here to consider the nature of the cumulative effects produced by errors in the values of , or of the limits , in the course of a succession of step by step displacements .
There may be a small error in the starting point , e.g. , in the above example , or in the value adopted for .
We will consider these separately , taking the case where the displacement is on the left , or the first term .
1 .
The correct , but slightly too Then calculated from this is also slightly too large .
It will , however , serve to identify a large series of steps in succession , i.e. , to reproduce the successive differences of the wave numbers of the lines .
But the errors will all be , and if the last line of a set be calculated direct from the first , its denominator is too large anl its wave number too small .
In this case a more correct value of can be obtained by using these extreme lines , and this corrected value must satisfy all the other lines .
In general a new correction will only affect an extra significant figure in the value of 2 .
correct , but limit \mdash ; In this case a slight error in the limit will be of no importance unless the and its multiples are considerable ; and , as a rule , the limits are known with very considerable accuracy , except possibly in the alkaline earths and a few others .
Let us suppose the limit adopted ( say ) is too large , that is , its denominator too small .
If the second line is due to a positive displacement , its denominator is than that of the first , and the wave number less .
Suppose the denominators for the two lines , if the displacement is positive , separation is .
If the limit is chosen too large and are chosen too small , although .
is correct since is supposed correct .
If becomes the error in the separation is , which is positive since is supposed , i.e. , the calculated separation is too large .
If the displacement is negative one , , the true separation is now and the error , which is now negative since .
The effect would be that in any series of step by step displacements would appear to require continual decreases , and at the end the " " corrected values\ldquo ; would not at all fit the initial cases .
If , then , it is found that when is corrected as in Case l.the corrections tend to alter the former corrected one , and not to produce additional significant figures only , it may be surmised that the limit has been wrongly chosen .
It is clear , then , that where there are a number of successive collaterals with a large .
multiple of between the extreme ones , we have at disposal a means whereby much more accurate values for and the limits are obtainable .
Cases are given below , e.g. , in For low atomic weights is always a small quantity and except for orders where or 2 , the alteration in wave number is small .
For the present purpose which is to obtain proof of the existence of the displacements here indicated , no evidence can be admitted in which the change in wave number produced by a displacement , is comparable with the possible error of observation .
The evidence , therefore , is of V0L .
DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
The latter is Principal series , and the difference between wave numbers of its first line and of its limit gives the limit of the other two .
Of the other two series , one shows a Zeeman effect of the same nature as that in the Principal .
This is called the Sharp series\mdash ; or ( by KAYSER and RUNG ) the 2nd associated series .
The third series is called the Diffuse\mdash ; or the 1st associated series .
It has in fact a negative kind of criterion .
The preceding definitions apply to the three series in all elements , including such elements as Li , He , and others which show singlet series .
When doublets and triplets appear , we have a simple physical criterion for the Principal series in that i.t is that series in which the doublets or triplets converge with increasing order .
This criterion can be applied even when the 1st line has not been observed .
In certain elements the constant separations are shown between satellites .
In these cases the series is certainly a -series , at least in those recognised up to the present\mdash ; but further knowledge may show that in certain cases such satellites may appear in other series .
* If , passing beyond the mere physical appearance of the series or their visible arrangement in the spectrum , we attempt to represent their wave numbers by formulae .
the recognised types , we have further criteria for the Principal and Sharp , viz. , that the 1st line of the Principal may also , very nearly at least , be calculated from the formula for the Sharp\mdash ; or vice \mdash ; and that the denominators in their formulae differ , roughly indeed but sufficiently closely for use as a criterion , by a number not far from .
But when an attempt is made to deal in the same way with a line of the diffuse series , no general type of formula has , at least as yet , been foumd .
In the alkali metals , as was seen in [ I. ] all the -series take a positive value for \mdash ; in other words , the fractional parts of the denominators decrease with increasing order , and the general conclusion might be drawn that this was a common feature of all diffuse series .
But the opposite occurs in the triplet spectra of the 2nd group of elements , whilst a similar rule of a positive value of recurs in the 3rd group .
This suggests that the series giving doublets have positive and triplets negative , but this is contradicted by the triplet series of and Se , which behave in the same way as the doublets of Groups 1 and 3 .
The question naturally arises , is there a typical -sequence with positive , and the diffuse series 2nd group do not really belong to this type , or is there no actual -sequence , i.e. , no regular type of formula to which the -series conform .
The difficulty of finding formulae to accurately represent any particular -series would point to the latter supposition , a supposition also which is strengthened when we study comparatively the series of numerical values of the denominators found directly from observations as is done below .
In the case of the alkalies the formulae given in [ I. ] ( as well as those in ) do not reproduce well the high orders and are probably only within the limits of error because the lines are so diffuse that the observation errors are very large .
In fact one of the few excessive deviations found in [ I. ] was that of , in which it is *E.g .
, in , see Appendix not probable that the error is one of observation .
In Group 2 the sub-group can be reproduced fairly well with a formula in in which is negative .
Mg can also be reproduced within error limits by a formula of the same kind , but it is impossible to do so for Ca , and Sr and Ba require additional terms in .
In Group 3 Al is quite intractable , and if really depending on a formula , appears to require complicated algebraic or circular functions .
In and Tl also are not amenable to formulae in only or .
Nevertheless , the general build of the series is so similar to that of the others that it would seem probable that the wave should also be of the form .
If so it is possible to calculate from the observations and a comparative study may throw some light on the origin of the different lines .
The attempt to deal with these series from the formulae point of view , however , brought out the fact that the satellites are related to the strong lines in a similar way to that in which the Principal line doublets are , viz. , by a constant difference in the denominators and that their differences probably depend on multiples of the " " oun as is the case in the Principal series .
As the evidence depends also on a comparison of the numerical values of , this point will also be considered now .
The actual values of will depend on the accuracy of the value ( or of the limit .
In the calculations below the most probable value has been used ( see note under each element ) and the true value has been taken to be that .
In order to be free from mental bias these have been in general taken to be the same as , which !
nvolves the theorem that .
But of this little doubt can be felt .
The true values of can then be given in the form where is sman For high orders of is comparatively large and can only be used when is very small .
It is however generally the case that errors made in this way are only a fraction of the observational errors .
As in the normal type where there are no satellites , and where there are satellites , it is only necessary to tabulate the values of for the case of or respectively .
When this is done certain regularities are clearly apparent , which can be made more exact by allowing small observational errors and giving a small permissible value to .
, It would cumber the space at disposal to give both sets of values , especially as it is possible to easily indicate the differences on the one set of tables .
Table II .
gives the dlues of with the modified value of , with the maximum errors attabhed in the usual way in , and the calculated value given a correction to the selected value .
Thus for , is the selected value , possible change last three digits in this , is change for , and observed value is less than the selected .
The values of the errors of observd over calculated ( O-C ) , and of possible observed errors ( O ) are given in each case on the right .
The tables.for the different elements are collected together and aeon of each is given later when considering the ordinal.relations of the The tellite \mdash ; As the values of the satellite differences are practically independent of the exact value of , their consideration may be taken up at once and the details of the calculations respecting the tables postponed until the consideration of the order differences .
An examination of the tables will show conclusively that these differences are multiples of the " " oun Dealing first with the differences for the first lines , the following figures , contained implicitly in the tables , will show how closely this is the case .
The nearest multiples of the oun are appended , as calculated from the first approximations of Table I. The possible errors are those of the lines ( except Zn ) .
Cs . . . .
Cu . . .
Ag . . .
Ca. . . .
Sr. . . .
Ba . . . .
Not observed , as deduced below ( p. 388 ) , or . . . .
Cd . . .
Eu . . .
Hg . . .
Al . . . .
In . . . .
Tl . . . .
Se . . . .
The only case of " " failure\ldquo ; is the first difference for Sr in which the estimated possible error is extremely small .
If the possible error be the sum of those of each line , the value is 15 in place of 7 , and if be calculated from the most probable value of the oun it should be about greater , i.e. , .
It will be noted that where triplets enter , the .
two satellite differences , and consequently the two satellite separations , are extremely close to the ratio 5 : 3 .
This ratio ssems to persist also in Hg where the separations are in reverse order , and we find a ratio of 3 : 5 in place of 5 : 3 .
The for this ratio is in fact much more closely obeyed than the corresponding one for the ratio 2 : 1 for the triplet separations .
It is therefore of great assistance in searching for the lines of series whose limits are VD ( 2 ) , and which consequently possess constant triplet separations in this ratio .
Its explanation should be expected to throw some valuable light on the constitution of the atom .
The general dependence of the differences on the small " " oun\ldquo ; should also be noted .
Passing now to the consideration of the satellite differences for orders beyond the first , it is seen that they still depend on multiples of the oun , but different from those of the first order .
In a large number of cases the multiples are the same DR W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
for different orders within limits of errors , especialy in the doublets and differences of the second and third satellites .
Thus we find Cs , ; Ca , ; Sr , for and for for first separations , and for all orders in the second ; Ba and show too few for comparison ( see discussion below ) ; Cd , ; Hg is irregular , Al is anomalous ; In , Tl , for , and for ; Ss , as amended later , shows for , 6 , 7 , and for , the lines for being outside region of observation .
The evidence points to a normal rule that the differences for the orders beyond the first in any spectrum are the same , but different from\mdash ; in general greater than\mdash ; that of the first .
The Order Differences.\mdash ; The order differences change very considerably with a change in the value taken for the limit , i.e. , in the value given to .
No doubt with unlimited choice of it would be possible to arrange a set of denominator differences multiples of the oun within error limits , for a series of values of could be found making the first difference a multiple .
Out of these one or two would probably give the second such a multiple .
After the second the error limits as a rule come to be very large , in fact larger than half the oun itself , except in case of very high atomic weight .
No conclusions could be drawn from any such arrangement .
But in the present cases the choice of is bounded by very narrow limits , for the relation is supposed to hold , and , as a rule , the values of are known with very considerable accuracy , and the possible limits of variation are known .
They were given in [ I. ] and [ II .
] .
Before proceeding to draw general conclusions from Table II .
, it will be well to consider in more detail the data for the different elements on which the table is based .
Na .
Although the readings for are very inexact , the peculiarity of the large depression shown for , as well as the large recovery afterwards to mantissae close to unity , must be real effects .
It is of course possible that is a collateral from the normal type .
If be calculated from , the mantissa becomes 989054 , in other words , the line begins to show the rise to the large ffial value at , whilst does not do so until .
The lines would seem to succumb to the disturbing effects sooner than the .
It was pointed out in [ I. , p. 83 ] that in the Na the series apparently belongs to the type , in which the mantissa is 998613 .
It would almost seem that the peculiar rise shown is due to the fact that it reverts to this sequence .
Here , as we shall see in other cases , the values of the first members of these series appear to be subject to some kind of displacement which affects their ( supposedly ) normal relations to other lines .
If now the first mantissa be supposed normally to be this 998613 , it is 9691 above that in the modified table , and this is , thus completing the order differences as multiples of .
But in any case the data for Na are of small use for the pressllt purpose , as the errors are so large , and so small .
The arrangement in the table VOL. CCXIII.\mdash ; A. gives the values of least .
If be taken about6 , the order differences can run within error limits .
K. It is seen that the numbers , with the exception as in other cases of the first difference , fall into multiples of quite naturally without a change in , though possibly a small change in might make the values of O-C less .
is so large , that the theory of the dependence on multiples receives considerable support .
The first also is close to .
This element is one in which the value given in the first discussion above for the oun is , which is presumably too large by 8 to 9 .
If it be should be less and .
This would scarcely effect the other intervals , but it would make the first one .
Again there is a sudden fall .
It doubtless corresponds to a real effect , for SAUNDERS as well as LIVING and DEWAR make anomalous here .
S. observes and L.D. but both give the same .
If this be taken as correct and the normal found from , the mantissa is 733756 , giving the difference .
This shows that the set have not participated in the sudden fall\mdash ; at least to the same degree as result analogous to what happens in Na .
Rb .
In Rb there is some doubt whether a satellite series exists .
The question has already been discussed in [ I. , pp. 71 , 86 ] .
SAUNDERS has given for lines whose wave numbers are , and , with normal separations and satellite separations .
Also is a doublet having a separation which certainly points to an unobserved satellite about .
But RANDALL'S observations for show only a doublet of normal separation\mdash ; that is clearly with no satellite .
Moreover the series , which depends on and for its limits , is a singlet series and not a double one .
In the table the series is taken as if it were without a satellite , the reading for being corrected from .
In other words it represents the satellite lines if they actually do exist .
In the latter case the strong lines would show denominators about 609 , and 1100 above those in the table for and .
The first is about and the second of the order of , whilst if normal , judging from other cases , they should be equal .
Moreover , in all other cases ( In and Tl excepted ) the satellite separations are practically the same multiples of the oun in the same group .
Cs shows a difference of , so that the supposed ones here are far too small , as well as irregular .
The observed separations , moreover , are equal within errors of observation , which would rather point to an alteration in .
Now a lateral displacement of on would produce a separation of , which is about the observed value .
The table shows a stationary point at and 9 and then the large fall shown in the other elements .
They could be accounted for by a lateral displacement DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
, the mantissa of being at the same time subject to the fall of multiples of which would scarcely stop at .
In the table , however , the errors are inserted on the supposition that the mantissae remain constant .
Cs .
The mantissae appear to run down by equal intervals from to 6 , are equal for 6 , 7 , 8 , then a large drop of about to the same value for 9 , 10 , and return to the value at for .
The possible errors are so large that the regularity is curioua It is possible that they.might run down at equal intervals of to the last one for .
Or , if there are very small observational errors , the drop for 9 , 10 may be due to a lateral displacement , about .
It should be noticed that with Cs all the order differences but one are multiples of , or the group multiple .
Cu .
The two first doublets of are strong .
The third is much weaker than would be expected .
Moreover , it gives a separation between and of , whereas it should be somewhat less than .
This can therefore scarcely be the normal chief line of this doublet .
Now EDER and VALENTA give a spark line at which gives a separation with of , leaving a satellite separation of , which is within limits in fair order with the corresponding separations in the two previous doublets , viz. , and .
