( 351 ) 
even an error twice or three times the one now accepted for 
Fresnel’s rhomb, cannot disturb our results. 
6. Sources of light. Vacuum tubes charged with mercury, thallium 
or sodium, and heated to the required temperatures, were used as 
sources of light. The tubes were excited by means of an induction 
coil with high-speed interruptor. 
In order to render the luminosity of the tubes as high as possible, 
the current in the coil was gradually increased to the maximum 
just possible without impairing the narrowness of the spectral lines. 
In the case of sodium the troublesome separate heating of the 
tube was eliminated by following a somewhat modified procedure. 
Mercury tubes require only a preliminary heating, the current once 
started being sufficient to supply the necessary, heat. It was tried to 
mix up some fine powder of previously heated chloride of sodium 
with the mercury. The tube now exhibited, besides the mercury lines, 
the sodium lines with extremely great intensity. Often only some 
parts of the capillary are radiating; the most suitable part is then 
selected. In the experiments with sodium the coil was actuated by an 
alternate current, a maximum of 12 Amp. being possible before the 
widening of the sodium lines considerably interfered with the work. 
7. Arrangement of the experiments. A few words will suffice to 
describe the disposition of the experiments. 
A dd Bois-electromagnet with one perforated polar piece was used. 
The light emitted parallel to the field and made parallel by means 
of an achromatic lens, traversed Fresnel’s rhomb, then a second 
achromatic lens, which forms an image on the slit of the auxiliary 
spectroscope. 
This preliminary analysis enables us to select a definite spectral 
line for further investigation by means of the echelon spectroscope, 
a more detailed description of which I gave on a former occasion 1 ). 
Between the second lens and the slit of the auxiliary spectroscope 
a Nicol’s prism was introduced, mounted upon a divided circle, which 
gave the rotation of the Nicol in degrees. 
The front plane of the rhomb was placed accurately perpendicular 
to the beam. 
The plane of incidence of the Fresnel rhomb was made vertical 
in most experiments. In a few observations it was set at an azimuth 
of 45° with the horizontal plane. 
In the first case the emergent linear vibrations are at an angle of 
*) Zeeman. These Proceedings Nov. 30, 1901, p. 247. 
