392 Prof. W. G. Duffield : A Comparison of the 



of light from the centre outwards, a reversal is only possible 

 when a certain diminution of temperature has occurred, and 

 it depends upon the structure of the source what the density 

 of the material is in this critical layer, and this determines 

 the position of one edge of the absorption line. The width 

 of this absorption line is determined by the amounts of 

 material in the successive outer layers and their temperatures. 

 The second reason why the position of the absorption line is 

 more affected is that it is impossible to select the part of it 

 which corresponds to the position of maximum absorption, as 

 can be done in the case of emission lines ; there is usually no 

 alternative other than to set the cross-wires upon the geome- 

 trical centre of the reversal, an operation which may involve 

 a considerable error if it is broad. 



We thus find both the existence of reversals and their 

 actual positions very sensitive to the density and temperature 

 gradients. Since these are probably different in the arc and 

 spark, the spectra show corresponding discrepancies. 



That bright lines are similarly, but probably to a less 

 extent, affected we have experimental evidence in the dif- 

 ferences in wave-length which have been found when 

 capacity or self-induction has been introduced into a 

 spark circuit or when the length of an arc is changed, or a 

 different part of it examined*, all of which, it can scarcely 

 be doubted, affect the temperature and density gradients. 

 I noted some years ago a curious phenomenon which seems 

 to support this — when the iron arc under pressure was in 

 such a condition that reversals were more numerous than 

 usual f, the displacements not only of those lines which were 

 reversed, but also of those which remained bright, were 

 larger than when the reversals were fewer. It suggested 

 the dependence of the displacements upon the tendency to 

 reverse, which is in turn dependent upon the temperature 

 and density gradients from the core of the arc outwards, 

 consequently we see the probability of a connexion between 

 the displacements and the gradients. In the iron arc the 

 displacements of lines greatly broadened to the red (Group III. 

 in the original paper) were found to be particularly sensitive 

 to the tendency to reverse. It is perhaps worth noting that 

 in agreement with this the nickel spark and arc displace- 

 ments are less in accord for lines of this type than for those 

 only slightly broadened, the ratios of spark to arc displace- 

 ments being 0'84 and 0*96 respectively for these two classes 

 (see section 3). I do not, however, wish to urge this point 

 unduly since these figures are scarcely beyond the limits of 



* Royds, Kodaikanal Observatory Bulletin, Nos. 38 and 40. 

 t Duffield, he. cit, 



