46 
DR. W. M. HICKS: A CRITICAL STUDY OF SPECTRAL SERIES. 
evidence for the dependence on atomic volume.* In doing so, howevei, the best 
result is obtained by 1 i.e., by taking the values of the denominatoi of VS(l) 
witliout any modification, although (2) is nearly as good. 
The density found for lla, using as constant '002740, is l'02x, where .x is a whole 
number. If the’number for Ba is 6, i.e., the same as for Sr, we might expect that 6 
micrlit he a limit and x = 6 for Ra. If for Ba the ratio is 5, ohen x might possibly 
he 5. If, however, the series runs down in regular order, then allowing foi the gap 
between Ba and Ra, x would be 3, giving a density 3'06, i.e., less than that of Ba. 
If X = 5 or 6, the densities calculated are 5'10 and 6'12 ; either are in fair order with 
' the other elements of the group. 
If the elements in the third group are treated in the same way, using the densities 
A1 = 2'583, Ga = 5'95, In = 7'26,t T1 = 11'85, A1 does not fall into step with the 
others. The results are 
A1 ='002744 x6'5u, 
Ga = '002613 X 7u, 
In ='002760 x 5u, 
? = ? 
T1 ='002816x47:. 
The agreement with the supposed law is very unsatisfactory, and in the case of A1 
is clear against it. A1 is the case where, in order to produce the '21520 constant, the 
fji liad to be referred to the F-sequence (p. 42), whicli produced an addition to ij. of 
'043761. If this he done in this case also, the value of the denominator comes out to 
he '002751 x8u. This is so close to the law that it suggests the possibility of the 
elements of the tliird group behaving like the alkalies—themselves elements with 
doulfiet series. In that case it was found necessary to refer the quantities in question 
to '086+jf, i.e., add '014 to the /x + a, where '986 was a constant for the whole group. 
If so the result fixr A1 would indicate that the constant for the third group should he 
'956239, I.e., (f— '043761), or that the quantity '043761 should be added to the /x + a 
as found. If this is done the following scheme results ;— 
Al. Ga. In. Tl. 
'002751 x 87:, '002752 x 8t:, '002758 x 6i:, '002761 x 5y, 
which recalls the close agreement found with the Zn group. 
Probably the F-sequence depends on the group constant ('956) plus a multiple of 
the atomic weight term. For Al this is very small, so that the group constant would 
be somewhat smaller than '9562 determined from the F. The difference would not 
appreciably afihct the coefficients alcove. This would explain, however, the apparent 
* These results for the two groups were given in a paper read before Section A of the British Asso¬ 
ciation at the Sheffield meeting (1910). ‘Rep.,’ p. -574. 
t Mean of values given l)y Landolt and Buknsteix. 
