I S INFLUENCE OF A MAGNETIC FIELD UPON THE SPARK SPECTRA OF IRON AND TITANIUM. 



also by the list of enhanced lines given by Lockyer (58) and by plates of the arc and spark spectra of iron 

 taken in this laboratory. For the titanium spectrum the tables and charts of Hasselberg (59) were use- 

 ful as far as Xsgoo. This was supplemented for the red end by the measures of arc hues by Fiebig (60). 

 The spark tables of Exner and Haschek and of Lockyer were used as for iron. The identifications of 

 solar lines in Rowland's Tables are in most cases so close to the values in the tables of arc and spark 

 spectra that there is no doubt of the correspondence of the lines. The wave-lengths given in this 

 pubhcation are entirely on the Rowland system. 



The chart of the iron arc spectrum by Buisson and Fabry (61) was of great assistance in the approxi- 

 mate identification of lines, the scale being almost the same as that of my plates taken in the third order 

 with the 13-foot focal length. In addition to using this chart for the iron spectrum, it served also for 

 titanium when used in conjunction with a set of plates which I made of the spectra of the titanium spark 

 and iron arc side by side. 



The definitive identification of lines was in the usual way by measurement from neighboring lines 

 whose identity was certain. On account of the incompleteness of the general tables of spectra for the 

 red region, a few lines are entered in my titanium table which may belong to other substances. Some 

 of these, in all probability, are lines given stronger in the spark than in the arc, which explains their 

 absence from Fiebig's fist. The doubtful origin of such lines is indicated in the column headed " Remarks." 



The spectrum given by the plane grating spectrograph not being quite normal, the reduction factor 

 of the plate, expressed in Angstrom units per millimeter, was determined at intervals usually of 2 to 3 cm. 

 The change in the factor between successive determinations was thus almost always less than 5 in the 

 third decimal place. This factor was multiphed by the distance in miUimeters between the Zeeman 

 components, which was the mean of at least four differential measurements taken alternately right and 

 left, setting first on one, then on the other of the components whose separation was desired. The accu- 

 racy of setting on first-class lines was usually well within 0.005 ^m- From such lines there are all grada- 

 tions up to those for which the measurements recorded can be taken only as indicating the order of mag- 

 nitude of the separation. Frequently a fine has its components on one side blended with those of an 

 adjacent line. In such a case it is usually possible to make a more or less accurate measurement of half 

 the separation by measuring from the clear component to the no-field line which was always photo- 

 graphed in juxtaposition. The accuracy of measurement will be discussed further in the explanation 

 of the tables when the weight of measurements is considered. 



After measurement by a member of the Computing Division each plate was carefully gone over by 

 the author. In this examination the identification of fines was checked, the character of the separation 

 and weight of the measurement as determined by the quality of the fine were decided upon, and many 

 check measurements with the machine were made, including all measures for determination of the mag- 

 netic field by a comparison of the separation of lines on different plates. 



