440 0. Barus — Certain generic Electrical 



represent the relation here involved. The constants of the 

 lines drawn through the points are given at the end of the 

 tables in §§ 3, 4 and 7. In the cases of silver, of gold, of 

 platinum and of steel, the distribution of the points with 

 reference to these lines is satisfactory, when the errors intro- 

 duced by the mechanical treatment, by variations of hardness, 

 and particularly by imperfect homogeneity are justly taken into 

 account. In many cases, moreover, the percentage presence of 

 foreign ingredient is greater than that specified in § 1. As all 

 this is even more frequently the case with alloys of the oxyd- 

 izable metal copper, the line computed by the method of least 

 squares does not fairly represent these observations. The ex- 

 ceptional points here are the alloy of Cu with 22*4 per cent 

 Ag, and the brasses with 23*6 per cent, 29*4 per cent and 42*1 

 per cent Zn. If these high per cents are rejected, the line for 

 copper* will agree more nearly in character with the lines for 

 gold and for silver, as it will tend more nearly to intersect the 

 origin of coordinates (smaller numeric m). Applying the 

 method of least squares for the case in which the inadmissi- 

 ble copper alloys are withdrawn, I find m=— 0'278 and 

 ^=+0*005655, and of course a better agreement between ob- 

 served and calculated a throughout. In figure 2, however, I 

 have nevertheless inserted the line calculated for all the copper 

 alloys in hand. 



An interesting peculiarity of the steel line is that it leads to 

 a larger value for the temperature-coefficient of ironf than 

 that hitherto accepted. Comparisons of absolute values must 

 however be made with caution, because of the great variety of 

 electrical standards used by different observers. The high 

 temperature-coefficient of iron is in conformity with relatively 

 high values usually shown by alloys containing iron (cf. fig. 1.) 



I desire finally to advert to the occurrence of the relatively 

 small values of the constant, m, as computed for each of the 

 series of silver, copper, gold, platinum and steel alloys. There 

 is a marked tendency in all the cases stated to intersect the 

 coordinate axes very near the origin. Inasmuch therefore as 

 (§l)a=/'(jr,0)//(ft0) and J=l//(jr,0), the slopes of these 

 lines are very nearly equal to /*'(;£, 0); or more rigorously to 

 /'(OjO), since their true nature is that of an initial tangent 

 (cf. §3.) In §11 it appears that I am not asserting, how- 

 ever, that these lines do pass through the origin. 



* Those who have worked with copper alloys, will know the extreme difficulty 

 encountered in making the individual points conform to any uniform curve, The 

 data usually make up a diagram of very broken lines, as in the above work of 

 Matthiessen, and in results of Dr. Strouhal and myself. Further comment is 

 made in the Bulletin. 



f In his research on the conductivity of iron, Auerbach (Wied. Ann., viii, p. 479, 

 1879). discusses reasons for the exceptionally high value of the temperature- 

 coefficient of iron. 



