102 PROCEEDINGS OF THE AMERICAN ACADEMY. 



then immediately closed in order to prevent further heatiug of the face 

 of the thermopile. This throw of the galvanometer was taken as an 

 indication of the intensity of the flame. 



It was found that the thermopile would cool sufficiently within two 

 minutes to admit of the repetition of the reading. These observations 

 were taken by an assistant simultaneously with each setting of the 

 spectrophotometer, the intention being to reject any spectrophotometry 

 readings made at a time when the flame showed marked deviation from 

 its standard intensity, and to reduce the readings to a uniform flame 

 intensity under the assumption that for the small range of variation 

 occurring from reading to reading, the change in the brightness of the 

 flame would be proportional to the variations of this galvanometer read- 

 ing from the mean of the whole set. In point of fact it was found that 

 the flame rarely varied from the mean in the course of a set of observa- 

 tions by more than one per cent. From day to day, indeed, its intensity 

 was usually within the limits stated above. Occasionally a larger varia- 

 tion was detected. None of these variations in the course of the present 

 investigation reached values so great as to lead me to hesitate to apply 

 the correction already referred to, and all the observations described in 

 this paper have been reduced to a constant flame intensity by means of a 

 correction factor obtained from the readings of the galvanometer. 



Control and Measurement of the Temperature of the Carbon Rod. 



The carbon rod, having been brought to the desired degree of incan- 

 descence by means of the current from a storage battery, was held at a 

 constant temperature by varying the resistance placed in the battery 

 circuit. The indications of the thermo-element inserted in the rod were 

 noted by means of the potentiometer. The cells used in the measure- 

 ment of the temperature of the carbon rod were the same as those em- 

 ployed in the calibration pf the thermo-elements and in the study of the 

 temperature of the acetylene flame. 



The potentiometer having been balanced by looping the circuit con- 

 taining the thermo-element around a sufficient portion of the resistance 

 box to balance its current against that of the Clark cells, a condition 

 which was indicated by the reduction of the galvanometer deflection to 

 zero, the current was maintained at such a value as to hold the carbon 

 at a constant temperature during the time necessary to complete meas- 

 urements of the intensity of eight different portions of the spectrum, 

 ranging from the extreme red to violet, with the corresponding portions 

 of the spectrum of the flame. In order to insure the maintenance of this 



