5 2 INFLUENCE OF A MAGNETIC FIELD UPON THE SPARK SPECTRA OF IRON AND TITANIUM. 



These measurements are intended only as a preliminary test of the reality of the difference in triplet 

 spacings. The evidence, however, points strongly to the existence of a true difference for many, if not 

 all triplets. Only 3 out of 26 lines fail to show a larger interval for the violet component. Although the 

 settings on a component seldom show a range greater than 0.004 A, which would indicate a very small 

 probable error in the mean of 8 determinations, it is likely that the actual probable error of the indi- 

 vidual diiTerences shown in Table 6 may amount to 0.003 or 0-004 A as a result of systematic errors in 

 the settings due to the character of the lines. The mean of all the differences is + 0.006 A, with a calcu- 

 lated probable error of o.ooi A, which can scarcely leave any doubt as to the reality of the difference. 



The measurements show that the magnitude of the difference can hardly be the same for all of the 

 lines. The true probable error will then be somewhat smaller than that given above, which would only 

 make the e\'idence for the reahty of the dissymmetry predicted by Voigt the stronger. The Unes from 

 X 3930.450 toward the violet, 17 in number, are with one exception either normal triplets or have the separ- 

 ation 3a, usually the latter. Of the 9 lines showing a difference greater than 0.008 A, 3 are normal trip- 

 lets and 4 have a separation of 2,(i- The question of dissymmetry seems worthy of investigation through 

 a long range of field-strengths for these hues, especially to test the generality of the change of spacing 

 with the square of the field-strength observed for one of the Hues in the mercury spectrum (see p. 5). 



An element which might sometimes affect the spacing of Zeeman components is the apparent differ- 

 ence in the wave-lengths of arc and spark lines. The spark is made more disruptive by the magnetic 

 field, and a greater disruptiveness seems in general to cause the lines of the spark to be moved slightly 

 toward the red as compared with their positions in the arc spectrum. The reality of this effect is still 

 a disputed question, but evidence published by a number of observers, as well as some photographs of 

 the arc and spark which I have taken for this portion of the iron spectrum, indicate that measurements 

 taken in the regular way will give a sHghtly greater wave-length for the spark lines, the difference being 

 greatest for a very disruptive spark. If this effect has a part in the Zeeman phenomenon, we should 

 expect all components of the triplet to be displaced alike. The greater strength of the middle component, 

 however, would probably make the effect more perceptible for this, as the apparent displacement is more 

 or less combined with unsymmetrical widening and is usually more distinct for strong lines. However, 

 in the photographs from which the measurements of Table 6 were taken, triplets to the violet of X 4000 

 show the middle component only about as strong as either side component on account of the polariza- 

 tion given by the angle of the grating used, so that the conditions of the spark discharge would not seem 

 to be adequate to explain the difference in spacing, unless the direction of vibration of the electrons, 

 parallel or perpendicular to the hues of force, affects their susceptibility to the displacing action of the 

 spark discharge. On this point we have no evidence. 



The other point of dissymmetry predicted by Voigt, a greater strength for the red component of the 

 triplet, is quite perceptible for many lines, especially in the iron spectrum. The difference is rarely greater 

 than 10 per cent., and, to be clearly detected, the two components must be distinct but not of full density, 

 since blackness of the components in the negative destroys so shght a difference. On account of this 

 necessity for just the right degree of exposure, it is difficult to say how general the phenomenon is, but it 

 is certainly present for many lines. 



