]\gt ; The Value of the for Spectral By W. E. CURTIS , B.Sc. , Demonstrator of Astrophysics at the College of Science , South Kensington .
( Communicated by Prof. A. Fowler , F.R.S. Received Apri19 , 1919 .
) troductory .
In a previous paper , results of a series of measurements of the lvavelengths of the first six lines of the Balmer series of hydrogen iveu , together with a determination of the Rydberg constant for spectral series .
It has recently been pointed out to me Prof. Fowler that the yalue of the series constant there obtained is not quite correct , in consequence of errors in the corrections applied to the erved reduce them to vacuo , the data employed for this purpose been taken from a Table appropriate to wave-lengths in air at C. , whereas the tertiary of iron upon which the wave-length determinations were based referred to C. Before the introduction of the International system , had always been given for air at C. , and there was no explicit mention the change of standard temperature in the by Burns , from which the iron arc wave-lengths used as standards were taken .
The oversight possibly have been detected earlier but for the or 's absence on military service from September , 1914 , to January , 1919 .
In any case , however , revision of the previous work would have been necessary in view of the recent accurate determinations at the Bureau of Standards of the refiactive index of air , S and also of other work , both theoretical and experimental , on the Balmer series , which has been carried out since 1914 .
The correction in question is of the order of A. , and results , as will be seen , in a decrease of nearly one unit in the value of the constant .
Corrected W/ 7tb .
The corrected values of the wave-lengths and wave-numbers are tabulated below .
The last column gives the values of derived from each line by applying the Balmer formula .
the observed wave-lenothso .
in air are only recorded to the third place of decimals , it is advisable to take 'Roy .
Soc. Proc , vol. 90 , p. 606 ( 1914 ) .
Kayser , 'Spectroscopie , ' vol. 2 , p. 514 .
'Lick Obs. Bull , ' vol. 8 , No. 247 ( S Bureau of Standards , , ' Scientific Papers , ' No. ( 1918 ) .
Mr. W. E. Curtis .
The Value of the out the corrections to the fourth place , in order to prevent the possibility accumulated error affecting the third figure .
Table I. The systematic variation of , of magnitnde exceeding that which could be attributed to experimental error , indicates , as was previously found to be the case , that the Balmer formula is only an approximation to the truth .
Discussion of ) estigated .
The formulae examined may all be written in the form Since Balmer 's law is so neal .
lie exact , is of the nature of a small COl'rection to never exceeding per cent. of the latter quantity in any of the formulae under discussion .
The exact form of this function , however , exercises an important influence the value obtained for N. It is natural to try first a formula of the ydberg type , .
:\mdash ; .
( I ) The most convenient method of solution is to assume an approximate value , , such that the true .
Then , neglecting the second and powers of and , we obtain A least-squares solution gives This , as will , seen from Table II , represents the results satisfactorily , but , before } ) ting this formula , it is necessary to examine whether an equally good representation of the series can be obtained by employing only constants .
Rydberg Constant for Spectral Series .
Retaining and , the formula becomes The least-square values The observed minus calculated values suggest a slight inferiority to formula ( I ) , but not more than would be expected in view of the elimination of one constant .
Since the representation is within the limits of probable experimental error , one must conclude that the introduction of the third constant is justifiable .
Retaining instead of , the formula becomes A least-square solution ives This again appears to be quite satisfactory as regards the numerical agreement with the observed wave-lengths .
The superiority over ( II ) , however , is too slight to warrant the definite adoption of ( III ) in preference to ( II ) , but the conclusion is confirmed that the present data do not necessitate a formula including more than two constants .
Turning now to theoretical investigations in connection with the radiation of the hydrogen atom , Dr. Bohr has modified his well-known derivation of the Balmer formula by introducing a small correction due to the dependence of the mass of the electron upon its velocity .
* He thus obtains the formula Applying this to the observed wave-numbers , the method of least-squnres gives It will be seen from Table II that this formula is inferior to either ( II ) or ( III ) , both of which employ the same number of constants , namely , two .
The residuals exceed the probable error of the wave-length determination in two cases , viz. , and , the former being the best determined line of Avhe six .
Moreover , the deviation tends to be systematic , a feature which is brought out very distinctly by calculating the constants from two lines instead of by the method of least-squares .
