334 Messrs. A. Schuster and G. Hemsalech. 



tempted to set it on the edge of the -most refrangible component. 

 There is reason to believe that this is the cause of the greater 

 deflection of the double line, and the photographs show some signs 

 that if this source of error is eliminated, the molecule giving out 

 the double line moves more quickly than that giving rise to the 

 triplet. We reserve the decision of this point until we have been able 

 to apply greater dispersion. 



Comparing the sparks obtained with different capacities, it is found 

 that when the spark gap is small, there seems a very curious diminution 

 of velocity as the capacity increases ; this is not what should have been 

 expected at first sight, as with the large number of jars we should 

 expect higher temperatures, and therefore greater velocity of diffusion. 

 When the spark gap is 1 cm., the experiments do not reveal any 

 marked change due to capacity. When the gap is increased still 

 further the sparks become very irregular and unsteady, and no certain 

 conclusions can be drawn from our measurements ; the numbers marked 

 with a query are specially doubtful. When six jars are used practically 

 identical numbers are obtained for all sparking distances, but with 

 small capacity the centimetre spark seems to give a lower result than 

 in the two other cases. While we should not like at present to con- 

 sider this as an established result, the table serves to show that the 

 centimetre spark and the highest capacity used gives the most con- 

 sistent numbers, and our experiments with other metals were all made 

 under these conditions, except in the case of bismuth, where clearer 

 spectra were obtained with only two jars. 



Comparing different metals with each other, we find in the first place 

 that those having comparatively low atomic weights, viz., aluminium 

 and magnesium, have higher molecular velocities. With magnesium 

 the metal vapour is scattered about to such an extent that no measure- 

 ments could be made, but the average velocity of the aluminium mole- 

 cule was found to be over three times as great as that of zinc, the 

 numbers not laying any claim to accuracy. Comparing zinc and 

 cadmium with each other, we obtain almost identical numbers, both 

 for the corresponding doublet and triplets. 



Bismuth gave remarkable results. In spite of its high atomic weight 

 some of the lines are but little displaced, indicating an average mole- 

 cular velocity of 1420 metres/second. For other lines the velocity 

 falls down to that of zinc and cadmium, while one line (A = 3793) has 

 a still smaller velocity. 



We have not obtained satisfactory results with mercury ; the best 

 were those in which poles used were of zinc or cadmium, which were 

 covered with amalgam. Differences in molecular velocities were ob- 

 tained for different lines, but the result here is not so certain as with 

 bismuth. There is obviously no simple law connecting these velocities 

 with the atomic weight. 



