428 L. Bell — Ultra-violet Spectrum of Cadmium. 



very exactly determined by a simple linear measurement. 

 Several photographs of this region were taken, and measured 

 very carefully by means of a dividing engine provided with an 

 accurate screw and reading microscope. To make sure of accu- 

 racy the slit was narrowed until only the strongest lines ap- 

 peared, while the definition was exceedingly sharp. Five 

 strong lines were thus measured, taking mean of measurements 

 from four negatives, with the following resulting wave-lengths, 

 2321-14, 2312-83, 2288-01, 2264-88, 2264-42. It is highly im- 

 probable that these are in error by as much as one part in fifty 

 thousand. The last two lines of this group form a sharp double 

 which with the dispersion and resolving power usually em- 

 ployed appears as a single strong, broad line with a conspicu- 

 ous nimbus, Mascart's twenty -fourth line. 



Having thus determined both ends of the spectrum with a 

 satisfactory degree of exactness, the next step was to connect 

 them by a series of negatives overlapping each other considera- 

 bly and sharp enough to yield accurate measurements for the 

 positions of the principal lines. As the only object in view 

 was the determination of a series of standard wave-lengths, no 

 attempt was made to photograph the fainter lines, but the slit 

 was narrowed to secure the best definition practicable. 



The main difficulty to be met in photographing spark spectra 

 with the concave grating is lack of light. In spite of the large 

 size of the grating its great focal length reduces the light so 

 much that with the spark apparatus available it was best to 

 photograph mainly in the first spectrum, although several nega- 

 tives were taken in the second as a check upon the work. A 

 string of seven negatives was then taken, overlapping by a con- 

 siderable amount, and covering in all a space of about 90 cm . 

 Besides these, odd photographs were taken in both the first and 

 second orders, to secure sufficient checks. 



These photographs were then placed on the dividing engine 

 previously mentioned and measured with the utmost care, each 

 measurement being repeated several times independently. 

 Various portions of the screw were used to eliminate errors, 

 and the measurements were made as nearly as possible at the 

 same temperature, so that errors due to expansion of the glass 

 became practically inappreciable. The dispersion used was so 

 great that one millimeter on the negative equalled 2-72 tenth- 

 meters of wave-length, so that a small error in measuring could 

 produce but a slight difference in the resulting wave-length. In 

 fact, very good results could be obtained by simply measuring 

 the photographs with an accurate scale, wave-lengths thus ob- 

 tained agreeing to within one part in ten thousand. In case of 

 each line measures were taken from both ends of the spectrum, 

 and they agreed so closely that the probable error of the deter- 