'Phil .
Mag vol. 29 , p. 332 ( 1915 ) . .
W. E. Curtis .
The Value of the The theoretical expression for is , where of electron , velocity of light , Planck 's element of action .
This , when evaluated , yives k , about one-third of the value found above .
The sign and order of magnitude are certainly right , and the discrepancy is possibly no more than could be attributed to error arising from the compound nature of the lines , a point which is considered in the succeeding section .
Dr. Allen , by account of the netic moment of the atomic nucleus , has obtained an expression of the form ( V ) The least-square values derived from this formula are It is singular that the value of should agree precisely with that obtained by Allen on the assumption of five magnetons , but the exactness of the agreement can only be accidental .
Moreover , the residual for is distinctly too large to permit us to yard this type of formula as satisfactory .
In fact , neither this nor formula ( IV ) seems to be of a quite suitable shape to represent the observed values .
It is considered that these latter are hardly likely to be in error ] ) more than the amounts shown in the third column of Table II , in view of the fact that the type of formula gives such good results .
The formula crinally proposed by Dr. Allen was more eneral than the above , and was of the form .
( VI ) With this type of formula , is necessarily negative , but there would seem to be no objection to this on theoretical grounds .
A least-square solution gives The representation is much better than in the case of , the residuals being comparatively small , but they exhibit a distinct tendency towards systematic variation , so that the formula cannot be considered so as any of the ydberg forms .
It may be however , that the above value of is Rydberg Constant for Spectral Series .
exactly that obtained theoretically on the assumption of three magnetons in the core of the atom .
One other type of formula has been examined , and the results are perhaps worth , although it does not give a quite satisfactory fit , nor has it any theoretical basis so far as is known .
The only point of interest in connection with it is that only one constant is employed .
If we write for in formula ( II ) we get .
( VII ) The average value of is , and the only sensible residual is that corresponding to , which , however , is too large to be admissible .
Table II.\mdash ; Comparison of Formulae .
( ?
obs .
calc .
) Eff'ect of Composite Structure of Lines .
Before attempting to decide upon the best value to adopt for , in view of the foregoing discussion of formulae , it is necessary to take account of the structure of the lines .
It has long been known that at least two components existed in the case of and , and presumably this complexity exists also in the other members .
Obviously , then , for the complete representation of the series as many formulae as there are components will be required , and the value of derived from measurements of the unresolved lines ( as they were in the present case ) will not necessarily be the true one .
If the intensity distribution were very unsymmetrical an appreciable error might be introduced in this way .
An attempt has been made to estimate the amount of the uncertainty due to this cause , but it is difficult to do so with any degree of exactitude .
In the first place , the precise structure would still appear to be in some doubt .
Bohr mentions ' the fact that the hydrogen Mr. W. E. Curtis .
The Value of the lines , when observed by instruments of high dispersive power , aresplit up in a number of components situated close to each other but the reference is not given , and has not yet been traced .
According to the observations of Merton and Nicholson , and are doublets , the separations ( and respectively ) corresponding to a principal type of series .
They agree with previous investigators that the ratio of intensities is about 10 : 7 , the less refrangible component being the stronger .
If it be assumed that the wave-length determinations really refer to the " " optical centre of yravity , \ldquo ; i.e. , a point 7/ 17 of the separation from the less refrangible component , and if it be further supposed that the series is of the principal type , it is possible to caloulate the wave-lengths of the individual components .
The former assumption would hardly be permissible if their actual wave-lengbhs were required , but the object here is merely to estimate to what extent the duplicity of the lines may affect value of N. If we adopt formulae of type II we shall have The wave-numbe interval , , approx. , and the interval or , where is a constant which can best be determined from the separation of Assuming for this the value , the separations of , etc. , come out , and respectively .
* Phil. Trans , vol. , p. 237(1917 ) .
Rydberg Constant for Series .
By applying the first assumption the thso iven in Table III were obtained .
The calculation of formulae of type ( II ) .
these As will be seen , the fit is satisfactory in both cases , and there is no appreciable in the value of N. If instead of formula ( II ) , either ( I ) or ( III ) be used , a similar result follows , that is to say , the values of respectively obtained differ but little from those already given .
