( 'M 



in mind that the proximity of the well conducting copper strips 

 must cause a decrease of the occurring differences of temperature 

 and of potential. 



Some observations were made in magnetic fields differing a little 

 from those mentioned below. The communicated numbers were then 

 obtained by graphical interpolation, just as the numbers of voN 

 Ettingshausen and J^Jernst, entered under v. E. and N., which are 

 given for the sake of comparison. 



a. GalvanonuKjnetic difference of temperature. The method of ob- 

 servation was the following : when the current through the plate and 

 the magnetizing current had been closed during about 15 minutes 

 and the temperature had become approximately constant, the deflec- 

 tions with both thermo-electi'ic couples were alternately read at con- 

 stant intervals, f. i. 4 times that with the first and 3 times that with 

 the second. Then the magnetic field was reversed, after some minutes 

 the series of experiments repeated, thereupon the magnetic field 

 reversed again and a third series of experiments made. Afterwards 

 this whole series was repeated with the direction of the current 

 through the plate reversed. 



Each series of 7 observations furnishes a mean difference of tem.- 

 perature between the upper and the lower border of the plate. The 

 difference between the mean of series 1 and 3 and series 2 gives 

 double the galvanoraagnetic difference of temperature. Finally the 

 mean was taken of the results with both directions of the current. 



During these experiments water at the temperature of the room 

 was flowing through both tubes. But for this precaution a difference 

 of temperature of about 30° might arise between the two borders 

 where the current entered and left the plate, caused by Peltier- 

 effect; in consequence of the rotation of the isothermal lines a sen- 

 sible error might be caused by this difference of temperature. 



Current through the plate 3,0 ampères. 



