EXPERIMENTS Wmi lONTZED AIR. 69 



Series 1, x/ r= .13, Series 3, x/ V = .10, 



Series 2, x/ V = .12, Series 4, «/ K = .16. 



Quantitatively, therefore, the assumption (7 ■= I makes the computed data of a;/ V 

 from 1.6 to 2.6 times too large. To improve this result one would thus have to 

 wi'ite U'<k Now since there is evidence in my second paper that h decreases as 

 the degree of dilution increases, and since I- when experimentally (f/</fc) found 

 comes out smaller than the value assumed for U, this inference is not tenable, k 

 bein" simply the absorpti(ni velocity for the average concentration in the tube. 

 Again U> k, though presumable from the earliei- work, is strongly negatived here. 



Moi-eover the vai'iability of the slopes x/ V, observed for the different series, 1 

 to 4, clearly with an inherent cause since the slope for a given series is determinate, 

 is additional evidence in this direction. 



In a color tube left entirely without disturbance, the color must depend on 

 Vn, the total nunibei- of nuclei entrapped. Thus the color may be reckoned 

 as Vn = V)i. s"-^''. The same color should appear if Fw = F'/i' = const., if the 

 saturation decreases as the volume supplied increases. The ratio of the electric cur- 

 rent in the condenser or preferably the conduction, x, of the ionized medium to the 

 color constant is thus x/ Vn = 434 Ue (e^"'-l)/(7/?</'3 -r J, which is independent 

 of the initial saturation 7?o but varies with V. Thus, although two volumes, 

 Fand F, produce the same color ( Vn = Fy<^'),they will nevertheless correspond 

 to diffei'ent conductions, x. Hence the variable slopes of figure 3 for the different 

 seiies. There, is however, an outstanding difiiculty requiring further observation 

 for singularly enough the volumes alone appear to determine the coloi's in the chart 

 rather than the (piantities T"/i. 



Solid conductions and other leakages were not found in the tests made, in 

 quantity to be menacing. 



11. The number of part ides.— The probable explanation of the discrepancy 

 (leaving the ratio U/km abeyance) is that the value n^ = 4 X 10'» inserted is too 

 large ; for the emanation within the condenser is not saturated to begin with. Tt 

 enters and leaves unavoidably from a certain length of influx tubulure, c in figure 2, 

 insuring absorption at the outset. Unfortunately my notes do not supply the 

 details T)f adjustment, as I did not suspect their importance at the time. But if 

 they did the loss at the elbow at both ends of the condenser would remain as a 

 vague but definite cause of ion dissipation. 



On the other hand the value of //„ which the observations demand is easily 

 computed from the observed values of x/ V = 2QK, nearly, and from these very 

 satisfactory inferences may be drawn. 



Solving therefore, 



^ _ (.x/V)kC(r,-r, i 



"0 434 Ue 



where as before Z7= /I" may be assumed for orientation. 



The extreme values of x/ F experimentally obtained were, x/ V= .10 in series 

 3, and x/ F= .16 in series 4. Hence the values n, = 1.5 X 10^ and «„ = 2.4 X 10* 

 are computed for these extremes. Recalling that 4 X 10' is the number of particles 

 per cubic ceutim. inferied for complete saturation, and that n, is the number at the 



