228 DR. W. GEOFFREY DUFFIELD ON THE 



pressures from 1 to 200 atmospheres, where it was shown that the pressure-displacement 

 curves were slightly concave to the axis of pressures and where the curves representing 

 the mean displacements per atmosphere had a general downward tendency as the 

 pressure increased -as in this research. Dealing with the spectrum produced by the 

 copper arc when subjected to the highest pressure, the same features appeared, so that 

 the evidence favours the general conclusion : " That though the relationship between the 

 pressure and the displacement is approximately linear, the displacement does not 

 increase quite as rapidly as the pressure." 



(8) The Relation between Displacement and Wave-length : In Diagrams 4 and 5 

 each black circle represents the mean displacement per atmosphere of the nickel 

 line whose wave-length is given by the horizontal scale, the data being derived from 

 Table 111. Inasmuch as the displacements at 10 atmospheres pressure are dispro- 

 portionately large they are treated separately in Diagram 5, whereas they are 

 excluded from the calculations of the mean displacements which are plotted in 

 Diagram 4. The prominent feature of these diagrams is the increase of the displace- 

 ments as the wave-lengths increase, but the division of the lines into two groups 

 is also indicated. 



Treating the diagrams critically it is scarcely open to doubt that the displacement 

 is dependent upon the wave-length, though Diagram 4 alone is perhaps not 

 conclusive in this respect as there are not many lines in the region X 3900 to X 4200. 

 Many more lines have been measured at 10 atmospheres pressure than at higher 

 pressures, so Diagram 5 is able to provide more information about this region of the 

 spectrum, though on the whole the values are not so reliable since they are derived 

 from readings at only one pressure. 



Granted then that the occurrence of larger displacements in the less refrangible 

 parts of the spectrum is not fortuitous, it remains to discuss the actual relationship 

 between these two variables. The diagram at 10 atmospheres points to a steep 

 descent which might be regarded as approximately linear if it were not that it 

 would involve the displacements becoming zero and subsequently negative in the 

 more refrangible regions of the spectrum. 



Though negative values for displacements due to pressure have been recorded, the 

 crossing of the axis has not been observed in any spectrum, and the asymptotic trend 

 of the black dots in Diagram 4 is contrary to this occurring in the case of nickel. 

 This conclusion is supported by HUMPHREYS' measurements of lines of smaller wave- 

 length down to X = 3000 which are included as open circles in the diagram. Though 

 the majority of these readings are based upon observations at only one high pressure 

 (either 42 or 101 atmospheres) they are in such good agreement with the results of 

 the present research as to leave no doubt that the axis is not crossed in this region of 

 the spectrum. If we make the assumption that the origin is on the curve relating 

 to displacement and wave-length, we can at least say that it is not contradicted by the 

 result's of this research. On the assumption that the graph is of the form d = a\" the 



