RELATION OF SEPARATIONS TO THE NORMAL INTERVAL. 



I. Summaries for Various Types. 



The study of how generally the separations observed show a simple relation to the fundamental 

 interval, the theory of which was summarized on p. 4, has been gone into in some detail. The relation 



H 



a = - 



gives a value for a of 0.753 ^or H= 16,000, and of 0.812 for H= 17,500, if e/m be taken equal to 1.75 X 10^. 

 The "normal triplets" for iron and titanium, with the standard field-strengths used in this work, should 

 accordingly show values of AX/X- for the distance between the side components of about 1.500 and 1.600 

 respectively. 



In the following summaries an attempt has been made to show to what extent the separations for 

 various classes of lines may be considered as multiples of the interval a. In Table 4 the clear triplets 

 for iron and titanium are thus classified, those triplets given in Tables 1 and 2 as doubtful not being 

 included. The allowable deviation for any line from the exact multiple was estimated as closely as pos- 

 sible according to the weight of the measurement, knowing the probable error for each weight. Lines 

 not falling into any class are placed in the "Odd" column. In the case of titanium a large proportion 

 of such lines appeared to be definite odd multiples of a/4, while the regular classes consider only multiples 

 of a/2. As in all of the following work relating to the interval a, greater field strength is desirable, as the 

 accuracy of the classification increases with the numerical value of a; but Table 4 shows in a general way 

 how the magnitudes of the separations may be grouped. 



Table 4. Separation op Triplets as Related to the Normal Interval a. 



The relation of the separation to the normal interval was also studied for those fines which appear 

 on my plates as quadruplets with components in many cases diffuse, indicating a compound structure. 

 The two ^-components are usually fairly sharp, but the w-components are often formed of two or more 

 pairs blended. Close agreement with exact multiples of the normal interval can not be expected for 

 fines of this class, but in the majority of cases the distance between the components of the n and p pairs 

 could be expressed as multiples of a or a/2 closely enough to show a real relation; 66 fines of iron and 



47 