Moreover , the satellite separation of gives a satellite difference of 1317 and , so that the normal satellite differences would run .
This , then , would seem to be the wanting normal chief line , and it is then interesting to note that the line usually accepted as is a collateral of this .
The denominator difference of the two is 2474 and .
Hence the K.R. line 3688 is the collateral , and apparently the small intensity is due to the usual decrease produced by a negative lateral displacement .
The modified table is taken on .
It is remarkable how close the observed differences come to multiples , but little reliance must be placed on deductions from them both on account of the large possible errors and the smallness of Ag .
Unfortunately the series in Silver is poorly developed\mdash ; only the first satellite ha.s been observed , and the three chief lines after the first have very large possible observation errors .
Nothing , therefore , can be learnt as to the march of the satellite differences beyond the fact that the \mdash ; is very close to In the modified table , the differences are very close as is seen to multiples of , but there can be no certainty with such large possible errors .
DR. W. M. HICKS : A CRITICAL STUDy OF SPECTRAL SERIES .
Mg .
In the series of Magnesium , as arranged by KAYSER and RUNG and as generally accepted , there are clearly certain abnormalities .
is more intense than we should expect , and its separation from is in place of , whilst that of and is very close to the exact value .
This cannot be due to observational error , for this is very small ( 03 ) .
Either , therefore , the true line is hidden by this bright one , which can scarcely be the case , or it is a collateral .
In the former case the true line ought to be that found by deducting and from the satellites .
In the second case it would require the addition of to the denominator of , and the addition would explain the increased intensity .
The two results agree , the wave numbers resulting bein respectively and 71 .
The former would give a denominator832041 in place of that in the table , but its observational error would be that of , viz. , 945 , while that of the collateral depends on the observed , and is 190 .
gives normal separations within limits .
gives and but normal within the observation errors [ see Note 1 at end ] .
But there is another question which arises in connection with Mg .
In the Ca sub-group the first lines have a denominator about , i.e. , with .
In the sub-group the lowest value of is 2 .
In the series , as generally accepted , the first line is , which requires .
If there is a line corresponding to it should be in the neighbourhood of 14900 .
Now PASCHEN has observed a strong line at , but there is no triplet , which would be decisive against the allocation if we could be certain all the lines must exist .
But there are cases where normal lines are observed weaker than we should expect , or are not seen at all .
The well-known case of KD ( 3 ) is one example , and it is curious that if 14877 be taken as the denominator comes out as given in the table in a very natural order with the other denominators .
The question is considered later under the series , and the evidence there adduced is rather against the present suggestion ( p. 398 ) .
Ca. The value of is calculated from as would give , practically the same .
To bring the differences of the first three denominators to multiples of it is necessary to diminish the limit given from the consideration of the series by ( variation limits given in [ II .
] ) .
The values can then be arranged as in the table .
One result of this is to increase the value of ( for the given ) to 2793 from 2791 , which gives in place of , and thus closer to the adopted value .
A noticeable peculiarity in this series is the very rapid falling down of the denominators after .
It is so large and at the same time so irregular that they cannot be brought into line with the others without diminishing the limit by a large amount and by different amounts .
It clearly points to the existence DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
of collaterals , formed by the addition of ouns to the limit .
As such increase tends to increase intensity it may account for these surviving when the typical ones are either too faint or are destroyed to form the collaterals .
It is useless to attempt to determine these multiples , because the observation errors are so large themselves as to be a large multiple of the oun , and at the same time we have no knowledge of what the typical VD ( m ) should be .
In general in the 2nd group the successive denominators are formed by the successive addition of smaller and smaller multiples of the oun until probably a constant value is reached .
In the present case , with the quantity , that limit is certainly not reached .
But it may be instructive , in order to illustrate the nature of the suggestion , to find what the collaterals ought to be if the denominators of VD ( m ) remain the same for , viz. , 091707 .
The multiples are found to be .
The series of the observed lines may then be exhibited by the following scheme , where stands for , , , , , .
The line is interesting .
PASCHEN gave it as with the remark " " Wahrscheinlich doppelt \ldquo ; and he allotted to and to the Principal series .
But in [ II .
, p. 56 ] reasons were given against the latter allocation .
In fact the line is very close ( probably within observation errors ) to the collateral formed by adding one oun to .
The of such would be [ see Note 2 at end ] .
Sr. The value of is calculated as from , which gives .
The differences as shown in the table are extremely close to multiples of .
Moreover , the limits of variation for the first two are so small that the variations of ROWLAND'S standards from the correct values for his scale may become of importance .
For the values should be lsss , whilst for , failing direct observations for reduction to , recourse must be had to extrapolation on KAYSER and RUNGr'S formula , *which has been done .
In order to bring the differences for and and of and to multiples of within error limits , it is necessary to take about or .
When this is done the denominators can be arranged as in the table .
The difference of the two *RANDALL appears to have done this for but not for , which also makes his value of instead of , which is close to the true value .
first denominators .
It is possible the real errors attached to by RANDALL may be greater and the difference slightly less ; but if we suppose to be the real value of it makes , and , therefore , very close to the adopted value .
It would appear that has been displaced from its normal value , judging from the irregularity introduced into the separations .
If so the separations might be in place of Ba .
Starting with the uncertain value of , as given in [ II .
] , the value of as calculated from is , and from is , both being near the most probable values but on opposite sides .
The value 683 is taken at first as a rough approximation .
Apparently , the first set of lines have not been observed .
RANDALL* has observed two lines , , which give a separation , which is but no signs of satellites\mdash ; or , rather , if there are satelites , the separation observed should be much smaler .
If , however , the satelites have gone here , and this pair denote the first two lines of the first triplet , they depend on , and the value of the denominator is which would range well with those of Ca and Sr , viz. , , but the second lines of these give , and of Ba , which would rather point to a less value than for the first lin. But if VD ( 2 ) is larger than , the lines would be for the first and for the second , giving a denominator for the first of .
The differences of the denominators of the lines for , 3 , 4 , will then be 267632 , 38612 instead of 6528 , 38612 , and are therefore more in agreement with the type of the other elements of this group .
Moreover , the former , as we shall see , is a multiple of , whilst the other ( 6528 ) is as far out as it can be .
Both values , however , are inserted in the tables ( see also p. 389 ) .
The satellite differences for are and .
The values of and are respectively and , and hence the first cannot be within limits of error , although it is so close as to produce a conviction that it really is so .
Now for small variations of the limit the separation differences are scarcely affected , but , as we saw in [ II .
] , there was evidence to show that the limit was considerably less than that found , and , in that case , the separation differences would be slightly changed .
A decrease of would increase those differences .
If , however , it is so large as to bring up the first to , the second is increased so much that it is not within limits .
Consequently , the two conditions confine the choice of within very narrow limits .
It is found to be close to a decrease of 32 , i.e. , .
This again changes the values of , with the given values of and to 29379 and 11997 , giving from .
The table is * Ann. der Phys 33 ( 1910 ) , p. 745 .
The values in this and in the table are calculated from the limit as modified below .
DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
drawn up on this basis .
With these data , the suggested allocations for give the following differences between their denominators and that of , viz. :\mdash ; With The first , therefore , cannot be a multiple within error limits .
The values shown for agree very well , but the regularity is upset .
Also the actual lines have changed in appearance , and their intensities are not normal .
The intensities for are 10 , 6 , 4 , those for , 4 , 6 .
We should expect to be stronger , and certainly to be much weaker .
It would seem that collaterals must displace the normal lines .
We have , in the foregoing pages , been led to expect that an addition of an oun increases the intensity and a deduction diminishes it .
If so , we should expect a deduction in and an addition in .
To bring above requires the deduction to be made in .
This would make the typical value of the denominator greater by 2733 , viz. , .
In the case of , to bring it closer to , i.e. , distant , the addition would have to be to , and if so , the value of , given in the modified table , would have been a mere coincidence .
But no such addition of a multiple of ( nor of ) will do this .
If , however , be added to the denominator in , it is brought to separation of , giving an error in for in place .
If , then , be also deducted from the separations will be .
The separations would then take the form This arrangement is to be preferred in that ( 1 ) it explains the abnormal intensities of , ( 2 ) brings the separations into line with other elements .
The arrangement suggested may be stated thus : if and represent the typical lines , the observed is with decreased intensity , and is , the increased intensity due to the in being greater than the decrease due to in The results for orders are similar to those of the other elements of their groups , probably collaterak of additions to .
Zn .
On account of the small values of , and the considerable observation errors , the satellite separations in Zinc do not give decisive results .
If we take the observed values for the and sets , the denominator differences are 489 , 419 for and 584 , 399 for .
Now the second sets , the , have much smaller observation .360 DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
errors than the satellites of the first\mdash ; see remark in the table .
If the latter be calculated from them , using the accurate value of , the differences come out to be 525 , 369 and 456 , 399 , quite reversing the order of magnitude for the first satellite of the lines for , 3 , and , moreover , their differences are larger than The differences best consonant with the measures\mdash ; using the derived values from are and , giving a ratio of .
Those entered in the table , however , are , both within the limits of observation , and adopted because they give a ratio 5 : 3 , the same as the other elements of Group 2 .
The satellite separations for may be the same as the latter within limits , but not necessarily so .
The order differences do not work in well with the above when .
If , however , be put , they fall into line with extreme accuracy , as shown in the table .
It is of interest to notice that the differences are multiples of , which seems to be specially associated with zinc .
The denominators for , 7 are now ( 4068 ) and ( 11320 ) , showing too large a rise to be due to error observations .
Treated as collaterals with and they become , clearly near the probable limit .
Judging from analogy , we should not expect the differences to stop at .
The series would probably be continued , but the errors are too large to settle the question .
If the series were continued , e.g. , by , the denominators would be and .
But the best agreement is to take them as they are .
This would also be in analogy with Ca , in which begins to change when VD stops changing .
Cd .
In the table is taken 25 above , as given in [ II .
] .
It is seen that the arrangement fits in with great accuracy , and as is as large as 114 , the arrangement may be considered to have some weight .
The denominators calculated from the and lines ( more accurately determined ) do not agree with the observed satellites .
It would therefore appear that the second and third members of .
the triplets may also be subject to special displacements .
Here , for instance , the lines of order 3 are brought into line if the observed is .
The value of , calculated from K.R. , cannot be the normal one , even when his extreme possible error is allowed .
This shows again that must be displaced from , in this case by The denominator for shows the sudden large rise after a slow change which Zn also exhibits , but it cannot be explained , as in that case , by treating it as a collateral due to a modification of alone .
The closest collateral of this type would be due to and this changes by producing in the denominator a change too far in the opposite direction .
In fact it becomes , 3471 below DR. W. M. HICKS : A ITICAL STUDY OF SPECTRAL SERIES .
that of in place of 5060 above .
But the observation error in this line is so large that as it stands it may correspond really to a denominator equal to that of .
If further the series of differences , be continued diminishing further as looks probable , it would come more nearly in line .
For instance , as each term is about the previous , the next is .
The value of as thus calculated differs from the observed by while the possible error is 30 .
This would make the constancy of the denominator begin at an order one higher than in .
The case of Hg will be seen to support this tendency .
There are two sets of doublets , , and 300553 , with separations and 117195 which are clearly associated with the series .
If we write down the observed satellite separations in the and the lines and of with the above we get the following scheme : and .
At first sight it makes the new lines appear as collaterals by the change of about on , but this cannot be the case , because a change of this amount would very considerably diminish the doublet separations below 1171 .
If the separation be deducted from that of the lines in question there results and .
Now the separations depend , as is seen above , on , so that the new lines may be written , where A is a constant which on more accurate calculation is found to be .
In other words , the VD of the new lines is derived from in the same way as that is from .
This formula is of a type of which there has been no example hitherto .
If it remained there the evidence , in spite of the curious connection with the other satellites , would scarcely be weighty enough to cause the introduction of a new departure .
I hope however to show in a future communication that this expresses a very common relation between sets of lines , the constant A being in reality a composite one .
The question naturally arises do the new terms give rise to an series in the same way as the ordinary ?
It should be at a distance about in wave number above that of F. The line is above that of the line which is allotted by PASCHIIN to and is clearly the first of the lines in question .
There is an line at and aIlother at is above this .
This may be the corresponding line sought for , but if so the line 11630 must be and the lines , would then be absent .
These lines were at first assigned by PASCHEN to a new doublet set of series , but to combinations of his new singlet series with , .
This question will be considered as a whole later , but the suggested explanation given above points rather to the fact that we have to do with a triplet series in which the third number is too to be observed .
stands as usual for or .
'Ann . .
Phys 29 , p. 650 ( 1909 ) .
'Ann . .
Phys 30 , p. 749 ( 1909 ) .
VOL. CCXIII.\mdash ; A. 3 Hg .
The series of Mercury shows a marked divergence from those of the other elements so far considered in that ( 1 ) the separations of the satellites increase as they go from the chief line , ( 2 ) the satellites do not seem to correspond in the different orders , and ( 3 ) there are a larger number .
The increased tendency which this element has shown to break up into collaterals appears also here .
One is led to infer that with varying conditions of the production of the light different collaterals appear .
The dependence on the oun is , however , here clearly shown , and the evidence is all the stronger because the nitude of itself is large ( 363 ) and because all the apparently unconnected differences come within close multiples of .
This is clearly seen in the following table where the denominator differences are exhibited together:\mdash ; It is still possible within errors that the differences for the first satellites shall be the same as for the second and succeeding , viz. , , but it is very unlikely .
For the second satellites this cannot hold .
It is clear that the regular law is not contradicted , but is upset by the formation of new configurations or aggregations in the oscillators .
The table is drawn on .
This brings in the best agreement and , moreover , brings \mdash ; supposin it and are the same\mdash ; closer to the value found from the first three lines .
The value given in [ II .
] being one modified slightly to bring all calculated ( even for ) within limits .
The agreement is to be remarkably close .
It is to be expected that the differences for the satellites of the same lines will be more accurate than the differences between the chief lines themselves , and this is exemplified in the table .
The observation errors after are too considerable to draw certain conclusions from .
Apparently the denominators increase by small multiples of to about and then remain constant .
Al .
