PROF. A. FOWLER ON SERIES LINES IN SPARK SPECTRA. 
229 
very probable, indicating a condition somewhat more advanced than the ordinary arc, 
in the direction of the spark.] 
The Line 3104.—This line was easily resolved in the 4th order of the grating 
(Plate 3, fig. 7), and has two components exactly resembling those of 4481, but 
o 
separated by less than one-tenth of an Angstrom unit. The iron arc comparison was 
photographed through a glycerine solution of para-nitroso-dimethyl-aniline contained 
in a quartz cell, as recommended by Prof. P. W. Wood. # This solution absorbs the 
superposed 3rd order spectrum, but transmits the 4th order lines with reference to 
which the magnesium lines were measured. The mean results from three excellent 
plates with five comparison spectra are as follows :—- 
International scale. Rowland scale. 6 k. 
3104-805 3104-929 
0-092 
3104-713 3104-837 
The Line 2661.—On the assumption of a constant wave-number interval between 
o 
the components of the resolved lines, the separation at 2661 would be about 0"07 A. IT. 
which is very near the limit of resolution of the grating. Hence with the exposures 
of 3 to 5 hours required in the 3rd and 4th orders, a very slight unsteadiness of 
conditions, such as might arise from small variations of temperature, would tend to 
mask any resolution which might otherwise be effected. Many 3rd order plates 
were taken, all of which show the line broadened, but only one suggests resolution, 
and on none of the 4th order plates was this result improved upon. Measurements 
of the best 3rd order plate gave the following wave-lengths :— 
International scale. 
Rowland scale. 
8k. 
2660-821 
2660-909 
0-066 
2660755 
2660-843 
These values are in good accordance with the mean of several determinations from 
1st order plates, namely, 2660785 (I.A.), 2660"873 (R). 
The Remaining Lines .—-It has not been found possible to photograph beyond the first 
three lines in the higher orders of the grating, but the wave-lengths of the next two, 
2449 and 2329, have been checked by a photograph in the first order, giving a 
o 
dispersion of 5"5 A. U. per mm. Wave-lengths of iron lines given by Burns were 
used for the line 2449, but were not available for 2329. The lines 2253, 2202, 2166 
do not appear on any of the grating photographs, but are well shown on the original 
plates taken with a quartz spectrograph, giving an average dispersion in this region 
o 
of about 6 A. IT. per mm. Further measurements have been made, but in the absence 
of better standards it has not been found necessary to modify the values given in the 
* ‘Phil. Mag.,’ Series 6, vol. 6, p. 257 (1903). 
