MR CHAKI.KS I>K \\.MTKVU.LK ON FI.AMK SPKCTHA. 141 



copper, perforated with 30 holes, each of about 2 millims. diameter, arranged along a 

 straight line, jwvrallel to the axis of the tube. In order to avoid any accumulation of 

 water in the indiarubl)er tul>e which conveys the mixture to the burner, it is advisable 

 to fix the tul>e GH at a level slightly higher than that of the burner. The mixture 

 arrives at the burner from the branch K of a three-way glass tube. Any liquid which 

 may condense here will fall into the closed tube R, the contents of which may be 

 easily removed. 



The flame is really composed of 30 small flames, each of which, by carefully 

 regulating the supply of gas, may be made to have the same height. Owing to the 

 transparency of the flames, the total intensity of the light will be considerably 

 increased in the direction of the length of the burner. 



If reference be made to the description which M. Gouv has given of his apparatus, 

 it will lie noticed that I have suppressed the regulator, which had Ixjen found 

 necessary when making delicate photometrical measurements. But this accessory 

 would have made it imperative for the experimenter to Ixj in constant attendance, 

 and in the present case this was impossible on account of the long exposures which 

 had to be made. However, the apparatus was found to be very steady in its working, 

 owing chiefly to the employment of a very constant air supply. In one stage of the 

 work I used a large gas-holder which had a capacity of 1 cubic metre, and which 

 could l>e filled with air under a pressure of four atmospheres. At another time a 

 public supply of compressed air at constant pressure was utilized. All that was 

 necessary was to adjust the gas supply from time to time in order to keep the flame 

 at a fixed height. The quantity of air used was altout 1 cubic metre (under 

 atmospheric pressure) per hour. 



Solutions employed. In all cases where this has l>een possible, a solution of the 

 chloride of the metal to Ixj investigated has l>een used. Only in those cases where 

 the chloride is but slightly soluble have other solutions l>een found necessary. 

 However, the particular salt taken is mentioned at the commencement of all the 

 descriptions of the different spectra examined. The solutions have been made as 

 strong as possible, but not so strong that the salt has crystallized out during the 

 working of the apparatus. When crystallization has occurred, a small addition of 

 warm, distilled water has been sufficient to overcome the inconvenience. The 

 concentration of the solution is of considerable importance, for upon it depends the 

 intensity of the lines of the spectrum ; certain of them may even disappear completely 

 if the solution is too dilute. For instance, the green lithium line 4972, which had 

 not hitherto been noticed in the flame, becomes extremely bright when a strong 

 solution of the chloride of this metal is used. Again, an experiment made with the 

 chloride of calcium showed a very marked difference in the intensity of the lines on 

 two photographs, obtained under precisely similar conditions, except that in one case 

 a dilute solution (10 per cent.) and in the other a saturated solution was employed. 

 It is, alx>ve all, necessary to secure that as much as possible of the spray of the 