In Al , the satellite differences deviate from ths ordinary rule in that they increase with increasing order for , 3 , 4 .
They are 94 , 800 and 1380 , and by no stretching to the extreme possible errors can the two last be made equal .
The inequality is certain .
Moreover , the observed differences are very close to multiples of .
If the first satellite position be calculated from , its difference is 110 , the observed is 94 and , 94 can be 106 within limits .
gives in the same way 800 for , the same as observed and 1468 for instead of 1380 .
The last may be the same as the observed within limits , but aS and DR. W. M. HICKS : A CRITICAL STUDY OF SPhCTRAL SERIES .
it is possible there may be this difference and is not as in the typical cases .
The real difference may be any multiple between and .
In fact is so small that there is not absolute certainty .
has proved itself the most intractable series to bring into any simple formula of the ordinary kind .
It was , in fact , the difficulty with this element which first led me to seek another solution\mdash ; on the lines now being considered .
It will be seen that it lends strong support to the theory suggested .
The table is arranged with .
The exactness of the relations there shown is very remarkable , and when it is remembered that is a large number like 1754 , the practically exact multiples referring to the first five lines must carry very great weight in the argument that at least is subject to a modification of successive denominators by multiples of certain units .
The objections to the arrangement are two : ( 1 ) that is outside the error limits of , and ( 2 ) the denominators appear to on dimimishing without reaching a limit .
A slight alteration , however , in will get over the first .
For instance , if would be about less , would be within limits of , and the same arrangement would also hold , but it could not be much more diminished because with and 6 the changes introduced into the denominators by would upset the multiples 54 and 29 .
change of by would change the ratio to from to is , therefore , probably very close to If , the denominators tend to a limit about 107 for and beyond .
But this is far outside permissible limits of , and , moreover , the striking arrangement with multiples of is quite upset .
We must therefore conclude either that the limit is not reached until an order , or beyond , or collaterals enter .
If the former , multiples of can enter , but the observation errors are too large to give certainty .
If collaterals based on are used with or the mantissae for 7 , 8 , 9 come respectively to 113569 , 113590 , 113956 , and for 10 for a , 113700 .
The separation observed for is instead of , either an observation error or a displacement of or A displacement of by on brings the separation very nearly correct , although the observation error in the wave number of these two lines is as large as , it is probable their difference is much more exact and that the defect shown by is reaL practically The results therefore to show that the and lines for , s , 9 , 10 are collaterals , , except that for an extra displacement of is added .
Although the numbers above are so nearly equal we must not place too much reliance them , as the observation errors have a very large effect on the denominators for such high orders .
If the suggested arrangement is correct it must mean that K. and R. measurements must have been of a very high order of exactness , which would further mean that the measures for the lines would not be so exact since the observed values of the for , 8 , 9 are respectively ( possibly real on ) , 7 , 13 in error .
.364 DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
I am inclined therefore to think that the exact equality for is a coincidence , especially as the difference for and 7 is not a multiple of like the others .
Taking the corrected , the mantissa for is 116154 , giving with a diflerence , which is as far as it can be from a multiple of If be taken for the collateral , and for the difference , the limiting denominator is 107384 and corresponding , 07 .
Now these make , as against the observed lines , the values of about correct , which gives a certain amount of weight to this arrangement of collaterals based on .
In the foregoing the conclusions up to may be taken as well based .
No definite answer can be given to the question of what happens beyond , although balance of evidence perhaps points to the last , viz. , collaterals based on , and this is strengthened by considerations which follow .
MANNING* has recently observed under diminished pressure certain groups of lines which by their look suggest doublets and satellites related to the type .
The strongest set , apparently a and doublet , , giving a separation of .
If be treated as having the same limiting term as the series , the denominator of the VD part comes out to be .
Now this mantissa has the limiting value according to the supposition of the preceding paragraph , , 107384 .
If this is not a mere coincidence , the connection should throw a great deal of light on the relations of these series , and would warra1lt a more searching discussion .
This would , however , lead too far from the immediate point at issue .
It will be sufficient merely to indicate somewhat more clearly the connection .
With the limit the mantissa of is below that of ( see Table II and it should therefore be with 63132,8 ( 25 ) whilst that of MANNING'S 4260 is .
If these are the same ( 46)-41 .
and this gives , which is too small .
But in this of the actepted series is referred to , whereas 4260 is referred to .
If it be referred also to , its mantissa is 107868 ( 21 ) and and therefore of the right order with MANNING'S new lines suggest , on a superficial glance , a series of bands , but there can be little doubt of their connection with the Diffuse series .
Their relations to one another can be discussed either on the basis of the old or of .
With * Observed in my laboratory , since published in ' Astrophys .
Journ May .
1913 .
DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
as ( intensity 10 ) goes as ( intensity 9 ) .
should give .
This has not been observed , but it is .
In fact , the error between this calculated value and that deduced from is only , and a satellite difference of is more in accordance with that of other elements than the small one of in the accepted series .
Amongst the other lines are the collaterals ( intensity ) with , and ( intensity ) ( 4260 ) with In .
As in the case of Al , so In shows an increase of satellite differences with the order .
The first three , be considered as certain , but the next , , although it is close to the observation , may , as in the case of aluminium , be the same as the others within error limits , owing to the large error in gives the difference in wave-lengths as , whilst HARTLEY and ADENEY in the spark give it as 4 , i.e. , clover .
In the table it is entered as as being closer to the observations , but if it really is , the O-C is against .
It is possible that many cases of diffuseness may be due to the simultaneous existence of several collaterals based on differences of , which for ]ines where is large or for small wave-lengths give differerlces in too small to resolve .
In this case , for instance , with a displacement by produces collaterals differing by about and several would give the impression of a nebulous line , broadened on one side or the other .
For there is clearly some collateral change different in and For if be calculated from it gives a position for of longer wave-length than , or the inverse of the typical order .
No conclusions therefore can be drawn as to the satelite differences for , except that is probably of the form .
Beyond this it is curious that the lines persist while the lines do not , which may be accounted for by their being also like additive collaterals .
Again , also , the order differences show themselves as close multiples of .
The table is based on , but it may be brought into still closer agreement by taking a small negative number , about to .
The difference between 5 and 6 becoming suddenly so large ( 59463 order ) and the entrance of the peculiarity mentioned above , suggest that some collateral influence comes in .
Further , if we regard the denominators of , or of calculated from , after a small difference of 8824 , the differences begin again to increase .
This has always in the previous cases pointed to a collateral displacement in .
The first object is to see by what displacement the denominators may be brought to a limiting uniform value .
If be put , ths denominators for become 715818 , 714394 , 716198 , 715743 , 7222 , 7319 .
omitting the values for and 11 , in which the probable errors are very large , it is clear that , allowing for quite reasonable observation errors , the denominators are in the neighbourhood of a limiting value .
Now , a collateral of in produces a displacement , and this makes the denominator DR. W. M. HICKS : A CRITICAL STUDy OF SPECTRAL SERIES .
for to be 716474 , and this is 23931 below the observed denominator for .
Now , , which is the same as the difference for and 5 instead of being less , as analogy with others would lead us to expect .
But the observation errors are large maximum order about ) so that room is allowed for this .
There would then be one step from to 6 of something less than combined with a collateral displacement of on .
To indicate the explanation this is entered in the table , a difference of making the denominator for The observed abnormality as between and is that the wave number of is less than , whereas it should be 8 or 9 greater , with a denominator about greater instead of 1721 less .
There is , in fact , a further defect beyond the normal value of about .
Thus , if the difference for is , that of is .
The lines would then be represented as follows:\mdash ; The collateral addition intensifies and explains its continuance , but in the increase , owing to addition of to , is overweighted by the diminution of the excess in the VD part , and so , after the first , the rest are too faint to observe .
At least , that is a suggestion of a possible explanation .
Tl .
In Thallium the satellite separations appear to be the following multiples of : 24 , 27 , 28 , 28 , 28 .
But so far as limits of error permit , they might be 24 , 27 , 27 , 27 , 27 .
A peculiarity appears in the lines in that the doublet.separation for the first set is , whereas the normal value is very close to .
The difference is therefore real and not attributable to observation errors .
Moreover , the next four show a gradual diminution , although still remaining normal within extreme permissible errors .
The doublet values beyond this depend for the urements of the second line on measurements of CORNU .
They give separations 15 to 20 less than normal , but little reliance can be placed on deductions from them , for CORNU'S results may err possibly by several units in the first decimal place , and with these small wave-lengths any error in is multiplied by 22 to 23 in the wave numbers .
I have , therefore , not brought them into the discussion .
The table is based on , though no attempt has been made to find the best value .
The mantissa of the first line is abnormal , since it is less than the second instead of greater , and , moreover , its difference from it cannot be a multiple even of Since the satellite difference is very close to such a multiple it is probable that the abnormality affects both in the same manner .
Now the arrangement may be made normal by regarding the first line as a collateral .
The DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
addition of to the denominator of produces a change , and this changes the denominator of to those given in the table , anrl as is seen now , produces a difference of between it and the next .
Now , this alteration in diminishes the value of by , whilst , as we have seen above , it is apparently too much\mdash ; or is below the valus of .
Now , this is just the change made by deductin from the denominator of .
The exact value is , which is within the limits .
The way in which , with the considerable numbers involved , all the different abnormalities are simultaneously made to fit in with a normal scheme gives some confidence that this is the real explanation .
The scheme of actual lines may be represented thus : Actual , , .
Contrary to the case in other elements the successive differences are equal after the first , and the limiting value of the denominator is reached at .
They can all from 7 to 14 be , within limits , equal ; but there is an apparent rise with the high orders .
O. Two series\mdash ; one of doublets and one of triplets\mdash ; have been recognised in oxygen .
The table shows that the lines of both sets fall into lin quite naturally with multiples of closely except in the case in the doublet sets of , 8 .
In both these cases the denominators are equal within limits , but much larger than those for instead of being less , and the deviation is real since the difference is more than 15 times the probable error for , and times that for , which latter has a very considerable probable error as against07.* The divergence for can be accounted for , as it is probable that there are two close lines here due to different series , viz. , that for this series and the other for a parallel series for which , and may therefore be stronger .
As it throws some light on the subject it may be well to say a few words about it here .
RUNG and PASCHEN give three lines at , with separations , and intensities 1 , 3 , 1 , so that the centre is the strongest .
There is a corresponding set at , with the same separations within error nits and intensities 3 , 4 , 3 , again with the centre strongest .
The strongest lines of these two triplets form a series with the observed value of .
They are of a diffuse type and in fact are not to be confounded with K.R. possible errors .
The possible errors are probably larger .
come between the and series .
The limit taken is and the scheme is as follows:\mdash ; o-C .
O. 969545-8 . . . .
01 03 . . .
5 Moreover , the difference between the first denominator and the corresponding one for is 2924 , and this is .
It is of course understood that the digits in the limit have been chosen so that the , come very close .
The argument depends on the possibility of doing this .
In fact RYDBERG'S tables give the limit 22926 , so that the modification by 11 is extremely slight .
Thus the observed line is the line corresponding to of this series , and it probably hides the weaker line of .
This accounts for the deviation noted aboye between calculated and observed in .
I have no explanation to offer for the corresponding deviation for .
All the others come so close that it is difficult to imagine that this does not fall in with the rule .
It is equivalent to an error in of about .
The doublet separation for is 62 very closely , and the corresponding doublet difference is say .
A lateral displacement of on the limit would just make the change , but that explanation seems out of place here .
The separations of the new lines require denominator differences in the limit of 373 and 473 , and and .
There is another line at , showing a separation of .
If this has the same VD as 6261 it requires a denominator difference in the limit of 277 and .
The four lines are therefore , ( 6261 ) , 6261 , and ( 6261 ) .
S. If RUNG and PASCHEN'S estimates of.their errors are valid the , value of the limit of the series is determinable very accurately .
It is , but to bring as calculated within limits it is necessary to take more than lless .
Accordingly the Dlines havs been calculated on the supposition that and it carmot be far from this .
To bring the differences within multiples it has been necessary to diminish this limit by putting .
The multiples then come in partly as multiples of and partly of .
The value of given in the first part of the paper is 651 , but this gives , whereas it should be , if the rule there established is correct , in the neighbourhood of , or larger , say .
This value has been adopted in the table , although the old one can be made to fit in though not so well .
The agreement is good , especially when- it is remembered that K. and P. estimates are less than Sjble errors ; DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
Se .
The lines allotted by RUNG and PASCHEN to the series present a quite different appearance from the normal , although there can be little doubt but that they form the .
The weak satellite lines after the first appear on the violet side of the strong lines , whereas in all other yet known cases they lie on the red side .
Moreover , the strong lines instead of standing by themselves are each the first members of complete triplets for , 5 , 6 , 7 ( is the first set observed ) .
The numbers in the table are calculated with .
The value of calculated from is , but it requires , as in , a further diminution of over 1 ( i.e. , within error limits ) to bring in the fifth line .
Hence cannot be far from 19274 .
The value of is calculated from , and consequently must be considered very exact .
and are calculated by transferring from calculated to , making the values and .
The numbers calculated from the observed values are given in a separate list .
A glance shows that the usual regularity is here quite upset , and one feels convinced that some disturbing influence must have been at work .
If we examine the wave numbers of the first four sets as exhibited in the following table , we notice that for the first ( 4 ) 1595 89 ( 3 ) 16917.33 set , we should expect a weak satellite about behind 14156 , which is not likely to have been observed in that region far in the red .
Its difference , 3129 , is close to , and provisionally we may regard this as normal .
The next two triplets give separations respectively and , corresponding to a lateral displacement in of ( actually gives925 displacement ) .
For the separation is corresponding to a displacement of ( gives ) .
Moreover , for this line the intensity has increased from 5 to 7 and gives suspicion of a displacement by addition .
If we suppose that the chief lines have a lateral displacement it means adding to their wave numbers , i. e. , they are I1 and , and they come in front of their satellites , which , allowing for errors in observation , is in fair order with the first separation .
For , not only is the strong line abllormally more intense than for , but the faint line is so V0L .
CCXIIr .
also\mdash ; which suggests they are both displaced\mdash ; a suspicion increased by the abnormal increase of the difference shown in the table of denominators between 6 and 7 .
