26 
metals which, perhaps, have not exactly the same temperature and 
therefore also yield a thermo-current. This influence may be elimi- 
nated by placing between the galvanometer and the thermoelement 
a commutator; on commuting, only that part of the current which 
has to be measured, namely the current of the thermoelement itself, 
takes another direction ; it may, therefore, be determined from the 
difference. Only care must be taken that no differences in tempera- 
ture occur in the commutator itself. As such served two three-limbed 
glass tubes well wrapped up in cotton-wool and placed in a little 
box, which was suspended and moveable round a horizontal axis. 
In both tubes was poured a little mercury, while in each of the 
limbs were introduced wires which effected communication with the 
galvanometer and the thermoelement, respectively. By inclining the 
box to 45° in any direction, the current is closed, but this, in both 
cases, passes through the galvanometer in a different direction. There 
is still another source of error due to the thermoelement itself whose 
wires are often not quite homogeneous; and if there should be no 
equal temperature over their whole length, a thermo-current may be 
generated. Although these irregularities seem to occur but rarely 
with copper wires (and only these were here at different tempera- 
tures), care was taken all the same that they should not influence 
the final result, by changing the junctions in the two flasks after 
a series of measurements. 
We then must take again the difference of the resulting figures 
of different series to obtain the thermo-electric force of the copper- 
constantan and to calculate thence the difference in temperature 
between the potassium and sodium chloride. I refrain from giving 
a detailed communication of the results of the measurements because, 
anyhow, my conclusion must be that the two salts do not show a 
difference in temperature, at least none exceeding 0,001°. As I look’ 
upon this figure as representing the ‘accuracy attained, I do not 
attach any importance to the fact that the final result showed sodium 
chloride to be about 0.001° warmer than potassium chloride. Four 
experiments were carried out, the junctions of the thermoelement 
being changed after each; an experiment consisted of five to six 
measurements which were each composed of three to seven readings, 
carried out one after another with continuous commutation. 
From these experiments also, I must conclude to the absence of 
a-rays in potassium compounds; this result did not afford reason to 
make also an experiment with rubidium chloride. 
My best thanks are due to Dr. A. H. W. Aren, who placed his 
galvanometer at my disposal for these experiments. 
Inorg. Chem. Laboratory University of Av torah 
