The Magnetic Separation of Lines 4I 
4. The interval .37 (=a/3) times (4, 2) and (5, 3, I) gives 
respectively the quadruplet and sextuplet principal series and sec- 
ond subordinate series found by Professor Runge in Na, Cu, Ag, 
Al, Tl, Mg, Ca, Sr, Ba, Ra. The quadruplet has one doubtful 
representative in yttrium, and the sextet one line in zircon. How- 
ever, the same interval, combined with factors in other propor- 
tions, gives nine new types. One of the Hg first subordinate series 
types has a solitary representative in these substances, but the in- 
tervals yield at least eighteen new types. 
5. The substances yttrium and zircon yield a great number of 
new types. 
6. The most prominent characteristic of the numerous new 
types is the number which are unrepeated in the spectral range of 
these experiments. It would be interesting to extend the meas- 
urements far into the ultra violet with a much stronger field, to 
see if there are not more repetitions and even series. 
7. The interval (.42(=6a/16) ) in zircon'is the only one which 
would promise series types. There are eight quadruplets of one 
type and six of another, but no series found. 
8. There are six lines in yttrium like seven iines in zircon, and 
these are represented by three types. These are scarcely enough 
terms to suggest similarity of the substances. Chemically, how- 
ever, there is a similarity. The substances are parallel terms in 
two adjacent (third and fourth) Mendelejeff’s groups. 
g. The one quadruplet of osmium has an interval of the first 
subordinate series, but it is not of the latter type. 
Io. An investigation of triplets for series is always tedious. Di- 
viding them into groups, as in barium ahove, is advantageous fre- 
quently. They then look like other types with suppressed com- 
ponents. But in yttrium and zirconium one would be at a loss to 
know which type had the component suppressed. This would sug- 
gest that a given separation in such triplets may represent more 
than one type. The investigation of yttrium triplets for series has 
been reasonably complete and negative. In zircon, time has per- 
mitted the study of only a few triplet magnitudes. The results 
have been likewise negative. The triplet values are extended over 
AI 
