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



itself is slightly diffuse, which may account for the lack of sharpness in the components. "Quadruple?" 

 means that the two ^-components are fairly sharp, but the /i-component is probably double. A doubtful 

 quintuplet will usually have five components measurable, with indications that others are possibly present. 

 Doubtful sextuplets are very common. As a rule such a line has its two H-components widened so that 

 there are probably two pairs, while the /^-component is either distinctly double, or unresolved and con- 

 siderably widened. The decision between doubtful sextuplets and septuplets is frequently difficult and 

 often quite uncertain. The /(-component in such cases is not resolved, but the character of its widening 

 will often show whether it is double or triple. A widening with strong central maximum means usually 

 three /(-components, but there may be five. Such a line, if it has two widened ^-components, is classed 

 as a probable septuplet. Octuplets and hues of higher separation are classified in a similar way, the widen- 

 ing given in the two AX columns, together with the remarks, showing in what respect the given char- 

 acter may be doubtful. Lines whose H-components are "fringed" are difficult to classify. Such "fringes" 

 indicate very close, unresolved components, and these may be numerous. A field double that available 

 here would probably show the full structure. Many hnes were fully resolved by a field of 20,000 which 

 had to be described as "fringed" for a field of 16,000. The degree of widening due to the fringes is given 

 in the AX column and a remark tells whether the fringes are toward the center or outwards. The number 

 of components is estimated as closely as possible from the width of the fringes, but when the structure 

 is very complex, an interrogation point is used without any attempt to give the number of components. 

 Although the doubtful elements which have been mentioned come into the estimates as to the char- 

 acter of hnes, the large number of plates from which the material was taken gave an opportunity to study 

 each fine under various conditions of intensity and degree of separation, so that the classification as to 

 character is probably as accurate as can be made without very much greater field-strength combined 

 with as high resolving power as was here used. 



4. Weight. 



Under this heading, each fine for wliich measurement was possible is given the weight 3, 2, or i, 

 according as the quahty of the Zeeman components for measurement is good, fair, or poor. This grading 

 should be of much service in any use which is made of these tables. In attempts which the author has 

 made to compare his measurements with those of others, the discordances were nearly always found to 

 occur in the case of lines of such character that one or both sets of measurements were poor. If hnes 

 of high weight in each set are compared, a good check on the observations is obtained. 



Lines of weight 3 have sharply defined components, and for such hnes measurements of the same 

 plate by different observers or different sets by the same observer usually give differences in the third 

 decimal place only, while for many lines of this class the probable error is not greater than two or three 

 thousandths of an Angstrom. Only hnes of weight 3 should be used in comparisons of field-strength. 



Lines are weighted 2, when the fine is reasonably strong, because the components are widened and 

 probably compound, fringed, or perhaps single and poorly defined for some reason, so that the measure- 

 ment is not so close as for lines weighted 3. Measurements of weight 2 have usually a probable error 

 not greater than 10 per cent and may be used for quantitative comparisons where a high degree of pre- 

 cision is not required. When a component is measured from the no-field fine, it is never weighted higher 

 than 2. A fine whose components are uniformly widened, each consisting of two or more components 

 of about equal intensity, gives a better measurement than a line whose components are fringed, since 

 in the latter case photographic conditions affect the distinctness of the maximum of each shaded com- 

 ponent, this maximum being the part measured. 



Weight I is given to hnes which are very faint, much disturbed by blends, or of such complex struc- 

 ture that the components are extremely diffuse. The error of measurement for such hnes may be large 

 and the three decimal places are entered only for the sake of uniformity. However, the figures given 



