14 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 8/ 



what too small for more intense lines and too large for very weak 

 lines it will indicate the accnracy of the measnrements. For the high- 

 intensity values, Table 2, the average galvanometer deflection was 

 40 mm. The probable error of the mean of 10 readings was 0.7 mm 

 or ± 1.4 per cent. Values in Table i, column 3, are subject to a 

 probable error of ± 0.5 per cent and values in column 2 to ± 2 

 per cent. The probable error of the value 1.33 (the thermocouple 

 sensitivity) is ± 1.4 per cent. Hence the probable error of the mean 

 value, 100 mew cm"-, is about 3 per cent or ± 3 mew cm"-. For the 

 values given that are of the same order of magnitude as the probable 

 error of the mean, ± 3 mew cm"-, the predominant error is in the 

 galvanometer reading. Hence the probable error for these entries 

 in Table 2, column 4, is about ±1.5 mew cm"-. For the low-intensity 

 values in Table 2, the average galvanometer deflection was 1.9 mm 

 and the probable error of the mean of 10 readings was 1/20 mm, or 

 zt 2.6 per cent. Here the deflections were read with a telescope, and 

 the arcs output was much steadier than for the high-intensity obser- 

 vations. The other errors are the same so the probable error of the 

 mean value, 4.6 mew cm"-, in Table 2, column 6, is about ± 4 per cent 

 or .18 mew cm'-. As before the probable errors of values of this 

 order of magnitude in Table 2, column 6, are determined by the error 

 of reading the galvanometer; hence the probable error of these entries 

 is about 0.12 mew cm""'. The average deviation of the values in 

 Table 2, column 5 (new arc) from those in column 4 (arc 400 hours 

 old) is 5.3 per cent, which is about twice the probable error. Thus 

 it is reasonable to assume that the differences are real — especially so 

 in the short wave length region where the new quartz is more 

 transparent. 



Figure i shows 50 maxima. Thirty-two of the strongest and most 

 clearly resolved of these are included in Table 2. The intensity of 

 the weaker lines and of the unresolved components omitted from 

 Table 2 will be reported in the near future. A new crystal quartz 

 spectrograph of higher dispersion now in the process of construction 

 will be used for this work. 



These absolute intensity measurements are interesting from a 

 spectroscopic standpoint. The pressure of the mercury vapor in the 

 high-intensity arc is about one atmosphere, and in the low-intensity 

 arc it is several millimeters of mercury. This amount of vapor will 

 cause considerable self reversal as is most evident for the resonance 

 line A 2,536.5 A. and so make difficult the comparison of experimental 

 and theoretical intensities of related spectral lines. However, because 

 these measurements cover a wide spectral range it is interesting to 