Provisionally the least change is to suppose the faint line displaced by and the strong line by , as it must , as was noted above , be more than the faint line .
For it is curious that only one line occurs and no triplets .
This suggests that there is no intensification by lateral displacement , and that provisionally it should be taken as normal .
The table of difference shows an abnormal increase instead of a decrease , but this may be due to observation errors .
If we now calculate the denominators for , 5 , 6 , 7 , on the above suppositions , displacing the lines for , 10 also by , we get 2139 622797 ( ) .
3475 617055 ?
Thus the changes indicated by the appearance and arrangement of the lines have brought the denominators and satellites into greater accordance with the general rule .
The practical constancy of denominator differences is exhibited also in Tl .
The only outstanding irregularity appears to be the satellite difference for A lateral displacement of in would decrease the denominator by 743 , and increase the difference from 2139 to 2882 .
It is better to leave the difference without an attempt of explanation at present .
The second list has been drawn up on this basis , taking as the errors are somewhat smaller with this value .
The denominator for is left without further change .
Another displacement of would bring it also below that for , but the observation errors render any deductions quite unreliable .
The DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL suggested scheme of actual lines may therefore be represented as follows where stand for the normal type , and .
The order ( 4 ) of the first line is so large that the error limits are too wide for absolute certainty .
In fact better agreement on the whole for the satellites would be obtained by taking the difference as , i.e. , being specially associated with this group ( see p. 331 ) .
The line is separated from by , and is therefore possibly the lateral .
The table shows of course the known essential difference between the behaviour of the elements of Group 2 and that of Groups 1 , 3 , 6 , signified by the signs of in the formula .
It consists in the fact that in Group 2 the orders are formed in succession by the addition of multiples , whilst in the others it is by subtraction , with the exception that Cu and Ag of Group 1 are additive .
But there are certain other features which appear between the different sub-groups when higher orders are looked at .
The alkalies all show a gradually decreasing decrement with a sudden dive .
Na then shows a sudden rise continued for several lines , and Cs has a similar indication .
Cu and Ag with only a few lines observed show decreasing increments .
The alkaline earths show decreasing increments and a sudden dive ( Mg excepted ) .
The sub-group shows decreasing increments and then a sudden ascent .
The Al Sub-group 3 show decreasing decrements ( Sc decreasing increments ) .
with and Se show decreasing increments .
In fact , were it not for the very clear behaviour of Zn , Cd , and Hg , the evidence would rather point to the conclusion that in each group , the low melting-point sub-group show subtraction ( a positive ) and the high melting-point addition ( a negative ) .
If this series depends on a formula sequence , it is difficult to see how it can be any simple algebraic one\mdash ; the mantissa would rather seem to depend on a term similar to or ma .
In the detailed discussion above , however , it is seen how these changes of direction can be explained by lateral displacements .
It is noticeable that where the irregularity observed in the first lines as compared with the others in the satellite differences appears , a similar irregularity exists also in connection with the first order differences .
This is evident especially in the alkalies , where the first differences are so close to exact multiples of or as to cause the conviction that they really are so .
DR. W. M. HICKS : A CRITICAL STUDy OF SPECTRAL SERIES .
It is a remarkable fact also , and one which will probably be of importance in throwing light on molecular constitution , that all those elements which do not exhibit satellites have order differences depending on multiples of , whereas all the others ( Al excepted ) depend on multiples of the oun , or .
The elements without satellites are Na , , Mg , possibly Al , both series of , and S. All these depend on multiples of or .
None of the others do so , and it may be regarded as an argument in favour of Rb possessing satellite series that its differences do not depend on directly .
It would appear that Rb only begins to show them for .
For the line is not split up into a chief line and satellite , the doublet separation is normal , and it is instructive to observe that the order separation between the first set and second line is close to , and only deviates from it in the same way that is mentioned in the previous paragraph .
Also Ba seems to have in the same way no satellite for , the separation is quite normal , and this also shows a first order difference very close to But Ca , on the contrary , which has a first difference , possesses satellites .
It is noticeable also that the high atomic weight elements appear to follow more regular and simple rules .
Thus both Cs and Tl show descent by equal steps in both cases The result of the discussion would seem to be that there can be no doubt but that satellite differences as well as the doublet and triplet differences depend on multiples of the oun .
For the other supposition , viz. , that in the Diffuse series the order differences also depend on differences of the oun , it can only be said that a case has been made out .
The supposition in all cases fits conditions , but the conditions are not all sufficiently definite to give certainty .
After the first two or three orders the observation errors are larger than the , and even for these the value of for the low atomic weights is comparable with the errors .
In some of these cases , however , multiples of which is much larger enter and strengthen the argument .
The strongest examples are those of the alkalin earths ( small errors and large or , first lines of Cd , and Hg , Al ( series in ) , In , Tl , and the series of and S. The \mdash ; If the foregoing theory of the constitution of the Diffuse series is correct , it is further necessary , in order to complete discussion , to determine the origin of the first term .
The apparently close relation of the series to the series , and the several cases of collaterals of the former which had been noted with large multiples of , suggested a trial to see if the denominators were multiples of this quantity .
As in cases where satellites are present , the separations depend on them and not on the strong line , it is natural to expect that the satellite is a normal line and the strong lin a collateral , and this is found to be justified by the calculations on this theory .
In Table 1II .
the first column of figures gives the value of the denominator taken from Table II .
The second column gives the factors together with possible variations .
Thus the denominator of .
This has * But see discussion of below .
DR. W. M. HICKS : A CRITICAL STUDy OF SPECTRAL SERIES .
See discussion of BaF\mdash ; two different triplets in quostion .
See below under discussion of observation errors and also possible error due to incorrectness of , i.e. , These give the denominator as .
Now is very close to 2939 , which is given as the approximate value of in Table I. The denominator is then written , and with this new value of the co1Tesponding value of the oun is calculated as instead of the old value , which for comparison is entered next to it .
Notes on the Tables .
Na .
is so small that several multiples of it might be taken for within limits .
Rb .
does not give a of although close to it .
If , however , Rb has satellites , the denominator for will be a few multiples of less than that of That for Cs is less in the corresponding case .
The values in the table are given for and .
Judging from the value of the oun it is probably near .
In any case the multiple would be .
This is a very strong argument that Rb does possess satellites .
Cs .
Neither or give multiples of Mg .
As in Na , is too smafl , and the denominator too large to give anything definite .
Sr. gives the oun clearly too small , although better than in the original table .
, however , gives a value quite close to the probable value .
A similar result is shown also by Cd which occupies an analogous position in the next sub-group .
If behaves in what appears to be the normal manner , it would appear necessary to take the atomic weight to be less than BRAUNER'S value , viz. , in place of , which is probably too large a change to be acceptable .
Ba .
In Barium the first set is doubtful .
That taken above shows no satellites .
The denominator is therefore that for , and this is a multiple which gives a value of the oun much nearer the probable value than that in Table I. Evidence will be given later however under that there is a normal satellite triplet , outside the region of observation , where has the denominator , which , from analogy with the other elements of this group , has a " " mantissa\ldquo ; , and this is again a multiple of .
This , therefore , is probably the corre value , and the other set will be collaterally displaced from this by Ra .
The first lins should be far in the ultra red , and has not been observed .
The multiple is dete1mined indirectly ( see below ) .
Cd .
This element shows the same irregularity as in Sr , in that does not give an exact multiple of , although one close to it .
Here , however , we have to go back to before finding the exact multiple , and gives almost the precise value of the oun as in Table I. , which was itself very exact .
Eu .
is 917794 , which is having the value 18329 of Table If it is exactly , would be , making the oun in place of of Table I. , a great improvement .
DR W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
gives denominator , and the latter factor is which is much closer to the probable value of the oun , and moreover 82 is the correct multiple to give for , which has been taken as a basis for This value of is supported by the discussion of the series below .
Al .
As the order differences are all multiples of , and there may therefore be some doubt as to the real existence of satellites the values for and are inserted .
The denominators for the two only differ by , or the observed by 96 .
As the differences can only refer to the set , it would seem that these should be taken as the normal lines giving as the value of the oun .
In .
Neither nor are exact multiples of although they are very close to is , or .
If these be taken as multiples of the oun , they give the oun as and in place of of Table I. , but the multiples are too large to found any conclusions upon .
It would rather seem that there is some displacement from a typical multiple .
Using as given in Table I. , viz. , 37684 , .
So that and .
If it is exactly , becomes less and the oun in place of .
The value of is therefore insertffi .
If the typical term were higher , the order differences would run , in place of \amp ; c. , and hence more in line with others .
Tl .
Neither the observed nor the supposed collaterals are multiples of .
They are expressed as multiples of .
Although they are large multiples , their values are quite definite provided we know a priori that the denominators are multiples as a fact .
If the multiple be altered by umty , the resultant quotient cannot come within the limiting values of the oun .
If the normal , the order difference over would be , and , so that the order difference would come out as usual a close multiple of .
All this group seem to show the same kind of irregularity .
O. There are three separate series , see data for Table II .
, differing by multiples of , just as in the order differences .
is too small to test the multiples of the denominators themselves .
S. The , lines for and Se are beyond observed regions .
Sulphur however shows no satellites , and we may surmise therefore in analogy with others that the differences for , , are like the others multiples of or As a fact , is a clear multiple of , and the surmise is justified so far as is concerned .
The value of is not very certain .
Se .
Se apparently has ] lites , and the order differences are only multiples of It should not therefore be expected that or should be a multiple of Nevertheless is clearly such a multiple and is entered in the table .
The table shows that where triplets occur the multiples are those of and not of , except in the case of the oxygen group of elements , in which clearly takes the place of .
If the law of multiples is correct , the values of the obtained in this way must clearly be far more exact than those obtained direct from the separations .
A glance at the deduced values of the oun compared with the former values shows how much closer to the mean value the new ones are than the old , and to some extent this adds to the weight of the evidence .
The cases where the multiples do not appear to enter are those of Rb , Cs , In and Tl .
The case of Rb has been considered above and a natural explanation offered .
Cs , In and Tl have all large values of , in which case we have already seen a tendency for the spectra to depend on smaller multiples of the oun than the .
In the case of Cs , the oun is smaller than the multiple it can give no evidence nor data for the oun .
The case is different however for Tl .
If the oun enters , the multiple can be no other than that given , and as is seen the value of the oun is improved .
All the elements of the Al group show a deviation from the normal type in that the first satellite separations are much smaller for the first order lines than for the second , and seem to point to some displacement .
As the Al orders differ by multiples of , any irregularity in the multiple between the first and second orders does not alter the dependence of the denominator on the multiple of .
In In and Tl , however , the differences go by multiples of or , and any irregularity on them will throw out the dependence of the first denominator on a multiple of .
As was shown above the addition of in In not only produces the multiple , but at the same time shows a more usual march of differences for the orders .
Tn Tl the observed denominator for is less than that for and quite abnormal .
The other anomalies occur in that in Sr , appears to take the place of , and in Cd , is in the ultra red and has not been observed .
The elements Na and Mg must be left out of account because the ratio denom .
must be so large that a number of multiples can be found all giving within observation limits .
Cu shows a multiple , but the theory of the constitution of the series of Cu and Ag is doubtful and must also be left out .
With the above doubtful cases the values for , Ca , Sr , Ba , , Cd , Eu , Hg , Al , and Se , are clearly exact multiples , and the large values of in Ba , Cd , Eu and Hg show that these multiples are real .
This ruIe , exhibited as it is in so many cases , and in by far the majority of the elements comparable , must correspond to a real relation and cannot be due to mere coincidence .
Against the reality of the relation is the antecedent improbability that those elements with the smallest value of should have the largest values of the denominator , as e.g. , in the case of Na and Mg .
possible explanation is that the mantissa is the nearest multiple of to some group constant .
But see also under discussion of the series .
It might however have been expected on this ground that the denominators would be of the form .
But the case of Na is clearly against this .
Its denominator and 11344 is and cannot be a multiple .
It would seem conclusive that the DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
denominators of the extreme satellites of the first line are multiples of or and that explanations should be sought for apparent exceptions .
The and Series .
The relationships between the doublet and triplet sets of the series and between the and series were discussed in [ II .
, p. 51 ] by comparing the differences between the corresponding denominators .
It is now possible to see how , if at all , these differences are related to the oun .
The Series.\mdash ; In the alkalies the differences between the corresponding denominators of the two sets were found to be constant within error limits and of course equal to .
In the other elements in which the series have been allocated , there was always a drop in the difference , which in several cases then remained constant for the succeeding orders .
The values were given on [ IL , pp. 51-53 ] .
They are reproduced here , and it is seen at once how they proceed on quite analogous Iines with successive satellite differences of the series considered above .
The possible errors of the single lines from which they are deduced are given in brackets .
Thus are , , , and the differences are given as 7209 , 3474 .
The higher orders , in which the possible errors are so large as to be themselves multiples of , are not included .
The value of is given with the symbol for the element .
It will be noticed that the more accurate the observations the closer are the differences to the multiples of the oun .
But the observed variations from true multiples in the case of the large separations would seem to point to a difference in the as well as in the .
In any case it would seem that must alter per saltum from order to order , unless the sequence formula is a complicated function of Zn .
2414 Cd . .
VOL. CCXIII.\mdash ; A. 3 DR. W. M. HICKS : A CRITICAL STUDy OF SPECTRAL SERIES .
Note how Zn still affects .
The variations from multiples in Hg seem to have relation to the transference properly noted above , , from to Al .
0\mdash ; 3 Tl .
In Al the value after the first is 1381 within limits of error for all .
The and Connections.\mdash ; The differences of the corresponding denominators in the and are also given in [ II .
] .
The values are , however , subject to uncertainties due to uncertain limits in both and , in which the are not the same for both .
In the case of the alkalies there seems a very clear connection with the , except in Cs , where as often before enters .
In the other elements it was shown that the sequences.are inverted and the differences are to be taken between the first of the and the second of the P. In Al , Tl , and , there is again a clear relation , but it is now to the denominator differences of the and , or 1329 , 94018 , 2525 respectively , say for each .
