EVAPORATION OF ATOMS 3 55 



negligible. The rate was changed by aging from an irregular to a uniform 

 variation with 6 (Fig. 7). These experiments may now be explained more 

 fully. Since [la is negligible, there is no mechanism for stopping the growth 

 of newly formed nuclei after ^'s lower than those corresponding to the level 

 of point C are reached ( ^0.18 per cent for Fig. 19). The rate of evapora- 

 tion will increase rapidly as the perimeter (boundary) of the nucleus in- 

 creases, and fall ofif when two advancing boundaries meet. Each peak in 

 Fig. 7 may be formed by this process. A new peak is formed as other nuclei 

 increase their perimeters. Diffusion at the boundaries from the concen- 

 trated to the dilute phases speeds the process. 



These conclusions were verified by experiments in which, after one or 

 more peaks in the evaporation rate had been passed, the field was reversed 

 so that only atoms could evaporate (very slowly in the region B—C). After 

 waiting several minutes the field was again reversed and ions allowed to 

 evaporate. The ion evaporation, however, did not proceed at the previous 

 rate but at a rate two or three times as great. This was because 6 had been 

 made uniform, by migration from areas of high 6 to those already cleared 

 by the first peaks. A uniform low 6 could also be produced by coating the 

 clean filament in a retarding field for ions. Here again, on reversing the 

 field, the values of Vp were larger than when the same total d was reached 

 by ion evaporation. 



It is emphasized that the inhomogeneities needed to serve as nuclei for 

 these discontinuous boundary effects do not compose any appreciable part of 

 the tungsten surface. They are probably crystal boundaries or irregularities 

 in these boundaries and need occupy no more than the one-half percent 

 of surface discovered in the analysis of the atom evaporation data. That the 

 effects are not accidental, such as might be caused by gas covered areas, is 

 shown by the exact repetition of the peaks in numbers of experiments 

 made at any stage in the aging process. Aging probably removed certain of 

 the inhomogeneities and caused the rest to become sufficiently uniformly 

 distributed to produce a regular evaporation rate. 



As further evidence that the smooth curve finally obtained is still dis- 

 turbed by the formation and growth of nuclei and does not represent the 

 true variation of Vp with 6, the observed values of Vp have also been plotted 

 (circles) in Fig. 19. At higher values of 6, v^ obs>Vp caic- This is due to 

 the formation of nuclei and boundaries as just explained. At lower values 

 of ^,Vp obs<Vp caic- Tliis is bccausc the high rate of evaporation from the 

 few remaining concentrated patches is finally overbalanced by the prac- 

 tically zero rate from the large areas of nearly bare tungsten. Thus the 

 observed maximum at ^ '~' 0.04 and the whole behavior of Vp(obs) depends 

 on the average of the rates from these two (concentrated and dilute) 

 phases and on the displacement of the phase boundary. The surface at 



