159 



replaced by an acetylene-oxygen flame. The temperature of the gas- 

 oxygen flame is estimated at 2000° C, that of the acetylene-oxygen 

 flame at 2400° C. 



The temperature of the sodium tube, used for the experiment to 

 which Fig. 4 relates, was determined at about 300° C. by means of 

 a thermo-couple. 



A comparison of the photographs 3 and 4 clearly exhibits a dif- 

 ference of the ratio of the intensities of the outer components. At 

 the higher temperature the outer components appear much fainter 

 in relation to the inner ones than at the temperature of 300° C. 

 A closer inspection shows that in the latter case also, the inner 

 components surpass the outer ones in intensity, but the difference is 

 less. If the density of the sodium vapour is increased, there is a 

 greater difference in the behaviour of inner and outer components 

 at the higher than at the lower temperature. We have the intention 

 to return to this point on a further occasion. It would appear that 

 we are able by merely changing the temperature to influence the 

 separate magnetic components and to change within certain limits 

 their intensities. 



With the electric spark the outer components are relatively feeble 

 just as with the oxygen-acetylene flame. 



It should be noticed that in the experiment to which Figui'es 3 

 and 4 relate, the temperature has been changed, but the pressure 

 as well. 



In the glass tube the pressure is of the order of some millimetres, 

 whereas the total pressure — which according to Humphreys is deci- 

 sive in the phenomena of pressure-shift of spectrum lines, — in the 

 experiment with the gas-oxygen flame is equal to one atmosphere. 



In a few cases Humphreys observed also a change in the relative 

 intensities of spectrum lines. Hence there would be a possibility that 

 the difference between figures 3 and 4 is partially due to a change 

 of the total pressure. 



The effect now under review is of a rather complicate character. 

 Besides the DopPLER-effect, also the density of the vapour, the change 

 of pressure and the length of the path of the light rays must play 

 a part. Though apparently there exists an influence of temperature 

 on the aspect of the resolved lines under the circumstances of our 

 experiments, it cannot be denied that the interpretation of observations 

 concerning the amount of the separation is more easy. The compli- 

 cations mentioned are then eliminated. 