In the case of Cd and Hg no clear relation is apparent , although they behave approximately like Al and Tl .
This want of exact agreement may be due to the effect of the transference inequalities considered above ( p. 333 ) in connection with the oun .
The relations indicated above are shown in the following scheme , in which the differences for the and are taken from [ II .
, p. 51-53 ] .
Na . . . . . . .
Rb . . . .
Cs . . . .
( roughly ) .
Al . . . .
Tl . . . .
Cd . . . .
Hg . . .
DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
The In Part I. the symbol was used to denote the series whose limit depends on the values of VD ( 2 ) in a similar way to that in which the limits of the and series depend on UP ( 1 ) .
Where the series show satellites the series in consequence consist of or triplet series with constant separations .
They comprise some of the strongest lines in the respective spectra , but as in general they occur in the ultra-red region they have not received the same attention as the other better known ones .
In the alkaline earths , however , they come well within the visible regions , and show strong sharply defined lines .
They are related also to other strong lines by collateral and other displacements depending on considerable multiples of , and so naturally come under discussion in the present communication .
As will be seen later , the discussion gives the means of obtaining very accurate determinations of the and consequently of the oun\mdash ; as well as of settling other questions .
I propose , howe.ver , not to attempt an exhaustive discussion in the present communication , partly because the main object now is only to illustrate the influence of the oun , and partly because it would seem that a large number of lines which clearly belong to the cycle are related in a manner neither ordinal nor collateral , nor according to RITZ'S combining theory .
* For convenience of reference the wave-lengths of these lines are given in the Appendix , together with short.historical notes .
The Alkalies.\mdash ; The table below gives the denominators for the two first lines in each as calculated from PASCHEN'S and from RANDALL'S results .
BERGMANN'S measurements for other lines are too much in error for the present objects .
The limits used are the calculated vahles of , using the limits given in Table II .
above and the values of in the Appendix .
Na .
K. Rb .
Cs .
5 .
9710 The question that first arises is , do these refer to actually the first lines of the series ?
If , like , the lowest value of were 2 , the wave numbers of the lines would be somewhat above , Na , , 1200 ; Rb , 2060 ; Cs , 4450 .
For Na it would be outside , but the others come within regions observed by PASOHEN .
He gave for lines at wave numbers and .
The former requires a The reIation is extremely common in certain types of spectra , e.g. , the rare gases other than He .
I hope to return to this in a future lication .
" " Kenntnis uItraroter Linienspektra III ' Ann. .
Phys 33 ( 1910 ) , p. 717 .
denominator and the latter , the one apparently too large and the other too small to fit in with the progression of the lines for , 4 .
But the mantissa for the latter is within limits below that for [ see Note 3 at end ] .
In Rb there is a line at wave number which would require a denominator , well in step with the other two .
PASCHEN identifies it as , assuming the existence of satellites in .
It would seem to be more probably the sought for .
There is another line given as or .
If the former is the more correct it gives denominator .
In Cs no line appears with wave number near 4450 .
There are two lines , however , with wave numbers and which differ by , and certainly suggest the doublet series depending on the satellite .
This requires a separation of , and if they belong to the series there must be a satellite with a separation which we should not expect to observe as being too faint .
The lines give a mantissa with a satellite difference 1003 .
The may be , within limits , , a value which in the alkaline earths seems to be closely associated with satellites .
But the mantissa is less than that for , when a larger value should be expected .
Even if not itself it may be related to the cycle in a similar way to certain displacements found in the alkaline earths ( see pp. 383 , 413 ) , and it should be noted that if so there seem to be lines in corresponding positions in and Rb .
They are ( in wave numbers ) the referred to above for and in Rb .
The latter requires a denominator .
In this connection it is interesting to note that PASCHEN makes the remark that this line at times shows itself double .
The separation calculated from his numbers is , giving a denominator difference of 107 for and , i.e. , for and .
This would indicate a sort of incipient satellite in .
These considerations seem to show that there is some likelihood that does not give the first line of the series , and they will be felt to have greater weight when the curious irregularity in the of the alkaline earths to be noticed immediately is taken into account .
The question is further discussed on p. 397 in connection with the other elements .
The next question is , is there any indication of satellites in accepted lines ?
If so we should only expect to find it in Cs .
Now RANDALL gives weak lines close to , and close to .
They look like satellites only on the wrong side .
The first changes the denominator by 2000 , which is within easy limits of .
It will be shown that this is a common satellite difference in the alkaline earths .
Further , it makes the denominator , thus bringing the values for , 4 , 5 , 6 in order , which is not the case in the table above .
There is , , something to be said in favour of taking the normal doublet to be at , and that that is then collaterally displaced by to the stronger lines .
It is also quite in keeping with in the alkalin earths that a similar displacement is not shown in the case of the first lines ( if are the first lines ) .
DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
Group II .
The Alkaline Earths.\mdash ; The series are most fully and regularly developed in Ca and Sr. In Ba and Ra the configurations which give rise to the normal type seem to be so modified that displaced lines become common , and in cases the normal line has disappeared .
On the other hand , Mg seems to range itself with the sub-group .
It will be best therefore to deal with Ca and Sr first , and as they are built on a precisely similar plan to consider them together .
The following table gives the wave numbers of the series together with certain others which are clearly similarly related in the different elements .
The separations are indicated by thick figures .
The wave-lengths are given in Appendix II .
Ca. Sr. 5854 5925 10047 Analogous Sets in Ca and Sr. 14 .
Ba . ?
Ra .
( 1 ) .
( 2 ) . .
It is clear that these lines also show satellites .
Also , it is curious that the first sets of triplets apparently have the lines corresponding to the second separation displaced below those forming the first .
Thus in Ca the second set ( giving ) have not been observed , in Sr the two are separated by a gap of 143 , and a similar effect will be found later in Ba .
Owing to this fact , the formulae constants are calculated from the 2nd , 3rd , and 4th sets .
They give for These give the following values of .
2 .
6 .
7 .
Ca . . .
22 Sr . . .
\mdash ; 42 05 02 The agreement is good , except for , and in this case the agreement is sufficient to show that the allocation for is correct .
The limits are close to those of , which is not known with great exactness because the values of or given in [ II .
] for the second group are subject to possible errors of some units .
With formulae in the values of are given [ II .
, p. 36 ] as 3399 @5 for Ca and for Sr , whilst with formulae in , the respective limits are .
The values of deduced from these are respectively with and with .
The limits , therefore , found above for lie ch between their corresponding values as deduced from the series direct .
Assuming that the are more accurate , the values of deduced from them are for Ca and for Sr , in both cases close to the mean of those in [ II .
] .
If the series depend on formulae sequences , these limits may be taken as close to the correct values .
If , however , the different orders proceed by multiples of or in the way illustrated in Table II .
for the series , the limits may require modification by a few units .
As the separations of the series depend on the separations of the satellites of the first lines of the series , and these depend on displacements by definite multiples of , as given in Table II .
, it is possible to calculate the values of the former with extreme accuracy .
Table II .
gives and as the multiples in question for both the elements Ca and Sr. Using the values of and of the denominators of there given , the separations in question calculate out to for Ca and for Sr. These may be regarded as exact to the 2nd decimal place and independent of any possible variation of DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
In the case of Ca the observed lines , with differences somewhat less than and , seem to indicate the presence of close satellites .
If is really , the separation of the first satellite is , with possible errors , which form a very considerable proportion of the .
total amount .
A displacement of produces a separation of and it may be this .
But has an excessive intensity for a satellite lin , viz. , 6 , as against 4 for , and , moreover , it may possibly be the collateral which gives with .
If the latter allocation is correct , it would hide , which should be , giving a separation of due to which gives .
The same considerations applied to the second set give a separation of for the first satellite , in which again gives , and 22 for the second , giving 32 .
The separations are so small that no certain conclusions can be drawn as to their origin .
The actually observed numbers may be due to and , but and are just possible [ but see Note 4 ] .
For Sr the first doublet is useless , as the line is due to the early measurements of LEHMANN , which are affected with considerable errors .
The observed separation for and gives instead of something less than .
The second triplet gives for the separation of the second satellite from the first and for the separation of the satellite of the second line of the triplet , and from analogy with other satellite series , this would be the separation of the first and second satellites of the first line .
Differences of and give separations of and , so that it may be concluded that the satellites depend on these differences , a conclusion supported by the fact that a similar result is indicated as possible for Ca. Returning to the curious fact noticed above that the first triplets of the series seem to be dislocated , the second fragment in Sr is found at a distance below its normal position .
For the present we note this can be explained by one of two possible collateral displacements , viz. or , where stands for the normal or .
The case of Ba below will give evidence in favour of the latter explanation .
In addition to the lines of the series itself , there are two sets of triplets and a doublet which are clearly analogous in the two elements .
They are given in the list above , following the series lines .
The first triplets in each are curious as having the middle line the strongest .
* They are also related to others in the way indicated in the following scheme:\mdash ; ( 8 ) 17847.46 21.94 Ca ( 8 ) ( 8 ) 1785650 ( 10 ) ( 8 ) A similar peculiarity has already been noted in the associated OD series .
Sr ( 6 ) ( 8 ) ( 10 ) 21.57 Ca ( 8 ) ( 6 ) 14.01 ( 4 ) 1900 11 ( 10 ) Sr ( 10 ) ( 8 ) 59.75 ( 4 ) 1917 73 ( 6 ) ( 10 ) 56.61 ( 8 ) ( 6 ) 1899 25 ( 8 ) ( 8 ) The first lines of the triplets give as denominators , supposing the true limits to be Ca. Sr. 1st triplet 2nd , , where lie between .
Clearly the differences are multiples for the two elements respectively .
The first lines of the doublets give for Ca and for Sr These differ from the denominator of by Ca. Sr. . . .
That is , they apparently differ by respectively .
We shall see shortly that the best value for makes the relation for Sr very exact , whilst that for Ca is more doubtful .
DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
The following list contains the wave numbers of certain lines related to the series in the two elements with the denominators appended for the chief lin 13.92 21.75 20.04 99.09 .
VOL. CCXIII.\mdash ; A. From these we find the following differences , denoting the mantissa only : Ca. Sr. . . . .
within limits . . . .
of of . . . .
7 of of . . .
562 88 H-F . . . . .
126 33 D-C . . . . . . .
, ( 2 ) . . . . .
158 89 . . . . .
158 89 E-D . . . . .
169 Also the first triplet A in Ca shows the same kind of dislocation as in of the other elements .
The dislocation is , corresponding to a denominator difference of 931 , and is 930 .
on 13133 may be as calculated from the formula .
Allowing for the satellite difference DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
The number of the cases where multiples of enter , as well as their appearance in the corresponding position in the two elements where corresponding lines are observed must produce a conviction that they represent real and not chance relations .
In the case of Ca it makes close to corresponding to .
If has any but a very small value , the first two multiples are upset , but these may be due to chance .
If be made as suggested below , will be about with , which is considerably greater than the most probable value .
The probability is that can only be a small quantity , changing by A similar reasoning applied to Sr rather tends to show that here the value is to be preferred .
It makes the first multiple , and the other values of become , and 5556 gives ) , the uncertainty of this being due to a possible error in the atomic weight of , with the first relation gives Again the most probable values of the denominators of are Ca Sr The numbers on the right are practically equal .
If analogous relations are found in Ba and Ra i.t points to the existence of a group constant about 937300 .
On the other hand it would seem that the denominators of are , like those of multiples of also , for denominator of , , and , 5 in Sr llakes in line with those above .
The denominator of is less than that of , .
, , Which of these two interpretations is the more likely must be left until cases of Ba and Ra are considered .
It should however be noted that there be some uncertainty as to what lines really represent , or , i.e. , as to which of them the multiple law is to be attached .
here remains to consider the question of the real limits .
The reasons for supposing them to be are so strong that it is necessary to see whether the values obtained direct from the series , and those required in Table Tl .
cannot be brought into agreement .
If the series possess what has been called in [ II .
] a formula sequence , the values obtained for above cannot be more than few units in error , in this case it * Calculated from formuln .
The observed is probat ) the } tiou with is the fnll } 3 .
88 DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
must be possible to raise the limits for to agree with those calculated from .
That is , to raise that for Ca from to and for Sr from 3102725 to , or Ca by 800 and Sr by or thereabouts .
It may probably be possible to find numbers near those which would still make the order differences of Ca Sr multiples of , but only by supposing that the successive mantissadifferences in the series after rising begin to decrease with higher orders , which is ainst the rule in other es .
This is so far an argument against this way of reconciling the different values of the limits .
If however the order differences in the series behave in a similar manner to that considered above for the , i.e. , by multiples of or , the exactness of the found by means of a formula is no longer so close , and the question becomes one of seeing if , when they are made less for and for Sr , it becomes possible to arrange the denominators in the same way .
If the attempt be made to reduce by 8 in , a similar objection to that raised above will enter , viz. , the successive mantissa-differences after falling begin to after .
If however a reduction of about be made , reducing the limit to that found in [ II .
] for , the order mantissae differ successively within observation limits by .
Further , in the case of Sr a fall of produces a similar fall and rise in successive denominators .
If however be put the mantissae differences become within limits .
If this is justified , it is curious that as in the series where there are no satellites , the diffel.ences proceed by multiples of the same rule should hold for , where satellites are at least not certain .
The difficulty can only be stated and the solution left open .
It is possible that the order differences must be compared from the of one line to the of the next , for which there is evidence in Ba and Ra .
.\mdash ; In discussing Ba we start under the disadvantage that the lines belonging to , with the corresponding satellite separations have not been observed , for the ultra-red doublet treated in the discussion .
on the series does not seem to belong to the normal .
Moreover , the observed lines which are clearly related to the series are so dispersed by collateral displacements that it is questionable whether it is possible to arrange a series proceeding by an algebraical sequence as in the other cases .
The lines exhibited in the table above run on parallel lines with the corresponding lines in Ca and Sr , and are clearly closely related to the uccessive orders of the series , even if they are not the typical ones themselves .
An attempt to obtain a formula from the first three gives a limit , and gives value of the wave number for of close to the strong line It is behind the strong line , which indicates that the last is probably normal , and makes the normal about 250 behind .
