( ^^<' ) 



in opposite senses from the bismuth of voN Ettijs'üSHAUSEN and 

 Nernst; also with respect to c, if we regard only the relation 

 with the magnetic force, whilst the discrepancy of sign is not found 

 in the alloys. 



That indeed the bismuth of the physicists mentioned was not 

 completely pure is moreover proved by what follows : the temperature 

 coefficient of the resistance was found to be negative, wliereas with 

 the electrolytic bismuth from Hartmann and Braun according to 

 Fleming and Dewar^) it is positive and just as in other pure 

 metals; the conductivity was 4,8. lO""" instead of 8,6.10~° with 

 electrolytic bismuth, and the increase of resistance in a magnetic 

 field 8400 was 307o against 40o/o with the electrolytic bismuth 

 according to Henderson ^). 



Hence very likely the bismuth used by me was very pure. By 

 this result the confidence in the results of the experiments is aug- 

 mented ; yet I will indicate a possible, though improbable disturbance, 

 which might account for the change of sign in the transverse effect. 

 It is namely possible that the thermo-electric needles did not follow 

 completely the variations of temperature of the plate, and that 

 therefore the disturbances caused by rotation of isothermal lines was 

 greater than was calculated. If we consider however, that I found 

 for that rotation greater values than voN Ettingshausen found 

 with soldered thermo-electric needles, that after 1 minute the deflections 

 varied rarely more than one scale-division, and that my values for 

 the rotation would have to become at least (3 times greater in order 

 to render the effect positive in all cases, then it will be agreed that 

 this explanation is very improbable. I intend however to repeat one 

 of the experiments with soldered electrodes. 



5. Before proceeding to the calculation of ?«, v etc. from the new 

 data I wish to point out a remarkable discrepancy between Riecke's 

 theory and my observations. 



According to formulae 1 and 3 the ratio of the coefficients of 

 thermomagnetic difference of potential and gal vanomagnetic difference 



of temperature is equal to 



— and — however are equal to 



1) Pi-oc. Roy. Soc. 60, p. 73 and 435, 1896. A similar result I obtained with 

 bismuth from Oberschlema. (See my dissertation p. 9'.)). 



2) Wied. Ann. 53, p. 912, 1894. 



