144 RALPH S. LILLIE. 



at its extremity; this process is accompanied by a flow of solution 

 outward through the tube. Filaments of iron ferricyanide show 

 most clearly the nature of the whole process. In a 2 per cent, 

 solution of KsFeCye in dilute egg-white containing a little NaCl 

 (0.25 per cent.) large hollow filaments grow out rapidly from 

 the metal within a few seconds. Each such filament advances 

 into the solution at an apparently rapid rate (of a millimeter or 

 more per minute) by a uniform motion; its contour is smooth 

 and regular and its shape cylindrical or slightly tapering; it 

 usually follows a straight course, but is easily diverted by en- 

 countering obstacles, such as solid particles or other filaments, 

 or by shaking the watch-glass so as to make irregular the deposi- 

 tion of precipitate at the extremity. In a growing filament 

 this extremity is always open, and from the orifice a stream of 

 liquid containing fine suspended particles flows continually out- 

 wards for a short distance in advance, suggesting somewhat the 

 smoke emerging from a chimney; this stream of finely divided 

 material leaves behind and on either side a trail of particles 

 through which the growing tubule advances; while at the edges 

 of the open orifice a dense uniform layer of precipitate is con- 

 tinually being deposited in continuation of that already laid down. 

 Growth proceeds rapidly in this manner until some condition 

 arrests the outflow of fluid; this may happen as a direct result 

 of increasing length, the flow through the long capillary tube 

 being slowed more and more by frictional resistance as the length 

 increases, until finally the end is sealed by the precipitate. 

 Evidently a wide tube will grow longer, under these conditions, 

 than a narrow one, and in fact the great majority of filaments 

 cease growth at a length of one or two millimeters or less; it is 

 only the exceptional wider tubes that reach the length of a 

 centimeter or more. 



From these facts it is clear that elongation is dependent upon 

 the maintenance of a flow of solution from the metal along the 

 tube. The precipitation at the open end shows that this solution 

 contains a dissolved salt of the metal. Apparently the process 

 takes place somewhat as follows. At the anodic portions of the 

 metallic surface ferrous ions enter solution; these form with the 

 ferricyanide ions of the solution already present a precipitate of 



