812 
a little para-azoxyanisol is put and the plates are pressed together 
and to the strip of copper K by two steel springs. The strip of 
copper is heated on a little gas-flame, till the para-azoxyanisol is 
melted. By capillary forces the isotropic liquid is drawn to the 
narrowest part in the centre, where it absolutely fills the space 
between A and C, and even by placing K in a perpendicular position 
hardly moves down at all. Air-bubbles, if they are present, are moved 
to the border by tapping carefully. After that, K is pushed into 
the oven, which before is brought to the temperature desired. The 
phases ex-liquid and ex-solid are obtained by regulating the heating- 
current around the oven, without moving the cuvet from its place. 
With the preparation obtained in this way the above mentioned 
measurements are taken. 
The source of light was a small Nitra-lamp, for which a battery 
of accumulators provided the constant current; the perpendicular 
incandescent wire of this lamp is placed at the distance of the focus 
before a lens which provides a parallel beam of rays filling the 
Opening in the wall of the oven and penetrating the substance. At 
some distance behind the oven, in the centre of the parallel beam 
of rays the narrow slit of the spectrograph is placed, which may 
be closed by a little valve. The spectrum is photographed; the plate 
is put in a chassis which can be moved up and down, and renders: 
it possible to take several spectra on the same photographic plate. 
6. Each of the 30 spectra on the photographic plate is photometered 
in the length-direction; and out of the registered curves the blackening 
is calculated for the various colours. For this purpose the deviation 
of the galvanometer U is measured in definite points e.g. on the 
right side of each zero-point, while the situation of these points is 
fixed with respect to the He-line 2 = 4718. 
Suppose the maximum deviation of the galvanometer obtained 
through ‘the unlighted part of the plate to be U,, then the blackening 
is defined by the formula 
2 == log: Ue 
D 
For each spectrum the blackening in about 15 points is calculated 
and marked as function of the wavelength. By comparison of the 
curves for two or more spectra obtained in this way, which repre- 
sent the same state in various spots of the plate, their mutual 
concurrence shows the degree of reliability of the method used. 
The greatest deviations from the average values all appear to be 
