( 838 ) 
scarcely any variation of the effect. However, in my experiments 
the angle between the incident and reflected beams was only 2° or 3°. 
The incident light was polarized horizontally in the plane of incidence. 
From Righi’s observations it is known that even a normal incident 
beam of linearly polarized light when reflected from a magnetized 
mirror becomes elliptically polarized, the ellipticity however being 
only slight; Zeeman ») later measured this ellipticity in the case of 
iron and cobalt. Up to the present the evaluation of the ellipticity 
in my experiments has not been attempted; l considered that the 
slight reflecting power of some of my mirrors would not warrant 
such an attempt, and in addition it must be borne in mind, that the 
rotations themselves are small. Moreover the ellipticity, if any, must 
be nearly inappreciable, for by employing the best of my mirrors 
and by carefully avoiding diffused light, 1 have never beefi confronted 
with any difficulties, while the extinction of light in each half of 
the field of view was satisfactory. Even when the rotations are 
very small it is possible by means of the half-shade arrangement to 
observe and to measure them with sufficient accuracy. Hence it was 
thought unnecessary to use the method of multiple reflections, there¬ 
by avoiding new complications and further sources of error. The 
azimuth of the analyser was determined by means of a vertical 
scale seen through a combination of mirrors. 
For the production of the magnetic field a small du Bois semicir¬ 
cular electromagnet of resistance 9 & was employed. To avoid the 
danger of sparking with reversal of current about 60 12 were shunted 
across its terminals. The field was determined by means of a standar¬ 
dised thin glass-plate silvered at the back, which could be placed 
immediately in front of the mirror. The light (A = 589pp), being 
reflected by the mirror as described above, suffers a double magnetic 
rotation m the glass. The ensuing very slight double rotation of the 
light in its passage to and fro through the magnetized air could be 
computed from the data of Sikrtsema 2 ), but proved quite negligible. 
Indeed, by using a silver-mirror, it was found that the rotation lies 
withm the limits of experimental error. All the measurements were 
made with “polar” magnetization and at ordinary temperature. 
Test-Specimens. The following substances were experimen¬ 
ted upon : Cupriferrite (Cu O . Fe 2 0 8 ), Cakiumferrite (Ca 0 . Fe, O,), 
Magnetite (Ferroferrite) (Fe0.Fe,0,), llmenite (Ti, O,. Fe.0,), ferro- 
magnetic chro mic oxide (Or. 0 # ), “Invar” <36 Ni, 64 Fe), theHEUsLEK 
2 > p - Zeeman, Leiden-Comm. No 15, 1895. 
*> L- H. Siertsema, Versl. Kon. Akad. Wet. Amsterdam 7 p. 289, 1899. 