This is in fair order with the march of the others .
We may therefore feel justified in settling that the limit of is near .
The separations are close to 260 and 157 , they are therefore based on satellite differences in the series of and is so DR. W. M. HICKS : A CRITICAL STUDY SPECTRAL SERIES .
large that there can be no doubt .
These numbers are in analogy with the values for Ca and Sr , viz. , and , are in the usual ratio 5 : 3 , and stand to the observed values for given in Table II .
in a similar relation to those in Sr. Now must be , and if the general rule found above that the mantissa of is a multiple of holds , it is possible to obtain a very accurate value .
As a fact , with the mantissa of is very nearly the multiple of .
If it is made so exactly , taking , then becomes about 25922 .
This value , with the satellite differences of , give and it is seen how close the separations come to those observed .
If we put , the mantissa of The lines calculated from these and found above give the following scheme in lengths on ROWLAND'S scale in The last one only comes within the region in which RANDALL'S ultra-red lines lie , his longest wave-length being 29223 , belonging to the doublet treated as a possible line above .
If the rules employed are valid , these values can only err by a few units .
The denominator of is .
In the cases of Ca and Sr there is apparent the existence of satellites , viz. , and , and in both cases the mantissa of one of the sets thus found are multiples of .
If this is general the mantissa for will be .
This is 77 , sufficiently near to give some weight to the allocation .
If this is and the value for given above becomes , giving a valus of within limits of error .
makes the oun with .
The satellites thus constituted would , if existing , have separations for of and , and for of and \mdash ; but they have not been observed for ( 2)\mdash ; as indeed is the case in Ca and Sr the first set in which show separations of the full amount .
The curious dislocation of the second half of the triplet from the first seen in Sr shows itself here also .
The analogue appears to be shown in the triplet coming next in the list which appears to have kept its first member and satellite .
The displacement is This cannot be due to a displacement in , for if so the separations due to , would be considerably larger .
If it is treated as a displacement on VF the denominator difference is , while .
It is probably therefore this .
The corresponding displacement in Sr was found to be The separation between the first line and the satellite is , the satellite being due to LEHMANN whose measures are not very accurate , it may well be corresponding to a satellite difference of .
The lines may therefore be represented , .
If the next two lines are correctly allocated , 18686 should have an unobserved satellite with a difference .
This would make or ahead of the satellite , so that this supports the allocation .
The line corresponds to a satellite with his makes , on the supposition of satellite differences of or ahead of the satellite , the satellite being absent .
The line for appears to be displaced to 22706 .
The value calculated from the rough formula gives a line 250 behind the strong line , clearly showing that the latter is a line , and 22706 is very close to a displacement of on the calculated .
If this be made exact the undisplaced line would be at behind .
This is within limits of Hence has been altogether displaced to , and is .
For the formula gives .
There is a doublet at , with a separation , and no others in the neighbourhood .
If these are the displaced , the normal would be and and the observed lines and .
The calculated normal lines have separation , or practically .
A line at is ahead of the calculated .
It is therefore the undisplaced .
There are a large number of other lines clearly related to the type .
Their complete discussion would require a more searching investigation than can be given now .
Several sets are related in a manner which is quite common in spark and rich arc spectra , indicated by the fact that a number of lines 11ry differ in succession by nearly the same separation\mdash ; a kind of relation which cannot be due to collateral displacement by equal denominator differences .
There are a few also which seem to be attached parasitically to and lines .
There may be uncertainty also as to DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
whether the separations shown which differ from 260 and 158 , differ through a satellite effect , or by successive collaterals of and .
For instance , putting all which separations occur .
In the lines now to be referred to , however , the separations will be supposed to owe their defect from 260 to the satellite effect , and thus treated it is clearly seen what an important role the term plays .
Amongst the ultra-red lines observed by RANDALL* appear the following in wave numbers :\mdash ; ( 70 ) 938 55 159.93 ( 60 ) 1 .
( 60 ) 977 54 ( 95 ) ( 5 ) 980 159.76 ( 60 ) 996 527 ( 3 ) The figures in brackets before the numbers give intensities and those after the estimated maximum errors .
They clearly belong to the cycle , and show within error limits the normal separations .
The run of the intensities would point to negative values , with the first four respectively for , but also the 2nd , 4th and 5th might be , whereby 9771 would not come in and he small intensity for would be abnormal .
On the first supposition , gives a denominator and .
On the second , gives a denominator .
Now or denominator of In the following the wave numbers of some sets of lines with their separations are given .
The low frequencies have been observed by LEHMANN and by HERMANN and HoHlLLER .
LEHMANN gives many weaker ones not observed by HERMANN and vice versed .
LEHMANN'S observations were earlier but are not nearly so accurate as those of the others .
2 .
3 .
4 .
5 .
'Ann . .
Phys 33 , p. 745 ( 1910 ) .
DR. W. M. HICKS : A CRITICAL STUDy OF SPECTRAL SERIES .
6 .
7 .
8 .
9 .
255 .
lO .
The following numbers give the corresponding denominators calculated from .
Where the separation differs from 260 , the satellite value\mdash ; or , which is the same thing , the denominator for \mdash ; is inserted as well , but , in order to distinguish it , it is printed further to the right\mdash ; the changes due to being the same for both .
( 2 ) @3 11895 ( 11969 ) ( 3 ) 108209 12039 ( 11982 ) ( 4 ) ( 5 ) ( 6 ) ( 7 ) DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL RIES .
( 8 ) ( 9 ) ( 10 ) \mdash ; @\mdash ; 14 852480 11932 11962 11970 11960 ( 10 ) The differences are given in thick figures .
The last column gives the corresponding value of without regard to observation errors when .
In the case of ( 3 ) if the differences be referred to a hypothetical , displaced from 14141 , the two abnormal values come to 11969 and 11982 .
It will be remembered that we had an indication aboye of with in treating both and as depending on multiples of .
It would seem , therefore , that the value of is close to and the value of The actual differences of successive denominators in the normal series may thus be represented:\mdash ; in which the last column also gives the value of , where The first multiple of 81 with addition of suggests ( 1 ) a real , displaced from 13089 , or that there is a normal type about behind .
The latter may well not be a typical line since it makes the exact separation 260 with .
There VOL. CCXIII.\mdash ; A. 3 is a line by LEHMANN at , but its collateral displacement cannot be within any likely limits of even LEHMANN'S measurements .
As to the second supposition , there is a line at by LEHMANN which gives denominator This gives a difference with of , which with would make .
It would thus appear that the normal line is 12992 , and the system receives a double displacement , first to 13089 , and again to 13471 .
The mantissa is .
The addition of makes it , which with is 937529 , well within error limits of the same quantity in the case of Ca and Sr. Again we are met with the apparent simultaneous existence of two explanations which cannot be compatible .
the true explanation that the typical first line is , but that the corresponding configuration is not very stable and transforms to one depending on the nearest complete multiple of .
Certainly such instability is indicated in Ba .
.\mdash ; The discussion for radium is rendered even more uncertain than that for barium , in that the ultra-red region has not been observed , the process of disintegration and re-aggregation has proceeded further , and , in addition , there is some uncertainty about adopted .
and PRECHT'S plates were only sensitive up to 6500 , and EXNER and HASHEK give only two lines above this , and , both of which belong to the cycle .
The number of lines , however , coming within this cycle is very large , but a complete discussion would involve the consideration of the new kinds of relationships referred to under barium , and cannot therefore be undertaken here .
It will be sufficient to deal only with some generalities , specially bearing on the series proper , which will also give some further light on the general series .
There are a large number of triplets with separations in the neighbourhood of 692 and 432 , which are roughly in the proper ratio 5 : 3 , allowing for.the fact that the actual separations must be larger .
Those in the table of lines above are roughly parallel to the , and give a limit somewhere about 24520 .
would therefore be about this , and more than larger .
The denominator of should be a multiple of .
Using the most probable value of , it is found that the denominator comes out very close to a multiple of .
If this be made exact it is found that .
The value of is then taken .
The values of are so large that there can be no ambiguity about the multiples to be chosen to give the separations , viz. , .
These multiples march well with those for Ca , Sr , Ba .
The separations resulting are , with and .
If we apply the rule shown in the preceding elements for , the denominator is .
Satellites depending on would give separations and , and the fact that these separations occur in connection with renders the tifioation of that for rather doubtful , a doubt which is increased when we test the allocation by the law in.d.icated above that the DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
S95 differences of the denominators of is always the same multiple of , as is done below .
The line 19897 has separation , and therefore should have a satellite ( too faint ) above it .
Allowing for observational errors on 19897 this is on the denominator .
The following scheme will then illustrate the law of formation :\mdash ; 9S , 9S , 9S , 9S , 9S , 9S , 9S , 9S , 9S , 9S , Calculated Satellite , 34312 In the above the first is the denominator calculated from .
It is affected with an uncertainty of about 400 on the 937300 .
The line has a separation with , and therefore should have a satellite above it .
Its denominator difference is 6493 behind and .
The line is associated with 17300 .
Its denominator is 16949 behind that of 17300 and 16695\mdash ; the same within limits .
The satellite of 19897 is displaced There is no evidence of a satellite behind it , but the difference of is made with this suppositious one .
It is seen that a value of about makes these the same within limits .
The corresponding values of are appended in the last column .
It may be taken that the discussion has established that the satellite differences in the lines are and .
This is the only result of which there can be certainty .
The Sub-group.\mdash ; Using the limits given in Table II .
above and the corresponding values of , the limits come as follows :\mdash ; . . . .
, with separations Hg 1275 07 Cd . . . .
PASCHEN allots the following for Zn and Cd , viz. , in wave numbers:\mdash ; 3.05 2.43 11.98 ( 4 ) 859 57 The line 8595 must be allotted to , because it makes the difference 266 with 8872 , so that the two are lines connected with and the companion to the line 267 above it .
This relation has already been discussed under the series above .
For Hg he assigns 5814 to , 5843 to , to , to , to .
Now 5843 is double with , corresponding to a displacement , and may well be , whilst 5908 is .
Again , and are separated by , which is so close to as to indicate that they are and and that has not been observed .
With these allocations VF ( 3 ) and VF ( 4 ) and the denominators for the lines calculated from the first lines ( except ) are Cd .
Hg .
The differences of the mantissae of the two orders in Cd and Hg are Cd .
Hg .
; ; in which the probable variations of are small fractions .
In fact , the greatest uncertainties are due to observation errors .
There is not much material to throw light on the origin of the term here , nor in fact is there evidence that the fundamental lines , or the first lines , of the series are On ths basis of Rrrz 's combining theory PASCHEN gives the following allocations ( lines in waveleIlgths ) There can be little doubt about the correctness of this allocation .
Using the value of [ II .
] 40139 55 and the wave numbers 3320019 , of the first lines of each set there results , which are practically the same as those found direct .
DR W. M. HICKS : A CRITICAL STUDy OF SPECTRAL SERIES .
for .
Any lines corresponding to would have wave numbers about 500 , and to negative wave numbers in the neighbourhood of 16000 .
Now PASCHEN has noted lines which may be treated as the actual lines in question .
They depend on terms where is the limit of his singlet series , and of course VD ( 2 ) is the of the above .
Using sequences for the series of the form , the limits of the series are Zn .
Cd .
Hg .
The lines in question are .
Cd .
Hg .
If these wave numbers be added to in each element , there results and , i. e. , the value of the aboye .
The corresponding lines for do not seem to exist .
There is no a priori reason to take rather than for .
In Cd , however , the case is settled in favour of , as the other lines exist , , giving the differences corresponding therefore to the companion series to , to and , not appearing .
But in Hg , with differences would seem to assign 17327 to the term .
Nevertheless to get the limit of PASCHEN'S series it is necessary to take .
If these be regarded as the first lines of the series , the denominators are 943072-3 5 ; Cd , 949840-3 9 ; Hg , .
In Hg the Iine would seem to stand in a normal relation to the , as it comes into line with the others as is seen below .
With this the apparent limit with would be , giving denominator .
The question now is , are these denominators related in any way to those for .
The differences of their mantissae are , using our new Hg lin Cd .
Hg .
well within errors , it being also remembered that can only be a fraction .
The value of for Hg adopted is the corrected one 29765 , from .
This is a striking connection .
It shows that the limits for PASCHEN'S singlet series are either VF ( 1 ) or are formed from VF ( 3 ) by deducting for for Cd , and apparently for a normal type in Hg which then receives some displacement .
DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
Magnesium.\mdash ; We are now in a better position to take up the consideration of the place Mg is to occupy in the second group of elements , viz. , whether it is allied with the Ca or the Zn sub-groups .
In the discussion of ( p. 356 ) there was evidence in favour of either view .
If it belongs to the former , then the line is ; if to the latter we have PASCHEN'S allocations of 14877 to and to .
Take first the supposition that Mg is analogous to the earths .
In this case , 6719 being the wave number of 14877 .
If the series is formed on the type of the Ca set the denominator of the first line will be .
This gives a line .
No line has been observed sufficiently near to this to be identified with it .
In the other case .
PASCHEN'S allocations then give denominators with a mantissa-difference .
With this is 3717 or .
The value will upset the difference in Table II .
between and the supposed which in this case does not exist .
It still leaves the difference between the denominators of and .
If Mg is completely analogous with the Zn set the combination lines , should exist .
They should be at , , and , .
Now EDER and VALENTA give two spark lines of weak intensity at , and SAUNDERS a weak arc at 3051 .
The wave numbers in vacuo are 3276948 and separated by , which is clearly .
These are therefore the looked for , and , the third not havin been seen .
As to the other set , SAUNDERS has observed a line at 2833 giving which is clearly .
The existence of these combination lines seems to settle the question in favour of Mg belonging spectroscopically to the group of metals rather than the alkaline earths .
It is possible that as a transition element it belongs to both types .
Judging from PASCHEN'S various readings it might well be that is double so that one might be and the other .
Group III.\mdash ; In Al and Tl alone have the ultra-red lines been observed , and here the lines are found in a similar position to those in the groups , and with them RITZ'S combinations and -VP ( 3 ) .
Using the values of of [ II .
] the values of are for Al and with separation for for Tl .
