bakus.1 PYRO-ELECTRIC PROPERTIES OF ALLOYS 159 
III interpreting these results graphically it is necessary to proceed 
with caution; for inasmuch as specific resistance enters into them recip- 
rocally, large values of resistance will be only inadequately represented. 
The tables contain many such values, nevertheless, enough data re- 
main to exhibit the striking linear character of the curves on which 
the gold, silver, and copper points, respectively, lie. It is certain that 
the initial tangent coincides with the initial curves throughout an enor- 
mous extent of their course. Matthiessen, 1 who expressed this result 
under a somewhat involved form, was well justified in computing 
by means of it the conductivity of a pure metal from data found for 
metals slightly impure. This computation premises the truth of Mat- 
thiessen's other principle that, with certain distinct exceptions, the 
electrical temperature-coefficients of all pure metals are the same. 
After Matthiessen and Yogt the curious relation in question seems to 
have failed to enlist further attention. It will be desirable therefore 
to endeavor to add to these results of Matthiessen and Vogt 2 and of 
myself, and to make a test of the corresponding properties of a group 
of metals whose electrical relations to temperature differ very largely 
from those relations which hold for platinum, silver, gold, copper, zinc, 
cadmium, tin, lead, arsenic, antimony, bismuth. I refer to the group of 
iron metals (iron, nickel, cobalt), more especially to the iron-carburets.' 2 
The latter also present an unusually large range of electrical variation ; 
and the fairly complete set of results investigated by Dr. Strouhal 3 and 
tnyself is available for discussion. If to these be added Matthiessen's 
result for pure iron, and a mean curve be passed through all the ob- 
servations graphically, the following principal coordinates of this curve, 
#=15, 45, 70, 
#=0.00420, 0.00166, 0.00130, 
if interpreted by the equation (x+l)(y+m)=n, lead to the constants, 
Z = _3.73, m= -0.000706, rc=0.0394. 
rhese constants do not reproduce the graphic curve satisfactorily,neither 
lias the constant I here the negligibly small value found for it in work- 
ing with platinum alloys. In view of these results, it is desirable to 
?xclude the interpolated value for pure iron, and to use only our own 
values for steel and cast-iron. It is not possible to reduce Matthiessen's 
^alue to our own satisfactorily, and the small datum for pure iron (large 
reciprocal) has an undue influence in modifying the constants to be ob- 
tained by the method of least squares. If cast-iron be also withdrawn 
:he results for steel form a unique series, the data being obtainable by 
1 Matthiessen and Vogt: Pbilos. Mag., London, 4th series, vol. 27, 1863, p. 4G7; 
Pogg. Ann., vol. 122, 1864, p. 19. 
2 This is particularly desirable here, because Matthiessen's and Vogt's iron alloys do 
lot seem to conform to the principle here in question, 
3 Ball. 11.. S. Geol. Survey, No. 14, }885, p. 15. 
(813) 
