556 HENEY A. ROWLAND 



2. Determination of about 15 lines in the visible spectrum by coin- 

 cidences by Rowland, using several gratings of 21^ feet focus. 



3. Interpolation by direct eye observations with concave gratings of 

 21 feet focus and micrometer of 5 inches range and of almost perfect 

 accuracy. 



4. Interpolation by means of flat gratings. 



5. Measurement of photographic plates from 10 to 19 inches long, 

 having two or three portions of the spectrum in different orders on 

 them, thus connecting the ultra violet and blue with the visible spec- 

 trum. The fact that nearly the same values are obtained for the violet 

 and ultra violet by use of different parts of the visible spectrum proves 

 the accuracy of the latter. 



6. Measurement of photographic plates having the solar visible spec- 

 trum in coincidence with the metal lines of different orders of spectra. 

 The fact that the wave-lengths of the metal lines are very nearly the 

 same as obtained from any portion of the visible or ultra violet spec- 

 trum proves the accuracy of the latter, as well as that of the metallic 

 wave-lengths. 



7. Measurement of plates having metallic spectra of different orders. 



ADVANTAGES OF THE PKOCESS 



The only other process of obtaining relative wave-lengths is by 

 means of angular measures. Supposing the angle to be about 45, an 

 error of 1" will make an error of about 1 in 200,000 in the sine of the 

 angle. When one considers the changes of temperature and barometer 

 measuring on one line and then another, together with the errors of 

 graduation, it would be a difficult matter to measure this angle to 2", 

 making an error of 1 in 100,000, or about ^ division of Angstrom. 



Looking over the observations of principal standards made under 

 the direction of Professor Vogel in Potsdam, with very poor gratings 

 but an excellent spectrometer, we find the average probable error to 

 be about =b T -g- uVinr f ^ ne wave-length, which is not far from the other 

 estimate. This does not include constant errors, and I believe the 

 probable error to be really greater than this. 



The method of coincidences by the concave grating gives far superior 

 results. The distance to be measured is very small, and the equivalent 

 focal length of a telescope to correspond would be very great (21 

 feet). Furthermore, all changes of barometer and thermometer are 

 eliminated at once, except the small effect on the dispersion of the air, 

 which, when known, can be corrected for. It is not to be wondered at 