For Alumimum PASCHEN gives and for and , from which result VF ( 3 ) and VF ( 4 ) .
The combination gives VF ( 3 ) .
For Thalium PASCHHIN gives , for , and , for , .
These give separations and instead of , but the same within limits .
He also gives , for and for DR W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES . .
The possible errors of the latter , however , are so large that they cannot be used to improve the values foumd from the direct lines .
The limit calculated from of [ II .
] and is , but there is some uncertainty owing to the abnormality of as explained above under the discussion of the series .
The lines and give for and 444174 ( 37 ) for VF ( 4 ) .
In the case of In no ultra-red lines have been observed .
In K.R. list there appears a doublet , which shows a separation , the true doublet separation being about .
Its relative position in the spectrum compared with that of Al and Tl point it out as the Ritz combination . .
also give a line at or , which its position might be -VF ( 3 ) .
If so , the value of VF ( 3 ) is and clearly in line with those of Al and Tl .
We shall adopt it provisionally .
K.R. mark all these lines as doubtful , but the existence of the doublet separation points to their real existence as In lines .
Collecting these give the following:\mdash ; Al .
In .
Tl .
Denom . . .
VF ( 4 ) Denom . . .
It is seen that Al and In may be the same within limits .
In Tl the uncertainty in referred to above is such as to raise the limit\mdash ; and by the explanation there given is correct .
A rise of 10 would make the denominator for the same as for Al and In .
One of the most striking results of this discussion of the series is the distinct divergence in type between the spectra of the high melting-point elements and those of the low melting point , and at the same time the close resemblance between the individual elements in each division .
So close indeed is the resemblance between all the low melting-point elements of Groups II .
and III .
that the differences between them appear to be almost wholly due to the difference of the linits , or the value of VD ( 2 ) and the values of VF ( 3 ) are almost the same .
To see how closely they agree the denominators to four places of decimals are collected here , and for comparison those of the alkalies .
, Cd , Eu ?
Hg , Ga ?
In , ?
Tl , Na , Rb , Ca It is seen how the elements in each group agree in spite of a very vide difference in atomic weight , and moreover the mantissae in all are very close to unity .
DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
It was shown in [ I. ] that the series of the alkalies could be represented by a series of the form .
The same is the case with the Al and groups .
As is so small and varies so little it can scarcely be a function of the atomic weight .
The atomic volumes of the elements are much more even and it may be a function of them as is the case the -sequence .
In fact , in the case of the alkalies , the are not far from bein proportional to , for the four elements considered , but the data are so inexact and uncertain that it seems not worth while to undertake an exhaustive numerical discussion .
We do not know that the chief lines of these sets are those depending on If lines exist depending on they would all be in extremest red , in fact with wave-lengths comparable with those of electro-magnetic waves capable of being experimentally excited , and it is possible that VF ( 2 ) might be the same for all low melting-point elements and as for He ( see [ I. ] ) .
For we should expect the lines of negative wave number in regions which have been observed and in which no such have been seen .
The series in the high melting-point elements , on the contrary , are profoundly influenced by the atomic weight term .
Either the lines observed belong to a different type from those of the others , or they are based on a normal type of aggregation which is modified by collateral and other types of displacement due to the splitting up of the typical aggregations , or to a more complex system of new aggregations .
The notation for these series was adopted in [ I. ] under the idea that the sequence for it was of a more fundamental nature than the others , and that impression is rather strengthened by the present discussion .
It has been seen , for instance , how the limits of PASCHEN'S singlet series in the Zn group depend on it .
It would be interesting to know whether similar series appear in the alkalies and alumimum group .
The Value of the Oun .
The fiirther knowledge now gained as to ways in which the oun\mdash ; or the \mdash ; enters in the constitution of spectra , enables a much closer approximation to its actual value to be obtained than was possible from the consideration of the doublet and triplet separations themselves .
Amongst the principal aids are ( 1 ) the separations themselves , ( 2 ) the dependence of the first denominator on a multiple of , ( 3 ) in the triplet elements , on the collateral relations between associated lines , ( 4 ) the satellite separations in the series , ( 5 ) the order separations in the series show no satellites , ( 6 ) collaterals depending on .
Of these , Nos. ( 1 ) and ( 4 ) have the great advantage that the values depend only slightly on the exactness of the.limit ( the value of ) , but ( 4 ) has the disadvantage that only small multiples of are involved , and ( 1 ) only itself .
There are also various uncertainties which show themselves when a high order of accuracy is desired\mdash ; chiefly in the elements of the 3rd group .
No. ( 2 ) has the great advantage of giving considerable multiples of DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
but they depend to a larger extent than ( 1 ) and ( 4 ) on the exactness of the limit .
This inexactness is , however , in general more than compensated by the largeness of the quantities dealt with .
Collateral relations also are capable of giving very exact values , but always subject to uncertainty as to the actuality of the relations indicated by the numerical coincidences .
This is less apparent in the series of the high melting-point elements in Group III .
, where the relations are largely established by analogy between the different elements involved .
No. ( 5 ) is affected by the exactness of the limit , and is only useful when the separation is taken between the first two orders and it is a considerable multiple of , as , for instance , in Al .
For the special purpose of obtaining as exact a value as possible of the ratio it will be better to exclude from consideration Na , Ga , He , Sc , , and Se .
Na is excluded on account of the uncertainty as to whether F. and P. interferometer measures of the lines are to be taken as giving the value of for the and series , in which a somewhat larger value is indicated by observers using ordinary methods .
Ga is omitted on account of its poor spectroscopic data .
He because its although very accurately determined , is so small that slight errors are very large proportionate ones .
because is small and the observations not so exact , and Sc , , and Se because their spectra have not been sufficiently discussed .
There remain 17 elements for consideration .
In the following the case of each element is considered first , with estimates of its possible error .
Then using these possible errors as probable errors , the most probable value of is deduced by least squares .
The ratio is denoted by K. The observations determining are very bad .
The adopted gives and gives , and is about .
The value of from this is .
This is adopted with probable error Rb .
The only source is from , since there is no light from the as the satellites are doubtful .
The value in Table I. is .
Probable error taken Cs .
Table I. gives is so close to that it seems justifiable to adopt it .
The observations seem to show thab should be in Table II .
The denominator is subject also to an observation error of 228 .
With the consequent value of is , but this value of makes the former value much less .
The relation may be a coincidence , as it ought to be near it , and it will therefore be safer to take the first adopted value , Cu and Ag as in Table I. , viz. , Mg .
With gives The actual first line has been seen to be uncertain , and in any case so small that the actual multiple cannot be obtained .
There is an order difference of between and 3 , but the observation errors , and those due to , give with far less exactness than from .
Value therefore adopted , VOL. CCXIII.\mdash ; A. Ca. From The denominator of gives From the discussion of Mantissa difference of and , the great uncertainty being due to The most reliable appears to be that from , and is included in the others .
Value adopted , Sr. From The denominator of does not appear as a multiple of , whereas that of If this is a real relation , From the collaterals and the denominator of , cannot be far from Adopted value , Ba .
From , with , as modified in Table II .
, and allowed , and From the collateral From , as found from the series , From the discussion , The most reliable is probably the mean of those depending on .
Adopted value , Ra .
From From the discussion , , but as there is some uncertainty in the theory , the limits of error should be greater .
Adopted value , .
From , with From , , and from the values lying between , and .
Value adopted , Cd .
From , with In the theory Cd appears to occupy a similar position to that of Sr in the other sub-group , in that the multiple of is carried back to or .
The most accurate is that from .
Adopted value , Eu .
From , with From , .
Adopted value , .
From .
There is some uncertainty as to the ratio of best agrees with the transference value from to discussed at the commencement , and it gives a value of which is in close agreement with the value found in the discussion , viz. , HoNlGSCHMID Sitz .
Akad .
Wiss .
Wien , Novomber , has recently made a carefuI determination of , and gives 225 97 in place of 226 4 .
This would make the oun 363 value quite inadmissible if the spectroscopic data are reliable .
Altbough they are not good , they can hardly be so in this value of would indicate .
For what it may be worth , the spectroscopic data would seem , therefore , to weigh against the acceptance of the new atomic weight , DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
The former gives with , and the latter with a large uncertainty owing to gives .
The large possible error is due to the inty in the atomic weight of Hg .
Adopted value , Al .
In Al also there is a large possible variation due to the uncertain weight . .
From .
There is an order diffelGnce between and 3 for the series ( see Table II This gives .
The denominator of gives , or if there is a satellite gives .
Adopted value , In .
From , with From , if , but the theory is uncertain .
Adopted value , Tl .
From .
From , , but with somewhat doubtful theory .
Value adopted , These values for the 17 elements hted according to the possible errors now give .
This is the same as our first approximate value , but its probable error is much less .
If the determination of the value depended only on questions of errors of calculation and of observation in spectral and atomic weight data , ) number would probably be extremely close to the actual one .
It must be bered , however , that our theory of the constitution is not yet complete .
instance , in [ 11 .
] it was seen that the supposition that was not constant for the -sequence , but that the value for the first line was slightly larger explained the introduction of a term in the denominator .
A similar explanation might } ) the fact that the value of appears to deviate from the mean by about the amount in each group of elements , and if it were justified , the value of bove would receive a slight modification .
I believe it will be found ultimately true value will lie within the limits given by or If the existence of the oun as a definite proportion of the ( atomic weight be considered as.established , the best and most direct method of determining the value of the factor would be from the discussion of an element in which the spectroscopic data are good and in which the atomic weight has been determined with great accuracy .
For this purpose we naturally turn to silver .
Regarded as the ultimate standard of atomic weight determinations , no error in the atomic weight enters\mdash ; the value of is determined in terms of .
Moreover its separation is large , so that any error of measurement is a small fraction of its total , and in addition the actual error is extremely small .
It is therefore tantalising to find that the lines , excepted , are not susceptible of such exact measures as in many others , that the typical series are not well developed , and that in fact there may be a doubt whether the lines generally accepted as the series follow laws altogether analogous to those in other groups .
In KAYSER and RUNGE'S measures four lines are assigned to and three to ( 2-4 ) , the possible errors for being much less than for .
Usin for and calculating the formula constants from the first three , there results for the series with the large possible variation in of .
Using this value of the limit and calculating the formula for , the line is reproduced with O-C being 5 .
If , however , the lines be used with the limit comes to and the fourth line is reproduced within limits .
and cannot be the same within error limits .
If and the limits can be the same , viz : , with on , but cannot be if the doublet usually assigned to belongs really to a series .
I had intended to supplement the direct determination of from by a discussion of collaterals , of which both Ag and Cu afford a large number .
The doubt however about the form , and the presence of the numerous collaterals , gives a suspicion that the series are related to the series with limits based on the typical , that they are analogous to the terms in the high melting-point elements of Group IL , and that the doublet usually allocated to the series is really analogous to the triplets with negative wave numbers found in Ca and Sr. That discussion is therefore held back for the present .
* But certain points not open to doubt and forming a portion of the work of the accurate determination command a place here .
The lines are sharp .
and have measured the lines and with their interferometer .
Their measures give a separation of In order to get full advantage of their accuracy , and to avoid the uncertainty due to the last significant figure it is necessary to use logarithmic tables with more than 7 figures .
This has been done on the supposition that and 's errors are not larger than001 , i.e. , unity in the seventh significant figure .
.The old measures are sufficient to show with certainty that the satellite difference is , and the old approximation to will give with an inappreciable error , whence can be found .
Taking the mantissa of is and .
Hence the mantissa of is the difference or calculated from this is giving with the value in which the correction for is only effective if .
The value of calculated from this is , in which the uncertainty of 20 is due to the uncertainty 001 in .
This value is 4 less than that of Table I. obtained by supposing K.R. values for and had no errors .
It gives .
With the mean value suggested above this becomes The foregoing is interesting also because it shows how the application of the laws developed in the present discussion can help towards more accurate determinations of * The value obtained for from the collaterals was , which with , as indicated in the text , gives a value surprisingly close to that deduced from all the elements combined .
DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
quantities involved .
, For instance , it has enabled us to obtain a value of correct to about a unit in the sixth significant figure .
In the case of Au , the knowledge is still more fragmentary than in Ag and the value of has not been determined .
By the application of our new laws , however , it is possible to obtain a good deal of information based on evidence of weight , and it will be interesting to consider it shortly here .
Although the spectrum of Au shows many analogies with those of Cu and Ag , no lines have been assigned to the or series .
There is a strong doublet in the ultra-violet analogous to the lines allocated to the series in Cu and Ag .
There is only one other doublet in K.R. list with the same separation , viz. , .
This is clearly analogous to the doublets in Cu and in Ag , which have the respective doublet separations but which do not belong to the or series .
and however give an arc line at , which gives a separation of with K.R. line at .
This has the appearance of a set , being at with a satellite separation of .
But if so it is quite out of step with the progression of the lines for Cu and Ag , , 5220 ( Cu ) , 5471 ( Ag ) .
But gives with the above a separation the same as that between 4792 and 4065 , and they are in step with Cu and Ag as and , the fainter satellite being unobserved .
This would seem the more probable allocation .
In any case , the curious doubling of a type would have to be explained .
There is , however , here not sufficient data to determine the limits , or the other formulae constants or the value of .
But it is possible to arrive at a probable estimate by the following considerations .
The limit will probably be in step with those of Cu and Ag , viz. , 31515 , 30644 , i. e. , will be in the neighbourhood of 30000 .
Now must give and must itself be a multiple of the oun , in fact if it is similar to Cu and Ag of .
Now with an uncertainty of a few units in the second decimal place .
The ratio , where is probably not greater than 1 in the first decimal place and it will be regarded as a correction on the From this it follows that .
The uncertainty 097 due to the uncertainty in produces so small an effect that it may be neglected here .
Now must be a multiple of and must give with the proper value of , , 30 being the maximum error of and therefore between This condition gives the following sets of possible limits in the neighbourhood of :\mdash ; with about 30542 , , The actual calculations made before the last most probable 361 890 was obtained , but 1lothing is to be gained by recalculating to it .
If now the lines be taken as , , their wave numbers are , , giving for the wave number of , or , the value .
