16 KENT AND TAYLOR. 



With Li X6708, which usually appears widely reversed in the grat- 

 ing, the grid is more difficult to obtain in the echelon, while with 

 Na X4972 — given as an unreversed line by the grating at either edge 

 or centre of the arc image — the echelon shows the grid at both edge 

 and centre. 



(i) We are now in a position to discuss jn detail Figures 6a and 6b. 

 These were obtained with the 131 mm. Lummer plate set between the 

 collimator and prism of Figure lb and crossed with echelon No. 2. 

 The source was that described on page 10: the arc current being from 

 10 to 25 amps. The plate dispersed vertically, the echelon horizon- 

 tally. Both figures are drawings based on visual filar micrometer 

 measurements, a single cross hair being moved successively along the 

 axes, vv' (vertical), hh' (horizontal), aa' (across the structure) pp' 

 (parallel to it), as shown below the two figures. 



Two Lummer plate orders are shown in each figure, the primes 

 distinguishing these. The numerals indicate the two components of 

 the spectroscopic doublet, the breadth along axis aa' their approximate 

 relative intensity. Xi is the weaker line, Xo the stronger in both 

 figures — X2 being the component of longer wavelength. 



In Figure 6a X2 is in double order condition; in 6b both Xi and X2 

 are between double and single order. The echelon grid structure is not 

 indicated in Figure 6a: in 6b its approximate position is shown. It 

 was difficult to observe at the ends of the lines and so is not there 

 indicated: it is slanted at an angle of about 2.4° (see gg' in Figure 6b) 

 with the vertical. The slant of the lines themselves as well as that of 

 the grid changes with the positions of both plate and echelon: further, 

 the grid slant is not due to the curvature of the echelon image. This 

 may throw some 'light on the disappearance of the grid at a breadth 

 of line greater than 2Ao. For, as the echelon action alone is given by 

 the projection, on the pp' axis, of the grids of the lines Xi and X2, it is 

 evident that lack of coincidence owing to slant would tend to obliter- 

 ate the grid altogether, this indicating that two broad lines, the centers 

 of which lie as far as 0.1 t. m. apart (the AX of the two components of 

 Li X6104), may not give coincident grid structures; or, in other words, 

 the grid maxima do not (for any one position and temperature of the 

 echelon) necessarily fall together. This is not inconsistent with shift 

 of intensity for small changes of wavelength (0.015 to 0.020 t. m.) as 

 noted on page 14. Shift of intensity and position probably both enter 

 with change of wavelength of the center of gravity of a primary echelon 

 image. 



These two figures show that the grid is unquestionably a secondary 