So that if the doublets are really of principal type , or , more generally , if they show a rapid decrease of wave-number interval as increases , the consequent uncertainty in the value of is not serious .
If instead of dealing with the " " optical centre of gravity\ldquo ; it be supposed that the settings were made midway between the edges of the lines ( and is certainly what was attempted ) , the values of become and , and there is no appreciable in the value of N. On the other hand , the values obtained for are considerably affected by the question of the structure of the lines , and cannot be determined with any degree of accuracy from measurements on unresolved doublets .
Two interesting possibilities present themselves , however , and may be noticed here .
One is that the Balmer law may hold rigorously for one of the components of each line .
The fact that the of formula ( II ) comes out positive excludes this possibility as far as the more refrangible component is concerned , and a simple calculation shows that for the less ible component , could only be zero if the settings had been made on a point situated about one-fifth of the separation from the latter component .
This would appear to require a much larger difference of intensity than has been found experimentally , so that in all probability neither component is exactly represented by the Balmer formula .
The second possibility in question is that the components of each line may be symmetrically situated on either side of the } ) ositions indicated Balmer 's law .
If this were so , and would be equal and of opposite signs , and it is found that the measurements would to be taken as referring to a point five-sevenths of the separation from the less refrangible component .
This would only be the case if the more refrangible were the stronger , which is again in disagreement with the experimental evidence .
The Value of the Series From the preceding remarks , it will be seen that the chief difficulty in determining the most probable value of arises not so much from the nonhomogeneity of the lines as from the uncertainty as to the proper type of Mr. W. E. Curtis .
The Value of the formula to be adopted .
The results given by the various formulae are sun marised below:\mdash ; ( 1 .
) ( II .
) ' ( III .
) ( V. ) ' It will be observed that the two theoretical formulae ( IV ) and ( VI ) give th lowest values , but neither oan be arded as definitely supported by th experimental results .
From the latter point of view , ( I ) , ( II ) , and ( 111 ) are satisfactory and about equally goo For the present , until accurate determi nations of the wave-lengths of the components ( particularly of ) become available , the adoption of as a provisional value is suggested .
Thi is in agreement with that derived from ( I ) , and is roughly intermediat between ( II ) and The amount of the uncertainty , about at most , is not serious consequence , having regaxd to the degree of accuracy of our presen data in the case of most of the series lines of other elements .
The same true with reference to the correction now applied to the valu ' give in the previous paper .
Thus , for example , the conclusion Nicholso that the Rydberg constant derived from the Balmer series is alsc applicable to the helium series , does not require modification on this account The wave e of all the lines down to have been recalculatec ' in the hope that they may possibly prove of servlce in certain cases , such as for example , the reduction of eclipse spectra , particularly those obtained with slitless raphs .
Formula ( I ) was used for the purpose , and the differences from the previous values are extremely small , never more than , in fact .
The new values are given in Table below:\mdash ; 'Roy .
Soc. Proc , vol. 91 , p. ( 1916 ) .
Rydberg for Series .
Table \mdash ; Wave-lengths ( in Air ) of Balmer Series Lines calculated by Formula ( I ) .
above formula gives the convergence wave-number in vacuo as orresponding to a wave-length in air of I.A. , which is exactly the \amp ; me as that obtained in the previous paper .
Summary .
( 1 ) The results given in a previous paper have been modified on account a slight alteration now found to be necessary in the values of the correcions to vacuum .
( 2 ) A number of formulae , empirical and theoretical , have been compared , it is found that a two-constant Bydberg formula is capable of repre .
enting the series satisfactorily , and shows an appreciable superiority in this espect over the theoretical formulae of both Bohr and Allen .
( 3 ) The probable effect upon the determination of arising from the jomposite structure of the lines is investigated , and found to be inappreciable , xcept in so far as this non-homogeneity renders it somewhat uncertain which of formula should be employed to represent the observed wave- Bngths .
( 4 ) The provisional adoption of the value for is suggested .
[ his is probably sufficiently accurate for present purposes , but could be mproved upon if accurate wave-lengths of the individual components of were available .
( 5 ) The calculated wave-lengths of the members of the series down to are given .
These do not differ by more than A from the corre- ponding values given previously .
\mdash ; VOL. XCVI.\mdash ; A.

