A CONTINUOUS RECORD OF ATMOSPHERIC NUCLEATION. 



109 



method often gave serviceable results, some of which are inserted in the follow- 

 ing tables ; but at times the n -values are out of proportion. The reason of this 

 is threefold: In the first place n is found from n l/i with the usual difficulty. 

 Again, in a simple arrangement like the above, air currents cannot be quite 

 excluded. They may arise incidentally in the apparatus or the motion of the 

 plate even if parallel to itself may stir the air unless some form of guard ring 

 attachment is added. Particles are thus swept down upon the disc before and 

 after the exposure, as was actually observed. The difficulty may be removed 

 by adding a capsule above the plate or simply by decreasing the distance be- 

 tween the shield and the plate to a millimeter or less. Finally, if the oil film is 

 semi-fluid and not qviite fixed, if there is slight creeping, as was usually the case, 

 the particles are redistributed after falling along stream lines where they cohere 

 in strings and bunches, but without coalescing. This was also observed, and in 

 fact the capillary forces involved are apt to be strong enough to counteract 

 viscosity. 



I have not thus far spent much time in correcting these defects, chiefly 

 because the new results for the diameters of fog particles are more immediately 

 interesting. The data are given in table 3. 



TABLE 3. OBSERVED DIAMETERS AND NUMBERS (per cub. cm.) OF CLOUD PAR- 

 TICLES, m =4.7 X io~ 6 g per cub. cm. ; If c = 144 X io" 6 sq. cm., and t = 30 sec. ; ^ 

 = 1.75 n'. Generally $ / ri = 2.11 n'/tc w 2/3 io 6 . Micrometer factor, .00004 cm. 



Particles as small as .0003 cm. present throughout. 



Particles as small as .0003 cm. always present. 



Particles graded as usual. 