In this is the possible observation error in .
The satellite separation of must therefore be caused by a denominator difference which is also a multiple of .
If this be tested , it is found at once that only the first and last of the above set can satisfy this condition , 30819 taking or , and 29465 taking or .
The corresponding values for Cu and Ag are both The differences between the values calculated from the lines and from the multiples of are ( between giving the observation error in ) .
with It is clear that either can easily be made to vanish well within possible errors , more especially the latter .
The limit 30819 is higher than that of Ag instead of lower as might be inferred from the fact that the limit of Ag is lower than that of Cu .
The limit 29465 is 1179 below that of Ag , which is itself 931 below that of Cu .
This seems a probable order of magnitude , especially when it is remembered that there is a gap in the Periodic Table between Ag and Au .
But there is further evidence in favour of the latter .
If the lines are collaterals of as the corresponding lines in Cu and Ag appear to be , should be .
With the limit 30819 this cannot possibly be the case .
The oun is so large that there can be no doubt .
If however the limit is 29465 , the line is .
Further , with neither limit is the mantissa of a multiple of , and as this is also not the case with Ag or Cu , it may be regarded that in this group either these lines are not of the type , or possibly like the high melting-point elements of Group II .
the first lines correspond to and not .
The actual values of the denominators as found are so dose to the same value for all three elements as to suggest the existence of a group constant .
If the limit 29465 is used the denominators are as given below , and as is seen they differ from such a constant by very small multiples of .
Ag .
Au .
Density . . .
978276 ( 21 ) 977162 ( 19 ) 971409 ( 26 ) or say a group constant 978430 .
Whether this apparent equality corresponds to a real relation or not must be left for further evidence .
In any case a limit would .
throw this relation quite out .
DR. W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES .
As a final result the evidence would seem conclusive that for Au is , that , and that the satellite separation is produced by It must be confessed that much of the foregoing discussion is of a problematica ] nature , and that , in fact , some of the suggestions offered are incompatible with one another .
This is no objection in a preliminary search for general principles , as the raising of questions is only next in importance to answering them .
Nsvertheless , some results appear to be well established and others to have considerable evidence in their favour .
Amongst the first are\mdash ; ( 1 ) The dependence of the spectrum of an element on its oun , a quantity proportional to the square of its atomic weight and which probably does not differ from by more than where denotes one-hundredth of the atomic weight ; ( 2 ) The direct dependence of the ordinary doublet and triplet separations on multiples of the oun ; ( S ) A similar dependence of the satellite separations in the Diffuse\mdash ; or the 1st associated\mdash ; series on multiples of the same quantity ; ( 4 ) The existence of collateral displacement , whereby new lines are formed by the addition or subtraction of multiples of the oun .
Until , however , the laws which govern the formation of collaterals are more fully known , it is not safe to assume that any displacement indicated by mere numerical coincidence corresponds to the physical change such collateral indicates .
Nevertheless , many cases of clear displacement of this kind , involving considerable multiples of , especially in the series , are given which serve to give more accurate values of the oun .
The conditions which govern the various multiples of the oun which enter in the various separations have not been determined .
It is probable , however , that the multiple for the doublet , or first two of a triplet , in the two sub-groups of the group of elements contain and respectively as factors .
It is probable that the mantissa of the normal first line of the Diffuse series , the last satellite , when such exists , being considered as the normal , is a multiple of and it is possible that its magnitude has some general relation of approximation to that of the corresponding series , which again may depend directly on a group constant .
It is possible that the wave numbers of the lines in the Diffuse and series may not depend directly on a mathematical function of the order of the line , and it is probable that this is the case when there are no satellites , the diiiere ces now being multiples of the themselves .
In the discussion of the material it has been attempted to keep the mind as free as possible from any preconceived theories as to the origin of the vibrations which give the lines .
The aim has been to discover relations , which it must be the object of theories to explain .
Nevertheless , the way in which multiples of a quantity depending directly on the element enter , and indeed multiples of these multiples , irresistibly suggests that each line is due to a special configuration built up of aggregates of the same kind .
Thus , in the group appear multiples of , in Mg , of \amp ; c. These smaller aggregates peculiar to a group then appear to enter like radicals into more complex aggregates , e.g. , in Zn , and again , multiples of occur in collaterals .
In cases , a certain aggregate , normally to be expected , appears to be affected with instability , a certain number of ouns are expelled or added and we get a stable collateral .
In the case of rich spectra and of spark spectra , a very large proportion of the lines appear to be collaterally connected .
It suggests systems in which a greater freedom of aggregation is permissible .
But there is another way in which the matter may be looked at .
The actual multiples may be determined by the number of electrons taking part in the vibrations , and the quantity enters into the formula as the product of this number by a fundamental quantity of the atom .
But it is difficult to see how .
this quantity should depend on the square of the mass .
It would almost look as if the gravitational pressure of two atoms always at the same distance produced some change in the configuration of the surrounding aether proportional to the pressure , and that the vibrations were conditioned by this change and by definite numbers of electrons .
In any case , the existence of the oun , and the extent in which its influence is shown in a spectrum , point to the conclusion that the positive atom plays an essential part in at least those vibrations emitted which are slow enough for us to observe .
APPENDIX I. The Value of in Scandium .
The value of as a multiple of in Scandium is of importance in connection with the evidence as to the curious relation , that the 's of the first elements of the two sub-groups in the group are multiples of and .
The lines in the visible part have been measured by FOWLER , lines both in visible and ultraviolet by EXNER and HASHEX .
I do not altogether feel full confidence in the allocation suggested below , but it gives related series , even if not the typical ones , and so will serve to determine The doublet separation is small fraction .
There are over 34 doublets with this separation\mdash ; two , and containing some of the strongest ]ines in the spectrum .
The lines suggested for the series appear to show ' Pbil .
Trans 209 , p. 66 .
'Spektren der Element : Bogenspektren .
' DR. W. M. HICKS : A CRITICAL STUDy OF SPECTRAL SERIES .
With the value of above the value of calculates with this limit to with .
The denominator differences of and for and 2 are respectively 4188 and 3964 , i. e. , close to and , with errors in of 31 and , divided tween the two sets of lines .
It is possible they may be the same within limits of error , when the value of in the second would be The formula for , calculated from the first three lines , is This formula gives the following values of for the lines for , 8 , 9 , 10 , , 85 , 26 , .
If the denominator be treated in the same way as , deducting the group constant 043761 , it may be put into the form , which reproduces the 215 constant .
Also VS ( 1 ) within error limits .
If had been taken as this would not have been the case .
But UP ( 1 ) extrapolates to with a denominator , whilst with a denominator , and they cannot be the same even approximately if the typical formula holds .
The extrapolated value of UP ( 1 ) requires to give or , again giving a multiple of and at the same time more in line with other elements as being a multiple of itself , and a multiple more in step with them .
It points to the likelihood that the series chosen for is a parallel series to the true , i.e. , the VS ( m ) is correct .
If so , using the values of VS with extrapolated limit 40717 , we should expect doublets with the first lines at 18440 and 30217 .
There is a doublet at , no line observed at 30217 , but the doublet companion expected at 30537 is found at 18457 appears as a satellite to a stronger line in a corresponding position -head .
The second set of lines above is , therefore , probably the true series , the first being a parallel one .
With limit and , the mean of all the doublets , the value for is with .
The value of is subject to considerable uncertainty owing to uncertainties in the limit value and the atomic weight , and its agreement with the final estimate for the oun is a mere coincidence .
The spectrum of Sc is a most interesting one , but its discussion must be postponed .
The object of touching upon it here is to obtain some indication of the nature of its oun as Sc occupies the first place in its sub-group .
It would clearly appear that the multiple of the oun in contains 7 as a factor , viz. , 13 The separation of and is , corresponding to a denominator difference of .
This is in fair agreement with the case of other series in which the differences for orders below the first are about satellite .
But EDER and VALENTA observe and not 7757 , whereas the former if a satellite should be fainter than the latter .
The question is discussed in [ I. ] and also in the present communication , and the weight of evidence would appear to be against the existence of satellites in Rubidium .
I have entered the third line as instead of K.R. value of , because the former value agrees with independent observations by , S. , and E.H. , and gives a better value for Cs .
The first were observed and allocated by PASCHEN , the second first by LEHMANN , but the measures are those of PASCHEN .
SAUNDERS was the first to draw attention to the satellites .
The last line , 5118 , was observed by HARTLEY , and is clearly the line .
The corresponding would be in the neighbourhood of , or .
Mg , Ca. The ] ines were all assigned by RYDBERG .
He also wrongly assigned two lines about J to .
The measures are by K.R. , except the ultra-red in Ca due to PASCHEN Phys .
Sr. Assigned and observed by K.R. , except ultra-red due to PASCHEN .
Ba .
Assigned by SAUNDERS ( ' Astrophys .
Journ 28 , p. 223 ) .
The measures are those of K.R. , except those with S. attached . .
The lines down to with the exception of the satellites were assigned by RYDBERG .
The measures , as well as allocations of the others are by K.R. Cd .
Lines to , satellites excepted , assigned by .
The remainder allocated and all the measures by K.R. Eu .
The spectrum gives evidence of much collateral disturbance , and the unobserved lines are possibly displaced in this way .
Hg .
The lines to assigned by RYDBERG , to satellite by K.R. to by MILNER ( ' Phil. Mag ( 6 .
) , 20 , p. 636 ) .
Al .
RYDBERG gives the first two doublets without the satellites and assigns 11280 wrongly to .
Further he assigns lines to the series , but the observations at his disposal were too inexact .
The measures and allocations are by K.R. In .
RyDBERG down to with satellites of first two .
The remainder and all the measures by K.R. As in the series it should be noted how the lines are more persistent than the Tl .
RYDBERG gives all except from , and he gives down to The measures given are by K.R. except those below 2105 , which are due to CORNU .
The Lines .
BERGMANN in 1908 Wiss .
Phot also BEIBL xxxii .
measured lines in the ultra-red spectra of the Alkalies and observed in Cs a number of doublets which clearly formed a series , and a few lines in Na , , Rb which were evidently analogous .
It was , however , RUNG ( ' Astrophys .
Journ 27 , p. 158 ; ' Phys. Z. 9 , p. 1 ) who pointed out the dependence of the limit of the series on DR W. M. HICKS : A CRITICAL STUDY OF SPECTRAL SERIES . .
The lines in the far ultra-red have since been observed by PASCHEN ( ' Ann. .
Phys 33 , 1910 , p. 717 ) .
BERGMANN'S lines began with the order , but in the lists above more recent and more accurate measures by others have been inserted .
BN .
refers to BERGMANN , .
to RANDALL Phys , S. to SAUNDERS ( ' Astrophys .
Journ 20 , p. 188 ) .
In 1905 FOWLER ( ' Astrophys .
Journ 21 , p. 84 ) discovered the lines in the series for Sr , beginning with K.R. 4892 , adding two sets of his own observations and also the two connected triplets given in the table .
But he attempted to combine them all in one formula .
The disarranged triplet for is assigned in the text above from K.R. and LEHMANN Phys .
In the same volume SAUNDERS ( ' Astrophys .
Journ 21 , p. 195 ) added the last line and gave a similar series for Ca , commencing with 4586 .
The measures inserted in the table , however , are from later observations by BARNES ( Bs .
) ( ' Astrophys .
Journ 30 , p. 14 ) .
SAUNDERS also suggested a corresponding series for Ba in which the separations are larger than those of the series assigned in the text .
It is possible they may form a similar series connected with the enhanced series .
The lines for and allied sets are assigned also in the text above .
The lines for Mg , Zn , Cd , Hg , Al , and Tl are all due to PASCHEN ( ' Ann. Phys 29 , p. 651 , and 35 , p. 860 ) .
In Hg , however , he assigns 17195 to a combination line , whereas in the text above it is assigned to [ Notes , September 2 , 1913 .
] Note 1.\mdash ; Since the present communication was read Messrs. and REYNOLDS ( ' Roy Soc. Proc 89 , p. 137 ) have published more accurate and extended measurements of the series , and other lines of Mg .
The limit of series appears to be somewhat higher than that adopted in the Table II .
( p. 344 ) , and the table will require the limit to be slightly raised to bring in muItiples of for the ordor differences .
With FOWLER and limit the mantissae show rising values for the first few orders and then decreasing , an effect which has been explained in the text in cases by a collateral displacement in the limit after a particular order .
If the limit be taken to be 39757 78 higher than text ) , there results an order difference of between and 3 , and equal mantissae afterw down to , when there is a sudden change to a rising .
If limit is now raised by unity for these , the mantissae can again be all equal .
Now , a displacement of on the limit makes a change of , so that the change can be explained by a collateral change of for .
The scheme is then as follows :\mdash ; 3829715(79 ) 4 .
The rangement is seen to be exceedingly simple .
The value for has been omitted as the evidence sesms strong that Mg conforms to the Zn type and has no .
If , however , it is retained , it is now below that for .
The values of are so small that Mg can give no positive evidence for any arrangement .
This is evidenced by the changed arrangements called for by more accurate values .
The measures of and REYNOLDS for the lines are here adduced in order to complete Appendix II .
:\mdash ; Note 2 , p. 357.\mdash ; The effect of positive collatel.als on for , 6 , 7 is to diminish the separations of the triplets , so that from onwards they would show diminishing values .
It is interesting to note that the observations bear this out .
, p. 380.\mdash ; The series of the alkalies .
In K. denominator has its mantissa 1 007542 , and this is 14725 above that of .
( 3 ) , and .
This suggests that if .
( 3 ) has a denominator , that of is , and the 1182 9 ( is with a satellite difference 28 and 1346 3 is the collateraI .
In Rb 3 001138 has its mantissa 13289 above that of } ) a , so that should have a similar arrangement to that in K. and 2156 would correspond to .
Note 4 , p. .
The line with wave number 16203 40 is possibly the and combined .
If so , is 16202 66 and is behind 16203 40 , and the satellite might have displaced the observation towards itself from 16203 66 .
The actual separation would easily be 1 00 corresponding to a rence , a usual satellite difference .
16024 72 would be on the other side and would correspond to a collateral forming on the violet side .

