230 DE. W. GEOFFREY DUFFIELD ON THE 



values of d/X, dfx" x d/\ 3 have been calculated (excluding measurements at 10 

 atmospheres) ; they will be found in the final columns of Table III. 



In the discussion upon the copper arc, a test of the rate of displacement with wave- 

 length was made by finding the value of n which enabled the quantities djx* to be 

 most distinctly separated into groups. This did not appear so likely to yield 

 promising results in the case of nickel, because in Diagram 1 two groups are 

 remarkably distinct without reference to the wave-length, a point which is further 

 illustrated in Diagram 6, fig. 1 , in which each number represents the presence in 

 the spectrum of a line with a mean displacement corresponding to its position upon 

 the horizontal scale. The distribution resolves itself into a well-defined group with 

 a mean displacement per atmosphere of about 1'75 thousandths of an Angstrom 

 Unit and another very diffuse and ill-defined group with a flat-topped maximum 

 extending from about 9 to 12 thousandths of an Angstrom Unit. It might be 

 argued that the probable errors of the measurements of the different types of line 

 are in accord with this grouping, because those lines whose displacements are small 

 have a smaller probable error than those with larger values, the latter being in general 

 (but not invariably) associated with a greater width of the lines. But though this is 

 a consideration which must be given due weight, to regard it as satisfactorily 

 accounting for the distribution in fig. 1 would be to disregard the significance of the 

 tendency for the displacement to increase with wave-length which has been 

 demonstrated in Diagrams 4 and 5. 



It is further important to note that lines as 208 (\ = 4762'87), 213 (\ = 4829'! 8), 

 217 (X = 4856-57), 228 (\ = 4936'02) and 238 (X = 501775) do not clearly fall within 

 either group, also that the line 24? (X = 5477 '13) is assigned to the group with large 

 displacement in spite of the fact that it is reversed unsymmetrically, and would not 

 be expected, in accordance with previous experience of similarly reversed lines in the 

 iron spectrum, to belong to a group with the highest rate of displacement. 



Plotting d/\ against wave-length, fig. 2 is obtained, and again the lines 208, 213, 

 217, 228 and 247 occupy anomalous positions. There is still less reason for regarding 

 this distribution as satisfactory. Fig. 3 shows the distributions of the values of d/\ 2 

 when the lines 247, 213, and 217 form a group by themselves, though 208 has 

 attached itself to the outskirts of the first group. 



In fig. 4, in which the values of d/X 3 are given, these lines have been absorbed by 

 Group 1 , and the resolution into two groups is clearer. We note also that the second 

 group is more compact ; this is obviously the most satisfactory diagram, and we may 

 therefore conclude that as far as values above 10 atmospheres are concerned, the rate 

 of displacement is not far removed from that of the cube of the wave-length. It may 

 appear remarkable that any order can be arrived at considering the apparently chaotic 

 distribution of dots in Diagram 4 in the region of wave-lengths X 4200 to X 4500, and 

 it is necessary to emphasize the fact that the method consists in the elimination of the 

 dots with displacements varying from 5 to 8 thousandths of an Angstrom Unit from 



