382 BELL SYSTEM TECHNICAL JOURNAL 



For example, in a 1 per cent solution of silver the spark spectrum 

 contains forty-five lines, under the conditions of Hartley's experiments. 

 In a 0.1 per cent solution of silver twenty-five lines remain; in a 0.01 

 per cent solution only nine lines remain; and at 0.001 per cent but one 

 of these persists. Pollock and Leonard, following Hartley, also used 

 solutions with an improved sparking cell and extended this method of 

 quantitative analysis to most of the common metallic elements. 

 Almost simultaneously De Gramont began his monumental work of 

 determining the persistent and ultimate lines {raies ultimes) of all the 

 known elements under the conditions of a condensed spark discharge. 



In 1922 Meggers, Kiess and Stimson ^ published the first paper on 

 what might be called the modern method of spectrochemical analysis. 

 Their departure from previous methods was that they did not rely 

 solely upon the presence or absence of certain lines as a measure of the 

 amount of an element present in the sample. Taking cognizance of the 

 fact that the concentration at which a line will just be visible depends 

 upon a number of other factors besides the nature of the element, 

 another means of standardization was adopted as follows: A graded 

 series of standard samples was prepared with known amounts of 

 impurity in the same base material as the alloy to be analyzed. This 

 series of standards was sparked under the same conditions as the alloy 

 in question and all spectra recorded on the same plate. This practi- 

 cally eliminated all variables of development and plate sensitivity and 

 reduced the variables of excitation materially. The impurity content 

 was determined by a visual comparison of the spectra of samples and 

 standards. 



The method of Meggers, Kiess and Stimson, known as the comparison 

 standard method, is still used extensively. It is subject, however, to 

 limitations as to reliability, sensitivity, and precision. It is not always 

 possible to make up homogeneous alloys for standards, and since only a 

 minute amount of sample is excited by the spark it may be difificult to 

 excite a representative sample. This can be obviated by using solu- 

 tions instead of alloys. Spark excitation is limited in its sensitivity 

 and it is frequently necessary to analyze for smaller amounts than can 

 be detected in the spark spectrum. The use of arc excitation usually 

 overcomes this difficulty and furthermore the arc is somewhat more 

 suitable for work with solutions, especially with small volumes. Visual 

 comparison of sample with standard is limited in precision. In recent 

 years, however, the use of the densitometer {vide infra) or of the 

 logarithmic sector has increased the precision of comparison so that it is 

 possible to make spectrochemical analyses with a precision of better 



^ Meggers, Kiess and Stimson, U. S. Bureau of Standards Paper No. 444 (1922). 



